support flash_attn v2 - Githubissues

intel / intel-extension-for-deepspeed

Intel® Extension for DeepSpeed* is an extension to DeepSpeed that brings feature support with SYCL kernels on Intel GPU(XPU) device. Note XPU is already supported in stock DeepSpeed (upstream).

MIT License

56 stars 19 forks source link

support flash_attn v2 #49

Closed YizhouZ closed 11 months ago

YizhouZ commented 1 year ago

Flash attn v2 would replace v1 implementation.

1pikachu commented 11 months ago

test_LLM_pr

1pikachu commented 11 months ago

test_LLM_pr

YizhouZ commented 11 months ago

verified by CI.