biondizzle d9dc042ff7 Fix compressor kv_score: use forward() for NVFP4 quantized weights ...

Raw torch.mm doesn't work with packed uint8 NVFP4 weights.
Use MergedColumnParallelLinear.forward() which handles dequantization.

2026-05-19 00:29:43 +00:00

..

patches

Fix compressor kv_score: use forward() for NVFP4 quantized weights

2026-05-19 00:29:43 +00:00

cutedsl_quant_method.py

Fix: add abstract create_weights to CuTeDSLNvfp4LinearMethod

2026-05-18 20:40:48 +00:00

nvfp4_cutedsl.py

Fix torch.compile: use custom autograd Function instead of @torch.compiler.disable

2026-05-18 21:38:28 +00:00