[AMD] Register 2 recently-added CPU/ROCm-safe tests for AMD 1-GPU PR CI#29680
Open
michaelzhang-ai wants to merge 1 commit into
Open
[AMD] Register 2 recently-added CPU/ROCm-safe tests for AMD 1-GPU PR CI#29680michaelzhang-ai wants to merge 1 commit into
michaelzhang-ai wants to merge 1 commit into
Conversation
These tests landed recently on NVIDIA per-commit CI and are AMD-safe with no code changes: - models/test_vit_pos_embed_interpolate.py: pure torch embedding math, asserted bit-exact, already runs on CPU and CUDA. - unit/mem_cache/test_minimax_sparse_pool_host_unit.py: CPU integration class + a device/host transfer class that already enables is_hip(). Adding register_amd_ci(...) next to register_cuda_ci(...) is sufficient. AST collector: stage-a-test-1-gpu-small-amd 5 -> 6, stage-b-test-1-gpu-small-amd 112 -> 113.
Contributor
|
Warning You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again! |
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Motivation
These two tests were recently added to NVIDIA per-commit CI and are AMD-safe with no code changes — they only need a
register_amd_ci(...)line next to the existingregister_cuda_ci(...). This closes part of the AMD-vs-NVIDIA per-commit coverage gap surfaced by the ROCm upstream-ci dashboard.Registered files
test/registered/models/test_vit_pos_embed_interpolate.pystage-a-test-1-gpu-small-amdtorchembedding lookups + arithmetic, asserted bit-exact (rtol=0/atol=0). Already runs on CPU and CUDA; no fp8 / FlashInfer / custom-kernel paths. Heavy model deps are guarded withtry/except ... skipTest.test/registered/unit/mem_cache/test_minimax_sparse_pool_host_unit.pystage-b-test-1-gpu-small-amddevice="cpu"). The device↔host transfer class is already ROCm-aware —setUpenables it foris_cuda() or is_hip()and skips NPU/XPU.Both files keep their existing
register_cuda_ci(...)andregister_cpu_ci(...)calls unchanged; onlyregister_amd_ci(...)(legacysuite=shape) is added so the effective suite resolves to the canonical AMD per-commit suite name (AMD suites were not renamed in the upstreamstage-*→base-*rename).Local verification (AST collector)
Using the vendored
python/sglang/test/ci/ci_register.pyparser:Test plan
stage-a-test-1-gpu-small-amdpasses ontest_vit_pos_embed_interpolate.pystage-b-test-1-gpu-small-amdpasses ontest_minimax_sparse_pool_host_unit.pyCI States
Latest PR Test (Base): ⏳ Run #28405097298
Latest PR Test (Extra): ❌ Run #28405113548