wah

Install

pip install wah

Requirements

You might want to manually install PyTorch for GPU computation.

lightning
matplotlib
numpy
pandas
pyperclip
PyYAML
selenium
tensorboard
timm
torch
torchaudio
torchmetrics
torchvision
webdriver_manager

Examples

• Model Training
• Model Evaluation
• Geodesic Optimization

Structure

classification

• attacks
  • fgsm: FGSM, IFGSM
• datasets
  • base: ClassificationDataset
  • cifar10: CIFAR10
  • cifar100: CIFAR100
  • dataloader
    • __init__: to_dataloader
    • transforms: CollateFunction
  • imagenet: ImageNet
  • stl10: STL10
  • utils: compute_mean_and_std, DeNormalize, Normalize, portion_dataset, tensor_to_dataset
• models
  • feature_extraction: FeatureExtractor
  • load: add_preprocess, load_model, load_state_dict
  • replace:
    • __init__: Replacer
• test
  • accuracy: AccuracyTest
  • eval: EvalTest
  • hessian_max_eigval_spectrum: HessianMaxEigValSpectrum
  • loss: LossTest
  • pred: PredTest
  • tid: TIDTest
• train
  • plot: proj_train_path_to_2d, TrainPathPlot2D
  • train: Wrapper, load_trainer

module

_getattr, get_attrs, get_module_name, get_module_params, get_named_modules, get_valid_attr

np

path

basename, clean, dirname, exists, isdir, join, ls, mkdir, rmdir, rmfile, split, splitext

plot

• dist: DistPlot2D
• hist: HistPlot2D
• image: ImShow
• mat: MatShow2D
• quiver: QuiverPlot2D, TrajPlot2D
• scatter: GridPlot2D, ScatterPlot2D

riemann

• geodesic: optimize_geodesic
• grad: compute_jacobian, compute_hessian
• jacobian_sigvals: JacobianSigVals

tensor

broadcasted_elementwise_mul, create_1d_traj, create_2d_grid, flatten_batch, repeat, stretch

torch

utils

• args: ArgumentParser
• dictionary: dict_to_df, dict_to_tensor, load_csv_to_dict, load_yaml_to_dict, save_dict_to_csv
• download: disable_ssl_verification, download_url, md5_check
• logs: disable_lightning_logging
• lst: load_txt_to_list, save_list_to_txt, sort_str_list
• random: seed, unseed
• time: time
• zip: extract