openvinotoolkit / openvinotoolkit/nncf

[TorchFX] Use FakeTensors in meta instead of actual tensors

Open
#2,934 1 comment 0 reactions 1 assignee View on GitHub

@daniil-lyakhov is already working on this.

Since Aug 30, 2024.

enhancement
Dominant language
Python
Stars
1.2k
Forks
305
Avg merge
1d 9h
Merged PRs (30d)
27

Description

🚀 Feature request

Current implementation is setting actual tensors as meta value (An exapmle: https://github.com/openvinotoolkit/nncf/blob/develop/nncf/experimental/torch/fx/transformations.py#L30-L46). The task is to optimize this process by creating fake tensors instead. Other problem - not every constant in captured graphs has meta val. The solution should cover such constants as well

Feature Use Case

No response

Are you going to submit a PR?
  • Yes I'd like to help by submitting a PR!

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.