pytorch / pytorch/executorch

Question Regarding Code Execution in the Calibration Process.

Open
#6,629 8 comments 0 reactions 1 assignee View on GitHub

@cccclai is already working on this.

Since Nov 5, 2024.

module: quantization triaged
Dominant language
Python
Stars
5k
Forks
1.2k
Avg merge
2d 10h
Merged PRs (30d)
581

Description

Dear @cccclai

I’m reviewing the code while following the guide(Export with Spinquant) you provided for converting the Llama3.2-3B-Instruct model with Qualcomm SpinQuant. When I execute the _export_llama function in the export_llama_lib.py file, the pt2e_quantize(quantizers) function is called. Within this function, the pt2e_calibrate function is executed before the convert_pt2e function. Why is pt2e_calibrate performed before convert_pt2e here?
Generally, wouldn't it make more sense to perform calibration after quantization?

Thank you

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.