Repository navigation
[minimax_m3_vl] Gate the integration tests on total accelerator memory - #50025
Merged
Merged
Conversation
MiniMaxAI/MiniMax-M3-MXFP8 is 413 GiB of weights. On smaller runners `device_map="auto"` offloads the MoE weights to disk, which fails since they need re-saving. Skip unless 480 GiB of accelerator memory is visible. Fixes #50023 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
|
[For maintainers] Suggested jobs to run (before merge) run-slow: minimax_m3_vl |
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
Collaborator
Author
|
run-slow: minimax_m3_vl |
Contributor
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #50023
All 6
MiniMaxM3VLIntegrationTesttests fail on the daily multi-GPU CI (2×A10G, 44 GiB) with "weights offloaded to the disk, which needed to be re-saved".MiniMaxAI/MiniMax-M3-MXFP8is 413 GiB of weights, sodevice_map="auto"pushes almost everything to disk, and the MoE weights can't be offloaded without re-saving. Passing anoffload_folderwould mean writing ~400 GB to disk per run, which isn't a realistic CI setup.This gates the class with
require_torch_accelerator_memory(memory=480)(total memory over all visible accelerators, as for longcat_flash and muse_spark), so it still runs on a node big enough for the model and skips elsewhere.As a follow-up, we'll add a fast integration test using
transformers-tmicso MiniMax-M3-VL keeps real coverage on regular CI.