Issues

• l1_pruning and fpgm are registered as kind "other", so prune → quantize combinations are never tested

  wncc/Hello-FOSS-26_ModelShrinker#34 · Filter by author dip-era opened on Nov 1, 2026

• Pruning crashes on CUDA models with a Conv → Flatten → Linear boundary

  wncc/Hello-FOSS-26_ModelShrinker#33 · Filter by author dip-era opened on Nov 1, 2026

• Add shrinker.export_onnx with an output-parity check

  enhancementhelp wanted
  wncc/Hello-FOSS-26_ModelShrinker#21 · Filter by author bandiarham07-bot opened on Oct 16, 2026
  2

  

• Add a ptq_quantization method (static post-training quantization with calibration)

  compressionenhancementhelp wanted
  wncc/Hello-FOSS-26_ModelShrinker#20 · Filter by author bandiarham07-bot opened on Oct 16, 2026
  2

  

• Add loss_fn, scheduler and weight_decay to finetune

  enhancementgood first issue
  wncc/Hello-FOSS-26_ModelShrinker#19 · Filter by author bandiarham07-bot opened on Oct 16, 2026
  2

  

• Add shrinker.summary(model): a per-layer table with prunability

  enhancementgood first issue
  wncc/Hello-FOSS-26_ModelShrinker#18 · Filter by author bandiarham07-bot opened on Oct 16, 2026
  4

  

• dfq_quantization and qat_quantization fail to restore quantized engine when default is 'none' (macOS Apple Silicon)

  wncc/Hello-FOSS-26_ModelShrinker#14 · Filter by author bhattyuvraj22 opened on Oct 9, 2026
  1

  

• Support grouped pruning for layers that share channels

  enhancementhardpruning
  wncc/Hello-FOSS-26_ModelShrinker#12 · Filter by author bandiarham07-bot opened on Oct 8, 2026
  6

  

• Add compression methods for RNNs and Transformers

  compressionenhancementhard
  wncc/Hello-FOSS-26_ModelShrinker#11 · Filter by author bandiarham07-bot opened on Oct 8, 2026
  2

  

• Add Tucker / SVD low-rank decomposition

  compressionenhancementhelp wanted
  wncc/Hello-FOSS-26_ModelShrinker#10 · Filter by author bandiarham07-bot opened on Oct 8, 2026
  2

  

• Add HRank filter pruning

  compressionhelp wantedpruning
  wncc/Hello-FOSS-26_ModelShrinker#9 · Filter by author bandiarham07-bot opened on Oct 8, 2026
  2

  

• Quantized DFQ and QAT models cannot be saved and reloaded

  bughelp wantedserialisation
  wncc/Hello-FOSS-26_ModelShrinker#8 · Filter by author bandiarham07-bot opened on Oct 8, 2026
  2

  

• Replace deprecated torch.ao quantization APIs before PyTorch 2.10 removal

  good first issuehelp wantedmigration
  wncc/Hello-FOSS-26_ModelShrinker#7 · Filter by author bandiarham07-bot opened on Oct 8, 2026
  4

  

• Add a pipeline to combine methods

  enhancementmedium
  wncc/Hello-FOSS-26_ModelShrinker#6 · Filter by author bandiarham07-bot opened on Oct 8, 2026
  2

  

• Allow a different pruning ratio for each layer

  enhancementhelp wanted
  wncc/Hello-FOSS-26_ModelShrinker#5 · Filter by author bandiarham07-bot opened on Oct 8, 2026
  2

  

• Add a total memory estimate for the compressed model

  enhancementgood first issue
  wncc/Hello-FOSS-26_ModelShrinker#3 · Filter by author bandiarham07-bot opened on Oct 8, 2026
  2

  

• Use RL to predict the pruning ratio for each layer

  enhancementhard
  wncc/Hello-FOSS-26_ModelShrinker#2 · Filter by author bandiarham07-bot opened on Oct 8, 2026
  4

  

• Fix knowledge distillation method

  buggood first issue
  wncc/Hello-FOSS-26_ModelShrinker#1 · Filter by author bandiarham07-bot opened on Oct 8, 2026
  6