Fix the failing test for basic model tests - #47077
Closed
aarushjain29 wants to merge 5 commits into
Closed
aarushjain29 wants to merge 5 commits into
aarushjain29 wants to merge 5 commits into
Conversation
aarushjain29
requested review from
AndreasKaratzas,
DarkLight1337,
njhill and
ywang96
as code owners
June 29, 2026 21:26
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
AndreasKaratzas
added
rocm
AndreasKaratzas
force-pushed
the
fix-basic-models-extra-init-amd-v2
branch
from
July 2, 2026 03:00
55d735c to
c6acc09
Compare
Member
|
Looks like |
AndreasKaratzas
force-pushed
the
fix-basic-models-extra-init-amd-v2
branch
from
July 7, 2026 23:06
6b6ab01 to
3d73907
Compare
Signed-off-by: aarushjain29 <aarushi.jain2@amd.com>
Signed-off-by: aarushjain29 <aarushi.jain2@amd.com>
Signed-off-by: aarushjain29 <aarushi.jain2@amd.com>
Signed-off-by: aarushjain29 <aarushi.jain2@amd.com>
| "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4" | ||
| "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4", | ||
| # Lower utilization so the KV cache fits on one GPU; OOMs at default 0.8. | ||
| gpu_memory_utilization=0.2, |
Member
There was a problem hiding this comment.
Maybe we can try without this after #47366
I dont like that we have to set 0.2 utilization to avoid ooming, i mean it should not oom at anything other than 100% utilization so there is something else going on there
Member
There was a problem hiding this comment.
Sry about this, let's wait for this one too:
Signed-off-by: aarushjain29 <aarushi.jain2@amd.com>
aarushjain29
force-pushed
the
fix-basic-models-extra-init-amd-v2
branch
3 times, most recently
from
July 10, 2026 17:37
9812ff9 to
bf876ea
Compare
Contributor
|
This pull request has merge conflicts that must be resolved before it can be |
4 tasks
Contributor
|
Heads-up: #48527 merged (runs the init-test engine in-process on ROCm) and fixes this class of failures at the root. Might be worth re-checking whether this PR is still needed on the test side. |
Member
|
Closing as #48527 was already merged. |
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
6 participants
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.
Purpose
Fix the three failing models in
test_can_initialize_large_subsetfor the "Basic Models Tests (Extra Initialization)" group (AMD MI300 nightly, also reproducible upstream):2 cannot be divided by 8becausedummy_hf_overridesignoresrouter_n_groups.hidden_states, _ = outputs; CUDA graph capture then trips an MLA decode-only assertion.Test Plan
pytest -v -s models/test_initialization.py -k 'test_can_initialize_large_subset[<model>]'on MI300X (gfx942).Test Result
cc @AndreasKaratzas