All pull requests

Pull requests

• Add vision_only observer mode for text/vision expert-usage skew analysis

  #31·aquilarubra opened Jul 19, 2026·3 tasks done·

  ·

• Add multimodal (image) calibration batches for layerwise saliency

  #30·aquilarubra opened Jul 19, 2026·

  ·

• Force eager fused-experts dispatch for layerwise replay

  #29·aquilarubra opened Jul 19, 2026·

  ·

• Add general MTP (multi-token-prediction head) pruning support

  #28·aquilarubra opened Jul 19, 2026·

  ·

• Add generic local-dataset fallback for calibration corpora

  #27·aquilarubra opened Jul 19, 2026·

  ·

• Stream checkpoints from disk during layerwise observation/pruning

  #26·aquilarubra opened Jul 19, 2026·

  ·

• Add Qwen3.5-MoE architecture support

  #25·aquilarubra opened Jul 19, 2026·

  ·

• feat: add MiniMax M2 (MiniMaxM2ForCausalLM) support

  #19·scottgl9 opened Apr 15, 2026·

  ·