transformers
0c9e543f - Triton finegrained fp8/fp4 (#46407)

Commit
31 days ago
Triton finegrained fp8/fp4 (#46407) * finegrained fp8/fp4 * drop pref * order * full testing * remove errors * better ep + batched_mm * add batched as well * add ability to disable deepgemm linear fp8 * compilable * torch compile tests and cold cache hack
Parents
Loading