Skip to content

Latest commit

 

History

History
21 lines (18 loc) · 530 Bytes

README.md

File metadata and controls

21 lines (18 loc) · 530 Bytes

torchcompress

  • A deep learning compression framework in Pytorch (work in progress)
    • Support Filter pruning:
      • Conv -> Conv
      • Linear -> Linear
      • Conv -> Flatten -> Linear
      • Batchnorm2d/1d

To do

  • Pruning
    • Unstructured pruning
    • Structured pruning
  • Quantization
    • Post Training Quantization
    • Quantization aware training
  • Knowledge distillation
    • Offline distillation
    • Online distillation
    • Self distillation