Repository navigation
fix(tests): support ROCm in AIME GPU memory checks - #763
Merged
Merged
Conversation
KarrAcaRn
pushed a commit
to KarrAcaRn/FreeToken-ByAI
that referenced
this pull request
Oct 20, 2026
FlashML-org#764 FlashML-org#763 FlashML-org#708 FlashML-org#687 deferred, FlashML-org#688 kept as a feature; FlashML-org#169 kept after its upstream close, FlashML-org#494/FlashML-org#154/FlashML-org#315 replaced upstream by FlashML-org#690/FlashML-org#698; FlashML-org#278 dropped with upstream's auto->hybrid pick (FlashML-org#734). Assisted-by: Claude Opus 5.5
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The AIME memory check calls
nvidia-smi, failing on ROCm before model loading. Querytorch.cuda.mem_get_info(0)in a subprocess on ROCm. The NVIDIA path stays unchanged.Validation on base
4d2e2e2:Qwen/Qwen3.5-35B-A3B-FP8, offload with auto cache, AIME25 question 1, pass@3, up to 16384 tokens per sample. Original sampling and answer assertions are unchanged.nvidia-sminvidia-smiOn the 32 GiB R9700, set
FREETOKEN_AIME_MIN_FREE_GIB=30to match available VRAM andFREETOKEN_TEST_KV_RESERVE=18432to preserve the 16384-token budget.AMD validation: ROCm 10.1.0, Triton 3.8, PyTorch 2.14. Related to #535.