Repository: spytensor/pytorch_img_classification_for_competition
Branch: master
Commit: 75cdbf998216
Files: 23
Total size: 73.3 KB
Directory structure:
gitextract_vee5ftx8/
├── .gitignore
├── LICENSE
├── README.md
├── config.py
├── ensemble.py
├── main.py
├── models/
│ ├── __init__.py
│ └── model.py
├── test.py
└── utils/
├── __init__.py
├── logger.py
├── losses/
│ ├── __init__.py
│ ├── focalloss.py
│ └── label_smoothing.py
├── misc.py
├── optimizers/
│ ├── __init__.py
│ ├── lookahead.py
│ ├── novograd.py
│ ├── over9000.py
│ ├── radam.py
│ ├── ralamb.py
│ └── ranger.py
└── reader.py
================================================
FILE CONTENTS
================================================
================================================
FILE: .gitignore
================================================
# Byte-compiled / optimized / DLL files
__pycache__/
*.py[cod]
*$py.class
# C extensions
*.so
# Distribution / packaging
.Python
build/
develop-eggs/
dist/
downloads/
eggs/
.eggs/
lib/
lib64/
parts/
sdist/
var/
wheels/
pip-wheel-metadata/
share/python-wheels/
*.egg-info/
.installed.cfg
*.egg
MANIFEST
# PyInstaller
# Usually these files are written by a python script from a template
# before PyInstaller builds the exe, so as to inject date/other infos into it.
*.manifest
*.spec
# Installer logs
pip-log.txt
pip-delete-this-directory.txt
# Unit test / coverage reports
htmlcov/
.tox/
.nox/
.coverage
.coverage.*
.cache
nosetests.xml
coverage.xml
*.cover
*.py,cover
.hypothesis/
.pytest_cache/
# Translations
*.mo
*.pot
# Django stuff:
*.log
local_settings.py
db.sqlite3
db.sqlite3-journal
# Flask stuff:
instance/
.webassets-cache
# Scrapy stuff:
.scrapy
# Sphinx documentation
docs/_build/
# PyBuilder
target/
# Jupyter Notebook
.ipynb_checkpoints
# IPython
profile_default/
ipython_config.py
# pyenv
.python-version
# pipenv
# According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
# However, in case of collaboration, if having platform-specific dependencies or dependencies
# having no cross-platform support, pipenv may install dependencies that don't work, or not
# install all needed dependencies.
#Pipfile.lock
# PEP 582; used by e.g. github.com/David-OConnor/pyflow
__pypackages__/
# Celery stuff
celerybeat-schedule
celerybeat.pid
# SageMath parsed files
*.sage.py
# Environments
.env
.venv
env/
venv/
ENV/
env.bak/
venv.bak/
# Spyder project settings
.spyderproject
.spyproject
# Rope project settings
.ropeproject
# mkdocs documentation
/site
# mypy
.mypy_cache/
.dmypy.json
dmypy.json
# Pyre type checker
.pyre/
================================================
FILE: LICENSE
================================================
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
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.
================================================
FILE: README.md
================================================
### pytorch 图像分类竞赛框架
### 1. 更新日志
- (2020年5月2日) 基础版本上线
### 2. 依赖库
- pretrainedmodels
- progress
- efficientnet-pytorch
- apex
### 3. 支持功能
- [x] pytorch官网模型
- [x] [pretrained-models.pytorch](https://github.com/Cadene/pretrained-models.pytorch) 复现的部分模型
- [x] [EfficientNet-PyTorch](https://github.com/lukemelas/EfficientNet-PyTorch)
- [x] fp16混合精度训练
- [x] TTA
- [x] 固定验证集/随机划分验证集
- [x] 多种优化器：adam、radam、novograd、sgd、ranger、ralamb、over9000、lookahead、lamb
- [x] OneCycle训练策略
- [x] LabelSmoothLoss
- [x] Focal Loss
- [ ] AotuAgument
### 4. 使用方法
更改`config.py`中的参数，训练执行 `python main.py`，预测执行`python test.py`
### 5. submit_example.csv
每一行：filename,label
样例：
```
0001.jpg,dog
0002.jpg,dog
0003.jpg,dog
```
注：预测图像可能没有label，所以label可以随意给个临时的，但一些比赛平台对都会给个提交样例，随意给个label
### 6.TODO
- [ ] 优化模型融合策略
- [ ] 优化online数据增强
- [ ] 优化pytorch官方模型调用接口
- [ ] 增加模型全连接层初始化
- [ ] 增加更多学习率衰减策略
- [ ] 增加find lr
- [ ] 增加dali
- [ ] 增加wsl模型
- [ ] 增加tensorboardX
- [ ] 优化文件夹创建
================================================
FILE: config.py
================================================
class DefaultConfigs(object):
# set default configs, if you don't understand, don't modify
seed = 666 # set random seed
workers = 4 # set number of data loading workers (default: 4)
beta1 = 0.9 # adam parameters beta1
beta2 = 0.999 # adam parameters beta2
mom = 0.9 # momentum parameters
wd = 1e-4 # weight-decay
resume = None # path to latest checkpoint (default: none),should endswith ".pth" or ".tar" if used
evaluate = False # just do evaluate
start_epoch = 0 # deault start epoch is zero,if use resume change it
split_online = False # split dataset to train and val online or offline
# set changeable configs, you can change one during your experiment
dataset = "/dataset/df/cloud/data/dataset/" # dataset folder with train and val
test_folder = "/dataset/df/cloud/data/test/" # test images' folder
submit_example = "/dataset/df/cloud/data/submit_example.csv" # submit example file
checkpoints = "./checkpoints/" # path to save checkpoints
log_dir = "./logs/" # path to save log files
submits = "./submits/" # path to save submission files
bs = 32 # batch size
lr = 2e-3 # learning rate
epochs = 40 # train epochs
input_size = 512 # model input size or image resied
num_classes = 9 # num of classes
gpu_id = "0" # default gpu id
model_name = "se_resnext50_32x4d-model-sgd-512" # model name to use
optim = "sgd" # "adam","radam","novograd",sgd","ranger","ralamb","over9000","lookahead","lamb"
fp16 = True # use float16 to train the model
opt_level = "O1" # if use fp16, "O0" means fp32，"O1" means mixed，"O2" means except BN，"O3" means only fp16
keep_batchnorm_fp32 = False # if use fp16,keep BN layer as fp32
loss_func = "CrossEntropy" # "CrossEntropy"、"FocalLoss"、"LabelSmoothCE"
lr_scheduler = "step" # lr scheduler method,"adjust","on_loss","on_acc","step"
configs = DefaultConfigs()
================================================
FILE: ensemble.py
================================================
import pandas as pd
import numpy as np
import os
from IPython import embed
file1 = pd.read_csv("./csvs/efficientnet-b3-model_512-_adam_aug_confidence.csv",header=None)
file2 = pd.read_csv("./csvs/efficientnet-b5-model_456_ranger_aug_confidence.csv",header=None)
file3 = pd.read_csv("./csvs/efficientnet-b4-model_380_ranger_aug_confidence.csv",header=None)
filenames,labels = [],[]
# embed()
# for (filename1,label1),(filename2,label2),(filename3,label3),(filename4,label4),(filename5,label5) in zip(file1.values,file2.values,file3.values,file4.values,file5.values):
for (filename1,label1) ,(filename2,label2),(filename3,label3) in zip(file1.values,file2.values,file3.values):
filename = filename1
filenames.append(filename)
#embed()
label1 = np.array(list(map(float,label1.split("-"))))
label2 = np.array(list(map(float,label2.split("-"))))
label3 = np.array(list(map(float,label3.split("-"))))
# label4 = np.array(list(map(float,label4.split("[")[1].split("]")[0].split(","))))
# label5 = np.array(list(map(float,label5.split("[")[1].split("]")[0].split(","))))
label = np.argmax((label1 + label2 + label3) / 3.0) + 1
labels.append(label)
submission = pd.DataFrame({'FileName': filenames, 'type': labels})
submission.to_csv("./ensemble_efficientnets.csv", header=None, index=False)
================================================
FILE: main.py
================================================
import random
import time
import warnings
import torch.nn as nn
import torch.nn.parallel
import torch.backends.cudnn as cudnn
import torch.optim
import torch.utils.data
import torch.utils.data.distributed
import torchvision.transforms as transforms
import torchvision.datasets as datasets
import numpy as np
from PIL import ImageFile
from config import configs
from models.model import get_model
from sklearn.model_selection import train_test_split
from utils.misc import *
from utils.logger import *
from utils.losses import *
from progress.bar import Bar
from utils.reader import WeatherDataset
# for train fp16
if configs.fp16:
try:
import apex
from apex.parallel import DistributedDataParallel as DDP
from apex.fp16_utils import *
from apex import amp, optimizers
from apex.multi_tensor_apply import multi_tensor_applier
except ImportError:
raise ImportError("Please install apex from https://www.github.com/nvidia/apex to run this example.")
ImageFile.LOAD_TRUNCATED_IMAGES = True
warnings.filterwarnings("ignore")
os.environ['CUDA_VISIBLE_DEVICES'] = configs.gpu_id
# set random seed
def seed_everything(seed):
random.seed(seed)
os.environ['PYTHONHASHSEED'] = str(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.backends.cudnn.deterministic = True
seed_everything(configs.seed)
# make dir for use
def makdir():
if not os.path.exists(configs.checkpoints):
os.makedirs(configs.checkpoints)
if not os.path.exists(configs.log_dir):
os.makedirs(configs.log_dir)
if not os.path.exists(configs.submits):
os.makedirs(configs.submits)
makdir()
best_acc = 0 # best test accuracy
best_loss = 999 # lower loss
def main():
global best_acc
global best_loss
start_epoch = configs.start_epoch
# set normalize configs for imagenet
normalize_imgnet = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
transform_train = transforms.Compose([
transforms.RandomResizedCrop(configs.input_size),
transforms.RandomHorizontalFlip(p=0.5),
transforms.RandomVerticalFlip(p=0.5),
transforms.ToTensor(),
normalize_imgnet
])
transform_val = transforms.Compose([
transforms.Resize(int(configs.input_size * 1.2)),
transforms.CenterCrop(configs.input_size),
transforms.ToTensor(),
normalize_imgnet
])
# Data loading code
if configs.split_online:
# use online random split dataset method
total_files = get_files(configs.dataset,"train")
train_files,val_files = train_test_split(total_files,test_size = 0.1,stratify=total_files["label"])
train_dataset = WeatherDataset(train_files,transform_train)
val_dataset = WeatherDataset(val_files,transform_val)
else:
# use offline split dataset
train_files = get_files(configs.dataset+"/train/","train")
val_files = get_files(configs.dataset+"/val/","train")
train_dataset = WeatherDataset(train_files,transform_train)
val_dataset = WeatherDataset(val_files,transform_val)
train_loader = torch.utils.data.DataLoader(
train_dataset, batch_size=configs.bs, shuffle=True,
num_workers=configs.workers, pin_memory=True,
)
val_loader = torch.utils.data.DataLoader(
val_dataset, batch_size=configs.bs, shuffle=False,
num_workers=configs.workers, pin_memory=True
)
# get model
model = get_model()
model.cuda()
# choose loss func,default is CE
if configs.loss_func == "LabelSmoothCE":
criterion = LabelSmoothingLoss(0.1, configs.num_classes).cuda()
elif configs.loss_func == "CrossEntropy":
criterion = nn.CrossEntropyLoss().cuda()
elif configs.loss_func == "FocalLoss":
criterion = FocalLoss(gamma=2).cuda()
else:
criterion = nn.CrossEntropyLoss().cuda()
optimizer = get_optimizer(model)
# set lr scheduler method
if configs.lr_scheduler == "step":
scheduler = torch.optim.lr_scheduler.StepLR(optimizer,step_size=10,gamma=0.1)
elif configs.lr_scheduler == "on_loss":
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, mode='min', factor=0.2, patience=5, verbose=False)
elif configs.lr_scheduler == "on_acc":
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, mode='max', factor=0.2, patience=5, verbose=False)
else:
scheduler = torch.optim.lr_scheduler.StepLR(optimizer,step_size=6,gamma=0.1)
# for fp16
if configs.fp16:
model, optimizer = amp.initialize(model, optimizer,
opt_level=configs.opt_level,
keep_batchnorm_fp32= None if configs.opt_level == "O1" else configs.keep_batchnorm_fp32
)
if configs.resume:
# Load checkpoint.
print('==> Resuming from checkpoint..')
assert os.path.isfile(configs.resume), 'Error: no checkpoint directory found!'
configs.checkpoint = os.path.dirname(configs.resume)
checkpoint = torch.load(configs.resume)
best_acc = checkpoint['best_acc']
start_epoch = checkpoint['epoch']
model.module.load_state_dict(checkpoint['state_dict'])
optimizer.load_state_dict(checkpoint['optimizer'])
logger = Logger(os.path.join(configs.log_dir, '%s_log.txt'%configs.model_name), title=configs.model_name, resume=True)
else:
logger = Logger(os.path.join(configs.log_dir, '%s_log.txt'%configs.model_name), title=configs.model_name)
logger.set_names(['Learning Rate', 'Train Loss', 'Valid Loss', 'Train Acc.', 'Valid Acc.'])
if configs.evaluate:
print('\nEvaluation only')
val_loss, val_acc = validate(val_loader, model, criterion, start_epoch)
print(' Test Loss: %.8f, Test Acc: %.2f' % (val_loss, val_acc))
return
# Train and val
for epoch in range(start_epoch, configs.epochs):
print('\nEpoch: [%d | %d] LR: %f' % (epoch + 1, configs.epochs, optimizer.param_groups[0]['lr']))
train_loss, train_acc, train_5 = train(train_loader, model, criterion, optimizer, epoch)
val_loss, val_acc, test_5 = validate(val_loader, model, criterion, epoch)
# adjust lr
if configs.lr_scheduler == "on_loss":
scheduler.step(val_loss)
elif configs.lr_scheduler == "on_acc":
scheduler.step(val_acc)
elif configs.lr_scheduler == "step":
scheduler.step(epoch)
elif configs.lr_scheduler == "adjust":
adjust_learning_rate(optimizer,epoch)
else:
scheduler.step(epoch)
# append logger file
lr_current = get_lr(optimizer)
logger.append([lr_current,train_loss, val_loss, train_acc, val_acc])
print('train_loss:%f, val_loss:%f, train_acc:%f, train_5:%f, val_acc:%f, val_5:%f' % (train_loss, val_loss, train_acc, train_5, val_acc, test_5))
# save model
is_best = val_acc > best_acc
is_best_loss = val_loss < best_loss
best_acc = max(val_acc, best_acc)
best_loss = min(val_loss,best_loss)
save_checkpoint({
'fold': 0,
'epoch': epoch + 1,
'state_dict': model.state_dict(),
'train_acc': train_acc,
'acc': val_acc,
'best_acc': best_acc,
'best_loss': best_loss,
'optimizer': optimizer.state_dict(),
}, is_best,is_best_loss)
logger.close()
print('Best acc:')
print(best_acc)
def train(train_loader, model, criterion, optimizer, epoch):
# switch to train mode
model.train()
batch_time = AverageMeter()
data_time = AverageMeter()
losses = AverageMeter()
top1 = AverageMeter()
top5 = AverageMeter()
end = time.time()
bar = Bar('Training: ', max=len(train_loader))
for batch_idx, (inputs, targets) in enumerate(train_loader):
# measure data loading time
data_time.update(time.time() - end)
inputs, targets = inputs.cuda(), targets.cuda()
inputs, targets = torch.autograd.Variable(inputs), torch.autograd.Variable(targets)
# compute output
outputs = model(inputs)
loss = criterion(outputs, targets)
# measure accuracy and record loss
prec1, prec5 = accuracy(outputs.data, targets.data, topk=(1, 5))
losses.update(loss.item(), inputs.size(0))
top1.update(prec1.item(), inputs.size(0))
top5.update(prec5.item(), inputs.size(0))
# compute gradient and do SGD step
optimizer.zero_grad()
if configs.fp16:
with amp.scale_loss(loss, optimizer) as scaled_loss:
scaled_loss.backward()
else:
loss.backward()
# clip gradient
torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=5.0, norm_type=2)
optimizer.step()
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
# plot progress
bar.suffix = '({batch}/{size}) Data: {data:.3f}s | Batch: {bt:.3f}s | Total: {total:} | ETA: {eta:} | Loss: {loss:.4f} | top1: {top1: .4f} | top5: {top5: .4f}'.format(
batch=batch_idx + 1,
size=len(train_loader),
data=data_time.val,
bt=batch_time.val,
total=bar.elapsed_td,
eta=bar.eta_td,
loss=losses.avg,
top1=top1.avg,
top5=top5.avg,
)
bar.next()
bar.finish()
return (losses.avg, top1.avg, top5.avg)
def validate(val_loader, model, criterion, epoch):
global best_acc
batch_time = AverageMeter()
data_time = AverageMeter()
losses = AverageMeter()
top1 = AverageMeter()
top5 = AverageMeter()
# switch to evaluate mode
model.eval()
end = time.time()
bar = Bar('Validating: ', max=len(val_loader))
with torch.no_grad():
for batch_idx, (inputs, targets) in enumerate(val_loader):
# measure data loading time
data_time.update(time.time() - end)
inputs, targets = inputs.cuda(), targets.cuda()
inputs, targets = torch.autograd.Variable(inputs), torch.autograd.Variable(targets)
# compute output
outputs = model(inputs)
loss = criterion(outputs, targets)
# measure accuracy and record loss
prec1, prec5 = accuracy(outputs.data, targets.data, topk=(1, 5))
losses.update(loss.item(), inputs.size(0))
top1.update(prec1.item(), inputs.size(0))
top5.update(prec5.item(), inputs.size(0))
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
# plot progress
bar.suffix = '({batch}/{size}) Data: {data:.3f}s | Batch: {bt:.3f}s | Total: {total:} | ETA: {eta:} | Loss: {loss:.4f} | top1: {top1: .4f} | top5: {top5: .4f}'.format(
batch=batch_idx + 1,
size=len(val_loader),
data=data_time.avg,
bt=batch_time.avg,
total=bar.elapsed_td,
eta=bar.eta_td,
loss=losses.avg,
top1=top1.avg,
top5=top5.avg,
)
bar.next()
bar.finish()
return (losses.avg, top1.avg, top5.avg)
if __name__ == '__main__':
main()
================================================
FILE: models/__init__.py
================================================
from .model import *
================================================
FILE: models/model.py
================================================
from pretrainedmodels import models as pm
import pretrainedmodels
from torch import nn
from torchvision import models as tm
from config import configs
from efficientnet_pytorch import EfficientNet
import torch
from torch.nn.parameter import Parameter
import torch.nn.functional as F
from torch.nn.parameter import Parameter
weights = {
"efficientnet-b3":"/data/dataset/detection/pretrainedmodels/efficientnet-b3-c8376fa2.pth",
"efficientnet-b4":"/data/dataset/detection/pretrainedmodels/efficientnet-b4-6ed6700e.pth",
"efficientnet-b5":"/data/dataset/detection/pretrainedmodels/efficientnet-b5-b6417697.pth",
"efficientnet-b6":"/data/dataset/detection/pretrainedmodels/efficientnet-b6-c76e70fd.pth",
}
def gem(x, p=3, eps=1e-6):
return F.avg_pool2d(x.clamp(min=eps).pow(p), (x.size(-2), x.size(-1))).pow(1./p)
class GeM(nn.Module):
def __init__(self, p=3, eps=1e-6):
super(GeM,self).__init__()
self.p = Parameter(torch.ones(1)*p)
self.eps = eps
def forward(self, x):
return gem(x, p=self.p, eps=self.eps)
def __repr__(self):
return self.__class__.__name__ + '(' + 'p=' + '{:.4f}'.format(self.p.data.tolist()[0]) + ', ' + 'eps=' + str(self.eps) + ')'
def get_model():
if configs.model_name.startswith("resnext50_32x4d"):
model = tm.resnext50_32x4d(pretrained=True)
model.avgpool = nn.AdaptiveAvgPool2d(1)
model.fc = nn.Linear(2048,configs.num_classes)
model.cuda()
elif configs.model_name.startswith("efficient"):
# efficientNet
model_name = configs.model_name[:15]
model = EfficientNet.from_name(model_name)
model.load_state_dict(torch.load(weights[model_name]))
in_features = model._fc.in_features
model._fc = nn.Sequential(
nn.BatchNorm1d(in_features),
nn.Dropout(0.5),
nn.Linear(in_features, configs.num_classes),
)
model.cuda()
else:
pretrained = "imagenet+5k" if configs.model_name.startswith("dpn") else "imagenet"
model = pretrainedmodels.__dict__[configs.model_name.split("-model")[0]](num_classes=1000, pretrained=pretrained)
if configs.model_name.startswith("pnasnet"):
model.last_linear = nn.Linear(4320, configs.num_classes)
model.avg_pool = nn.AdaptiveAvgPool2d(1)
elif configs.model_name.startswith("inception"):
model.last_linear = nn.Linear(1536, configs.num_classes)
model.avgpool_1a = nn.AdaptiveAvgPool2d(1)
else:
model.last_linear = nn.Linear(2048, configs.num_classes)
model.avg_pool = nn.AdaptiveAvgPool2d(1)
model.cuda()
return model
================================================
FILE: test.py
================================================
import os
import torch
import warnings
import pandas as pd
import numpy as np
import torch.backends.cudnn as cudnn
from tqdm import tqdm
from glob import glob
from PIL import Image,ImageFile
from config import configs
from models.model import get_model
from torch.utils.data import Dataset,DataLoader
from torchvision import transforms
from utils.misc import get_files
from IPython import embed
ImageFile.LOAD_TRUNCATED_IMAGES = True
warnings.filterwarnings("ignore")
os.environ['CUDA_VISIBLE_DEVICES'] = configs.gpu_id
len_data = 0
class WeatherTTADataset(Dataset):
def __init__(self,labels_file,aug):
imgs = []
for index, row in labels_file.iterrows():
imgs.append((row["FileName"],row["type"]))
self.imgs = imgs
self.length = len(imgs)
global len_data
len_data = self.length
self.aug = aug
self.Hflip = transforms.RandomHorizontalFlip(p=1)
self.Vflip = transforms.RandomVerticalFlip(p=1)
self.Rotate = transforms.functional.rotate
self.resize = transforms.Resize((configs.input_size,configs.input_size))
self.randomCrop = transforms.Compose([transforms.Resize(int(configs.input_size * 1.2)),
transforms.CenterCrop(configs.input_size),
])
def __getitem__(self,index):
filename,label_tmp = self.imgs[index]
img = Image.open(configs.test_folder + os.sep + filename).convert('RGB')
img = self.transform_(img,self.aug)
return img,filename
def __len__(self):
return self.length
def transform_(self,data_torch,aug):
if aug == 'Ori':
data_torch = data_torch
data_torch = self.resize(data_torch)
if aug == 'Ori_Hflip':
data_torch = self.Hflip(data_torch)
data_torch = self.resize(data_torch)
if aug == 'Ori_Vflip':
data_torch = self.Vflip(data_torch)
data_torch = self.resize(data_torch)
if aug == 'Ori_Rotate_90':
data_torch = self.Rotate(data_torch, 90)
data_torch = self.resize(data_torch)
if aug == 'Ori_Rotate_180':
data_torch = self.Rotate(data_torch, 180)
data_torch = self.resize(data_torch)
if aug == 'Ori_Rotate_270':
data_torch = self.Rotate(data_torch, 270)
data_torch = self.resize(data_torch)
if aug == 'Crop':
# print(data_torch.size)
data_torch = self.randomCrop(data_torch)
data_torch = data_torch
if aug == 'Crop_Hflip':
data_torch = self.randomCrop(data_torch)
data_torch = self.Hflip(data_torch)
if aug == 'Crop_Vflip':
data_torch = self.randomCrop(data_torch)
data_torch = self.Vflip(data_torch)
if aug == 'Crop_Rotate_90':
data_torch = self.randomCrop(data_torch)
data_torch = self.Rotate(data_torch, 90)
if aug == 'Crop_Rotate_180':
data_torch = self.randomCrop(data_torch)
data_torch = self.Rotate(data_torch, 180)
if aug == 'Crop_Rotate_270':
data_torch = self.randomCrop(data_torch)
data_torch = self.Rotate(data_torch, 270)
data_torch = transforms.ToTensor()(data_torch)
data_torch = transforms.Normalize(mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225])(data_torch)
return data_torch
#aug = ['Ori','Ori_Hflip','Ori_Vflip','Ori_Rotate_90','Ori_Rotate_180','Ori_Rotate_270',
# 'Crop','Crop_Hflip','Crop_Vflip','Crop_Rotate_90','Crop_Rotate_180','Crop_Rotate_270']
aug = ['Ori_Hflip']
cpk_filename = configs.checkpoints + os.sep + configs.model_name + "-checkpoint.pth.tar"
best_cpk = cpk_filename.replace("-checkpoint.pth.tar","-best_model.pth.tar")
checkpoint = torch.load(best_cpk)
cudnn.benchmark = True
model = get_model()
model.load_state_dict(checkpoint['state_dict'])
model.eval()
test_files = pd.read_csv(configs.submit_example)
with torch.no_grad():
y_pred_prob = torch.FloatTensor([])
for a in tqdm(aug):
print(a)
test_set = WeatherTTADataset(test_files, a)
test_loader = DataLoader(dataset=test_set, batch_size=configs.bs, shuffle=False,
num_workers=4, pin_memory=True, sampler=None)
total = 0
correct = 0
for inputs, labels in tqdm(test_loader):
inputs = inputs.cuda()
outputs = model(inputs)
outputs = torch.nn.functional.softmax(outputs, dim=1)
# print(outputs.shape)
y_pred_prob = torch.cat([y_pred_prob, outputs.to("cpu")], dim=0)
#embed()
y_pred_prob = y_pred_prob.reshape((len(aug), len_data, configs.num_classes))
y_pred_prob = torch.sum(y_pred_prob, 0) / (len(aug) * 1.0)
_, predicted_all = torch.max(y_pred_prob, 1)
predicted = predicted_all + 1 # If the category starts with 1 ,else delet 1
test_files.type = predicted.data.cpu().numpy().tolist()
test_files.to_csv('./submits/%s_baseline.csv' % configs.model_name, index=False)
================================================
FILE: utils/__init__.py
================================================
from .optimizers import *
from .logger import *
from .losses import *
================================================
FILE: utils/logger.py
================================================
# A simple torch style logger
# (C) Wei YANG 2017
from __future__ import absolute_import
import matplotlib.pyplot as plt
import os
import sys
import numpy as np
__all__ = ['Logger', 'LoggerMonitor', 'savefig']
def savefig(fname, dpi=None):
dpi = 150 if dpi == None else dpi
plt.savefig(fname, dpi=dpi)
def plot_overlap(logger, names=None):
names = logger.names if names == None else names
numbers = logger.numbers
for _, name in enumerate(names):
x = np.arange(len(numbers[name]))
plt.plot(x, np.asarray(numbers[name]))
return [logger.title + '(' + name + ')' for name in names]
class Logger(object):
'''Save training process to log file with simple plot function.'''
def __init__(self, fpath, title=None, resume=False):
self.file = None
self.resume = resume
self.title = '' if title == None else title
if fpath is not None:
if resume:
self.file = open(fpath, 'r')
name = self.file.readline()
self.names = name.rstrip().split('\t')
self.numbers = {}
for _, name in enumerate(self.names):
self.numbers[name] = []
for numbers in self.file:
numbers = numbers.rstrip().split('\t')
for i in range(0, len(numbers)):
self.numbers[self.names[i]].append(numbers[i])
self.file.close()
self.file = open(fpath, 'a')
else:
self.file = open(fpath, 'w')
def set_names(self, names):
if self.resume:
pass
# initialize numbers as empty list
self.numbers = {}
self.names = names
for _, name in enumerate(self.names):
self.file.write(name)
self.file.write('\t')
self.numbers[name] = []
self.file.write('\n')
self.file.flush()
def append(self, numbers):
assert len(self.names) == len(numbers), 'Numbers do not match names'
for index, num in enumerate(numbers):
self.file.write("{0:.6f}".format(num))
self.file.write('\t')
self.numbers[self.names[index]].append(num)
self.file.write('\n')
self.file.flush()
def plot(self, names=None):
names = self.names if names == None else names
numbers = self.numbers
for _, name in enumerate(names):
x = np.arange(len(numbers[name]))
plt.plot(x, np.asarray(numbers[name]))
plt.legend([self.title + '(' + name + ')' for name in names])
plt.grid(True)
def close(self):
if self.file is not None:
self.file.close()
class LoggerMonitor(object):
'''Load and visualize multiple logs.'''
def __init__ (self, paths):
'''paths is a distionary with {name:filepath} pair'''
self.loggers = []
for title, path in paths.items():
logger = Logger(path, title=title, resume=True)
self.loggers.append(logger)
def plot(self, names=None):
plt.figure()
plt.subplot(121)
legend_text = []
for logger in self.loggers:
legend_text += plot_overlap(logger, names)
plt.legend(legend_text, bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)
plt.grid(True)
if __name__ == '__main__':
# Example: logger monitor
paths = {
'temp':'./logs/efficientnet-b3/log.txt',
}
field = ['Valid Acc.']
monitor = LoggerMonitor(paths)
monitor.plot(names=field)
savefig('test.eps')
================================================
FILE: utils/losses/__init__.py
================================================
from .label_smoothing import LabelSmoothingLoss
from .focalloss import FocalLoss
================================================
FILE: utils/losses/focalloss.py
================================================
import torch
from torch import nn
class FocalLoss(nn.Module):
def __init__(self, gamma=2., reduction='mean'):
super().__init__()
self.gamma = gamma
self.reduction = reduction
def forward(self, inputs, targets):
CE_loss = nn.CrossEntropyLoss(reduction='none')(inputs, targets)
pt = torch.exp(-CE_loss)
F_loss = ((1 - pt)**self.gamma) * CE_loss
if self.reduction == 'sum':
return F_loss.sum()
elif self.reduction == 'mean':
return F_loss.mean()
================================================
FILE: utils/losses/label_smoothing.py
================================================
import torch
from torch import nn
import torch.nn.functional as F
class LabelSmoothingLoss(nn.Module):
def __init__(self, label_smoothing, class_nums, ignore_index=-100):
assert 0.0 < label_smoothing <= 1.0
self.ignore_index = ignore_index
super(LabelSmoothingLoss, self).__init__()
smoothing_value = label_smoothing / (class_nums - 1)
one_hot = torch.full((class_nums,), smoothing_value)
if self.ignore_index >= 0:
one_hot[self.ignore_index] = 0
self.register_buffer('one_hot', one_hot.unsqueeze(0))
self.confidence = 1.0 - label_smoothing
def forward(self, output, target):
"""
output (FloatTensor): batch_size x n_classes
target (LongTensor): batch_size
"""
log_output = F.log_softmax(output, dim=1)
model_prob = self.one_hot.repeat(target.size(0), 1)
model_prob.scatter_(1, target.unsqueeze(1), self.confidence)
if self.ignore_index >= 0:
model_prob.masked_fill_((target == self.ignore_index).unsqueeze(1), 0)
# print("model_prob:{}".format(model_prob))
# print("log_output:{}".format(log_output))
return -torch.sum(model_prob * log_output) / target.size(0)
================================================
FILE: utils/misc.py
================================================
import os
import torch
import shutil
import pandas as pd
from .optimizers import *
from config import configs
from torch import optim as optim_t
from tqdm import tqdm
from glob import glob
from itertools import chain
def get_optimizer(model):
if configs.optim == "adam":
return optim_t.Adam(model.parameters(),
configs.lr,
betas=(configs.beta1,configs.beta2),
weight_decay=configs.wd)
elif configs.optim == "radam":
return RAdam(model.parameters(),
configs.lr,
betas=(configs.beta1,configs.beta2),
weight_decay=configs.wd)
elif configs.optim == "ranger":
return Ranger(model.parameters(),
lr = configs.lr,
betas=(configs.beta1,configs.beta2),
weight_decay=configs.wd)
elif configs.optim == "over9000":
return Over9000(model.parameters(),
lr = configs.lr,
betas=(configs.beta1,configs.beta2),
weight_decay=configs.wd)
elif configs.optim == "ralamb":
return Ralamb(model.parameters(),
lr = configs.lr,
betas=(configs.beta1,configs.beta2),
weight_decay=configs.wd)
elif configs.optim == "sgd":
return optim_t.SGD(model.parameters(),
lr = configs.lr,
momentum=configs.mom,
weight_decay=configs.wd)
else:
print("%s optimizer will be add later"%configs.optim)
def save_checkpoint(state,is_best,is_best_loss):
filename = configs.checkpoints + os.sep + configs.model_name + "-checkpoint.pth.tar"
torch.save(state, filename)
if is_best:
message = filename.replace("-checkpoint.pth.tar","-best_model.pth.tar")
shutil.copyfile(filename, message)
if is_best_loss:
message = filename.replace("-checkpoint.pth.tar","-best_loss.pth.tar")
shutil.copyfile(filename, message)
def get_lr(optimizer):
for param_group in optimizer.param_groups:
return param_group['lr']
class AverageMeter(object):
"""Computes and stores the average and current value
Imported from https://github.com/pytorch/examples/blob/master/imagenet/main.py#L247-L262
"""
def __init__(self):
self.reset()
def reset(self):
self.val = 0
self.avg = 0
self.sum = 0
self.count = 0
def update(self, val, n=1):
self.val = val
self.sum += val * n
self.count += n
self.avg = self.sum / self.count
def accuracy(output, target, topk=(1,)):
"""Computes the precision@k for the specified values of k"""
maxk = max(topk)
batch_size = target.size(0)
_, pred = output.topk(maxk, 1, True, True)
pred = pred.t()
correct = pred.eq(target.view(1, -1).expand_as(pred))
res = []
for k in topk:
correct_k = correct[:k].view(-1).float().sum(0)
res.append(correct_k.mul_(100.0 / batch_size))
return res
def get_files(root,mode):
if mode == "test":
files = []
for img in os.listdir(root):
files.append(root + img)
files = pd.DataFrame({"filename":files})
return files
else:
all_data_path, labels = [], []
image_folders = list(map(lambda x: root + x, os.listdir(root)))
all_images = list(chain.from_iterable(list(map(lambda x: glob(x + "/*"), image_folders))))
print("loading train dataset")
for file in tqdm(all_images):
all_data_path.append(file)
labels.append(int(file.split(os.sep)[-2]))
all_files = pd.DataFrame({"filename": all_data_path, "label": labels})
return all_files
def adjust_learning_rate(optimizer, epoch):
"""Sets the learning rate to the initial LR decayed by 10 every 30 epochs"""
lrs = [5e-4, 1e-4, 1e-5, 1e-6]
if epoch<=10:
lr = lrs[0]
elif epoch>10 and epoch<=16:
lr = lrs[1]
elif epoch>16 and epoch<=22:
lr = lrs[2]
else:
lr = lrs[-1]
for param_group in optimizer.param_groups:
param_group['lr'] = lr
================================================
FILE: utils/optimizers/__init__.py
================================================
from .lookahead import *
from .novograd import *
from .over9000 import *
from .radam import *
from .ralamb import *
from .ranger import *
================================================
FILE: utils/optimizers/lookahead.py
================================================
# Lookahead implementation from https://github.com/rwightman/pytorch-image-models/blob/master/timm/optim/lookahead.py
""" Lookahead Optimizer Wrapper.
Implementation modified from: https://github.com/alphadl/lookahead.pytorch
Paper: `Lookahead Optimizer: k steps forward, 1 step back` - https://arxiv.org/abs/1907.08610
"""
import torch
from torch.optim.optimizer import Optimizer
from collections import defaultdict
class Lookahead(Optimizer):
def __init__(self, base_optimizer, alpha=0.5, k=6):
if not 0.0 <= alpha <= 1.0:
raise ValueError(f'Invalid slow update rate: {alpha}')
if not 1 <= k:
raise ValueError(f'Invalid lookahead steps: {k}')
defaults = dict(lookahead_alpha=alpha, lookahead_k=k, lookahead_step=0)
self.base_optimizer = base_optimizer
self.param_groups = self.base_optimizer.param_groups
self.defaults = base_optimizer.defaults
self.defaults.update(defaults)
self.state = defaultdict(dict)
# manually add our defaults to the param groups
for name, default in defaults.items():
for group in self.param_groups:
group.setdefault(name, default)
def update_slow(self, group):
for fast_p in group["params"]:
if fast_p.grad is None:
continue
param_state = self.state[fast_p]
if 'slow_buffer' not in param_state:
param_state['slow_buffer'] = torch.empty_like(fast_p.data)
param_state['slow_buffer'].copy_(fast_p.data)
slow = param_state['slow_buffer']
slow.add_(group['lookahead_alpha'], fast_p.data - slow)
fast_p.data.copy_(slow)
def sync_lookahead(self):
for group in self.param_groups:
self.update_slow(group)
def step(self, closure=None):
# print(self.k)
#assert id(self.param_groups) == id(self.base_optimizer.param_groups)
loss = self.base_optimizer.step(closure)
for group in self.param_groups:
group['lookahead_step'] += 1
if group['lookahead_step'] % group['lookahead_k'] == 0:
self.update_slow(group)
return loss
def state_dict(self):
fast_state_dict = self.base_optimizer.state_dict()
slow_state = {
(id(k) if isinstance(k, torch.Tensor) else k): v
for k, v in self.state.items()
}
fast_state = fast_state_dict['state']
param_groups = fast_state_dict['param_groups']
return {
'state': fast_state,
'slow_state': slow_state,
'param_groups': param_groups,
}
def load_state_dict(self, state_dict):
fast_state_dict = {
'state': state_dict['state'],
'param_groups': state_dict['param_groups'],
}
self.base_optimizer.load_state_dict(fast_state_dict)
# We want to restore the slow state, but share param_groups reference
# with base_optimizer. This is a bit redundant but least code
slow_state_new = False
if 'slow_state' not in state_dict:
print('Loading state_dict from optimizer without Lookahead applied.')
state_dict['slow_state'] = defaultdict(dict)
slow_state_new = True
slow_state_dict = {
'state': state_dict['slow_state'],
'param_groups': state_dict['param_groups'], # this is pointless but saves code
}
super(Lookahead, self).load_state_dict(slow_state_dict)
self.param_groups = self.base_optimizer.param_groups # make both ref same container
if slow_state_new:
# reapply defaults to catch missing lookahead specific ones
for name, default in self.defaults.items():
for group in self.param_groups:
group.setdefault(name, default)
def LookaheadAdam(params, alpha=0.5, k=6, *args, **kwargs):
adam = Adam(params, *args, **kwargs)
return Lookahead(adam, alpha, k)
================================================
FILE: utils/optimizers/novograd.py
================================================
# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
#
# 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.
import torch
from torch.optim import Optimizer
import math
class AdamW(Optimizer):
"""Implements AdamW algorithm.
It has been proposed in `Adam: A Method for Stochastic Optimization`_.
Arguments:
params (iterable): iterable of parameters to optimize or dicts defining
parameter groups
lr (float, optional): learning rate (default: 1e-3)
betas (Tuple[float, float], optional): coefficients used for computing
running averages of gradient and its square (default: (0.9, 0.999))
eps (float, optional): term added to the denominator to improve
numerical stability (default: 1e-8)
weight_decay (float, optional): weight decay (L2 penalty) (default: 0)
amsgrad (boolean, optional): whether to use the AMSGrad variant of this
algorithm from the paper `On the Convergence of Adam and Beyond`_
Adam: A Method for Stochastic Optimization:
https://arxiv.org/abs/1412.6980
On the Convergence of Adam and Beyond:
https://openreview.net/forum?id=ryQu7f-RZ
"""
def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8,
weight_decay=0, amsgrad=False):
if not 0.0 <= lr:
raise ValueError("Invalid learning rate: {}".format(lr))
if not 0.0 <= eps:
raise ValueError("Invalid epsilon value: {}".format(eps))
if not 0.0 <= betas[0] < 1.0:
raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0]))
if not 0.0 <= betas[1] < 1.0:
raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1]))
defaults = dict(lr=lr, betas=betas, eps=eps,
weight_decay=weight_decay, amsgrad=amsgrad)
super(AdamW, self).__init__(params, defaults)
def __setstate__(self, state):
super(AdamW, self).__setstate__(state)
for group in self.param_groups:
group.setdefault('amsgrad', False)
def step(self, closure=None):
"""Performs a single optimization step.
Arguments:
closure (callable, optional): A closure that reevaluates the model
and returns the loss.
"""
loss = None
if closure is not None:
loss = closure()
for group in self.param_groups:
for p in group['params']:
if p.grad is None:
continue
grad = p.grad.data
if grad.is_sparse:
raise RuntimeError('Adam does not support sparse gradients, please consider SparseAdam instead')
amsgrad = group['amsgrad']
state = self.state[p]
# State initialization
if len(state) == 0:
state['step'] = 0
# Exponential moving average of gradient values
state['exp_avg'] = torch.zeros_like(p.data)
# Exponential moving average of squared gradient values
state['exp_avg_sq'] = torch.zeros_like(p.data)
if amsgrad:
# Maintains max of all exp. moving avg. of sq. grad. values
state['max_exp_avg_sq'] = torch.zeros_like(p.data)
exp_avg, exp_avg_sq = state['exp_avg'], state['exp_avg_sq']
if amsgrad:
max_exp_avg_sq = state['max_exp_avg_sq']
beta1, beta2 = group['betas']
state['step'] += 1
# Decay the first and second moment running average coefficient
exp_avg.mul_(beta1).add_(1 - beta1, grad)
exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad)
if amsgrad:
# Maintains the maximum of all 2nd moment running avg. till now
torch.max(max_exp_avg_sq, exp_avg_sq, out=max_exp_avg_sq)
# Use the max. for normalizing running avg. of gradient
denom = max_exp_avg_sq.sqrt().add_(group['eps'])
else:
denom = exp_avg_sq.sqrt().add_(group['eps'])
bias_correction1 = 1 - beta1 ** state['step']
bias_correction2 = 1 - beta2 ** state['step']
step_size = group['lr'] * math.sqrt(bias_correction2) / bias_correction1
p.data.add_(-step_size, torch.mul(p.data, group['weight_decay']).addcdiv_(1, exp_avg, denom) )
return loss
class Novograd(Optimizer):
"""
Implements Novograd algorithm.
Args:
params (iterable): iterable of parameters to optimize or dicts defining
parameter groups
lr (float, optional): learning rate (default: 1e-3)
betas (Tuple[float, float], optional): coefficients used for computing
running averages of gradient and its square (default: (0.95, 0))
eps (float, optional): term added to the denominator to improve
numerical stability (default: 1e-8)
weight_decay (float, optional): weight decay (L2 penalty) (default: 0)
grad_averaging: gradient averaging
amsgrad (boolean, optional): whether to use the AMSGrad variant of this
algorithm from the paper `On the Convergence of Adam and Beyond`_
(default: False)
"""
def __init__(self, params, lr=1e-3, betas=(0.95, 0), eps=1e-8,
weight_decay=0, grad_averaging=False, amsgrad=False):
if not 0.0 <= lr:
raise ValueError("Invalid learning rate: {}".format(lr))
if not 0.0 <= eps:
raise ValueError("Invalid epsilon value: {}".format(eps))
if not 0.0 <= betas[0] < 1.0:
raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0]))
if not 0.0 <= betas[1] < 1.0:
raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1]))
defaults = dict(lr=lr, betas=betas, eps=eps,
weight_decay=weight_decay,
grad_averaging=grad_averaging,
amsgrad=amsgrad)
super(Novograd, self).__init__(params, defaults)
def __setstate__(self, state):
super(Novograd, self).__setstate__(state)
for group in self.param_groups:
group.setdefault('amsgrad', False)
def step(self, closure=None):
"""Performs a single optimization step.
Arguments:
closure (callable, optional): A closure that reevaluates the model
and returns the loss.
"""
loss = None
if closure is not None:
loss = closure()
for group in self.param_groups:
for p in group['params']:
if p.grad is None:
continue
grad = p.grad.data
if grad.is_sparse:
raise RuntimeError('Sparse gradients are not supported.')
amsgrad = group['amsgrad']
state = self.state[p]
# State initialization
if len(state) == 0:
state['step'] = 0
# Exponential moving average of gradient values
state['exp_avg'] = torch.zeros_like(p.data)
# Exponential moving average of squared gradient values
state['exp_avg_sq'] = torch.zeros([]).to(state['exp_avg'].device)
if amsgrad:
# Maintains max of all exp. moving avg. of sq. grad. values
state['max_exp_avg_sq'] = torch.zeros([]).to(state['exp_avg'].device)
exp_avg, exp_avg_sq = state['exp_avg'], state['exp_avg_sq']
if amsgrad:
max_exp_avg_sq = state['max_exp_avg_sq']
beta1, beta2 = group['betas']
state['step'] += 1
norm = torch.sum(torch.pow(grad, 2))
if exp_avg_sq == 0:
exp_avg_sq.copy_(norm)
else:
exp_avg_sq.mul_(beta2).add_(1 - beta2, norm)
if amsgrad:
# Maintains the maximum of all 2nd moment running avg. till now
torch.max(max_exp_avg_sq, exp_avg_sq, out=max_exp_avg_sq)
# Use the max. for normalizing running avg. of gradient
denom = max_exp_avg_sq.sqrt().add_(group['eps'])
else:
denom = exp_avg_sq.sqrt().add_(group['eps'])
grad.div_(denom)
if group['weight_decay'] != 0:
grad.add_(group['weight_decay'], p.data)
if group['grad_averaging']:
grad.mul_(1 - beta1)
exp_avg.mul_(beta1).add_(grad)
p.data.add_(-group['lr'], exp_avg)
return loss
================================================
FILE: utils/optimizers/over9000.py
================================================
import torch, math
from torch.optim.optimizer import Optimizer
import itertools as it
from .lookahead import *
from .ralamb import *
# RAdam + LARS + LookAHead
# Lookahead implementation from https://github.com/lonePatient/lookahead_pytorch/blob/master/optimizer.py
# RAdam + LARS implementation from https://gist.github.com/redknightlois/c4023d393eb8f92bb44b2ab582d7ec20
def Over9000(params, alpha=0.5, k=6, *args, **kwargs):
ralamb = Ralamb(params, *args, **kwargs)
return Lookahead(ralamb, alpha, k)
RangerLars = Over9000
================================================
FILE: utils/optimizers/radam.py
================================================
# from https://github.com/LiyuanLucasLiu/RAdam/blob/master/radam.py
import math
import torch
from torch.optim.optimizer import Optimizer, required
class RAdam(Optimizer):
def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weight_decay=0):
defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay)
self.buffer = [[None, None, None] for ind in range(10)]
super(RAdam, self).__init__(params, defaults)
def __setstate__(self, state):
super(RAdam, self).__setstate__(state)
def step(self, closure=None):
loss = None
if closure is not None:
loss = closure()
for group in self.param_groups:
for p in group['params']:
if p.grad is None:
continue
grad = p.grad.data.float()
if grad.is_sparse:
raise RuntimeError('RAdam does not support sparse gradients')
p_data_fp32 = p.data.float()
state = self.state[p]
if len(state) == 0:
state['step'] = 0
state['exp_avg'] = torch.zeros_like(p_data_fp32)
state['exp_avg_sq'] = torch.zeros_like(p_data_fp32)
else:
state['exp_avg'] = state['exp_avg'].type_as(p_data_fp32)
state['exp_avg_sq'] = state['exp_avg_sq'].type_as(p_data_fp32)
exp_avg, exp_avg_sq = state['exp_avg'], state['exp_avg_sq']
beta1, beta2 = group['betas']
exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad)
exp_avg.mul_(beta1).add_(1 - beta1, grad)
state['step'] += 1
buffered = self.buffer[int(state['step'] % 10)]
if state['step'] == buffered[0]:
N_sma, step_size = buffered[1], buffered[2]
else:
buffered[0] = state['step']
beta2_t = beta2 ** state['step']
N_sma_max = 2 / (1 - beta2) - 1
N_sma = N_sma_max - 2 * state['step'] * beta2_t / (1 - beta2_t)
buffered[1] = N_sma
# more conservative since it's an approximated value
if N_sma >= 5:
step_size = math.sqrt((1 - beta2_t) * (N_sma - 4) / (N_sma_max - 4) * (N_sma - 2) / N_sma * N_sma_max / (N_sma_max - 2)) / (1 - beta1 ** state['step'])
else:
step_size = 1.0 / (1 - beta1 ** state['step'])
buffered[2] = step_size
if group['weight_decay'] != 0:
p_data_fp32.add_(-group['weight_decay'] * group['lr'], p_data_fp32)
# more conservative since it's an approximated value
if N_sma >= 5:
denom = exp_avg_sq.sqrt().add_(group['eps'])
p_data_fp32.addcdiv_(-step_size * group['lr'], exp_avg, denom)
else:
p_data_fp32.add_(-step_size * group['lr'], exp_avg)
p.data.copy_(p_data_fp32)
return loss
class PlainRAdam(Optimizer):
def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weight_decay=0):
defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay)
super(PlainRAdam, self).__init__(params, defaults)
def __setstate__(self, state):
super(PlainRAdam, self).__setstate__(state)
def step(self, closure=None):
loss = None
if closure is not None:
loss = closure()
for group in self.param_groups:
for p in group['params']:
if p.grad is None:
continue
grad = p.grad.data.float()
if grad.is_sparse:
raise RuntimeError('RAdam does not support sparse gradients')
p_data_fp32 = p.data.float()
state = self.state[p]
if len(state) == 0:
state['step'] = 0
state['exp_avg'] = torch.zeros_like(p_data_fp32)
state['exp_avg_sq'] = torch.zeros_like(p_data_fp32)
else:
state['exp_avg'] = state['exp_avg'].type_as(p_data_fp32)
state['exp_avg_sq'] = state['exp_avg_sq'].type_as(p_data_fp32)
exp_avg, exp_avg_sq = state['exp_avg'], state['exp_avg_sq']
beta1, beta2 = group['betas']
exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad)
exp_avg.mul_(beta1).add_(1 - beta1, grad)
state['step'] += 1
beta2_t = beta2 ** state['step']
N_sma_max = 2 / (1 - beta2) - 1
N_sma = N_sma_max - 2 * state['step'] * beta2_t / (1 - beta2_t)
if group['weight_decay'] != 0:
p_data_fp32.add_(-group['weight_decay'] * group['lr'], p_data_fp32)
# more conservative since it's an approximated value
if N_sma >= 5:
step_size = group['lr'] * math.sqrt((1 - beta2_t) * (N_sma - 4) / (N_sma_max - 4) * (N_sma - 2) / N_sma * N_sma_max / (N_sma_max - 2)) / (1 - beta1 ** state['step'])
denom = exp_avg_sq.sqrt().add_(group['eps'])
p_data_fp32.addcdiv_(-step_size, exp_avg, denom)
else:
step_size = group['lr'] / (1 - beta1 ** state['step'])
p_data_fp32.add_(-step_size, exp_avg)
p.data.copy_(p_data_fp32)
return loss
class AdamW(Optimizer):
def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weight_decay=0, warmup = 0):
defaults = dict(lr=lr, betas=betas, eps=eps,
weight_decay=weight_decay, warmup = warmup)
super(AdamW, self).__init__(params, defaults)
def __setstate__(self, state):
super(AdamW, self).__setstate__(state)
def step(self, closure=None):
loss = None
if closure is not None:
loss = closure()
for group in self.param_groups:
for p in group['params']:
if p.grad is None:
continue
grad = p.grad.data.float()
if grad.is_sparse:
raise RuntimeError('Adam does not support sparse gradients, please consider SparseAdam instead')
p_data_fp32 = p.data.float()
state = self.state[p]
if len(state) == 0:
state['step'] = 0
state['exp_avg'] = torch.zeros_like(p_data_fp32)
state['exp_avg_sq'] = torch.zeros_like(p_data_fp32)
else:
state['exp_avg'] = state['exp_avg'].type_as(p_data_fp32)
state['exp_avg_sq'] = state['exp_avg_sq'].type_as(p_data_fp32)
exp_avg, exp_avg_sq = state['exp_avg'], state['exp_avg_sq']
beta1, beta2 = group['betas']
state['step'] += 1
exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad)
exp_avg.mul_(beta1).add_(1 - beta1, grad)
denom = exp_avg_sq.sqrt().add_(group['eps'])
bias_correction1 = 1 - beta1 ** state['step']
bias_correction2 = 1 - beta2 ** state['step']
if group['warmup'] > state['step']:
scheduled_lr = 1e-8 + state['step'] * group['lr'] / group['warmup']
else:
scheduled_lr = group['lr']
step_size = group['lr'] * math.sqrt(bias_correction2) / bias_correction1
if group['weight_decay'] != 0:
p_data_fp32.add_(-group['weight_decay'] * scheduled_lr, p_data_fp32)
p_data_fp32.addcdiv_(-step_size, exp_avg, denom)
p.data.copy_(p_data_fp32)
return loss
================================================
FILE: utils/optimizers/ralamb.py
================================================
import torch, math
from torch.optim.optimizer import Optimizer
# RAdam + LARS
class Ralamb(Optimizer):
def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weight_decay=0):
defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay)
self.buffer = [[None, None, None] for ind in range(10)]
super(Ralamb, self).__init__(params, defaults)
def __setstate__(self, state):
super(Ralamb, self).__setstate__(state)
def step(self, closure=None):
loss = None
if closure is not None:
loss = closure()
for group in self.param_groups:
for p in group['params']:
if p.grad is None:
continue
grad = p.grad.data.float()
if grad.is_sparse:
raise RuntimeError('Ralamb does not support sparse gradients')
p_data_fp32 = p.data.float()
state = self.state[p]
if len(state) == 0:
state['step'] = 0
state['exp_avg'] = torch.zeros_like(p_data_fp32)
state['exp_avg_sq'] = torch.zeros_like(p_data_fp32)
else:
state['exp_avg'] = state['exp_avg'].type_as(p_data_fp32)
state['exp_avg_sq'] = state['exp_avg_sq'].type_as(p_data_fp32)
exp_avg, exp_avg_sq = state['exp_avg'], state['exp_avg_sq']
beta1, beta2 = group['betas']
# Decay the first and second moment running average coefficient
# m_t
exp_avg.mul_(beta1).add_(1 - beta1, grad)
# v_t
exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad)
state['step'] += 1
buffered = self.buffer[int(state['step'] % 10)]
if state['step'] == buffered[0]:
N_sma, radam_step_size = buffered[1], buffered[2]
else:
buffered[0] = state['step']
beta2_t = beta2 ** state['step']
N_sma_max = 2 / (1 - beta2) - 1
N_sma = N_sma_max - 2 * state['step'] * beta2_t / (1 - beta2_t)
buffered[1] = N_sma
# more conservative since it's an approximated value
if N_sma >= 5:
radam_step_size = math.sqrt((1 - beta2_t) * (N_sma - 4) / (N_sma_max - 4) * (N_sma - 2) / N_sma * N_sma_max / (N_sma_max - 2)) / (1 - beta1 ** state['step'])
else:
radam_step_size = 1.0 / (1 - beta1 ** state['step'])
buffered[2] = radam_step_size
if group['weight_decay'] != 0:
p_data_fp32.add_(-group['weight_decay'] * group['lr'], p_data_fp32)
# more conservative since it's an approximated value
radam_step = p_data_fp32.clone()
if N_sma >= 5:
denom = exp_avg_sq.sqrt().add_(group['eps'])
radam_step.addcdiv_(-radam_step_size * group['lr'], exp_avg, denom)
else:
radam_step.add_(-radam_step_size * group['lr'], exp_avg)
radam_norm = radam_step.pow(2).sum().sqrt()
weight_norm = p.data.pow(2).sum().sqrt().clamp(0, 10)
if weight_norm == 0 or radam_norm == 0:
trust_ratio = 1
else:
trust_ratio = weight_norm / radam_norm
state['weight_norm'] = weight_norm
state['adam_norm'] = radam_norm
state['trust_ratio'] = trust_ratio
if N_sma >= 5:
p_data_fp32.addcdiv_(-radam_step_size * group['lr'] * trust_ratio, exp_avg, denom)
else:
p_data_fp32.add_(-radam_step_size * group['lr'] * trust_ratio, exp_avg)
p.data.copy_(p_data_fp32)
return loss
================================================
FILE: utils/optimizers/ranger.py
================================================
import math
import torch
from torch.optim.optimizer import Optimizer, required
import itertools as it
from .lookahead import *
from .radam import *
def Ranger(params, alpha=0.5, k=6, *args, **kwargs):
radam = RAdam(params, *args, **kwargs)
return Lookahead(radam, alpha, k)
================================================
FILE: utils/reader.py
================================================
from torch.utils.data import Dataset
from PIL import Image
class WeatherDataset(Dataset):
# define dataset
def __init__(self,label_list,transforms=None,mode="train"):
super(WeatherDataset,self).__init__()
self.label_list = label_list
self.transforms = transforms
self.mode = mode
imgs = []
if self.mode == "test":
for index,row in label_list.iterrows():
imgs.append((row["filename"]))
self.imgs = imgs
else:
for index,row in label_list.iterrows():
imgs.append((row["filename"],row["label"]))
self.imgs = imgs
def __len__(self):
return len(self.imgs)
def __getitem__(self,index):
if self.mode == "test":
filename = self.imgs[index]
img = Image.open(filename).convert('RGB')
img = self.transforms(img)
return img,filename
else:
filename,label = self.imgs[index]
img = Image.open(filename).convert('RGB')
img = self.transforms(img)
return img,label
gitextract_vee5ftx8/
├── .gitignore
├── LICENSE
├── README.md
├── config.py
├── ensemble.py
├── main.py
├── models/
│ ├── __init__.py
│ └── model.py
├── test.py
└── utils/
├── __init__.py
├── logger.py
├── losses/
│ ├── __init__.py
│ ├── focalloss.py
│ └── label_smoothing.py
├── misc.py
├── optimizers/
│ ├── __init__.py
│ ├── lookahead.py
│ ├── novograd.py
│ ├── over9000.py
│ ├── radam.py
│ ├── ralamb.py
│ └── ranger.py
└── reader.py
SYMBOL INDEX (82 symbols across 15 files)
FILE: config.py
class DefaultConfigs (line 1) | class DefaultConfigs(object):
FILE: main.py
function seed_everything (line 40) | def seed_everything(seed):
function makdir (line 50) | def makdir():
function main (line 62) | def main():
function train (line 196) | def train(train_loader, model, criterion, optimizer, epoch):
function validate (line 255) | def validate(val_loader, model, criterion, epoch):
FILE: models/model.py
function gem (line 19) | def gem(x, p=3, eps=1e-6):
class GeM (line 22) | class GeM(nn.Module):
method __init__ (line 23) | def __init__(self, p=3, eps=1e-6):
method forward (line 27) | def forward(self, x):
method __repr__ (line 29) | def __repr__(self):
function get_model (line 32) | def get_model():
FILE: test.py
class WeatherTTADataset (line 22) | class WeatherTTADataset(Dataset):
method __init__ (line 23) | def __init__(self,labels_file,aug):
method __getitem__ (line 39) | def __getitem__(self,index):
method __len__ (line 45) | def __len__(self):
method transform_ (line 47) | def transform_(self,data_torch,aug):
FILE: utils/logger.py
function savefig (line 11) | def savefig(fname, dpi=None):
function plot_overlap (line 15) | def plot_overlap(logger, names=None):
class Logger (line 23) | class Logger(object):
method __init__ (line 25) | def __init__(self, fpath, title=None, resume=False):
method set_names (line 47) | def set_names(self, names):
method append (line 61) | def append(self, numbers):
method plot (line 70) | def plot(self, names=None):
method close (line 79) | def close(self):
class LoggerMonitor (line 83) | class LoggerMonitor(object):
method __init__ (line 85) | def __init__ (self, paths):
method plot (line 92) | def plot(self, names=None):
FILE: utils/losses/focalloss.py
class FocalLoss (line 4) | class FocalLoss(nn.Module):
method __init__ (line 5) | def __init__(self, gamma=2., reduction='mean'):
method forward (line 10) | def forward(self, inputs, targets):
FILE: utils/losses/label_smoothing.py
class LabelSmoothingLoss (line 8) | class LabelSmoothingLoss(nn.Module):
method __init__ (line 9) | def __init__(self, label_smoothing, class_nums, ignore_index=-100):
method forward (line 22) | def forward(self, output, target):
FILE: utils/misc.py
function get_optimizer (line 12) | def get_optimizer(model):
function save_checkpoint (line 46) | def save_checkpoint(state,is_best,is_best_loss):
function get_lr (line 56) | def get_lr(optimizer):
class AverageMeter (line 60) | class AverageMeter(object):
method __init__ (line 64) | def __init__(self):
method reset (line 67) | def reset(self):
method update (line 73) | def update(self, val, n=1):
function accuracy (line 79) | def accuracy(output, target, topk=(1,)):
function get_files (line 94) | def get_files(root,mode):
function adjust_learning_rate (line 111) | def adjust_learning_rate(optimizer, epoch):
FILE: utils/optimizers/lookahead.py
class Lookahead (line 11) | class Lookahead(Optimizer):
method __init__ (line 12) | def __init__(self, base_optimizer, alpha=0.5, k=6):
method update_slow (line 28) | def update_slow(self, group):
method sync_lookahead (line 40) | def sync_lookahead(self):
method step (line 44) | def step(self, closure=None):
method state_dict (line 54) | def state_dict(self):
method load_state_dict (line 68) | def load_state_dict(self, state_dict):
function LookaheadAdam (line 94) | def LookaheadAdam(params, alpha=0.5, k=6, *args, **kwargs):
FILE: utils/optimizers/novograd.py
class AdamW (line 19) | class AdamW(Optimizer):
method __init__ (line 42) | def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8,
method __setstate__ (line 56) | def __setstate__(self, state):
method step (line 61) | def step(self, closure=None):
class Novograd (line 118) | class Novograd(Optimizer):
method __init__ (line 137) | def __init__(self, params, lr=1e-3, betas=(0.95, 0), eps=1e-8,
method __setstate__ (line 154) | def __setstate__(self, state):
method step (line 159) | def step(self, closure=None):
FILE: utils/optimizers/over9000.py
function Over9000 (line 12) | def Over9000(params, alpha=0.5, k=6, *args, **kwargs):
FILE: utils/optimizers/radam.py
class RAdam (line 7) | class RAdam(Optimizer):
method __init__ (line 9) | def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weig...
method __setstate__ (line 14) | def __setstate__(self, state):
method step (line 17) | def step(self, closure=None):
class PlainRAdam (line 82) | class PlainRAdam(Optimizer):
method __init__ (line 84) | def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weig...
method __setstate__ (line 89) | def __setstate__(self, state):
method step (line 92) | def step(self, closure=None):
class AdamW (line 147) | class AdamW(Optimizer):
method __init__ (line 149) | def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weig...
method __setstate__ (line 154) | def __setstate__(self, state):
method step (line 157) | def step(self, closure=None):
FILE: utils/optimizers/ralamb.py
class Ralamb (line 5) | class Ralamb(Optimizer):
method __init__ (line 7) | def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weig...
method __setstate__ (line 12) | def __setstate__(self, state):
method step (line 15) | def step(self, closure=None):
FILE: utils/optimizers/ranger.py
function Ranger (line 9) | def Ranger(params, alpha=0.5, k=6, *args, **kwargs):
FILE: utils/reader.py
class WeatherDataset (line 4) | class WeatherDataset(Dataset):
method __init__ (line 6) | def __init__(self,label_list,transforms=None,mode="train"):
method __len__ (line 20) | def __len__(self):
method __getitem__ (line 22) | def __getitem__(self,index):
Condensed preview — 23 files, each showing path, character count, and a content snippet. Download the .json file or copy for the full structured content (79K chars).
[
{
"path": ".gitignore",
"chars": 1799,
"preview": "# Byte-compiled / optimized / DLL files\n__pycache__/\n*.py[cod]\n*$py.class\n\n# C extensions\n*.so\n\n# Distribution / packagi"
},
{
"path": "LICENSE",
"chars": 11357,
"preview": " Apache License\n Version 2.0, January 2004\n "
},
{
"path": "README.md",
"chars": 950,
"preview": "### pytorch 图像分类竞赛框架\n\n### 1. 更新日志\n- (2020年5月2日) 基础版本上线\n\n### 2. 依赖库\n- pretrainedmodels\n- progress\n- efficientnet-pytorch\n"
},
{
"path": "config.py",
"chars": 2140,
"preview": "class DefaultConfigs(object):\n # set default configs, if you don't understand, don't modify\n seed = 666 "
},
{
"path": "ensemble.py",
"chars": 1329,
"preview": "import pandas as pd \nimport numpy as np \nimport os\nfrom IPython import embed\n\nfile1 = pd.read_csv(\"./csvs/efficientnet-b"
},
{
"path": "main.py",
"chars": 11779,
"preview": "import random\nimport time\nimport warnings\n\nimport torch.nn as nn\nimport torch.nn.parallel\nimport torch.backends.cudnn as"
},
{
"path": "models/__init__.py",
"chars": 20,
"preview": "from .model import *"
},
{
"path": "models/model.py",
"chars": 2817,
"preview": "from pretrainedmodels import models as pm\nimport pretrainedmodels\nfrom torch import nn\nfrom torchvision import models as"
},
{
"path": "test.py",
"chars": 5135,
"preview": "import os\nimport torch\nimport warnings\nimport pandas as pd\nimport numpy as np\nimport torch.backends.cudnn as cudnn\nfrom "
},
{
"path": "utils/__init__.py",
"chars": 69,
"preview": "from .optimizers import *\nfrom .logger import *\nfrom .losses import *"
},
{
"path": "utils/logger.py",
"chars": 3652,
"preview": "# A simple torch style logger\n# (C) Wei YANG 2017\nfrom __future__ import absolute_import\nimport matplotlib.pyplot as plt"
},
{
"path": "utils/losses/__init__.py",
"chars": 80,
"preview": "from .label_smoothing import LabelSmoothingLoss\nfrom .focalloss import FocalLoss"
},
{
"path": "utils/losses/focalloss.py",
"chars": 540,
"preview": "import torch\nfrom torch import nn\n\nclass FocalLoss(nn.Module):\n def __init__(self, gamma=2., reduction='mean'):\n "
},
{
"path": "utils/losses/label_smoothing.py",
"chars": 1254,
"preview": "import torch\n\nfrom torch import nn\n\nimport torch.nn.functional as F\n\n\nclass LabelSmoothingLoss(nn.Module):\n def __ini"
},
{
"path": "utils/misc.py",
"chars": 4274,
"preview": "import os\nimport torch\nimport shutil\nimport pandas as pd\nfrom .optimizers import *\nfrom config import configs\nfrom torch"
},
{
"path": "utils/optimizers/__init__.py",
"chars": 137,
"preview": "from .lookahead import *\nfrom .novograd import *\nfrom .over9000 import *\nfrom .radam import *\nfrom .ralamb import *\nfrom"
},
{
"path": "utils/optimizers/lookahead.py",
"chars": 4047,
"preview": "# Lookahead implementation from https://github.com/rwightman/pytorch-image-models/blob/master/timm/optim/lookahead.py\n\n\""
},
{
"path": "utils/optimizers/novograd.py",
"chars": 9569,
"preview": "# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.\n#\n# Licensed under the Apache License, Version 2.0 (the \""
},
{
"path": "utils/optimizers/over9000.py",
"chars": 540,
"preview": "import torch, math\nfrom torch.optim.optimizer import Optimizer\nimport itertools as it\nfrom .lookahead import *\nfrom .ral"
},
{
"path": "utils/optimizers/radam.py",
"chars": 8100,
"preview": "# from https://github.com/LiyuanLucasLiu/RAdam/blob/master/radam.py\n\nimport math\nimport torch\nfrom torch.optim.optimizer"
},
{
"path": "utils/optimizers/ralamb.py",
"chars": 4026,
"preview": "import torch, math\nfrom torch.optim.optimizer import Optimizer\n\n# RAdam + LARS\nclass Ralamb(Optimizer):\n\n def __init_"
},
{
"path": "utils/optimizers/ranger.py",
"chars": 288,
"preview": "\nimport math\nimport torch\nfrom torch.optim.optimizer import Optimizer, required\nimport itertools as it\nfrom .lookahead i"
},
{
"path": "utils/reader.py",
"chars": 1121,
"preview": "from torch.utils.data import Dataset\nfrom PIL import Image\n\nclass WeatherDataset(Dataset):\n # define dataset\n def "
}
]
About this extraction
This page contains the full source code of the spytensor/pytorch_img_classification_for_competition GitHub repository, extracted and formatted as plain text for AI agents and large language models (LLMs). The extraction includes 23 files (73.3 KB), approximately 18.3k tokens, and a symbol index with 82 extracted functions, classes, methods, constants, and types. Use this with OpenClaw, Claude, ChatGPT, Cursor, Windsurf, or any other AI tool that accepts text input. You can copy the full output to your clipboard or download it as a .txt file.
Extracted by GitExtract — free GitHub repo to text converter for AI. Built by Nikandr Surkov.