Commit History

Автор SHA1 Съобщение Дата
  Tri Dao ccbb14f38e Implement rotary embedding in flash_attn_with_kvcache преди 1 година
  Tri Dao a157cc8c9b [FT] Implement MQA/GQA преди 1 година
  Tri Dao 2800efc71f [FT] rotary_cos/sin should have batch_size dimension преди 1 година
  Tri Dao 3a9bfd076f [FT] rotary_cos/sin should have shape (dim) instead of (seqlen, dim) преди 1 година
  Tri Dao 62e9814466 [Rotary] Make sure frequency calculation is in fp32 преди 1 година
  Tri Dao 48bc6eacd6 [Gen] Add rotary base as an argument to FT attention kernel преди 1 година
  Tri Dao f1e01c27ba [Gen] Pass qkv_stride to ft_attention kernel for batched generation преди 1 година
  Tri Dao 7c2191542a [Gen] Make generation work with Tensor Parallel преди 1 година
  Tri Dao a01d1213d7 [Gen] Add kernel from FasterTransformer for benchmarking преди 1 година