Skip to content

Update llama_factory.rst which might be more user-friendly for beginners#850

Open
SuperAZHE wants to merge 1 commit intoQwenLM:mainfrom
SuperAZHE:patch-1
Open

Update llama_factory.rst which might be more user-friendly for beginners#850
SuperAZHE wants to merge 1 commit intoQwenLM:mainfrom
SuperAZHE:patch-1

Conversation

@SuperAZHE
Copy link
Copy Markdown

I have a suggestion to enhance the script's description, which might be more user-friendly for beginners, reducing the level of confusion

I have a suggestion to enhance the script's description, which might be more user-friendly for beginners, reducing the level of confusion
@jklj077 jklj077 requested a review from yangjianxin1 August 19, 2024 06:27
@jklj077
Copy link
Copy Markdown
Collaborator

jklj077 commented Aug 19, 2024

IMO, enabling fa2 by default is prone to problems. however, I agree that there should be section on how fa2 works in llama-factory that includes at least:

  1. what kinds of devices are required
  2. what improvements should be expected (if users are with pytorch like 2.3 and flash_attn=auto in llama_factory, there are chances that fa2 is already enabled through sdpa).

@jklj077
Copy link
Copy Markdown
Collaborator

jklj077 commented Aug 19, 2024

ping @yangjianxin1 who is also working on this

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants