Steve Luo 7806842f2d add paged-attetionv2: support seq length split across thread block (#5707)

2024-05-14 12:46:54 +08:00

..

benchmark_context_attn_unpad.py

…

benchmark_decoding_attn.py

…

benchmark_flash_decoding_attention.py

add paged-attetionv2: support seq length split across thread block (#5707)

2024-05-14 12:46:54 +08:00

benchmark_fused_rotary_embdding_unpad.py

…

benchmark_kv_cache_memcopy.py

…

benchmark_rmsnorm.py

…

benchmark_rotary_embedding.py

…

benchmark_xine_copy.py

…

test_ci.sh

…