Historia zmian

Autor SHA1 Wiadomość Data
  jayhshah a5a75274bc FA3 kvcache + split kv + gqa parallelization (#1236) 2 miesięcy temu
  Ying Zhang 7b4e68e04f hopper local attention 3 miesięcy temu
  jayhshah c92ca63268 FA3 FP8 qkv descales + restore max offset for h128 causal + added sync for producer WG (#1173) 3 miesięcy temu
  Ying Zhang a3a257c71d Fix out-of-bound writes for var-seq-len zero-length KVs 4 miesięcy temu
  jayhshah 5018ac6ac5 Fp8 kernel with "in-kernel" transpose of V in producer (#1100) 4 miesięcy temu
  Ying Zhang dfe1a59e4b Add var-seq-len to FA3 fp16 / bf16 fwd (#1072) 4 miesięcy temu
  Tri Dao 74b0761ff7 [FA3] BF16 forward 5 miesięcy temu
  Tri Dao 7f67966cc7 FA3 initial code release 5 miesięcy temu