Woosuk Kwon b9926f7f66 Support block size 32 (#35)		2023-04-09 23:07:18 -07:00
..
activation.py	Optimize data movement (#20)	2023-04-02 00:30:17 -07:00
attention.py	Support block size 32 (#35)	2023-04-09 23:07:18 -07:00
cache.py	Implement block copy kernel to optimize beam search (#32)	2023-04-07 17:45:07 -07:00
layernorm.py	Add custom kernel for RMS normalization (#16)	2023-04-01 00:51:22 +08:00
pos_encoding.py	Optimize data movement (#20)	2023-04-02 00:30:17 -07:00