Repository navigation

Qwen3-MoE: fused-expert quantisation, BF16 router and expert-wise GPTQ - #324

Open

Shreyas8612 wants to merge 4 commits into

feat/phase-split-quantisationfrom

feat/qwen3-moe-quantisation

Open

Qwen3-MoE: fused-expert quantisation, BF16 router and expert-wise GPTQ#324
Shreyas8612 wants to merge 4 commits into
feat/phase-split-quantisationfrom
feat/qwen3-moe-quantisation

Commits on Sep 16, 2026
- Add fused Qwen3-MoE experts, BF16 router and phase-aware attention
  Shreyas8612
  committed
  30e8182
- Wire Qwen3-MoE modules into replacement, phase hooks and rotation
  Shreyas8612
  committed
  3fb8809
- Extend GPTQ to fused Qwen3-MoE experts and sharded device maps
  Shreyas8612
  committed
  ec32347
- Add CPU tests for Qwen3-MoE phase quantisation and the router lane
  Shreyas8612
  committed
  2b19da7