vllm/csrc
Zhuohan Li db09d4ad83
[FIX] Fix Alibi implementation in PagedAttention kernel (#945)
* [FIX] Fix Alibi implementation in PagedAttention kernel

* Fix test_attention

* Fix

---------

Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Co-authored-by: Oliver-ss <yuansongwx@outlook.com>
2023-09-07 15:53:14 -07:00
..
attention [FIX] Fix Alibi implementation in PagedAttention kernel (#945) 2023-09-07 15:53:14 -07:00
activation_kernels.cu Avoid compiling kernels for double data type (#933) 2023-09-02 14:59:47 +09:00
activation.cpp Implement approximate GELU kernels (#828) 2023-08-23 07:43:21 +09:00
attention.cpp Optimize MQA Kernel (#452) 2023-07-14 20:06:40 -04:00
cache_kernels.cu Avoid compiling kernels for double data type (#933) 2023-09-02 14:59:47 +09:00
cache.cpp Memcpy kernel for flash attention (#29) 2023-04-10 18:22:49 -07:00
dispatch_utils.h Avoid compiling kernels for double data type (#933) 2023-09-02 14:59:47 +09:00
layernorm_kernels.cu Avoid compiling kernels for double data type (#933) 2023-09-02 14:59:47 +09:00
layernorm.cpp Add custom kernel for RMS normalization (#16) 2023-04-01 00:51:22 +08:00
pos_encoding_kernels.cu [BugFix] Implement RoPE for GPT-J (#941) 2023-09-06 11:54:33 +09:00
pos_encoding.cpp [BugFix] Implement RoPE for GPT-J (#941) 2023-09-06 11:54:33 +09:00
reduction_utils.cuh Change the name to vLLM (#150) 2023-06-17 03:07:40 -07:00