google / google/qwix

How to pass Qwix quantization scales into a Pallas kernel?

Open
#192 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
134
Forks
27
Avg merge
2d 2h
Merged PRs (30d)
12

Description

Hi Qwix team,

I’m trying to use Qwix quantized weights together with a custom JAX Pallas kernel.
In the Pallas kernel, I pass the quantized weight tensor as an input, but I also need the corresponding scale to correctly dequantize or fuse the computation.

I’m not sure what the intended way is to get and pass these scales from Qwix.

Any guidance or pointers to examples would be very helpful. Thanks!

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.