Commit History

Auteur SHA1 Bericht Datum
  Jay Shah bb169e2a54 revert to single tile scheduler 1 maand geleden
  Jay Shah 483b26e541 add working fp8 varlen 1 maand geleden
  jayhshah a5a75274bc FA3 kvcache + split kv + gqa parallelization (#1236) 2 maanden geleden
  Ying Zhang dff976a84a fixes 3 maanden geleden
  Ying Zhang 7b4e68e04f hopper local attention 3 maanden geleden
  jayhshah c92ca63268 FA3 FP8 qkv descales + restore max offset for h128 causal + added sync for producer WG (#1173) 3 maanden geleden
  jayhshah 5018ac6ac5 Fp8 kernel with "in-kernel" transpose of V in producer (#1100) 4 maanden geleden
  Tri Dao 3aae9c18c1 Revert "Changes For FP8 (#1075)" 4 maanden geleden
  ganeshcolfax 1899c970c8 Changes For FP8 (#1075) 4 maanden geleden
  Ying Zhang dfe1a59e4b Add var-seq-len to FA3 fp16 / bf16 fwd (#1072) 4 maanden geleden
  Tri Dao 74b0761ff7 [FA3] BF16 forward 5 maanden geleden
  Tri Dao 7f67966cc7 FA3 initial code release 5 maanden geleden