Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Using Mirage for different models #27

Open
ramyaprabhu-alt opened this issue Jun 25, 2024 · 0 comments
Open

Using Mirage for different models #27

ramyaprabhu-alt opened this issue Jun 25, 2024 · 0 comments

Comments

@ramyaprabhu-alt
Copy link

Hi, I just found this repository and I really like the idea. I wanted to try it out for a different model that what's on the readme for this repo, like say LLama 3 8B TP2.
But I'm a novice and am struggling to understand how the inputs in the given example must be modified

image

1.) From all the looking around I assume this is a prefill kernel for cl 4096 with input size of 256. am I correct?
2.) for each of the Q K and V tensors, why is the first of the input dims tuple 2 * batch size?
3.) what is 64 in the input dims tuple supposed to be? because the readme says this is a kernel for Llama 70B tp4...

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

No branches or pull requests

1 participant