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

Port torch GPTQ to 3.x #1408

Merged
merged 30 commits into from
Nov 27, 2023
Merged

Port torch GPTQ to 3.x #1408

merged 30 commits into from
Nov 27, 2023

Conversation

yiliu30
Copy link
Contributor

@yiliu30 yiliu30 commented Nov 22, 2023

Type of Change

Port torch GPTQ to 3.x

The 3rd PR for 3.X API

  1. Introduce INC 3.0 quantization API and port torch RTN into 3.0 #1380
  2. Enhance 3.x API #1397

API changed:

  • Add torch 3.x GPTQ API
    Add calib_dataloader as a new arg for quantize as GPTQ needs a dataloader for calibration.
  • To align with torch quantize, rename calib_func to run_fn, calib_func_args to run_args

Description

  • Port GPTQ impl and add GPTQConfig
  • UTs for GPTQConfig
  • GPTQConfig + RTNConfig
  • Clean lwy import

How has this PR been tested?

Pre-CI

Dependency Change?

None

@chensuyue chensuyue added this to the v2.4 milestone Nov 22, 2023
@yiliu30 yiliu30 requested review from YIYANGCAI and xin3he November 24, 2023 06:03
@yiliu30 yiliu30 removed the WIP label Nov 24, 2023
@yiliu30 yiliu30 requested a review from ftian1 November 24, 2023 06:46
@yiliu30 yiliu30 added the WIP label Nov 24, 2023
@yiliu30 yiliu30 removed the WIP label Nov 27, 2023
Signed-off-by: yiliu30 <[email protected]>
@yiliu30 yiliu30 requested a review from chensuyue November 27, 2023 09:05
@chensuyue chensuyue merged commit 9150181 into master Nov 27, 2023
17 of 19 checks passed
@chensuyue chensuyue deleted the ly/gptq_3x branch November 27, 2023 12:50
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants