!pip install flash-attn --no-build-isolation
#15
by
NivYO
- opened
!pip install flash-attn --no-build-isolation takes forever on google colab...
40 min and still running. any suggestions of what can i do?
Hi @NivYO ! compiling can take long if you don't have ninja installed (> 2 hours according to the flash attention installation instructions) -- can you check if ninja is installed in you runtime?
Alternatively, if you prefer not to use flash attention, you can set trust_remote_code=False when you load the model form HF hub.
Hope this helps!:)
Any other solutions if I already installed ninja and want to use flash attention?