Skip to content

[DO NOT MERGE] Copy of #2793 - #2798

Closed
Xiaoyu Z (xiaoyu-work) wants to merge 2 commits into
mainfrom
quark-0.13
Closed

Xiaoyu Z (xiaoyu-work) wants to merge 2 commits into
mainfrom
quark-0.13

Conversation

@xiaoyu-work

Copy link
Copy Markdown
Member

Describe your changes

Independent, destination-owned copy of #2793: [AMD] Require and test amd-quark 0.13 for QuarkQuantization. Do not merge this draft.

  • Source: thiagocrepaldi/Olive:quark-0.13
  • Destination: microsoft/Olive:quark-0.13
  • Base: main
  • Preserved head: 3de106e6f5c0722322eaf4c589c3633900395b8d

The source change raises the QuarkQuantization minimum version to amd-quark>=0.13.0 and updates the AMD test requirements, runtime version gates, tests, and quantization documentation. Original commits are preserved without edits, and the original PR remains unchanged.

Copy verification: the copied branch head matches the source PR head. Tests, model execution, and lintrunner -a were not run for this copy-only operation; the upstream implementation has not been revalidated.

Checklist before requesting a review

  • Add unit tests for this change.
  • Make sure all tests can pass.
  • Update documents if necessary.
  • Lint and apply fixes to your code by running lintrunner -a
  • Is this a user-facing change? If yes, give a description of this change to be included in the release notes.

(Optional) Issue link

Original PR: #2793

The AMD CI installed amd-quark==0.10 (test/requirements-test-amd.txt), three
releases behind the API the QuarkQuantization pass has used since #2532
(amd-quark>=0.12), so the pass's torch path was never tested against the
Quark it targets. Test against the current release, amd-quark 0.13, and make
it the pass's minimum (olive_config.json extra, the runtime version gates and
their tests, and the docs).

Verified with amd-quark 0.13, onnxruntime 1.28, torch 2.13 and transformers
5.15 / 5.22: test/passes/quark_quantizer passes (10 tests), and the pass runs
AWQ uint4_wo_128 with a Hugging Face export on a tiny Llama.

Co-Authored-By: Claude <noreply@anthropic.com>
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

Successfully merging this pull request may close these issues.

2 participants