biondizzle cf0731cf4b fix: warmup with 128 tokens (fills MMA tile), better error handling ...

The CuTeDSL kernel uses MMA tiler (128,128,256). With only 1 token,
the kernel can't fill a tile and may access illegal memory. Using 128
tokens for the warmup.

Also improved error message — after CUDA illegal memory access, the
context is corrupted and can't recover.

2026-05-16 04:56:45 +00:00

..

patches

fix: warmup with 128 tokens (fills MMA tile), better error handling

2026-05-16 04:56:45 +00:00

nvfp4_cutedsl.py

fix: L1 gate/up split — intermediate_size is per-projection, not fused

2026-05-16 04:04:40 +00:00