Skip to content

jitdee-ai/pooraka

Repository files navigation

Pooraka project

PyTorch wrapper for image classification

PyPI version Downloads PyPI Release

Supported:

  • Calculating Params and FLOPs
  • Calculate mean std
  • Get mean std (for CIFAR-10, CIFAR-100, and ImageNet)
  • Cross entropy label smooth
  • Get learning rate
  • Avgrage meter (including accuracy)
  • Create exp dir
  • Save checkpoint

Install

The library can be installed with pip:

pip install pooraka

Calculating Params and FLOPs

import pooraka as prk

flops, params = prk.get_flops_params(model, (224, 224))

CrossEntropyLabelSmooth

import pooraka as prk

CLASSES = 1000
criterion_smooth = prk.CrossEntropyLabelSmooth(CLASSES, 0.1)
criterion_smooth = criterion_smooth.cuda()