MONAI 0.7 : tutorials : モジュール – 公開データセットと新しいデータセットのクイックスタート (翻訳/解説)
翻訳 : (株)クラスキャット セールスインフォメーション
作成日時 : 11/05/2021 (0.7.0)
* 本ページは、MONAI の以下のドキュメントを翻訳した上で適宜、補足説明したものです：
* サンプルコードの動作確認はしておりますが、必要な場合には適宜、追加改変しています。
* ご自由にリンクを張って頂いてかまいませんが、sales-info@classcat.com までご一報いただけると嬉しいです。
MONAI 0.7 : tutorials : モジュール – 公開データセットと新しいデータセットのクイックスタート
このノートブックは MedNISTDataset と DecathlonDataset に基づいた訓練ワークフローを素早くセットアップする方法と、新しいデータセットを作成する方法を示します。
このチュートリアルでは、MONAI 公開データセットでワークフローを素早くセットアップする方法と新しいデータセットを追加する方法を紹介します。現在、MONAI は MedNISTDataset と DecathlonDataset を提供していて、MedNIST と Decathlon データセットを自動的にダウンロードして抽出し、そして訓練/検証/テストデータを生成する PyTorch データセットとして機能します。
このチュートリアルでは以下のトピックをカバーします :
- MedNISTDataset による訓練実験とワークフローを作成する
- DecathlonDataset による訓練実験とワークフローを作成する
- 他の公開データを共有して MONAI でデータセットを追加する
環境のセットアップ
!python -c "import monai" || pip install -q "monai-weekly[nibabel, pillow, ignite, tqdm]"
!python -c "import matplotlib" || pip install -q matplotlib
%matplotlib inline
from monai.transforms import (
AddChanneld,
AsDiscreted,
Compose,
LoadImaged,
Orientationd,
Randomizable,
Resized,
ScaleIntensityd,
Spacingd,
EnsureTyped,
)
from monai.networks.nets import UNet, DenseNet121
from monai.networks.layers import Norm
from monai.losses import DiceLoss
from monai.inferers import SimpleInferer
from monai.handlers import MeanDice, StatsHandler, from_engine
from monai.engines import SupervisedTrainer
from monai.data import CacheDataset, DataLoader
from monai.config import print_config
from monai.apps import DecathlonDataset, MedNISTDataset, download_and_extract
import torch
import matplotlib.pyplot as plt
import ignite
import tempfile
import sys
import shutil
import os
import logging
インポートのセットアップ
# Copyright 2020 MONAI Consortium
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
# http://www.apache.org/licenses/LICENSE-2.0
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
print_config()
MONAI version: 0.6.0rc1+23.gc6793fd0
Numpy version: 1.20.3
Pytorch version: 1.9.0a0+c3d40fd
MONAI flags: HAS_EXT = True, USE_COMPILED = False
MONAI rev id: c6793fd0f316a448778d0047664aaf8c1895fe1c
Optional dependencies:
Pytorch Ignite version: 0.4.5
Nibabel version: 3.2.1
scikit-image version: 0.15.0
Pillow version: 7.0.0
Tensorboard version: 2.5.0
gdown version: 3.13.0
TorchVision version: 0.10.0a0
ITK version: 5.1.2
tqdm version: 4.53.0
lmdb version: 1.2.1
psutil version: 5.8.0
pandas version: 1.1.4
einops version: 0.3.0
For details about installing the optional dependencies, please visit:
https://docs.monai.io/en/latest/installation.html#installing-the-recommended-dependencies
データディレクトリのセットアップ
MONAI_DATA_DIRECTORY 環境変数でディレクトリを指定できます。これは結果をセーブしてダウンロードを再利用することを可能にします。指定されない場合、一時ディレクトリが使用されます。
directory = os.environ.get("MONAI_DATA_DIRECTORY")
root_dir = tempfile.mkdtemp() if directory is None else directory
print(root_dir)
/workspace/data/medical
ロギングのセットアップ
logging.basicConfig(stream=sys.stdout, level=logging.INFO)
MedNISTDataset による訓練実験とワークフローを作成する
MedMNIST データセットは TCIA, RSNA Bone Age チャレンジ と NIH Chest X-ray データセット からの様々なセットから集められました。
前処理変換のセットアップ
transform = Compose(
[
LoadImaged(keys="image"),
AddChanneld(keys="image"),
ScaleIntensityd(keys="image"),
EnsureTyped(keys="image"),
]
)
訓練のために MedNISTDataset を作成する
MedNISTDataset は MONAI CacheDataset から継承して期待する動作を実現するために豊富なパラメータを提供しています :
- root_dir: MedNIST データセットをダウンロードしてロードするターゲットディレクトリ。
- section: 想定されるデータ・セクション、以下のいずれかです : training, validation or test.
- transform: 入力データ上で操作を実行する変換。デフォルト変換は LoadPNGd と AddChanneld から成り、これはデータを [C, H, W] shape を持つ numpy 配列にロードできます。
- download: リソース・リンクから MedNIST をダウンロードして抽出するか否か、デフォルトは False です。想定されるファイルが既に存在していれば、それが True に設定されていてもダウンロードはスキップします。ユーザは MedNIST.tar.gz ファイルか MedNIST フォルダをルートディレクトリに手動でコピーできます。
- seed: 訓練、検証とテストデータセットをランダムに分割するためのランダムシード、デフォルトは 0 です。
- val_frac: データセット全体の検証比率のパーセンテージ、デフォルトは 0.1 です。
- test_frac: データセット全体のテスト比率のパーセンテージ、デフォルトは 0.1 です。
- cache_num: キャッシュされる項目の数。デフォルトは sys.maxsize です。(cache_num, data_length x cache_rate, data_length) の最小値を取ります。
- cache_rate: トータルでキャッシュされるデータのパーセンテージで、デフォルトは 1.0 (cache all) です。(cache_num, data_length x cache_rate, data_length) の最小値を取ります。
- num_workers: 使用するワーカースレッドの数です。0 であればシングルスレッドが使用されます。デフォルトは 0 です。
“tar フアイル” は最初にダウンロードした後でキャッシュされることに注意してください。self.__getitem__() API は、self.__len__() 内の指定されたインデックスに従って 1 {“image”: XXX, “label”: XXX} 辞書を生成します。
train_ds = MedNISTDataset(
root_dir=root_dir, transform=transform, section="training", download=True)
# the dataset can work seamlessly with the pytorch native dataset loader,
# but using monai.data.DataLoader has additional benefits of mutli-process
# random seeds handling, and the customized collate functions
train_loader = DataLoader(train_ds, batch_size=300,
shuffle=True, num_workers=10)
Verified 'MedNIST.tar.gz', md5: 0bc7306e7427e00ad1c5526a6677552d. file /workspace/data/medical/MedNIST.tar.gz exists, skip downloading. extracted file /workspace/data/medical/MedNIST exists, skip extracting. 100%|██████████| 46946/46946 [00:27<00:00, 1684.89it/s]
可視化して確認するために MedNISTDataset から画像をピックアップする
plt.subplots(3, 3, figsize=(8, 8))
for i in range(9):
plt.subplot(3, 3, i + 1)
plt.imshow(train_ds[i * 5000]["image"][0].detach().cpu(), cmap="gray")
plt.tight_layout()
plt.show()
訓練コンポーネントの作成
device = torch.device("cuda:0")
net = DenseNet121(spatial_dims=2, in_channels=1, out_channels=6).to(device)
loss = torch.nn.CrossEntropyLoss()
opt = torch.optim.Adam(net.parameters(), 1e-5)
最も簡単な訓練ワークフローを定義して実行する
訓練ワークフローを素早くセットアップするために MONAI SupervisedTrainer ハンドラを使用します。
trainer = SupervisedTrainer(
device=device,
max_epochs=5,
train_data_loader=train_loader,
network=net,
optimizer=opt,
loss_function=loss,
inferer=SimpleInferer(),
key_train_metric={
"train_acc": ignite.metrics.Accuracy(
output_transform=from_engine(["pred", "label"]))
},
train_handlers=StatsHandler(
tag_name="train_loss", output_transform=from_engine(["loss"], first=True)),
)
trainer.run()
INFO:ignite.engine.engine.SupervisedTrainer:Engine run resuming from iteration 0, epoch 0 until 5 epochs INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 1/157 -- train_loss: 1.8252 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 2/157 -- train_loss: 1.7637 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 3/157 -- train_loss: 1.7338 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 4/157 -- train_loss: 1.7239 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 5/157 -- train_loss: 1.6950 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 6/157 -- train_loss: 1.6414 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 7/157 -- train_loss: 1.6280 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 8/157 -- train_loss: 1.6190 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 9/157 -- train_loss: 1.5820 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 10/157 -- train_loss: 1.5858 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 11/157 -- train_loss: 1.5342 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 12/157 -- train_loss: 1.4917 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 13/157 -- train_loss: 1.4970 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 14/157 -- train_loss: 1.4540 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 15/157 -- train_loss: 1.4422 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 16/157 -- train_loss: 1.4415 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 17/157 -- train_loss: 1.4168 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 18/157 -- train_loss: 1.4373 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 19/157 -- train_loss: 1.3618 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 20/157 -- train_loss: 1.3188 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 21/157 -- train_loss: 1.3052 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 22/157 -- train_loss: 1.3105 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 23/157 -- train_loss: 1.2936 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 24/157 -- train_loss: 1.2629 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 25/157 -- train_loss: 1.2414 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 26/157 -- train_loss: 1.2488 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 27/157 -- train_loss: 1.1862 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 28/157 -- train_loss: 1.1873 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 29/157 -- train_loss: 1.1685 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 30/157 -- train_loss: 1.1491 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 31/157 -- train_loss: 1.1658 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 32/157 -- train_loss: 1.1465 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 33/157 -- train_loss: 1.1273 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 34/157 -- train_loss: 1.1121 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 35/157 -- train_loss: 1.1016 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 36/157 -- train_loss: 1.0937 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 37/157 -- train_loss: 1.0664 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 38/157 -- train_loss: 1.0472 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 39/157 -- train_loss: 0.9981 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 40/157 -- train_loss: 0.9918 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 41/157 -- train_loss: 0.9644 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 42/157 -- train_loss: 0.9875 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 43/157 -- train_loss: 0.9691 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 44/157 -- train_loss: 0.9285 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 45/157 -- train_loss: 0.9872 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 46/157 -- train_loss: 0.9350 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 47/157 -- train_loss: 0.8944 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 48/157 -- train_loss: 0.9034 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 49/157 -- train_loss: 0.8550 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 50/157 -- train_loss: 0.8690 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 51/157 -- train_loss: 0.8494 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 52/157 -- train_loss: 0.8600 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 53/157 -- train_loss: 0.7888 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 54/157 -- train_loss: 0.8017 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 55/157 -- train_loss: 0.8211 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 56/157 -- train_loss: 0.8050 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 57/157 -- train_loss: 0.7900 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 58/157 -- train_loss: 0.7971 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 59/157 -- train_loss: 0.7693 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 60/157 -- train_loss: 0.7425 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 61/157 -- train_loss: 0.7386 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 62/157 -- train_loss: 0.7208 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 63/157 -- train_loss: 0.7203 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 64/157 -- train_loss: 0.7036 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 65/157 -- train_loss: 0.6863 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 66/157 -- train_loss: 0.6678 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 67/157 -- train_loss: 0.6729 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 68/157 -- train_loss: 0.6435 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 69/157 -- train_loss: 0.6373 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 70/157 -- train_loss: 0.6616 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 71/157 -- train_loss: 0.6325 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 72/157 -- train_loss: 0.6357 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 73/157 -- train_loss: 0.6333 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 74/157 -- train_loss: 0.5840 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 75/157 -- train_loss: 0.6075 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 76/157 -- train_loss: 0.5632 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 77/157 -- train_loss: 0.5624 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 78/157 -- train_loss: 0.5698 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 79/157 -- train_loss: 0.5504 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 80/157 -- train_loss: 0.5456 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 81/157 -- train_loss: 0.5512 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 82/157 -- train_loss: 0.5329 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 83/157 -- train_loss: 0.5306 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 84/157 -- train_loss: 0.5312 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 85/157 -- train_loss: 0.5226 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 86/157 -- train_loss: 0.5166 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 87/157 -- train_loss: 0.5152 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 88/157 -- train_loss: 0.4779 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 89/157 -- train_loss: 0.5044 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 90/157 -- train_loss: 0.5009 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 91/157 -- train_loss: 0.4685 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 92/157 -- train_loss: 0.4510 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 93/157 -- train_loss: 0.4815 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 94/157 -- train_loss: 0.4358 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 95/157 -- train_loss: 0.4570 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 96/157 -- train_loss: 0.4495 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 97/157 -- train_loss: 0.4120 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 98/157 -- train_loss: 0.4414 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 99/157 -- train_loss: 0.4252 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 100/157 -- train_loss: 0.4197 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 101/157 -- train_loss: 0.4312 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 102/157 -- train_loss: 0.4120 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 103/157 -- train_loss: 0.4054 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 104/157 -- train_loss: 0.4341 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 105/157 -- train_loss: 0.3731 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 106/157 -- train_loss: 0.3637 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 107/157 -- train_loss: 0.3850 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 108/157 -- train_loss: 0.3813 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 109/157 -- train_loss: 0.3464 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 110/157 -- train_loss: 0.3810 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 111/157 -- train_loss: 0.3641 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 112/157 -- train_loss: 0.3643 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 113/157 -- train_loss: 0.3592 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 114/157 -- train_loss: 0.3635 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 115/157 -- train_loss: 0.3190 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 116/157 -- train_loss: 0.3745 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 117/157 -- train_loss: 0.3255 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 118/157 -- train_loss: 0.3126 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 119/157 -- train_loss: 0.3331 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 120/157 -- train_loss: 0.3332 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 121/157 -- train_loss: 0.3412 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 122/157 -- train_loss: 0.2994 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 123/157 -- train_loss: 0.3147 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 124/157 -- train_loss: 0.3394 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 125/157 -- train_loss: 0.3133 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 126/157 -- train_loss: 0.3016 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 127/157 -- train_loss: 0.2861 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 128/157 -- train_loss: 0.2973 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 129/157 -- train_loss: 0.2843 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 130/157 -- train_loss: 0.2936 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 131/157 -- train_loss: 0.2902 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 132/157 -- train_loss: 0.2588 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 133/157 -- train_loss: 0.2815 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 134/157 -- train_loss: 0.2529 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 135/157 -- train_loss: 0.2647 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 136/157 -- train_loss: 0.2909 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 137/157 -- train_loss: 0.2883 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 138/157 -- train_loss: 0.3009 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 139/157 -- train_loss: 0.2295 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 140/157 -- train_loss: 0.2391 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 141/157 -- train_loss: 0.2902 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 142/157 -- train_loss: 0.2217 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 143/157 -- train_loss: 0.2387 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 144/157 -- train_loss: 0.2454 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 145/157 -- train_loss: 0.2214 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 146/157 -- train_loss: 0.2477 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 147/157 -- train_loss: 0.2534 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 148/157 -- train_loss: 0.2287 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 149/157 -- train_loss: 0.2381 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 150/157 -- train_loss: 0.2240 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 151/157 -- train_loss: 0.2288 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 152/157 -- train_loss: 0.2337 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 153/157 -- train_loss: 0.2109 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 154/157 -- train_loss: 0.2132 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 155/157 -- train_loss: 0.1772 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 156/157 -- train_loss: 0.2177 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 157/157 -- train_loss: 0.1833 INFO:ignite.engine.engine.SupervisedTrainer:Got new best metric of train_acc: 0.8910450304605291 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[1] Metrics -- train_acc: 0.8910 INFO:ignite.engine.engine.SupervisedTrainer:Key metric: train_acc best value: 0.8910450304605291 at epoch: 1 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[1] Complete. Time taken: 00:00:22 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 1/157 -- train_loss: 0.2187 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 2/157 -- train_loss: 0.1953 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 3/157 -- train_loss: 0.2099 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 4/157 -- train_loss: 0.1911 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 5/157 -- train_loss: 0.1970 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 6/157 -- train_loss: 0.1962 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 7/157 -- train_loss: 0.1935 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 8/157 -- train_loss: 0.2117 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 9/157 -- train_loss: 0.1942 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 10/157 -- train_loss: 0.2168 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 11/157 -- train_loss: 0.1968 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 12/157 -- train_loss: 0.1700 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 13/157 -- train_loss: 0.2183 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 14/157 -- train_loss: 0.2052 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 15/157 -- train_loss: 0.1826 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 16/157 -- train_loss: 0.1790 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 17/157 -- train_loss: 0.2115 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 18/157 -- train_loss: 0.1955 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 19/157 -- train_loss: 0.1888 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 20/157 -- train_loss: 0.1706 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 21/157 -- train_loss: 0.1812 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 22/157 -- train_loss: 0.1542 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 23/157 -- train_loss: 0.1554 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 24/157 -- train_loss: 0.1728 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 25/157 -- train_loss: 0.1681 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 26/157 -- train_loss: 0.1579 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 27/157 -- train_loss: 0.1803 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 28/157 -- train_loss: 0.1819 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 29/157 -- train_loss: 0.1789 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 30/157 -- train_loss: 0.1591 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 31/157 -- train_loss: 0.1464 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 32/157 -- train_loss: 0.1687 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 33/157 -- train_loss: 0.1740 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 34/157 -- train_loss: 0.1438 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 35/157 -- train_loss: 0.1432 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 36/157 -- train_loss: 0.1790 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 37/157 -- train_loss: 0.1615 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 38/157 -- train_loss: 0.1457 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 39/157 -- train_loss: 0.1639 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 40/157 -- train_loss: 0.1887 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 41/157 -- train_loss: 0.1642 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 42/157 -- train_loss: 0.1487 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 43/157 -- train_loss: 0.1519 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 44/157 -- train_loss: 0.1506 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 45/157 -- train_loss: 0.1562 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 46/157 -- train_loss: 0.1436 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 47/157 -- train_loss: 0.1527 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 48/157 -- train_loss: 0.1184 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 49/157 -- train_loss: 0.1562 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 50/157 -- train_loss: 0.1344 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 51/157 -- train_loss: 0.1734 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 52/157 -- train_loss: 0.1416 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 53/157 -- train_loss: 0.1601 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 54/157 -- train_loss: 0.1302 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 55/157 -- train_loss: 0.1560 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 56/157 -- train_loss: 0.1393 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 57/157 -- train_loss: 0.1624 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 58/157 -- train_loss: 0.1227 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 59/157 -- train_loss: 0.1354 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 60/157 -- train_loss: 0.1270 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 61/157 -- train_loss: 0.1595 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 62/157 -- train_loss: 0.1158 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 63/157 -- train_loss: 0.1426 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 64/157 -- train_loss: 0.1158 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 65/157 -- train_loss: 0.1218 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 66/157 -- train_loss: 0.1408 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 67/157 -- train_loss: 0.1318 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 68/157 -- train_loss: 0.1169 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 69/157 -- train_loss: 0.1218 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 70/157 -- train_loss: 0.1448 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 71/157 -- train_loss: 0.1177 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 72/157 -- train_loss: 0.1105 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 73/157 -- train_loss: 0.1308 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 74/157 -- train_loss: 0.1065 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 75/157 -- train_loss: 0.1464 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 76/157 -- train_loss: 0.1296 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 77/157 -- train_loss: 0.1276 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 78/157 -- train_loss: 0.1239 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 79/157 -- train_loss: 0.1269 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 80/157 -- train_loss: 0.1382 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 81/157 -- train_loss: 0.1123 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 82/157 -- train_loss: 0.1223 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 83/157 -- train_loss: 0.1120 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 84/157 -- train_loss: 0.1225 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 85/157 -- train_loss: 0.1137 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 86/157 -- train_loss: 0.1216 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 87/157 -- train_loss: 0.1397 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 88/157 -- train_loss: 0.1087 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 89/157 -- train_loss: 0.1158 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 90/157 -- train_loss: 0.1171 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 91/157 -- train_loss: 0.1165 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 92/157 -- train_loss: 0.1073 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 93/157 -- train_loss: 0.1082 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 94/157 -- train_loss: 0.1055 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 95/157 -- train_loss: 0.1054 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 96/157 -- train_loss: 0.0869 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 97/157 -- train_loss: 0.1184 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 98/157 -- train_loss: 0.1490 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 99/157 -- train_loss: 0.1033 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 100/157 -- train_loss: 0.0991 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 101/157 -- train_loss: 0.1009 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 102/157 -- train_loss: 0.1224 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 103/157 -- train_loss: 0.1098 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 104/157 -- train_loss: 0.0965 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 105/157 -- train_loss: 0.1097 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 106/157 -- train_loss: 0.1287 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 107/157 -- train_loss: 0.1126 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 108/157 -- train_loss: 0.1096 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 109/157 -- train_loss: 0.1235 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 110/157 -- train_loss: 0.0929 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 111/157 -- train_loss: 0.0996 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 112/157 -- train_loss: 0.0982 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 113/157 -- train_loss: 0.0942 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 114/157 -- train_loss: 0.1076 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 115/157 -- train_loss: 0.0978 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 116/157 -- train_loss: 0.1017 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 117/157 -- train_loss: 0.0882 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 118/157 -- train_loss: 0.1133 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 119/157 -- train_loss: 0.0938 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 120/157 -- train_loss: 0.0902 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 121/157 -- train_loss: 0.1098 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 122/157 -- train_loss: 0.0852 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 123/157 -- train_loss: 0.0835 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 124/157 -- train_loss: 0.0857 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 125/157 -- train_loss: 0.0822 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 126/157 -- train_loss: 0.0956 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 127/157 -- train_loss: 0.0966 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 128/157 -- train_loss: 0.1105 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 129/157 -- train_loss: 0.0925 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 130/157 -- train_loss: 0.0875 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 131/157 -- train_loss: 0.0856 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 132/157 -- train_loss: 0.0854 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 133/157 -- train_loss: 0.0728 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 134/157 -- train_loss: 0.0928 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 135/157 -- train_loss: 0.0884 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 136/157 -- train_loss: 0.1160 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 137/157 -- train_loss: 0.0679 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 138/157 -- train_loss: 0.1008 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 139/157 -- train_loss: 0.0959 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 140/157 -- train_loss: 0.0868 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 141/157 -- train_loss: 0.0797 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 142/157 -- train_loss: 0.0869 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 143/157 -- train_loss: 0.0884 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 144/157 -- train_loss: 0.0675 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 145/157 -- train_loss: 0.0856 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 146/157 -- train_loss: 0.1024 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 147/157 -- train_loss: 0.0965 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 148/157 -- train_loss: 0.0864 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 149/157 -- train_loss: 0.0700 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 150/157 -- train_loss: 0.0698 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 151/157 -- train_loss: 0.0871 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 152/157 -- train_loss: 0.0680 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 153/157 -- train_loss: 0.0719 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 154/157 -- train_loss: 0.0706 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 155/157 -- train_loss: 0.0620 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 156/157 -- train_loss: 0.0814 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 157/157 -- train_loss: 0.0689 INFO:ignite.engine.engine.SupervisedTrainer:Got new best metric of train_acc: 0.9833425637967026 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[2] Metrics -- train_acc: 0.9833 INFO:ignite.engine.engine.SupervisedTrainer:Key metric: train_acc best value: 0.9833425637967026 at epoch: 2 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[2] Complete. Time taken: 00:00:23 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 1/157 -- train_loss: 0.0757 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 2/157 -- train_loss: 0.0663 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 3/157 -- train_loss: 0.0792 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 4/157 -- train_loss: 0.0586 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 5/157 -- train_loss: 0.0749 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 6/157 -- train_loss: 0.0906 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 7/157 -- train_loss: 0.0684 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 8/157 -- train_loss: 0.0677 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 9/157 -- train_loss: 0.0870 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 10/157 -- train_loss: 0.0835 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 11/157 -- train_loss: 0.0652 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 12/157 -- train_loss: 0.0805 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 13/157 -- train_loss: 0.0663 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 14/157 -- train_loss: 0.0744 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 15/157 -- train_loss: 0.0653 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 16/157 -- train_loss: 0.0717 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 17/157 -- train_loss: 0.0520 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 18/157 -- train_loss: 0.0743 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 19/157 -- train_loss: 0.0975 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 20/157 -- train_loss: 0.0981 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 21/157 -- train_loss: 0.0632 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 22/157 -- train_loss: 0.0904 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 23/157 -- train_loss: 0.0762 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 24/157 -- train_loss: 0.0700 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 25/157 -- train_loss: 0.0915 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 26/157 -- train_loss: 0.1116 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 27/157 -- train_loss: 0.0731 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 28/157 -- train_loss: 0.0694 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 29/157 -- train_loss: 0.0657 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 30/157 -- train_loss: 0.0826 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 31/157 -- train_loss: 0.0822 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 32/157 -- train_loss: 0.0687 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 33/157 -- train_loss: 0.0553 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 34/157 -- train_loss: 0.0591 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 35/157 -- train_loss: 0.0676 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 36/157 -- train_loss: 0.0588 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 37/157 -- train_loss: 0.0668 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 38/157 -- train_loss: 0.0736 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 39/157 -- train_loss: 0.0560 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 40/157 -- train_loss: 0.0557 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 41/157 -- train_loss: 0.0574 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 42/157 -- train_loss: 0.0946 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 43/157 -- train_loss: 0.0600 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 44/157 -- train_loss: 0.0700 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 45/157 -- train_loss: 0.0600 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 46/157 -- train_loss: 0.0573 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 47/157 -- train_loss: 0.0596 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 48/157 -- train_loss: 0.0803 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 49/157 -- train_loss: 0.0676 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 50/157 -- train_loss: 0.0669 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 51/157 -- train_loss: 0.0731 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 52/157 -- train_loss: 0.0735 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 53/157 -- train_loss: 0.0578 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 54/157 -- train_loss: 0.0675 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 55/157 -- train_loss: 0.0642 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 56/157 -- train_loss: 0.0646 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 57/157 -- train_loss: 0.0613 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 58/157 -- train_loss: 0.0472 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 59/157 -- train_loss: 0.0723 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 60/157 -- train_loss: 0.0661 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 61/157 -- train_loss: 0.0651 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 62/157 -- train_loss: 0.0490 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 63/157 -- train_loss: 0.0557 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 64/157 -- train_loss: 0.0714 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 65/157 -- train_loss: 0.0462 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 66/157 -- train_loss: 0.0475 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 67/157 -- train_loss: 0.0600 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 68/157 -- train_loss: 0.0479 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 69/157 -- train_loss: 0.0484 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 70/157 -- train_loss: 0.0610 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 71/157 -- train_loss: 0.0740 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 72/157 -- train_loss: 0.0574 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 73/157 -- train_loss: 0.0637 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 74/157 -- train_loss: 0.0448 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 75/157 -- train_loss: 0.0600 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 76/157 -- train_loss: 0.0663 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 77/157 -- train_loss: 0.0511 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 78/157 -- train_loss: 0.0641 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 79/157 -- train_loss: 0.0644 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 80/157 -- train_loss: 0.0644 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 81/157 -- train_loss: 0.0608 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 82/157 -- train_loss: 0.0516 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 83/157 -- train_loss: 0.0659 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 84/157 -- train_loss: 0.0429 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 85/157 -- train_loss: 0.0508 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 86/157 -- train_loss: 0.0713 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 87/157 -- train_loss: 0.0608 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 88/157 -- train_loss: 0.0534 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 89/157 -- train_loss: 0.0461 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 90/157 -- train_loss: 0.0522 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 91/157 -- train_loss: 0.0565 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 92/157 -- train_loss: 0.0464 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 93/157 -- train_loss: 0.0716 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 94/157 -- train_loss: 0.0456 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 95/157 -- train_loss: 0.0469 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 96/157 -- train_loss: 0.0645 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 97/157 -- train_loss: 0.0618 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 98/157 -- train_loss: 0.0767 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 99/157 -- train_loss: 0.0418 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 100/157 -- train_loss: 0.0559 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 101/157 -- train_loss: 0.0430 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 102/157 -- train_loss: 0.0373 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 103/157 -- train_loss: 0.0594 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 104/157 -- train_loss: 0.0677 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 105/157 -- train_loss: 0.0366 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 106/157 -- train_loss: 0.0457 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 107/157 -- train_loss: 0.0385 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 108/157 -- train_loss: 0.0568 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 109/157 -- train_loss: 0.0623 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 110/157 -- train_loss: 0.0580 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 111/157 -- train_loss: 0.0548 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 112/157 -- train_loss: 0.0379 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 113/157 -- train_loss: 0.0395 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 114/157 -- train_loss: 0.0444 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 115/157 -- train_loss: 0.0362 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 116/157 -- train_loss: 0.0416 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 117/157 -- train_loss: 0.0429 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 118/157 -- train_loss: 0.0332 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 119/157 -- train_loss: 0.0452 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 120/157 -- train_loss: 0.0638 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 121/157 -- train_loss: 0.0389 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 122/157 -- train_loss: 0.0532 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 123/157 -- train_loss: 0.0561 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 124/157 -- train_loss: 0.0622 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 125/157 -- train_loss: 0.0387 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 126/157 -- train_loss: 0.0620 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 127/157 -- train_loss: 0.0426 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 128/157 -- train_loss: 0.0361 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 129/157 -- train_loss: 0.0439 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 130/157 -- train_loss: 0.0472 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 131/157 -- train_loss: 0.0363 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 132/157 -- train_loss: 0.0461 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 133/157 -- train_loss: 0.0365 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 134/157 -- train_loss: 0.0414 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 135/157 -- train_loss: 0.0546 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 136/157 -- train_loss: 0.0396 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 137/157 -- train_loss: 0.0347 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 138/157 -- train_loss: 0.0411 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 139/157 -- train_loss: 0.0563 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 140/157 -- train_loss: 0.0579 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 141/157 -- train_loss: 0.0415 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 142/157 -- train_loss: 0.0435 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 143/157 -- train_loss: 0.0674 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 144/157 -- train_loss: 0.0372 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 145/157 -- train_loss: 0.0447 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 146/157 -- train_loss: 0.0427 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 147/157 -- train_loss: 0.0406 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 148/157 -- train_loss: 0.0423 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 149/157 -- train_loss: 0.0391 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 150/157 -- train_loss: 0.0345 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 151/157 -- train_loss: 0.0534 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 152/157 -- train_loss: 0.0304 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 153/157 -- train_loss: 0.0364 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 154/157 -- train_loss: 0.0612 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 155/157 -- train_loss: 0.0383 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 156/157 -- train_loss: 0.0389 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 157/157 -- train_loss: 0.0441 INFO:ignite.engine.engine.SupervisedTrainer:Got new best metric of train_acc: 0.9922464107698207 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[3] Metrics -- train_acc: 0.9922 INFO:ignite.engine.engine.SupervisedTrainer:Key metric: train_acc best value: 0.9922464107698207 at epoch: 3 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[3] Complete. Time taken: 00:00:23 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 1/157 -- train_loss: 0.0391 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 2/157 -- train_loss: 0.0445 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 3/157 -- train_loss: 0.0479 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 4/157 -- train_loss: 0.0516 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 5/157 -- train_loss: 0.0459 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 6/157 -- train_loss: 0.0306 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 7/157 -- train_loss: 0.0654 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 8/157 -- train_loss: 0.0536 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 9/157 -- train_loss: 0.0547 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 10/157 -- train_loss: 0.0461 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 11/157 -- train_loss: 0.0283 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 12/157 -- train_loss: 0.0377 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 13/157 -- train_loss: 0.0543 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 14/157 -- train_loss: 0.0476 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 15/157 -- train_loss: 0.0265 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 16/157 -- train_loss: 0.0320 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 17/157 -- train_loss: 0.0531 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 18/157 -- train_loss: 0.0324 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 19/157 -- train_loss: 0.0416 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 20/157 -- train_loss: 0.0318 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 21/157 -- train_loss: 0.0399 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 22/157 -- train_loss: 0.0348 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 23/157 -- train_loss: 0.0313 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 24/157 -- train_loss: 0.0442 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 25/157 -- train_loss: 0.0339 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 26/157 -- train_loss: 0.0383 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 27/157 -- train_loss: 0.0463 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 28/157 -- train_loss: 0.0483 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 29/157 -- train_loss: 0.0471 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 30/157 -- train_loss: 0.0477 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 31/157 -- train_loss: 0.0327 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 32/157 -- train_loss: 0.0444 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 33/157 -- train_loss: 0.0424 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 34/157 -- train_loss: 0.0361 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 35/157 -- train_loss: 0.0399 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 36/157 -- train_loss: 0.0337 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 37/157 -- train_loss: 0.0505 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 38/157 -- train_loss: 0.0297 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 39/157 -- train_loss: 0.0241 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 40/157 -- train_loss: 0.0298 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 41/157 -- train_loss: 0.0579 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 42/157 -- train_loss: 0.0244 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 43/157 -- train_loss: 0.0372 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 44/157 -- train_loss: 0.0493 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 45/157 -- train_loss: 0.0465 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 46/157 -- train_loss: 0.0396 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 47/157 -- train_loss: 0.0517 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 48/157 -- train_loss: 0.0388 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 49/157 -- train_loss: 0.0513 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 50/157 -- train_loss: 0.0411 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 51/157 -- train_loss: 0.0274 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 52/157 -- train_loss: 0.0413 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 53/157 -- train_loss: 0.0366 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 54/157 -- train_loss: 0.0321 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 55/157 -- train_loss: 0.0336 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 56/157 -- train_loss: 0.0469 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 57/157 -- train_loss: 0.0430 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 58/157 -- train_loss: 0.0381 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 59/157 -- train_loss: 0.0321 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 60/157 -- train_loss: 0.0414 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 61/157 -- train_loss: 0.0393 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 62/157 -- train_loss: 0.0410 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 63/157 -- train_loss: 0.0292 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 64/157 -- train_loss: 0.0327 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 65/157 -- train_loss: 0.0270 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 66/157 -- train_loss: 0.0320 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 67/157 -- train_loss: 0.0384 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 68/157 -- train_loss: 0.0270 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 69/157 -- train_loss: 0.0411 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 70/157 -- train_loss: 0.0295 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 71/157 -- train_loss: 0.0388 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 72/157 -- train_loss: 0.0290 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 73/157 -- train_loss: 0.0374 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 74/157 -- train_loss: 0.0511 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 75/157 -- train_loss: 0.0455 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 76/157 -- train_loss: 0.0369 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 77/157 -- train_loss: 0.0359 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 78/157 -- train_loss: 0.0416 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 79/157 -- train_loss: 0.0287 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 80/157 -- train_loss: 0.0357 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 81/157 -- train_loss: 0.0333 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 82/157 -- train_loss: 0.0307 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 83/157 -- train_loss: 0.0355 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 84/157 -- train_loss: 0.0364 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 85/157 -- train_loss: 0.0355 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 86/157 -- train_loss: 0.0336 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 87/157 -- train_loss: 0.0302 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 88/157 -- train_loss: 0.0312 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 89/157 -- train_loss: 0.0238 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 90/157 -- train_loss: 0.0235 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 91/157 -- train_loss: 0.0381 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 92/157 -- train_loss: 0.0218 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 93/157 -- train_loss: 0.0404 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 94/157 -- train_loss: 0.0338 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 95/157 -- train_loss: 0.0236 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 96/157 -- train_loss: 0.0356 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 97/157 -- train_loss: 0.0354 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 98/157 -- train_loss: 0.0321 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 99/157 -- train_loss: 0.0340 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 100/157 -- train_loss: 0.0281 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 101/157 -- train_loss: 0.0402 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 102/157 -- train_loss: 0.0439 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 103/157 -- train_loss: 0.0291 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 104/157 -- train_loss: 0.0262 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 105/157 -- train_loss: 0.0255 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 106/157 -- train_loss: 0.0483 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 107/157 -- train_loss: 0.0235 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 108/157 -- train_loss: 0.0342 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 109/157 -- train_loss: 0.0294 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 110/157 -- train_loss: 0.0316 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 111/157 -- train_loss: 0.0318 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 112/157 -- train_loss: 0.0307 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 113/157 -- train_loss: 0.0303 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 114/157 -- train_loss: 0.0298 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 115/157 -- train_loss: 0.0254 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 116/157 -- train_loss: 0.0207 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 117/157 -- train_loss: 0.0212 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 118/157 -- train_loss: 0.0233 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 119/157 -- train_loss: 0.0296 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 120/157 -- train_loss: 0.0286 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 121/157 -- train_loss: 0.0424 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 122/157 -- train_loss: 0.0263 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 123/157 -- train_loss: 0.0291 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 124/157 -- train_loss: 0.0218 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 125/157 -- train_loss: 0.0293 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 126/157 -- train_loss: 0.0395 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 127/157 -- train_loss: 0.0274 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 128/157 -- train_loss: 0.0302 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 129/157 -- train_loss: 0.0240 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 130/157 -- train_loss: 0.0281 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 131/157 -- train_loss: 0.0172 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 132/157 -- train_loss: 0.0321 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 133/157 -- train_loss: 0.0272 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 134/157 -- train_loss: 0.0163 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 135/157 -- train_loss: 0.0419 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 136/157 -- train_loss: 0.0296 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 137/157 -- train_loss: 0.0374 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 138/157 -- train_loss: 0.0451 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 139/157 -- train_loss: 0.0190 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 140/157 -- train_loss: 0.0241 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 141/157 -- train_loss: 0.0219 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 142/157 -- train_loss: 0.0354 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 143/157 -- train_loss: 0.0297 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 144/157 -- train_loss: 0.0222 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 145/157 -- train_loss: 0.0282 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 146/157 -- train_loss: 0.0288 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 147/157 -- train_loss: 0.0236 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 148/157 -- train_loss: 0.0301 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 149/157 -- train_loss: 0.0240 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 150/157 -- train_loss: 0.0291 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 151/157 -- train_loss: 0.0379 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 152/157 -- train_loss: 0.0239 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 153/157 -- train_loss: 0.0225 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 154/157 -- train_loss: 0.0254 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 155/157 -- train_loss: 0.0185 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 156/157 -- train_loss: 0.0252 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 157/157 -- train_loss: 0.0317 INFO:ignite.engine.engine.SupervisedTrainer:Got new best metric of train_acc: 0.9951433561964811 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[4] Metrics -- train_acc: 0.9951 INFO:ignite.engine.engine.SupervisedTrainer:Key metric: train_acc best value: 0.9951433561964811 at epoch: 4 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[4] Complete. Time taken: 00:00:23 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 1/157 -- train_loss: 0.0224 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 2/157 -- train_loss: 0.0260 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 3/157 -- train_loss: 0.0490 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 4/157 -- train_loss: 0.0172 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 5/157 -- train_loss: 0.0202 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 6/157 -- train_loss: 0.0324 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 7/157 -- train_loss: 0.0305 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 8/157 -- train_loss: 0.0186 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 9/157 -- train_loss: 0.0262 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 10/157 -- train_loss: 0.0171 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 11/157 -- train_loss: 0.0207 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 12/157 -- train_loss: 0.0129 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 13/157 -- train_loss: 0.0281 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 14/157 -- train_loss: 0.0236 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 15/157 -- train_loss: 0.0233 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 16/157 -- train_loss: 0.0265 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 17/157 -- train_loss: 0.0257 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 18/157 -- train_loss: 0.0246 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 19/157 -- train_loss: 0.0229 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 20/157 -- train_loss: 0.0194 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 21/157 -- train_loss: 0.0141 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 22/157 -- train_loss: 0.0279 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 23/157 -- train_loss: 0.0380 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 24/157 -- train_loss: 0.0220 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 25/157 -- train_loss: 0.0164 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 26/157 -- train_loss: 0.0230 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 27/157 -- train_loss: 0.0185 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 28/157 -- train_loss: 0.0323 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 29/157 -- train_loss: 0.0307 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 30/157 -- train_loss: 0.0435 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 31/157 -- train_loss: 0.0355 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 32/157 -- train_loss: 0.0288 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 33/157 -- train_loss: 0.0358 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 34/157 -- train_loss: 0.0286 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 35/157 -- train_loss: 0.0164 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 36/157 -- train_loss: 0.0358 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 37/157 -- train_loss: 0.0260 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 38/157 -- train_loss: 0.0189 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 39/157 -- train_loss: 0.0420 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 40/157 -- train_loss: 0.0239 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 41/157 -- train_loss: 0.0363 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 42/157 -- train_loss: 0.0280 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 43/157 -- train_loss: 0.0262 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 44/157 -- train_loss: 0.0288 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 45/157 -- train_loss: 0.0512 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 46/157 -- train_loss: 0.0249 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 47/157 -- train_loss: 0.0244 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 48/157 -- train_loss: 0.0218 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 49/157 -- train_loss: 0.0194 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 50/157 -- train_loss: 0.0209 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 51/157 -- train_loss: 0.0280 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 52/157 -- train_loss: 0.0287 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 53/157 -- train_loss: 0.0250 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 54/157 -- train_loss: 0.0183 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 55/157 -- train_loss: 0.0156 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 56/157 -- train_loss: 0.0171 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 57/157 -- train_loss: 0.0230 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 58/157 -- train_loss: 0.0166 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 59/157 -- train_loss: 0.0224 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 60/157 -- train_loss: 0.0192 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 61/157 -- train_loss: 0.0333 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 62/157 -- train_loss: 0.0315 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 63/157 -- train_loss: 0.0229 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 64/157 -- train_loss: 0.0155 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 65/157 -- train_loss: 0.0307 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 66/157 -- train_loss: 0.0289 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 67/157 -- train_loss: 0.0233 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 68/157 -- train_loss: 0.0293 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 69/157 -- train_loss: 0.0194 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 70/157 -- train_loss: 0.0206 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 71/157 -- train_loss: 0.0259 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 72/157 -- train_loss: 0.0241 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 73/157 -- train_loss: 0.0561 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 74/157 -- train_loss: 0.0294 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 75/157 -- train_loss: 0.0353 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 76/157 -- train_loss: 0.0126 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 77/157 -- train_loss: 0.0128 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 78/157 -- train_loss: 0.0187 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 79/157 -- train_loss: 0.0310 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 80/157 -- train_loss: 0.0166 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 81/157 -- train_loss: 0.0184 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 82/157 -- train_loss: 0.0288 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 83/157 -- train_loss: 0.0189 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 84/157 -- train_loss: 0.0292 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 85/157 -- train_loss: 0.0161 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 86/157 -- train_loss: 0.0167 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 87/157 -- train_loss: 0.0239 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 88/157 -- train_loss: 0.0185 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 89/157 -- train_loss: 0.0161 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 90/157 -- train_loss: 0.0214 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 91/157 -- train_loss: 0.0251 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 92/157 -- train_loss: 0.0332 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 93/157 -- train_loss: 0.0324 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 94/157 -- train_loss: 0.0172 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 95/157 -- train_loss: 0.0178 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 96/157 -- train_loss: 0.0195 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 97/157 -- train_loss: 0.0441 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 98/157 -- train_loss: 0.0202 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 99/157 -- train_loss: 0.0158 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 100/157 -- train_loss: 0.0134 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 101/157 -- train_loss: 0.0196 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 102/157 -- train_loss: 0.0286 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 103/157 -- train_loss: 0.0247 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 104/157 -- train_loss: 0.0245 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 105/157 -- train_loss: 0.0146 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 106/157 -- train_loss: 0.0165 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 107/157 -- train_loss: 0.0156 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 108/157 -- train_loss: 0.0221 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 109/157 -- train_loss: 0.0239 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 110/157 -- train_loss: 0.0198 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 111/157 -- train_loss: 0.0227 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 112/157 -- train_loss: 0.0125 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 113/157 -- train_loss: 0.0236 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 114/157 -- train_loss: 0.0132 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 115/157 -- train_loss: 0.0209 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 116/157 -- train_loss: 0.0190 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 117/157 -- train_loss: 0.0330 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 118/157 -- train_loss: 0.0175 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 119/157 -- train_loss: 0.0316 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 120/157 -- train_loss: 0.0182 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 121/157 -- train_loss: 0.0197 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 122/157 -- train_loss: 0.0214 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 123/157 -- train_loss: 0.0177 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 124/157 -- train_loss: 0.0173 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 125/157 -- train_loss: 0.0298 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 126/157 -- train_loss: 0.0126 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 127/157 -- train_loss: 0.0141 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 128/157 -- train_loss: 0.0275 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 129/157 -- train_loss: 0.0158 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 130/157 -- train_loss: 0.0203 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 131/157 -- train_loss: 0.0123 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 132/157 -- train_loss: 0.0263 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 133/157 -- train_loss: 0.0141 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 134/157 -- train_loss: 0.0117 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 135/157 -- train_loss: 0.0192 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 136/157 -- train_loss: 0.0240 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 137/157 -- train_loss: 0.0187 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 138/157 -- train_loss: 0.0209 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 139/157 -- train_loss: 0.0224 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 140/157 -- train_loss: 0.0154 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 141/157 -- train_loss: 0.0277 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 142/157 -- train_loss: 0.0274 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 143/157 -- train_loss: 0.0243 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 144/157 -- train_loss: 0.0132 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 145/157 -- train_loss: 0.0158 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 146/157 -- train_loss: 0.0138 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 147/157 -- train_loss: 0.0167 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 148/157 -- train_loss: 0.0135 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 149/157 -- train_loss: 0.0240 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 150/157 -- train_loss: 0.0194 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 151/157 -- train_loss: 0.0225 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 152/157 -- train_loss: 0.0229 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 153/157 -- train_loss: 0.0226 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 154/157 -- train_loss: 0.0224 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 155/157 -- train_loss: 0.0127 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 156/157 -- train_loss: 0.0165 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 157/157 -- train_loss: 0.0191 INFO:ignite.engine.engine.SupervisedTrainer:Got new best metric of train_acc: 0.9965918289098112 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[5] Metrics -- train_acc: 0.9966 INFO:ignite.engine.engine.SupervisedTrainer:Key metric: train_acc best value: 0.9965918289098112 at epoch: 5 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[5] Complete. Time taken: 00:00:22 INFO:ignite.engine.engine.SupervisedTrainer:Engine run complete. Time taken: 00:01:54
DecathlonDataset による訓練実験とワークフロー
Decathlon データセットは Medical Segmentation Decathlon AI challenge に由来します。
前処理変換のセットアップ
transform = Compose(
[
LoadImaged(keys=["image", "label"]),
AddChanneld(keys=["image", "label"]),
Spacingd(keys=["image", "label"], pixdim=(
1.0, 1.0, 1.0), mode=("bilinear", "nearest")),
Orientationd(keys=["image", "label"], axcodes="RAS"),
ScaleIntensityd(keys="image"),
Resized(keys=["image", "label"], spatial_size=(
32, 64, 32), mode=("trilinear", "nearest")),
EnsureTyped(keys=["image", "label"]),
]
)
訓練のために DeccathlonDataset を作成する
DecathlonDataset は MONAI CacheDataset から継承して期待する動作を実現するために豊富なパラメータを提供しています :
- root_dir: MSD データセットをキャッシュしてロードするためのユーザのローカルディレクトリ。
- task: どのタスクをダウンロードして実行するか : 次のリストの一つです (“Task01_BrainTumour”, “Task02_Heart”, “Task03_Liver”, “Task04_Hippocampus”, “Task05_Prostate”, “Task06_Lung”, “Task07_Pancreas”, “Task08_HepaticVessel”, “Task09_Spleen”, “Task10_Colon”).
- section: 想定されるデータ・セクション、以下のいずれかです : training, validation or test.
- transform: 入力データ上で操作を実行する変換。デフォルト変換は LoadPNGd と AddChanneld から成り、これはデータを [C, H, W, D] shape を持つ numpy 配列にロードできます。
- download: リソース・リンクから Decathlon をダウンロードして抽出するか否か、デフォルトは False です。想定されるファイルが既に存在していれば、それが True に設定されていてもダウンロードはスキップします。ユーザは tar ファイルか dataset フォルダをルートディレクトリに手動でコピーできます。
- seed: 訓練、検証とテストデータセットをランダムに分割するためのランダムシード、デフォルトは 0 です。
- val_frac: 訓練セクションからの検証比率のパーセンテージで、デフォルトは 0.2 です。Decathlon データはラベルを持つ訓練セクションとラベルなしのテストセクションだけを含みますので、検証セクションとして訓練セクションからランダムに一部を選択します。
- cache_num: キャッシュされる項目の数。デフォルトは sys.maxsize です。(cache_num, data_length x cache_rate, data_length) の最小値を取ります。
- cache_rate: トータルでキャッシュされるデータのパーセンテージで、デフォルトは 1.0 (cache all) です。(cache_num, data_length x cache_rate, data_length) の最小値を取ります。
- num_workers: 使用するワーカースレッドの数です。0 であればシングルスレッドが使用されます。デフォルトは 0 です。
“tar フアイル” は最初にダウンロードした後でキャッシュされることに注意してください。self.__getitem__() API は、self.__len__() 内の指定されたインデックスに従って 1 {“image”: XXX, “label”: XXX} 辞書を生成します。
train_ds = DecathlonDataset(
root_dir=root_dir,
task="Task04_Hippocampus",
transform=transform,
section="training",
download=True,
)
# the dataset can work seamlessly with the pytorch native dataset loader,
# but using monai.data.DataLoader has additional benefits of mutli-process
# random seeds handling, and the customized collate functions
train_loader = DataLoader(train_ds, batch_size=32,
shuffle=True, num_workers=16)
Task04_Hippocampus.tar: 100%|██████████| 27.1M/27.1M [01:15<00:00, 377kB/s] Verified 'Task04_Hippocampus.tar.part', md5: 9d24dba78a72977dbd1d2e110310f31b. Verified 'Task04_Hippocampus.tar', md5: 9d24dba78a72977dbd1d2e110310f31b. Verified 'Task04_Hippocampus.tar', md5: 9d24dba78a72977dbd1d2e110310f31b. 0%| | 0/208 [00:00<?, ?it/s]Default upsampling behavior when mode=trilinear is changed to align_corners=False since 0.4.0. Please specify align_corners=True if the old behavior is desired. See the documentation of nn.Upsample for details. 100%|██████████| 208/208 [00:34<00:00, 6.03it/s]
可視化して確認するために DecathlonDataset から画像をピックアップする
plt.subplots(3, 3, figsize=(8, 8))
for i in range(9):
plt.subplot(3, 3, i + 1)
plt.imshow(train_ds[i * 20]["image"]
[0, :, :, 10].detach().cpu(), cmap="gray")
plt.tight_layout()
plt.show()
訓練コンポーネントの作成
device = torch.device("cuda:0")
net = UNet(
spatial_dims=3,
in_channels=1,
out_channels=3,
channels=(16, 32, 64, 128, 256),
strides=(2, 2, 2, 2),
num_res_units=2,
norm=Norm.BATCH,
).to(device)
loss = DiceLoss(to_onehot_y=True, softmax=True)
opt = torch.optim.Adam(net.parameters(), 1e-2)
最も簡単な訓練ワークフローを定義して実行する
訓練ワークフローを素早くセットアップするために MONAI SupervisedTrainer ハンドラを使用します。
trainer = SupervisedTrainer(
device=device,
max_epochs=5,
train_data_loader=train_loader,
network=net,
optimizer=opt,
loss_function=loss,
inferer=SimpleInferer(),
postprocessing=AsDiscreted(
keys=["pred", "label"], argmax=(True, False),
to_onehot=True, num_classes=3,
),
key_train_metric={
"train_meandice": MeanDice(
output_transform=from_engine(["pred", "label"]))
},
train_handlers=StatsHandler(
tag_name="train_loss", output_transform=from_engine(["loss"], first=True)),
)
trainer.run()
INFO:ignite.engine.engine.SupervisedTrainer:Engine run resuming from iteration 0, epoch 0 until 5 epochs INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 1/7 -- train_loss: 0.7959 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 2/7 -- train_loss: 0.7505 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 3/7 -- train_loss: 0.7073 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 4/7 -- train_loss: 0.6903 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 5/7 -- train_loss: 0.6436 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 6/7 -- train_loss: 0.5983 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 1/5, Iter: 7/7 -- train_loss: 0.5609 INFO:ignite.engine.engine.SupervisedTrainer:Got new best metric of train_meandice: 0.3706986674895653 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[1] Metrics -- train_meandice: 0.3707 INFO:ignite.engine.engine.SupervisedTrainer:Key metric: train_meandice best value: 0.3706986674895653 at epoch: 1 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[1] Complete. Time taken: 00:00:01 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 1/7 -- train_loss: 0.5261 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 2/7 -- train_loss: 0.5211 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 3/7 -- train_loss: 0.5054 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 4/7 -- train_loss: 0.4820 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 5/7 -- train_loss: 0.4693 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 6/7 -- train_loss: 0.4541 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 2/5, Iter: 7/7 -- train_loss: 0.4388 INFO:ignite.engine.engine.SupervisedTrainer:Got new best metric of train_meandice: 0.5458252521661612 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[2] Metrics -- train_meandice: 0.5458 INFO:ignite.engine.engine.SupervisedTrainer:Key metric: train_meandice best value: 0.5458252521661612 at epoch: 2 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[2] Complete. Time taken: 00:00:03 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 1/7 -- train_loss: 0.4413 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 2/7 -- train_loss: 0.4305 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 3/7 -- train_loss: 0.4200 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 4/7 -- train_loss: 0.4210 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 5/7 -- train_loss: 0.4044 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 6/7 -- train_loss: 0.3963 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 3/5, Iter: 7/7 -- train_loss: 0.3781 INFO:ignite.engine.engine.SupervisedTrainer:Got new best metric of train_meandice: 0.5662448773017297 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[3] Metrics -- train_meandice: 0.5662 INFO:ignite.engine.engine.SupervisedTrainer:Key metric: train_meandice best value: 0.5662448773017297 at epoch: 3 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[3] Complete. Time taken: 00:00:03 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 1/7 -- train_loss: 0.3717 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 2/7 -- train_loss: 0.3557 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 3/7 -- train_loss: 0.3526 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 4/7 -- train_loss: 0.3365 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 5/7 -- train_loss: 0.3163 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 6/7 -- train_loss: 0.3026 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 4/5, Iter: 7/7 -- train_loss: 0.2737 INFO:ignite.engine.engine.SupervisedTrainer:Got new best metric of train_meandice: 0.6949328046578628 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[4] Metrics -- train_meandice: 0.6949 INFO:ignite.engine.engine.SupervisedTrainer:Key metric: train_meandice best value: 0.6949328046578628 at epoch: 4 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[4] Complete. Time taken: 00:00:03 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 1/7 -- train_loss: 0.2623 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 2/7 -- train_loss: 0.2445 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 3/7 -- train_loss: 0.2299 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 4/7 -- train_loss: 0.2190 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 5/7 -- train_loss: 0.2207 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 6/7 -- train_loss: 0.2051 INFO:ignite.engine.engine.SupervisedTrainer:Epoch: 5/5, Iter: 7/7 -- train_loss: 0.1910 INFO:ignite.engine.engine.SupervisedTrainer:Got new best metric of train_meandice: 0.7996270152238699 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[5] Metrics -- train_meandice: 0.7996 INFO:ignite.engine.engine.SupervisedTrainer:Key metric: train_meandice best value: 0.7996270152238699 at epoch: 5 INFO:ignite.engine.engine.SupervisedTrainer:Epoch[5] Complete. Time taken: 00:00:03 INFO:ignite.engine.engine.SupervisedTrainer:Engine run complete. Time taken: 00:00:15
他の公開データを共有して MONAI にデータセットを追加する
MedNISTDataset or DecathlonDataset を参考にして、他の公開データのための新しいデータセットを作成することは簡単です。主として以下のステップを含みます :
- MONAI CacheDataset を継承して、訓練を高速化するためにキャッシング機構を利用します。
- データセットのライセンスが公開アクセスと共有を許容していることを確認してください。
- データセットでデータをダウンロードして抽出するために monai.apps.download_and_extract を使用します。
- 訓練、検証とテストセクションをランダムに分割するロジックを定義します。
- 辞書項目でデータリストを構築します :
[ {'image': image1_path, 'label': label1_path}, {'image': image2_path, 'label': label2_path}, {'image': image3_path, 'label': label3_path}, ... ... ] - データセット固有のロジックを定義します。
例として IXIDataset を定義する
ここでは新しい IXIDataset を作成する方法を示すために例として IXI データセット を使用します。
class IXIDataset(Randomizable, CacheDataset):
resource = "http://biomedic.doc.ic.ac.uk/" \
+ "brain-development/downloads/IXI/IXI-T1.tar"
md5 = "34901a0593b41dd19c1a1f746eac2d58"
def __init__(
self,
root_dir,
section,
transform,
download=False,
seed=0,
val_frac=0.2,
test_frac=0.2,
cache_num=sys.maxsize,
cache_rate=1.0,
num_workers=0,
):
if not os.path.isdir(root_dir):
raise ValueError("Root directory root_dir must be a directory.")
self.section = section
self.val_frac = val_frac
self.test_frac = test_frac
self.set_random_state(seed=seed)
dataset_dir = os.path.join(root_dir, "ixi")
tarfile_name = f"{dataset_dir}.tar"
if download:
download_and_extract(
self.resource, tarfile_name, dataset_dir, self.md5)
# as a quick demo, we just use 10 images to show
self.datalist = [
{"image": os.path.join(
dataset_dir, "IXI314-IOP-0889-T1.nii.gz"), "label": 0},
{"image": os.path.join(
dataset_dir, "IXI249-Guys-1072-T1.nii.gz"), "label": 0},
{"image": os.path.join(
dataset_dir, "IXI609-HH-2600-T1.nii.gz"), "label": 0},
{"image": os.path.join(
dataset_dir, "IXI173-HH-1590-T1.nii.gz"), "label": 1},
{"image": os.path.join(
dataset_dir, "IXI020-Guys-0700-T1.nii.gz"), "label": 0},
{"image": os.path.join(
dataset_dir, "IXI342-Guys-0909-T1.nii.gz"), "label": 0},
{"image": os.path.join(
dataset_dir, "IXI134-Guys-0780-T1.nii.gz"), "label": 0},
{"image": os.path.join(
dataset_dir, "IXI577-HH-2661-T1.nii.gz"), "label": 1},
{"image": os.path.join(
dataset_dir, "IXI066-Guys-0731-T1.nii.gz"), "label": 1},
{"image": os.path.join(
dataset_dir, "IXI130-HH-1528-T1.nii.gz"), "label": 0},
]
data = self._generate_data_list()
super().__init__(
data, transform, cache_num=cache_num,
cache_rate=cache_rate, num_workers=num_workers,
)
def randomize(self, data=None):
self.rann = self.R.random()
def _generate_data_list(self):
data = []
for d in self.datalist:
self.randomize()
if self.section == "training":
if self.rann < self.val_frac + self.test_frac:
continue
elif self.section == "validation":
if self.rann >= self.val_frac:
continue
elif self.section == "test":
if self.rann < self.val_frac or \
self.rann >= self.val_frac + self.test_frac:
continue
else:
raise ValueError(
f"Unsupported section: {self.section}, "
"available options are ['training', 'validation', 'test']."
)
data.append(d)
return data
可視化して確認するために IXIDataset から画像をピックアップする
train_ds = IXIDataset(
root_dir=root_dir,
section="training",
transform=Compose([LoadImaged("image"), EnsureTyped("image")]),
download=True,
)
plt.figure("check", (18, 6))
for i in range(3):
plt.subplot(1, 3, i + 1)
plt.imshow(train_ds[i]["image"][:, :, 80].detach().cpu(), cmap="gray")
plt.show()
Verified 'ixi.tar', md5: 34901a0593b41dd19c1a1f746eac2d58. file /workspace/data/medical/ixi.tar exists, skip downloading. Verified 'ixi.tar', md5: 34901a0593b41dd19c1a1f746eac2d58. 100%|██████████| 9/9 [00:02<00:00, 3.51it/s]
データディレクトリのクリーンアップ
一時ディレクトリが使用された場合にはディレクトリを削除します。
if directory is None:
shutil.rmtree(root_dir)
以上
