Small utilities for PyTorch
Project description
Weaver PyTorch 🧶🧵
from weaver import get_classifier, get_optimizer, get_scheduler, get_transforms
from torchvision.transforms import Compose
model = get_classifier('torchvision', 'resnet50')
optim = get_optimizer(model.parameters(), name='SGD', lr=1e-3)
sched = get_scheduler(optim, name='CosineAnnealingLR', T_max=10)
transform = Compose(get_transforms([
{'name': 'RandAugment', 'num_ops': 2, 'magnitude': 10},
{"name": "ToTensor"},
{"name": "Normalize", "mean": "cifar10", "std": "cifar10"}
]))
Install
pip install .
API
get_classifier(src, name, **kwargs)
- weaver:
wide_resnet{depth}_{width}
,preact_resnet{depth}
- torchvision: https://pytorch.org/vision/stable/models.html
get_optimizer(params, name, **kwargs)
- PyTorch: https://pytorch.org/docs/stable/optim.html#algorithms
- AdaBelief: https://github.com/juntang-zhuang/Adabelief-Optimizer
get_scheduler(optim, name, **kwargs)
- PyTorch: https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
- Custom:
HalfCosineAnnealingLR
get_transform(name, **kwargs)
- PyTorch: https://pytorch.org/vision/stable/transforms.html
- Custom:
AllRandAugment
,Cutout
,Contain
get_transforms(kwargs_list)
- get list of transforms
Others
weaver.optimizers.exclude_wd(module: Module, skip_list=['bias', 'bn'])
weaver.optimizers.EMAModel(model: Module, alpha: float)
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Close
Hashes for weaver-pytorch-rnx0dvmdxk-0.0.1.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | d5c63f4116b2227a0a9460df6e29653e54a34cc5c64a2e254088d7a418edd17d |
|
MD5 | a3309b7f388eac178efcd833935a5661 |
|
BLAKE2b-256 | c0842d66118ba7c587d61dd9df2b42e305924d4f9fc5b96e1a7642a82a6084d6 |
Close
Hashes for weaver_pytorch_rnx0dvmdxk-0.0.1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 5f42df4735c22eda27279eb204de4c89cfdb02302128f03976c74edbed87df2c |
|
MD5 | c224b4ae2f66ec22567ea1f03841f525 |
|
BLAKE2b-256 | 74e63f0d221e20c38ebbaefab1e6751d65b4453d27aac0621df1895bf15e493d |