swin transformer分类MNIST

安装依赖

• timm库中提供了swin transformer使用的DropPath层等结构

• torch库是构建神经网络和实现自动反向传播的基础库

• sys库提供了一些系统信息和操作的接口

• logging库提供了日志记录的功能

!pip install timm

Collecting timm
  Downloading timm-0.6.7-py3-none-any.whl (509 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 510.0/510.0 kB 799.5 kB/s eta 0:00:0000:0100:01
[?25hRequirement already satisfied: torchvision in /opt/conda/lib/python3.7/site-packages (from timm) (0.12.0)
Requirement already satisfied: torch>=1.4 in /opt/conda/lib/python3.7/site-packages (from timm) (1.11.0)
Requirement already satisfied: typing-extensions in /opt/conda/lib/python3.7/site-packages (from torch>=1.4->timm) (4.3.0)
Requirement already satisfied: requests in /opt/conda/lib/python3.7/site-packages (from torchvision->timm) (2.28.1)
Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in /opt/conda/lib/python3.7/site-packages (from torchvision->timm) (9.1.1)
Requirement already satisfied: numpy in /opt/conda/lib/python3.7/site-packages (from torchvision->timm) (1.21.6)
Requirement already satisfied: charset-normalizer<3,>=2 in /opt/conda/lib/python3.7/site-packages (from requests->torchvision->timm) (2.1.0)
Requirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.7/site-packages (from requests->torchvision->timm) (1.26.12)
Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests->torchvision->timm) (2022.6.15.2)
Requirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests->torchvision->timm) (3.3)
Installing collected packages: timm
Successfully installed timm-0.6.7
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv


from torch import nn, optim
import torch.backends.cudnn as cudnn
from timm.data import Mixup
import torch.utils.checkpoint as checkpoint

from timm.loss import LabelSmoothingCrossEntropy, SoftTargetCrossEntropy
from timm.utils import accuracy, AverageMeter
from timm.models.layers import DropPath, to_2tuple, trunc_normal_
from timm.optim.nadam import Nadam
from timm.scheduler.cosine_lr import CosineLRScheduler

import sys
import logging
import functools
from termcolor import colored

from torchvision.datasets import MNIST

创建日志记录的功能，可将输出按照一定格式重定向到文件中，在大型网络调试的时候对于掌握网络信息很有用

@functools.lru_cache()
def create_logger(output_dir, dist_rank=0, name=''):
    # create logger，如果名字有层级结构，则logger也会向对应的父级结构传递日志
    logger = logging.getLogger(name)
    # 设置记录的日志等级，有CRITICAL、ERROR、WARNING、INFO、DEBUG、NOTSET
    logger.setLevel(logging.DEBUG)
    # 禁止层级传递
    logger.propagate = False

    # create formatter
    # 格式化输出
    fmt = '[%(asctime)s %(name)s] (%(filename)s %(lineno)d): %(levelname)s %(message)s'
    color_fmt = colored('[%(asctime)s %(name)s]', 'green') + colored('(%(filename)s %(lineno)d)',
                                                                     'yellow') + ': %(levelname)s %(message)s'

    # create console handlers for master process
    if dist_rank == 0:
        # 主进程的控制句柄为向标准输出输出信息
        console_handler = logging.StreamHandler(sys.stdout)
        console_handler.setLevel(logging.DEBUG)
        console_handler.setFormatter(
            logging.Formatter(fmt=color_fmt, datefmt='%Y-%m-%d %H:%M:%S'))
        logger.addHandler(console_handler)

    # create file handlers
    # 文件句柄，向文件进行输出日志信息
    file_handler = logging.FileHandler(os.path.join(output_dir, f'log_rank{dist_rank}.txt'), mode='a')
    file_handler.setLevel(logging.DEBUG)
    file_handler.setFormatter(logging.Formatter(fmt=fmt, datefmt='%Y-%m-%d %H:%M:%S'))
    logger.addHandler(file_handler)

    return logger

swin transformer基本结构：

包括一个用于将图像打成patch的Patch Partition层、对patch后图像channel数目进行变换的Linear Embedding层，以及4个Basic Block用于层级transformer变换。数据流向为：

• Patch Partition和Linear Embedding层在代码中合并为Patch Embed层，用于将图片打成patch和变换channels数目。输入图像从(B,H,W,3)变换至(B,H//num_patch, W//num_patch, num_patchnum_patch3)，再变换至(B,H//num_patch, W//num_patch, 96)，代码实现是通过步长为num_patch的卷积操作一次性实现两个操作。此时形状记为(B, H0, W0, C)，注意该形状为代码实现中的写法，和图中表示有所不同。

• BasicBlock层包括多个Swin Transformer Block和一个Patch Merging层。Swin Transformer网络一共包含4个Basic Block，每层具有的Swin Transformer Block数目为[2,2,6,2]，每个Basic Block层中仅第一个Swin Transformer不需要实现shifted window操作。

  1. Swin Transformer Block中首先通过window_partition操作将图片划分为H0//window_size * W0//window_size个(B, window_size, window_size, C)形状的小窗口，并将其组织成(H0//window_size * W0//window_size * B, window_size * window_size, C)的形状，记为(nWB, N, C)，其中nW=H0//window_sizeW0//window_size为划分的小窗口数量，然后执行多头注意力WindowsAttention操作。

  2. WindowsAttention中先对矩阵通过全连接层生成形状均为的(nWB, N, C)的三个矩阵Q,K,V，然后按照多头注意力的数目将其形状变为(nWB, num_heads, N, C//num_heads)，将其视为nWBnum_heads个维度为(N,C//num_heads)的矩阵，通过 attn=Q K^\top 计算得到形状为(nWB, num_heads, N, N)的注意力矩阵，然后将通过attn和V矩阵相乘得到形状为(nWB, num_heads, N, C//num_heads)的矩阵，最后将其组织为(nW*B, N, C)形状的矩阵进行输出。因此WindowsAttention输入和输出的维度相同，可以任意堆叠多层。

  3. PatchMerging层之前，首先将多层WindowsAttention堆叠输出得到的(nWB, N, C)矩阵先按照小窗口的划分方式将小窗口合并为(B, NnW, C)，用输入图片的形状表示即为(B, H0W0, C)，然后将其变换为Swin Transformer Block层最初输入的图片形状(B, H0, W0, C)。PatchMerging层中通过间隔采样得到4个形状为(B, H0//2, W0//2, C)的矩阵，拼接得到的(B, H0//2, W0//2, C4)再送到全连接降维成(B, H0//2, W0//2, C2)的矩阵。该层输入为(B, H0, W0, C)，输出为(B, H0//2, W0//2, C2)从而实现下采样的目的。

• 由于最后一层Basic Block没有下采样层，因此经过4层Basic Block得到的输出为(B, H0//2^3, W0//2^3, C2^3)，经过AvgPool转化为(B, 1, C2^3)的形式，然后送入全连接进行分类。

class Mlp(nn.Module):
    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        self.fc1 = nn.Linear(in_features, hidden_features)
        self.act = act_layer()
        self.fc2 = nn.Linear(hidden_features, out_features)
        self.drop = nn.Dropout(drop)

    def forward(self, x):
        x = self.fc1(x)
        x = self.act(x)
        x = self.drop(x)
        x = self.fc2(x)
        x = self.drop(x)
        return x


def window_partition(x, window_size):
    """
    Args:
        x: (B, H, W, C)
        window_size (int): window size

    Returns:
        windows: (num_windows*B, window_size, window_size, C)
    """
    B, H, W, C = x.shape
    x = x.view(B, H // window_size, window_size, W // window_size, window_size, C)
    windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
    return windows


def window_reverse(windows, window_size, H, W):
    """
    Args:
        windows: (num_windows*B, window_size, window_size, C)
        window_size (int): Window size
        H (int): Height of image
        W (int): Width of image

    Returns:
        x: (B, H, W, C)
    """
    # 将nW*B个窗口合并为一个
    # B = B*nW/nW = (windows.shape[0]) / (H*W/window_size/window_size)
    B = int(windows.shape[0] / (H * W / window_size / window_size))
    x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1)
    x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1)
    return x


class WindowAttention(nn.Module):
    r""" Window based multi-head self attention (W-MSA) module with relative position bias.
    It supports both of shifted and non-shifted window.

    Args:
        dim (int): Number of input channels.
        window_size (tuple[int]): The height and width of the window.
        num_heads (int): Number of attention heads.
        qkv_bias (bool, optional):  If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set
        attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0
        proj_drop (float, optional): Dropout ratio of output. Default: 0.0
    """

    def __init__(self, dim, window_size, num_heads, qkv_bias=True, qk_scale=None, attn_drop=0., proj_drop=0.):

        super().__init__()
        self.dim = dim
        self.window_size = window_size  # Wh, Ww
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = qk_scale or head_dim ** -0.5

        # define a parameter table of relative position bias
        self.relative_position_bias_table = nn.Parameter(
            torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads))  # 2*Wh-1 * 2*Ww-1, nH

        # get pair-wise relative position index for each token inside the window
        coords_h = torch.arange(self.window_size[0])
        coords_w = torch.arange(self.window_size[1])
        # 避免报错信息
        coords = torch.stack(torch.meshgrid([coords_h, coords_w], indexing='ij'))  # 2, Wh, Ww
        coords_flatten = torch.flatten(coords, 1)  # 2, Wh*Ww
        relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :]  # 2, Wh*Ww, Wh*Ww
        relative_coords = relative_coords.permute(1, 2, 0).contiguous()  # Wh*Ww, Wh*Ww, 2
        relative_coords[:, :, 0] += self.window_size[0] - 1  # shift to start from 0
        relative_coords[:, :, 1] += self.window_size[1] - 1
        relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1
        relative_position_index = relative_coords.sum(-1)  # Wh*Ww, Wh*Ww
        # 注册为不可变参数，在保存模型时该参数也会被保存
        self.register_buffer("relative_position_index", relative_position_index)
        # 96 -> 96*3
        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)
        # 截断到(\mu-3\sigma,\mu+3\sigma)之内的正态分布
        trunc_normal_(self.relative_position_bias_table, std=.02)
        self.softmax = nn.Softmax(dim=-1)

    def forward(self, x, mask=None):
        """
        Args:
            x: input features with shape of (num_windows*B, N, C)
            mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None
        """
        B_, N, C = x.shape
        qkv = self.qkv(x).reshape(B_, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        # (3, B_, num_heads, N, C//num_heads)
        q, k, v = qkv[0], qkv[1], qkv[2]  # make torchscript happy (cannot use tensor as tuple)

        # 注意力机制
        q = q * self.scale
        attn = (q @ k.transpose(-2, -1))
        # (B_, num_heads, N, N)

        relative_position_bias = self.relative_position_bias_table[self.relative_position_index.view(-1)].view(
            self.window_size[0] * self.window_size[1], self.window_size[0] * self.window_size[1], -1)  # Wh*Ww,Wh*Ww,nH
        relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous()  # nH, Wh*Ww, Wh*Ww
        attn = attn + relative_position_bias.unsqueeze(0)

        if mask is not None:
            nW = mask.shape[0]
            # mask用于屏蔽不应connect的注意力矩阵部分
            attn = attn.view(B_ // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0)
            attn = attn.view(-1, self.num_heads, N, N)
            attn = self.softmax(attn)
        else:
            attn = self.softmax(attn)
        # 对注意力图进行dropout操作
        attn = self.attn_drop(attn)

        x = (attn @ v).transpose(1, 2).reshape(B_, N, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x

    def extra_repr(self) -> str:
        return f'dim={self.dim}, window_size={self.window_size}, num_heads={self.num_heads}'

    def flops(self, N):
        # 计算复杂度，并不是每秒的浮点运算数
        # calculate flops for 1 window with token length of N
        flops = 0
        # qkv = self.qkv(x)
        # N是小块拉直的数目(token长度)，dim是每个序列的长度
        # (N,dim)(dim,dim*3)->(N,3)计算复杂度为(N*dim*dim*3)
        flops += N * self.dim * 3 * self.dim
        # attn = (q @ k.transpose(-2, -1))
        # 先拆分为多头注意力，计算复杂度为num_heads*N*(dim//N)*N
        # (num_heads, N, dim//num_heads) (num_heads, N, dim//num_heads).T -> (num_heads, N, N)
        flops += self.num_heads * N * (self.dim // self.num_heads) * N
        #  x = (attn @ v)，计算复杂度为num_heads*N^2*dim//num_heads
        # (num_heads, N, N)(num_heads, N, dim//num_heads) -> (num_heads, N, dim//num_heads)
        flops += self.num_heads * N * N * (self.dim // self.num_heads)
        # x = self.proj(x)，计算复杂度为(N, dim, dim)
        # (N, dim) -> (N, dim)
        flops += N * self.dim * self.dim
        return flops


class SwinTransformerBlock(nn.Module):
    r""" Swin Transformer Block.

    Args:
        dim (int): Number of input channels.
        input_resolution (tuple[int]): Input resulotion.
        num_heads (int): Number of attention heads.
        window_size (int): Window size.
        shift_size (int): Shift size for SW-MSA.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
        qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
        drop (float, optional): Dropout rate. Default: 0.0
        attn_drop (float, optional): Attention dropout rate. Default: 0.0
        drop_path (float, optional): Stochastic depth rate. Default: 0.0
        act_layer (nn.Module, optional): Activation layer. Default: nn.GELU
        norm_layer (nn.Module, optional): Normalization layer.  Default: nn.LayerNorm
        fused_window_process (bool, optional): If True, use one kernel to fused window shift & window partition for acceleration, similar for the reversed part. Default: False
    """

    def __init__(self, dim, input_resolution, num_heads, window_size=7, shift_size=0,
                 mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0., drop_path=0.,
                 act_layer=nn.GELU, norm_layer=nn.LayerNorm,
                 fused_window_process=False):
        super().__init__()
        self.dim = dim
        self.input_resolution = input_resolution
        self.num_heads = num_heads
        self.window_size = window_size
        self.shift_size = shift_size
        self.mlp_ratio = mlp_ratio
        # 在小范围内做自注意力
        if min(self.input_resolution) <= self.window_size:
            # if window size is larger than input resolution, we don't partition windows
            self.shift_size = 0
            self.window_size = min(self.input_resolution)
        assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size"

        self.norm1 = norm_layer(dim)
        self.attn = WindowAttention(
            dim, window_size=to_2tuple(self.window_size), num_heads=num_heads,
            qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop)
        # 一种正则化手段，再batch维度随机设置一定样本不进行主干而直接由分支进行恒等映射
        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()
        self.norm2 = norm_layer(dim)
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)

        if self.shift_size > 0:
            # calculate attention mask for SW-MSA
            H, W = self.input_resolution
            img_mask = torch.zeros((1, H, W, 1))  # 1 H W 1
            h_slices = (slice(0, -self.window_size),
                        slice(-self.window_size, -self.shift_size),
                        slice(-self.shift_size, None))
            w_slices = (slice(0, -self.window_size),
                        slice(-self.window_size, -self.shift_size),
                        slice(-self.shift_size, None))
            # 给每块区域划分标记序号，每一个维度分为0:-7,-7:-2,-2:三部分
            cnt = 0
            for h in h_slices:
                for w in w_slices:
                    img_mask[:, h, w, :] = cnt
                    cnt += 1
            # 生成掩码窗口，64, 7, 7, 1
            mask_windows = window_partition(img_mask, self.window_size)  # nW, window_size, window_size, 1
            # mask_windows (64, 49)
            mask_windows = mask_windows.view(-1, self.window_size * self.window_size)
            attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2)
            attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0))
        else:
            attn_mask = None
        # 获得掩码矩阵，具体看论文中所说
        self.register_buffer("attn_mask", attn_mask)
        self.fused_window_process = fused_window_process

    def forward(self, x):
        H, W = self.input_resolution
        B, L, C = x.shape
        assert L == H * W, "input feature has wrong size"

        shortcut = x
        # 标准化层
        x = self.norm1(x)
        x = x.view(B, H, W, C)

        # cyclic shift
        if self.shift_size > 0:
            shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))
            # partition windows
            x_windows = window_partition(shifted_x, self.window_size)  # nW*B, window_size, window_size, C
        else:
            shifted_x = x
            # partition windows，nW为窗口数目
            x_windows = window_partition(shifted_x, self.window_size)  # nW*B, window_size, window_size, C
        # 窗口数目*每个窗口的token长度，通道数目
        x_windows = x_windows.view(-1, self.window_size * self.window_size, C)  # nW*B, window_size*window_size, C

        # W-MSA/SW-MSA，多个窗口可以并行地做注意力
        attn_windows = self.attn(x_windows, mask=self.attn_mask)  # nW*B, window_size*window_size, C

        # merge windows
        attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C)

        # reverse cyclic shift
        if self.shift_size > 0:
            # 将所有小窗口合并为一个大窗口
            shifted_x = window_reverse(attn_windows, self.window_size, H, W)  # B H' W' C
            # 对整个窗口进行左上的平移操作
            x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2))
        else:
            shifted_x = window_reverse(attn_windows, self.window_size, H, W)  # B H' W' C
            x = shifted_x
        x = x.view(B, H * W, C)
        x = shortcut + self.drop_path(x)

        # FFN
        x = x + self.drop_path(self.mlp(self.norm2(x)))

        return x


class PatchMerging(nn.Module):
    r""" Patch Merging Layer.

    Args:
        input_resolution (tuple[int]): Resolution of input feature.
        dim (int): Number of input channels.
        norm_layer (nn.Module, optional): Normalization layer.  Default: nn.LayerNorm
    """

    def __init__(self, input_resolution, dim, norm_layer=nn.LayerNorm):
        super().__init__()
        # 下采样，每行、每列间隔采样，通道数扩增到4倍，然后再将回到2倍
        self.input_resolution = input_resolution
        self.dim = dim
        self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False)
        self.norm = norm_layer(4 * dim)

    def forward(self, x):
        """
        x: B, H*W, C
        """
        H, W = self.input_resolution
        B, L, C = x.shape
        assert L == H * W, "input feature has wrong size"
        assert H % 2 == 0 and W % 2 == 0, f"x size ({H}*{W}) are not even."

        x = x.view(B, H, W, C)

        x0 = x[:, 0::2, 0::2, :]  # B H/2 W/2 C
        x1 = x[:, 1::2, 0::2, :]  # B H/2 W/2 C
        x2 = x[:, 0::2, 1::2, :]  # B H/2 W/2 C
        x3 = x[:, 1::2, 1::2, :]  # B H/2 W/2 C
        x = torch.cat([x0, x1, x2, x3], -1)  # B H/2 W/2 4*C
        x = x.view(B, -1, 4 * C)  # B H/2*W/2 4*C

        x = self.norm(x)
        x = self.reduction(x)

        return x


class BasicLayer(nn.Module):
    """ A basic Swin Transformer layer for one stage.

    Args:
        dim (int): Number of input channels.
        input_resolution (tuple[int]): Input resolution.
        depth (int): Number of blocks.
        num_heads (int): Number of attention heads.
        window_size (int): Local window size.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
        qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
        drop (float, optional): Dropout rate. Default: 0.0
        attn_drop (float, optional): Attention dropout rate. Default: 0.0
        drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0
        norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm
        downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None
        use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False.
        fused_window_process (bool, optional): If True, use one kernel to fused window shift & window partition for acceleration, similar for the reversed part. Default: False
    """

    def __init__(self, dim, input_resolution, depth, num_heads, window_size,
                 mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0.,
                 drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False,
                 fused_window_process=False):

        super().__init__()
        self.dim = dim
        self.input_resolution = input_resolution
        self.depth = depth
        self.use_checkpoint = use_checkpoint

        # build blocks
        self.blocks = nn.ModuleList([
            SwinTransformerBlock(dim=dim, input_resolution=input_resolution,
                                 num_heads=num_heads, window_size=window_size,
                                 shift_size=1,
                                 mlp_ratio=mlp_ratio,
                                 qkv_bias=qkv_bias, qk_scale=qk_scale,
                                 drop=drop, attn_drop=attn_drop,
                                 drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path,
                                 norm_layer=norm_layer,
                                 fused_window_process=fused_window_process)
            for i in range(depth)])

        # patch merging layer
        if downsample is not None:
            self.downsample = downsample(input_resolution, dim=dim, norm_layer=norm_layer)
        else:
            self.downsample = None

    def forward(self, x):
        for blk in self.blocks:
            if self.use_checkpoint:
                x = checkpoint.checkpoint(blk, x)
            else:
                x = blk(x)
        if self.downsample is not None:
            x = self.downsample(x)
        return x


class PatchEmbed(nn.Module):
    r""" Image to Patch Embedding

    Args:
        img_size (int): Image size.  Default: 224.
        patch_size (int): Patch token size. Default: 4.
        in_chans (int): Number of input image channels. Default: 3.
        embed_dim (int): Number of linear projection output channels. Default: 96.
        norm_layer (nn.Module, optional): Normalization layer. Default: None
    """

    def __init__(self, img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None):
        super().__init__()
        # 224 -> (224, 224)
        img_size = to_2tuple(img_size)
        # 4 -> (4, 4)
        patch_size = to_2tuple(patch_size)
        # 切割成[56,56]个块
        patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]]
        self.img_size = img_size
        self.patch_size = patch_size
        self.patches_resolution = patches_resolution
        self.num_patches = patches_resolution[0] * patches_resolution[1]
        # 输入通道数目为3
        self.in_chans = in_chans
        # 嵌入维度为96
        self.embed_dim = embed_dim
        # (b,3,224,224) -> (b,96,224//4,224//4)通道数变化
        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
        # 层归一化，常用于nlp，因此可能来源于transformer自带的
        if norm_layer is not None:
            self.norm = norm_layer(embed_dim)
        else:
            self.norm = None

    def forward(self, x):
        B, C, H, W = x.shape
        # FIXME look at relaxing size constraints
        assert H == self.img_size[0] and W == self.img_size[1], \
            f"Input image size ({H}*{W}) doesn't match model ({self.img_size[0]}*{self.img_size[1]})."

        x = self.proj(x).flatten(2).transpose(1, 2)  # B Ph*Pw C
        if self.norm is not None:
            x = self.norm(x)
        return x


class SwinTransformer(nn.Module):
    r""" Swin Transformer
        A PyTorch impl of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows`  -
          https://arxiv.org/pdf/2103.14030

    Args:
        img_size (int | tuple(int)): Input image size. Default 224
        patch_size (int | tuple(int)): Patch size. Default: 4
        in_chans (int): Number of input image channels. Default: 3
        num_classes (int): Number of classes for classification head. Default: 1000
        embed_dim (int): Patch embedding dimension. Default: 96
        depths (tuple(int)): Depth of each Swin Transformer layer.
        num_heads (tuple(int)): Number of attention heads in different layers.
        window_size (int): Window size. Default: 7
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4
        qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float): Override default qk scale of head_dim ** -0.5 if set. Default: None
        drop_rate (float): Dropout rate. Default: 0
        attn_drop_rate (float): Attention dropout rate. Default: 0
        drop_path_rate (float): Stochastic depth rate. Default: 0.1
        norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm.
        ape (bool): If True, add absolute position embedding to the patch embedding. Default: False
        patch_norm (bool): If True, add normalization after patch embedding. Default: True
        use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False
        fused_window_process (bool, optional): If True, use one kernel to fused window shift & window partition for acceleration, similar for the reversed part. Default: False
    """

    def __init__(self, img_size=224, patch_size=4, in_chans=3, num_classes=1000,
                 embed_dim=96, depths=[2, 2, 6, 2], num_heads=[3, 6, 12, 24],
                 window_size=7, mlp_ratio=4., qkv_bias=True, qk_scale=None,
                 drop_rate=0., attn_drop_rate=0., drop_path_rate=0.1,
                 norm_layer=nn.LayerNorm, ape=False, patch_norm=True,
                 use_checkpoint=False, fused_window_process=False, **kwargs):
        super().__init__()

        self.num_classes = num_classes
        self.num_layers = len(depths)  # depths每一个swin transformer层的深度，因为swin transformer可以随意堆叠
        self.embed_dim = embed_dim
        self.ape = ape  # 绝对位置编码
        self.patch_norm = patch_norm  # patch归一化
        self.num_features = int(embed_dim * 2 ** (self.num_layers - 1))  # 经过所有层后的通道数量
        self.mlp_ratio = mlp_ratio  # mlp隐藏层维度与嵌入维度的商

        # split image into non-overlapping patches
        # 对图片进行切割的块
        self.patch_embed = PatchEmbed(
            img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim,
            norm_layer=norm_layer if self.patch_norm else None)
        num_patches = self.patch_embed.num_patches
        patches_resolution = self.patch_embed.patches_resolution
        self.patches_resolution = patches_resolution

        # absolute position embedding
        if self.ape:
            self.absolute_pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim))
            trunc_normal_(self.absolute_pos_embed, std=.02)

        self.pos_drop = nn.Dropout(p=drop_rate)

        # stochastic depth
        # swin transformer层数一共为12，其中drop_path_rate=0.2，每一层的dropout概率从0增加到0.2
        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))]  # stochastic depth decay rule

        # build layers
        self.layers = nn.ModuleList()
        # 每经过一层BasicLayer，图片宽和高减半，channels加倍，96->96*2->96*4
        # 初始的图片为(b, 96, 3136)
        for i_layer in range(self.num_layers):
            layer = BasicLayer(dim=int(embed_dim * 2 ** i_layer),
                               input_resolution=(patches_resolution[0] // (2 ** i_layer),
                                                 patches_resolution[1] // (2 ** i_layer)),
                               depth=depths[i_layer],  # [2, 2, 6, 2]
                               num_heads=num_heads[i_layer],
                               window_size=window_size,  # 窗口尺寸是7
                               mlp_ratio=self.mlp_ratio,
                               qkv_bias=qkv_bias, qk_scale=qk_scale,
                               drop=drop_rate, attn_drop=attn_drop_rate,
                               drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])],
                               norm_layer=norm_layer,
                               downsample=PatchMerging if (i_layer < self.num_layers - 1) else None,
                               use_checkpoint=use_checkpoint,
                               fused_window_process=fused_window_process)
            self.layers.append(layer)

        self.norm = norm_layer(self.num_features)
        self.avgpool = nn.AdaptiveAvgPool1d(1)
        self.head = nn.Linear(self.num_features, num_classes) if num_classes > 0 else nn.Identity()

        self.apply(self._init_weights)

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            trunc_normal_(m.weight, std=.02)
            if isinstance(m, nn.Linear) and m.bias is not None:
                nn.init.constant_(m.bias, 0)
        elif isinstance(m, nn.LayerNorm):
            nn.init.constant_(m.bias, 0)
            nn.init.constant_(m.weight, 1.0)

    def forward_features(self, x):
        x = self.patch_embed(x)
        if self.ape:
            x = x + self.absolute_pos_embed
        x = self.pos_drop(x)

        for layer in self.layers:
            x = layer(x)

        x = self.norm(x)  # B L C
        x = self.avgpool(x.transpose(1, 2))  # B C 1
        x = torch.flatten(x, 1)
        return x

    def forward(self, x):
        x = self.forward_features(x)
        x = self.head(x)
        return x

pytorch官方训练参考中给出的相关代码

图片预处理相关

from torchvision.transforms import autoaugment, transforms
from torchvision.transforms.functional import InterpolationMode

class ClassificationPresetTrain:
    def __init__(
        self,
        *,
        crop_size,
        mean=(0.485, 0.456, 0.406),
        std=(0.229, 0.224, 0.225),
        interpolation=InterpolationMode.BILINEAR,
        hflip_prob=0.5,
        auto_augment_policy=None,
        ra_magnitude=9,
        augmix_severity=3,
        random_erase_prob=0.0,
    ):
        trans = [transforms.RandomResizedCrop(crop_size, interpolation=interpolation)]
        if hflip_prob > 0:
            trans.append(transforms.RandomHorizontalFlip(hflip_prob))
        if auto_augment_policy is not None:
            if auto_augment_policy == "ra":
                trans.append(autoaugment.RandAugment(interpolation=interpolation, magnitude=ra_magnitude))
            elif auto_augment_policy == "ta_wide":
                trans.append(autoaugment.TrivialAugmentWide(interpolation=interpolation))
            elif auto_augment_policy == "augmix":
                trans.append(autoaugment.AugMix(interpolation=interpolation, severity=augmix_severity))
            else:
                aa_policy = autoaugment.AutoAugmentPolicy(auto_augment_policy)
                trans.append(autoaugment.AutoAugment(policy=aa_policy, interpolation=interpolation))
        trans.extend(
            [
                transforms.PILToTensor(),
                transforms.ConvertImageDtype(torch.float),
                transforms.Normalize(mean=mean, std=std),
            ]
        )
        if random_erase_prob > 0:
            trans.append(transforms.RandomErasing(p=random_erase_prob))

        self.transforms = transforms.Compose(trans)

    def __call__(self, img):
        return self.transforms(img)


class ClassificationPresetEval:
    def __init__(
        self,
        *,
        crop_size,
        resize_size=256,
        mean=(0.485, 0.456, 0.406),
        std=(0.229, 0.224, 0.225),
        interpolation=InterpolationMode.BILINEAR,
    ):

        self.transforms = transforms.Compose(
            [
                transforms.Resize(resize_size, interpolation=interpolation),
                transforms.CenterCrop(crop_size),
                transforms.PILToTensor(),
                transforms.ConvertImageDtype(torch.float),
                transforms.Normalize(mean=mean, std=std),
            ]
        )

    def __call__(self, img):
        return self.transforms(img)

分布式数据采样

import torch
import torch.distributed as dist


class RASampler(torch.utils.data.Sampler):
    """Sampler that restricts data loading to a subset of the dataset for distributed,
    with repeated augmentation.
    It ensures that different each augmented version of a sample will be visible to a
    different process (GPU).
    Heavily based on 'torch.utils.data.DistributedSampler'.

    This is borrowed from the DeiT Repo:
    https://github.com/facebookresearch/deit/blob/main/samplers.py
    """

    def __init__(self, dataset, num_replicas=None, rank=None, shuffle=True, seed=0, repetitions=3):
        if num_replicas is None:
            if not dist.is_available():
                raise RuntimeError("Requires distributed package to be available!")
            num_replicas = dist.get_world_size()
        if rank is None:
            if not dist.is_available():
                raise RuntimeError("Requires distributed package to be available!")
            rank = dist.get_rank()
        self.dataset = dataset
        self.num_replicas = num_replicas
        self.rank = rank
        self.epoch = 0
        self.num_samples = int(math.ceil(len(self.dataset) * float(repetitions) / self.num_replicas))
        self.total_size = self.num_samples * self.num_replicas
        self.num_selected_samples = int(math.floor(len(self.dataset) // 256 * 256 / self.num_replicas))
        self.shuffle = shuffle
        self.seed = seed
        self.repetitions = repetitions

    def __iter__(self):
        if self.shuffle:
            # Deterministically shuffle based on epoch
            g = torch.Generator()
            g.manual_seed(self.seed + self.epoch)
            indices = torch.randperm(len(self.dataset), generator=g).tolist()
        else:
            indices = list(range(len(self.dataset)))

        # Add extra samples to make it evenly divisible
        indices = [ele for ele in indices for i in range(self.repetitions)]
        indices += indices[: (self.total_size - len(indices))]
        assert len(indices) == self.total_size

        # Subsample
        indices = indices[self.rank : self.total_size : self.num_replicas]
        assert len(indices) == self.num_samples

        return iter(indices[: self.num_selected_samples])

    def __len__(self):
        return self.num_selected_samples

    def set_epoch(self, epoch):
        self.epoch = epoch

数据增强措施CutMix和MixUp

import math
from typing import Tuple

import torch
from torch import Tensor
from torchvision.transforms import functional as F


class RandomMixup(torch.nn.Module):
    """Randomly apply Mixup to the provided batch and targets.
    The class implements the data augmentations as described in the paper
    `"mixup: Beyond Empirical Risk Minimization" <https://arxiv.org/abs/1710.09412>`_.

    Args:
        num_classes (int): number of classes used for one-hot encoding.
        p (float): probability of the batch being transformed. Default value is 0.5.
        alpha (float): hyperparameter of the Beta distribution used for mixup.
            Default value is 1.0.
        inplace (bool): boolean to make this transform inplace. Default set to False.
    """

    def __init__(self, num_classes: int, p: float = 0.5, alpha: float = 1.0, inplace: bool = False) -> None:
        super().__init__()

        if num_classes < 1:
            raise ValueError(
                f"Please provide a valid positive value for the num_classes. Got num_classes={num_classes}"
            )

        if alpha <= 0:
            raise ValueError("Alpha param can't be zero.")

        self.num_classes = num_classes
        self.p = p
        self.alpha = alpha
        self.inplace = inplace

    def forward(self, batch: Tensor, target: Tensor) -> Tuple[Tensor, Tensor]:
        """
        Args:
            batch (Tensor): Float tensor of size (B, C, H, W)
            target (Tensor): Integer tensor of size (B, )

        Returns:
            Tensor: Randomly transformed batch.
        """
        if batch.ndim != 4:
            raise ValueError(f"Batch ndim should be 4. Got {batch.ndim}")
        if target.ndim != 1:
            raise ValueError(f"Target ndim should be 1. Got {target.ndim}")
        if not batch.is_floating_point():
            raise TypeError(f"Batch dtype should be a float tensor. Got {batch.dtype}.")
        if target.dtype != torch.int64:
            raise TypeError(f"Target dtype should be torch.int64. Got {target.dtype}")

        if not self.inplace:
            batch = batch.clone()
            target = target.clone()

        if target.ndim == 1:
            target = torch.nn.functional.one_hot(target, num_classes=self.num_classes).to(dtype=batch.dtype)

        if torch.rand(1).item() >= self.p:
            return batch, target

        # It's faster to roll the batch by one instead of shuffling it to create image pairs
        batch_rolled = batch.roll(1, 0)
        target_rolled = target.roll(1, 0)

        # Implemented as on mixup paper, page 3.
        lambda_param = float(torch._sample_dirichlet(torch.tensor([self.alpha, self.alpha]))[0])
        batch_rolled.mul_(1.0 - lambda_param)
        batch.mul_(lambda_param).add_(batch_rolled)

        target_rolled.mul_(1.0 - lambda_param)
        target.mul_(lambda_param).add_(target_rolled)

        return batch, target

    def __repr__(self) -> str:
        s = (
            f"{self.__class__.__name__}("
            f"num_classes={self.num_classes}"
            f", p={self.p}"
            f", alpha={self.alpha}"
            f", inplace={self.inplace}"
            f")"
        )
        return s


class RandomCutmix(torch.nn.Module):
    """Randomly apply Cutmix to the provided batch and targets.
    The class implements the data augmentations as described in the paper
    `"CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features"
    <https://arxiv.org/abs/1905.04899>`_.

    Args:
        num_classes (int): number of classes used for one-hot encoding.
        p (float): probability of the batch being transformed. Default value is 0.5.
        alpha (float): hyperparameter of the Beta distribution used for cutmix.
            Default value is 1.0.
        inplace (bool): boolean to make this transform inplace. Default set to False.
    """

    def __init__(self, num_classes: int, p: float = 0.5, alpha: float = 1.0, inplace: bool = False) -> None:
        super().__init__()
        if num_classes < 1:
            raise ValueError("Please provide a valid positive value for the num_classes.")
        if alpha <= 0:
            raise ValueError("Alpha param can't be zero.")

        self.num_classes = num_classes
        self.p = p
        self.alpha = alpha
        self.inplace = inplace

    def forward(self, batch: Tensor, target: Tensor) -> Tuple[Tensor, Tensor]:
        """
        Args:
            batch (Tensor): Float tensor of size (B, C, H, W)
            target (Tensor): Integer tensor of size (B, )

        Returns:
            Tensor: Randomly transformed batch.
        """
        if batch.ndim != 4:
            raise ValueError(f"Batch ndim should be 4. Got {batch.ndim}")
        if target.ndim != 1:
            raise ValueError(f"Target ndim should be 1. Got {target.ndim}")
        if not batch.is_floating_point():
            raise TypeError(f"Batch dtype should be a float tensor. Got {batch.dtype}.")
        if target.dtype != torch.int64:
            raise TypeError(f"Target dtype should be torch.int64. Got {target.dtype}")

        if not self.inplace:
            batch = batch.clone()
            target = target.clone()

        if target.ndim == 1:
            target = torch.nn.functional.one_hot(target, num_classes=self.num_classes).to(dtype=batch.dtype)

        if torch.rand(1).item() >= self.p:
            return batch, target

        # It's faster to roll the batch by one instead of shuffling it to create image pairs
        batch_rolled = batch.roll(1, 0)
        target_rolled = target.roll(1, 0)

        # Implemented as on cutmix paper, page 12 (with minor corrections on typos).
        lambda_param = float(torch._sample_dirichlet(torch.tensor([self.alpha, self.alpha]))[0])
        _, H, W = F.get_dimensions(batch)

        r_x = torch.randint(W, (1,))
        r_y = torch.randint(H, (1,))

        r = 0.5 * math.sqrt(1.0 - lambda_param)
        r_w_half = int(r * W)
        r_h_half = int(r * H)

        x1 = int(torch.clamp(r_x - r_w_half, min=0))
        y1 = int(torch.clamp(r_y - r_h_half, min=0))
        x2 = int(torch.clamp(r_x + r_w_half, max=W))
        y2 = int(torch.clamp(r_y + r_h_half, max=H))

        batch[:, :, y1:y2, x1:x2] = batch_rolled[:, :, y1:y2, x1:x2]
        lambda_param = float(1.0 - (x2 - x1) * (y2 - y1) / (W * H))

        target_rolled.mul_(1.0 - lambda_param)
        target.mul_(lambda_param).add_(target_rolled)

        return batch, target

    def __repr__(self) -> str:
        s = (
            f"{self.__class__.__name__}("
            f"num_classes={self.num_classes}"
            f", p={self.p}"
            f", alpha={self.alpha}"
            f", inplace={self.inplace}"
            f")"
        )
        return s

数值平滑，减轻训练过程中损失函数数值的抖动

import copy
import datetime
import errno
import hashlib
import os
import time
from collections import defaultdict, deque, OrderedDict
from typing import List, Optional, Tuple

import torch
import torch.distributed as dist


class SmoothedValue:
    """Track a series of values and provide access to smoothed values over a
    window or the global series average.
    """

    def __init__(self, window_size=20, fmt=None):
        if fmt is None:
            fmt = "{median:.4f} ({global_avg:.4f})"
        self.deque = deque(maxlen=window_size)
        self.total = 0.0
        self.count = 0
        self.fmt = fmt

    def update(self, value, n=1):
        self.deque.append(value)
        self.count += n
        self.total += value * n

    def synchronize_between_processes(self):
        """
        Warning: does not synchronize the deque!
        """
        t = reduce_across_processes([self.count, self.total])
        t = t.tolist()
        self.count = int(t[0])
        self.total = t[1]

    @property
    def median(self):
        d = torch.tensor(list(self.deque))
        return d.median().item()

    @property
    def avg(self):
        d = torch.tensor(list(self.deque), dtype=torch.float32)
        return d.mean().item()

    @property
    def global_avg(self):
        return self.total / self.count

    @property
    def max(self):
        return max(self.deque)

    @property
    def value(self):
        return self.deque[-1]

    def __str__(self):
        return self.fmt.format(
            median=self.median, avg=self.avg, global_avg=self.global_avg, max=self.max, value=self.value
        )


class MetricLogger:
    def __init__(self, delimiter="\t"):
        self.meters = defaultdict(SmoothedValue)
        self.delimiter = delimiter

    def update(self, **kwargs):
        for k, v in kwargs.items():
            if isinstance(v, torch.Tensor):
                v = v.item()
            assert isinstance(v, (float, int))
            self.meters[k].update(v)

    def __getattr__(self, attr):
        if attr in self.meters:
            return self.meters[attr]
        if attr in self.__dict__:
            return self.__dict__[attr]
        raise AttributeError(f"'{type(self).__name__}' object has no attribute '{attr}'")

    def __str__(self):
        loss_str = []
        for name, meter in self.meters.items():
            loss_str.append(f"{name}: {str(meter)}")
        return self.delimiter.join(loss_str)

    def synchronize_between_processes(self):
        for meter in self.meters.values():
            meter.synchronize_between_processes()

    def add_meter(self, name, meter):
        self.meters[name] = meter

    def log_every(self, iterable, print_freq, header=None):
        i = 0
        if not header:
            header = ""
        start_time = time.time()
        end = time.time()
        iter_time = SmoothedValue(fmt="{avg:.4f}")
        data_time = SmoothedValue(fmt="{avg:.4f}")
        space_fmt = ":" + str(len(str(len(iterable)))) + "d"
        if torch.cuda.is_available():
            log_msg = self.delimiter.join(
                [
                    header,
                    "[{0" + space_fmt + "}/{1}]",
                    "eta: {eta}",
                    "{meters}",
                    "time: {time}",
                    "data: {data}",
                    "max mem: {memory:.0f}",
                ]
            )
        else:
            log_msg = self.delimiter.join(
                [header, "[{0" + space_fmt + "}/{1}]", "eta: {eta}", "{meters}", "time: {time}", "data: {data}"]
            )
        MB = 1024.0 * 1024.0
        for obj in iterable:
            data_time.update(time.time() - end)
            yield obj
            iter_time.update(time.time() - end)
            if i % print_freq == 0:
                eta_seconds = iter_time.global_avg * (len(iterable) - i)
                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))
                if torch.cuda.is_available():
                    print(
                        log_msg.format(
                            i,
                            len(iterable),
                            eta=eta_string,
                            meters=str(self),
                            time=str(iter_time),
                            data=str(data_time),
                            memory=torch.cuda.max_memory_allocated() / MB,
                        )
                    )
                else:
                    print(
                        log_msg.format(
                            i, len(iterable), eta=eta_string, meters=str(self), time=str(iter_time), data=str(data_time)
                        )
                    )
            i += 1
            end = time.time()
        total_time = time.time() - start_time
        total_time_str = str(datetime.timedelta(seconds=int(total_time)))
        print(f"{header} Total time: {total_time_str}")


class ExponentialMovingAverage(torch.optim.swa_utils.AveragedModel):
    """Maintains moving averages of model parameters using an exponential decay.
    ``ema_avg = decay * avg_model_param + (1 - decay) * model_param``
    `torch.optim.swa_utils.AveragedModel <https://pytorch.org/docs/stable/optim.html#custom-averaging-strategies>`_
    is used to compute the EMA.
    """

    def __init__(self, model, decay, device="cpu"):
        def ema_avg(avg_model_param, model_param, num_averaged):
            return decay * avg_model_param + (1 - decay) * model_param

        super().__init__(model, device, ema_avg, use_buffers=True)


def accuracy(output, target, topk=(1,)):
    """Computes the accuracy over the k top predictions for the specified values of k"""
    with torch.inference_mode():
        maxk = max(topk)
        batch_size = target.size(0)
        if target.ndim == 2:
            target = target.max(dim=1)[1]

        _, pred = output.topk(maxk, 1, True, True)
        pred = pred.t()
        correct = pred.eq(target[None])

        res = []
        for k in topk:
            correct_k = correct[:k].flatten().sum(dtype=torch.float32)
            res.append(correct_k * (100.0 / batch_size))
        return res


def mkdir(path):
    try:
        os.makedirs(path)
    except OSError as e:
        if e.errno != errno.EEXIST:
            raise


def setup_for_distributed(is_master):
    """
    This function disables printing when not in master process
    """
    import builtins as __builtin__

    builtin_print = __builtin__.print

    def print(*args, **kwargs):
        force = kwargs.pop("force", False)
        if is_master or force:
            builtin_print(*args, **kwargs)

    __builtin__.print = print


def is_dist_avail_and_initialized():
    if not dist.is_available():
        return False
    if not dist.is_initialized():
        return False
    return True


def get_world_size():
    if not is_dist_avail_and_initialized():
        return 1
    return dist.get_world_size()


def get_rank():
    if not is_dist_avail_and_initialized():
        return 0
    return dist.get_rank()


def is_main_process():
    return get_rank() == 0


def save_on_master(*args, **kwargs):
    if is_main_process():
        torch.save(*args, **kwargs)


def init_distributed_mode(args):
    if "RANK" in os.environ and "WORLD_SIZE" in os.environ:
        args.rank = int(os.environ["RANK"])
        args.world_size = int(os.environ["WORLD_SIZE"])
        args.gpu = int(os.environ["LOCAL_RANK"])
    elif "SLURM_PROCID" in os.environ:
        args.rank = int(os.environ["SLURM_PROCID"])
        args.gpu = args.rank % torch.cuda.device_count()
    elif hasattr(args, "rank"):
        pass
    else:
        print("Not using distributed mode")
        args.distributed = False
        return

    args.distributed = True

    torch.cuda.set_device(args.gpu)
    args.dist_backend = "nccl"
    print(f"| distributed init (rank {args.rank}): {args.dist_url}", flush=True)
    torch.distributed.init_process_group(
        backend=args.dist_backend, init_method=args.dist_url, world_size=args.world_size, rank=args.rank
    )
    torch.distributed.barrier()
    setup_for_distributed(args.rank == 0)


def average_checkpoints(inputs):
    """Loads checkpoints from inputs and returns a model with averaged weights. Original implementation taken from:
    https://github.com/pytorch/fairseq/blob/a48f235636557b8d3bc4922a6fa90f3a0fa57955/scripts/average_checkpoints.py#L16

    Args:
      inputs (List[str]): An iterable of string paths of checkpoints to load from.
    Returns:
      A dict of string keys mapping to various values. The 'model' key
      from the returned dict should correspond to an OrderedDict mapping
      string parameter names to torch Tensors.
    """
    params_dict = OrderedDict()
    params_keys = None
    new_state = None
    num_models = len(inputs)
    for fpath in inputs:
        with open(fpath, "rb") as f:
            state = torch.load(
                f,
                map_location=(lambda s, _: torch.serialization.default_restore_location(s, "cpu")),
            )
        # Copies over the settings from the first checkpoint
        if new_state is None:
            new_state = state
        model_params = state["model"]
        model_params_keys = list(model_params.keys())
        if params_keys is None:
            params_keys = model_params_keys
        elif params_keys != model_params_keys:
            raise KeyError(
                f"For checkpoint {f}, expected list of params: {params_keys}, but found: {model_params_keys}"
            )
        for k in params_keys:
            p = model_params[k]
            if isinstance(p, torch.HalfTensor):
                p = p.float()
            if k not in params_dict:
                params_dict[k] = p.clone()
                # NOTE: clone() is needed in case of p is a shared parameter
            else:
                params_dict[k] += p
    averaged_params = OrderedDict()
    for k, v in params_dict.items():
        averaged_params[k] = v
        if averaged_params[k].is_floating_point():
            averaged_params[k].div_(num_models)
        else:
            averaged_params[k] //= num_models
    new_state["model"] = averaged_params
    return new_state


def store_model_weights(model, checkpoint_path, checkpoint_key="model", strict=True):
    """
    This method can be used to prepare weights files for new models. It receives as
    input a model architecture and a checkpoint from the training script and produces
    a file with the weights ready for release.

    Examples:
        from torchvision import models as M

        # Classification
        model = M.mobilenet_v3_large(weights=None)
        print(store_model_weights(model, './class.pth'))

        # Quantized Classification
        model = M.quantization.mobilenet_v3_large(weights=None, quantize=False)
        model.fuse_model(is_qat=True)
        model.qconfig = torch.ao.quantization.get_default_qat_qconfig('qnnpack')
        _ = torch.ao.quantization.prepare_qat(model, inplace=True)
        print(store_model_weights(model, './qat.pth'))

        # Object Detection
        model = M.detection.fasterrcnn_mobilenet_v3_large_fpn(weights=None, weights_backbone=None)
        print(store_model_weights(model, './obj.pth'))

        # Segmentation
        model = M.segmentation.deeplabv3_mobilenet_v3_large(weights=None, weights_backbone=None, aux_loss=True)
        print(store_model_weights(model, './segm.pth', strict=False))

    Args:
        model (pytorch.nn.Module): The model on which the weights will be loaded for validation purposes.
        checkpoint_path (str): The path of the checkpoint we will load.
        checkpoint_key (str, optional): The key of the checkpoint where the model weights are stored.
            Default: "model".
        strict (bool): whether to strictly enforce that the keys
            in :attr:`state_dict` match the keys returned by this module's
            :meth:`~torch.nn.Module.state_dict` function. Default: ``True``

    Returns:
        output_path (str): The location where the weights are saved.
    """
    # Store the new model next to the checkpoint_path
    checkpoint_path = os.path.abspath(checkpoint_path)
    output_dir = os.path.dirname(checkpoint_path)

    # Deep copy to avoid side-effects on the model object.
    model = copy.deepcopy(model)
    checkpoint = torch.load(checkpoint_path, map_location="cpu")

    # Load the weights to the model to validate that everything works
    # and remove unnecessary weights (such as auxiliaries, etc)
    if checkpoint_key == "model_ema":
        del checkpoint[checkpoint_key]["n_averaged"]
        torch.nn.modules.utils.consume_prefix_in_state_dict_if_present(checkpoint[checkpoint_key], "module.")
    model.load_state_dict(checkpoint[checkpoint_key], strict=strict)

    tmp_path = os.path.join(output_dir, str(model.__hash__()))
    torch.save(model.state_dict(), tmp_path)

    sha256_hash = hashlib.sha256()
    with open(tmp_path, "rb") as f:
        # Read and update hash string value in blocks of 4K
        for byte_block in iter(lambda: f.read(4096), b""):
            sha256_hash.update(byte_block)
        hh = sha256_hash.hexdigest()

    output_path = os.path.join(output_dir, "weights-" + str(hh[:8]) + ".pth")
    os.replace(tmp_path, output_path)

    return output_path


def reduce_across_processes(val):
    if not is_dist_avail_and_initialized():
        # nothing to sync, but we still convert to tensor for consistency with the distributed case.
        return torch.tensor(val)

    t = torch.tensor(val, device="cuda")
    dist.barrier()
    dist.all_reduce(t)
    return t


def set_weight_decay(
    model: torch.nn.Module,
    weight_decay: float,
    norm_weight_decay: Optional[float] = None,
    norm_classes: Optional[List[type]] = None,
    custom_keys_weight_decay: Optional[List[Tuple[str, float]]] = None,
):
    if not norm_classes:
        norm_classes = [
            torch.nn.modules.batchnorm._BatchNorm,
            torch.nn.LayerNorm,
            torch.nn.GroupNorm,
            torch.nn.modules.instancenorm._InstanceNorm,
            torch.nn.LocalResponseNorm,
        ]
    norm_classes = tuple(norm_classes)

    params = {
        "other": [],
        "norm": [],
    }
    params_weight_decay = {
        "other": weight_decay,
        "norm": norm_weight_decay,
    }
    custom_keys = []
    if custom_keys_weight_decay is not None:
        for key, weight_decay in custom_keys_weight_decay:
            params[key] = []
            params_weight_decay[key] = weight_decay
            custom_keys.append(key)

    def _add_params(module, prefix=""):
        for name, p in module.named_parameters(recurse=False):
            if not p.requires_grad:
                continue
            is_custom_key = False
            for key in custom_keys:
                target_name = f"{prefix}.{name}" if prefix != "" and "." in key else name
                if key == target_name:
                    params[key].append(p)
                    is_custom_key = True
                    break
            if not is_custom_key:
                if norm_weight_decay is not None and isinstance(module, norm_classes):
                    params["norm"].append(p)
                else:
                    params["other"].append(p)

        for child_name, child_module in module.named_children():
            child_prefix = f"{prefix}.{child_name}" if prefix != "" else child_name
            _add_params(child_module, prefix=child_prefix)

    _add_params(model)

    param_groups = []
    for key in params:
        if len(params[key]) > 0:
            param_groups.append({"params": params[key], "weight_decay": params_weight_decay[key]})
    return param_groups

训练和评估函数，本次分类没有用官方给定的主函数

import datetime
import os
import time
import warnings

import torch
import torch.utils.data
import torchvision
from torch import nn
from torch.utils.data.dataloader import default_collate
from torchvision.transforms.functional import InterpolationMode


def train_one_epoch(model, criterion, optimizer, data_loader, device, epoch, print_freq=10, model_ema=None, scaler=None, clip_grad_norm=None, lr_warmup_epochs=10, model_ema_steps=5):
    model.train()
    metric_logger = MetricLogger(delimiter="  ")
    metric_logger.add_meter("lr", SmoothedValue(window_size=1, fmt="{value}"))
    metric_logger.add_meter("img/s", SmoothedValue(window_size=10, fmt="{value}"))

    header = f"Epoch: [{epoch}]"
    for i, (image, target) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):
        start_time = time.time()
        image, target = image.to(device), target.to(device)
        with torch.cuda.amp.autocast(enabled=scaler is not None):
            output = model(image)
            loss = criterion(output, target)

        optimizer.zero_grad()
        if scaler is not None:
            scaler.scale(loss).backward()
            if clip_grad_norm is not None:
                # we should unscale the gradients of optimizer's assigned params if do gradient clipping
                scaler.unscale_(optimizer)
                nn.utils.clip_grad_norm_(model.parameters(), clip_grad_norm)
            scaler.step(optimizer)
            scaler.update()
        else:
            loss.backward()
            if clip_grad_norm is not None:
                nn.utils.clip_grad_norm_(model.parameters(), clip_grad_norm)
            optimizer.step()

        if model_ema and i % model_ema_steps == 0:
            model_ema.update_parameters(model)
            if epoch < lr_warmup_epochs:
                # Reset ema buffer to keep copying weights during warmup period
                model_ema.n_averaged.fill_(0)

        acc1, acc5 = accuracy(output, target, topk=(1, 5))
        batch_size = image.shape[0]
        metric_logger.update(loss=loss.item(), lr=optimizer.param_groups[0]["lr"])
        metric_logger.meters["acc1"].update(acc1.item(), n=batch_size)
        metric_logger.meters["acc5"].update(acc5.item(), n=batch_size)
        metric_logger.meters["img/s"].update(batch_size / (time.time() - start_time))


def evaluate(model, criterion, data_loader, device, print_freq=100, log_suffix=""):
    model.eval()
    metric_logger = MetricLogger(delimiter="  ")
    header = f"Test: {log_suffix}"

    num_processed_samples = 0
    with torch.inference_mode():
        for image, target in metric_logger.log_every(data_loader, print_freq, header):
            image = image.to(device, non_blocking=True)
            target = target.to(device, non_blocking=True)
            output = model(image)
            loss = criterion(output, target)

            acc1, acc5 = accuracy(output, target, topk=(1, 5))
            # FIXME need to take into account that the datasets
            # could have been padded in distributed setup
            batch_size = image.shape[0]
            metric_logger.update(loss=loss.item())
            metric_logger.meters["acc1"].update(acc1.item(), n=batch_size)
            metric_logger.meters["acc5"].update(acc5.item(), n=batch_size)
            num_processed_samples += batch_size
    # gather the stats from all processes

    num_processed_samples = reduce_across_processes(num_processed_samples)
    if (
        hasattr(data_loader.dataset, "__len__")
        and len(data_loader.dataset) != num_processed_samples
        and torch.distributed.get_rank() == 0
    ):
        # See FIXME above
        warnings.warn(
            f"It looks like the dataset has {len(data_loader.dataset)} samples, but {num_processed_samples} "
            "samples were used for the validation, which might bias the results. "
            "Try adjusting the batch size and / or the world size. "
            "Setting the world size to 1 is always a safe bet."
        )

    metric_logger.synchronize_between_processes()

    print(f"{header} Acc@1 {metric_logger.acc1.global_avg:.3f} Acc@5 {metric_logger.acc5.global_avg:.3f}")
    return metric_logger.acc1.global_avg


def _get_cache_path(filepath):
    import hashlib

    h = hashlib.sha1(filepath.encode()).hexdigest()
    cache_path = os.path.join("~", ".torch", "vision", "datasets", "imagefolder", h[:10] + ".pt")
    cache_path = os.path.expanduser(cache_path)
    return cache_path


def load_data(traindir, valdir, args):
    # Data loading code
    print("Loading data")
    val_resize_size, val_crop_size, train_crop_size = (
        args.val_resize_size,
        args.val_crop_size,
        args.train_crop_size,
    )
    interpolation = InterpolationMode(args.interpolation)

    print("Loading training data")
    st = time.time()
    cache_path = _get_cache_path(traindir)
    if args.cache_dataset and os.path.exists(cache_path):
        # Attention, as the transforms are also cached!
        print(f"Loading dataset_train from {cache_path}")
        dataset, _ = torch.load(cache_path)
    else:
        auto_augment_policy = getattr(args, "auto_augment", None)
        random_erase_prob = getattr(args, "random_erase", 0.0)
        ra_magnitude = args.ra_magnitude
        augmix_severity = args.augmix_severity
        dataset = torchvision.datasets.ImageFolder(
            traindir,
            ClassificationPresetTrain(
                crop_size=train_crop_size,
                interpolation=interpolation,
                auto_augment_policy=auto_augment_policy,
                random_erase_prob=random_erase_prob,
                ra_magnitude=ra_magnitude,
                augmix_severity=augmix_severity,
            ),
        )
        if args.cache_dataset:
            print(f"Saving dataset_train to {cache_path}")
            mkdir(os.path.dirname(cache_path))
            save_on_master((dataset, traindir), cache_path)
    print("Took", time.time() - st)

    print("Loading validation data")
    cache_path = _get_cache_path(valdir)
    if args.cache_dataset and os.path.exists(cache_path):
        # Attention, as the transforms are also cached!
        print(f"Loading dataset_test from {cache_path}")
        dataset_test, _ = torch.load(cache_path)
    else:
        if args.weights and args.test_only:
            weights = torchvision.models.get_weight(args.weights)
            preprocessing = weights.transforms()
        else:
            preprocessing = ClassificationPresetEval(
                crop_size=val_crop_size, resize_size=val_resize_size, interpolation=interpolation
            )

        dataset_test = torchvision.datasets.ImageFolder(
            valdir,
            preprocessing,
        )
        if args.cache_dataset:
            print(f"Saving dataset_test to {cache_path}")
            mkdir(os.path.dirname(cache_path))
            save_on_master((dataset_test, valdir), cache_path)

    print("Creating data loaders")
    if args.distributed:
        if hasattr(args, "ra_sampler") and args.ra_sampler:
            train_sampler = RASampler(dataset, shuffle=True, repetitions=args.ra_reps)
        else:
            train_sampler = torch.utils.data.distributed.DistributedSampler(dataset)
        test_sampler = torch.utils.data.distributed.DistributedSampler(dataset_test, shuffle=False)
    else:
        train_sampler = torch.utils.data.RandomSampler(dataset)
        test_sampler = torch.utils.data.SequentialSampler(dataset_test)

    return dataset, dataset_test, train_sampler, test_sampler


def main(args):
    if args.output_dir:
        mkdir(args.output_dir)

    init_distributed_mode(args)
    print(args)

    device = torch.device(args.device)

    if args.use_deterministic_algorithms:
        torch.backends.cudnn.benchmark = False
        torch.use_deterministic_algorithms(True)
    else:
        torch.backends.cudnn.benchmark = True

    train_dir = os.path.join(args.data_path, "train")
    val_dir = os.path.join(args.data_path, "val")
    dataset, dataset_test, train_sampler, test_sampler = load_data(train_dir, val_dir, args)

    collate_fn = None
    num_classes = len(dataset.classes)
    mixup_transforms = []
    if args.mixup_alpha > 0.0:
        mixup_transforms.append(RandomMixup(num_classes, p=1.0, alpha=args.mixup_alpha))
    if args.cutmix_alpha > 0.0:
        mixup_transforms.append(RandomCutmix(num_classes, p=1.0, alpha=args.cutmix_alpha))
    if mixup_transforms:
        mixupcutmix = torchvision.transforms.RandomChoice(mixup_transforms)

        def collate_fn(batch):
            return mixupcutmix(*default_collate(batch))

    data_loader = torch.utils.data.DataLoader(
        dataset,
        batch_size=args.batch_size,
        sampler=train_sampler,
        num_workers=args.workers,
        pin_memory=True,
        collate_fn=collate_fn,
    )
    data_loader_test = torch.utils.data.DataLoader(
        dataset_test, batch_size=args.batch_size, sampler=test_sampler, num_workers=args.workers, pin_memory=True
    )

    print("Creating model")
    model = torchvision.models.get_model(args.model, weights=args.weights, num_classes=num_classes)
    model.to(device)

    if args.distributed and args.sync_bn:
        model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model)

    criterion = nn.CrossEntropyLoss(label_smoothing=args.label_smoothing)

    custom_keys_weight_decay = []
    if args.bias_weight_decay is not None:
        custom_keys_weight_decay.append(("bias", args.bias_weight_decay))
    if args.transformer_embedding_decay is not None:
        for key in ["class_token", "position_embedding", "relative_position_bias_table"]:
            custom_keys_weight_decay.append((key, args.transformer_embedding_decay))
    parameters = set_weight_decay(
        model,
        args.weight_decay,
        norm_weight_decay=args.norm_weight_decay,
        custom_keys_weight_decay=custom_keys_weight_decay if len(custom_keys_weight_decay) > 0 else None,
    )

    opt_name = args.opt.lower()
    if opt_name.startswith("sgd"):
        optimizer = torch.optim.SGD(
            parameters,
            lr=args.lr,
            momentum=args.momentum,
            weight_decay=args.weight_decay,
            nesterov="nesterov" in opt_name,
        )
    elif opt_name == "rmsprop":
        optimizer = torch.optim.RMSprop(
            parameters, lr=args.lr, momentum=args.momentum, weight_decay=args.weight_decay, eps=0.0316, alpha=0.9
        )
    elif opt_name == "adamw":
        optimizer = torch.optim.AdamW(parameters, lr=args.lr, weight_decay=args.weight_decay)
    else:
        raise RuntimeError(f"Invalid optimizer {args.opt}. Only SGD, RMSprop and AdamW are supported.")

    scaler = torch.cuda.amp.GradScaler() if args.amp else None

    args.lr_scheduler = args.lr_scheduler.lower()
    if args.lr_scheduler == "steplr":
        main_lr_scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=args.lr_step_size, gamma=args.lr_gamma)
    elif args.lr_scheduler == "cosineannealinglr":
        main_lr_scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(
            optimizer, T_max=args.epochs - args.lr_warmup_epochs, eta_min=args.lr_min
        )
    elif args.lr_scheduler == "exponentiallr":
        main_lr_scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=args.lr_gamma)
    else:
        raise RuntimeError(
            f"Invalid lr scheduler '{args.lr_scheduler}'. Only StepLR, CosineAnnealingLR and ExponentialLR "
            "are supported."
        )

    if args.lr_warmup_epochs > 0:
        if args.lr_warmup_method == "linear":
            warmup_lr_scheduler = torch.optim.lr_scheduler.LinearLR(
                optimizer, start_factor=args.lr_warmup_decay, total_iters=args.lr_warmup_epochs
            )
        elif args.lr_warmup_method == "constant":
            warmup_lr_scheduler = torch.optim.lr_scheduler.ConstantLR(
                optimizer, factor=args.lr_warmup_decay, total_iters=args.lr_warmup_epochs
            )
        else:
            raise RuntimeError(
                f"Invalid warmup lr method '{args.lr_warmup_method}'. Only linear and constant are supported."
            )
        lr_scheduler = torch.optim.lr_scheduler.SequentialLR(
            optimizer, schedulers=[warmup_lr_scheduler, main_lr_scheduler], milestones=[args.lr_warmup_epochs]
        )
    else:
        lr_scheduler = main_lr_scheduler

    model_without_ddp = model
    if args.distributed:
        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu])
        model_without_ddp = model.module

    model_ema = None
    if args.model_ema:
        # Decay adjustment that aims to keep the decay independent from other hyper-parameters originally proposed at:
        # https://github.com/facebookresearch/pycls/blob/f8cd9627/pycls/core/net.py#L123
        #
        # total_ema_updates = (Dataset_size / n_GPUs) * epochs / (batch_size_per_gpu * EMA_steps)
        # We consider constant = Dataset_size for a given dataset/setup and ommit it. Thus:
        # adjust = 1 / total_ema_updates ~= n_GPUs * batch_size_per_gpu * EMA_steps / epochs
        adjust = args.world_size * args.batch_size * args.model_ema_steps / args.epochs
        alpha = 1.0 - args.model_ema_decay
        alpha = min(1.0, alpha * adjust)
        model_ema = ExponentialMovingAverage(model_without_ddp, device=device, decay=1.0 - alpha)

    if args.resume:
        checkpoint = torch.load(args.resume, map_location="cpu")
        model_without_ddp.load_state_dict(checkpoint["model"])
        if not args.test_only:
            optimizer.load_state_dict(checkpoint["optimizer"])
            lr_scheduler.load_state_dict(checkpoint["lr_scheduler"])
        args.start_epoch = checkpoint["epoch"] + 1
        if model_ema:
            model_ema.load_state_dict(checkpoint["model_ema"])
        if scaler:
            scaler.load_state_dict(checkpoint["scaler"])

    if args.test_only:
        # We disable the cudnn benchmarking because it can noticeably affect the accuracy
        torch.backends.cudnn.benchmark = False
        torch.backends.cudnn.deterministic = True
        if model_ema:
            evaluate(model_ema, criterion, data_loader_test, device=device, log_suffix="EMA")
        else:
            evaluate(model, criterion, data_loader_test, device=device)
        return

    print("Start training")
    start_time = time.time()
    for epoch in range(args.start_epoch, args.epochs):
        if args.distributed:
            train_sampler.set_epoch(epoch)
        train_one_epoch(model, criterion, optimizer, data_loader, device, epoch, args, model_ema, scaler)
        lr_scheduler.step()
        evaluate(model, criterion, data_loader_test, device=device)
        if model_ema:
            evaluate(model_ema, criterion, data_loader_test, device=device, log_suffix="EMA")
        if args.output_dir:
            checkpoint = {
                "model": model_without_ddp.state_dict(),
                "optimizer": optimizer.state_dict(),
                "lr_scheduler": lr_scheduler.state_dict(),
                "epoch": epoch,
                "args": args,
            }
            if model_ema:
                checkpoint["model_ema"] = model_ema.state_dict()
            if scaler:
                checkpoint["scaler"] = scaler.state_dict()
            save_on_master(checkpoint, os.path.join(args.output_dir, f"model_{epoch}.pth"))
            save_on_master(checkpoint, os.path.join(args.output_dir, "checkpoint.pth"))

    total_time = time.time() - start_time
    total_time_str = str(datetime.timedelta(seconds=int(total_time)))
    print(f"Training time {total_time_str}")

训练和评估

from torchvision.transforms import Compose, ToTensor, PILToTensor

data_train = MNIST('.data/mnist', train=True, download=True, transform=ToTensor())
data_val = MNIST('./data/mnist', train=False, download=True, transform=ToTensor())
num_classes = len(data_train.classes)
sampler_train = torch.utils.data.RandomSampler(data_train, replacement=False)
data_loader_train = torch.utils.data.DataLoader(
    data_train, sampler=sampler_train,
    batch_size=128,
    num_workers=2,
    pin_memory=True,
    drop_last=True,  # 当不足一个batch时候，使丢弃后面一部分还是随机增加一部分样本
)
sampler_val = torch.utils.data.SequentialSampler(data_val)
data_loader_val = torch.utils.data.DataLoader(
    data_val, sampler=sampler_val,
    batch_size=128,
    shuffle=False,
    num_workers=2,
    pin_memory=True,
    drop_last=False
)
# timm库里面自带的mixup，0.8的MIXUP和0.8的CUTMIX，此处没用到
mixup_fn = Mixup(
    # mixup=0.8,cutmix=0.8,cutmix_minmax is None,mixup_prob=1, mixup_switch_prob=0.5, mixup_mode='batch', label_smoothing=0.1
    mixup_alpha=0.8, cutmix_alpha=0.8, prob=0.8, switch_prob=0.5, mode='batch',
    label_smoothing=0.1, num_classes=num_classes)

device = 'cuda' if torch.cuda.is_available() else 'cpu'
model = SwinTransformer(img_size=28, patch_size=7, in_chans=1, num_classes=10, embed_dim=96, depths=[2, 2, 4], num_heads=[3, 6, 12], window_size=2, mlp_ratio=4., qkv_bias=True, qk_scale=True, drop_rate=0., drop_path_rate=0., ape=False, patch_norm=True, use_checkpoint=False, fused_window_process=False).to(device)
optimizer = Nadam(model.parameters())
loss_fn = LabelSmoothingCrossEntropy()
lr_scheduler = CosineLRScheduler(optimizer, t_initial=15)

print("Start training")
start_time = time.time()
for epoch in range(100):
    train_one_epoch(model, loss_fn, optimizer, data_loader_train, device, epoch)
    lr_scheduler.step(epoch)
    evaluate(model, loss_fn, data_loader_val, device='cuda')
total_time = time.time() - start_time
total_time_str = str(datetime.timedelta(seconds=int(total_time)))
print(f"Training time {total_time_str}")

Start training
Epoch: [0]  [  0/468]  eta: 0:53:11  lr: 0.002  img/s: 19.040995153557684  loss: 2.3678 (2.3678)  acc1: 8.5938 (8.5938)  acc5: 49.2188 (49.2188)  time: 6.8184  data: 0.0960  max mem: 138
Epoch: [0]  [ 10/468]  eta: 0:05:01  lr: 0.002  img/s: 3501.2711431105545  loss: 2.6789 (2.8085)  acc1: 11.7188 (12.9972)  acc5: 52.3438 (52.9830)  time: 0.6580  data: 0.0090  max mem: 190
Epoch: [0]  [ 20/468]  eta: 0:02:43  lr: 0.002  img/s: 3585.1146043405674  loss: 2.4055 (2.5948)  acc1: 11.7188 (12.7232)  acc5: 53.9062 (56.5848)  time: 0.0412  data: 0.0003  max mem: 190
Epoch: [0]  [ 30/468]  eta: 0:01:53  lr: 0.002  img/s: 3649.6010441592343  loss: 2.2687 (2.4918)  acc1: 12.5000 (13.7853)  acc5: 64.0625 (59.5010)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [0]  [ 40/468]  eta: 0:01:27  lr: 0.002  img/s: 3546.464652336473  loss: 2.2370 (2.4254)  acc1: 17.1875 (14.8438)  acc5: 67.1875 (61.1852)  time: 0.0361  data: 0.0003  max mem: 190
Epoch: [0]  [ 50/468]  eta: 0:01:11  lr: 0.002  img/s: 3591.8786094682473  loss: 2.2039 (2.3840)  acc1: 17.1875 (15.4105)  acc5: 67.1875 (62.3775)  time: 0.0362  data: 0.0002  max mem: 190
Epoch: [0]  [ 60/468]  eta: 0:01:00  lr: 0.002  img/s: 3516.176414340542  loss: 2.0975 (2.3266)  acc1: 21.0938 (17.5717)  acc5: 74.2188 (65.4201)  time: 0.0363  data: 0.0002  max mem: 190
Epoch: [0]  [ 70/468]  eta: 0:00:53  lr: 0.002  img/s: 3595.655457400995  loss: 1.9138 (2.2534)  acc1: 32.8125 (20.4005)  acc5: 88.2812 (69.0361)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [0]  [ 80/468]  eta: 0:00:47  lr: 0.002  img/s: 3523.9311585165738  loss: 1.6661 (2.1731)  acc1: 46.0938 (24.2091)  acc5: 92.9688 (72.1065)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [0]  [ 90/468]  eta: 0:00:42  lr: 0.002  img/s: 3560.978423374125  loss: 1.5143 (2.0945)  acc1: 54.6875 (27.6700)  acc5: 94.5312 (74.6738)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [0]  [100/468]  eta: 0:00:38  lr: 0.002  img/s: 3397.5516052070348  loss: 1.3297 (2.0148)  acc1: 61.7188 (31.3738)  acc5: 96.8750 (76.8487)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [0]  [110/468]  eta: 0:00:35  lr: 0.002  img/s: 3430.11246062728  loss: 1.2297 (1.9384)  acc1: 65.6250 (34.9733)  acc5: 97.6562 (78.7796)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [0]  [120/468]  eta: 0:00:32  lr: 0.002  img/s: 3561.07290346973  loss: 1.1007 (1.8659)  acc1: 75.0000 (38.4362)  acc5: 98.4375 (80.3525)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [0]  [130/468]  eta: 0:00:30  lr: 0.002  img/s: 3507.3555366825635  loss: 0.9622 (1.7933)  acc1: 80.4688 (41.7820)  acc5: 99.2188 (81.7987)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [0]  [140/468]  eta: 0:00:27  lr: 0.002  img/s: 3631.1864186675684  loss: 0.9119 (1.7340)  acc1: 82.0312 (44.5867)  acc5: 99.2188 (83.0286)  time: 0.0364  data: 0.0003  max mem: 190
Epoch: [0]  [150/468]  eta: 0:00:26  lr: 0.002  img/s: 3552.660252253206  loss: 0.8685 (1.6751)  acc1: 83.5938 (47.2941)  acc5: 99.2188 (84.1060)  time: 0.0363  data: 0.0003  max mem: 190
Epoch: [0]  [160/468]  eta: 0:00:24  lr: 0.002  img/s: 3542.01905365107  loss: 0.8460 (1.6237)  acc1: 86.7188 (49.7234)  acc5: 99.2188 (85.0398)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [0]  [170/468]  eta: 0:00:22  lr: 0.002  img/s: 3549.3250826391645  loss: 0.8027 (1.5745)  acc1: 89.0625 (52.0514)  acc5: 99.2188 (85.8872)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [0]  [180/468]  eta: 0:00:21  lr: 0.002  img/s: 3582.865593550626  loss: 0.7574 (1.5300)  acc1: 89.0625 (54.1523)  acc5: 99.2188 (86.6238)  time: 0.0364  data: 0.0002  max mem: 190
Epoch: [0]  [190/468]  eta: 0:00:20  lr: 0.002  img/s: 3565.306025952637  loss: 0.7399 (1.4878)  acc1: 89.8438 (56.1109)  acc5: 99.2188 (87.3037)  time: 0.0364  data: 0.0002  max mem: 190
Epoch: [0]  [200/468]  eta: 0:00:18  lr: 0.002  img/s: 3605.6771975069846  loss: 0.7423 (1.4527)  acc1: 89.8438 (57.7231)  acc5: 99.2188 (87.9120)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [0]  [210/468]  eta: 0:00:17  lr: 0.002  img/s: 3401.2969342954707  loss: 0.7311 (1.4173)  acc1: 90.6250 (59.3676)  acc5: 99.2188 (88.4590)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [0]  [220/468]  eta: 0:00:16  lr: 0.002  img/s: 3480.7953422632554  loss: 0.7156 (1.3896)  acc1: 91.4062 (60.6582)  acc5: 99.2188 (88.9352)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [0]  [230/468]  eta: 0:00:15  lr: 0.002  img/s: 3562.7034746370077  loss: 0.7577 (1.3609)  acc1: 89.8438 (61.9690)  acc5: 99.2188 (89.3939)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [0]  [240/468]  eta: 0:00:14  lr: 0.002  img/s: 3469.951602895553  loss: 0.6970 (1.3335)  acc1: 92.1875 (63.1905)  acc5: 100.0000 (89.8178)  time: 0.0375  data: 0.0002  max mem: 190
Epoch: [0]  [250/468]  eta: 0:00:14  lr: 0.002  img/s: 2312.135436719682  loss: 0.6869 (1.3075)  acc1: 92.9688 (64.3800)  acc5: 100.0000 (90.2141)  time: 0.0424  data: 0.0008  max mem: 190
Epoch: [0]  [260/468]  eta: 0:00:13  lr: 0.002  img/s: 3446.49529764465  loss: 0.6936 (1.2856)  acc1: 92.9688 (65.3766)  acc5: 100.0000 (90.5741)  time: 0.0492  data: 0.0031  max mem: 190
Epoch: [0]  [270/468]  eta: 0:00:12  lr: 0.002  img/s: 3455.8797038944317  loss: 0.6986 (1.2633)  acc1: 91.4062 (66.3947)  acc5: 100.0000 (90.9162)  time: 0.0451  data: 0.0025  max mem: 190
Epoch: [0]  [280/468]  eta: 0:00:11  lr: 0.002  img/s: 3372.855566863935  loss: 0.6628 (1.2424)  acc1: 92.9688 (67.3349)  acc5: 100.0000 (91.2339)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [0]  [290/468]  eta: 0:00:10  lr: 0.002  img/s: 3639.0379784587644  loss: 0.6992 (1.2249)  acc1: 92.1875 (68.1352)  acc5: 100.0000 (91.5190)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [0]  [300/468]  eta: 0:00:10  lr: 0.002  img/s: 3276.9800098882383  loss: 0.7055 (1.2073)  acc1: 91.4062 (68.9369)  acc5: 100.0000 (91.7956)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [0]  [310/468]  eta: 0:00:09  lr: 0.002  img/s: 3596.450327576736  loss: 0.6860 (1.1903)  acc1: 92.1875 (69.7171)  acc5: 100.0000 (92.0418)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [0]  [320/468]  eta: 0:00:08  lr: 0.002  img/s: 3440.2452452965604  loss: 0.6749 (1.1755)  acc1: 92.9688 (70.3831)  acc5: 100.0000 (92.2873)  time: 0.0365  data: 0.0003  max mem: 190
Epoch: [0]  [330/468]  eta: 0:00:08  lr: 0.002  img/s: 3599.681596309607  loss: 0.6807 (1.1608)  acc1: 92.9688 (71.0583)  acc5: 100.0000 (92.5038)  time: 0.0365  data: 0.0003  max mem: 190
Epoch: [0]  [340/468]  eta: 0:00:07  lr: 0.002  img/s: 3264.029960907339  loss: 0.6687 (1.1463)  acc1: 93.7500 (71.7238)  acc5: 100.0000 (92.7144)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [0]  [350/468]  eta: 0:00:06  lr: 0.002  img/s: 2509.434432857657  loss: 0.6476 (1.1321)  acc1: 93.7500 (72.3669)  acc5: 100.0000 (92.9153)  time: 0.0386  data: 0.0006  max mem: 190
Epoch: [0]  [360/468]  eta: 0:00:06  lr: 0.002  img/s: 2476.787392566006  loss: 0.6356 (1.1186)  acc1: 93.7500 (72.9787)  acc5: 100.0000 (93.1073)  time: 0.0489  data: 0.0037  max mem: 190
Epoch: [0]  [370/468]  eta: 0:00:05  lr: 0.002  img/s: 2908.151346900747  loss: 0.6492 (1.1062)  acc1: 93.7500 (73.5428)  acc5: 100.0000 (93.2867)  time: 0.0485  data: 0.0035  max mem: 190
Epoch: [0]  [380/468]  eta: 0:00:04  lr: 0.002  img/s: 3440.5098049268154  loss: 0.6544 (1.0938)  acc1: 93.7500 (74.0957)  acc5: 100.0000 (93.4568)  time: 0.0406  data: 0.0004  max mem: 190
Epoch: [0]  [390/468]  eta: 0:00:04  lr: 0.002  img/s: 3392.3991482209317  loss: 0.6606 (1.0837)  acc1: 92.9688 (74.5524)  acc5: 100.0000 (93.6141)  time: 0.0395  data: 0.0003  max mem: 190
Epoch: [0]  [400/468]  eta: 0:00:03  lr: 0.002  img/s: 3594.0427104392884  loss: 0.6609 (1.0728)  acc1: 93.7500 (75.0487)  acc5: 100.0000 (93.7695)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [0]  [410/468]  eta: 0:00:03  lr: 0.002  img/s: 3482.601694365521  loss: 0.6539 (1.0631)  acc1: 92.9688 (75.4733)  acc5: 100.0000 (93.9154)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [0]  [420/468]  eta: 0:00:02  lr: 0.002  img/s: 3460.6245576490455  loss: 0.6279 (1.0527)  acc1: 95.3125 (75.9538)  acc5: 100.0000 (94.0525)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [0]  [430/468]  eta: 0:00:02  lr: 0.002  img/s: 3554.1184196589343  loss: 0.6099 (1.0425)  acc1: 95.3125 (76.4084)  acc5: 100.0000 (94.1905)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [0]  [440/468]  eta: 0:00:01  lr: 0.002  img/s: 3696.159833667238  loss: 0.6209 (1.0333)  acc1: 94.5312 (76.8282)  acc5: 100.0000 (94.3187)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [0]  [450/468]  eta: 0:00:00  lr: 0.002  img/s: 3577.42225065302  loss: 0.6355 (1.0245)  acc1: 94.5312 (77.2208)  acc5: 100.0000 (94.4429)  time: 0.0362  data: 0.0002  max mem: 190
Epoch: [0]  [460/468]  eta: 0:00:00  lr: 0.002  img/s: 3550.005699889573  loss: 0.6098 (1.0152)  acc1: 95.3125 (77.6318)  acc5: 100.0000 (94.5618)  time: 0.0362  data: 0.0002  max mem: 190
Epoch: [0] Total time: 0:00:24
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5895 (0.5895)  acc1: 96.0938 (96.0938)  acc5: 100.0000 (100.0000)  time: 0.1171  data: 0.0978  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 95.010 Acc@5 99.930
Epoch: [1]  [  0/468]  eta: 0:01:13  lr: 0.002  img/s: 2390.055122491953  loss: 0.6364 (0.6364)  acc1: 94.5312 (94.5312)  acc5: 100.0000 (100.0000)  time: 0.1565  data: 0.1029  max mem: 190
Epoch: [1]  [ 10/468]  eta: 0:00:21  lr: 0.002  img/s: 3516.7061566980865  loss: 0.5902 (0.6113)  acc1: 96.0938 (95.3125)  acc5: 100.0000 (99.9290)  time: 0.0476  data: 0.0096  max mem: 190
Epoch: [1]  [ 20/468]  eta: 0:00:18  lr: 0.002  img/s: 3492.6156808660126  loss: 0.6155 (0.6309)  acc1: 94.5312 (94.4568)  acc5: 100.0000 (99.8140)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [1]  [ 30/468]  eta: 0:00:21  lr: 0.002  img/s: 1832.5428361749698  loss: 0.6361 (0.6283)  acc1: 94.5312 (94.7329)  acc5: 100.0000 (99.8236)  time: 0.0500  data: 0.0016  max mem: 190
Epoch: [1]  [ 40/468]  eta: 0:00:20  lr: 0.002  img/s: 3413.4288221156903  loss: 0.6217 (0.6312)  acc1: 94.5312 (94.6837)  acc5: 100.0000 (99.8285)  time: 0.0549  data: 0.0024  max mem: 190
Epoch: [1]  [ 50/468]  eta: 0:00:19  lr: 0.002  img/s: 3562.2306916503  loss: 0.6316 (0.6319)  acc1: 94.5312 (94.5925)  acc5: 100.0000 (99.7702)  time: 0.0420  data: 0.0011  max mem: 190
Epoch: [1]  [ 60/468]  eta: 0:00:18  lr: 0.002  img/s: 3474.66773671607  loss: 0.6233 (0.6315)  acc1: 94.5312 (94.6337)  acc5: 100.0000 (99.8079)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [1]  [ 70/468]  eta: 0:00:17  lr: 0.002  img/s: 3406.649398775342  loss: 0.6054 (0.6279)  acc1: 95.3125 (94.8393)  acc5: 100.0000 (99.8349)  time: 0.0382  data: 0.0002  max mem: 190
Epoch: [1]  [ 80/468]  eta: 0:00:16  lr: 0.002  img/s: 3441.9214771124502  loss: 0.6031 (0.6286)  acc1: 95.3125 (94.7531)  acc5: 100.0000 (99.8553)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [1]  [ 90/468]  eta: 0:00:16  lr: 0.002  img/s: 3423.2448431751377  loss: 0.6496 (0.6324)  acc1: 94.5312 (94.5999)  acc5: 100.0000 (99.8369)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [1]  [100/468]  eta: 0:00:15  lr: 0.002  img/s: 3509.3960164987807  loss: 0.6414 (0.6306)  acc1: 93.7500 (94.6473)  acc5: 100.0000 (99.8453)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [1]  [110/468]  eta: 0:00:14  lr: 0.002  img/s: 3536.1166606290135  loss: 0.6056 (0.6282)  acc1: 95.3125 (94.7494)  acc5: 100.0000 (99.8592)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [1]  [120/468]  eta: 0:00:14  lr: 0.002  img/s: 3551.7201338996283  loss: 0.6056 (0.6265)  acc1: 96.0938 (94.8670)  acc5: 100.0000 (99.8580)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [1]  [130/468]  eta: 0:00:13  lr: 0.002  img/s: 3513.990038028289  loss: 0.6123 (0.6287)  acc1: 95.3125 (94.7758)  acc5: 100.0000 (99.8628)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [1]  [140/468]  eta: 0:00:13  lr: 0.002  img/s: 2365.0075857025804  loss: 0.6116 (0.6272)  acc1: 95.3125 (94.8637)  acc5: 100.0000 (99.8670)  time: 0.0377  data: 0.0002  max mem: 190
Epoch: [1]  [150/468]  eta: 0:00:13  lr: 0.002  img/s: 2116.48142804204  loss: 0.6052 (0.6264)  acc1: 96.0938 (94.9141)  acc5: 100.0000 (99.8707)  time: 0.0510  data: 0.0014  max mem: 190
Epoch: [1]  [160/468]  eta: 0:00:13  lr: 0.002  img/s: 2346.6791619860214  loss: 0.6072 (0.6246)  acc1: 95.3125 (94.9874)  acc5: 100.0000 (99.8690)  time: 0.0622  data: 0.0044  max mem: 190
Epoch: [1]  [170/468]  eta: 0:00:12  lr: 0.002  img/s: 3253.033634881875  loss: 0.5855 (0.6233)  acc1: 95.3125 (95.0612)  acc5: 100.0000 (99.8629)  time: 0.0514  data: 0.0036  max mem: 190
Epoch: [1]  [180/468]  eta: 0:00:12  lr: 0.002  img/s: 3258.898336773097  loss: 0.6087 (0.6239)  acc1: 95.3125 (95.0233)  acc5: 100.0000 (99.8576)  time: 0.0408  data: 0.0006  max mem: 190
Epoch: [1]  [190/468]  eta: 0:00:11  lr: 0.002  img/s: 3504.1734623945067  loss: 0.6345 (0.6238)  acc1: 94.5312 (95.0344)  acc5: 100.0000 (99.8527)  time: 0.0392  data: 0.0003  max mem: 190
Epoch: [1]  [200/468]  eta: 0:00:11  lr: 0.002  img/s: 3558.052024998509  loss: 0.6168 (0.6241)  acc1: 95.3125 (95.0249)  acc5: 100.0000 (99.8445)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [1]  [210/468]  eta: 0:00:10  lr: 0.002  img/s: 3303.73967410033  loss: 0.6099 (0.6240)  acc1: 95.3125 (95.0163)  acc5: 100.0000 (99.8408)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [1]  [220/468]  eta: 0:00:10  lr: 0.002  img/s: 3516.1533856844394  loss: 0.6165 (0.6246)  acc1: 94.5312 (94.9625)  acc5: 100.0000 (99.8374)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [1]  [230/468]  eta: 0:00:09  lr: 0.002  img/s: 3532.231380598979  loss: 0.6224 (0.6240)  acc1: 94.5312 (95.0081)  acc5: 100.0000 (99.8444)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [1]  [240/468]  eta: 0:00:09  lr: 0.002  img/s: 3485.9936626668746  loss: 0.6139 (0.6245)  acc1: 95.3125 (94.9721)  acc5: 100.0000 (99.8412)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [1]  [250/468]  eta: 0:00:09  lr: 0.002  img/s: 3558.971905866755  loss: 0.6313 (0.6250)  acc1: 95.3125 (94.9608)  acc5: 100.0000 (99.8381)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [1]  [260/468]  eta: 0:00:08  lr: 0.002  img/s: 3356.2403070729297  loss: 0.6108 (0.6250)  acc1: 95.3125 (94.9683)  acc5: 100.0000 (99.8414)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [1]  [270/468]  eta: 0:00:08  lr: 0.002  img/s: 3548.9027618027735  loss: 0.6029 (0.6244)  acc1: 96.0938 (94.9983)  acc5: 100.0000 (99.8443)  time: 0.0386  data: 0.0003  max mem: 190
Epoch: [1]  [280/468]  eta: 0:00:07  lr: 0.002  img/s: 3527.7286478388287  loss: 0.6134 (0.6240)  acc1: 96.0938 (95.0317)  acc5: 100.0000 (99.8443)  time: 0.0386  data: 0.0003  max mem: 190
Epoch: [1]  [290/468]  eta: 0:00:07  lr: 0.002  img/s: 3131.92185230342  loss: 0.6139 (0.6236)  acc1: 95.3125 (95.0279)  acc5: 100.0000 (99.8443)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [1]  [300/468]  eta: 0:00:06  lr: 0.002  img/s: 2843.4905061782665  loss: 0.6021 (0.6229)  acc1: 95.3125 (95.0607)  acc5: 100.0000 (99.8469)  time: 0.0470  data: 0.0026  max mem: 190
Epoch: [1]  [310/468]  eta: 0:00:06  lr: 0.002  img/s: 3476.4452214905036  loss: 0.6042 (0.6231)  acc1: 95.3125 (95.0437)  acc5: 100.0000 (99.8493)  time: 0.0499  data: 0.0033  max mem: 190
Epoch: [1]  [320/468]  eta: 0:00:06  lr: 0.002  img/s: 3499.6506808685394  loss: 0.6194 (0.6227)  acc1: 95.3125 (95.0594)  acc5: 100.0000 (99.8491)  time: 0.0403  data: 0.0009  max mem: 190
Epoch: [1]  [330/468]  eta: 0:00:05  lr: 0.002  img/s: 3577.5414448213132  loss: 0.6167 (0.6226)  acc1: 95.3125 (95.0600)  acc5: 100.0000 (99.8513)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [1]  [340/468]  eta: 0:00:05  lr: 0.002  img/s: 3553.3421493292035  loss: 0.6114 (0.6226)  acc1: 94.5312 (95.0536)  acc5: 100.0000 (99.8534)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [1]  [350/468]  eta: 0:00:04  lr: 0.002  img/s: 3620.0947519605133  loss: 0.6088 (0.6230)  acc1: 94.5312 (95.0521)  acc5: 100.0000 (99.8442)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [1]  [360/468]  eta: 0:00:04  lr: 0.002  img/s: 3531.1394576391585  loss: 0.6052 (0.6223)  acc1: 96.0938 (95.0918)  acc5: 100.0000 (99.8442)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [1]  [370/468]  eta: 0:00:03  lr: 0.002  img/s: 3390.620891751926  loss: 0.5937 (0.6220)  acc1: 96.0938 (95.0998)  acc5: 100.0000 (99.8484)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [1]  [380/468]  eta: 0:00:03  lr: 0.002  img/s: 3559.184253618048  loss: 0.5937 (0.6215)  acc1: 95.3125 (95.1013)  acc5: 100.0000 (99.8483)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [1]  [390/468]  eta: 0:00:03  lr: 0.002  img/s: 3451.502838370396  loss: 0.6005 (0.6211)  acc1: 96.0938 (95.1247)  acc5: 100.0000 (99.8461)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [1]  [400/468]  eta: 0:00:02  lr: 0.002  img/s: 3515.7389214498544  loss: 0.6005 (0.6209)  acc1: 95.3125 (95.1391)  acc5: 100.0000 (99.8402)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [1]  [410/468]  eta: 0:00:02  lr: 0.002  img/s: 3433.4687779795863  loss: 0.6419 (0.6247)  acc1: 94.5312 (94.9856)  acc5: 100.0000 (99.8270)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [1]  [420/468]  eta: 0:00:01  lr: 0.002  img/s: 3529.1896162973385  loss: 0.6978 (0.6257)  acc1: 90.6250 (94.9358)  acc5: 100.0000 (99.8256)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [1]  [430/468]  eta: 0:00:01  lr: 0.002  img/s: 3377.545010160236  loss: 0.6389 (0.6257)  acc1: 94.5312 (94.9355)  acc5: 100.0000 (99.8242)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [1]  [440/468]  eta: 0:00:01  lr: 0.002  img/s: 3556.0723573089226  loss: 0.6291 (0.6259)  acc1: 94.5312 (94.9157)  acc5: 100.0000 (99.8228)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [1]  [450/468]  eta: 0:00:00  lr: 0.002  img/s: 3517.881372369145  loss: 0.6064 (0.6255)  acc1: 95.3125 (94.9349)  acc5: 100.0000 (99.8268)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [1]  [460/468]  eta: 0:00:00  lr: 0.002  img/s: 3552.3546592029434  loss: 0.6011 (0.6259)  acc1: 94.5312 (94.9109)  acc5: 100.0000 (99.8288)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [1] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5358 (0.5358)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1236  data: 0.0995  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 96.310 Acc@5 99.900
Epoch: [2]  [  0/468]  eta: 0:01:13  lr: 0.0019781476007338056  img/s: 2821.315311811909  loss: 0.5854 (0.5854)  acc1: 96.0938 (96.0938)  acc5: 100.0000 (100.0000)  time: 0.1570  data: 0.1116  max mem: 190
Epoch: [2]  [ 10/468]  eta: 0:00:22  lr: 0.0019781476007338056  img/s: 3476.2426314426316  loss: 0.6098 (0.6028)  acc1: 96.0938 (95.8097)  acc5: 100.0000 (99.9290)  time: 0.0488  data: 0.0105  max mem: 190
Epoch: [2]  [ 20/468]  eta: 0:00:20  lr: 0.0019781476007338056  img/s: 3316.966389877422  loss: 0.6098 (0.6142)  acc1: 95.3125 (95.2381)  acc5: 100.0000 (99.8140)  time: 0.0404  data: 0.0008  max mem: 190
Epoch: [2]  [ 30/468]  eta: 0:00:19  lr: 0.0019781476007338056  img/s: 3366.1095596671953  loss: 0.6261 (0.6167)  acc1: 94.5312 (95.1613)  acc5: 100.0000 (99.8236)  time: 0.0425  data: 0.0007  max mem: 190
Epoch: [2]  [ 40/468]  eta: 0:00:18  lr: 0.0019781476007338056  img/s: 3201.510569910491  loss: 0.6153 (0.6128)  acc1: 95.3125 (95.4649)  acc5: 100.0000 (99.8095)  time: 0.0400  data: 0.0003  max mem: 190
Epoch: [2]  [ 50/468]  eta: 0:00:17  lr: 0.0019781476007338056  img/s: 3468.8081875803605  loss: 0.5954 (0.6109)  acc1: 96.0938 (95.6342)  acc5: 100.0000 (99.8315)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [2]  [ 60/468]  eta: 0:00:16  lr: 0.0019781476007338056  img/s: 3455.4570860338936  loss: 0.5947 (0.6105)  acc1: 96.0938 (95.6327)  acc5: 100.0000 (99.8335)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [2]  [ 70/468]  eta: 0:00:17  lr: 0.0019781476007338056  img/s: 2209.2453098830915  loss: 0.6161 (0.6150)  acc1: 95.3125 (95.4445)  acc5: 100.0000 (99.8129)  time: 0.0469  data: 0.0014  max mem: 190
Epoch: [2]  [ 80/468]  eta: 0:00:17  lr: 0.0019781476007338056  img/s: 3254.7888546693503  loss: 0.6118 (0.6127)  acc1: 95.3125 (95.4572)  acc5: 100.0000 (99.8264)  time: 0.0529  data: 0.0034  max mem: 190
Epoch: [2]  [ 90/468]  eta: 0:00:16  lr: 0.0019781476007338056  img/s: 3281.928012519562  loss: 0.6002 (0.6121)  acc1: 96.0938 (95.5100)  acc5: 100.0000 (99.7940)  time: 0.0450  data: 0.0025  max mem: 190
Epoch: [2]  [100/468]  eta: 0:00:15  lr: 0.0019781476007338056  img/s: 3343.969205663069  loss: 0.6085 (0.6118)  acc1: 96.0938 (95.5523)  acc5: 100.0000 (99.7912)  time: 0.0400  data: 0.0005  max mem: 190
Epoch: [2]  [110/468]  eta: 0:00:15  lr: 0.0019781476007338056  img/s: 2319.247087284273  loss: 0.5869 (0.6091)  acc1: 96.0938 (95.6503)  acc5: 100.0000 (99.8029)  time: 0.0431  data: 0.0003  max mem: 190
Epoch: [2]  [120/468]  eta: 0:00:15  lr: 0.0019781476007338056  img/s: 2519.728123078656  loss: 0.5794 (0.6075)  acc1: 96.0938 (95.6999)  acc5: 100.0000 (99.8063)  time: 0.0471  data: 0.0007  max mem: 190
Epoch: [2]  [130/468]  eta: 0:00:14  lr: 0.0019781476007338056  img/s: 2833.570376000169  loss: 0.5886 (0.6080)  acc1: 96.0938 (95.6763)  acc5: 100.0000 (99.8151)  time: 0.0471  data: 0.0007  max mem: 190
Epoch: [2]  [140/468]  eta: 0:00:14  lr: 0.0019781476007338056  img/s: 3309.3399577140954  loss: 0.5939 (0.6070)  acc1: 96.0938 (95.7170)  acc5: 100.0000 (99.8227)  time: 0.0441  data: 0.0003  max mem: 190
Epoch: [2]  [150/468]  eta: 0:00:13  lr: 0.0019781476007338056  img/s: 3178.707086017431  loss: 0.6062 (0.6089)  acc1: 96.0938 (95.6178)  acc5: 100.0000 (99.8241)  time: 0.0411  data: 0.0003  max mem: 190
Epoch: [2]  [160/468]  eta: 0:00:13  lr: 0.0019781476007338056  img/s: 3495.025792591628  loss: 0.6187 (0.6087)  acc1: 95.3125 (95.6619)  acc5: 100.0000 (99.8205)  time: 0.0391  data: 0.0003  max mem: 190
Epoch: [2]  [170/468]  eta: 0:00:12  lr: 0.0019781476007338056  img/s: 3553.836100299203  loss: 0.6058 (0.6102)  acc1: 96.0938 (95.6232)  acc5: 100.0000 (99.8173)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [2]  [180/468]  eta: 0:00:12  lr: 0.0019781476007338056  img/s: 3498.419220518568  loss: 0.6069 (0.6115)  acc1: 95.3125 (95.5413)  acc5: 100.0000 (99.8187)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [2]  [190/468]  eta: 0:00:11  lr: 0.0019781476007338056  img/s: 3494.0022257655137  loss: 0.6069 (0.6111)  acc1: 95.3125 (95.5334)  acc5: 100.0000 (99.8282)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [2]  [200/468]  eta: 0:00:11  lr: 0.0019781476007338056  img/s: 3423.7469516861383  loss: 0.5874 (0.6098)  acc1: 96.0938 (95.5729)  acc5: 100.0000 (99.8368)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [2]  [210/468]  eta: 0:00:10  lr: 0.0019781476007338056  img/s: 3504.1963343950706  loss: 0.5874 (0.6095)  acc1: 96.0938 (95.5791)  acc5: 100.0000 (99.8408)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [2]  [220/468]  eta: 0:00:10  lr: 0.0019781476007338056  img/s: 3503.6932193434704  loss: 0.5977 (0.6110)  acc1: 95.3125 (95.5246)  acc5: 100.0000 (99.8409)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [2]  [230/468]  eta: 0:00:09  lr: 0.0019781476007338056  img/s: 3421.2596831546884  loss: 0.6120 (0.6113)  acc1: 94.5312 (95.4917)  acc5: 100.0000 (99.8410)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [2]  [240/468]  eta: 0:00:09  lr: 0.0019781476007338056  img/s: 3483.8672567520216  loss: 0.6175 (0.6122)  acc1: 94.5312 (95.4487)  acc5: 100.0000 (99.8476)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [2]  [250/468]  eta: 0:00:09  lr: 0.0019781476007338056  img/s: 3258.5818543785963  loss: 0.6107 (0.6115)  acc1: 95.3125 (95.4681)  acc5: 100.0000 (99.8444)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [2]  [260/468]  eta: 0:00:08  lr: 0.0019781476007338056  img/s: 1981.9730431155099  loss: 0.6107 (0.6116)  acc1: 95.3125 (95.4622)  acc5: 100.0000 (99.8443)  time: 0.0385  data: 0.0002  max mem: 190
Epoch: [2]  [270/468]  eta: 0:00:08  lr: 0.0019781476007338056  img/s: 3555.4600494043007  loss: 0.6134 (0.6117)  acc1: 95.3125 (95.4595)  acc5: 100.0000 (99.8443)  time: 0.0386  data: 0.0002  max mem: 190
Epoch: [2]  [280/468]  eta: 0:00:07  lr: 0.0019781476007338056  img/s: 3548.410181164449  loss: 0.5971 (0.6111)  acc1: 96.0938 (95.4932)  acc5: 100.0000 (99.8443)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [2]  [290/468]  eta: 0:00:07  lr: 0.0019781476007338056  img/s: 3422.0447458664253  loss: 0.5899 (0.6107)  acc1: 96.0938 (95.4924)  acc5: 100.0000 (99.8416)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [2]  [300/468]  eta: 0:00:06  lr: 0.0019781476007338056  img/s: 3448.1997739184053  loss: 0.6042 (0.6106)  acc1: 96.0938 (95.5124)  acc5: 100.0000 (99.8443)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [2]  [310/468]  eta: 0:00:06  lr: 0.0019781476007338056  img/s: 3380.6714607760414  loss: 0.6186 (0.6109)  acc1: 95.3125 (95.5110)  acc5: 100.0000 (99.8468)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [2]  [320/468]  eta: 0:00:06  lr: 0.0019781476007338056  img/s: 3389.9786070594178  loss: 0.6108 (0.6104)  acc1: 95.3125 (95.5242)  acc5: 100.0000 (99.8515)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [2]  [330/468]  eta: 0:00:05  lr: 0.0019781476007338056  img/s: 3506.7369837423335  loss: 0.5978 (0.6102)  acc1: 96.0938 (95.5249)  acc5: 100.0000 (99.8513)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [2]  [340/468]  eta: 0:00:05  lr: 0.0019781476007338056  img/s: 1529.856215929718  loss: 0.6067 (0.6103)  acc1: 95.3125 (95.5187)  acc5: 100.0000 (99.8534)  time: 0.0402  data: 0.0006  max mem: 190
Epoch: [2]  [350/468]  eta: 0:00:04  lr: 0.0019781476007338056  img/s: 2602.5688343796205  loss: 0.6111 (0.6111)  acc1: 94.5312 (95.4950)  acc5: 100.0000 (99.8397)  time: 0.0504  data: 0.0036  max mem: 190
Epoch: [2]  [360/468]  eta: 0:00:04  lr: 0.0019781476007338056  img/s: 3450.659845100749  loss: 0.6074 (0.6105)  acc1: 95.3125 (95.5181)  acc5: 100.0000 (99.8377)  time: 0.0477  data: 0.0034  max mem: 190
Epoch: [2]  [370/468]  eta: 0:00:04  lr: 0.0019781476007338056  img/s: 3490.935119318551  loss: 0.6074 (0.6108)  acc1: 95.3125 (95.5041)  acc5: 100.0000 (99.8357)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [2]  [380/468]  eta: 0:00:03  lr: 0.0019781476007338056  img/s: 3499.1032581420964  loss: 0.5872 (0.6100)  acc1: 96.0938 (95.5504)  acc5: 100.0000 (99.8380)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [2]  [390/468]  eta: 0:00:03  lr: 0.0019781476007338056  img/s: 3348.140069473461  loss: 0.5739 (0.6090)  acc1: 96.8750 (95.5942)  acc5: 100.0000 (99.8362)  time: 0.0384  data: 0.0003  max mem: 190
Epoch: [2]  [400/468]  eta: 0:00:02  lr: 0.0019781476007338056  img/s: 3369.2571543327645  loss: 0.5725 (0.6083)  acc1: 96.8750 (95.6242)  acc5: 100.0000 (99.8363)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [2]  [410/468]  eta: 0:00:02  lr: 0.0019781476007338056  img/s: 3479.306511820821  loss: 0.5725 (0.6077)  acc1: 96.8750 (95.6490)  acc5: 100.0000 (99.8384)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [2]  [420/468]  eta: 0:00:01  lr: 0.0019781476007338056  img/s: 3559.3966267105125  loss: 0.5968 (0.6074)  acc1: 96.0938 (95.6632)  acc5: 100.0000 (99.8404)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [2]  [430/468]  eta: 0:00:01  lr: 0.0019781476007338056  img/s: 2184.8355973726834  loss: 0.5968 (0.6067)  acc1: 96.0938 (95.6805)  acc5: 100.0000 (99.8423)  time: 0.0443  data: 0.0014  max mem: 190
Epoch: [2]  [440/468]  eta: 0:00:01  lr: 0.0019781476007338056  img/s: 2506.7746442045495  loss: 0.5753 (0.6061)  acc1: 96.0938 (95.7022)  acc5: 100.0000 (99.8441)  time: 0.0546  data: 0.0031  max mem: 190
Epoch: [2]  [450/468]  eta: 0:00:00  lr: 0.0019781476007338056  img/s: 3444.5714872321314  loss: 0.5954 (0.6062)  acc1: 96.0938 (95.7023)  acc5: 100.0000 (99.8424)  time: 0.0545  data: 0.0039  max mem: 190
Epoch: [2]  [460/468]  eta: 0:00:00  lr: 0.0019781476007338056  img/s: 3485.7446938364747  loss: 0.5979 (0.6060)  acc1: 95.3125 (95.7074)  acc5: 100.0000 (99.8424)  time: 0.0444  data: 0.0021  max mem: 190
Epoch: [2] Total time: 0:00:19
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5610 (0.5610)  acc1: 96.8750 (96.8750)  acc5: 100.0000 (100.0000)  time: 0.1196  data: 0.0968  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 96.960 Acc@5 99.950
Epoch: [3]  [  0/468]  eta: 0:01:21  lr: 0.001913545457642601  img/s: 2668.437331318684  loss: 0.6060 (0.6060)  acc1: 94.5312 (94.5312)  acc5: 100.0000 (100.0000)  time: 0.1742  data: 0.1262  max mem: 190
Epoch: [3]  [ 10/468]  eta: 0:00:23  lr: 0.001913545457642601  img/s: 3215.315721703509  loss: 0.6060 (0.5948)  acc1: 95.3125 (96.0227)  acc5: 100.0000 (99.9290)  time: 0.0504  data: 0.0117  max mem: 190
Epoch: [3]  [ 20/468]  eta: 0:00:19  lr: 0.001913545457642601  img/s: 3411.194916923468  loss: 0.5929 (0.5942)  acc1: 96.0938 (96.0193)  acc5: 100.0000 (99.9256)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [3]  [ 30/468]  eta: 0:00:18  lr: 0.001913545457642601  img/s: 3474.3529289948488  loss: 0.5953 (0.5982)  acc1: 96.0938 (96.1190)  acc5: 100.0000 (99.8236)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [3]  [ 40/468]  eta: 0:00:17  lr: 0.001913545457642601  img/s: 3487.4429143065013  loss: 0.6071 (0.5976)  acc1: 95.3125 (96.0938)  acc5: 100.0000 (99.8666)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [3]  [ 50/468]  eta: 0:00:16  lr: 0.001913545457642601  img/s: 3424.161848088833  loss: 0.5653 (0.5907)  acc1: 97.6562 (96.3848)  acc5: 100.0000 (99.8928)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [3]  [ 60/468]  eta: 0:00:16  lr: 0.001913545457642601  img/s: 3444.3946929453127  loss: 0.5667 (0.5880)  acc1: 96.8750 (96.4267)  acc5: 100.0000 (99.9103)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [3]  [ 70/468]  eta: 0:00:15  lr: 0.001913545457642601  img/s: 3553.5538257876624  loss: 0.5666 (0.5839)  acc1: 96.8750 (96.6329)  acc5: 100.0000 (99.9120)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [3]  [ 80/468]  eta: 0:00:15  lr: 0.001913545457642601  img/s: 3473.8358687001364  loss: 0.5538 (0.5815)  acc1: 97.6562 (96.7207)  acc5: 100.0000 (99.9228)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [3]  [ 90/468]  eta: 0:00:14  lr: 0.001913545457642601  img/s: 3484.319466258226  loss: 0.5678 (0.5810)  acc1: 96.8750 (96.7119)  acc5: 100.0000 (99.9227)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [3]  [100/468]  eta: 0:00:14  lr: 0.001913545457642601  img/s: 3542.579987858632  loss: 0.5647 (0.5788)  acc1: 96.8750 (96.7899)  acc5: 100.0000 (99.9149)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [3]  [110/468]  eta: 0:00:14  lr: 0.001913545457642601  img/s: 2912.3950960182274  loss: 0.5647 (0.5789)  acc1: 96.8750 (96.7483)  acc5: 100.0000 (99.9226)  time: 0.0464  data: 0.0015  max mem: 190
Epoch: [3]  [120/468]  eta: 0:00:14  lr: 0.001913545457642601  img/s: 3437.117709573746  loss: 0.5790 (0.5800)  acc1: 96.0938 (96.7265)  acc5: 100.0000 (99.9290)  time: 0.0512  data: 0.0029  max mem: 190
Epoch: [3]  [130/468]  eta: 0:00:13  lr: 0.001913545457642601  img/s: 3497.894972765891  loss: 0.6027 (0.5813)  acc1: 96.0938 (96.6961)  acc5: 100.0000 (99.9165)  time: 0.0423  data: 0.0016  max mem: 190
Epoch: [3]  [140/468]  eta: 0:00:13  lr: 0.001913545457642601  img/s: 3335.4720610345557  loss: 0.5973 (0.5822)  acc1: 96.0938 (96.6201)  acc5: 100.0000 (99.9113)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [3]  [150/468]  eta: 0:00:12  lr: 0.001913545457642601  img/s: 3523.792381002389  loss: 0.5937 (0.5826)  acc1: 96.0938 (96.6215)  acc5: 100.0000 (99.9120)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [3]  [160/468]  eta: 0:00:12  lr: 0.001913545457642601  img/s: 3514.9793240712856  loss: 0.6063 (0.5881)  acc1: 96.0938 (96.4431)  acc5: 100.0000 (99.8932)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [3]  [170/468]  eta: 0:00:11  lr: 0.001913545457642601  img/s: 3534.068262755656  loss: 0.6085 (0.5878)  acc1: 96.0938 (96.4638)  acc5: 100.0000 (99.8995)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [3]  [180/468]  eta: 0:00:11  lr: 0.001913545457642601  img/s: 3485.1111154387945  loss: 0.5896 (0.5887)  acc1: 96.8750 (96.4132)  acc5: 100.0000 (99.9050)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [3]  [190/468]  eta: 0:00:10  lr: 0.001913545457642601  img/s: 3553.4832641660546  loss: 0.5913 (0.5890)  acc1: 95.3125 (96.3719)  acc5: 100.0000 (99.9059)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [3]  [200/468]  eta: 0:00:10  lr: 0.001913545457642601  img/s: 3421.3905019245967  loss: 0.5847 (0.5893)  acc1: 95.3125 (96.3386)  acc5: 100.0000 (99.9067)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [3]  [210/468]  eta: 0:00:10  lr: 0.001913545457642601  img/s: 3595.2701922626184  loss: 0.5847 (0.5894)  acc1: 96.0938 (96.3344)  acc5: 100.0000 (99.9111)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [3]  [220/468]  eta: 0:00:09  lr: 0.001913545457642601  img/s: 3537.258275353152  loss: 0.5805 (0.5894)  acc1: 96.8750 (96.3518)  acc5: 100.0000 (99.9152)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [3]  [230/468]  eta: 0:00:09  lr: 0.001913545457642601  img/s: 3386.7281008314303  loss: 0.5732 (0.5886)  acc1: 96.8750 (96.3846)  acc5: 100.0000 (99.9154)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [3]  [240/468]  eta: 0:00:08  lr: 0.001913545457642601  img/s: 3616.8023821393444  loss: 0.5657 (0.5874)  acc1: 96.8750 (96.4341)  acc5: 100.0000 (99.9157)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [3]  [250/468]  eta: 0:00:08  lr: 0.001913545457642601  img/s: 3618.533177863002  loss: 0.5598 (0.5867)  acc1: 96.8750 (96.4704)  acc5: 100.0000 (99.9128)  time: 0.0363  data: 0.0003  max mem: 190
Epoch: [3]  [260/468]  eta: 0:00:08  lr: 0.001913545457642601  img/s: 3545.551224731048  loss: 0.5767 (0.5865)  acc1: 96.8750 (96.4559)  acc5: 100.0000 (99.9162)  time: 0.0392  data: 0.0003  max mem: 190
Epoch: [3]  [270/468]  eta: 0:00:07  lr: 0.001913545457642601  img/s: 3522.613213303851  loss: 0.5849 (0.5870)  acc1: 96.0938 (96.4310)  acc5: 100.0000 (99.9193)  time: 0.0415  data: 0.0003  max mem: 190
Epoch: [3]  [280/468]  eta: 0:00:07  lr: 0.001913545457642601  img/s: 3410.609813737199  loss: 0.5820 (0.5864)  acc1: 96.8750 (96.4552)  acc5: 100.0000 (99.9166)  time: 0.0389  data: 0.0003  max mem: 190
Epoch: [3]  [290/468]  eta: 0:00:06  lr: 0.001913545457642601  img/s: 3503.327408219464  loss: 0.5733 (0.5860)  acc1: 96.8750 (96.4696)  acc5: 100.0000 (99.9168)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [3]  [300/468]  eta: 0:00:06  lr: 0.001913545457642601  img/s: 3625.130232212671  loss: 0.5620 (0.5857)  acc1: 96.8750 (96.4805)  acc5: 100.0000 (99.9195)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [3]  [310/468]  eta: 0:00:06  lr: 0.001913545457642601  img/s: 3454.3232016471497  loss: 0.5660 (0.5855)  acc1: 96.8750 (96.4957)  acc5: 100.0000 (99.9196)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [3]  [320/468]  eta: 0:00:05  lr: 0.001913545457642601  img/s: 3379.0330746524173  loss: 0.6028 (0.5871)  acc1: 96.0938 (96.4393)  acc5: 100.0000 (99.9124)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [3]  [330/468]  eta: 0:00:05  lr: 0.001913545457642601  img/s: 3490.503884688152  loss: 0.6150 (0.5877)  acc1: 95.3125 (96.4171)  acc5: 100.0000 (99.9103)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [3]  [340/468]  eta: 0:00:04  lr: 0.001913545457642601  img/s: 3462.0081379977432  loss: 0.5826 (0.5875)  acc1: 96.0938 (96.4214)  acc5: 100.0000 (99.9084)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [3]  [350/468]  eta: 0:00:04  lr: 0.001913545457642601  img/s: 3005.1716606306222  loss: 0.5865 (0.5878)  acc1: 96.0938 (96.4009)  acc5: 100.0000 (99.9110)  time: 0.0377  data: 0.0002  max mem: 190
Epoch: [3]  [360/468]  eta: 0:00:04  lr: 0.001913545457642601  img/s: 3514.818239549576  loss: 0.5899 (0.5879)  acc1: 96.0938 (96.4140)  acc5: 100.0000 (99.9134)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [3]  [370/468]  eta: 0:00:03  lr: 0.001913545457642601  img/s: 3496.049959300622  loss: 0.5930 (0.5880)  acc1: 96.0938 (96.3907)  acc5: 100.0000 (99.9137)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [3]  [380/468]  eta: 0:00:03  lr: 0.001913545457642601  img/s: 3580.0702315935478  loss: 0.6009 (0.5906)  acc1: 94.5312 (96.2803)  acc5: 100.0000 (99.9118)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [3]  [390/468]  eta: 0:00:03  lr: 0.001913545457642601  img/s: 2281.0239118981663  loss: 0.6759 (0.5998)  acc1: 92.1875 (95.9359)  acc5: 100.0000 (99.8441)  time: 0.0419  data: 0.0004  max mem: 190
Epoch: [3]  [400/468]  eta: 0:00:02  lr: 0.001913545457642601  img/s: 3228.949118290952  loss: 0.8094 (0.6080)  acc1: 86.7188 (95.6223)  acc5: 99.2188 (99.8208)  time: 0.0525  data: 0.0031  max mem: 190
Epoch: [3]  [410/468]  eta: 0:00:02  lr: 0.001913545457642601  img/s: 3199.450015196572  loss: 0.7277 (0.6102)  acc1: 89.0625 (95.5254)  acc5: 100.0000 (99.8213)  time: 0.0491  data: 0.0029  max mem: 190
Epoch: [3]  [420/468]  eta: 0:00:01  lr: 0.001913545457642601  img/s: 3448.9086237023334  loss: 0.6970 (0.6125)  acc1: 91.4062 (95.4313)  acc5: 100.0000 (99.8126)  time: 0.0384  data: 0.0003  max mem: 190
Epoch: [3]  [430/468]  eta: 0:00:01  lr: 0.001913545457642601  img/s: 3289.6703533722634  loss: 0.6822 (0.6141)  acc1: 92.1875 (95.3705)  acc5: 100.0000 (99.8133)  time: 0.0375  data: 0.0002  max mem: 190
Epoch: [3]  [440/468]  eta: 0:00:01  lr: 0.001913545457642601  img/s: 3540.1505552185267  loss: 0.6592 (0.6149)  acc1: 93.7500 (95.3426)  acc5: 100.0000 (99.8104)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [3]  [450/468]  eta: 0:00:00  lr: 0.001913545457642601  img/s: 3581.431529512221  loss: 0.6415 (0.6156)  acc1: 93.7500 (95.3090)  acc5: 100.0000 (99.8095)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [3]  [460/468]  eta: 0:00:00  lr: 0.001913545457642601  img/s: 3241.6413290906125  loss: 0.6232 (0.6155)  acc1: 95.3125 (95.3125)  acc5: 100.0000 (99.8119)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [3] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5786 (0.5786)  acc1: 97.6562 (97.6562)  acc5: 100.0000 (100.0000)  time: 0.1213  data: 0.0950  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 95.390 Acc@5 99.820
Epoch: [4]  [  0/468]  eta: 0:01:13  lr: 0.0018090169943749475  img/s: 2566.57445811701  loss: 0.6653 (0.6653)  acc1: 92.1875 (92.1875)  acc5: 100.0000 (100.0000)  time: 0.1563  data: 0.1064  max mem: 190
Epoch: [4]  [ 10/468]  eta: 0:00:24  lr: 0.0018090169943749475  img/s: 3494.7982814737666  loss: 0.6109 (0.6174)  acc1: 96.0938 (95.3125)  acc5: 100.0000 (99.8580)  time: 0.0527  data: 0.0101  max mem: 190
Epoch: [4]  [ 20/468]  eta: 0:00:21  lr: 0.0018090169943749475  img/s: 3517.005646904684  loss: 0.6020 (0.6088)  acc1: 96.0938 (95.7217)  acc5: 100.0000 (99.9256)  time: 0.0418  data: 0.0006  max mem: 190
Epoch: [4]  [ 30/468]  eta: 0:00:19  lr: 0.0018090169943749475  img/s: 3463.392825117893  loss: 0.5943 (0.6025)  acc1: 96.0938 (95.9173)  acc5: 100.0000 (99.9244)  time: 0.0390  data: 0.0005  max mem: 190
Epoch: [4]  [ 40/468]  eta: 0:00:18  lr: 0.0018090169943749475  img/s: 3438.5706453513694  loss: 0.5790 (0.5999)  acc1: 96.0938 (95.9794)  acc5: 100.0000 (99.9428)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [4]  [ 50/468]  eta: 0:00:17  lr: 0.0018090169943749475  img/s: 3453.900964365442  loss: 0.5844 (0.5983)  acc1: 96.0938 (96.0478)  acc5: 100.0000 (99.9081)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [4]  [ 60/468]  eta: 0:00:16  lr: 0.0018090169943749475  img/s: 3443.908602219514  loss: 0.6074 (0.6035)  acc1: 95.3125 (95.7992)  acc5: 100.0000 (99.8975)  time: 0.0384  data: 0.0004  max mem: 190
Epoch: [4]  [ 70/468]  eta: 0:00:16  lr: 0.0018090169943749475  img/s: 3476.0850777289297  loss: 0.6247 (0.6063)  acc1: 94.5312 (95.7416)  acc5: 100.0000 (99.9010)  time: 0.0386  data: 0.0004  max mem: 190
Epoch: [4]  [ 80/468]  eta: 0:00:15  lr: 0.0018090169943749475  img/s: 3564.5721949632502  loss: 0.6247 (0.6091)  acc1: 94.5312 (95.6019)  acc5: 100.0000 (99.8939)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [4]  [ 90/468]  eta: 0:00:14  lr: 0.0018090169943749475  img/s: 3499.81037809648  loss: 0.6081 (0.6080)  acc1: 95.3125 (95.6731)  acc5: 100.0000 (99.8884)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [4]  [100/468]  eta: 0:00:14  lr: 0.0018090169943749475  img/s: 3433.578572387902  loss: 0.6030 (0.6081)  acc1: 95.3125 (95.6142)  acc5: 100.0000 (99.8917)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [4]  [110/468]  eta: 0:00:14  lr: 0.0018090169943749475  img/s: 3602.9912151778103  loss: 0.6003 (0.6076)  acc1: 94.5312 (95.6011)  acc5: 100.0000 (99.8944)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [4]  [120/468]  eta: 0:00:13  lr: 0.0018090169943749475  img/s: 3575.6114766763462  loss: 0.5878 (0.6064)  acc1: 96.0938 (95.6353)  acc5: 100.0000 (99.9032)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [4]  [130/468]  eta: 0:00:13  lr: 0.0018090169943749475  img/s: 3571.3301048374224  loss: 0.5812 (0.6045)  acc1: 96.8750 (95.7180)  acc5: 100.0000 (99.9046)  time: 0.0364  data: 0.0002  max mem: 190
Epoch: [4]  [140/468]  eta: 0:00:12  lr: 0.0018090169943749475  img/s: 3436.017817828068  loss: 0.5775 (0.6025)  acc1: 96.8750 (95.7890)  acc5: 100.0000 (99.9058)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [4]  [150/468]  eta: 0:00:12  lr: 0.0018090169943749475  img/s: 3567.675283422602  loss: 0.5696 (0.6018)  acc1: 96.8750 (95.8506)  acc5: 100.0000 (99.9069)  time: 0.0385  data: 0.0003  max mem: 190
Epoch: [4]  [160/468]  eta: 0:00:11  lr: 0.0018090169943749475  img/s: 3576.2307456601966  loss: 0.5856 (0.6003)  acc1: 96.8750 (95.9045)  acc5: 100.0000 (99.9078)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [4]  [170/468]  eta: 0:00:11  lr: 0.0018090169943749475  img/s: 2341.4799443492216  loss: 0.5834 (0.5991)  acc1: 96.8750 (95.9430)  acc5: 100.0000 (99.9086)  time: 0.0456  data: 0.0003  max mem: 190
Epoch: [4]  [180/468]  eta: 0:00:11  lr: 0.0018090169943749475  img/s: 3425.844300372658  loss: 0.5871 (0.5986)  acc1: 96.0938 (95.9384)  acc5: 100.0000 (99.9007)  time: 0.0495  data: 0.0010  max mem: 190
Epoch: [4]  [190/468]  eta: 0:00:11  lr: 0.0018090169943749475  img/s: 3642.025045790652  loss: 0.5852 (0.5970)  acc1: 96.0938 (96.0079)  acc5: 100.0000 (99.9059)  time: 0.0403  data: 0.0009  max mem: 190
Epoch: [4]  [200/468]  eta: 0:00:10  lr: 0.0018090169943749475  img/s: 3513.277177185038  loss: 0.5691 (0.5967)  acc1: 96.8750 (96.0199)  acc5: 100.0000 (99.9028)  time: 0.0365  data: 0.0003  max mem: 190
Epoch: [4]  [210/468]  eta: 0:00:10  lr: 0.0018090169943749475  img/s: 3409.310302783987  loss: 0.5779 (0.5966)  acc1: 96.0938 (96.0086)  acc5: 100.0000 (99.9000)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [4]  [220/468]  eta: 0:00:09  lr: 0.0018090169943749475  img/s: 2335.2163617541387  loss: 0.5842 (0.5966)  acc1: 96.0938 (96.0230)  acc5: 100.0000 (99.9010)  time: 0.0430  data: 0.0008  max mem: 190
Epoch: [4]  [230/468]  eta: 0:00:09  lr: 0.0018090169943749475  img/s: 2321.1723311989554  loss: 0.5957 (0.5973)  acc1: 96.0938 (96.0058)  acc5: 100.0000 (99.8985)  time: 0.0547  data: 0.0040  max mem: 190
Epoch: [4]  [240/468]  eta: 0:00:09  lr: 0.0018090169943749475  img/s: 3229.8426923030647  loss: 0.5963 (0.5970)  acc1: 96.0938 (96.0095)  acc5: 100.0000 (99.8930)  time: 0.0570  data: 0.0054  max mem: 190
Epoch: [4]  [250/468]  eta: 0:00:08  lr: 0.0018090169943749475  img/s: 3209.798589023078  loss: 0.5701 (0.5961)  acc1: 96.8750 (96.0471)  acc5: 100.0000 (99.8911)  time: 0.0471  data: 0.0021  max mem: 190
Epoch: [4]  [260/468]  eta: 0:00:08  lr: 0.0018090169943749475  img/s: 3546.6755101636354  loss: 0.5701 (0.6003)  acc1: 96.8750 (95.8752)  acc5: 100.0000 (99.8743)  time: 0.0393  data: 0.0003  max mem: 190
Epoch: [4]  [270/468]  eta: 0:00:08  lr: 0.0018090169943749475  img/s: 3492.1840309623703  loss: 0.6600 (0.6041)  acc1: 92.9688 (95.7074)  acc5: 100.0000 (99.8645)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [4]  [280/468]  eta: 0:00:07  lr: 0.0018090169943749475  img/s: 3504.7453519949863  loss: 0.6688 (0.6056)  acc1: 92.1875 (95.6350)  acc5: 100.0000 (99.8638)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [4]  [290/468]  eta: 0:00:07  lr: 0.0018090169943749475  img/s: 3551.9786168432056  loss: 0.6537 (0.6070)  acc1: 93.7500 (95.5944)  acc5: 100.0000 (99.8523)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [4]  [300/468]  eta: 0:00:06  lr: 0.0018090169943749475  img/s: 3557.674775521023  loss: 0.6653 (0.6094)  acc1: 93.7500 (95.5046)  acc5: 99.2188 (99.8391)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [4]  [310/468]  eta: 0:00:06  lr: 0.0018090169943749475  img/s: 3551.7906255168537  loss: 0.6758 (0.6117)  acc1: 92.9688 (95.4130)  acc5: 100.0000 (99.8342)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [4]  [320/468]  eta: 0:00:05  lr: 0.0018090169943749475  img/s: 3595.39057874928  loss: 0.6588 (0.6142)  acc1: 92.9688 (95.3149)  acc5: 100.0000 (99.8248)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [4]  [330/468]  eta: 0:00:05  lr: 0.0018090169943749475  img/s: 3503.7846840614516  loss: 0.6509 (0.6150)  acc1: 93.7500 (95.2960)  acc5: 100.0000 (99.8230)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [4]  [340/468]  eta: 0:00:05  lr: 0.0018090169943749475  img/s: 3458.952349045177  loss: 0.6186 (0.6149)  acc1: 95.3125 (95.3033)  acc5: 100.0000 (99.8236)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [4]  [350/468]  eta: 0:00:04  lr: 0.0018090169943749475  img/s: 3440.5318533993836  loss: 0.5989 (0.6143)  acc1: 96.0938 (95.3281)  acc5: 100.0000 (99.8264)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [4]  [360/468]  eta: 0:00:04  lr: 0.0018090169943749475  img/s: 3385.6815685087436  loss: 0.5974 (0.6144)  acc1: 96.0938 (95.3276)  acc5: 100.0000 (99.8247)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [4]  [370/468]  eta: 0:00:03  lr: 0.0018090169943749475  img/s: 3553.3421493292035  loss: 0.5979 (0.6139)  acc1: 96.0938 (95.3651)  acc5: 100.0000 (99.8273)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [4]  [380/468]  eta: 0:00:03  lr: 0.0018090169943749475  img/s: 3550.9214244139903  loss: 0.5998 (0.6138)  acc1: 96.0938 (95.3494)  acc5: 100.0000 (99.8319)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [4]  [390/468]  eta: 0:00:03  lr: 0.0018090169943749475  img/s: 3523.838638959266  loss: 0.6247 (0.6144)  acc1: 94.5312 (95.3165)  acc5: 100.0000 (99.8342)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [4]  [400/468]  eta: 0:00:02  lr: 0.0018090169943749475  img/s: 3421.870256351422  loss: 0.6127 (0.6145)  acc1: 94.5312 (95.3281)  acc5: 100.0000 (99.8266)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [4]  [410/468]  eta: 0:00:02  lr: 0.0018090169943749475  img/s: 3300.7132484491526  loss: 0.6100 (0.6147)  acc1: 94.5312 (95.3125)  acc5: 100.0000 (99.8251)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [4]  [420/468]  eta: 0:00:01  lr: 0.0018090169943749475  img/s: 3615.414067813731  loss: 0.6168 (0.6146)  acc1: 95.3125 (95.3106)  acc5: 100.0000 (99.8200)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [4]  [430/468]  eta: 0:00:01  lr: 0.0018090169943749475  img/s: 3585.21036955912  loss: 0.5991 (0.6142)  acc1: 95.3125 (95.3252)  acc5: 100.0000 (99.8187)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [4]  [440/468]  eta: 0:00:01  lr: 0.0018090169943749475  img/s: 3062.4962893243205  loss: 0.5911 (0.6138)  acc1: 95.3125 (95.3444)  acc5: 100.0000 (99.8211)  time: 0.0415  data: 0.0007  max mem: 190
Epoch: [4]  [450/468]  eta: 0:00:00  lr: 0.0018090169943749475  img/s: 2291.2944530513128  loss: 0.5909 (0.6136)  acc1: 95.3125 (95.3419)  acc5: 100.0000 (99.8233)  time: 0.0487  data: 0.0029  max mem: 190
Epoch: [4]  [460/468]  eta: 0:00:00  lr: 0.0018090169943749475  img/s: 3341.2845069020027  loss: 0.5903 (0.6131)  acc1: 95.3125 (95.3498)  acc5: 100.0000 (99.8271)  time: 0.0447  data: 0.0024  max mem: 190
Epoch: [4] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.6161 (0.6161)  acc1: 95.3125 (95.3125)  acc5: 100.0000 (100.0000)  time: 0.1243  data: 0.0979  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 95.730 Acc@5 99.930
Epoch: [5]  [  0/468]  eta: 0:01:14  lr: 0.0016691306063588583  img/s: 2564.9313560617643  loss: 0.6228 (0.6228)  acc1: 96.0938 (96.0938)  acc5: 99.2188 (99.2188)  time: 0.1600  data: 0.1101  max mem: 190
Epoch: [5]  [ 10/468]  eta: 0:00:24  lr: 0.0016691306063588583  img/s: 3533.300725257657  loss: 0.6159 (0.6111)  acc1: 96.0938 (95.5966)  acc5: 100.0000 (99.7869)  time: 0.0539  data: 0.0111  max mem: 190
Epoch: [5]  [ 20/468]  eta: 0:00:21  lr: 0.0016691306063588583  img/s: 3536.978628086542  loss: 0.5970 (0.5997)  acc1: 96.0938 (95.9077)  acc5: 100.0000 (99.8140)  time: 0.0420  data: 0.0008  max mem: 190
Epoch: [5]  [ 30/468]  eta: 0:00:19  lr: 0.0016691306063588583  img/s: 3382.5245370749562  loss: 0.5810 (0.6025)  acc1: 96.0938 (95.8417)  acc5: 100.0000 (99.7984)  time: 0.0390  data: 0.0004  max mem: 190
Epoch: [5]  [ 40/468]  eta: 0:00:18  lr: 0.0016691306063588583  img/s: 3408.271406805485  loss: 0.6031 (0.6021)  acc1: 95.3125 (95.8460)  acc5: 100.0000 (99.8285)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [5]  [ 50/468]  eta: 0:00:17  lr: 0.0016691306063588583  img/s: 3385.361330760597  loss: 0.5901 (0.6014)  acc1: 96.0938 (95.8946)  acc5: 100.0000 (99.8468)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [5]  [ 60/468]  eta: 0:00:16  lr: 0.0016691306063588583  img/s: 3415.5135444632474  loss: 0.5825 (0.5997)  acc1: 96.8750 (96.0041)  acc5: 100.0000 (99.8719)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [5]  [ 70/468]  eta: 0:00:16  lr: 0.0016691306063588583  img/s: 3336.840315242523  loss: 0.5662 (0.5967)  acc1: 97.6562 (96.0827)  acc5: 100.0000 (99.8790)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [5]  [ 80/468]  eta: 0:00:15  lr: 0.0016691306063588583  img/s: 3558.1935141798613  loss: 0.5666 (0.5931)  acc1: 97.6562 (96.2288)  acc5: 100.0000 (99.8939)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [5]  [ 90/468]  eta: 0:00:15  lr: 0.0016691306063588583  img/s: 3518.388570679599  loss: 0.5748 (0.5911)  acc1: 96.8750 (96.2826)  acc5: 100.0000 (99.9056)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [5]  [100/468]  eta: 0:00:14  lr: 0.0016691306063588583  img/s: 3507.9972295186944  loss: 0.5749 (0.5911)  acc1: 96.0938 (96.2871)  acc5: 100.0000 (99.9149)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [5]  [110/468]  eta: 0:00:14  lr: 0.0016691306063588583  img/s: 3445.898023106547  loss: 0.5652 (0.5887)  acc1: 96.8750 (96.3612)  acc5: 100.0000 (99.9155)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [5]  [120/468]  eta: 0:00:13  lr: 0.0016691306063588583  img/s: 3503.6703539101095  loss: 0.5570 (0.5868)  acc1: 98.4375 (96.4941)  acc5: 100.0000 (99.9161)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [5]  [130/468]  eta: 0:00:13  lr: 0.0016691306063588583  img/s: 3515.8770653376905  loss: 0.5673 (0.5860)  acc1: 97.6562 (96.5649)  acc5: 100.0000 (99.9105)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [5]  [140/468]  eta: 0:00:12  lr: 0.0016691306063588583  img/s: 3546.5115074646583  loss: 0.5682 (0.5845)  acc1: 97.6562 (96.6312)  acc5: 100.0000 (99.9169)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [5]  [150/468]  eta: 0:00:12  lr: 0.0016691306063588583  img/s: 3463.392825117893  loss: 0.5682 (0.5837)  acc1: 96.8750 (96.6474)  acc5: 100.0000 (99.9224)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [5]  [160/468]  eta: 0:00:11  lr: 0.0016691306063588583  img/s: 3087.843005532997  loss: 0.5680 (0.5827)  acc1: 96.8750 (96.7052)  acc5: 100.0000 (99.9175)  time: 0.0391  data: 0.0003  max mem: 190
Epoch: [5]  [170/468]  eta: 0:00:11  lr: 0.0016691306063588583  img/s: 3348.808373410181  loss: 0.5629 (0.5817)  acc1: 97.6562 (96.7608)  acc5: 100.0000 (99.9132)  time: 0.0408  data: 0.0003  max mem: 190
Epoch: [5]  [180/468]  eta: 0:00:11  lr: 0.0016691306063588583  img/s: 3404.726617792547  loss: 0.5687 (0.5814)  acc1: 97.6562 (96.7800)  acc5: 100.0000 (99.9050)  time: 0.0401  data: 0.0003  max mem: 190
Epoch: [5]  [190/468]  eta: 0:00:10  lr: 0.0016691306063588583  img/s: 3387.732525634958  loss: 0.5621 (0.5802)  acc1: 97.6562 (96.8136)  acc5: 100.0000 (99.9018)  time: 0.0392  data: 0.0003  max mem: 190
Epoch: [5]  [200/468]  eta: 0:00:10  lr: 0.0016691306063588583  img/s: 3349.100840283713  loss: 0.5525 (0.5793)  acc1: 96.8750 (96.8595)  acc5: 100.0000 (99.8989)  time: 0.0384  data: 0.0003  max mem: 190
Epoch: [5]  [210/468]  eta: 0:00:10  lr: 0.0016691306063588583  img/s: 3505.157228105454  loss: 0.5715 (0.5793)  acc1: 96.8750 (96.8269)  acc5: 100.0000 (99.9037)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [5]  [220/468]  eta: 0:00:09  lr: 0.0016691306063588583  img/s: 2838.7243924620884  loss: 0.5872 (0.5800)  acc1: 96.8750 (96.8149)  acc5: 100.0000 (99.9081)  time: 0.0469  data: 0.0017  max mem: 190
Epoch: [5]  [230/468]  eta: 0:00:09  lr: 0.0016691306063588583  img/s: 3548.808926375907  loss: 0.5927 (0.5808)  acc1: 96.0938 (96.7803)  acc5: 100.0000 (99.9087)  time: 0.0489  data: 0.0030  max mem: 190
Epoch: [5]  [240/468]  eta: 0:00:09  lr: 0.0016691306063588583  img/s: 3558.052024998509  loss: 0.5956 (0.5810)  acc1: 96.0938 (96.7680)  acc5: 100.0000 (99.9060)  time: 0.0394  data: 0.0015  max mem: 190
Epoch: [5]  [250/468]  eta: 0:00:08  lr: 0.0016691306063588583  img/s: 3512.0361100571745  loss: 0.5814 (0.5812)  acc1: 96.8750 (96.7661)  acc5: 100.0000 (99.9035)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [5]  [260/468]  eta: 0:00:08  lr: 0.0016691306063588583  img/s: 3292.7977233139522  loss: 0.5690 (0.5806)  acc1: 96.8750 (96.7762)  acc5: 100.0000 (99.9072)  time: 0.0388  data: 0.0003  max mem: 190
Epoch: [5]  [270/468]  eta: 0:00:07  lr: 0.0016691306063588583  img/s: 3578.6384039567793  loss: 0.5666 (0.5808)  acc1: 96.8750 (96.7626)  acc5: 100.0000 (99.9106)  time: 0.0388  data: 0.0003  max mem: 190
Epoch: [5]  [280/468]  eta: 0:00:07  lr: 0.0016691306063588583  img/s: 3536.978628086542  loss: 0.5828 (0.5810)  acc1: 96.0938 (96.7471)  acc5: 100.0000 (99.9083)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [5]  [290/468]  eta: 0:00:06  lr: 0.0016691306063588583  img/s: 3547.8473992717563  loss: 0.5722 (0.5804)  acc1: 96.8750 (96.7730)  acc5: 100.0000 (99.9114)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [5]  [300/468]  eta: 0:00:06  lr: 0.0016691306063588583  img/s: 3528.3778177946606  loss: 0.5726 (0.5804)  acc1: 97.6562 (96.7712)  acc5: 100.0000 (99.9143)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [5]  [310/468]  eta: 0:00:06  lr: 0.0016691306063588583  img/s: 3517.6047803753013  loss: 0.5726 (0.5806)  acc1: 96.8750 (96.7620)  acc5: 100.0000 (99.9146)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [5]  [320/468]  eta: 0:00:05  lr: 0.0016691306063588583  img/s: 3548.433634284657  loss: 0.5642 (0.5801)  acc1: 96.8750 (96.7825)  acc5: 100.0000 (99.9173)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [5]  [330/468]  eta: 0:00:05  lr: 0.0016691306063588583  img/s: 3526.8480134538577  loss: 0.5545 (0.5798)  acc1: 96.8750 (96.7782)  acc5: 100.0000 (99.9198)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [5]  [340/468]  eta: 0:00:04  lr: 0.0016691306063588583  img/s: 3395.2740082087994  loss: 0.5802 (0.5803)  acc1: 96.0938 (96.7536)  acc5: 100.0000 (99.9198)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [5]  [350/468]  eta: 0:00:04  lr: 0.0016691306063588583  img/s: 3578.161382555435  loss: 0.5988 (0.5803)  acc1: 96.0938 (96.7548)  acc5: 100.0000 (99.9176)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [5]  [360/468]  eta: 0:00:04  lr: 0.0016691306063588583  img/s: 3519.4264774328885  loss: 0.5805 (0.5803)  acc1: 96.8750 (96.7430)  acc5: 100.0000 (99.9178)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [5]  [370/468]  eta: 0:00:03  lr: 0.0016691306063588583  img/s: 3484.5908483157004  loss: 0.5814 (0.5806)  acc1: 96.8750 (96.7297)  acc5: 100.0000 (99.9179)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [5]  [380/468]  eta: 0:00:03  lr: 0.0016691306063588583  img/s: 3446.738691079981  loss: 0.5775 (0.5806)  acc1: 96.8750 (96.7315)  acc5: 100.0000 (99.9180)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [5]  [390/468]  eta: 0:00:03  lr: 0.0016691306063588583  img/s: 3408.812419441887  loss: 0.5775 (0.5808)  acc1: 96.8750 (96.7291)  acc5: 100.0000 (99.9161)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [5]  [400/468]  eta: 0:00:02  lr: 0.0016691306063588583  img/s: 3489.619052571369  loss: 0.5905 (0.5811)  acc1: 96.8750 (96.7191)  acc5: 100.0000 (99.9104)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [5]  [410/468]  eta: 0:00:02  lr: 0.0016691306063588583  img/s: 3493.979486645494  loss: 0.5783 (0.5811)  acc1: 96.8750 (96.7115)  acc5: 100.0000 (99.9088)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [5]  [420/468]  eta: 0:00:01  lr: 0.0016691306063588583  img/s: 3499.240097767639  loss: 0.5662 (0.5808)  acc1: 96.8750 (96.7191)  acc5: 100.0000 (99.9109)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [5]  [430/468]  eta: 0:00:01  lr: 0.0016691306063588583  img/s: 3484.4777673211097  loss: 0.5600 (0.5802)  acc1: 96.8750 (96.7391)  acc5: 100.0000 (99.9112)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [5]  [440/468]  eta: 0:00:01  lr: 0.0016691306063588583  img/s: 3520.742038717801  loss: 0.5597 (0.5799)  acc1: 97.6562 (96.7528)  acc5: 100.0000 (99.9114)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [5]  [450/468]  eta: 0:00:00  lr: 0.0016691306063588583  img/s: 3472.4877398824115  loss: 0.5598 (0.5798)  acc1: 96.8750 (96.7503)  acc5: 100.0000 (99.9134)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [5]  [460/468]  eta: 0:00:00  lr: 0.0016691306063588583  img/s: 3504.310698876654  loss: 0.5630 (0.5794)  acc1: 97.6562 (96.7615)  acc5: 100.0000 (99.9153)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [5] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5471 (0.5471)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1204  data: 0.1010  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 97.440 Acc@5 99.980
Epoch: [6]  [  0/468]  eta: 0:01:42  lr: 0.0015  img/s: 1579.9847319707471  loss: 0.6669 (0.6669)  acc1: 93.7500 (93.7500)  acc5: 100.0000 (100.0000)  time: 0.2198  data: 0.1388  max mem: 190
Epoch: [6]  [ 10/468]  eta: 0:00:37  lr: 0.0015  img/s: 3150.1517482558515  loss: 0.5509 (0.5569)  acc1: 98.4375 (97.9403)  acc5: 100.0000 (100.0000)  time: 0.0820  data: 0.0176  max mem: 190
Epoch: [6]  [ 20/468]  eta: 0:00:32  lr: 0.0015  img/s: 2385.509817600142  loss: 0.5436 (0.5496)  acc1: 98.4375 (98.1027)  acc5: 100.0000 (99.9256)  time: 0.0648  data: 0.0048  max mem: 190
Epoch: [6]  [ 30/468]  eta: 0:00:28  lr: 0.0015  img/s: 3463.6386111146953  loss: 0.5402 (0.5503)  acc1: 97.6562 (98.0091)  acc5: 100.0000 (99.9496)  time: 0.0559  data: 0.0043  max mem: 190
Epoch: [6]  [ 40/468]  eta: 0:00:24  lr: 0.0015  img/s: 3532.8589609449546  loss: 0.5531 (0.5517)  acc1: 97.6562 (97.8849)  acc5: 100.0000 (99.9428)  time: 0.0438  data: 0.0024  max mem: 190
Epoch: [6]  [ 50/468]  eta: 0:00:22  lr: 0.0015  img/s: 3459.219793814433  loss: 0.5620 (0.5530)  acc1: 97.6562 (97.8248)  acc5: 100.0000 (99.9540)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [6]  [ 60/468]  eta: 0:00:20  lr: 0.0015  img/s: 3523.537983946655  loss: 0.5487 (0.5516)  acc1: 97.6562 (97.8996)  acc5: 100.0000 (99.9616)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [6]  [ 70/468]  eta: 0:00:19  lr: 0.0015  img/s: 3478.6978118459674  loss: 0.5407 (0.5519)  acc1: 98.4375 (97.9093)  acc5: 100.0000 (99.9670)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [6]  [ 80/468]  eta: 0:00:18  lr: 0.0015  img/s: 3461.6509791025915  loss: 0.5368 (0.5515)  acc1: 97.6562 (97.9649)  acc5: 100.0000 (99.9614)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [6]  [ 90/468]  eta: 0:00:17  lr: 0.0015  img/s: 3463.593920156899  loss: 0.5368 (0.5517)  acc1: 98.4375 (97.9653)  acc5: 100.0000 (99.9657)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [6]  [100/468]  eta: 0:00:16  lr: 0.0015  img/s: 3570.3563367449406  loss: 0.5376 (0.5519)  acc1: 97.6562 (97.9502)  acc5: 100.0000 (99.9691)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [6]  [110/468]  eta: 0:00:16  lr: 0.0015  img/s: 3560.765861488055  loss: 0.5520 (0.5527)  acc1: 97.6562 (97.8885)  acc5: 100.0000 (99.9578)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [6]  [120/468]  eta: 0:00:15  lr: 0.0015  img/s: 2785.958465226822  loss: 0.5322 (0.5515)  acc1: 97.6562 (97.9468)  acc5: 100.0000 (99.9548)  time: 0.0393  data: 0.0003  max mem: 190
Epoch: [6]  [130/468]  eta: 0:00:15  lr: 0.0015  img/s: 2802.054875025444  loss: 0.5300 (0.5520)  acc1: 97.6562 (97.8948)  acc5: 100.0000 (99.9523)  time: 0.0415  data: 0.0003  max mem: 190
Epoch: [6]  [140/468]  eta: 0:00:14  lr: 0.0015  img/s: 3535.5810547389497  loss: 0.5530 (0.5526)  acc1: 97.6562 (97.8723)  acc5: 100.0000 (99.9501)  time: 0.0394  data: 0.0003  max mem: 190
Epoch: [6]  [150/468]  eta: 0:00:13  lr: 0.0015  img/s: 3535.7207623714125  loss: 0.5523 (0.5529)  acc1: 97.6562 (97.8477)  acc5: 100.0000 (99.9483)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [6]  [160/468]  eta: 0:00:13  lr: 0.0015  img/s: 3605.6287660008866  loss: 0.5583 (0.5537)  acc1: 97.6562 (97.8164)  acc5: 100.0000 (99.9515)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [6]  [170/468]  eta: 0:00:12  lr: 0.0015  img/s: 3558.688814943458  loss: 0.5609 (0.5539)  acc1: 97.6562 (97.8207)  acc5: 100.0000 (99.9543)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [6]  [180/468]  eta: 0:00:12  lr: 0.0015  img/s: 3531.743416681468  loss: 0.5429 (0.5535)  acc1: 97.6562 (97.8289)  acc5: 100.0000 (99.9568)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [6]  [190/468]  eta: 0:00:11  lr: 0.0015  img/s: 3577.5176053522405  loss: 0.5429 (0.5534)  acc1: 97.6562 (97.8321)  acc5: 100.0000 (99.9550)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [6]  [200/468]  eta: 0:00:11  lr: 0.0015  img/s: 3349.2470928781754  loss: 0.5524 (0.5535)  acc1: 97.6562 (97.8234)  acc5: 100.0000 (99.9572)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [6]  [210/468]  eta: 0:00:10  lr: 0.0015  img/s: 3514.5421290022714  loss: 0.5593 (0.5540)  acc1: 97.6562 (97.8044)  acc5: 100.0000 (99.9556)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [6]  [220/468]  eta: 0:00:10  lr: 0.0015  img/s: 3602.0967767907464  loss: 0.5576 (0.5543)  acc1: 97.6562 (97.7835)  acc5: 100.0000 (99.9540)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [6]  [230/468]  eta: 0:00:09  lr: 0.0015  img/s: 3557.816234700031  loss: 0.5456 (0.5544)  acc1: 97.6562 (97.7746)  acc5: 100.0000 (99.9527)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [6]  [240/468]  eta: 0:00:09  lr: 0.0015  img/s: 3482.5791033932496  loss: 0.5582 (0.5549)  acc1: 96.8750 (97.7470)  acc5: 100.0000 (99.9546)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [6]  [250/468]  eta: 0:00:08  lr: 0.0015  img/s: 3449.9502753555203  loss: 0.5683 (0.5558)  acc1: 96.8750 (97.7185)  acc5: 100.0000 (99.9533)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [6]  [260/468]  eta: 0:00:08  lr: 0.0015  img/s: 3443.1134769057117  loss: 0.5683 (0.5562)  acc1: 96.8750 (97.7011)  acc5: 100.0000 (99.9551)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [6]  [270/468]  eta: 0:00:08  lr: 0.0015  img/s: 3272.545531017415  loss: 0.5686 (0.5569)  acc1: 96.8750 (97.6764)  acc5: 100.0000 (99.9539)  time: 0.0466  data: 0.0031  max mem: 190
Epoch: [6]  [280/468]  eta: 0:00:07  lr: 0.0015  img/s: 3548.8793024808465  loss: 0.5686 (0.5571)  acc1: 96.8750 (97.6646)  acc5: 100.0000 (99.9555)  time: 0.0494  data: 0.0035  max mem: 190
Epoch: [6]  [290/468]  eta: 0:00:07  lr: 0.0015  img/s: 3544.848908227744  loss: 0.5674 (0.5572)  acc1: 97.6562 (97.6616)  acc5: 100.0000 (99.9570)  time: 0.0401  data: 0.0007  max mem: 190
Epoch: [6]  [300/468]  eta: 0:00:06  lr: 0.0015  img/s: 3525.4352825294677  loss: 0.5674 (0.5573)  acc1: 97.6562 (97.6640)  acc5: 100.0000 (99.9533)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [6]  [310/468]  eta: 0:00:06  lr: 0.0015  img/s: 3477.2558178697495  loss: 0.5506 (0.5572)  acc1: 97.6562 (97.6613)  acc5: 100.0000 (99.9548)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [6]  [320/468]  eta: 0:00:06  lr: 0.0015  img/s: 3415.2310892562928  loss: 0.5397 (0.5569)  acc1: 97.6562 (97.6830)  acc5: 100.0000 (99.9538)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [6]  [330/468]  eta: 0:00:05  lr: 0.0015  img/s: 3478.066792347709  loss: 0.5433 (0.5572)  acc1: 97.6562 (97.6657)  acc5: 100.0000 (99.9528)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [6]  [340/468]  eta: 0:00:05  lr: 0.0015  img/s: 2952.0947976751477  loss: 0.5600 (0.5575)  acc1: 96.8750 (97.6631)  acc5: 100.0000 (99.9496)  time: 0.0409  data: 0.0013  max mem: 190
Epoch: [6]  [350/468]  eta: 0:00:04  lr: 0.0015  img/s: 2809.577425871074  loss: 0.5623 (0.5582)  acc1: 96.8750 (97.6295)  acc5: 100.0000 (99.9510)  time: 0.0456  data: 0.0019  max mem: 190
Epoch: [6]  [360/468]  eta: 0:00:04  lr: 0.0015  img/s: 3442.8264385432767  loss: 0.5646 (0.5587)  acc1: 97.6562 (97.6151)  acc5: 100.0000 (99.9459)  time: 0.0457  data: 0.0021  max mem: 190
Epoch: [6]  [370/468]  eta: 0:00:04  lr: 0.0015  img/s: 3640.8888888888887  loss: 0.5475 (0.5584)  acc1: 97.6562 (97.6183)  acc5: 100.0000 (99.9474)  time: 0.0406  data: 0.0015  max mem: 190
Epoch: [6]  [380/468]  eta: 0:00:03  lr: 0.0015  img/s: 3591.614286956696  loss: 0.5605 (0.5589)  acc1: 96.8750 (97.5968)  acc5: 100.0000 (99.9467)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [6]  [390/468]  eta: 0:00:03  lr: 0.0015  img/s: 3500.0157245209953  loss: 0.5546 (0.5586)  acc1: 97.6562 (97.6143)  acc5: 100.0000 (99.9480)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [6]  [400/468]  eta: 0:00:02  lr: 0.0015  img/s: 3382.9508188457394  loss: 0.5446 (0.5587)  acc1: 97.6562 (97.6114)  acc5: 100.0000 (99.9493)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [6]  [410/468]  eta: 0:00:02  lr: 0.0015  img/s: 3362.062260074522  loss: 0.5734 (0.5591)  acc1: 96.8750 (97.5840)  acc5: 100.0000 (99.9487)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [6]  [420/468]  eta: 0:00:01  lr: 0.0015  img/s: 3414.861795237126  loss: 0.5567 (0.5587)  acc1: 96.8750 (97.5987)  acc5: 100.0000 (99.9499)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [6]  [430/468]  eta: 0:00:01  lr: 0.0015  img/s: 3449.7507614408905  loss: 0.5381 (0.5586)  acc1: 98.4375 (97.5964)  acc5: 100.0000 (99.9511)  time: 0.0389  data: 0.0003  max mem: 190
Epoch: [6]  [440/468]  eta: 0:00:01  lr: 0.0015  img/s: 3363.6210035649174  loss: 0.5545 (0.5587)  acc1: 97.6562 (97.5942)  acc5: 100.0000 (99.9504)  time: 0.0389  data: 0.0003  max mem: 190
Epoch: [6]  [450/468]  eta: 0:00:00  lr: 0.0015  img/s: 3469.7497689508755  loss: 0.5516 (0.5587)  acc1: 97.6562 (97.5956)  acc5: 100.0000 (99.9498)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [6]  [460/468]  eta: 0:00:00  lr: 0.0015  img/s: 3430.682352339751  loss: 0.5530 (0.5589)  acc1: 97.6562 (97.5851)  acc5: 100.0000 (99.9492)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [6] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5535 (0.5535)  acc1: 97.6562 (97.6562)  acc5: 100.0000 (100.0000)  time: 0.1226  data: 0.0983  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 97.780 Acc@5 99.980
Epoch: [7]  [  0/468]  eta: 0:01:16  lr: 0.0013090169943749475  img/s: 2489.0627005174047  loss: 0.5265 (0.5265)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1624  data: 0.1110  max mem: 190
Epoch: [7]  [ 10/468]  eta: 0:00:25  lr: 0.0013090169943749475  img/s: 3451.502838370396  loss: 0.5363 (0.5397)  acc1: 97.6562 (98.2955)  acc5: 100.0000 (100.0000)  time: 0.0547  data: 0.0117  max mem: 190
Epoch: [7]  [ 20/468]  eta: 0:00:21  lr: 0.0013090169943749475  img/s: 3377.842518198806  loss: 0.5363 (0.5402)  acc1: 98.4375 (98.2887)  acc5: 100.0000 (100.0000)  time: 0.0429  data: 0.0014  max mem: 190
Epoch: [7]  [ 30/468]  eta: 0:00:20  lr: 0.0013090169943749475  img/s: 1929.3512011931073  loss: 0.5474 (0.5520)  acc1: 97.6562 (97.7571)  acc5: 100.0000 (99.9496)  time: 0.0422  data: 0.0007  max mem: 190
Epoch: [7]  [ 40/468]  eta: 0:00:20  lr: 0.0013090169943749475  img/s: 3102.421348866506  loss: 0.5683 (0.5610)  acc1: 96.0938 (97.3133)  acc5: 100.0000 (99.9428)  time: 0.0472  data: 0.0019  max mem: 190
Epoch: [7]  [ 50/468]  eta: 0:00:19  lr: 0.0013090169943749475  img/s: 3340.3282148279036  loss: 0.5635 (0.5618)  acc1: 96.0938 (97.3039)  acc5: 100.0000 (99.9540)  time: 0.0484  data: 0.0025  max mem: 190
Epoch: [7]  [ 60/468]  eta: 0:00:18  lr: 0.0013090169943749475  img/s: 3524.926050673968  loss: 0.5575 (0.5615)  acc1: 97.6562 (97.3361)  acc5: 100.0000 (99.9616)  time: 0.0415  data: 0.0009  max mem: 190
Epoch: [7]  [ 70/468]  eta: 0:00:17  lr: 0.0013090169943749475  img/s: 3519.4264774328885  loss: 0.5377 (0.5585)  acc1: 98.4375 (97.5022)  acc5: 100.0000 (99.9670)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [7]  [ 80/468]  eta: 0:00:16  lr: 0.0013090169943749475  img/s: 3478.787976180448  loss: 0.5332 (0.5582)  acc1: 98.4375 (97.5598)  acc5: 100.0000 (99.9711)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [7]  [ 90/468]  eta: 0:00:16  lr: 0.0013090169943749475  img/s: 3446.9821188949027  loss: 0.5633 (0.5601)  acc1: 97.6562 (97.4931)  acc5: 100.0000 (99.9657)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [7]  [100/468]  eta: 0:00:15  lr: 0.0013090169943749475  img/s: 3419.908474749019  loss: 0.5747 (0.5619)  acc1: 97.6562 (97.4629)  acc5: 100.0000 (99.9536)  time: 0.0377  data: 0.0002  max mem: 190
Epoch: [7]  [110/468]  eta: 0:00:15  lr: 0.0013090169943749475  img/s: 3515.669853576761  loss: 0.5706 (0.5618)  acc1: 97.6562 (97.4944)  acc5: 100.0000 (99.9507)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [7]  [120/468]  eta: 0:00:14  lr: 0.0013090169943749475  img/s: 3549.489345665871  loss: 0.5600 (0.5613)  acc1: 97.6562 (97.5465)  acc5: 100.0000 (99.9548)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [7]  [130/468]  eta: 0:00:14  lr: 0.0013090169943749475  img/s: 2857.048879522753  loss: 0.5624 (0.5615)  acc1: 97.6562 (97.5310)  acc5: 100.0000 (99.9404)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [7]  [140/468]  eta: 0:00:13  lr: 0.0013090169943749475  img/s: 3495.7767894932185  loss: 0.5542 (0.5605)  acc1: 97.6562 (97.5842)  acc5: 100.0000 (99.9446)  time: 0.0405  data: 0.0003  max mem: 190
Epoch: [7]  [150/468]  eta: 0:00:13  lr: 0.0013090169943749475  img/s: 3554.0007811413934  loss: 0.5517 (0.5603)  acc1: 98.4375 (97.6407)  acc5: 100.0000 (99.9483)  time: 0.0396  data: 0.0003  max mem: 190
Epoch: [7]  [160/468]  eta: 0:00:12  lr: 0.0013090169943749475  img/s: 3441.9214771124502  loss: 0.5565 (0.5605)  acc1: 98.4375 (97.6271)  acc5: 100.0000 (99.9369)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [7]  [170/468]  eta: 0:00:12  lr: 0.0013090169943749475  img/s: 3572.7798651733247  loss: 0.5565 (0.5605)  acc1: 97.6562 (97.6243)  acc5: 100.0000 (99.9269)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [7]  [180/468]  eta: 0:00:11  lr: 0.0013090169943749475  img/s: 3475.117561007185  loss: 0.5568 (0.5610)  acc1: 97.6562 (97.6088)  acc5: 100.0000 (99.9309)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [7]  [190/468]  eta: 0:00:11  lr: 0.0013090169943749475  img/s: 3443.1797233249745  loss: 0.5642 (0.5609)  acc1: 97.6562 (97.6194)  acc5: 100.0000 (99.9346)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [7]  [200/468]  eta: 0:00:10  lr: 0.0013090169943749475  img/s: 3543.702389438944  loss: 0.5639 (0.5611)  acc1: 97.6562 (97.6096)  acc5: 100.0000 (99.9378)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [7]  [210/468]  eta: 0:00:10  lr: 0.0013090169943749475  img/s: 3572.684762861763  loss: 0.5502 (0.5611)  acc1: 97.6562 (97.5970)  acc5: 100.0000 (99.9408)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [7]  [220/468]  eta: 0:00:09  lr: 0.0013090169943749475  img/s: 3566.6087280022853  loss: 0.5558 (0.5613)  acc1: 96.8750 (97.5785)  acc5: 100.0000 (99.9434)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [7]  [230/468]  eta: 0:00:09  lr: 0.0013090169943749475  img/s: 3526.6626727626253  loss: 0.5505 (0.5606)  acc1: 97.6562 (97.6089)  acc5: 100.0000 (99.9459)  time: 0.0363  data: 0.0003  max mem: 190
Epoch: [7]  [240/468]  eta: 0:00:09  lr: 0.0013090169943749475  img/s: 3495.754027269531  loss: 0.5505 (0.5605)  acc1: 98.4375 (97.6109)  acc5: 100.0000 (99.9449)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [7]  [250/468]  eta: 0:00:08  lr: 0.0013090169943749475  img/s: 3506.9660519835124  loss: 0.5430 (0.5600)  acc1: 97.6562 (97.6251)  acc5: 100.0000 (99.9471)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [7]  [260/468]  eta: 0:00:08  lr: 0.0013090169943749475  img/s: 3502.4589128676184  loss: 0.5449 (0.5598)  acc1: 98.4375 (97.6503)  acc5: 100.0000 (99.9461)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [7]  [270/468]  eta: 0:00:07  lr: 0.0013090169943749475  img/s: 3502.9845296585563  loss: 0.5531 (0.5596)  acc1: 97.6562 (97.6534)  acc5: 100.0000 (99.9452)  time: 0.0378  data: 0.0002  max mem: 190
Epoch: [7]  [280/468]  eta: 0:00:07  lr: 0.0013090169943749475  img/s: 3531.3484970071695  loss: 0.5551 (0.5595)  acc1: 97.6562 (97.6590)  acc5: 100.0000 (99.9472)  time: 0.0379  data: 0.0002  max mem: 190
Epoch: [7]  [290/468]  eta: 0:00:06  lr: 0.0013090169943749475  img/s: 3533.8589012782877  loss: 0.5554 (0.5593)  acc1: 97.6562 (97.6562)  acc5: 100.0000 (99.9463)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [7]  [300/468]  eta: 0:00:06  lr: 0.0013090169943749475  img/s: 3543.655608506818  loss: 0.5470 (0.5591)  acc1: 97.6562 (97.6537)  acc5: 100.0000 (99.9481)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [7]  [310/468]  eta: 0:00:06  lr: 0.0013090169943749475  img/s: 1957.5682104910047  loss: 0.5547 (0.5592)  acc1: 97.6562 (97.6437)  acc5: 100.0000 (99.9498)  time: 0.0412  data: 0.0003  max mem: 190
Epoch: [7]  [320/468]  eta: 0:00:06  lr: 0.0013090169943749475  img/s: 1364.188460815254  loss: 0.5524 (0.5589)  acc1: 98.4375 (97.6611)  acc5: 100.0000 (99.9489)  time: 0.0678  data: 0.0030  max mem: 190
Epoch: [7]  [330/468]  eta: 0:00:05  lr: 0.0013090169943749475  img/s: 3207.1908050371576  loss: 0.5477 (0.5592)  acc1: 97.6562 (97.6586)  acc5: 100.0000 (99.9410)  time: 0.0775  data: 0.0048  max mem: 190
Epoch: [7]  [340/468]  eta: 0:00:05  lr: 0.0013090169943749475  img/s: 3364.8437321767688  loss: 0.5512 (0.5589)  acc1: 97.6562 (97.6631)  acc5: 100.0000 (99.9427)  time: 0.0515  data: 0.0021  max mem: 190
Epoch: [7]  [350/468]  eta: 0:00:04  lr: 0.0013090169943749475  img/s: 3425.0574935565364  loss: 0.5448 (0.5588)  acc1: 97.6562 (97.6629)  acc5: 100.0000 (99.9444)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [7]  [360/468]  eta: 0:00:04  lr: 0.0013090169943749475  img/s: 3450.127640432109  loss: 0.5366 (0.5583)  acc1: 98.4375 (97.6887)  acc5: 100.0000 (99.9459)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [7]  [370/468]  eta: 0:00:04  lr: 0.0013090169943749475  img/s: 3478.201991538875  loss: 0.5509 (0.5586)  acc1: 97.6562 (97.6773)  acc5: 100.0000 (99.9410)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [7]  [380/468]  eta: 0:00:03  lr: 0.0013090169943749475  img/s: 3473.7684373989  loss: 0.5645 (0.5588)  acc1: 96.8750 (97.6686)  acc5: 100.0000 (99.9426)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [7]  [390/468]  eta: 0:00:03  lr: 0.0013090169943749475  img/s: 3551.344225858944  loss: 0.5611 (0.5590)  acc1: 96.8750 (97.6443)  acc5: 100.0000 (99.9441)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [7]  [400/468]  eta: 0:00:02  lr: 0.0013090169943749475  img/s: 3478.990863023108  loss: 0.5611 (0.5593)  acc1: 96.8750 (97.6387)  acc5: 100.0000 (99.9454)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [7]  [410/468]  eta: 0:00:02  lr: 0.0013090169943749475  img/s: 3477.931603666634  loss: 0.5499 (0.5592)  acc1: 97.6562 (97.6448)  acc5: 100.0000 (99.9468)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [7]  [420/468]  eta: 0:00:01  lr: 0.0013090169943749475  img/s: 3323.3109373742627  loss: 0.5408 (0.5588)  acc1: 98.4375 (97.6581)  acc5: 100.0000 (99.9480)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [7]  [430/468]  eta: 0:00:01  lr: 0.0013090169943749475  img/s: 3464.0632327416556  loss: 0.5350 (0.5586)  acc1: 97.6562 (97.6581)  acc5: 100.0000 (99.9492)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [7]  [440/468]  eta: 0:00:01  lr: 0.0013090169943749475  img/s: 3507.7221895540138  loss: 0.5480 (0.5584)  acc1: 97.6562 (97.6704)  acc5: 100.0000 (99.9504)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [7]  [450/468]  eta: 0:00:00  lr: 0.0013090169943749475  img/s: 3492.6611239054346  loss: 0.5523 (0.5584)  acc1: 97.6562 (97.6649)  acc5: 100.0000 (99.9515)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [7]  [460/468]  eta: 0:00:00  lr: 0.0013090169943749475  img/s: 3556.143021792409  loss: 0.5525 (0.5584)  acc1: 97.6562 (97.6613)  acc5: 100.0000 (99.9509)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [7] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5532 (0.5532)  acc1: 97.6562 (97.6562)  acc5: 100.0000 (100.0000)  time: 0.1250  data: 0.1056  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 97.530 Acc@5 99.950
Epoch: [8]  [  0/468]  eta: 0:01:36  lr: 0.0011045284632676536  img/s: 2262.6431328916537  loss: 0.5521 (0.5521)  acc1: 97.6562 (97.6562)  acc5: 100.0000 (100.0000)  time: 0.2056  data: 0.1490  max mem: 190
Epoch: [8]  [ 10/468]  eta: 0:00:23  lr: 0.0011045284632676536  img/s: 3470.736735947248  loss: 0.5344 (0.5478)  acc1: 98.4375 (97.8693)  acc5: 100.0000 (100.0000)  time: 0.0523  data: 0.0138  max mem: 190
Epoch: [8]  [ 20/468]  eta: 0:00:20  lr: 0.0011045284632676536  img/s: 3469.368591110594  loss: 0.5449 (0.5517)  acc1: 97.6562 (97.7679)  acc5: 100.0000 (99.9256)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [8]  [ 30/468]  eta: 0:00:18  lr: 0.0011045284632676536  img/s: 3585.66532422342  loss: 0.5611 (0.5518)  acc1: 97.6562 (97.7823)  acc5: 100.0000 (99.9496)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [8]  [ 40/468]  eta: 0:00:17  lr: 0.0011045284632676536  img/s: 3444.195821064044  loss: 0.5630 (0.5562)  acc1: 96.8750 (97.5800)  acc5: 100.0000 (99.9619)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [8]  [ 50/468]  eta: 0:00:16  lr: 0.0011045284632676536  img/s: 3318.9554337007526  loss: 0.5700 (0.5597)  acc1: 96.8750 (97.4877)  acc5: 100.0000 (99.9387)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [8]  [ 60/468]  eta: 0:00:16  lr: 0.0011045284632676536  img/s: 3511.8293507767785  loss: 0.5537 (0.5556)  acc1: 97.6562 (97.6562)  acc5: 100.0000 (99.9488)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [8]  [ 70/468]  eta: 0:00:15  lr: 0.0011045284632676536  img/s: 3525.57418948115  loss: 0.5447 (0.5569)  acc1: 97.6562 (97.6012)  acc5: 100.0000 (99.9450)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [8]  [ 80/468]  eta: 0:00:15  lr: 0.0011045284632676536  img/s: 2674.5390018631624  loss: 0.5500 (0.5547)  acc1: 97.6562 (97.6852)  acc5: 100.0000 (99.9518)  time: 0.0420  data: 0.0008  max mem: 190
Epoch: [8]  [ 90/468]  eta: 0:00:15  lr: 0.0011045284632676536  img/s: 3554.4243162543116  loss: 0.5306 (0.5527)  acc1: 98.4375 (97.7850)  acc5: 100.0000 (99.9571)  time: 0.0483  data: 0.0032  max mem: 190
Epoch: [8]  [100/468]  eta: 0:00:15  lr: 0.0011045284632676536  img/s: 3394.6514239465832  loss: 0.5306 (0.5515)  acc1: 98.4375 (97.8342)  acc5: 100.0000 (99.9536)  time: 0.0435  data: 0.0027  max mem: 190
Epoch: [8]  [110/468]  eta: 0:00:14  lr: 0.0011045284632676536  img/s: 3309.278761280142  loss: 0.5341 (0.5502)  acc1: 98.4375 (97.8885)  acc5: 100.0000 (99.9578)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [8]  [120/468]  eta: 0:00:13  lr: 0.0011045284632676536  img/s: 3566.2059730045703  loss: 0.5408 (0.5502)  acc1: 98.4375 (97.8758)  acc5: 100.0000 (99.9613)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [8]  [130/468]  eta: 0:00:13  lr: 0.0011045284632676536  img/s: 3405.6121235957194  loss: 0.5434 (0.5499)  acc1: 97.6562 (97.8769)  acc5: 100.0000 (99.9642)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [8]  [140/468]  eta: 0:00:13  lr: 0.0011045284632676536  img/s: 3592.7439370415973  loss: 0.5348 (0.5492)  acc1: 97.6562 (97.9000)  acc5: 100.0000 (99.9612)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [8]  [150/468]  eta: 0:00:12  lr: 0.0011045284632676536  img/s: 3486.876656989394  loss: 0.5285 (0.5483)  acc1: 98.4375 (97.9408)  acc5: 100.0000 (99.9534)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [8]  [160/468]  eta: 0:00:12  lr: 0.0011045284632676536  img/s: 3574.635372763651  loss: 0.5372 (0.5475)  acc1: 98.4375 (97.9862)  acc5: 100.0000 (99.9515)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [8]  [170/468]  eta: 0:00:11  lr: 0.0011045284632676536  img/s: 3459.509572322424  loss: 0.5390 (0.5475)  acc1: 98.4375 (97.9989)  acc5: 100.0000 (99.9543)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [8]  [180/468]  eta: 0:00:11  lr: 0.0011045284632676536  img/s: 3559.8686576664986  loss: 0.5453 (0.5478)  acc1: 98.4375 (98.0016)  acc5: 100.0000 (99.9568)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [8]  [190/468]  eta: 0:00:10  lr: 0.0011045284632676536  img/s: 3072.609495902202  loss: 0.5619 (0.5487)  acc1: 97.6562 (97.9917)  acc5: 100.0000 (99.9591)  time: 0.0396  data: 0.0003  max mem: 190
Epoch: [8]  [200/468]  eta: 0:00:10  lr: 0.0011045284632676536  img/s: 3079.9600254717143  loss: 0.5619 (0.5487)  acc1: 97.6562 (97.9711)  acc5: 100.0000 (99.9611)  time: 0.0421  data: 0.0004  max mem: 190
Epoch: [8]  [210/468]  eta: 0:00:10  lr: 0.0011045284632676536  img/s: 3520.095675207847  loss: 0.5388 (0.5484)  acc1: 97.6562 (98.0006)  acc5: 100.0000 (99.9630)  time: 0.0402  data: 0.0004  max mem: 190
Epoch: [8]  [220/468]  eta: 0:00:09  lr: 0.0011045284632676536  img/s: 3515.3705908159322  loss: 0.5334 (0.5481)  acc1: 98.4375 (98.0098)  acc5: 100.0000 (99.9646)  time: 0.0375  data: 0.0002  max mem: 190
Epoch: [8]  [230/468]  eta: 0:00:09  lr: 0.0011045284632676536  img/s: 3547.4488700938286  loss: 0.5408 (0.5483)  acc1: 98.4375 (97.9978)  acc5: 100.0000 (99.9628)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [8]  [240/468]  eta: 0:00:08  lr: 0.0011045284632676536  img/s: 3518.2271735355216  loss: 0.5406 (0.5481)  acc1: 97.6562 (97.9999)  acc5: 100.0000 (99.9611)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [8]  [250/468]  eta: 0:00:08  lr: 0.0011045284632676536  img/s: 3403.798411178809  loss: 0.5354 (0.5477)  acc1: 98.4375 (98.0142)  acc5: 100.0000 (99.9626)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [8]  [260/468]  eta: 0:00:08  lr: 0.0011045284632676536  img/s: 3418.7324851309872  loss: 0.5354 (0.5476)  acc1: 98.4375 (98.0244)  acc5: 100.0000 (99.9581)  time: 0.0415  data: 0.0014  max mem: 190
Epoch: [8]  [270/468]  eta: 0:00:07  lr: 0.0011045284632676536  img/s: 3439.9366438136735  loss: 0.5399 (0.5473)  acc1: 98.4375 (98.0339)  acc5: 100.0000 (99.9596)  time: 0.0417  data: 0.0014  max mem: 190
Epoch: [8]  [280/468]  eta: 0:00:07  lr: 0.0011045284632676536  img/s: 3443.4005631345685  loss: 0.5378 (0.5470)  acc1: 98.4375 (98.0344)  acc5: 100.0000 (99.9611)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [8]  [290/468]  eta: 0:00:06  lr: 0.0011045284632676536  img/s: 3593.2248547640083  loss: 0.5373 (0.5468)  acc1: 97.6562 (98.0402)  acc5: 100.0000 (99.9597)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [8]  [300/468]  eta: 0:00:06  lr: 0.0011045284632676536  img/s: 3540.967780657842  loss: 0.5535 (0.5469)  acc1: 97.6562 (98.0456)  acc5: 100.0000 (99.9611)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [8]  [310/468]  eta: 0:00:06  lr: 0.0011045284632676536  img/s: 3562.325238208987  loss: 0.5366 (0.5469)  acc1: 98.4375 (98.0481)  acc5: 100.0000 (99.9623)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [8]  [320/468]  eta: 0:00:05  lr: 0.0011045284632676536  img/s: 3435.7319612699266  loss: 0.5502 (0.5472)  acc1: 96.8750 (98.0262)  acc5: 100.0000 (99.9635)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [8]  [330/468]  eta: 0:00:05  lr: 0.0011045284632676536  img/s: 3478.8105179943755  loss: 0.5403 (0.5468)  acc1: 97.6562 (98.0433)  acc5: 100.0000 (99.9646)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [8]  [340/468]  eta: 0:00:04  lr: 0.0011045284632676536  img/s: 3482.1725160043325  loss: 0.5320 (0.5465)  acc1: 98.4375 (98.0595)  acc5: 100.0000 (99.9656)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [8]  [350/468]  eta: 0:00:04  lr: 0.0011045284632676536  img/s: 3570.237620865309  loss: 0.5320 (0.5463)  acc1: 98.4375 (98.0747)  acc5: 100.0000 (99.9666)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [8]  [360/468]  eta: 0:00:04  lr: 0.0011045284632676536  img/s: 3473.813391308849  loss: 0.5514 (0.5467)  acc1: 98.4375 (98.0674)  acc5: 100.0000 (99.9675)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [8]  [370/468]  eta: 0:00:03  lr: 0.0011045284632676536  img/s: 3036.1365176132604  loss: 0.5527 (0.5467)  acc1: 97.6562 (98.0711)  acc5: 100.0000 (99.9684)  time: 0.0452  data: 0.0016  max mem: 190
Epoch: [8]  [380/468]  eta: 0:00:03  lr: 0.0011045284632676536  img/s: 3556.378590355061  loss: 0.5467 (0.5466)  acc1: 97.6562 (98.0746)  acc5: 100.0000 (99.9692)  time: 0.0491  data: 0.0032  max mem: 190
Epoch: [8]  [390/468]  eta: 0:00:03  lr: 0.0011045284632676536  img/s: 3614.294450690382  loss: 0.5329 (0.5464)  acc1: 98.4375 (98.0818)  acc5: 100.0000 (99.9700)  time: 0.0408  data: 0.0019  max mem: 190
Epoch: [8]  [400/468]  eta: 0:00:02  lr: 0.0011045284632676536  img/s: 3452.4350471045946  loss: 0.5323 (0.5460)  acc1: 99.2188 (98.1024)  acc5: 100.0000 (99.9708)  time: 0.0368  data: 0.0004  max mem: 190
Epoch: [8]  [410/468]  eta: 0:00:02  lr: 0.0011045284632676536  img/s: 3487.9640334977034  loss: 0.5392 (0.5461)  acc1: 98.4375 (98.1049)  acc5: 100.0000 (99.9696)  time: 0.0366  data: 0.0004  max mem: 190
Epoch: [8]  [420/468]  eta: 0:00:01  lr: 0.0011045284632676536  img/s: 3611.0611942908645  loss: 0.5458 (0.5460)  acc1: 98.4375 (98.1072)  acc5: 100.0000 (99.9703)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [8]  [430/468]  eta: 0:00:01  lr: 0.0011045284632676536  img/s: 3563.412884469873  loss: 0.5407 (0.5459)  acc1: 98.4375 (98.1130)  acc5: 100.0000 (99.9710)  time: 0.0362  data: 0.0002  max mem: 190
Epoch: [8]  [440/468]  eta: 0:00:01  lr: 0.0011045284632676536  img/s: 3511.278111694648  loss: 0.5311 (0.5457)  acc1: 98.4375 (98.1257)  acc5: 100.0000 (99.9699)  time: 0.0363  data: 0.0003  max mem: 190
Epoch: [8]  [450/468]  eta: 0:00:00  lr: 0.0011045284632676536  img/s: 3517.4895465475106  loss: 0.5311 (0.5458)  acc1: 99.2188 (98.1257)  acc5: 100.0000 (99.9688)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [8]  [460/468]  eta: 0:00:00  lr: 0.0011045284632676536  img/s: 3581.455421172358  loss: 0.5253 (0.5455)  acc1: 99.2188 (98.1375)  acc5: 100.0000 (99.9695)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [8] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5504 (0.5504)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1249  data: 0.1060  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.000 Acc@5 99.970
Epoch: [9]  [  0/468]  eta: 0:01:14  lr: 0.0008954715367323467  img/s: 2598.3995043946256  loss: 0.5086 (0.5086)  acc1: 100.0000 (100.0000)  acc5: 100.0000 (100.0000)  time: 0.1588  data: 0.1095  max mem: 190
Epoch: [9]  [ 10/468]  eta: 0:00:23  lr: 0.0008954715367323467  img/s: 2439.9008898462994  loss: 0.5330 (0.5354)  acc1: 98.4375 (98.5085)  acc5: 100.0000 (99.9290)  time: 0.0515  data: 0.0102  max mem: 190
Epoch: [9]  [ 20/468]  eta: 0:00:20  lr: 0.0008954715367323467  img/s: 2203.5869723151436  loss: 0.5330 (0.5314)  acc1: 98.4375 (98.7351)  acc5: 100.0000 (99.9628)  time: 0.0408  data: 0.0003  max mem: 190
Epoch: [9]  [ 30/468]  eta: 0:00:19  lr: 0.0008954715367323467  img/s: 3566.5376469806683  loss: 0.5352 (0.5309)  acc1: 98.4375 (98.6895)  acc5: 100.0000 (99.9748)  time: 0.0395  data: 0.0009  max mem: 190
Epoch: [9]  [ 40/468]  eta: 0:00:18  lr: 0.0008954715367323467  img/s: 3491.2756429848805  loss: 0.5292 (0.5306)  acc1: 98.4375 (98.7424)  acc5: 100.0000 (99.9809)  time: 0.0376  data: 0.0009  max mem: 190
Epoch: [9]  [ 50/468]  eta: 0:00:17  lr: 0.0008954715367323467  img/s: 3387.283666464769  loss: 0.5334 (0.5317)  acc1: 98.4375 (98.6520)  acc5: 100.0000 (99.9847)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [9]  [ 60/468]  eta: 0:00:16  lr: 0.0008954715367323467  img/s: 3530.6517953439434  loss: 0.5342 (0.5321)  acc1: 98.4375 (98.6680)  acc5: 100.0000 (99.9872)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [9]  [ 70/468]  eta: 0:00:15  lr: 0.0008954715367323467  img/s: 3613.8078768990517  loss: 0.5293 (0.5325)  acc1: 98.4375 (98.6356)  acc5: 100.0000 (99.9890)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [9]  [ 80/468]  eta: 0:00:15  lr: 0.0008954715367323467  img/s: 3584.276876856828  loss: 0.5376 (0.5332)  acc1: 98.4375 (98.5822)  acc5: 100.0000 (99.9904)  time: 0.0363  data: 0.0002  max mem: 190
Epoch: [9]  [ 90/468]  eta: 0:00:14  lr: 0.0008954715367323467  img/s: 3455.723990550796  loss: 0.5376 (0.5334)  acc1: 98.4375 (98.5834)  acc5: 100.0000 (99.9914)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [9]  [100/468]  eta: 0:00:14  lr: 0.0008954715367323467  img/s: 3509.9925599032395  loss: 0.5252 (0.5334)  acc1: 99.2188 (98.5767)  acc5: 100.0000 (99.9845)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [9]  [110/468]  eta: 0:00:14  lr: 0.0008954715367323467  img/s: 2140.45439576431  loss: 0.5335 (0.5337)  acc1: 98.4375 (98.5642)  acc5: 100.0000 (99.9859)  time: 0.0437  data: 0.0011  max mem: 190
Epoch: [9]  [120/468]  eta: 0:00:14  lr: 0.0008954715367323467  img/s: 2856.7904303775913  loss: 0.5341 (0.5336)  acc1: 98.4375 (98.5731)  acc5: 100.0000 (99.9871)  time: 0.0544  data: 0.0045  max mem: 190
Epoch: [9]  [130/468]  eta: 0:00:14  lr: 0.0008954715367323467  img/s: 1459.1936682403655  loss: 0.5305 (0.5335)  acc1: 99.2188 (98.5866)  acc5: 100.0000 (99.9881)  time: 0.0578  data: 0.0061  max mem: 190
Epoch: [9]  [140/468]  eta: 0:00:14  lr: 0.0008954715367323467  img/s: 2368.293897860965  loss: 0.5229 (0.5326)  acc1: 99.2188 (98.6370)  acc5: 100.0000 (99.9889)  time: 0.0550  data: 0.0053  max mem: 190
Epoch: [9]  [150/468]  eta: 0:00:13  lr: 0.0008954715367323467  img/s: 3473.1841424283202  loss: 0.5229 (0.5323)  acc1: 99.2188 (98.6445)  acc5: 100.0000 (99.9897)  time: 0.0454  data: 0.0028  max mem: 190
Epoch: [9]  [160/468]  eta: 0:00:13  lr: 0.0008954715367323467  img/s: 3528.6097221126797  loss: 0.5357 (0.5330)  acc1: 98.4375 (98.6122)  acc5: 100.0000 (99.9903)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [9]  [170/468]  eta: 0:00:12  lr: 0.0008954715367323467  img/s: 3583.1286298745936  loss: 0.5247 (0.5322)  acc1: 98.4375 (98.6431)  acc5: 100.0000 (99.9909)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [9]  [180/468]  eta: 0:00:12  lr: 0.0008954715367323467  img/s: 3462.253727493164  loss: 0.5234 (0.5326)  acc1: 99.2188 (98.6360)  acc5: 100.0000 (99.9871)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [9]  [190/468]  eta: 0:00:11  lr: 0.0008954715367323467  img/s: 3527.1955797620376  loss: 0.5289 (0.5326)  acc1: 98.4375 (98.6420)  acc5: 100.0000 (99.9877)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [9]  [200/468]  eta: 0:00:11  lr: 0.0008954715367323467  img/s: 3514.059039914124  loss: 0.5289 (0.5330)  acc1: 98.4375 (98.6435)  acc5: 100.0000 (99.9883)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [9]  [210/468]  eta: 0:00:10  lr: 0.0008954715367323467  img/s: 3603.982868574037  loss: 0.5418 (0.5334)  acc1: 98.4375 (98.6226)  acc5: 100.0000 (99.9852)  time: 0.0376  data: 0.0002  max mem: 190
Epoch: [9]  [220/468]  eta: 0:00:10  lr: 0.0008954715367323467  img/s: 3533.626306505542  loss: 0.5405 (0.5335)  acc1: 98.4375 (98.6178)  acc5: 100.0000 (99.9859)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [9]  [230/468]  eta: 0:00:09  lr: 0.0008954715367323467  img/s: 3551.649645080411  loss: 0.5366 (0.5336)  acc1: 98.4375 (98.6134)  acc5: 100.0000 (99.9831)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [9]  [240/468]  eta: 0:00:09  lr: 0.0008954715367323467  img/s: 3531.1626830134574  loss: 0.5348 (0.5340)  acc1: 98.4375 (98.5996)  acc5: 100.0000 (99.9838)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [9]  [250/468]  eta: 0:00:08  lr: 0.0008954715367323467  img/s: 3386.557194221914  loss: 0.5304 (0.5338)  acc1: 98.4375 (98.6118)  acc5: 100.0000 (99.9844)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [9]  [260/468]  eta: 0:00:08  lr: 0.0008954715367323467  img/s: 3556.143021792409  loss: 0.5290 (0.5342)  acc1: 98.4375 (98.5961)  acc5: 100.0000 (99.9790)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [9]  [270/468]  eta: 0:00:07  lr: 0.0008954715367323467  img/s: 3552.77780204218  loss: 0.5240 (0.5341)  acc1: 98.4375 (98.6076)  acc5: 100.0000 (99.9798)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [9]  [280/468]  eta: 0:00:07  lr: 0.0008954715367323467  img/s: 3469.7273444063853  loss: 0.5242 (0.5341)  acc1: 98.4375 (98.6043)  acc5: 100.0000 (99.9778)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [9]  [290/468]  eta: 0:00:07  lr: 0.0008954715367323467  img/s: 3458.595820341691  loss: 0.5285 (0.5343)  acc1: 98.4375 (98.6013)  acc5: 100.0000 (99.9758)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [9]  [300/468]  eta: 0:00:06  lr: 0.0008954715367323467  img/s: 3508.5245100281663  loss: 0.5416 (0.5351)  acc1: 98.4375 (98.5751)  acc5: 100.0000 (99.9740)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [9]  [310/468]  eta: 0:00:06  lr: 0.0008954715367323467  img/s: 3411.8886325649974  loss: 0.5515 (0.5356)  acc1: 97.6562 (98.5480)  acc5: 100.0000 (99.9749)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [9]  [320/468]  eta: 0:00:05  lr: 0.0008954715367323467  img/s: 3536.535943665312  loss: 0.5444 (0.5360)  acc1: 97.6562 (98.5349)  acc5: 100.0000 (99.9757)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [9]  [330/468]  eta: 0:00:05  lr: 0.0008954715367323467  img/s: 3589.9091407556  loss: 0.5402 (0.5362)  acc1: 98.4375 (98.5272)  acc5: 100.0000 (99.9740)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [9]  [340/468]  eta: 0:00:05  lr: 0.0008954715367323467  img/s: 3572.2093272384905  loss: 0.5437 (0.5365)  acc1: 98.4375 (98.5085)  acc5: 100.0000 (99.9725)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [9]  [350/468]  eta: 0:00:04  lr: 0.0008954715367323467  img/s: 3516.4758143221134  loss: 0.5372 (0.5365)  acc1: 98.4375 (98.4976)  acc5: 100.0000 (99.9733)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [9]  [360/468]  eta: 0:00:04  lr: 0.0008954715367323467  img/s: 3577.255392160129  loss: 0.5323 (0.5364)  acc1: 98.4375 (98.4981)  acc5: 100.0000 (99.9740)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [9]  [370/468]  eta: 0:00:03  lr: 0.0008954715367323467  img/s: 3451.6137892016304  loss: 0.5394 (0.5369)  acc1: 97.6562 (98.4712)  acc5: 100.0000 (99.9747)  time: 0.0375  data: 0.0002  max mem: 190
Epoch: [9]  [380/468]  eta: 0:00:03  lr: 0.0008954715367323467  img/s: 3403.3884345720335  loss: 0.5448 (0.5369)  acc1: 98.4375 (98.4724)  acc5: 100.0000 (99.9754)  time: 0.0378  data: 0.0002  max mem: 190
Epoch: [9]  [390/468]  eta: 0:00:03  lr: 0.0008954715367323467  img/s: 3540.5707954680347  loss: 0.5286 (0.5366)  acc1: 98.4375 (98.4895)  acc5: 100.0000 (99.9760)  time: 0.0375  data: 0.0002  max mem: 190
Epoch: [9]  [400/468]  eta: 0:00:02  lr: 0.0008954715367323467  img/s: 3487.0804884385557  loss: 0.5253 (0.5365)  acc1: 99.2188 (98.4901)  acc5: 100.0000 (99.9747)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [9]  [410/468]  eta: 0:00:02  lr: 0.0008954715367323467  img/s: 3550.0291741056667  loss: 0.5383 (0.5368)  acc1: 98.4375 (98.4793)  acc5: 100.0000 (99.9734)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [9]  [420/468]  eta: 0:00:01  lr: 0.0008954715367323467  img/s: 1396.214262494181  loss: 0.5412 (0.5369)  acc1: 98.4375 (98.4765)  acc5: 100.0000 (99.9740)  time: 0.0427  data: 0.0003  max mem: 190
Epoch: [9]  [430/468]  eta: 0:00:01  lr: 0.0008954715367323467  img/s: 3218.3997170484313  loss: 0.5426 (0.5372)  acc1: 97.6562 (98.4665)  acc5: 100.0000 (99.9746)  time: 0.0484  data: 0.0027  max mem: 190
Epoch: [9]  [440/468]  eta: 0:00:01  lr: 0.0008954715367323467  img/s: 3407.51427755387  loss: 0.5414 (0.5373)  acc1: 97.6562 (98.4588)  acc5: 100.0000 (99.9752)  time: 0.0437  data: 0.0030  max mem: 190
Epoch: [9]  [450/468]  eta: 0:00:00  lr: 0.0008954715367323467  img/s: 3551.7906255168537  loss: 0.5339 (0.5370)  acc1: 98.4375 (98.4618)  acc5: 100.0000 (99.9757)  time: 0.0384  data: 0.0006  max mem: 190
Epoch: [9]  [460/468]  eta: 0:00:00  lr: 0.0008954715367323467  img/s: 3539.6604009942444  loss: 0.5245 (0.5369)  acc1: 98.4375 (98.4697)  acc5: 100.0000 (99.9746)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [9] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:11  loss: 0.5483 (0.5483)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1414  data: 0.1171  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.280 Acc@5 99.960
Epoch: [10]  [  0/468]  eta: 0:01:13  lr: 0.0006909830056250527  img/s: 2420.8236928016163  loss: 0.5375 (0.5375)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1576  data: 0.1046  max mem: 190
Epoch: [10]  [ 10/468]  eta: 0:00:24  lr: 0.0006909830056250527  img/s: 3436.0398087643284  loss: 0.5144 (0.5181)  acc1: 99.2188 (99.2188)  acc5: 100.0000 (100.0000)  time: 0.0530  data: 0.0104  max mem: 190
Epoch: [10]  [ 20/468]  eta: 0:00:21  lr: 0.0006909830056250527  img/s: 3524.069946962138  loss: 0.5215 (0.5234)  acc1: 99.2188 (99.1071)  acc5: 100.0000 (100.0000)  time: 0.0421  data: 0.0007  max mem: 190
Epoch: [10]  [ 30/468]  eta: 0:00:19  lr: 0.0006909830056250527  img/s: 3538.447269731422  loss: 0.5242 (0.5217)  acc1: 99.2188 (99.1431)  acc5: 100.0000 (100.0000)  time: 0.0394  data: 0.0003  max mem: 190
Epoch: [10]  [ 40/468]  eta: 0:00:18  lr: 0.0006909830056250527  img/s: 3512.8633906955442  loss: 0.5161 (0.5239)  acc1: 99.2188 (99.0091)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [10]  [ 50/468]  eta: 0:00:17  lr: 0.0006909830056250527  img/s: 3348.3280029936386  loss: 0.5179 (0.5234)  acc1: 99.2188 (99.0656)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [10]  [ 60/468]  eta: 0:00:16  lr: 0.0006909830056250527  img/s: 3320.2280314415234  loss: 0.5163 (0.5226)  acc1: 99.2188 (99.0907)  acc5: 100.0000 (100.0000)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [10]  [ 70/468]  eta: 0:00:16  lr: 0.0006909830056250527  img/s: 3438.5706453513694  loss: 0.5123 (0.5220)  acc1: 99.2188 (99.1087)  acc5: 100.0000 (100.0000)  time: 0.0385  data: 0.0003  max mem: 190
Epoch: [10]  [ 80/468]  eta: 0:00:15  lr: 0.0006909830056250527  img/s: 3534.6034103627626  loss: 0.5209 (0.5231)  acc1: 99.2188 (99.0451)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [10]  [ 90/468]  eta: 0:00:15  lr: 0.0006909830056250527  img/s: 3577.0170498837356  loss: 0.5209 (0.5231)  acc1: 99.2188 (99.0556)  acc5: 100.0000 (100.0000)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [10]  [100/468]  eta: 0:00:14  lr: 0.0006909830056250527  img/s: 3524.069946962138  loss: 0.5268 (0.5248)  acc1: 98.4375 (99.0022)  acc5: 100.0000 (100.0000)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [10]  [110/468]  eta: 0:00:14  lr: 0.0006909830056250527  img/s: 3550.2639333421507  loss: 0.5263 (0.5245)  acc1: 99.2188 (99.0217)  acc5: 100.0000 (100.0000)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [10]  [120/468]  eta: 0:00:13  lr: 0.0006909830056250527  img/s: 3530.9072206985907  loss: 0.5166 (0.5249)  acc1: 99.2188 (99.0121)  acc5: 100.0000 (100.0000)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [10]  [130/468]  eta: 0:00:13  lr: 0.0006909830056250527  img/s: 3610.624055093751  loss: 0.5242 (0.5256)  acc1: 99.2188 (98.9742)  acc5: 100.0000 (100.0000)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [10]  [140/468]  eta: 0:00:12  lr: 0.0006909830056250527  img/s: 3567.912382370142  loss: 0.5208 (0.5253)  acc1: 99.2188 (99.0082)  acc5: 100.0000 (99.9945)  time: 0.0363  data: 0.0002  max mem: 190
Epoch: [10]  [150/468]  eta: 0:00:12  lr: 0.0006909830056250527  img/s: 2762.3779244768484  loss: 0.5192 (0.5255)  acc1: 99.2188 (99.0014)  acc5: 100.0000 (99.9948)  time: 0.0406  data: 0.0003  max mem: 190
Epoch: [10]  [160/468]  eta: 0:00:12  lr: 0.0006909830056250527  img/s: 3191.291160910658  loss: 0.5192 (0.5253)  acc1: 99.2188 (99.0149)  acc5: 100.0000 (99.9951)  time: 0.0430  data: 0.0003  max mem: 190
Epoch: [10]  [170/468]  eta: 0:00:11  lr: 0.0006909830056250527  img/s: 3518.8267232960393  loss: 0.5229 (0.5257)  acc1: 98.4375 (98.9766)  acc5: 100.0000 (99.9954)  time: 0.0413  data: 0.0003  max mem: 190
Epoch: [10]  [180/468]  eta: 0:00:11  lr: 0.0006909830056250527  img/s: 3413.4939311669073  loss: 0.5251 (0.5263)  acc1: 98.4375 (98.9555)  acc5: 100.0000 (99.9957)  time: 0.0395  data: 0.0003  max mem: 190
Epoch: [10]  [190/468]  eta: 0:00:10  lr: 0.0006909830056250527  img/s: 1556.440813486599  loss: 0.5210 (0.5262)  acc1: 99.2188 (98.9570)  acc5: 100.0000 (99.9959)  time: 0.0407  data: 0.0002  max mem: 190
Epoch: [10]  [200/468]  eta: 0:00:10  lr: 0.0006909830056250527  img/s: 2556.966489493437  loss: 0.5269 (0.5272)  acc1: 98.4375 (98.9195)  acc5: 100.0000 (99.9961)  time: 0.0485  data: 0.0027  max mem: 190
Epoch: [10]  [210/468]  eta: 0:00:10  lr: 0.0006909830056250527  img/s: 3502.276126608042  loss: 0.5357 (0.5273)  acc1: 98.4375 (98.9114)  acc5: 100.0000 (99.9963)  time: 0.0458  data: 0.0027  max mem: 190
Epoch: [10]  [220/468]  eta: 0:00:09  lr: 0.0006909830056250527  img/s: 3562.79804629433  loss: 0.5338 (0.5279)  acc1: 98.4375 (98.8758)  acc5: 100.0000 (99.9965)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [10]  [230/468]  eta: 0:00:09  lr: 0.0006909830056250527  img/s: 3597.125038525963  loss: 0.5187 (0.5277)  acc1: 99.2188 (98.8907)  acc5: 100.0000 (99.9966)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [10]  [240/468]  eta: 0:00:09  lr: 0.0006909830056250527  img/s: 3483.7994354498555  loss: 0.5226 (0.5282)  acc1: 99.2188 (98.8654)  acc5: 100.0000 (99.9968)  time: 0.0364  data: 0.0002  max mem: 190
Epoch: [10]  [250/468]  eta: 0:00:08  lr: 0.0006909830056250527  img/s: 3499.057647311856  loss: 0.5308 (0.5282)  acc1: 98.4375 (98.8546)  acc5: 100.0000 (99.9969)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [10]  [260/468]  eta: 0:00:08  lr: 0.0006909830056250527  img/s: 3570.4988062222756  loss: 0.5202 (0.5281)  acc1: 99.2188 (98.8685)  acc5: 100.0000 (99.9940)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [10]  [270/468]  eta: 0:00:07  lr: 0.0006909830056250527  img/s: 3405.525712509594  loss: 0.5211 (0.5282)  acc1: 99.2188 (98.8555)  acc5: 100.0000 (99.9942)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [10]  [280/468]  eta: 0:00:07  lr: 0.0006909830056250527  img/s: 3477.5711518904527  loss: 0.5305 (0.5286)  acc1: 98.4375 (98.8295)  acc5: 100.0000 (99.9944)  time: 0.0385  data: 0.0003  max mem: 190
Epoch: [10]  [290/468]  eta: 0:00:06  lr: 0.0006909830056250527  img/s: 3542.112530349414  loss: 0.5300 (0.5288)  acc1: 98.4375 (98.8187)  acc5: 100.0000 (99.9946)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [10]  [300/468]  eta: 0:00:06  lr: 0.0006909830056250527  img/s: 3544.1234734159834  loss: 0.5232 (0.5288)  acc1: 99.2188 (98.8268)  acc5: 100.0000 (99.9948)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [10]  [310/468]  eta: 0:00:06  lr: 0.0006909830056250527  img/s: 3441.08315707162  loss: 0.5262 (0.5292)  acc1: 99.2188 (98.8118)  acc5: 100.0000 (99.9950)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [10]  [320/468]  eta: 0:00:05  lr: 0.0006909830056250527  img/s: 3481.9466751412247  loss: 0.5417 (0.5296)  acc1: 98.4375 (98.7977)  acc5: 100.0000 (99.9951)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [10]  [330/468]  eta: 0:00:05  lr: 0.0006909830056250527  img/s: 3582.1245170975812  loss: 0.5399 (0.5302)  acc1: 97.6562 (98.7632)  acc5: 100.0000 (99.9953)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [10]  [340/468]  eta: 0:00:04  lr: 0.0006909830056250527  img/s: 3493.865795484866  loss: 0.5399 (0.5307)  acc1: 97.6562 (98.7422)  acc5: 100.0000 (99.9954)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [10]  [350/468]  eta: 0:00:04  lr: 0.0006909830056250527  img/s: 3524.671489909269  loss: 0.5373 (0.5309)  acc1: 98.4375 (98.7313)  acc5: 100.0000 (99.9955)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [10]  [360/468]  eta: 0:00:04  lr: 0.0006909830056250527  img/s: 3499.719120753044  loss: 0.5248 (0.5307)  acc1: 98.4375 (98.7383)  acc5: 100.0000 (99.9957)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [10]  [370/468]  eta: 0:00:03  lr: 0.0006909830056250527  img/s: 3413.9063461783035  loss: 0.5168 (0.5305)  acc1: 99.2188 (98.7449)  acc5: 100.0000 (99.9958)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [10]  [380/468]  eta: 0:00:03  lr: 0.0006909830056250527  img/s: 3589.9571508813224  loss: 0.5213 (0.5303)  acc1: 99.2188 (98.7533)  acc5: 100.0000 (99.9959)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [10]  [390/468]  eta: 0:00:03  lr: 0.0006909830056250527  img/s: 3470.4450736273257  loss: 0.5237 (0.5301)  acc1: 99.2188 (98.7612)  acc5: 100.0000 (99.9960)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [10]  [400/468]  eta: 0:00:02  lr: 0.0006909830056250527  img/s: 3534.463791014905  loss: 0.5256 (0.5300)  acc1: 99.2188 (98.7629)  acc5: 100.0000 (99.9961)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [10]  [410/468]  eta: 0:00:02  lr: 0.0006909830056250527  img/s: 2385.032927587739  loss: 0.5269 (0.5300)  acc1: 98.4375 (98.7625)  acc5: 100.0000 (99.9962)  time: 0.0418  data: 0.0006  max mem: 190
Epoch: [10]  [420/468]  eta: 0:00:01  lr: 0.0006909830056250527  img/s: 1669.8991660938293  loss: 0.5245 (0.5301)  acc1: 98.4375 (98.7622)  acc5: 100.0000 (99.9963)  time: 0.0561  data: 0.0036  max mem: 190
Epoch: [10]  [430/468]  eta: 0:00:01  lr: 0.0006909830056250527  img/s: 3159.8113780561016  loss: 0.5219 (0.5300)  acc1: 98.4375 (98.7583)  acc5: 100.0000 (99.9964)  time: 0.0600  data: 0.0044  max mem: 190
Epoch: [10]  [440/468]  eta: 0:00:01  lr: 0.0006909830056250527  img/s: 3148.636799230539  loss: 0.5261 (0.5301)  acc1: 98.4375 (98.7564)  acc5: 100.0000 (99.9929)  time: 0.0488  data: 0.0015  max mem: 190
Epoch: [10]  [450/468]  eta: 0:00:00  lr: 0.0006909830056250527  img/s: 3290.860071104573  loss: 0.5279 (0.5301)  acc1: 98.4375 (98.7562)  acc5: 100.0000 (99.9931)  time: 0.0413  data: 0.0004  max mem: 190
Epoch: [10]  [460/468]  eta: 0:00:00  lr: 0.0006909830056250527  img/s: 1988.366599260757  loss: 0.5279 (0.5301)  acc1: 99.2188 (98.7578)  acc5: 100.0000 (99.9932)  time: 0.0415  data: 0.0003  max mem: 190
Epoch: [10] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:10  loss: 0.5544 (0.5544)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1328  data: 0.1155  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.270 Acc@5 99.970
Epoch: [11]  [  0/468]  eta: 0:01:20  lr: 0.0005000000000000002  img/s: 2074.494628974173  loss: 0.5164 (0.5164)  acc1: 99.2188 (99.2188)  acc5: 100.0000 (100.0000)  time: 0.1717  data: 0.1100  max mem: 190
Epoch: [11]  [ 10/468]  eta: 0:00:25  lr: 0.0005000000000000002  img/s: 3200.5181197651195  loss: 0.5164 (0.5185)  acc1: 99.2188 (99.1477)  acc5: 100.0000 (100.0000)  time: 0.0554  data: 0.0103  max mem: 190
Epoch: [11]  [ 20/468]  eta: 0:00:21  lr: 0.0005000000000000002  img/s: 3149.2278254544603  loss: 0.5217 (0.5232)  acc1: 99.2188 (98.9583)  acc5: 100.0000 (100.0000)  time: 0.0423  data: 0.0004  max mem: 190
Epoch: [11]  [ 30/468]  eta: 0:00:20  lr: 0.0005000000000000002  img/s: 3227.5514728868584  loss: 0.5288 (0.5235)  acc1: 99.2188 (98.9919)  acc5: 100.0000 (100.0000)  time: 0.0406  data: 0.0004  max mem: 190
Epoch: [11]  [ 40/468]  eta: 0:00:19  lr: 0.0005000000000000002  img/s: 3470.4226400946354  loss: 0.5152 (0.5219)  acc1: 99.2188 (99.0663)  acc5: 100.0000 (99.9809)  time: 0.0402  data: 0.0003  max mem: 190
Epoch: [11]  [ 50/468]  eta: 0:00:18  lr: 0.0005000000000000002  img/s: 3297.773387879458  loss: 0.5106 (0.5212)  acc1: 99.2188 (99.0656)  acc5: 100.0000 (99.9847)  time: 0.0390  data: 0.0003  max mem: 190
Epoch: [11]  [ 60/468]  eta: 0:00:17  lr: 0.0005000000000000002  img/s: 3524.5095158378467  loss: 0.5191 (0.5218)  acc1: 99.2188 (99.0523)  acc5: 100.0000 (99.9872)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [11]  [ 70/468]  eta: 0:00:16  lr: 0.0005000000000000002  img/s: 3443.8423277504444  loss: 0.5181 (0.5218)  acc1: 99.2188 (99.0427)  acc5: 100.0000 (99.9890)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [11]  [ 80/468]  eta: 0:00:15  lr: 0.0005000000000000002  img/s: 3546.980126849894  loss: 0.5163 (0.5221)  acc1: 99.2188 (99.0355)  acc5: 100.0000 (99.9904)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [11]  [ 90/468]  eta: 0:00:15  lr: 0.0005000000000000002  img/s: 3493.1610753975483  loss: 0.5179 (0.5222)  acc1: 99.2188 (99.0299)  acc5: 100.0000 (99.9914)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [11]  [100/468]  eta: 0:00:14  lr: 0.0005000000000000002  img/s: 3373.745770806626  loss: 0.5201 (0.5224)  acc1: 99.2188 (99.0176)  acc5: 100.0000 (99.9923)  time: 0.0377  data: 0.0004  max mem: 190
Epoch: [11]  [110/468]  eta: 0:00:14  lr: 0.0005000000000000002  img/s: 3415.90471342767  loss: 0.5171 (0.5224)  acc1: 99.2188 (99.0146)  acc5: 100.0000 (99.9930)  time: 0.0377  data: 0.0004  max mem: 190
Epoch: [11]  [120/468]  eta: 0:00:13  lr: 0.0005000000000000002  img/s: 3553.130498087334  loss: 0.5166 (0.5222)  acc1: 99.2188 (99.0121)  acc5: 100.0000 (99.9935)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [11]  [130/468]  eta: 0:00:13  lr: 0.0005000000000000002  img/s: 3494.0249651815116  loss: 0.5192 (0.5222)  acc1: 99.2188 (99.0160)  acc5: 100.0000 (99.9940)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [11]  [140/468]  eta: 0:00:12  lr: 0.0005000000000000002  img/s: 3481.5402354009275  loss: 0.5231 (0.5224)  acc1: 99.2188 (99.0193)  acc5: 100.0000 (99.9945)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [11]  [150/468]  eta: 0:00:12  lr: 0.0005000000000000002  img/s: 3100.5042389521586  loss: 0.5230 (0.5225)  acc1: 99.2188 (99.0273)  acc5: 100.0000 (99.9948)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [11]  [160/468]  eta: 0:00:12  lr: 0.0005000000000000002  img/s: 3358.0876940590715  loss: 0.5219 (0.5226)  acc1: 99.2188 (99.0392)  acc5: 100.0000 (99.9951)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [11]  [170/468]  eta: 0:00:11  lr: 0.0005000000000000002  img/s: 3427.9004456703574  loss: 0.5155 (0.5224)  acc1: 99.2188 (99.0451)  acc5: 100.0000 (99.9954)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [11]  [180/468]  eta: 0:00:11  lr: 0.0005000000000000002  img/s: 3247.5632096301  loss: 0.5159 (0.5225)  acc1: 99.2188 (99.0331)  acc5: 100.0000 (99.9957)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [11]  [190/468]  eta: 0:00:10  lr: 0.0005000000000000002  img/s: 3533.370487617067  loss: 0.5161 (0.5224)  acc1: 99.2188 (99.0388)  acc5: 100.0000 (99.9959)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [11]  [200/468]  eta: 0:00:10  lr: 0.0005000000000000002  img/s: 3471.7018145135216  loss: 0.5194 (0.5228)  acc1: 98.4375 (99.0244)  acc5: 100.0000 (99.9961)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [11]  [210/468]  eta: 0:00:10  lr: 0.0005000000000000002  img/s: 3124.103347143988  loss: 0.5306 (0.5233)  acc1: 98.4375 (99.0077)  acc5: 100.0000 (99.9963)  time: 0.0385  data: 0.0003  max mem: 190
Epoch: [11]  [220/468]  eta: 0:00:09  lr: 0.0005000000000000002  img/s: 3118.1337344709223  loss: 0.5267 (0.5233)  acc1: 99.2188 (99.0243)  acc5: 100.0000 (99.9965)  time: 0.0409  data: 0.0003  max mem: 190
Epoch: [11]  [230/468]  eta: 0:00:09  lr: 0.0005000000000000002  img/s: 2732.584679594849  loss: 0.5109 (0.5232)  acc1: 100.0000 (99.0294)  acc5: 100.0000 (99.9966)  time: 0.0512  data: 0.0011  max mem: 190
Epoch: [11]  [240/468]  eta: 0:00:09  lr: 0.0005000000000000002  img/s: 3494.184149381374  loss: 0.5244 (0.5235)  acc1: 98.4375 (99.0113)  acc5: 100.0000 (99.9935)  time: 0.0530  data: 0.0019  max mem: 190
Epoch: [11]  [250/468]  eta: 0:00:08  lr: 0.0005000000000000002  img/s: 3441.568450473089  loss: 0.5169 (0.5231)  acc1: 99.2188 (99.0382)  acc5: 100.0000 (99.9938)  time: 0.0464  data: 0.0010  max mem: 190
Epoch: [11]  [260/468]  eta: 0:00:08  lr: 0.0005000000000000002  img/s: 3567.485626951957  loss: 0.5087 (0.5230)  acc1: 99.2188 (99.0451)  acc5: 100.0000 (99.9940)  time: 0.0440  data: 0.0006  max mem: 190
Epoch: [11]  [270/468]  eta: 0:00:08  lr: 0.0005000000000000002  img/s: 3464.443245615167  loss: 0.5222 (0.5231)  acc1: 99.2188 (99.0515)  acc5: 100.0000 (99.9942)  time: 0.0395  data: 0.0006  max mem: 190
Epoch: [11]  [280/468]  eta: 0:00:07  lr: 0.0005000000000000002  img/s: 3437.4478144228246  loss: 0.5222 (0.5231)  acc1: 99.2188 (99.0547)  acc5: 100.0000 (99.9944)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [11]  [290/468]  eta: 0:00:07  lr: 0.0005000000000000002  img/s: 3604.3216069606315  loss: 0.5156 (0.5230)  acc1: 99.2188 (99.0630)  acc5: 100.0000 (99.9946)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [11]  [300/468]  eta: 0:00:06  lr: 0.0005000000000000002  img/s: 3499.4225672513476  loss: 0.5130 (0.5227)  acc1: 99.2188 (99.0734)  acc5: 100.0000 (99.9922)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [11]  [310/468]  eta: 0:00:06  lr: 0.0005000000000000002  img/s: 3521.6196261069204  loss: 0.5125 (0.5226)  acc1: 99.2188 (99.0756)  acc5: 100.0000 (99.9925)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [11]  [320/468]  eta: 0:00:05  lr: 0.0005000000000000002  img/s: 3499.6506808685394  loss: 0.5142 (0.5223)  acc1: 99.2188 (99.0873)  acc5: 100.0000 (99.9927)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [11]  [330/468]  eta: 0:00:05  lr: 0.0005000000000000002  img/s: 3439.870522127466  loss: 0.5142 (0.5224)  acc1: 99.2188 (99.0889)  acc5: 100.0000 (99.9906)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [11]  [340/468]  eta: 0:00:05  lr: 0.0005000000000000002  img/s: 3406.044244811988  loss: 0.5148 (0.5224)  acc1: 99.2188 (99.0927)  acc5: 100.0000 (99.9908)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [11]  [350/468]  eta: 0:00:04  lr: 0.0005000000000000002  img/s: 3536.6524288217547  loss: 0.5167 (0.5225)  acc1: 99.2188 (99.0919)  acc5: 100.0000 (99.9889)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [11]  [360/468]  eta: 0:00:04  lr: 0.0005000000000000002  img/s: 3499.1944833699413  loss: 0.5167 (0.5223)  acc1: 99.2188 (99.1041)  acc5: 100.0000 (99.9870)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [11]  [370/468]  eta: 0:00:03  lr: 0.0005000000000000002  img/s: 3468.449624322456  loss: 0.5129 (0.5222)  acc1: 99.2188 (99.1029)  acc5: 100.0000 (99.9874)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [11]  [380/468]  eta: 0:00:03  lr: 0.0005000000000000002  img/s: 3447.4469402170425  loss: 0.5251 (0.5225)  acc1: 99.2188 (99.0957)  acc5: 100.0000 (99.9877)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [11]  [390/468]  eta: 0:00:03  lr: 0.0005000000000000002  img/s: 3585.473750292183  loss: 0.5270 (0.5227)  acc1: 98.4375 (99.0869)  acc5: 100.0000 (99.9880)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [11]  [400/468]  eta: 0:00:02  lr: 0.0005000000000000002  img/s: 3455.968689248516  loss: 0.5184 (0.5230)  acc1: 99.2188 (99.0804)  acc5: 100.0000 (99.9864)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [11]  [410/468]  eta: 0:00:02  lr: 0.0005000000000000002  img/s: 3584.205089860336  loss: 0.5204 (0.5232)  acc1: 98.4375 (99.0705)  acc5: 100.0000 (99.9867)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [11]  [420/468]  eta: 0:00:01  lr: 0.0005000000000000002  img/s: 3539.0304021094266  loss: 0.5257 (0.5233)  acc1: 98.4375 (99.0684)  acc5: 100.0000 (99.9852)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [11]  [430/468]  eta: 0:00:01  lr: 0.0005000000000000002  img/s: 3503.738951105542  loss: 0.5203 (0.5233)  acc1: 99.2188 (99.0683)  acc5: 100.0000 (99.9855)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [11]  [440/468]  eta: 0:00:01  lr: 0.0005000000000000002  img/s: 3516.6140161266026  loss: 0.5149 (0.5232)  acc1: 99.2188 (99.0699)  acc5: 100.0000 (99.9858)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [11]  [450/468]  eta: 0:00:00  lr: 0.0005000000000000002  img/s: 3491.661650125522  loss: 0.5159 (0.5232)  acc1: 99.2188 (99.0698)  acc5: 100.0000 (99.9861)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [11]  [460/468]  eta: 0:00:00  lr: 0.0005000000000000002  img/s: 3486.265305592353  loss: 0.5175 (0.5231)  acc1: 99.2188 (99.0747)  acc5: 100.0000 (99.9864)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [11] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:10  loss: 0.5518 (0.5518)  acc1: 97.6562 (97.6562)  acc5: 100.0000 (100.0000)  time: 0.1369  data: 0.1105  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.430 Acc@5 99.990
Epoch: [12]  [  0/468]  eta: 0:02:14  lr: 0.0003308693936411421  img/s: 948.6122759093495  loss: 0.5233 (0.5233)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.2872  data: 0.1522  max mem: 190
Epoch: [12]  [ 10/468]  eta: 0:00:36  lr: 0.0003308693936411421  img/s: 3477.0756526751425  loss: 0.5233 (0.5273)  acc1: 99.2188 (98.9347)  acc5: 100.0000 (100.0000)  time: 0.0804  data: 0.0196  max mem: 190
Epoch: [12]  [ 20/468]  eta: 0:00:26  lr: 0.0003308693936411421  img/s: 3368.8977353304763  loss: 0.5224 (0.5260)  acc1: 99.2188 (99.0327)  acc5: 100.0000 (100.0000)  time: 0.0488  data: 0.0033  max mem: 190
Epoch: [12]  [ 30/468]  eta: 0:00:23  lr: 0.0003308693936411421  img/s: 3471.9487812922375  loss: 0.5112 (0.5222)  acc1: 99.2188 (99.2188)  acc5: 100.0000 (100.0000)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [12]  [ 40/468]  eta: 0:00:21  lr: 0.0003308693936411421  img/s: 3426.893938620232  loss: 0.5088 (0.5198)  acc1: 99.2188 (99.2950)  acc5: 100.0000 (100.0000)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [12]  [ 50/468]  eta: 0:00:19  lr: 0.0003308693936411421  img/s: 3567.248584717608  loss: 0.5111 (0.5197)  acc1: 99.2188 (99.3107)  acc5: 100.0000 (99.9694)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [12]  [ 60/468]  eta: 0:00:18  lr: 0.0003308693936411421  img/s: 3598.4511009082075  loss: 0.5193 (0.5200)  acc1: 99.2188 (99.2572)  acc5: 100.0000 (99.9744)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [12]  [ 70/468]  eta: 0:00:17  lr: 0.0003308693936411421  img/s: 3365.2866634907105  loss: 0.5134 (0.5187)  acc1: 99.2188 (99.2958)  acc5: 100.0000 (99.9780)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [12]  [ 80/468]  eta: 0:00:16  lr: 0.0003308693936411421  img/s: 3487.42026048264  loss: 0.5116 (0.5181)  acc1: 99.2188 (99.3248)  acc5: 100.0000 (99.9807)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [12]  [ 90/468]  eta: 0:00:16  lr: 0.0003308693936411421  img/s: 2787.694391077233  loss: 0.5146 (0.5188)  acc1: 99.2188 (99.2788)  acc5: 100.0000 (99.9742)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [12]  [100/468]  eta: 0:00:15  lr: 0.0003308693936411421  img/s: 3409.505166293034  loss: 0.5101 (0.5178)  acc1: 99.2188 (99.3038)  acc5: 100.0000 (99.9768)  time: 0.0386  data: 0.0002  max mem: 190
Epoch: [12]  [110/468]  eta: 0:00:14  lr: 0.0003308693936411421  img/s: 3557.1090512757655  loss: 0.5063 (0.5172)  acc1: 100.0000 (99.3243)  acc5: 100.0000 (99.9789)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [12]  [120/468]  eta: 0:00:14  lr: 0.0003308693936411421  img/s: 3579.1871358284775  loss: 0.5063 (0.5173)  acc1: 100.0000 (99.3285)  acc5: 100.0000 (99.9806)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [12]  [130/468]  eta: 0:00:13  lr: 0.0003308693936411421  img/s: 3457.0591318570223  loss: 0.5063 (0.5167)  acc1: 99.2188 (99.3380)  acc5: 100.0000 (99.9821)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [12]  [140/468]  eta: 0:00:13  lr: 0.0003308693936411421  img/s: 3483.822042257177  loss: 0.5083 (0.5166)  acc1: 99.2188 (99.3351)  acc5: 100.0000 (99.9834)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [12]  [150/468]  eta: 0:00:12  lr: 0.0003308693936411421  img/s: 3503.2359673735727  loss: 0.5090 (0.5161)  acc1: 99.2188 (99.3429)  acc5: 100.0000 (99.9845)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [12]  [160/468]  eta: 0:00:12  lr: 0.0003308693936411421  img/s: 3429.1046543563934  loss: 0.5088 (0.5158)  acc1: 99.2188 (99.3595)  acc5: 100.0000 (99.9854)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [12]  [170/468]  eta: 0:00:11  lr: 0.0003308693936411421  img/s: 3477.0981723034674  loss: 0.5088 (0.5158)  acc1: 99.2188 (99.3604)  acc5: 100.0000 (99.9863)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [12]  [180/468]  eta: 0:00:11  lr: 0.0003308693936411421  img/s: 2182.446439968292  loss: 0.5171 (0.5164)  acc1: 99.2188 (99.3482)  acc5: 100.0000 (99.9871)  time: 0.0454  data: 0.0003  max mem: 190
Epoch: [12]  [190/468]  eta: 0:00:11  lr: 0.0003308693936411421  img/s: 2737.196131315037  loss: 0.5142 (0.5163)  acc1: 99.2188 (99.3496)  acc5: 100.0000 (99.9877)  time: 0.0563  data: 0.0035  max mem: 190
Epoch: [12]  [200/468]  eta: 0:00:11  lr: 0.0003308693936411421  img/s: 3381.0972755784514  loss: 0.5142 (0.5168)  acc1: 99.2188 (99.3276)  acc5: 100.0000 (99.9883)  time: 0.0535  data: 0.0055  max mem: 190
Epoch: [12]  [210/468]  eta: 0:00:10  lr: 0.0003308693936411421  img/s: 3422.4374123466864  loss: 0.5139 (0.5166)  acc1: 99.2188 (99.3372)  acc5: 100.0000 (99.9889)  time: 0.0428  data: 0.0023  max mem: 190
Epoch: [12]  [220/468]  eta: 0:00:10  lr: 0.0003308693936411421  img/s: 3410.0898904951855  loss: 0.5099 (0.5163)  acc1: 100.0000 (99.3460)  acc5: 100.0000 (99.9894)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [12]  [230/468]  eta: 0:00:09  lr: 0.0003308693936411421  img/s: 3545.2936763695916  loss: 0.5056 (0.5161)  acc1: 100.0000 (99.3506)  acc5: 100.0000 (99.9899)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [12]  [240/468]  eta: 0:00:09  lr: 0.0003308693936411421  img/s: 3464.9798763408244  loss: 0.5121 (0.5160)  acc1: 99.2188 (99.3517)  acc5: 100.0000 (99.9903)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [12]  [250/468]  eta: 0:00:08  lr: 0.0003308693936411421  img/s: 3481.901510483887  loss: 0.5145 (0.5162)  acc1: 99.2188 (99.3464)  acc5: 100.0000 (99.9907)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [12]  [260/468]  eta: 0:00:08  lr: 0.0003308693936411421  img/s: 3417.1875067628207  loss: 0.5087 (0.5161)  acc1: 99.2188 (99.3534)  acc5: 100.0000 (99.9880)  time: 0.0384  data: 0.0003  max mem: 190
Epoch: [12]  [270/468]  eta: 0:00:08  lr: 0.0003308693936411421  img/s: 2321.513592984489  loss: 0.5086 (0.5161)  acc1: 99.2188 (99.3571)  acc5: 100.0000 (99.9885)  time: 0.0396  data: 0.0003  max mem: 190
Epoch: [12]  [280/468]  eta: 0:00:07  lr: 0.0003308693936411421  img/s: 2544.41190521327  loss: 0.5116 (0.5161)  acc1: 99.2188 (99.3550)  acc5: 100.0000 (99.9889)  time: 0.0479  data: 0.0041  max mem: 190
Epoch: [12]  [290/468]  eta: 0:00:07  lr: 0.0003308693936411421  img/s: 3506.5995571608655  loss: 0.5128 (0.5162)  acc1: 99.2188 (99.3449)  acc5: 100.0000 (99.9893)  time: 0.0457  data: 0.0041  max mem: 190
Epoch: [12]  [300/468]  eta: 0:00:06  lr: 0.0003308693936411421  img/s: 3373.660969233863  loss: 0.5145 (0.5163)  acc1: 99.2188 (99.3433)  acc5: 100.0000 (99.9896)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [12]  [310/468]  eta: 0:00:06  lr: 0.0003308693936411421  img/s: 3646.9483394583285  loss: 0.5117 (0.5164)  acc1: 99.2188 (99.3343)  acc5: 100.0000 (99.9900)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [12]  [320/468]  eta: 0:00:06  lr: 0.0003308693936411421  img/s: 2982.964190266642  loss: 0.5159 (0.5167)  acc1: 99.2188 (99.3137)  acc5: 100.0000 (99.9903)  time: 0.0384  data: 0.0003  max mem: 190
Epoch: [12]  [330/468]  eta: 0:00:05  lr: 0.0003308693936411421  img/s: 3034.3518772854954  loss: 0.5159 (0.5167)  acc1: 99.2188 (99.3202)  acc5: 100.0000 (99.9906)  time: 0.0444  data: 0.0004  max mem: 190
Epoch: [12]  [340/468]  eta: 0:00:05  lr: 0.0003308693936411421  img/s: 3081.7810433504774  loss: 0.5096 (0.5168)  acc1: 99.2188 (99.3173)  acc5: 100.0000 (99.9908)  time: 0.0476  data: 0.0014  max mem: 190
Epoch: [12]  [350/468]  eta: 0:00:04  lr: 0.0003308693936411421  img/s: 3615.1219269125363  loss: 0.5097 (0.5167)  acc1: 99.2188 (99.3234)  acc5: 100.0000 (99.9911)  time: 0.0413  data: 0.0013  max mem: 190
Epoch: [12]  [360/468]  eta: 0:00:04  lr: 0.0003308693936411421  img/s: 3503.578895161027  loss: 0.5110 (0.5166)  acc1: 99.2188 (99.3248)  acc5: 100.0000 (99.9913)  time: 0.0364  data: 0.0003  max mem: 190
Epoch: [12]  [370/468]  eta: 0:00:04  lr: 0.0003308693936411421  img/s: 3446.561674263337  loss: 0.5092 (0.5166)  acc1: 99.2188 (99.3240)  acc5: 100.0000 (99.9916)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [12]  [380/468]  eta: 0:00:03  lr: 0.0003308693936411421  img/s: 3527.2419271124195  loss: 0.5092 (0.5165)  acc1: 99.2188 (99.3233)  acc5: 100.0000 (99.9918)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [12]  [390/468]  eta: 0:00:03  lr: 0.0003308693936411421  img/s: 3609.555938037866  loss: 0.5093 (0.5164)  acc1: 99.2188 (99.3286)  acc5: 100.0000 (99.9920)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [12]  [400/468]  eta: 0:00:02  lr: 0.0003308693936411421  img/s: 3477.7288403487632  loss: 0.5089 (0.5163)  acc1: 99.2188 (99.3298)  acc5: 100.0000 (99.9922)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [12]  [410/468]  eta: 0:00:02  lr: 0.0003308693936411421  img/s: 3580.810458213833  loss: 0.5094 (0.5162)  acc1: 99.2188 (99.3309)  acc5: 100.0000 (99.9924)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [12]  [420/468]  eta: 0:00:01  lr: 0.0003308693936411421  img/s: 3572.375715312342  loss: 0.5097 (0.5163)  acc1: 99.2188 (99.3338)  acc5: 100.0000 (99.9926)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [12]  [430/468]  eta: 0:00:01  lr: 0.0003308693936411421  img/s: 3535.953632962748  loss: 0.5195 (0.5164)  acc1: 99.2188 (99.3293)  acc5: 100.0000 (99.9927)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [12]  [440/468]  eta: 0:00:01  lr: 0.0003308693936411421  img/s: 3534.1380554275556  loss: 0.5130 (0.5163)  acc1: 99.2188 (99.3339)  acc5: 100.0000 (99.9929)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [12]  [450/468]  eta: 0:00:00  lr: 0.0003308693936411421  img/s: 3457.749359164273  loss: 0.5076 (0.5163)  acc1: 99.2188 (99.3383)  acc5: 100.0000 (99.9931)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [12]  [460/468]  eta: 0:00:00  lr: 0.0003308693936411421  img/s: 3558.1227681826012  loss: 0.5090 (0.5163)  acc1: 99.2188 (99.3374)  acc5: 100.0000 (99.9932)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [12] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5338 (0.5338)  acc1: 99.2188 (99.2188)  acc5: 100.0000 (100.0000)  time: 0.1201  data: 0.1064  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.620 Acc@5 99.970
Epoch: [13]  [  0/468]  eta: 0:01:24  lr: 0.00019098300562505265  img/s: 1682.8015572007998  loss: 0.5060 (0.5060)  acc1: 100.0000 (100.0000)  acc5: 100.0000 (100.0000)  time: 0.1802  data: 0.1041  max mem: 190
Epoch: [13]  [ 10/468]  eta: 0:00:23  lr: 0.00019098300562505265  img/s: 3474.173064478555  loss: 0.5087 (0.5116)  acc1: 100.0000 (99.7159)  acc5: 100.0000 (100.0000)  time: 0.0514  data: 0.0097  max mem: 190
Epoch: [13]  [ 20/468]  eta: 0:00:20  lr: 0.00019098300562505265  img/s: 3444.748299668917  loss: 0.5079 (0.5102)  acc1: 100.0000 (99.7396)  acc5: 100.0000 (100.0000)  time: 0.0382  data: 0.0002  max mem: 190
Epoch: [13]  [ 30/468]  eta: 0:00:18  lr: 0.00019098300562505265  img/s: 3511.117366225001  loss: 0.5056 (0.5099)  acc1: 100.0000 (99.6976)  acc5: 100.0000 (100.0000)  time: 0.0375  data: 0.0002  max mem: 190
Epoch: [13]  [ 40/468]  eta: 0:00:18  lr: 0.00019098300562505265  img/s: 1363.658307488716  loss: 0.5043 (0.5086)  acc1: 100.0000 (99.7523)  acc5: 100.0000 (100.0000)  time: 0.0407  data: 0.0003  max mem: 190
Epoch: [13]  [ 50/468]  eta: 0:00:18  lr: 0.00019098300562505265  img/s: 3069.394789350012  loss: 0.5024 (0.5089)  acc1: 100.0000 (99.7243)  acc5: 100.0000 (100.0000)  time: 0.0470  data: 0.0023  max mem: 190
Epoch: [13]  [ 60/468]  eta: 0:00:17  lr: 0.00019098300562505265  img/s: 3438.2623442163103  loss: 0.5078 (0.5095)  acc1: 99.2188 (99.6926)  acc5: 100.0000 (99.9872)  time: 0.0461  data: 0.0028  max mem: 190
Epoch: [13]  [ 70/468]  eta: 0:00:17  lr: 0.00019098300562505265  img/s: 3492.479358842586  loss: 0.5070 (0.5098)  acc1: 99.2188 (99.6589)  acc5: 100.0000 (99.9890)  time: 0.0406  data: 0.0008  max mem: 190
Epoch: [13]  [ 80/468]  eta: 0:00:16  lr: 0.00019098300562505265  img/s: 3470.579680914333  loss: 0.5058 (0.5097)  acc1: 99.2188 (99.6431)  acc5: 100.0000 (99.9904)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [13]  [ 90/468]  eta: 0:00:15  lr: 0.00019098300562505265  img/s: 3411.7802208975713  loss: 0.5038 (0.5095)  acc1: 100.0000 (99.6394)  acc5: 100.0000 (99.9914)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [13]  [100/468]  eta: 0:00:15  lr: 0.00019098300562505265  img/s: 3476.692863618702  loss: 0.5039 (0.5095)  acc1: 100.0000 (99.6364)  acc5: 100.0000 (99.9923)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [13]  [110/468]  eta: 0:00:14  lr: 0.00019098300562505265  img/s: 3369.5955011046394  loss: 0.5042 (0.5100)  acc1: 99.2188 (99.6199)  acc5: 100.0000 (99.9930)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [13]  [120/468]  eta: 0:00:14  lr: 0.00019098300562505265  img/s: 2763.3305470342384  loss: 0.5049 (0.5098)  acc1: 100.0000 (99.6320)  acc5: 100.0000 (99.9935)  time: 0.0390  data: 0.0002  max mem: 190
Epoch: [13]  [130/468]  eta: 0:00:13  lr: 0.00019098300562505265  img/s: 3448.089042459586  loss: 0.5049 (0.5101)  acc1: 100.0000 (99.6243)  acc5: 100.0000 (99.9940)  time: 0.0398  data: 0.0003  max mem: 190
Epoch: [13]  [140/468]  eta: 0:00:13  lr: 0.00019098300562505265  img/s: 3369.5320559087686  loss: 0.5095 (0.5103)  acc1: 99.2188 (99.6121)  acc5: 100.0000 (99.9945)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [13]  [150/468]  eta: 0:00:12  lr: 0.00019098300562505265  img/s: 3471.432436277109  loss: 0.5060 (0.5102)  acc1: 99.2188 (99.6171)  acc5: 100.0000 (99.9948)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [13]  [160/468]  eta: 0:00:12  lr: 0.00019098300562505265  img/s: 3444.505187247775  loss: 0.5056 (0.5102)  acc1: 100.0000 (99.6167)  acc5: 100.0000 (99.9951)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [13]  [170/468]  eta: 0:00:11  lr: 0.00019098300562505265  img/s: 3307.5865570033575  loss: 0.5027 (0.5099)  acc1: 100.0000 (99.6254)  acc5: 100.0000 (99.9954)  time: 0.0389  data: 0.0003  max mem: 190
Epoch: [13]  [180/468]  eta: 0:00:11  lr: 0.00019098300562505265  img/s: 3425.866161277766  loss: 0.5031 (0.5099)  acc1: 100.0000 (99.6202)  acc5: 100.0000 (99.9957)  time: 0.0384  data: 0.0003  max mem: 190
Epoch: [13]  [190/468]  eta: 0:00:11  lr: 0.00019098300562505265  img/s: 3484.432536978264  loss: 0.5074 (0.5100)  acc1: 99.2188 (99.6073)  acc5: 100.0000 (99.9959)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [13]  [200/468]  eta: 0:00:10  lr: 0.00019098300562505265  img/s: 3348.119189273464  loss: 0.5049 (0.5099)  acc1: 100.0000 (99.6191)  acc5: 100.0000 (99.9961)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [13]  [210/468]  eta: 0:00:10  lr: 0.00019098300562505265  img/s: 3368.7074857250423  loss: 0.5055 (0.5099)  acc1: 100.0000 (99.6149)  acc5: 100.0000 (99.9963)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [13]  [220/468]  eta: 0:00:09  lr: 0.00019098300562505265  img/s: 3569.2644483595386  loss: 0.5026 (0.5098)  acc1: 100.0000 (99.6182)  acc5: 100.0000 (99.9965)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [13]  [230/468]  eta: 0:00:09  lr: 0.00019098300562505265  img/s: 3386.0659337887014  loss: 0.5020 (0.5096)  acc1: 100.0000 (99.6280)  acc5: 100.0000 (99.9966)  time: 0.0376  data: 0.0002  max mem: 190
Epoch: [13]  [240/468]  eta: 0:00:08  lr: 0.00019098300562505265  img/s: 3378.607779588806  loss: 0.5026 (0.5096)  acc1: 100.0000 (99.6304)  acc5: 100.0000 (99.9968)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [13]  [250/468]  eta: 0:00:08  lr: 0.00019098300562505265  img/s: 3548.785468294521  loss: 0.5026 (0.5094)  acc1: 100.0000 (99.6452)  acc5: 100.0000 (99.9969)  time: 0.0404  data: 0.0006  max mem: 190
Epoch: [13]  [260/468]  eta: 0:00:08  lr: 0.00019098300562505265  img/s: 3539.357040201469  loss: 0.5032 (0.5096)  acc1: 100.0000 (99.6288)  acc5: 100.0000 (99.9970)  time: 0.0427  data: 0.0006  max mem: 190
Epoch: [13]  [270/468]  eta: 0:00:07  lr: 0.00019098300562505265  img/s: 3462.633343437796  loss: 0.5123 (0.5099)  acc1: 99.2188 (99.6166)  acc5: 100.0000 (99.9971)  time: 0.0395  data: 0.0003  max mem: 190
Epoch: [13]  [280/468]  eta: 0:00:07  lr: 0.00019098300562505265  img/s: 3555.130432479323  loss: 0.5059 (0.5099)  acc1: 99.2188 (99.6191)  acc5: 100.0000 (99.9972)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [13]  [290/468]  eta: 0:00:07  lr: 0.00019098300562505265  img/s: 3626.9915214733046  loss: 0.5069 (0.5104)  acc1: 99.2188 (99.6000)  acc5: 100.0000 (99.9973)  time: 0.0364  data: 0.0003  max mem: 190
Epoch: [13]  [300/468]  eta: 0:00:06  lr: 0.00019098300562505265  img/s: 3500.6319084009283  loss: 0.5151 (0.5105)  acc1: 99.2188 (99.5977)  acc5: 100.0000 (99.9948)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [13]  [310/468]  eta: 0:00:06  lr: 0.00019098300562505265  img/s: 3429.871409588061  loss: 0.5056 (0.5106)  acc1: 100.0000 (99.5981)  acc5: 100.0000 (99.9950)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [13]  [320/468]  eta: 0:00:05  lr: 0.00019098300562505265  img/s: 3511.392938898845  loss: 0.5083 (0.5109)  acc1: 99.2188 (99.5887)  acc5: 100.0000 (99.9951)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [13]  [330/468]  eta: 0:00:05  lr: 0.00019098300562505265  img/s: 3008.3037492365365  loss: 0.5073 (0.5108)  acc1: 100.0000 (99.5917)  acc5: 100.0000 (99.9953)  time: 0.0458  data: 0.0019  max mem: 190
Epoch: [13]  [340/468]  eta: 0:00:05  lr: 0.00019098300562505265  img/s: 3446.9378567347017  loss: 0.5035 (0.5106)  acc1: 100.0000 (99.6014)  acc5: 100.0000 (99.9954)  time: 0.0487  data: 0.0032  max mem: 190
Epoch: [13]  [350/468]  eta: 0:00:04  lr: 0.00019098300562505265  img/s: 3523.006181507973  loss: 0.5034 (0.5106)  acc1: 100.0000 (99.5971)  acc5: 100.0000 (99.9955)  time: 0.0402  data: 0.0015  max mem: 190
Epoch: [13]  [360/468]  eta: 0:00:04  lr: 0.00019098300562505265  img/s: 3495.3898720002085  loss: 0.5038 (0.5106)  acc1: 100.0000 (99.5996)  acc5: 100.0000 (99.9957)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [13]  [370/468]  eta: 0:00:03  lr: 0.00019098300562505265  img/s: 3456.858794895239  loss: 0.5027 (0.5105)  acc1: 100.0000 (99.6062)  acc5: 100.0000 (99.9958)  time: 0.0376  data: 0.0002  max mem: 190
Epoch: [13]  [380/468]  eta: 0:00:03  lr: 0.00019098300562505265  img/s: 3468.7857752048176  loss: 0.5026 (0.5104)  acc1: 100.0000 (99.6125)  acc5: 100.0000 (99.9959)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [13]  [390/468]  eta: 0:00:03  lr: 0.00019098300562505265  img/s: 3539.4037077081302  loss: 0.5050 (0.5105)  acc1: 100.0000 (99.6084)  acc5: 100.0000 (99.9960)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [13]  [400/468]  eta: 0:00:02  lr: 0.00019098300562505265  img/s: 3395.0592981857044  loss: 0.5056 (0.5105)  acc1: 100.0000 (99.6103)  acc5: 100.0000 (99.9961)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [13]  [410/468]  eta: 0:00:02  lr: 0.00019098300562505265  img/s: 3485.880490607936  loss: 0.5043 (0.5105)  acc1: 100.0000 (99.6084)  acc5: 100.0000 (99.9962)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [13]  [420/468]  eta: 0:00:01  lr: 0.00019098300562505265  img/s: 3440.598000512689  loss: 0.5027 (0.5104)  acc1: 100.0000 (99.6122)  acc5: 100.0000 (99.9963)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [13]  [430/468]  eta: 0:00:01  lr: 0.00019098300562505265  img/s: 3459.130640962862  loss: 0.5031 (0.5104)  acc1: 100.0000 (99.6103)  acc5: 100.0000 (99.9964)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [13]  [440/468]  eta: 0:00:01  lr: 0.00019098300562505265  img/s: 3513.5071006923995  loss: 0.5042 (0.5105)  acc1: 100.0000 (99.6120)  acc5: 100.0000 (99.9947)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [13]  [450/468]  eta: 0:00:00  lr: 0.00019098300562505265  img/s: 3505.61498178209  loss: 0.5024 (0.5103)  acc1: 100.0000 (99.6189)  acc5: 100.0000 (99.9948)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [13]  [460/468]  eta: 0:00:00  lr: 0.00019098300562505265  img/s: 3506.6453648245274  loss: 0.5038 (0.5103)  acc1: 100.0000 (99.6153)  acc5: 100.0000 (99.9949)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [13] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:14  loss: 0.5582 (0.5582)  acc1: 97.6562 (97.6562)  acc5: 100.0000 (100.0000)  time: 0.1795  data: 0.1629  max mem: 190
Test:  Total time: 0:00:02
Test:  Acc@1 98.670 Acc@5 99.910
Epoch: [14]  [  0/468]  eta: 0:01:18  lr: 8.645454235739902e-05  img/s: 2322.0256738521157  loss: 0.5286 (0.5286)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1679  data: 0.1128  max mem: 190
Epoch: [14]  [ 10/468]  eta: 0:00:24  lr: 8.645454235739902e-05  img/s: 3280.4039594280825  loss: 0.5064 (0.5121)  acc1: 100.0000 (99.5739)  acc5: 100.0000 (100.0000)  time: 0.0536  data: 0.0106  max mem: 190
Epoch: [14]  [ 20/468]  eta: 0:00:20  lr: 8.645454235739902e-05  img/s: 3501.63652491521  loss: 0.5060 (0.5105)  acc1: 100.0000 (99.5908)  acc5: 100.0000 (100.0000)  time: 0.0406  data: 0.0003  max mem: 190
Epoch: [14]  [ 30/468]  eta: 0:00:19  lr: 8.645454235739902e-05  img/s: 3416.948268839104  loss: 0.5057 (0.5089)  acc1: 100.0000 (99.6472)  acc5: 100.0000 (100.0000)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [14]  [ 40/468]  eta: 0:00:17  lr: 8.645454235739902e-05  img/s: 3601.3718824208113  loss: 0.5030 (0.5086)  acc1: 100.0000 (99.6951)  acc5: 100.0000 (100.0000)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [14]  [ 50/468]  eta: 0:00:17  lr: 8.645454235739902e-05  img/s: 3491.0940220960706  loss: 0.5035 (0.5095)  acc1: 100.0000 (99.6630)  acc5: 100.0000 (100.0000)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [14]  [ 60/468]  eta: 0:00:16  lr: 8.645454235739902e-05  img/s: 3246.169034863894  loss: 0.5039 (0.5095)  acc1: 100.0000 (99.6670)  acc5: 100.0000 (100.0000)  time: 0.0383  data: 0.0008  max mem: 190
Epoch: [14]  [ 70/468]  eta: 0:00:16  lr: 8.645454235739902e-05  img/s: 3279.902935516388  loss: 0.5030 (0.5087)  acc1: 100.0000 (99.6919)  acc5: 100.0000 (100.0000)  time: 0.0399  data: 0.0008  max mem: 190
Epoch: [14]  [ 80/468]  eta: 0:00:15  lr: 8.645454235739902e-05  img/s: 1911.368477266335  loss: 0.5030 (0.5087)  acc1: 100.0000 (99.7010)  acc5: 100.0000 (99.9904)  time: 0.0412  data: 0.0003  max mem: 190
Epoch: [14]  [ 90/468]  eta: 0:00:16  lr: 8.645454235739902e-05  img/s: 2397.2909546369933  loss: 0.5029 (0.5091)  acc1: 100.0000 (99.6995)  acc5: 100.0000 (99.9914)  time: 0.0512  data: 0.0023  max mem: 190
Epoch: [14]  [100/468]  eta: 0:00:15  lr: 8.645454235739902e-05  img/s: 3118.043186860415  loss: 0.5028 (0.5085)  acc1: 100.0000 (99.7215)  acc5: 100.0000 (99.9923)  time: 0.0517  data: 0.0030  max mem: 190
Epoch: [14]  [110/468]  eta: 0:00:15  lr: 8.645454235739902e-05  img/s: 3220.793885619327  loss: 0.5028 (0.5086)  acc1: 100.0000 (99.7185)  acc5: 100.0000 (99.9930)  time: 0.0421  data: 0.0010  max mem: 190
Epoch: [14]  [120/468]  eta: 0:00:14  lr: 8.645454235739902e-05  img/s: 3145.684992587992  loss: 0.5033 (0.5085)  acc1: 100.0000 (99.7224)  acc5: 100.0000 (99.9935)  time: 0.0408  data: 0.0003  max mem: 190
Epoch: [14]  [130/468]  eta: 0:00:14  lr: 8.645454235739902e-05  img/s: 3213.7759393725346  loss: 0.5029 (0.5083)  acc1: 100.0000 (99.7197)  acc5: 100.0000 (99.9940)  time: 0.0420  data: 0.0003  max mem: 190
Epoch: [14]  [140/468]  eta: 0:00:13  lr: 8.645454235739902e-05  img/s: 2861.4496807410646  loss: 0.5028 (0.5083)  acc1: 100.0000 (99.7174)  acc5: 100.0000 (99.9945)  time: 0.0417  data: 0.0003  max mem: 190
Epoch: [14]  [150/468]  eta: 0:00:13  lr: 8.645454235739902e-05  img/s: 2756.477801680983  loss: 0.5024 (0.5081)  acc1: 100.0000 (99.7258)  acc5: 100.0000 (99.9948)  time: 0.0400  data: 0.0003  max mem: 190
Epoch: [14]  [160/468]  eta: 0:00:12  lr: 8.645454235739902e-05  img/s: 3562.7271170806484  loss: 0.5023 (0.5078)  acc1: 100.0000 (99.7428)  acc5: 100.0000 (99.9951)  time: 0.0386  data: 0.0003  max mem: 190
Epoch: [14]  [170/468]  eta: 0:00:12  lr: 8.645454235739902e-05  img/s: 3452.013271263599  loss: 0.5018 (0.5078)  acc1: 100.0000 (99.7350)  acc5: 100.0000 (99.9954)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [14]  [180/468]  eta: 0:00:11  lr: 8.645454235739902e-05  img/s: 3525.967818628417  loss: 0.5033 (0.5079)  acc1: 100.0000 (99.7194)  acc5: 100.0000 (99.9957)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [14]  [190/468]  eta: 0:00:11  lr: 8.645454235739902e-05  img/s: 3569.5254913433155  loss: 0.5049 (0.5080)  acc1: 100.0000 (99.7137)  acc5: 100.0000 (99.9959)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [14]  [200/468]  eta: 0:00:10  lr: 8.645454235739902e-05  img/s: 3462.164418190729  loss: 0.5022 (0.5078)  acc1: 100.0000 (99.7240)  acc5: 100.0000 (99.9961)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [14]  [210/468]  eta: 0:00:10  lr: 8.645454235739902e-05  img/s: 3587.1746846268975  loss: 0.5022 (0.5081)  acc1: 100.0000 (99.7186)  acc5: 100.0000 (99.9963)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [14]  [220/468]  eta: 0:00:10  lr: 8.645454235739902e-05  img/s: 3513.277177185038  loss: 0.5015 (0.5081)  acc1: 100.0000 (99.7207)  acc5: 100.0000 (99.9965)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [14]  [230/468]  eta: 0:00:09  lr: 8.645454235739902e-05  img/s: 3025.5794053335135  loss: 0.5026 (0.5081)  acc1: 100.0000 (99.7193)  acc5: 100.0000 (99.9966)  time: 0.0430  data: 0.0003  max mem: 190
Epoch: [14]  [240/468]  eta: 0:00:09  lr: 8.645454235739902e-05  img/s: 3155.651043319814  loss: 0.5033 (0.5079)  acc1: 100.0000 (99.7309)  acc5: 100.0000 (99.9968)  time: 0.0479  data: 0.0007  max mem: 190
Epoch: [14]  [250/468]  eta: 0:00:08  lr: 8.645454235739902e-05  img/s: 3337.5248633896767  loss: 0.5024 (0.5077)  acc1: 100.0000 (99.7385)  acc5: 100.0000 (99.9969)  time: 0.0429  data: 0.0007  max mem: 190
Epoch: [14]  [260/468]  eta: 0:00:08  lr: 8.645454235739902e-05  img/s: 3527.5432145813897  loss: 0.5022 (0.5075)  acc1: 100.0000 (99.7426)  acc5: 100.0000 (99.9970)  time: 0.0387  data: 0.0003  max mem: 190
Epoch: [14]  [270/468]  eta: 0:00:08  lr: 8.645454235739902e-05  img/s: 3607.8579627165573  loss: 0.5031 (0.5078)  acc1: 100.0000 (99.7348)  acc5: 100.0000 (99.9971)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [14]  [280/468]  eta: 0:00:07  lr: 8.645454235739902e-05  img/s: 3467.2176282920655  loss: 0.5059 (0.5078)  acc1: 100.0000 (99.7331)  acc5: 100.0000 (99.9972)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [14]  [290/468]  eta: 0:00:07  lr: 8.645454235739902e-05  img/s: 3557.085483336646  loss: 0.5026 (0.5077)  acc1: 100.0000 (99.7396)  acc5: 100.0000 (99.9973)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [14]  [300/468]  eta: 0:00:06  lr: 8.645454235739902e-05  img/s: 3545.7619739518664  loss: 0.5026 (0.5081)  acc1: 100.0000 (99.7275)  acc5: 100.0000 (99.9974)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [14]  [310/468]  eta: 0:00:06  lr: 8.645454235739902e-05  img/s: 3492.433920532903  loss: 0.5024 (0.5079)  acc1: 100.0000 (99.7337)  acc5: 100.0000 (99.9975)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [14]  [320/468]  eta: 0:00:05  lr: 8.645454235739902e-05  img/s: 3540.3840097070733  loss: 0.5016 (0.5078)  acc1: 100.0000 (99.7396)  acc5: 100.0000 (99.9976)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [14]  [330/468]  eta: 0:00:05  lr: 8.645454235739902e-05  img/s: 3359.495591557316  loss: 0.5020 (0.5077)  acc1: 100.0000 (99.7404)  acc5: 100.0000 (99.9976)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [14]  [340/468]  eta: 0:00:05  lr: 8.645454235739902e-05  img/s: 3507.4701074706823  loss: 0.5033 (0.5076)  acc1: 100.0000 (99.7411)  acc5: 100.0000 (99.9977)  time: 0.0376  data: 0.0002  max mem: 190
Epoch: [14]  [350/468]  eta: 0:00:04  lr: 8.645454235739902e-05  img/s: 3505.5920912587253  loss: 0.5036 (0.5078)  acc1: 100.0000 (99.7396)  acc5: 100.0000 (99.9978)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [14]  [360/468]  eta: 0:00:04  lr: 8.645454235739902e-05  img/s: 3241.112941006013  loss: 0.5040 (0.5078)  acc1: 100.0000 (99.7381)  acc5: 100.0000 (99.9978)  time: 0.0428  data: 0.0006  max mem: 190
Epoch: [14]  [370/468]  eta: 0:00:03  lr: 8.645454235739902e-05  img/s: 3499.468187595737  loss: 0.5020 (0.5078)  acc1: 100.0000 (99.7410)  acc5: 100.0000 (99.9979)  time: 0.0486  data: 0.0030  max mem: 190
Epoch: [14]  [380/468]  eta: 0:00:03  lr: 8.645454235739902e-05  img/s: 3273.283777192469  loss: 0.5021 (0.5079)  acc1: 100.0000 (99.7396)  acc5: 100.0000 (99.9979)  time: 0.0429  data: 0.0026  max mem: 190
Epoch: [14]  [390/468]  eta: 0:00:03  lr: 8.645454235739902e-05  img/s: 3569.5254913433155  loss: 0.5042 (0.5079)  acc1: 100.0000 (99.7383)  acc5: 100.0000 (99.9980)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [14]  [400/468]  eta: 0:00:02  lr: 8.645454235739902e-05  img/s: 3462.8343502883163  loss: 0.5022 (0.5080)  acc1: 100.0000 (99.7350)  acc5: 100.0000 (99.9981)  time: 0.0376  data: 0.0002  max mem: 190
Epoch: [14]  [410/468]  eta: 0:00:02  lr: 8.645454235739902e-05  img/s: 3365.64531235307  loss: 0.5032 (0.5081)  acc1: 100.0000 (99.7320)  acc5: 100.0000 (99.9962)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [14]  [420/468]  eta: 0:00:01  lr: 8.645454235739902e-05  img/s: 3430.1343760382324  loss: 0.5045 (0.5080)  acc1: 100.0000 (99.7328)  acc5: 100.0000 (99.9963)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [14]  [430/468]  eta: 0:00:01  lr: 8.645454235739902e-05  img/s: 3436.6776683864855  loss: 0.5019 (0.5079)  acc1: 100.0000 (99.7372)  acc5: 100.0000 (99.9964)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [14]  [440/468]  eta: 0:00:01  lr: 8.645454235739902e-05  img/s: 3557.2740355945457  loss: 0.5018 (0.5077)  acc1: 100.0000 (99.7431)  acc5: 100.0000 (99.9965)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [14]  [450/468]  eta: 0:00:00  lr: 8.645454235739902e-05  img/s: 3501.362481412882  loss: 0.5031 (0.5078)  acc1: 100.0000 (99.7384)  acc5: 100.0000 (99.9965)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [14]  [460/468]  eta: 0:00:00  lr: 8.645454235739902e-05  img/s: 3422.76470325719  loss: 0.5031 (0.5077)  acc1: 100.0000 (99.7407)  acc5: 100.0000 (99.9966)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [14] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5510 (0.5510)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1211  data: 0.1021  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.750 Acc@5 99.920
Epoch: [15]  [  0/468]  eta: 0:01:14  lr: 2.1852399266194312e-05  img/s: 2757.893798127058  loss: 0.5055 (0.5055)  acc1: 100.0000 (100.0000)  acc5: 100.0000 (100.0000)  time: 0.1598  data: 0.1133  max mem: 190
Epoch: [15]  [ 10/468]  eta: 0:00:22  lr: 2.1852399266194312e-05  img/s: 3457.9943447876076  loss: 0.5040 (0.5057)  acc1: 100.0000 (99.7869)  acc5: 100.0000 (100.0000)  time: 0.0488  data: 0.0105  max mem: 190
Epoch: [15]  [ 20/468]  eta: 0:00:19  lr: 2.1852399266194312e-05  img/s: 3491.298346924708  loss: 0.5027 (0.5084)  acc1: 100.0000 (99.7396)  acc5: 100.0000 (100.0000)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [15]  [ 30/468]  eta: 0:00:18  lr: 2.1852399266194312e-05  img/s: 3435.9738368  loss: 0.5024 (0.5065)  acc1: 100.0000 (99.7984)  acc5: 100.0000 (100.0000)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [15]  [ 40/468]  eta: 0:00:17  lr: 2.1852399266194312e-05  img/s: 3355.107689231077  loss: 0.5020 (0.5062)  acc1: 100.0000 (99.8285)  acc5: 100.0000 (100.0000)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [15]  [ 50/468]  eta: 0:00:16  lr: 2.1852399266194312e-05  img/s: 3455.2569347011804  loss: 0.5019 (0.5060)  acc1: 100.0000 (99.8468)  acc5: 100.0000 (100.0000)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [15]  [ 60/468]  eta: 0:00:16  lr: 2.1852399266194312e-05  img/s: 3422.2628827864046  loss: 0.5019 (0.5068)  acc1: 100.0000 (99.8207)  acc5: 100.0000 (100.0000)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [15]  [ 70/468]  eta: 0:00:15  lr: 2.1852399266194312e-05  img/s: 3367.4185823334233  loss: 0.5016 (0.5069)  acc1: 100.0000 (99.8239)  acc5: 100.0000 (100.0000)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [15]  [ 80/468]  eta: 0:00:15  lr: 2.1852399266194312e-05  img/s: 3468.8081875803605  loss: 0.5019 (0.5069)  acc1: 100.0000 (99.8264)  acc5: 100.0000 (99.9904)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [15]  [ 90/468]  eta: 0:00:14  lr: 2.1852399266194312e-05  img/s: 3418.9501999643376  loss: 0.5028 (0.5071)  acc1: 100.0000 (99.8025)  acc5: 100.0000 (99.9914)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [15]  [100/468]  eta: 0:00:14  lr: 2.1852399266194312e-05  img/s: 3447.491215452584  loss: 0.5026 (0.5069)  acc1: 100.0000 (99.8066)  acc5: 100.0000 (99.9923)  time: 0.0384  data: 0.0003  max mem: 190
Epoch: [15]  [110/468]  eta: 0:00:13  lr: 2.1852399266194312e-05  img/s: 3479.8252020663595  loss: 0.5016 (0.5068)  acc1: 100.0000 (99.8100)  acc5: 100.0000 (99.9930)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [15]  [120/468]  eta: 0:00:13  lr: 2.1852399266194312e-05  img/s: 3482.601694365521  loss: 0.5021 (0.5070)  acc1: 100.0000 (99.7998)  acc5: 100.0000 (99.9935)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [15]  [130/468]  eta: 0:00:13  lr: 2.1852399266194312e-05  img/s: 1422.856818766084  loss: 0.5022 (0.5067)  acc1: 100.0000 (99.8092)  acc5: 100.0000 (99.9940)  time: 0.0414  data: 0.0003  max mem: 190
Epoch: [15]  [140/468]  eta: 0:00:13  lr: 2.1852399266194312e-05  img/s: 2415.8780340734206  loss: 0.5016 (0.5069)  acc1: 100.0000 (99.8005)  acc5: 100.0000 (99.9945)  time: 0.0490  data: 0.0027  max mem: 190
Epoch: [15]  [150/468]  eta: 0:00:12  lr: 2.1852399266194312e-05  img/s: 3300.9161906752825  loss: 0.5014 (0.5066)  acc1: 100.0000 (99.8086)  acc5: 100.0000 (99.9948)  time: 0.0458  data: 0.0027  max mem: 190
Epoch: [15]  [160/468]  eta: 0:00:12  lr: 2.1852399266194312e-05  img/s: 3474.397898033937  loss: 0.5013 (0.5064)  acc1: 100.0000 (99.8156)  acc5: 100.0000 (99.9951)  time: 0.0396  data: 0.0003  max mem: 190
Epoch: [15]  [170/468]  eta: 0:00:11  lr: 2.1852399266194312e-05  img/s: 3440.487756736839  loss: 0.5015 (0.5061)  acc1: 100.0000 (99.8264)  acc5: 100.0000 (99.9954)  time: 0.0391  data: 0.0003  max mem: 190
Epoch: [15]  [180/468]  eta: 0:00:11  lr: 2.1852399266194312e-05  img/s: 3407.2763921150504  loss: 0.5015 (0.5062)  acc1: 100.0000 (99.8273)  acc5: 100.0000 (99.9957)  time: 0.0378  data: 0.0002  max mem: 190
Epoch: [15]  [190/468]  eta: 0:00:11  lr: 2.1852399266194312e-05  img/s: 3543.2814055095764  loss: 0.5026 (0.5062)  acc1: 100.0000 (99.8200)  acc5: 100.0000 (99.9959)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [15]  [200/468]  eta: 0:00:10  lr: 2.1852399266194312e-05  img/s: 3530.1642677257514  loss: 0.5032 (0.5064)  acc1: 100.0000 (99.8134)  acc5: 100.0000 (99.9961)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [15]  [210/468]  eta: 0:00:10  lr: 2.1852399266194312e-05  img/s: 3561.9234499917065  loss: 0.5021 (0.5062)  acc1: 100.0000 (99.8223)  acc5: 100.0000 (99.9963)  time: 0.0372  data: 0.0005  max mem: 190
Epoch: [15]  [220/468]  eta: 0:00:09  lr: 2.1852399266194312e-05  img/s: 3406.4332476761524  loss: 0.5014 (0.5061)  acc1: 100.0000 (99.8268)  acc5: 100.0000 (99.9965)  time: 0.0373  data: 0.0005  max mem: 190
Epoch: [15]  [230/468]  eta: 0:00:09  lr: 2.1852399266194312e-05  img/s: 2149.091168193809  loss: 0.5022 (0.5064)  acc1: 100.0000 (99.8174)  acc5: 100.0000 (99.9966)  time: 0.0398  data: 0.0003  max mem: 190
Epoch: [15]  [240/468]  eta: 0:00:09  lr: 2.1852399266194312e-05  img/s: 1784.0806850921997  loss: 0.5031 (0.5063)  acc1: 100.0000 (99.8185)  acc5: 100.0000 (99.9968)  time: 0.0456  data: 0.0003  max mem: 190
Epoch: [15]  [250/468]  eta: 0:00:08  lr: 2.1852399266194312e-05  img/s: 2532.445798761303  loss: 0.5021 (0.5064)  acc1: 100.0000 (99.8164)  acc5: 100.0000 (99.9969)  time: 0.0536  data: 0.0026  max mem: 190
Epoch: [15]  [260/468]  eta: 0:00:08  lr: 2.1852399266194312e-05  img/s: 2783.5606619934465  loss: 0.5015 (0.5062)  acc1: 100.0000 (99.8234)  acc5: 100.0000 (99.9970)  time: 0.0567  data: 0.0063  max mem: 190
Epoch: [15]  [270/468]  eta: 0:00:08  lr: 2.1852399266194312e-05  img/s: 3299.638071122147  loss: 0.5016 (0.5064)  acc1: 100.0000 (99.8155)  acc5: 100.0000 (99.9971)  time: 0.0464  data: 0.0040  max mem: 190
Epoch: [15]  [280/468]  eta: 0:00:07  lr: 2.1852399266194312e-05  img/s: 3415.40490231629  loss: 0.5020 (0.5065)  acc1: 100.0000 (99.8165)  acc5: 100.0000 (99.9972)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [15]  [290/468]  eta: 0:00:07  lr: 2.1852399266194312e-05  img/s: 3205.0463977839863  loss: 0.5019 (0.5065)  acc1: 100.0000 (99.8148)  acc5: 100.0000 (99.9973)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [15]  [300/468]  eta: 0:00:06  lr: 2.1852399266194312e-05  img/s: 3537.0485357578154  loss: 0.5014 (0.5064)  acc1: 100.0000 (99.8183)  acc5: 100.0000 (99.9974)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [15]  [310/468]  eta: 0:00:06  lr: 2.1852399266194312e-05  img/s: 3132.0132077893286  loss: 0.5016 (0.5064)  acc1: 100.0000 (99.8216)  acc5: 100.0000 (99.9975)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [15]  [320/468]  eta: 0:00:06  lr: 2.1852399266194312e-05  img/s: 3513.254186489369  loss: 0.5017 (0.5063)  acc1: 100.0000 (99.8223)  acc5: 100.0000 (99.9976)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [15]  [330/468]  eta: 0:00:05  lr: 2.1852399266194312e-05  img/s: 3602.870319168926  loss: 0.5017 (0.5062)  acc1: 100.0000 (99.8253)  acc5: 100.0000 (99.9976)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [15]  [340/468]  eta: 0:00:05  lr: 2.1852399266194312e-05  img/s: 3537.328194079316  loss: 0.5017 (0.5061)  acc1: 100.0000 (99.8305)  acc5: 100.0000 (99.9977)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [15]  [350/468]  eta: 0:00:04  lr: 2.1852399266194312e-05  img/s: 3544.029890550942  loss: 0.5014 (0.5060)  acc1: 100.0000 (99.8331)  acc5: 100.0000 (99.9978)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [15]  [360/468]  eta: 0:00:04  lr: 2.1852399266194312e-05  img/s: 3440.9067206747595  loss: 0.5013 (0.5060)  acc1: 100.0000 (99.8312)  acc5: 100.0000 (99.9978)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [15]  [370/468]  eta: 0:00:03  lr: 2.1852399266194312e-05  img/s: 3565.566490227202  loss: 0.5013 (0.5059)  acc1: 100.0000 (99.8315)  acc5: 100.0000 (99.9979)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [15]  [380/468]  eta: 0:00:03  lr: 2.1852399266194312e-05  img/s: 3579.234859596256  loss: 0.5012 (0.5059)  acc1: 100.0000 (99.8339)  acc5: 100.0000 (99.9979)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [15]  [390/468]  eta: 0:00:03  lr: 2.1852399266194312e-05  img/s: 3276.220102642965  loss: 0.5025 (0.5059)  acc1: 100.0000 (99.8322)  acc5: 100.0000 (99.9980)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [15]  [400/468]  eta: 0:00:02  lr: 2.1852399266194312e-05  img/s: 2341.0613270076046  loss: 0.5033 (0.5059)  acc1: 100.0000 (99.8344)  acc5: 100.0000 (99.9981)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [15]  [410/468]  eta: 0:00:02  lr: 2.1852399266194312e-05  img/s: 2289.818783587819  loss: 0.5022 (0.5059)  acc1: 100.0000 (99.8327)  acc5: 100.0000 (99.9981)  time: 0.0504  data: 0.0026  max mem: 190
Epoch: [15]  [420/468]  eta: 0:00:01  lr: 2.1852399266194312e-05  img/s: 3458.3730272227162  loss: 0.5017 (0.5060)  acc1: 100.0000 (99.8311)  acc5: 100.0000 (99.9981)  time: 0.0494  data: 0.0026  max mem: 190
Epoch: [15]  [430/468]  eta: 0:00:01  lr: 2.1852399266194312e-05  img/s: 3442.3407903257867  loss: 0.5016 (0.5060)  acc1: 100.0000 (99.8296)  acc5: 100.0000 (99.9982)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [15]  [440/468]  eta: 0:00:01  lr: 2.1852399266194312e-05  img/s: 3324.9780882662603  loss: 0.5012 (0.5059)  acc1: 100.0000 (99.8335)  acc5: 100.0000 (99.9982)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [15]  [450/468]  eta: 0:00:00  lr: 2.1852399266194312e-05  img/s: 3340.619202289839  loss: 0.5010 (0.5058)  acc1: 100.0000 (99.8354)  acc5: 100.0000 (99.9983)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [15]  [460/468]  eta: 0:00:00  lr: 2.1852399266194312e-05  img/s: 3465.6285269796595  loss: 0.5015 (0.5059)  acc1: 100.0000 (99.8339)  acc5: 100.0000 (99.9983)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [15] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5526 (0.5526)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1216  data: 0.1018  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.760 Acc@5 99.900
Epoch: [16]  [  0/468]  eta: 0:01:19  lr: 0.0  img/s: 1982.2950379015851  loss: 0.5016 (0.5016)  acc1: 100.0000 (100.0000)  acc5: 100.0000 (100.0000)  time: 0.1703  data: 0.1057  max mem: 190
Epoch: [16]  [ 10/468]  eta: 0:00:23  lr: 0.0  img/s: 3292.55539201256  loss: 0.5020 (0.5069)  acc1: 100.0000 (99.8580)  acc5: 100.0000 (100.0000)  time: 0.0522  data: 0.0099  max mem: 190
Epoch: [16]  [ 20/468]  eta: 0:00:20  lr: 0.0  img/s: 3561.07290346973  loss: 0.5019 (0.5053)  acc1: 100.0000 (99.8884)  acc5: 100.0000 (100.0000)  time: 0.0401  data: 0.0003  max mem: 190
Epoch: [16]  [ 30/468]  eta: 0:00:18  lr: 0.0  img/s: 3545.9024873518883  loss: 0.5019 (0.5074)  acc1: 100.0000 (99.7984)  acc5: 100.0000 (100.0000)  time: 0.0384  data: 0.0003  max mem: 190
Epoch: [16]  [ 40/468]  eta: 0:00:17  lr: 0.0  img/s: 3467.934319488405  loss: 0.5022 (0.5069)  acc1: 100.0000 (99.8095)  acc5: 100.0000 (100.0000)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [16]  [ 50/468]  eta: 0:00:17  lr: 0.0  img/s: 3523.9311585165738  loss: 0.5020 (0.5070)  acc1: 100.0000 (99.8162)  acc5: 100.0000 (100.0000)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [16]  [ 60/468]  eta: 0:00:16  lr: 0.0  img/s: 3601.251095056983  loss: 0.5017 (0.5061)  acc1: 100.0000 (99.8463)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [16]  [ 70/468]  eta: 0:00:15  lr: 0.0  img/s: 3460.6022508991996  loss: 0.5014 (0.5060)  acc1: 100.0000 (99.8460)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [16]  [ 80/468]  eta: 0:00:15  lr: 0.0  img/s: 3590.8215527850607  loss: 0.5012 (0.5058)  acc1: 100.0000 (99.8553)  acc5: 100.0000 (100.0000)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [16]  [ 90/468]  eta: 0:00:14  lr: 0.0  img/s: 3534.2311166115887  loss: 0.5014 (0.5057)  acc1: 100.0000 (99.8541)  acc5: 100.0000 (100.0000)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [16]  [100/468]  eta: 0:00:14  lr: 0.0  img/s: 3035.878988023207  loss: 0.5015 (0.5057)  acc1: 100.0000 (99.8608)  acc5: 100.0000 (100.0000)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [16]  [110/468]  eta: 0:00:13  lr: 0.0  img/s: 3506.8744210959494  loss: 0.5018 (0.5055)  acc1: 100.0000 (99.8663)  acc5: 100.0000 (100.0000)  time: 0.0386  data: 0.0003  max mem: 190
Epoch: [16]  [120/468]  eta: 0:00:13  lr: 0.0  img/s: 3458.4398621444907  loss: 0.5019 (0.5057)  acc1: 100.0000 (99.8644)  acc5: 100.0000 (100.0000)  time: 0.0383  data: 0.0002  max mem: 190
Epoch: [16]  [130/468]  eta: 0:00:13  lr: 0.0  img/s: 3514.818239549576  loss: 0.5016 (0.5056)  acc1: 100.0000 (99.8688)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [16]  [140/468]  eta: 0:00:12  lr: 0.0  img/s: 3442.8043606515325  loss: 0.5016 (0.5056)  acc1: 100.0000 (99.8670)  acc5: 100.0000 (100.0000)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [16]  [150/468]  eta: 0:00:12  lr: 0.0  img/s: 3508.180613458447  loss: 0.5014 (0.5055)  acc1: 100.0000 (99.8707)  acc5: 100.0000 (100.0000)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [16]  [160/468]  eta: 0:00:11  lr: 0.0  img/s: 3484.9753786034676  loss: 0.5014 (0.5057)  acc1: 100.0000 (99.8593)  acc5: 100.0000 (100.0000)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [16]  [170/468]  eta: 0:00:11  lr: 0.0  img/s: 1982.7416128698683  loss: 0.5026 (0.5058)  acc1: 100.0000 (99.8538)  acc5: 100.0000 (100.0000)  time: 0.0392  data: 0.0003  max mem: 190
Epoch: [16]  [180/468]  eta: 0:00:11  lr: 0.0  img/s: 3118.242398545632  loss: 0.5021 (0.5058)  acc1: 100.0000 (99.8532)  acc5: 100.0000 (100.0000)  time: 0.0478  data: 0.0035  max mem: 190
Epoch: [16]  [190/468]  eta: 0:00:11  lr: 0.0  img/s: 3482.8276202091497  loss: 0.5016 (0.5056)  acc1: 100.0000 (99.8568)  acc5: 100.0000 (100.0000)  time: 0.0482  data: 0.0044  max mem: 190
Epoch: [16]  [200/468]  eta: 0:00:10  lr: 0.0  img/s: 3515.3475727138198  loss: 0.5014 (0.5056)  acc1: 100.0000 (99.8601)  acc5: 100.0000 (100.0000)  time: 0.0395  data: 0.0011  max mem: 190
Epoch: [16]  [210/468]  eta: 0:00:10  lr: 0.0  img/s: 3514.680178853166  loss: 0.5012 (0.5054)  acc1: 100.0000 (99.8667)  acc5: 100.0000 (100.0000)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [16]  [220/468]  eta: 0:00:09  lr: 0.0  img/s: 3472.6674299316296  loss: 0.5012 (0.5054)  acc1: 100.0000 (99.8657)  acc5: 100.0000 (100.0000)  time: 0.0402  data: 0.0008  max mem: 190
Epoch: [16]  [230/468]  eta: 0:00:09  lr: 0.0  img/s: 3407.752196211852  loss: 0.5012 (0.5054)  acc1: 100.0000 (99.8647)  acc5: 100.0000 (100.0000)  time: 0.0424  data: 0.0012  max mem: 190
Epoch: [16]  [240/468]  eta: 0:00:09  lr: 0.0  img/s: 3496.163792654337  loss: 0.5015 (0.5055)  acc1: 100.0000 (99.8606)  acc5: 100.0000 (100.0000)  time: 0.0393  data: 0.0007  max mem: 190
Epoch: [16]  [250/468]  eta: 0:00:08  lr: 0.0  img/s: 3485.065868652182  loss: 0.5018 (0.5056)  acc1: 100.0000 (99.8568)  acc5: 100.0000 (100.0000)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [16]  [260/468]  eta: 0:00:08  lr: 0.0  img/s: 3557.4861807796547  loss: 0.5018 (0.5056)  acc1: 100.0000 (99.8593)  acc5: 100.0000 (100.0000)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [16]  [270/468]  eta: 0:00:07  lr: 0.0  img/s: 3434.3473299045572  loss: 0.5021 (0.5055)  acc1: 100.0000 (99.8616)  acc5: 100.0000 (100.0000)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [16]  [280/468]  eta: 0:00:07  lr: 0.0  img/s: 3463.7950385496306  loss: 0.5014 (0.5054)  acc1: 100.0000 (99.8665)  acc5: 100.0000 (100.0000)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [16]  [290/468]  eta: 0:00:06  lr: 0.0  img/s: 3488.0546788204033  loss: 0.5012 (0.5055)  acc1: 100.0000 (99.8631)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [16]  [300/468]  eta: 0:00:06  lr: 0.0  img/s: 3538.9370880134998  loss: 0.5015 (0.5054)  acc1: 100.0000 (99.8650)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [16]  [310/468]  eta: 0:00:06  lr: 0.0  img/s: 3528.9344393758133  loss: 0.5015 (0.5053)  acc1: 100.0000 (99.8669)  acc5: 100.0000 (100.0000)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [16]  [320/468]  eta: 0:00:05  lr: 0.0  img/s: 3418.710715172664  loss: 0.5019 (0.5056)  acc1: 100.0000 (99.8588)  acc5: 100.0000 (100.0000)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [16]  [330/468]  eta: 0:00:05  lr: 0.0  img/s: 3593.008425857142  loss: 0.5018 (0.5055)  acc1: 100.0000 (99.8631)  acc5: 100.0000 (100.0000)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [16]  [340/468]  eta: 0:00:04  lr: 0.0  img/s: 3582.7460443513137  loss: 0.5013 (0.5054)  acc1: 100.0000 (99.8648)  acc5: 100.0000 (100.0000)  time: 0.0365  data: 0.0003  max mem: 190
Epoch: [16]  [350/468]  eta: 0:00:04  lr: 0.0  img/s: 3636.6961917277445  loss: 0.5014 (0.5054)  acc1: 100.0000 (99.8665)  acc5: 100.0000 (100.0000)  time: 0.0363  data: 0.0003  max mem: 190
Epoch: [16]  [360/468]  eta: 0:00:04  lr: 0.0  img/s: 3500.449312781994  loss: 0.5021 (0.5054)  acc1: 100.0000 (99.8658)  acc5: 100.0000 (100.0000)  time: 0.0364  data: 0.0002  max mem: 190
Epoch: [16]  [370/468]  eta: 0:00:03  lr: 0.0  img/s: 3570.3563367449406  loss: 0.5021 (0.5054)  acc1: 100.0000 (99.8673)  acc5: 100.0000 (100.0000)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [16]  [380/468]  eta: 0:00:03  lr: 0.0  img/s: 3518.6422246836064  loss: 0.5018 (0.5054)  acc1: 100.0000 (99.8667)  acc5: 100.0000 (100.0000)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [16]  [390/468]  eta: 0:00:03  lr: 0.0  img/s: 3430.5946643662737  loss: 0.5021 (0.5054)  acc1: 100.0000 (99.8661)  acc5: 100.0000 (100.0000)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [16]  [400/468]  eta: 0:00:02  lr: 0.0  img/s: 3495.4126294818125  loss: 0.5019 (0.5054)  acc1: 100.0000 (99.8656)  acc5: 100.0000 (99.9981)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [16]  [410/468]  eta: 0:00:02  lr: 0.0  img/s: 3535.2085550229153  loss: 0.5014 (0.5054)  acc1: 100.0000 (99.8650)  acc5: 100.0000 (99.9981)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [16]  [420/468]  eta: 0:00:01  lr: 0.0  img/s: 3448.155480481445  loss: 0.5017 (0.5054)  acc1: 100.0000 (99.8645)  acc5: 100.0000 (99.9981)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [16]  [430/468]  eta: 0:00:01  lr: 0.0  img/s: 3298.2798867133984  loss: 0.5025 (0.5054)  acc1: 100.0000 (99.8659)  acc5: 100.0000 (99.9982)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [16]  [440/468]  eta: 0:00:01  lr: 0.0  img/s: 3572.4945734267594  loss: 0.5017 (0.5054)  acc1: 100.0000 (99.8671)  acc5: 100.0000 (99.9982)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [16]  [450/468]  eta: 0:00:00  lr: 0.0  img/s: 3348.2653561427687  loss: 0.5011 (0.5054)  acc1: 100.0000 (99.8666)  acc5: 100.0000 (99.9983)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [16]  [460/468]  eta: 0:00:00  lr: 0.0  img/s: 2339.847162961381  loss: 0.5018 (0.5054)  acc1: 100.0000 (99.8661)  acc5: 100.0000 (99.9983)  time: 0.0439  data: 0.0006  max mem: 190
Epoch: [16] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:12  loss: 0.5526 (0.5526)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1549  data: 0.1313  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.760 Acc@5 99.900
Epoch: [17]  [  0/468]  eta: 0:01:14  lr: 0.0  img/s: 2452.9998766351555  loss: 0.5031 (0.5031)  acc1: 100.0000 (100.0000)  acc5: 100.0000 (100.0000)  time: 0.1589  data: 0.1066  max mem: 190
Epoch: [17]  [ 10/468]  eta: 0:00:22  lr: 0.0  img/s: 3470.1086010871745  loss: 0.5013 (0.5042)  acc1: 100.0000 (99.9290)  acc5: 100.0000 (100.0000)  time: 0.0490  data: 0.0100  max mem: 190
Epoch: [17]  [ 20/468]  eta: 0:00:19  lr: 0.0  img/s: 2451.6332716852753  loss: 0.5011 (0.5029)  acc1: 100.0000 (99.9628)  acc5: 100.0000 (100.0000)  time: 0.0386  data: 0.0003  max mem: 190
Epoch: [17]  [ 30/468]  eta: 0:00:21  lr: 0.0  img/s: 1973.3548187899728  loss: 0.5012 (0.5050)  acc1: 100.0000 (99.8992)  acc5: 100.0000 (99.9748)  time: 0.0498  data: 0.0035  max mem: 190
Epoch: [17]  [ 40/468]  eta: 0:00:22  lr: 0.0  img/s: 2807.2709171054626  loss: 0.5012 (0.5047)  acc1: 100.0000 (99.9047)  acc5: 100.0000 (99.9809)  time: 0.0589  data: 0.0053  max mem: 190
Epoch: [17]  [ 50/468]  eta: 0:00:21  lr: 0.0  img/s: 3277.0000122077763  loss: 0.5015 (0.5047)  acc1: 100.0000 (99.9081)  acc5: 100.0000 (99.9847)  time: 0.0519  data: 0.0032  max mem: 190
Epoch: [17]  [ 60/468]  eta: 0:00:19  lr: 0.0  img/s: 3144.763687697327  loss: 0.5018 (0.5050)  acc1: 100.0000 (99.8847)  acc5: 100.0000 (99.9872)  time: 0.0432  data: 0.0014  max mem: 190
Epoch: [17]  [ 70/468]  eta: 0:00:18  lr: 0.0  img/s: 3306.7716547072773  loss: 0.5023 (0.5064)  acc1: 100.0000 (99.8460)  acc5: 100.0000 (99.9890)  time: 0.0393  data: 0.0003  max mem: 190
Epoch: [17]  [ 80/468]  eta: 0:00:17  lr: 0.0  img/s: 3440.2232004972543  loss: 0.5015 (0.5064)  acc1: 100.0000 (99.8457)  acc5: 100.0000 (99.9904)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [17]  [ 90/468]  eta: 0:00:17  lr: 0.0  img/s: 3507.928465484008  loss: 0.5013 (0.5062)  acc1: 100.0000 (99.8369)  acc5: 100.0000 (99.9914)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [17]  [100/468]  eta: 0:00:16  lr: 0.0  img/s: 3577.9944551076987  loss: 0.5023 (0.5063)  acc1: 100.0000 (99.8376)  acc5: 100.0000 (99.9923)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [17]  [110/468]  eta: 0:00:15  lr: 0.0  img/s: 3463.035380477201  loss: 0.5026 (0.5063)  acc1: 100.0000 (99.8311)  acc5: 100.0000 (99.9930)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [17]  [120/468]  eta: 0:00:15  lr: 0.0  img/s: 3505.0428083645074  loss: 0.5021 (0.5059)  acc1: 100.0000 (99.8450)  acc5: 100.0000 (99.9935)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [17]  [130/468]  eta: 0:00:14  lr: 0.0  img/s: 3500.129816280495  loss: 0.5015 (0.5058)  acc1: 100.0000 (99.8449)  acc5: 100.0000 (99.9940)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [17]  [140/468]  eta: 0:00:13  lr: 0.0  img/s: 3441.568450473089  loss: 0.5015 (0.5056)  acc1: 100.0000 (99.8559)  acc5: 100.0000 (99.9945)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [17]  [150/468]  eta: 0:00:13  lr: 0.0  img/s: 3556.6141901291817  loss: 0.5018 (0.5054)  acc1: 100.0000 (99.8655)  acc5: 100.0000 (99.9948)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [17]  [160/468]  eta: 0:00:12  lr: 0.0  img/s: 3476.3551785540844  loss: 0.5018 (0.5055)  acc1: 100.0000 (99.8593)  acc5: 100.0000 (99.9951)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [17]  [170/468]  eta: 0:00:12  lr: 0.0  img/s: 3535.7440480502632  loss: 0.5022 (0.5056)  acc1: 100.0000 (99.8584)  acc5: 100.0000 (99.9954)  time: 0.0375  data: 0.0002  max mem: 190
Epoch: [17]  [180/468]  eta: 0:00:11  lr: 0.0  img/s: 3560.3172030531923  loss: 0.5022 (0.5056)  acc1: 100.0000 (99.8532)  acc5: 100.0000 (99.9957)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [17]  [190/468]  eta: 0:00:11  lr: 0.0  img/s: 3294.737658639566  loss: 0.5019 (0.5057)  acc1: 100.0000 (99.8527)  acc5: 100.0000 (99.9959)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [17]  [200/468]  eta: 0:00:10  lr: 0.0  img/s: 3465.4048269140153  loss: 0.5015 (0.5056)  acc1: 100.0000 (99.8523)  acc5: 100.0000 (99.9961)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [17]  [210/468]  eta: 0:00:10  lr: 0.0  img/s: 3518.665285952103  loss: 0.5014 (0.5055)  acc1: 100.0000 (99.8556)  acc5: 100.0000 (99.9963)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [17]  [220/468]  eta: 0:00:10  lr: 0.0  img/s: 1131.3978588889825  loss: 0.5014 (0.5053)  acc1: 100.0000 (99.8621)  acc5: 100.0000 (99.9965)  time: 0.0452  data: 0.0003  max mem: 190
Epoch: [17]  [230/468]  eta: 0:00:09  lr: 0.0  img/s: 2135.9070318871713  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8647)  acc5: 100.0000 (99.9966)  time: 0.0558  data: 0.0021  max mem: 190
Epoch: [17]  [240/468]  eta: 0:00:09  lr: 0.0  img/s: 3016.535442981076  loss: 0.5014 (0.5053)  acc1: 100.0000 (99.8638)  acc5: 100.0000 (99.9968)  time: 0.0512  data: 0.0021  max mem: 190
Epoch: [17]  [250/468]  eta: 0:00:09  lr: 0.0  img/s: 2825.2497658215193  loss: 0.5018 (0.5052)  acc1: 100.0000 (99.8693)  acc5: 100.0000 (99.9969)  time: 0.0456  data: 0.0004  max mem: 190
Epoch: [17]  [260/468]  eta: 0:00:08  lr: 0.0  img/s: 2882.9463170499885  loss: 0.5016 (0.5051)  acc1: 100.0000 (99.8713)  acc5: 100.0000 (99.9970)  time: 0.0464  data: 0.0004  max mem: 190
Epoch: [17]  [270/468]  eta: 0:00:08  lr: 0.0  img/s: 3288.562068923273  loss: 0.5020 (0.5053)  acc1: 100.0000 (99.8616)  acc5: 100.0000 (99.9971)  time: 0.0437  data: 0.0003  max mem: 190
Epoch: [17]  [280/468]  eta: 0:00:07  lr: 0.0  img/s: 3306.242183506691  loss: 0.5015 (0.5052)  acc1: 100.0000 (99.8638)  acc5: 100.0000 (99.9972)  time: 0.0411  data: 0.0003  max mem: 190
Epoch: [17]  [290/468]  eta: 0:00:07  lr: 0.0  img/s: 3393.085239374309  loss: 0.5015 (0.5052)  acc1: 100.0000 (99.8631)  acc5: 100.0000 (99.9973)  time: 0.0401  data: 0.0003  max mem: 190
Epoch: [17]  [300/468]  eta: 0:00:07  lr: 0.0  img/s: 3480.4342966795025  loss: 0.5016 (0.5053)  acc1: 100.0000 (99.8598)  acc5: 100.0000 (99.9974)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [17]  [310/468]  eta: 0:00:06  lr: 0.0  img/s: 3473.678533069348  loss: 0.5016 (0.5053)  acc1: 100.0000 (99.8593)  acc5: 100.0000 (99.9975)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [17]  [320/468]  eta: 0:00:06  lr: 0.0  img/s: 3557.3211767824014  loss: 0.5016 (0.5053)  acc1: 100.0000 (99.8613)  acc5: 100.0000 (99.9976)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [17]  [330/468]  eta: 0:00:05  lr: 0.0  img/s: 3493.024710796497  loss: 0.5015 (0.5054)  acc1: 100.0000 (99.8607)  acc5: 100.0000 (99.9976)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [17]  [340/468]  eta: 0:00:05  lr: 0.0  img/s: 3491.5481097533216  loss: 0.5016 (0.5052)  acc1: 100.0000 (99.8648)  acc5: 100.0000 (99.9977)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [17]  [350/468]  eta: 0:00:04  lr: 0.0  img/s: 3523.5148587629947  loss: 0.5014 (0.5052)  acc1: 100.0000 (99.8687)  acc5: 100.0000 (99.9978)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [17]  [360/468]  eta: 0:00:04  lr: 0.0  img/s: 3506.46214135028  loss: 0.5014 (0.5051)  acc1: 100.0000 (99.8702)  acc5: 100.0000 (99.9978)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [17]  [370/468]  eta: 0:00:04  lr: 0.0  img/s: 3623.613226331171  loss: 0.5015 (0.5052)  acc1: 100.0000 (99.8652)  acc5: 100.0000 (99.9979)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [17]  [380/468]  eta: 0:00:03  lr: 0.0  img/s: 3479.9379812803027  loss: 0.5022 (0.5053)  acc1: 100.0000 (99.8606)  acc5: 100.0000 (99.9979)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [17]  [390/468]  eta: 0:00:03  lr: 0.0  img/s: 3509.7630961331024  loss: 0.5022 (0.5053)  acc1: 100.0000 (99.8621)  acc5: 100.0000 (99.9980)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [17]  [400/468]  eta: 0:00:02  lr: 0.0  img/s: 3538.9837444463487  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8636)  acc5: 100.0000 (99.9981)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [17]  [410/468]  eta: 0:00:02  lr: 0.0  img/s: 3372.5589366032614  loss: 0.5011 (0.5052)  acc1: 100.0000 (99.8650)  acc5: 100.0000 (99.9981)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [17]  [420/468]  eta: 0:00:01  lr: 0.0  img/s: 3518.20411800941  loss: 0.5013 (0.5052)  acc1: 100.0000 (99.8682)  acc5: 100.0000 (99.9981)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [17]  [430/468]  eta: 0:00:01  lr: 0.0  img/s: 3518.18106278547  loss: 0.5017 (0.5053)  acc1: 100.0000 (99.8659)  acc5: 100.0000 (99.9982)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [17]  [440/468]  eta: 0:00:01  lr: 0.0  img/s: 3516.79830209814  loss: 0.5017 (0.5053)  acc1: 100.0000 (99.8654)  acc5: 100.0000 (99.9982)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [17]  [450/468]  eta: 0:00:00  lr: 0.0  img/s: 3461.16000593116  loss: 0.5017 (0.5054)  acc1: 100.0000 (99.8632)  acc5: 100.0000 (99.9983)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [17]  [460/468]  eta: 0:00:00  lr: 0.0  img/s: 3561.0492829758164  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8661)  acc5: 100.0000 (99.9983)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [17] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:10  loss: 0.5526 (0.5526)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1292  data: 0.1093  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.760 Acc@5 99.900
Epoch: [18]  [  0/468]  eta: 0:01:43  lr: 0.0  img/s: 1582.0426281777382  loss: 0.5010 (0.5010)  acc1: 100.0000 (100.0000)  acc5: 100.0000 (100.0000)  time: 0.2210  data: 0.1400  max mem: 190
Epoch: [18]  [ 10/468]  eta: 0:00:26  lr: 0.0  img/s: 3425.472707667375  loss: 0.5014 (0.5025)  acc1: 100.0000 (99.9290)  acc5: 100.0000 (100.0000)  time: 0.0577  data: 0.0130  max mem: 190
Epoch: [18]  [ 20/468]  eta: 0:00:21  lr: 0.0  img/s: 3583.296036735947  loss: 0.5018 (0.5044)  acc1: 100.0000 (99.8512)  acc5: 100.0000 (100.0000)  time: 0.0393  data: 0.0003  max mem: 190
Epoch: [18]  [ 30/468]  eta: 0:00:19  lr: 0.0  img/s: 3592.9122436004686  loss: 0.5023 (0.5055)  acc1: 100.0000 (99.8236)  acc5: 100.0000 (100.0000)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [18]  [ 40/468]  eta: 0:00:18  lr: 0.0  img/s: 3484.839652340986  loss: 0.5023 (0.5059)  acc1: 100.0000 (99.8285)  acc5: 100.0000 (100.0000)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [18]  [ 50/468]  eta: 0:00:17  lr: 0.0  img/s: 3621.437806918137  loss: 0.5021 (0.5062)  acc1: 100.0000 (99.8162)  acc5: 100.0000 (100.0000)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [18]  [ 60/468]  eta: 0:00:16  lr: 0.0  img/s: 3508.08891909199  loss: 0.5019 (0.5055)  acc1: 100.0000 (99.8463)  acc5: 100.0000 (100.0000)  time: 0.0364  data: 0.0003  max mem: 190
Epoch: [18]  [ 70/468]  eta: 0:00:15  lr: 0.0  img/s: 3505.271655316954  loss: 0.5017 (0.5054)  acc1: 100.0000 (99.8460)  acc5: 100.0000 (100.0000)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [18]  [ 80/468]  eta: 0:00:15  lr: 0.0  img/s: 3068.0616960117036  loss: 0.5018 (0.5057)  acc1: 100.0000 (99.8457)  acc5: 100.0000 (100.0000)  time: 0.0387  data: 0.0002  max mem: 190
Epoch: [18]  [ 90/468]  eta: 0:00:15  lr: 0.0  img/s: 3119.982054336772  loss: 0.5018 (0.5058)  acc1: 100.0000 (99.8455)  acc5: 100.0000 (100.0000)  time: 0.0412  data: 0.0003  max mem: 190
Epoch: [18]  [100/468]  eta: 0:00:14  lr: 0.0  img/s: 3181.7023652192465  loss: 0.5017 (0.5054)  acc1: 100.0000 (99.8608)  acc5: 100.0000 (100.0000)  time: 0.0408  data: 0.0003  max mem: 190
Epoch: [18]  [110/468]  eta: 0:00:14  lr: 0.0  img/s: 3458.952349045177  loss: 0.5017 (0.5057)  acc1: 100.0000 (99.8522)  acc5: 100.0000 (100.0000)  time: 0.0400  data: 0.0003  max mem: 190
Epoch: [18]  [120/468]  eta: 0:00:13  lr: 0.0  img/s: 3476.737893250787  loss: 0.5021 (0.5061)  acc1: 100.0000 (99.8386)  acc5: 100.0000 (100.0000)  time: 0.0385  data: 0.0002  max mem: 190
Epoch: [18]  [130/468]  eta: 0:00:13  lr: 0.0  img/s: 3438.6367170737017  loss: 0.5021 (0.5058)  acc1: 100.0000 (99.8449)  acc5: 100.0000 (100.0000)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [18]  [140/468]  eta: 0:00:12  lr: 0.0  img/s: 3564.4065329969458  loss: 0.5016 (0.5057)  acc1: 100.0000 (99.8504)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [18]  [150/468]  eta: 0:00:12  lr: 0.0  img/s: 2829.6127294671987  loss: 0.5014 (0.5059)  acc1: 100.0000 (99.8500)  acc5: 100.0000 (100.0000)  time: 0.0400  data: 0.0003  max mem: 190
Epoch: [18]  [160/468]  eta: 0:00:12  lr: 0.0  img/s: 3410.263180628605  loss: 0.5017 (0.5058)  acc1: 100.0000 (99.8496)  acc5: 100.0000 (100.0000)  time: 0.0419  data: 0.0003  max mem: 190
Epoch: [18]  [170/468]  eta: 0:00:11  lr: 0.0  img/s: 3449.817263515033  loss: 0.5020 (0.5058)  acc1: 100.0000 (99.8492)  acc5: 100.0000 (100.0000)  time: 0.0401  data: 0.0003  max mem: 190
Epoch: [18]  [180/468]  eta: 0:00:11  lr: 0.0  img/s: 3422.459229793392  loss: 0.5015 (0.5058)  acc1: 100.0000 (99.8489)  acc5: 100.0000 (100.0000)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [18]  [190/468]  eta: 0:00:10  lr: 0.0  img/s: 3534.1147908973016  loss: 0.5017 (0.5058)  acc1: 100.0000 (99.8527)  acc5: 100.0000 (99.9959)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [18]  [200/468]  eta: 0:00:10  lr: 0.0  img/s: 3625.668829984805  loss: 0.5018 (0.5057)  acc1: 100.0000 (99.8562)  acc5: 100.0000 (99.9961)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [18]  [210/468]  eta: 0:00:10  lr: 0.0  img/s: 3558.1463498691055  loss: 0.5016 (0.5059)  acc1: 100.0000 (99.8519)  acc5: 100.0000 (99.9963)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [18]  [220/468]  eta: 0:00:09  lr: 0.0  img/s: 3487.5335325451474  loss: 0.5018 (0.5059)  acc1: 100.0000 (99.8480)  acc5: 100.0000 (99.9965)  time: 0.0404  data: 0.0006  max mem: 190
Epoch: [18]  [230/468]  eta: 0:00:09  lr: 0.0  img/s: 3458.0611650735577  loss: 0.5018 (0.5057)  acc1: 100.0000 (99.8546)  acc5: 100.0000 (99.9966)  time: 0.0432  data: 0.0013  max mem: 190
Epoch: [18]  [240/468]  eta: 0:00:09  lr: 0.0  img/s: 3391.713334470494  loss: 0.5017 (0.5057)  acc1: 100.0000 (99.8574)  acc5: 100.0000 (99.9968)  time: 0.0404  data: 0.0010  max mem: 190
Epoch: [18]  [250/468]  eta: 0:00:08  lr: 0.0  img/s: 3510.267956035909  loss: 0.5015 (0.5056)  acc1: 100.0000 (99.8599)  acc5: 100.0000 (99.9969)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [18]  [260/468]  eta: 0:00:08  lr: 0.0  img/s: 3547.191045979214  loss: 0.5015 (0.5057)  acc1: 100.0000 (99.8533)  acc5: 100.0000 (99.9970)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [18]  [270/468]  eta: 0:00:07  lr: 0.0  img/s: 2666.648017166018  loss: 0.5015 (0.5056)  acc1: 100.0000 (99.8587)  acc5: 100.0000 (99.9971)  time: 0.0450  data: 0.0015  max mem: 190
Epoch: [18]  [280/468]  eta: 0:00:07  lr: 0.0  img/s: 3161.4301815462345  loss: 0.5014 (0.5056)  acc1: 100.0000 (99.8610)  acc5: 100.0000 (99.9972)  time: 0.0500  data: 0.0033  max mem: 190
Epoch: [18]  [290/468]  eta: 0:00:07  lr: 0.0  img/s: 2749.0535349449287  loss: 0.5010 (0.5054)  acc1: 100.0000 (99.8658)  acc5: 100.0000 (99.9973)  time: 0.0479  data: 0.0037  max mem: 190
Epoch: [18]  [300/468]  eta: 0:00:06  lr: 0.0  img/s: 1436.9591023939017  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8676)  acc5: 100.0000 (99.9974)  time: 0.0534  data: 0.0050  max mem: 190
Epoch: [18]  [310/468]  eta: 0:00:06  lr: 0.0  img/s: 1795.242673514974  loss: 0.5012 (0.5053)  acc1: 100.0000 (99.8694)  acc5: 100.0000 (99.9975)  time: 0.0713  data: 0.0073  max mem: 190
Epoch: [18]  [320/468]  eta: 0:00:06  lr: 0.0  img/s: 3321.3372184381633  loss: 0.5012 (0.5052)  acc1: 100.0000 (99.8710)  acc5: 100.0000 (99.9976)  time: 0.0763  data: 0.0082  max mem: 190
Epoch: [18]  [330/468]  eta: 0:00:05  lr: 0.0  img/s: 3390.8564571240895  loss: 0.5013 (0.5052)  acc1: 100.0000 (99.8678)  acc5: 100.0000 (99.9976)  time: 0.0527  data: 0.0042  max mem: 190
Epoch: [18]  [340/468]  eta: 0:00:05  lr: 0.0  img/s: 3459.866289448415  loss: 0.5017 (0.5053)  acc1: 100.0000 (99.8671)  acc5: 100.0000 (99.9977)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [18]  [350/468]  eta: 0:00:05  lr: 0.0  img/s: 3461.9411650911156  loss: 0.5018 (0.5052)  acc1: 100.0000 (99.8709)  acc5: 100.0000 (99.9978)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [18]  [360/468]  eta: 0:00:04  lr: 0.0  img/s: 3527.357800817335  loss: 0.5018 (0.5054)  acc1: 100.0000 (99.8637)  acc5: 100.0000 (99.9978)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [18]  [370/468]  eta: 0:00:04  lr: 0.0  img/s: 3433.3809475084417  loss: 0.5015 (0.5055)  acc1: 100.0000 (99.8631)  acc5: 100.0000 (99.9979)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [18]  [380/468]  eta: 0:00:03  lr: 0.0  img/s: 3431.603346777544  loss: 0.5013 (0.5054)  acc1: 100.0000 (99.8667)  acc5: 100.0000 (99.9979)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [18]  [390/468]  eta: 0:00:03  lr: 0.0  img/s: 3442.7602057175104  loss: 0.5014 (0.5054)  acc1: 100.0000 (99.8661)  acc5: 100.0000 (99.9980)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [18]  [400/468]  eta: 0:00:02  lr: 0.0  img/s: 3547.4723105081966  loss: 0.5018 (0.5055)  acc1: 100.0000 (99.8636)  acc5: 100.0000 (99.9981)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [18]  [410/468]  eta: 0:00:02  lr: 0.0  img/s: 3446.229816734602  loss: 0.5017 (0.5055)  acc1: 100.0000 (99.8650)  acc5: 100.0000 (99.9981)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [18]  [420/468]  eta: 0:00:02  lr: 0.0  img/s: 3311.5649642240314  loss: 0.5014 (0.5055)  acc1: 100.0000 (99.8627)  acc5: 100.0000 (99.9981)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [18]  [430/468]  eta: 0:00:01  lr: 0.0  img/s: 3400.37059650634  loss: 0.5021 (0.5055)  acc1: 100.0000 (99.8622)  acc5: 100.0000 (99.9982)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [18]  [440/468]  eta: 0:00:01  lr: 0.0  img/s: 3415.687386275433  loss: 0.5019 (0.5054)  acc1: 100.0000 (99.8654)  acc5: 100.0000 (99.9982)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [18]  [450/468]  eta: 0:00:00  lr: 0.0  img/s: 3421.913877061928  loss: 0.5017 (0.5053)  acc1: 100.0000 (99.8683)  acc5: 100.0000 (99.9983)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [18]  [460/468]  eta: 0:00:00  lr: 0.0  img/s: 3543.0241867893274  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8678)  acc5: 100.0000 (99.9983)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [18] Total time: 0:00:19
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5526 (0.5526)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1238  data: 0.1070  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.760 Acc@5 99.900
Epoch: [19]  [  0/468]  eta: 0:01:14  lr: 0.0  img/s: 2553.5975951408145  loss: 0.5011 (0.5011)  acc1: 100.0000 (100.0000)  acc5: 100.0000 (100.0000)  time: 0.1595  data: 0.1094  max mem: 190
Epoch: [19]  [ 10/468]  eta: 0:00:22  lr: 0.0  img/s: 3504.425070823379  loss: 0.5017 (0.5064)  acc1: 100.0000 (99.8580)  acc5: 100.0000 (100.0000)  time: 0.0489  data: 0.0105  max mem: 190
Epoch: [19]  [ 20/468]  eta: 0:00:24  lr: 0.0  img/s: 3486.65020554751  loss: 0.5017 (0.5057)  acc1: 100.0000 (99.8884)  acc5: 100.0000 (100.0000)  time: 0.0484  data: 0.0032  max mem: 190
Epoch: [19]  [ 30/468]  eta: 0:00:21  lr: 0.0  img/s: 3525.944661539573  loss: 0.5015 (0.5067)  acc1: 100.0000 (99.8488)  acc5: 100.0000 (100.0000)  time: 0.0498  data: 0.0034  max mem: 190
Epoch: [19]  [ 40/468]  eta: 0:00:19  lr: 0.0  img/s: 3489.2788519657097  loss: 0.5016 (0.5076)  acc1: 100.0000 (99.8285)  acc5: 100.0000 (100.0000)  time: 0.0394  data: 0.0007  max mem: 190
Epoch: [19]  [ 50/468]  eta: 0:00:18  lr: 0.0  img/s: 3354.8351361315763  loss: 0.5016 (0.5069)  acc1: 100.0000 (99.8468)  acc5: 100.0000 (100.0000)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [19]  [ 60/468]  eta: 0:00:17  lr: 0.0  img/s: 3312.6683696765517  loss: 0.5014 (0.5068)  acc1: 100.0000 (99.8463)  acc5: 100.0000 (100.0000)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [19]  [ 70/468]  eta: 0:00:17  lr: 0.0  img/s: 2751.462736134316  loss: 0.5014 (0.5070)  acc1: 100.0000 (99.8349)  acc5: 100.0000 (100.0000)  time: 0.0392  data: 0.0003  max mem: 190
Epoch: [19]  [ 80/468]  eta: 0:00:16  lr: 0.0  img/s: 3579.8076440935642  loss: 0.5017 (0.5069)  acc1: 100.0000 (99.8360)  acc5: 100.0000 (100.0000)  time: 0.0396  data: 0.0003  max mem: 190
Epoch: [19]  [ 90/468]  eta: 0:00:15  lr: 0.0  img/s: 3475.297523336052  loss: 0.5018 (0.5067)  acc1: 100.0000 (99.8455)  acc5: 100.0000 (99.9914)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [19]  [100/468]  eta: 0:00:15  lr: 0.0  img/s: 3478.4273468832407  loss: 0.5018 (0.5063)  acc1: 100.0000 (99.8530)  acc5: 100.0000 (99.9923)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [19]  [110/468]  eta: 0:00:14  lr: 0.0  img/s: 3450.704202901345  loss: 0.5015 (0.5061)  acc1: 100.0000 (99.8522)  acc5: 100.0000 (99.9930)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [19]  [120/468]  eta: 0:00:14  lr: 0.0  img/s: 3475.7925158617118  loss: 0.5014 (0.5058)  acc1: 100.0000 (99.8644)  acc5: 100.0000 (99.9935)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [19]  [130/468]  eta: 0:00:13  lr: 0.0  img/s: 3526.477351550184  loss: 0.5016 (0.5058)  acc1: 100.0000 (99.8569)  acc5: 100.0000 (99.9940)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [19]  [140/468]  eta: 0:00:13  lr: 0.0  img/s: 3456.5917150620016  loss: 0.5015 (0.5055)  acc1: 100.0000 (99.8615)  acc5: 100.0000 (99.9945)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [19]  [150/468]  eta: 0:00:12  lr: 0.0  img/s: 3405.741748447382  loss: 0.5017 (0.5060)  acc1: 100.0000 (99.8448)  acc5: 100.0000 (99.9948)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [19]  [160/468]  eta: 0:00:12  lr: 0.0  img/s: 3511.1403289624277  loss: 0.5019 (0.5060)  acc1: 100.0000 (99.8447)  acc5: 100.0000 (99.9951)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [19]  [170/468]  eta: 0:00:11  lr: 0.0  img/s: 3486.763427591671  loss: 0.5012 (0.5060)  acc1: 100.0000 (99.8447)  acc5: 100.0000 (99.9954)  time: 0.0377  data: 0.0002  max mem: 190
Epoch: [19]  [180/468]  eta: 0:00:11  lr: 0.0  img/s: 3573.5646524754716  loss: 0.5015 (0.5060)  acc1: 100.0000 (99.8446)  acc5: 100.0000 (99.9957)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [19]  [190/468]  eta: 0:00:11  lr: 0.0  img/s: 2244.987965309314  loss: 0.5017 (0.5059)  acc1: 100.0000 (99.8487)  acc5: 100.0000 (99.9959)  time: 0.0387  data: 0.0003  max mem: 190
Epoch: [19]  [200/468]  eta: 0:00:10  lr: 0.0  img/s: 2122.91724958777  loss: 0.5017 (0.5059)  acc1: 100.0000 (99.8484)  acc5: 100.0000 (99.9961)  time: 0.0410  data: 0.0005  max mem: 190
Epoch: [19]  [210/468]  eta: 0:00:10  lr: 0.0  img/s: 3555.3894120607674  loss: 0.5016 (0.5058)  acc1: 100.0000 (99.8556)  acc5: 100.0000 (99.9963)  time: 0.0397  data: 0.0008  max mem: 190
Epoch: [19]  [220/468]  eta: 0:00:09  lr: 0.0  img/s: 3395.510220603117  loss: 0.5021 (0.5059)  acc1: 100.0000 (99.8480)  acc5: 100.0000 (99.9965)  time: 0.0376  data: 0.0006  max mem: 190
Epoch: [19]  [230/468]  eta: 0:00:09  lr: 0.0  img/s: 3594.1870765605336  loss: 0.5019 (0.5059)  acc1: 100.0000 (99.8512)  acc5: 100.0000 (99.9966)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [19]  [240/468]  eta: 0:00:08  lr: 0.0  img/s: 3571.2113241936236  loss: 0.5017 (0.5058)  acc1: 100.0000 (99.8541)  acc5: 100.0000 (99.9968)  time: 0.0363  data: 0.0003  max mem: 190
Epoch: [19]  [250/468]  eta: 0:00:08  lr: 0.0  img/s: 3560.0102913677174  loss: 0.5014 (0.5058)  acc1: 100.0000 (99.8537)  acc5: 100.0000 (99.9969)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [19]  [260/468]  eta: 0:00:08  lr: 0.0  img/s: 3507.401364099616  loss: 0.5014 (0.5058)  acc1: 100.0000 (99.8533)  acc5: 100.0000 (99.9970)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [19]  [270/468]  eta: 0:00:07  lr: 0.0  img/s: 3432.788209341731  loss: 0.5022 (0.5058)  acc1: 100.0000 (99.8530)  acc5: 100.0000 (99.9971)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [19]  [280/468]  eta: 0:00:07  lr: 0.0  img/s: 3539.5203818590576  loss: 0.5018 (0.5058)  acc1: 100.0000 (99.8526)  acc5: 100.0000 (99.9972)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [19]  [290/468]  eta: 0:00:06  lr: 0.0  img/s: 3547.2613579300682  loss: 0.5017 (0.5056)  acc1: 100.0000 (99.8577)  acc5: 100.0000 (99.9973)  time: 0.0376  data: 0.0002  max mem: 190
Epoch: [19]  [300/468]  eta: 0:00:06  lr: 0.0  img/s: 2699.906018667525  loss: 0.5016 (0.5055)  acc1: 100.0000 (99.8598)  acc5: 100.0000 (99.9974)  time: 0.0444  data: 0.0017  max mem: 190
Epoch: [19]  [310/468]  eta: 0:00:06  lr: 0.0  img/s: 3534.952506995885  loss: 0.5015 (0.5054)  acc1: 100.0000 (99.8643)  acc5: 100.0000 (99.9975)  time: 0.0490  data: 0.0022  max mem: 190
Epoch: [19]  [320/468]  eta: 0:00:05  lr: 0.0  img/s: 3387.0485972228357  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8661)  acc5: 100.0000 (99.9976)  time: 0.0421  data: 0.0008  max mem: 190
Epoch: [19]  [330/468]  eta: 0:00:05  lr: 0.0  img/s: 3504.2649521882445  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8678)  acc5: 100.0000 (99.9976)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [19]  [340/468]  eta: 0:00:05  lr: 0.0  img/s: 3549.5832170791214  loss: 0.5013 (0.5052)  acc1: 100.0000 (99.8694)  acc5: 100.0000 (99.9977)  time: 0.0370  data: 0.0002  max mem: 190
Epoch: [19]  [350/468]  eta: 0:00:04  lr: 0.0  img/s: 3632.6606130320047  loss: 0.5011 (0.5051)  acc1: 100.0000 (99.8731)  acc5: 100.0000 (99.9978)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [19]  [360/468]  eta: 0:00:04  lr: 0.0  img/s: 3434.5230942449907  loss: 0.5011 (0.5052)  acc1: 100.0000 (99.8723)  acc5: 100.0000 (99.9978)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [19]  [370/468]  eta: 0:00:03  lr: 0.0  img/s: 3486.7860728829082  loss: 0.5017 (0.5051)  acc1: 100.0000 (99.8737)  acc5: 100.0000 (99.9979)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [19]  [380/468]  eta: 0:00:03  lr: 0.0  img/s: 3491.0259191343816  loss: 0.5014 (0.5052)  acc1: 100.0000 (99.8729)  acc5: 100.0000 (99.9979)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [19]  [390/468]  eta: 0:00:03  lr: 0.0  img/s: 3496.4597940695685  loss: 0.5012 (0.5052)  acc1: 100.0000 (99.8721)  acc5: 100.0000 (99.9980)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [19]  [400/468]  eta: 0:00:02  lr: 0.0  img/s: 3354.898029070276  loss: 0.5026 (0.5053)  acc1: 100.0000 (99.8695)  acc5: 100.0000 (99.9981)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [19]  [410/468]  eta: 0:00:02  lr: 0.0  img/s: 3429.082752101377  loss: 0.5030 (0.5054)  acc1: 100.0000 (99.8631)  acc5: 100.0000 (99.9981)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [19]  [420/468]  eta: 0:00:01  lr: 0.0  img/s: 3449.2853187019346  loss: 0.5028 (0.5055)  acc1: 100.0000 (99.8608)  acc5: 100.0000 (99.9981)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [19]  [430/468]  eta: 0:00:01  lr: 0.0  img/s: 3384.3156428278753  loss: 0.5021 (0.5055)  acc1: 100.0000 (99.8604)  acc5: 100.0000 (99.9982)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [19]  [440/468]  eta: 0:00:01  lr: 0.0  img/s: 3651.8849617718283  loss: 0.5021 (0.5054)  acc1: 100.0000 (99.8636)  acc5: 100.0000 (99.9982)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [19]  [450/468]  eta: 0:00:00  lr: 0.0  img/s: 3311.217747953891  loss: 0.5016 (0.5054)  acc1: 100.0000 (99.8649)  acc5: 100.0000 (99.9983)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [19]  [460/468]  eta: 0:00:00  lr: 0.0  img/s: 3503.921890092677  loss: 0.5015 (0.5053)  acc1: 100.0000 (99.8661)  acc5: 100.0000 (99.9983)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [19] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5526 (0.5526)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1225  data: 0.1034  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.760 Acc@5 99.900
Epoch: [20]  [  0/468]  eta: 0:01:14  lr: 0.0  img/s: 2341.7148440226115  loss: 0.5010 (0.5010)  acc1: 100.0000 (100.0000)  acc5: 100.0000 (100.0000)  time: 0.1601  data: 0.1053  max mem: 190
Epoch: [20]  [ 10/468]  eta: 0:00:21  lr: 0.0  img/s: 3526.8248448021022  loss: 0.5020 (0.5086)  acc1: 100.0000 (99.7869)  acc5: 100.0000 (99.9290)  time: 0.0480  data: 0.0099  max mem: 190
Epoch: [20]  [ 20/468]  eta: 0:00:19  lr: 0.0  img/s: 3593.922414197064  loss: 0.5016 (0.5055)  acc1: 100.0000 (99.8512)  acc5: 100.0000 (99.9628)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [20]  [ 30/468]  eta: 0:00:17  lr: 0.0  img/s: 3534.9059567939844  loss: 0.5012 (0.5058)  acc1: 100.0000 (99.8488)  acc5: 100.0000 (99.9748)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [20]  [ 40/468]  eta: 0:00:17  lr: 0.0  img/s: 3485.608907644863  loss: 0.5011 (0.5047)  acc1: 100.0000 (99.8857)  acc5: 100.0000 (99.9809)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [20]  [ 50/468]  eta: 0:00:16  lr: 0.0  img/s: 3497.3936653941864  loss: 0.5017 (0.5054)  acc1: 100.0000 (99.8621)  acc5: 100.0000 (99.9847)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [20]  [ 60/468]  eta: 0:00:15  lr: 0.0  img/s: 3581.527098065377  loss: 0.5021 (0.5058)  acc1: 100.0000 (99.8591)  acc5: 100.0000 (99.9872)  time: 0.0377  data: 0.0002  max mem: 190
Epoch: [20]  [ 70/468]  eta: 0:00:15  lr: 0.0  img/s: 3528.6329142211152  loss: 0.5021 (0.5062)  acc1: 100.0000 (99.8349)  acc5: 100.0000 (99.9890)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [20]  [ 80/468]  eta: 0:00:15  lr: 0.0  img/s: 2350.4599690908058  loss: 0.5021 (0.5064)  acc1: 100.0000 (99.8167)  acc5: 100.0000 (99.9904)  time: 0.0438  data: 0.0006  max mem: 190
Epoch: [20]  [ 90/468]  eta: 0:00:16  lr: 0.0  img/s: 2334.2924002034842  loss: 0.5032 (0.5071)  acc1: 100.0000 (99.7940)  acc5: 100.0000 (99.9914)  time: 0.0559  data: 0.0035  max mem: 190
Epoch: [20]  [100/468]  eta: 0:00:16  lr: 0.0  img/s: 2650.677699823739  loss: 0.5027 (0.5071)  acc1: 100.0000 (99.7989)  acc5: 100.0000 (99.9923)  time: 0.0599  data: 0.0053  max mem: 190
Epoch: [20]  [110/468]  eta: 0:00:16  lr: 0.0  img/s: 3581.6704604587244  loss: 0.5015 (0.5067)  acc1: 100.0000 (99.8170)  acc5: 100.0000 (99.9930)  time: 0.0547  data: 0.0049  max mem: 190
Epoch: [20]  [120/468]  eta: 0:00:15  lr: 0.0  img/s: 3442.914752941931  loss: 0.5014 (0.5064)  acc1: 100.0000 (99.8192)  acc5: 100.0000 (99.9935)  time: 0.0442  data: 0.0027  max mem: 190
Epoch: [20]  [130/468]  eta: 0:00:14  lr: 0.0  img/s: 3491.4345767649966  loss: 0.5014 (0.5060)  acc1: 100.0000 (99.8330)  acc5: 100.0000 (99.9940)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [20]  [140/468]  eta: 0:00:14  lr: 0.0  img/s: 3387.8607929626614  loss: 0.5012 (0.5059)  acc1: 100.0000 (99.8338)  acc5: 100.0000 (99.9945)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [20]  [150/468]  eta: 0:00:13  lr: 0.0  img/s: 3553.7184804697067  loss: 0.5013 (0.5059)  acc1: 100.0000 (99.8344)  acc5: 100.0000 (99.9948)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [20]  [160/468]  eta: 0:00:13  lr: 0.0  img/s: 3362.335988776993  loss: 0.5016 (0.5056)  acc1: 100.0000 (99.8447)  acc5: 100.0000 (99.9951)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [20]  [170/468]  eta: 0:00:12  lr: 0.0  img/s: 3438.6587415453987  loss: 0.5016 (0.5055)  acc1: 100.0000 (99.8492)  acc5: 100.0000 (99.9954)  time: 0.0374  data: 0.0002  max mem: 190
Epoch: [20]  [180/468]  eta: 0:00:12  lr: 0.0  img/s: 3528.2155029080272  loss: 0.5015 (0.5054)  acc1: 100.0000 (99.8532)  acc5: 100.0000 (99.9957)  time: 0.0376  data: 0.0002  max mem: 190
Epoch: [20]  [190/468]  eta: 0:00:11  lr: 0.0  img/s: 1953.906248180283  loss: 0.5014 (0.5054)  acc1: 100.0000 (99.8527)  acc5: 100.0000 (99.9959)  time: 0.0404  data: 0.0003  max mem: 190
Epoch: [20]  [200/468]  eta: 0:00:11  lr: 0.0  img/s: 3480.118443228667  loss: 0.5014 (0.5053)  acc1: 100.0000 (99.8562)  acc5: 100.0000 (99.9961)  time: 0.0413  data: 0.0003  max mem: 190
Epoch: [20]  [210/468]  eta: 0:00:10  lr: 0.0  img/s: 3652.009169631378  loss: 0.5014 (0.5051)  acc1: 100.0000 (99.8593)  acc5: 100.0000 (99.9963)  time: 0.0400  data: 0.0003  max mem: 190
Epoch: [20]  [220/468]  eta: 0:00:10  lr: 0.0  img/s: 3518.8959152640136  loss: 0.5023 (0.5051)  acc1: 100.0000 (99.8621)  acc5: 100.0000 (99.9965)  time: 0.0387  data: 0.0003  max mem: 190
Epoch: [20]  [230/468]  eta: 0:00:09  lr: 0.0  img/s: 3478.3597372137924  loss: 0.5024 (0.5052)  acc1: 100.0000 (99.8647)  acc5: 100.0000 (99.9966)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [20]  [240/468]  eta: 0:00:09  lr: 0.0  img/s: 3593.104613263551  loss: 0.5015 (0.5052)  acc1: 100.0000 (99.8638)  acc5: 100.0000 (99.9968)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [20]  [250/468]  eta: 0:00:08  lr: 0.0  img/s: 3502.0705148694396  loss: 0.5011 (0.5050)  acc1: 100.0000 (99.8693)  acc5: 100.0000 (99.9969)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [20]  [260/468]  eta: 0:00:08  lr: 0.0  img/s: 3548.550904536231  loss: 0.5014 (0.5049)  acc1: 100.0000 (99.8713)  acc5: 100.0000 (99.9970)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [20]  [270/468]  eta: 0:00:08  lr: 0.0  img/s: 2910.8002667519695  loss: 0.5018 (0.5049)  acc1: 100.0000 (99.8732)  acc5: 100.0000 (99.9971)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [20]  [280/468]  eta: 0:00:07  lr: 0.0  img/s: 3321.871535791408  loss: 0.5018 (0.5048)  acc1: 100.0000 (99.8777)  acc5: 100.0000 (99.9972)  time: 0.0391  data: 0.0003  max mem: 190
Epoch: [20]  [290/468]  eta: 0:00:07  lr: 0.0  img/s: 3177.1833563147648  loss: 0.5011 (0.5048)  acc1: 100.0000 (99.8792)  acc5: 100.0000 (99.9973)  time: 0.0402  data: 0.0003  max mem: 190
Epoch: [20]  [300/468]  eta: 0:00:06  lr: 0.0  img/s: 3421.7175926220993  loss: 0.5013 (0.5047)  acc1: 100.0000 (99.8832)  acc5: 100.0000 (99.9974)  time: 0.0398  data: 0.0003  max mem: 190
Epoch: [20]  [310/468]  eta: 0:00:06  lr: 0.0  img/s: 3331.3326797304508  loss: 0.5014 (0.5048)  acc1: 100.0000 (99.8819)  acc5: 100.0000 (99.9975)  time: 0.0390  data: 0.0003  max mem: 190
Epoch: [20]  [320/468]  eta: 0:00:06  lr: 0.0  img/s: 3345.3444414673204  loss: 0.5013 (0.5048)  acc1: 100.0000 (99.8807)  acc5: 100.0000 (99.9976)  time: 0.0390  data: 0.0003  max mem: 190
Epoch: [20]  [330/468]  eta: 0:00:05  lr: 0.0  img/s: 3488.5759808699495  loss: 0.5024 (0.5050)  acc1: 100.0000 (99.8749)  acc5: 100.0000 (99.9976)  time: 0.0384  data: 0.0002  max mem: 190
Epoch: [20]  [340/468]  eta: 0:00:05  lr: 0.0  img/s: 3523.722996344161  loss: 0.5023 (0.5049)  acc1: 100.0000 (99.8740)  acc5: 100.0000 (99.9977)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [20]  [350/468]  eta: 0:00:04  lr: 0.0  img/s: 2873.1184416140427  loss: 0.5014 (0.5048)  acc1: 100.0000 (99.8776)  acc5: 100.0000 (99.9978)  time: 0.0456  data: 0.0013  max mem: 190
Epoch: [20]  [360/468]  eta: 0:00:04  lr: 0.0  img/s: 3543.959706645367  loss: 0.5016 (0.5051)  acc1: 100.0000 (99.8702)  acc5: 100.0000 (99.9978)  time: 0.0500  data: 0.0031  max mem: 190
Epoch: [20]  [370/468]  eta: 0:00:04  lr: 0.0  img/s: 3406.02263614678  loss: 0.5031 (0.5052)  acc1: 100.0000 (99.8652)  acc5: 100.0000 (99.9979)  time: 0.0420  data: 0.0021  max mem: 190
Epoch: [20]  [380/468]  eta: 0:00:03  lr: 0.0  img/s: 3502.6417182076775  loss: 0.5021 (0.5053)  acc1: 100.0000 (99.8647)  acc5: 100.0000 (99.9979)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [20]  [390/468]  eta: 0:00:03  lr: 0.0  img/s: 3555.130432479323  loss: 0.5016 (0.5053)  acc1: 100.0000 (99.8641)  acc5: 100.0000 (99.9980)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [20]  [400/468]  eta: 0:00:02  lr: 0.0  img/s: 3579.879255046043  loss: 0.5021 (0.5053)  acc1: 100.0000 (99.8656)  acc5: 100.0000 (99.9981)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [20]  [410/468]  eta: 0:00:02  lr: 0.0  img/s: 3541.8321150547567  loss: 0.5021 (0.5055)  acc1: 100.0000 (99.8593)  acc5: 100.0000 (99.9981)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [20]  [420/468]  eta: 0:00:01  lr: 0.0  img/s: 3626.5015232266737  loss: 0.5020 (0.5055)  acc1: 100.0000 (99.8608)  acc5: 100.0000 (99.9981)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [20]  [430/468]  eta: 0:00:01  lr: 0.0  img/s: 3511.5077736135368  loss: 0.5018 (0.5054)  acc1: 100.0000 (99.8622)  acc5: 100.0000 (99.9982)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [20]  [440/468]  eta: 0:00:01  lr: 0.0  img/s: 3442.4290798104607  loss: 0.5018 (0.5054)  acc1: 100.0000 (99.8636)  acc5: 100.0000 (99.9982)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [20]  [450/468]  eta: 0:00:00  lr: 0.0  img/s: 3436.281727642797  loss: 0.5014 (0.5054)  acc1: 100.0000 (99.8649)  acc5: 100.0000 (99.9983)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [20]  [460/468]  eta: 0:00:00  lr: 0.0  img/s: 3585.8808693677447  loss: 0.5015 (0.5053)  acc1: 100.0000 (99.8661)  acc5: 100.0000 (99.9983)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [20] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5526 (0.5526)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1171  data: 0.1000  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.760 Acc@5 99.900
Epoch: [21]  [  0/468]  eta: 0:01:14  lr: 0.0  img/s: 2728.0719123961485  loss: 0.5023 (0.5023)  acc1: 100.0000 (100.0000)  acc5: 100.0000 (100.0000)  time: 0.1582  data: 0.1113  max mem: 190
Epoch: [21]  [ 10/468]  eta: 0:00:21  lr: 0.0  img/s: 3509.7630961331024  loss: 0.5019 (0.5035)  acc1: 100.0000 (99.8580)  acc5: 100.0000 (100.0000)  time: 0.0480  data: 0.0104  max mem: 190
Epoch: [21]  [ 20/468]  eta: 0:00:19  lr: 0.0  img/s: 3548.363275854092  loss: 0.5011 (0.5025)  acc1: 100.0000 (99.9256)  acc5: 100.0000 (100.0000)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [21]  [ 30/468]  eta: 0:00:17  lr: 0.0  img/s: 3536.9320245075432  loss: 0.5011 (0.5033)  acc1: 100.0000 (99.8992)  acc5: 100.0000 (100.0000)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [21]  [ 40/468]  eta: 0:00:17  lr: 0.0  img/s: 3387.1127037803462  loss: 0.5015 (0.5049)  acc1: 100.0000 (99.8476)  acc5: 100.0000 (100.0000)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [21]  [ 50/468]  eta: 0:00:16  lr: 0.0  img/s: 3461.0038164002062  loss: 0.5020 (0.5049)  acc1: 100.0000 (99.8621)  acc5: 100.0000 (100.0000)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [21]  [ 60/468]  eta: 0:00:16  lr: 0.0  img/s: 3559.3258328637253  loss: 0.5020 (0.5063)  acc1: 100.0000 (99.8335)  acc5: 100.0000 (100.0000)  time: 0.0385  data: 0.0003  max mem: 190
Epoch: [21]  [ 70/468]  eta: 0:00:15  lr: 0.0  img/s: 3587.1507166004076  loss: 0.5018 (0.5066)  acc1: 100.0000 (99.8239)  acc5: 100.0000 (100.0000)  time: 0.0382  data: 0.0002  max mem: 190
Epoch: [21]  [ 80/468]  eta: 0:00:15  lr: 0.0  img/s: 3504.219206694211  loss: 0.5022 (0.5073)  acc1: 100.0000 (99.7975)  acc5: 100.0000 (99.9904)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [21]  [ 90/468]  eta: 0:00:14  lr: 0.0  img/s: 3514.7031882160395  loss: 0.5023 (0.5071)  acc1: 100.0000 (99.8111)  acc5: 100.0000 (99.9914)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [21]  [100/468]  eta: 0:00:14  lr: 0.0  img/s: 3495.3898720002085  loss: 0.5015 (0.5065)  acc1: 100.0000 (99.8298)  acc5: 100.0000 (99.9923)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [21]  [110/468]  eta: 0:00:13  lr: 0.0  img/s: 3589.837127974698  loss: 0.5015 (0.5065)  acc1: 100.0000 (99.8311)  acc5: 100.0000 (99.9930)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [21]  [120/468]  eta: 0:00:13  lr: 0.0  img/s: 3519.795658530509  loss: 0.5017 (0.5061)  acc1: 100.0000 (99.8386)  acc5: 100.0000 (99.9935)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [21]  [130/468]  eta: 0:00:13  lr: 0.0  img/s: 3226.63977354  loss: 0.5019 (0.5060)  acc1: 100.0000 (99.8449)  acc5: 100.0000 (99.9940)  time: 0.0443  data: 0.0013  max mem: 190
Epoch: [21]  [140/468]  eta: 0:00:13  lr: 0.0  img/s: 3418.7324851309872  loss: 0.5012 (0.5058)  acc1: 100.0000 (99.8504)  acc5: 100.0000 (99.9945)  time: 0.0494  data: 0.0030  max mem: 190
Epoch: [21]  [150/468]  eta: 0:00:12  lr: 0.0  img/s: 3301.728209196632  loss: 0.5011 (0.5056)  acc1: 100.0000 (99.8551)  acc5: 100.0000 (99.9948)  time: 0.0424  data: 0.0021  max mem: 190
Epoch: [21]  [160/468]  eta: 0:00:12  lr: 0.0  img/s: 3516.1994432982938  loss: 0.5016 (0.5054)  acc1: 100.0000 (99.8641)  acc5: 100.0000 (99.9951)  time: 0.0375  data: 0.0004  max mem: 190
Epoch: [21]  [170/468]  eta: 0:00:11  lr: 0.0  img/s: 3412.6910930865265  loss: 0.5020 (0.5054)  acc1: 100.0000 (99.8675)  acc5: 100.0000 (99.9954)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [21]  [180/468]  eta: 0:00:11  lr: 0.0  img/s: 3458.840925929505  loss: 0.5017 (0.5052)  acc1: 100.0000 (99.8748)  acc5: 100.0000 (99.9957)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [21]  [190/468]  eta: 0:00:10  lr: 0.0  img/s: 3420.583945512351  loss: 0.5025 (0.5056)  acc1: 100.0000 (99.8609)  acc5: 100.0000 (99.9959)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [21]  [200/468]  eta: 0:00:10  lr: 0.0  img/s: 3542.0657913835194  loss: 0.5025 (0.5056)  acc1: 100.0000 (99.8640)  acc5: 100.0000 (99.9961)  time: 0.0401  data: 0.0007  max mem: 190
Epoch: [21]  [210/468]  eta: 0:00:10  lr: 0.0  img/s: 2264.1699083992644  loss: 0.5019 (0.5054)  acc1: 100.0000 (99.8704)  acc5: 100.0000 (99.9963)  time: 0.0423  data: 0.0006  max mem: 190
Epoch: [21]  [220/468]  eta: 0:00:09  lr: 0.0  img/s: 3493.297450646123  loss: 0.5017 (0.5056)  acc1: 100.0000 (99.8657)  acc5: 100.0000 (99.9965)  time: 0.0395  data: 0.0003  max mem: 190
Epoch: [21]  [230/468]  eta: 0:00:09  lr: 0.0  img/s: 3334.0221079563803  loss: 0.5015 (0.5056)  acc1: 100.0000 (99.8647)  acc5: 100.0000 (99.9966)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [21]  [240/468]  eta: 0:00:08  lr: 0.0  img/s: 3511.278111694648  loss: 0.5009 (0.5056)  acc1: 100.0000 (99.8671)  acc5: 100.0000 (99.9968)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [21]  [250/468]  eta: 0:00:08  lr: 0.0  img/s: 3417.8183855360326  loss: 0.5010 (0.5054)  acc1: 100.0000 (99.8724)  acc5: 100.0000 (99.9969)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [21]  [260/468]  eta: 0:00:08  lr: 0.0  img/s: 3480.8630466495933  loss: 0.5011 (0.5052)  acc1: 100.0000 (99.8773)  acc5: 100.0000 (99.9970)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [21]  [270/468]  eta: 0:00:07  lr: 0.0  img/s: 3405.2881046315442  loss: 0.5010 (0.5053)  acc1: 100.0000 (99.8760)  acc5: 100.0000 (99.9971)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [21]  [280/468]  eta: 0:00:07  lr: 0.0  img/s: 3434.2814229147875  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8721)  acc5: 100.0000 (99.9972)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [21]  [290/468]  eta: 0:00:06  lr: 0.0  img/s: 3563.2236808920156  loss: 0.5016 (0.5055)  acc1: 100.0000 (99.8658)  acc5: 100.0000 (99.9973)  time: 0.0376  data: 0.0002  max mem: 190
Epoch: [21]  [300/468]  eta: 0:00:06  lr: 0.0  img/s: 3592.4794871623294  loss: 0.5019 (0.5055)  acc1: 100.0000 (99.8650)  acc5: 100.0000 (99.9974)  time: 0.0386  data: 0.0002  max mem: 190
Epoch: [21]  [310/468]  eta: 0:00:06  lr: 0.0  img/s: 3356.597030229141  loss: 0.5019 (0.5054)  acc1: 100.0000 (99.8694)  acc5: 100.0000 (99.9975)  time: 0.0393  data: 0.0003  max mem: 190
Epoch: [21]  [320/468]  eta: 0:00:05  lr: 0.0  img/s: 3536.978628086542  loss: 0.5018 (0.5054)  acc1: 100.0000 (99.8686)  acc5: 100.0000 (99.9976)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [21]  [330/468]  eta: 0:00:05  lr: 0.0  img/s: 3451.014739440377  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8702)  acc5: 100.0000 (99.9976)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [21]  [340/468]  eta: 0:00:04  lr: 0.0  img/s: 3516.4527817441085  loss: 0.5016 (0.5055)  acc1: 100.0000 (99.8671)  acc5: 100.0000 (99.9977)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [21]  [350/468]  eta: 0:00:04  lr: 0.0  img/s: 3412.734480084417  loss: 0.5018 (0.5055)  acc1: 100.0000 (99.8665)  acc5: 100.0000 (99.9978)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [21]  [360/468]  eta: 0:00:04  lr: 0.0  img/s: 3465.7851342104245  loss: 0.5018 (0.5056)  acc1: 100.0000 (99.8658)  acc5: 100.0000 (99.9978)  time: 0.0376  data: 0.0003  max mem: 190
Epoch: [21]  [370/468]  eta: 0:00:03  lr: 0.0  img/s: 3468.4272165800967  loss: 0.5020 (0.5056)  acc1: 100.0000 (99.8631)  acc5: 100.0000 (99.9979)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [21]  [380/468]  eta: 0:00:03  lr: 0.0  img/s: 3363.4102780962403  loss: 0.5013 (0.5055)  acc1: 100.0000 (99.8647)  acc5: 100.0000 (99.9979)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [21]  [390/468]  eta: 0:00:03  lr: 0.0  img/s: 2199.209044732099  loss: 0.5015 (0.5056)  acc1: 100.0000 (99.8581)  acc5: 100.0000 (99.9980)  time: 0.0394  data: 0.0003  max mem: 190
Epoch: [21]  [400/468]  eta: 0:00:02  lr: 0.0  img/s: 2553.67047351773  loss: 0.5015 (0.5055)  acc1: 100.0000 (99.8617)  acc5: 100.0000 (99.9981)  time: 0.0502  data: 0.0012  max mem: 190
Epoch: [21]  [410/468]  eta: 0:00:02  lr: 0.0  img/s: 1813.054809971835  loss: 0.5012 (0.5056)  acc1: 100.0000 (99.8612)  acc5: 100.0000 (99.9981)  time: 0.0729  data: 0.0018  max mem: 190
Epoch: [21]  [420/468]  eta: 0:00:01  lr: 0.0  img/s: 3504.1505906925136  loss: 0.5014 (0.5055)  acc1: 100.0000 (99.8645)  acc5: 100.0000 (99.9981)  time: 0.0660  data: 0.0021  max mem: 190
Epoch: [21]  [430/468]  eta: 0:00:01  lr: 0.0  img/s: 3498.3052402486546  loss: 0.5016 (0.5055)  acc1: 100.0000 (99.8641)  acc5: 100.0000 (99.9982)  time: 0.0414  data: 0.0015  max mem: 190
Epoch: [21]  [440/468]  eta: 0:00:01  lr: 0.0  img/s: 3455.47932650224  loss: 0.5026 (0.5055)  acc1: 100.0000 (99.8654)  acc5: 100.0000 (99.9982)  time: 0.0375  data: 0.0003  max mem: 190
Epoch: [21]  [450/468]  eta: 0:00:00  lr: 0.0  img/s: 3469.52553654864  loss: 0.5017 (0.5055)  acc1: 100.0000 (99.8649)  acc5: 100.0000 (99.9983)  time: 0.0379  data: 0.0002  max mem: 190
Epoch: [21]  [460/468]  eta: 0:00:00  lr: 0.0  img/s: 3408.40123417601  loss: 0.5012 (0.5054)  acc1: 100.0000 (99.8644)  acc5: 100.0000 (99.9983)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [21] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5526 (0.5526)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1211  data: 0.1003  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.760 Acc@5 99.900
Epoch: [22]  [  0/468]  eta: 0:01:15  lr: 0.0  img/s: 2650.978737692452  loss: 0.5289 (0.5289)  acc1: 99.2188 (99.2188)  acc5: 100.0000 (100.0000)  time: 0.1605  data: 0.1121  max mem: 190
Epoch: [22]  [ 10/468]  eta: 0:00:21  lr: 0.0  img/s: 3643.780071807193  loss: 0.5014 (0.5080)  acc1: 100.0000 (99.7869)  acc5: 100.0000 (100.0000)  time: 0.0478  data: 0.0105  max mem: 190
Epoch: [22]  [ 20/468]  eta: 0:00:19  lr: 0.0  img/s: 3613.248478974856  loss: 0.5017 (0.5065)  acc1: 100.0000 (99.8140)  acc5: 100.0000 (100.0000)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [22]  [ 30/468]  eta: 0:00:18  lr: 0.0  img/s: 3002.331487881533  loss: 0.5025 (0.5070)  acc1: 100.0000 (99.7984)  acc5: 100.0000 (100.0000)  time: 0.0383  data: 0.0002  max mem: 190
Epoch: [22]  [ 40/468]  eta: 0:00:17  lr: 0.0  img/s: 3119.275085118003  loss: 0.5025 (0.5063)  acc1: 100.0000 (99.8285)  acc5: 100.0000 (100.0000)  time: 0.0408  data: 0.0003  max mem: 190
Epoch: [22]  [ 50/468]  eta: 0:00:17  lr: 0.0  img/s: 3240.760777969601  loss: 0.5020 (0.5059)  acc1: 100.0000 (99.8315)  acc5: 100.0000 (100.0000)  time: 0.0409  data: 0.0003  max mem: 190
Epoch: [22]  [ 60/468]  eta: 0:00:16  lr: 0.0  img/s: 3273.243863478399  loss: 0.5018 (0.5053)  acc1: 100.0000 (99.8591)  acc5: 100.0000 (100.0000)  time: 0.0399  data: 0.0003  max mem: 190
Epoch: [22]  [ 70/468]  eta: 0:00:16  lr: 0.0  img/s: 3546.089855876564  loss: 0.5017 (0.5057)  acc1: 100.0000 (99.8460)  acc5: 100.0000 (100.0000)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [22]  [ 80/468]  eta: 0:00:15  lr: 0.0  img/s: 3495.6857423769866  loss: 0.5015 (0.5052)  acc1: 100.0000 (99.8553)  acc5: 100.0000 (100.0000)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [22]  [ 90/468]  eta: 0:00:15  lr: 0.0  img/s: 3509.327197615437  loss: 0.5012 (0.5051)  acc1: 100.0000 (99.8626)  acc5: 100.0000 (100.0000)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [22]  [100/468]  eta: 0:00:14  lr: 0.0  img/s: 3441.8111485078693  loss: 0.5012 (0.5050)  acc1: 100.0000 (99.8685)  acc5: 100.0000 (100.0000)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [22]  [110/468]  eta: 0:00:14  lr: 0.0  img/s: 3550.592648439877  loss: 0.5015 (0.5047)  acc1: 100.0000 (99.8733)  acc5: 100.0000 (100.0000)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [22]  [120/468]  eta: 0:00:13  lr: 0.0  img/s: 3530.7678931965406  loss: 0.5019 (0.5047)  acc1: 100.0000 (99.8773)  acc5: 100.0000 (100.0000)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [22]  [130/468]  eta: 0:00:13  lr: 0.0  img/s: 3508.9831436806776  loss: 0.5032 (0.5050)  acc1: 100.0000 (99.8628)  acc5: 100.0000 (100.0000)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [22]  [140/468]  eta: 0:00:12  lr: 0.0  img/s: 3529.468033212588  loss: 0.5011 (0.5050)  acc1: 100.0000 (99.8615)  acc5: 100.0000 (100.0000)  time: 0.0364  data: 0.0002  max mem: 190
Epoch: [22]  [150/468]  eta: 0:00:12  lr: 0.0  img/s: 2792.9587614385377  loss: 0.5011 (0.5051)  acc1: 100.0000 (99.8603)  acc5: 100.0000 (100.0000)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [22]  [160/468]  eta: 0:00:12  lr: 0.0  img/s: 2942.4194586181006  loss: 0.5013 (0.5049)  acc1: 100.0000 (99.8690)  acc5: 100.0000 (100.0000)  time: 0.0412  data: 0.0003  max mem: 190
Epoch: [22]  [170/468]  eta: 0:00:11  lr: 0.0  img/s: 2722.703843637636  loss: 0.5013 (0.5051)  acc1: 100.0000 (99.8675)  acc5: 100.0000 (100.0000)  time: 0.0481  data: 0.0007  max mem: 190
Epoch: [22]  [180/468]  eta: 0:00:11  lr: 0.0  img/s: 3555.0362674400894  loss: 0.5030 (0.5052)  acc1: 100.0000 (99.8576)  acc5: 100.0000 (100.0000)  time: 0.0497  data: 0.0020  max mem: 190
Epoch: [22]  [190/468]  eta: 0:00:11  lr: 0.0  img/s: 3453.900964365442  loss: 0.5035 (0.5053)  acc1: 100.0000 (99.8527)  acc5: 100.0000 (100.0000)  time: 0.0437  data: 0.0016  max mem: 190
Epoch: [22]  [200/468]  eta: 0:00:10  lr: 0.0  img/s: 3532.4173070849565  loss: 0.5012 (0.5051)  acc1: 100.0000 (99.8601)  acc5: 100.0000 (100.0000)  time: 0.0420  data: 0.0007  max mem: 190
Epoch: [22]  [210/468]  eta: 0:00:10  lr: 0.0  img/s: 3577.0647157981703  loss: 0.5012 (0.5052)  acc1: 100.0000 (99.8593)  acc5: 100.0000 (100.0000)  time: 0.0391  data: 0.0007  max mem: 190
Epoch: [22]  [220/468]  eta: 0:00:09  lr: 0.0  img/s: 3581.6704604587244  loss: 0.5013 (0.5052)  acc1: 100.0000 (99.8621)  acc5: 100.0000 (100.0000)  time: 0.0362  data: 0.0002  max mem: 190
Epoch: [22]  [230/468]  eta: 0:00:09  lr: 0.0  img/s: 3663.3725597232365  loss: 0.5018 (0.5053)  acc1: 100.0000 (99.8580)  acc5: 100.0000 (100.0000)  time: 0.0360  data: 0.0002  max mem: 190
Epoch: [22]  [240/468]  eta: 0:00:09  lr: 0.0  img/s: 3384.7210369697887  loss: 0.5011 (0.5051)  acc1: 100.0000 (99.8638)  acc5: 100.0000 (100.0000)  time: 0.0362  data: 0.0002  max mem: 190
Epoch: [22]  [250/468]  eta: 0:00:08  lr: 0.0  img/s: 3542.907281534177  loss: 0.5011 (0.5053)  acc1: 100.0000 (99.8599)  acc5: 100.0000 (99.9969)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [22]  [260/468]  eta: 0:00:08  lr: 0.0  img/s: 3555.4835958092162  loss: 0.5016 (0.5053)  acc1: 100.0000 (99.8623)  acc5: 100.0000 (99.9970)  time: 0.0364  data: 0.0002  max mem: 190
Epoch: [22]  [270/468]  eta: 0:00:07  lr: 0.0  img/s: 3492.0704566150644  loss: 0.5015 (0.5053)  acc1: 100.0000 (99.8645)  acc5: 100.0000 (99.9971)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [22]  [280/468]  eta: 0:00:07  lr: 0.0  img/s: 3521.4348345118  loss: 0.5016 (0.5054)  acc1: 100.0000 (99.8638)  acc5: 100.0000 (99.9972)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [22]  [290/468]  eta: 0:00:06  lr: 0.0  img/s: 3505.0428083645074  loss: 0.5013 (0.5054)  acc1: 100.0000 (99.8631)  acc5: 100.0000 (99.9973)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [22]  [300/468]  eta: 0:00:06  lr: 0.0  img/s: 3506.233138931158  loss: 0.5011 (0.5053)  acc1: 100.0000 (99.8676)  acc5: 100.0000 (99.9974)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [22]  [310/468]  eta: 0:00:06  lr: 0.0  img/s: 3525.0417722682564  loss: 0.5024 (0.5054)  acc1: 100.0000 (99.8618)  acc5: 100.0000 (99.9975)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [22]  [320/468]  eta: 0:00:05  lr: 0.0  img/s: 3452.812512862729  loss: 0.5018 (0.5053)  acc1: 100.0000 (99.8637)  acc5: 100.0000 (99.9976)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [22]  [330/468]  eta: 0:00:05  lr: 0.0  img/s: 3561.8052942347244  loss: 0.5023 (0.5053)  acc1: 100.0000 (99.8631)  acc5: 100.0000 (99.9976)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [22]  [340/468]  eta: 0:00:04  lr: 0.0  img/s: 3574.278394716519  loss: 0.5026 (0.5055)  acc1: 100.0000 (99.8580)  acc5: 100.0000 (99.9977)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [22]  [350/468]  eta: 0:00:04  lr: 0.0  img/s: 3384.678359328702  loss: 0.5016 (0.5055)  acc1: 100.0000 (99.8598)  acc5: 100.0000 (99.9978)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [22]  [360/468]  eta: 0:00:04  lr: 0.0  img/s: 3586.3839088291684  loss: 0.5012 (0.5054)  acc1: 100.0000 (99.8637)  acc5: 100.0000 (99.9978)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [22]  [370/468]  eta: 0:00:03  lr: 0.0  img/s: 3521.6196261069204  loss: 0.5016 (0.5053)  acc1: 100.0000 (99.8652)  acc5: 100.0000 (99.9979)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [22]  [380/468]  eta: 0:00:03  lr: 0.0  img/s: 3221.6056214633327  loss: 0.5016 (0.5054)  acc1: 100.0000 (99.8626)  acc5: 100.0000 (99.9979)  time: 0.0375  data: 0.0002  max mem: 190
Epoch: [22]  [390/468]  eta: 0:00:03  lr: 0.0  img/s: 3571.116305367275  loss: 0.5020 (0.5054)  acc1: 100.0000 (99.8641)  acc5: 100.0000 (99.9980)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [22]  [400/468]  eta: 0:00:02  lr: 0.0  img/s: 3317.9503609215863  loss: 0.5021 (0.5053)  acc1: 100.0000 (99.8675)  acc5: 100.0000 (99.9981)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [22]  [410/468]  eta: 0:00:02  lr: 0.0  img/s: 3300.0842861006613  loss: 0.5026 (0.5054)  acc1: 100.0000 (99.8669)  acc5: 100.0000 (99.9981)  time: 0.0388  data: 0.0003  max mem: 190
Epoch: [22]  [420/468]  eta: 0:00:01  lr: 0.0  img/s: 3266.949700001217  loss: 0.5020 (0.5054)  acc1: 100.0000 (99.8664)  acc5: 100.0000 (99.9981)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [22]  [430/468]  eta: 0:00:01  lr: 0.0  img/s: 3328.420585372507  loss: 0.5020 (0.5054)  acc1: 100.0000 (99.8695)  acc5: 100.0000 (99.9982)  time: 0.0382  data: 0.0002  max mem: 190
Epoch: [22]  [440/468]  eta: 0:00:01  lr: 0.0  img/s: 3564.5721949632502  loss: 0.5018 (0.5053)  acc1: 100.0000 (99.8707)  acc5: 100.0000 (99.9982)  time: 0.0384  data: 0.0003  max mem: 190
Epoch: [22]  [450/468]  eta: 0:00:00  lr: 0.0  img/s: 3527.3114504218024  loss: 0.5012 (0.5053)  acc1: 100.0000 (99.8718)  acc5: 100.0000 (99.9983)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [22]  [460/468]  eta: 0:00:00  lr: 0.0  img/s: 2299.469375867327  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8712)  acc5: 100.0000 (99.9983)  time: 0.0474  data: 0.0027  max mem: 190
Epoch: [22] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:09  loss: 0.5526 (0.5526)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1234  data: 0.1042  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.760 Acc@5 99.900
Epoch: [23]  [  0/468]  eta: 0:01:14  lr: 0.0  img/s: 2733.669966190069  loss: 0.5012 (0.5012)  acc1: 100.0000 (100.0000)  acc5: 100.0000 (100.0000)  time: 0.1601  data: 0.1132  max mem: 190
Epoch: [23]  [ 10/468]  eta: 0:00:22  lr: 0.0  img/s: 3446.274060712658  loss: 0.5017 (0.5046)  acc1: 100.0000 (99.8580)  acc5: 100.0000 (100.0000)  time: 0.0492  data: 0.0106  max mem: 190
Epoch: [23]  [ 20/468]  eta: 0:00:19  lr: 0.0  img/s: 3491.5708172370287  loss: 0.5015 (0.5045)  acc1: 100.0000 (99.8512)  acc5: 100.0000 (100.0000)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [23]  [ 30/468]  eta: 0:00:18  lr: 0.0  img/s: 3525.875192098039  loss: 0.5010 (0.5035)  acc1: 100.0000 (99.8992)  acc5: 100.0000 (100.0000)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [23]  [ 40/468]  eta: 0:00:17  lr: 0.0  img/s: 3545.949327626747  loss: 0.5010 (0.5039)  acc1: 100.0000 (99.9047)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [23]  [ 50/468]  eta: 0:00:16  lr: 0.0  img/s: 3406.779102602339  loss: 0.5022 (0.5054)  acc1: 100.0000 (99.8468)  acc5: 100.0000 (100.0000)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [23]  [ 60/468]  eta: 0:00:16  lr: 0.0  img/s: 3442.517373824165  loss: 0.5016 (0.5047)  acc1: 100.0000 (99.8719)  acc5: 100.0000 (100.0000)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [23]  [ 70/468]  eta: 0:00:15  lr: 0.0  img/s: 3544.2638569806436  loss: 0.5016 (0.5044)  acc1: 100.0000 (99.8900)  acc5: 100.0000 (100.0000)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [23]  [ 80/468]  eta: 0:00:15  lr: 0.0  img/s: 3308.952418519797  loss: 0.5017 (0.5041)  acc1: 100.0000 (99.9035)  acc5: 100.0000 (100.0000)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [23]  [ 90/468]  eta: 0:00:14  lr: 0.0  img/s: 3444.7925056143727  loss: 0.5013 (0.5038)  acc1: 100.0000 (99.9141)  acc5: 100.0000 (100.0000)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [23]  [100/468]  eta: 0:00:14  lr: 0.0  img/s: 3346.699946390056  loss: 0.5014 (0.5042)  acc1: 100.0000 (99.8994)  acc5: 100.0000 (100.0000)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [23]  [110/468]  eta: 0:00:13  lr: 0.0  img/s: 3565.8269925611053  loss: 0.5017 (0.5043)  acc1: 100.0000 (99.9015)  acc5: 100.0000 (100.0000)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [23]  [120/468]  eta: 0:00:13  lr: 0.0  img/s: 3452.7014849543066  loss: 0.5016 (0.5047)  acc1: 100.0000 (99.8902)  acc5: 100.0000 (100.0000)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [23]  [130/468]  eta: 0:00:13  lr: 0.0  img/s: 3512.19693966335  loss: 0.5024 (0.5048)  acc1: 100.0000 (99.8867)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [23]  [140/468]  eta: 0:00:12  lr: 0.0  img/s: 3590.485347029948  loss: 0.5020 (0.5048)  acc1: 100.0000 (99.8892)  acc5: 100.0000 (100.0000)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [23]  [150/468]  eta: 0:00:12  lr: 0.0  img/s: 3541.8087491176334  loss: 0.5026 (0.5050)  acc1: 100.0000 (99.8655)  acc5: 100.0000 (100.0000)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [23]  [160/468]  eta: 0:00:11  lr: 0.0  img/s: 3459.130640962862  loss: 0.5020 (0.5050)  acc1: 100.0000 (99.8690)  acc5: 100.0000 (100.0000)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [23]  [170/468]  eta: 0:00:11  lr: 0.0  img/s: 3506.5079454237884  loss: 0.5016 (0.5050)  acc1: 100.0000 (99.8721)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [23]  [180/468]  eta: 0:00:11  lr: 0.0  img/s: 2821.315311811909  loss: 0.5016 (0.5049)  acc1: 100.0000 (99.8748)  acc5: 100.0000 (100.0000)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [23]  [190/468]  eta: 0:00:10  lr: 0.0  img/s: 2042.9344353371844  loss: 0.5018 (0.5054)  acc1: 100.0000 (99.8609)  acc5: 100.0000 (100.0000)  time: 0.0486  data: 0.0024  max mem: 190
Epoch: [23]  [200/468]  eta: 0:00:10  lr: 0.0  img/s: 2757.029867302083  loss: 0.5019 (0.5053)  acc1: 100.0000 (99.8562)  acc5: 100.0000 (100.0000)  time: 0.0597  data: 0.0042  max mem: 190
Epoch: [23]  [210/468]  eta: 0:00:10  lr: 0.0  img/s: 3406.9088163063275  loss: 0.5013 (0.5052)  acc1: 100.0000 (99.8593)  acc5: 100.0000 (100.0000)  time: 0.0557  data: 0.0046  max mem: 190
Epoch: [23]  [220/468]  eta: 0:00:10  lr: 0.0  img/s: 2642.445376331384  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8586)  acc5: 100.0000 (100.0000)  time: 0.0512  data: 0.0032  max mem: 190
Epoch: [23]  [230/468]  eta: 0:00:09  lr: 0.0  img/s: 3511.8523228279496  loss: 0.5018 (0.5054)  acc1: 100.0000 (99.8580)  acc5: 100.0000 (100.0000)  time: 0.0497  data: 0.0029  max mem: 190
Epoch: [23]  [240/468]  eta: 0:00:09  lr: 0.0  img/s: 3404.683434166635  loss: 0.5018 (0.5052)  acc1: 100.0000 (99.8638)  acc5: 100.0000 (100.0000)  time: 0.0423  data: 0.0024  max mem: 190
Epoch: [23]  [250/468]  eta: 0:00:08  lr: 0.0  img/s: 3574.5639714498775  loss: 0.5021 (0.5055)  acc1: 100.0000 (99.8537)  acc5: 100.0000 (100.0000)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [23]  [260/468]  eta: 0:00:08  lr: 0.0  img/s: 3545.7151386265473  loss: 0.5029 (0.5056)  acc1: 100.0000 (99.8473)  acc5: 100.0000 (100.0000)  time: 0.0366  data: 0.0002  max mem: 190
Epoch: [23]  [270/468]  eta: 0:00:08  lr: 0.0  img/s: 3458.840925929505  loss: 0.5019 (0.5058)  acc1: 100.0000 (99.8443)  acc5: 100.0000 (100.0000)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [23]  [280/468]  eta: 0:00:07  lr: 0.0  img/s: 3398.4548947618296  loss: 0.5020 (0.5057)  acc1: 100.0000 (99.8471)  acc5: 100.0000 (100.0000)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [23]  [290/468]  eta: 0:00:07  lr: 0.0  img/s: 3445.367286169011  loss: 0.5016 (0.5056)  acc1: 100.0000 (99.8523)  acc5: 100.0000 (100.0000)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [23]  [300/468]  eta: 0:00:06  lr: 0.0  img/s: 3603.692572057619  loss: 0.5012 (0.5056)  acc1: 100.0000 (99.8521)  acc5: 100.0000 (100.0000)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [23]  [310/468]  eta: 0:00:06  lr: 0.0  img/s: 3618.4356242122785  loss: 0.5012 (0.5056)  acc1: 100.0000 (99.8543)  acc5: 100.0000 (99.9975)  time: 0.0364  data: 0.0002  max mem: 190
Epoch: [23]  [320/468]  eta: 0:00:05  lr: 0.0  img/s: 3546.9098260473165  loss: 0.5012 (0.5055)  acc1: 100.0000 (99.8588)  acc5: 100.0000 (99.9976)  time: 0.0362  data: 0.0002  max mem: 190
Epoch: [23]  [330/468]  eta: 0:00:05  lr: 0.0  img/s: 3492.774736677748  loss: 0.5013 (0.5054)  acc1: 100.0000 (99.8607)  acc5: 100.0000 (99.9976)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [23]  [340/468]  eta: 0:00:05  lr: 0.0  img/s: 3468.15834625323  loss: 0.5015 (0.5055)  acc1: 100.0000 (99.8602)  acc5: 100.0000 (99.9977)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [23]  [350/468]  eta: 0:00:04  lr: 0.0  img/s: 3601.323566502992  loss: 0.5015 (0.5054)  acc1: 100.0000 (99.8642)  acc5: 100.0000 (99.9978)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [23]  [360/468]  eta: 0:00:04  lr: 0.0  img/s: 2893.6212487064504  loss: 0.5011 (0.5053)  acc1: 100.0000 (99.8680)  acc5: 100.0000 (99.9978)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [23]  [370/468]  eta: 0:00:03  lr: 0.0  img/s: 3375.824741879095  loss: 0.5014 (0.5053)  acc1: 100.0000 (99.8694)  acc5: 100.0000 (99.9979)  time: 0.0392  data: 0.0003  max mem: 190
Epoch: [23]  [380/468]  eta: 0:00:03  lr: 0.0  img/s: 3237.5166558120463  loss: 0.5017 (0.5052)  acc1: 100.0000 (99.8708)  acc5: 100.0000 (99.9979)  time: 0.0395  data: 0.0003  max mem: 190
Epoch: [23]  [390/468]  eta: 0:00:03  lr: 0.0  img/s: 3385.7456233287926  loss: 0.5013 (0.5054)  acc1: 100.0000 (99.8621)  acc5: 100.0000 (99.9980)  time: 0.0392  data: 0.0003  max mem: 190
Epoch: [23]  [400/468]  eta: 0:00:02  lr: 0.0  img/s: 3163.553884682243  loss: 0.5014 (0.5055)  acc1: 100.0000 (99.8597)  acc5: 100.0000 (99.9981)  time: 0.0394  data: 0.0003  max mem: 190
Epoch: [23]  [410/468]  eta: 0:00:02  lr: 0.0  img/s: 3531.023335350293  loss: 0.5014 (0.5054)  acc1: 100.0000 (99.8631)  acc5: 100.0000 (99.9981)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [23]  [420/468]  eta: 0:00:01  lr: 0.0  img/s: 3491.4345767649966  loss: 0.5012 (0.5054)  acc1: 100.0000 (99.8645)  acc5: 100.0000 (99.9981)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [23]  [430/468]  eta: 0:00:01  lr: 0.0  img/s: 3502.344669219579  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8677)  acc5: 100.0000 (99.9982)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [23]  [440/468]  eta: 0:00:01  lr: 0.0  img/s: 3352.5516242241065  loss: 0.5015 (0.5053)  acc1: 100.0000 (99.8671)  acc5: 100.0000 (99.9982)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [23]  [450/468]  eta: 0:00:00  lr: 0.0  img/s: 3509.4418972538715  loss: 0.5017 (0.5054)  acc1: 100.0000 (99.8649)  acc5: 100.0000 (99.9983)  time: 0.0378  data: 0.0003  max mem: 190
Epoch: [23]  [460/468]  eta: 0:00:00  lr: 0.0  img/s: 3522.7056685235857  loss: 0.5021 (0.5054)  acc1: 100.0000 (99.8661)  acc5: 100.0000 (99.9983)  time: 0.0377  data: 0.0003  max mem: 190
Epoch: [23] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:10  loss: 0.5526 (0.5526)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1268  data: 0.1103  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.760 Acc@5 99.900
Epoch: [24]  [  0/468]  eta: 0:01:47  lr: 0.0  img/s: 1711.2187061733432  loss: 0.5025 (0.5025)  acc1: 100.0000 (100.0000)  acc5: 100.0000 (100.0000)  time: 0.2291  data: 0.1543  max mem: 190
Epoch: [24]  [ 10/468]  eta: 0:00:27  lr: 0.0  img/s: 3611.1340611148107  loss: 0.5018 (0.5043)  acc1: 100.0000 (99.9290)  acc5: 100.0000 (100.0000)  time: 0.0607  data: 0.0158  max mem: 190
Epoch: [24]  [ 20/468]  eta: 0:00:22  lr: 0.0  img/s: 3511.41590524092  loss: 0.5014 (0.5039)  acc1: 100.0000 (99.9256)  acc5: 100.0000 (100.0000)  time: 0.0407  data: 0.0011  max mem: 190
Epoch: [24]  [ 30/468]  eta: 0:00:19  lr: 0.0  img/s: 3481.3144765424895  loss: 0.5014 (0.5041)  acc1: 100.0000 (99.9244)  acc5: 100.0000 (100.0000)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [24]  [ 40/468]  eta: 0:00:18  lr: 0.0  img/s: 3515.209470431093  loss: 0.5016 (0.5042)  acc1: 100.0000 (99.9238)  acc5: 100.0000 (100.0000)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [24]  [ 50/468]  eta: 0:00:17  lr: 0.0  img/s: 3570.3563367449406  loss: 0.5016 (0.5043)  acc1: 100.0000 (99.9234)  acc5: 100.0000 (100.0000)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [24]  [ 60/468]  eta: 0:00:16  lr: 0.0  img/s: 3415.687386275433  loss: 0.5020 (0.5050)  acc1: 100.0000 (99.9103)  acc5: 100.0000 (100.0000)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [24]  [ 70/468]  eta: 0:00:16  lr: 0.0  img/s: 3545.2234424010303  loss: 0.5020 (0.5053)  acc1: 100.0000 (99.9010)  acc5: 100.0000 (100.0000)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [24]  [ 80/468]  eta: 0:00:15  lr: 0.0  img/s: 3526.755340673201  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8939)  acc5: 100.0000 (100.0000)  time: 0.0368  data: 0.0003  max mem: 190
Epoch: [24]  [ 90/468]  eta: 0:00:15  lr: 0.0  img/s: 3453.1012188454733  loss: 0.5016 (0.5053)  acc1: 100.0000 (99.8884)  acc5: 100.0000 (100.0000)  time: 0.0369  data: 0.0003  max mem: 190
Epoch: [24]  [100/468]  eta: 0:00:14  lr: 0.0  img/s: 3487.6014993146546  loss: 0.5016 (0.5054)  acc1: 100.0000 (99.8840)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [24]  [110/468]  eta: 0:00:14  lr: 0.0  img/s: 2461.0849347220183  loss: 0.5017 (0.5054)  acc1: 100.0000 (99.8874)  acc5: 100.0000 (100.0000)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [24]  [120/468]  eta: 0:00:14  lr: 0.0  img/s: 2650.651525848832  loss: 0.5020 (0.5053)  acc1: 100.0000 (99.8838)  acc5: 100.0000 (100.0000)  time: 0.0487  data: 0.0037  max mem: 190
Epoch: [24]  [130/468]  eta: 0:00:13  lr: 0.0  img/s: 3563.412884469873  loss: 0.5020 (0.5059)  acc1: 100.0000 (99.8628)  acc5: 100.0000 (100.0000)  time: 0.0480  data: 0.0042  max mem: 190
Epoch: [24]  [140/468]  eta: 0:00:13  lr: 0.0  img/s: 3567.5093329080532  loss: 0.5022 (0.5060)  acc1: 100.0000 (99.8559)  acc5: 100.0000 (100.0000)  time: 0.0371  data: 0.0008  max mem: 190
Epoch: [24]  [150/468]  eta: 0:00:12  lr: 0.0  img/s: 3569.383099527957  loss: 0.5025 (0.5060)  acc1: 100.0000 (99.8500)  acc5: 100.0000 (100.0000)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [24]  [160/468]  eta: 0:00:12  lr: 0.0  img/s: 3519.657206542761  loss: 0.5023 (0.5060)  acc1: 100.0000 (99.8447)  acc5: 100.0000 (100.0000)  time: 0.0367  data: 0.0002  max mem: 190
Epoch: [24]  [170/468]  eta: 0:00:11  lr: 0.0  img/s: 3487.624155493192  loss: 0.5023 (0.5062)  acc1: 100.0000 (99.8355)  acc5: 100.0000 (100.0000)  time: 0.0366  data: 0.0003  max mem: 190
Epoch: [24]  [180/468]  eta: 0:00:11  lr: 0.0  img/s: 3531.464640684098  loss: 0.5028 (0.5061)  acc1: 100.0000 (99.8403)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [24]  [190/468]  eta: 0:00:11  lr: 0.0  img/s: 3589.5011065275094  loss: 0.5028 (0.5063)  acc1: 100.0000 (99.8282)  acc5: 100.0000 (100.0000)  time: 0.0368  data: 0.0002  max mem: 190
Epoch: [24]  [200/468]  eta: 0:00:10  lr: 0.0  img/s: 3495.2533333333336  loss: 0.5031 (0.5062)  acc1: 100.0000 (99.8290)  acc5: 100.0000 (100.0000)  time: 0.0397  data: 0.0003  max mem: 190
Epoch: [24]  [210/468]  eta: 0:00:10  lr: 0.0  img/s: 3428.9075441330506  loss: 0.5015 (0.5061)  acc1: 100.0000 (99.8334)  acc5: 100.0000 (100.0000)  time: 0.0424  data: 0.0003  max mem: 190
Epoch: [24]  [220/468]  eta: 0:00:09  lr: 0.0  img/s: 3414.38400386675  loss: 0.5015 (0.5061)  acc1: 100.0000 (99.8339)  acc5: 100.0000 (100.0000)  time: 0.0392  data: 0.0003  max mem: 190
Epoch: [24]  [230/468]  eta: 0:00:09  lr: 0.0  img/s: 3399.0143147471654  loss: 0.5015 (0.5060)  acc1: 100.0000 (99.8377)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [24]  [240/468]  eta: 0:00:09  lr: 0.0  img/s: 3501.956296557212  loss: 0.5014 (0.5059)  acc1: 100.0000 (99.8412)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [24]  [250/468]  eta: 0:00:08  lr: 0.0  img/s: 3344.2399960133553  loss: 0.5023 (0.5060)  acc1: 100.0000 (99.8413)  acc5: 100.0000 (100.0000)  time: 0.0381  data: 0.0003  max mem: 190
Epoch: [24]  [260/468]  eta: 0:00:08  lr: 0.0  img/s: 3525.6899536362084  loss: 0.5022 (0.5059)  acc1: 100.0000 (99.8443)  acc5: 100.0000 (100.0000)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [24]  [270/468]  eta: 0:00:07  lr: 0.0  img/s: 3487.352300776886  loss: 0.5019 (0.5060)  acc1: 100.0000 (99.8414)  acc5: 100.0000 (100.0000)  time: 0.0374  data: 0.0003  max mem: 190
Epoch: [24]  [280/468]  eta: 0:00:07  lr: 0.0  img/s: 2152.0459854892374  loss: 0.5021 (0.5059)  acc1: 100.0000 (99.8443)  acc5: 100.0000 (100.0000)  time: 0.0529  data: 0.0012  max mem: 190
Epoch: [24]  [290/468]  eta: 0:00:07  lr: 0.0  img/s: 3302.459367522314  loss: 0.5020 (0.5060)  acc1: 100.0000 (99.8443)  acc5: 100.0000 (100.0000)  time: 0.0637  data: 0.0030  max mem: 190
Epoch: [24]  [300/468]  eta: 0:00:06  lr: 0.0  img/s: 3530.41962254225  loss: 0.5014 (0.5058)  acc1: 100.0000 (99.8495)  acc5: 100.0000 (100.0000)  time: 0.0480  data: 0.0020  max mem: 190
Epoch: [24]  [310/468]  eta: 0:00:06  lr: 0.0  img/s: 3457.0146105255026  loss: 0.5014 (0.5059)  acc1: 100.0000 (99.8468)  acc5: 100.0000 (99.9975)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [24]  [320/468]  eta: 0:00:06  lr: 0.0  img/s: 3494.411580543749  loss: 0.5013 (0.5058)  acc1: 100.0000 (99.8515)  acc5: 100.0000 (99.9976)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [24]  [330/468]  eta: 0:00:05  lr: 0.0  img/s: 3519.4956929894715  loss: 0.5011 (0.5058)  acc1: 100.0000 (99.8513)  acc5: 100.0000 (99.9976)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [24]  [340/468]  eta: 0:00:05  lr: 0.0  img/s: 3488.8480264098466  loss: 0.5015 (0.5058)  acc1: 100.0000 (99.8534)  acc5: 100.0000 (99.9977)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [24]  [350/468]  eta: 0:00:04  lr: 0.0  img/s: 3487.6014993146546  loss: 0.5016 (0.5057)  acc1: 100.0000 (99.8531)  acc5: 100.0000 (99.9978)  time: 0.0373  data: 0.0002  max mem: 190
Epoch: [24]  [360/468]  eta: 0:00:04  lr: 0.0  img/s: 3494.88928236642  loss: 0.5014 (0.5056)  acc1: 100.0000 (99.8572)  acc5: 100.0000 (99.9978)  time: 0.0371  data: 0.0003  max mem: 190
Epoch: [24]  [370/468]  eta: 0:00:03  lr: 0.0  img/s: 3483.6185915529513  loss: 0.5014 (0.5055)  acc1: 100.0000 (99.8610)  acc5: 100.0000 (99.9979)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [24]  [380/468]  eta: 0:00:03  lr: 0.0  img/s: 3589.2371338030994  loss: 0.5013 (0.5055)  acc1: 100.0000 (99.8626)  acc5: 100.0000 (99.9979)  time: 0.0370  data: 0.0003  max mem: 190
Epoch: [24]  [390/468]  eta: 0:00:03  lr: 0.0  img/s: 3577.636805874866  loss: 0.5017 (0.5054)  acc1: 100.0000 (99.8641)  acc5: 100.0000 (99.9980)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [24]  [400/468]  eta: 0:00:02  lr: 0.0  img/s: 3608.3186837559733  loss: 0.5017 (0.5054)  acc1: 100.0000 (99.8636)  acc5: 100.0000 (99.9981)  time: 0.0369  data: 0.0002  max mem: 190
Epoch: [24]  [410/468]  eta: 0:00:02  lr: 0.0  img/s: 3593.03247222594  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8669)  acc5: 100.0000 (99.9981)  time: 0.0365  data: 0.0002  max mem: 190
Epoch: [24]  [420/468]  eta: 0:00:01  lr: 0.0  img/s: 3527.8909179321718  loss: 0.5011 (0.5052)  acc1: 100.0000 (99.8701)  acc5: 100.0000 (99.9981)  time: 0.0371  data: 0.0002  max mem: 190
Epoch: [24]  [430/468]  eta: 0:00:01  lr: 0.0  img/s: 3469.6824961869556  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8695)  acc5: 100.0000 (99.9982)  time: 0.0372  data: 0.0002  max mem: 190
Epoch: [24]  [440/468]  eta: 0:00:01  lr: 0.0  img/s: 3280.2235731873475  loss: 0.5018 (0.5053)  acc1: 100.0000 (99.8671)  acc5: 100.0000 (99.9982)  time: 0.0373  data: 0.0003  max mem: 190
Epoch: [24]  [450/468]  eta: 0:00:00  lr: 0.0  img/s: 3400.090640219381  loss: 0.5016 (0.5053)  acc1: 100.0000 (99.8683)  acc5: 100.0000 (99.9983)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [24]  [460/468]  eta: 0:00:00  lr: 0.0  img/s: 3525.0186273415493  loss: 0.5012 (0.5053)  acc1: 100.0000 (99.8678)  acc5: 100.0000 (99.9983)  time: 0.0383  data: 0.0003  max mem: 190
Epoch: [24] Total time: 0:00:18
Test:   [ 0/79]  eta: 0:00:10  loss: 0.5526 (0.5526)  acc1: 98.4375 (98.4375)  acc5: 100.0000 (100.0000)  time: 0.1272  data: 0.1089  max mem: 190
Test:  Total time: 0:00:01
Test:  Acc@1 98.760 Acc@5 99.900
Epoch: [25]  [  0/468]  eta: 0:01:19  lr: 0.0  img/s: 2520.7810759796785  loss: 0.5107 (0.5107)  acc1: 99.2188 (99.2188)  acc5: 100.0000 (100.0000)  time: 0.1689  data: 0.1181  max mem: 190
Epoch: [25]  [ 10/468]  eta: 0:00:23  lr: 0.0  img/s: 3186.9908166475716  loss: 0.5024 (0.5065)  acc1: 100.0000 (99.7869)  acc5: 100.0000 (100.0000)  time: 0.0521  data: 0.0111  max mem: 190
Epoch: [25]  [ 20/468]  eta: 0:00:20  lr: 0.0  img/s: 3260.640089400675  loss: 0.5013 (0.5041)  acc1: 100.0000 (99.8884)  acc5: 100.0000 (100.0000)  time: 0.0400  data: 0.0003  max mem: 190
Epoch: [25]  [ 30/468]  eta: 0:00:20  lr: 0.0  img/s: 2113.99792093243  loss: 0.5013 (0.5043)  acc1: 100.0000 (99.8992)  acc5: 100.0000 (100.0000)  time: 0.0456  data: 0.0007  max mem: 190
Epoch: [25]  [ 40/468]  eta: 0:00:20  lr: 0.0  img/s: 3353.9967888847937  loss: 0.5018 (0.5048)  acc1: 100.0000 (99.8476)  acc5: 100.0000 (100.0000)  time: 0.0491  data: 0.0017  max mem: 190
Epoch: [25]  [ 50/468]  eta: 0:00:19  lr: 0.0  img/s: 3436.0398087643284  loss: 0.5022 (0.5044)  acc1: 100.0000 (99.8775)  acc5: 100.0000 (100.0000)  time: 0.0437  data: 0.0023  max mem: 190
Epoch: [25]  [ 60/468]  eta: 0:00:18  lr: 0.0  img/s: 3491.0713208136085  loss: 0.5016 (0.5045)  acc1: 100.0000 (99.8847)  acc5: 100.0000 (100.0000)  time: 0.0389  data: 0.0012  max mem: 190
Epoch: [25]  [ 70/468]  eta: 0:00:17  lr: 0.0  img/s: 3572.019188417754  loss: 0.5019 (0.5045)  acc1: 100.0000 (99.8900)  acc5: 100.0000 (100.0000)  time: 0.0380  data: 0.0003  max mem: 190
Epoch: [25]  [ 80/468]  eta: 0:00:16  lr: 0.0  img/s: 3435.6660011262993  loss: 0.5024 (0.5046)  acc1: 100.0000 (99.8843)  acc5: 100.0000 (100.0000)  time: 0.0384  data: 0.0003  max mem: 190
Epoch: [25]  [ 90/468]  eta: 0:00:16  lr: 0.0  img/s: 3535.301672593178  loss: 0.5018 (0.5043)  acc1: 100.0000 (99.8970)  acc5: 100.0000 (100.0000)  time: 0.0372  data: 0.0003  max mem: 190
Epoch: [25]  [100/468]  eta: 0:00:15  lr: 0.0  img/s: 3394.9304852060527  loss: 0.5015 (0.5049)  acc1: 100.0000 (99.8840)  acc5: 100.0000 (100.0000)  time: 0.0367  data: 0.0003  max mem: 190
Epoch: [25]  [110/468]  eta: 0:00:14  lr: 0.0  img/s: 3215.5082982457193  loss: 0.5013 (0.5053)  acc1: 100.0000 (99.8733)  acc5: 100.0000 (100.0000)  time: 0.0391  data: 0.0003  max mem: 190
Epoch: [25]  [120/468]  eta: 0:00:14  lr: 0.0  img/s: 3210.969634985855  loss: 0.5017 (0.5053)  acc1: 100.0000 (99.8709)  acc5: 100.0000 (100.0000)  time: 0.0409  data: 0.0003  max mem: 190
Epoch: [25]  [130/468]  eta: 0:00:14  lr: 0.0  img/s: 3372.6860574688094  loss: 0.5017 (0.5050)  acc1: 100.0000 (99.8807)  acc5: 100.0000 (100.0000)  time: 0.0408  data: 0.0003  max mem: 190
Epoch: [25]  [140/468]  eta: 0:00:13  lr: 0.0  img/s: 3508.8226082637284  loss: 0.5019 (0.5051)  acc1: 100.0000 (99.8726)  acc5: 100.0000 (100.0000)  time: 0.0401  data: 0.0004  max mem: 190
Epoch: [25]  [150/468]  eta: 0:00:13  lr: 0.0  img/s: 3403.971062459184  loss: 0.5014 (0.5051)  acc1: 100.0000 (99.8707)  acc5: 100.0000 (100.0000)  time: 0.0385  data: 0.0004  max mem: 190
Epoch: [25]  [160/468]  eta: 0:00:12  lr: 0.0  img/s: 3460.245382007554  loss: 0.5012 (0.5052)  acc1: 100.0000 (99.8690)  acc5: 100.0000 (99.9951)  time: 0.0382  data: 0.0003  max mem: 190
Epoch: [25]  [170/468]  eta: 0:00:12  lr: 0.0  img/s: 3412.3874149876056  loss: 0.5012 (0.5051)  acc1: 100.0000 (99.8766)  acc5: 100.0000 (99.9954)  time: 0.0379  data: 0.0003  max mem: 190
Epoch: [25]  [180/468]  eta: 0:00:11  lr: 0.0  img/s: 3046.127945440206  loss: 0.5021 (0.5051)  acc1: 100.0000 (99.8748)  acc5: 100.0000 (99.9957)  time: 0.0423  data: 0.0010  max mem: 190
Epoch: [25]  [190/468]  eta: 0:00:11  lr: 0.0  img/s: 3169.0065815492317  loss: 0.5019 (0.5049)  acc1: 100.0000 (99.8773)  acc5: 100.0000 (99.9959)  time: 0.0466  data: 0.0016  max mem: 190
Epoch: [25]  [200/468]  eta: 0:00:11  lr: 0.0  img/s: 3170.6160978461794  loss: 0.5016 (0.5048)  acc1: 100.0000 (99.8795)  acc5: 100.0000 (99.9961)  time: 0.0438  data: 0.0011  max mem: 190
Epoch: [25]  [210/468]  eta: 0:00:10  lr: 0.0  img/s: 3243.109978132438  loss: 0.5014 (0.5048)  acc1: 100.0000 (99.8741)  acc5: 100.0000 (99.9963)  time: 0.0413  data: 0.0005  max mem: 190
Epoch: [25]  [220/468]  eta: 0:00:10  lr: 0.0  img/s: 3291.707513274228  loss: 0.5015 (0.5048)  acc1: 100.0000 (99.8763)  acc5: 100.0000 (99.9965)  time: 0.0406  data: 0.0003  max mem: 190
Epoch: [25]  [230/468]  eta: 0:00:09  lr: 0.0  img/s: 3252.205986224777  loss: 0.5015 (0.5048)  acc1: 100.0000 (99.8782)  acc5: 100.0000 (99.9966)  time: 0.0402  data: 0.0004  max mem: 190
Epoch: [25]  [240/468]  eta: 0:00:09  lr: 0.0  img/s: 3223.036818692105  loss: 0.5016 (0.5047)  acc1: 100.0000 (99.8833)  acc5: 100.0000 (99.9968)  time: 0.0404  data: 0.0004  max mem: 190
Epoch: [25]  [250/468]  eta: 0:00:08  lr: 0.0  img/s: 3207.804019980402  loss: 0.5014 (0.5049)  acc1: 100.0000 (99.8786)  acc5: 100.0000 (99.9969)  time: 0.0403  data: 0.0003  max mem: 190
Epoch: [25]  [260/468]  eta: 0:00:08  lr: 0.0  img/s: 3453.0123810932664  loss: 0.5013 (0.5049)  acc1: 100.0000 (99.8803)  acc5: 100.0000 (99.9970)  time: 0.0397  data: 0.0003  max mem: 190
Epoch: [25]  [270/468]  eta: 0:00:08  lr: 0.0  img/s: 3455.723990550796  loss: 0.5015 (0.5050)  acc1: 100.0000 (99.8789)  acc5: 100.0000 (99.9971)  time: 0.0384  data: 0.0003  max mem: 190
Epoch: [25]  [280/468]  eta: 0:00:07  lr: 0.0  img/s: 3354.436868939318  loss: 0.5015 (0.5050)  acc1: 100.0000 (99.8804)  acc5: 100.0000 (99.9972)  time: 0.0377  data: 0.0003  max mem: 190