biondizzle bf22b6f0e4 Fix scale assembly: variable-size per-expert padding matching GEMM offsets ...

- Compute padded_expert_offsets from real expert_offsets (ceil to 128)
- Scatter x_sf into padded positions matching those offsets
- Per-expert swizzle in 128-row chunks (supports >128 tokens per expert)
- Pad slot_hidden/activated using same padded offsets for GEMM input
- Pre-allocated buffers sized for max_tokens*top_k (not num_experts*128)

2026-05-17 13:55:10 +00:00

..

patches

Fix warmup: pass local expert IDs (not global), remove incorrect _warmup_done guard

2026-05-17 11:38:19 +00:00

nvfp4_cutedsl.py

Fix scale assembly: variable-size per-expert padding matching GEMM offsets

2026-05-17 13:55:10 +00:00