[https://nvbugs/5593199][test] Enhance beam search tests deterministic dummy model #8625
Conversation
|
/bot run |
| generation_cache_indirection = attn_metadata.cache_indirection[ | ||
| num_context_requests:num_requests].view( | ||
| num_generation_requests * attn_metadata.beam_width, | ||
| attn_metadata.cache_indirection.shape[-1] | ||
| )[:num_generation_beams] |
There was a problem hiding this comment.
num_generation_beams = num_generation_requests * attn_metadata.beam_width, so why do we need [:num_generation_beams] again?
There was a problem hiding this comment.
We don't. I'll remove this.
| start_token: int, | ||
| num_iterations: int = 4, | ||
| vocab_size: int = 1000, | ||
| use_overlap_scheduler: bool = False) -> list[list[int]]: |
There was a problem hiding this comment.
| use_overlap_scheduler: bool = False) -> list[list[int]]: | |
| use_overlap_scheduler: bool = False) -> BeamSearchTestOutput: |
| beam_scores[0] = score_03 | ||
| beam_scores[1] = score_13 | ||
| expected_cache_indirection[1].append(1) | ||
| expected_cache_indirection[0].append(0) |
There was a problem hiding this comment.
IIUC beam 0 is independent of the condition here so we should move it outside the if. I suggest grouping everything for beam 0 together before we handle beam 1.
There was a problem hiding this comment.
Done, although beam 0 is moved to the bottom, as beam 1 relies on beam 0
| top_offset, second_best_offset = torch.topk(softmax_score, k=2, | ||
| dim=-1).indices | ||
|
|
||
| # iteration 1 |
There was a problem hiding this comment.
I suggest changing that since the loop below starts at 1.
| # iteration 1 | |
| # First iteration |
| # Avoid swapping in the last iteration, as the cache indirection provided by the model does not reflect this change | ||
| # in case of overlap scheduler one additional model forward is performed, so we do not need to skip the last swap |
There was a problem hiding this comment.
| # Avoid swapping in the last iteration, as the cache indirection provided by the model does not reflect this change | |
| # in case of overlap scheduler one additional model forward is performed, so we do not need to skip the last swap | |
| # Avoid swapping in the last iteration, as the cache indirection provided by the model does not reflect this change. | |
| # In case of overlap scheduler one additional model forward is performed, so we do not need to skip the last swap. |
I don't really understand what's going on here. The model just outputs the cache indirections it gets as the input. So does that mean the sampler doesn't perform the swap in the last iteration?
There was a problem hiding this comment.
The sampler does perform the swap. But the model output is created before the sampling step in which this swap happens. So, the cache indirection from the model output does not "contain" any changes to cache indirection, which may happen during the following sampling step. Therefore, this code should also skip the last update to the cache indirection.
There was a problem hiding this comment.
Right, we're not able to check the final cache indirection, so we expect to see the one before the final sampling step. I think it would be good to make this more clear in the function documentation.
| @pytest.mark.parametrize("num_output_beams", [1, 2]) | ||
| @pytest.mark.parametrize("num_prompts", [1, 2]) | ||
| @pytest.mark.threadleak(enabled=False) | ||
| def test_beam_search_output_shapes(gather_context_logits: bool, |
There was a problem hiding this comment.
Should we rename the test? It tests more than the shapes, no?
There was a problem hiding this comment.
Sure, I'll remove the shapes
…odel and validation checks to provide deterministic output - added copyright header - Introduced a DummyModel and associated configurations to facilitate deterministic outputs for beam search tests. - Updated test cases to validate generation logits, logprobs, and cache indirection. - Refactored input prompts and expected outputs to align with the new model structure. - Added checks for context logits and overall output validation to ensure correctness in beam search functionality. Signed-off-by: Stefan Niebler <[email protected]>
…or clarity and correctness - Removed reliance on additional output from overlap scheduler - Removed unused parameters from `get_expected_outputs` and `validate_output` functions to simplify the interface. - Updated comments for better clarity on the logic flow within the beam search output validation. - Adjusted cache indirection checks to ensure accuracy in the last token generation. - Enhanced readability by restructuring code and comments in the test cases. Signed-off-by: Stefan Niebler <[email protected]>
stnie
force-pushed
the
develop/tests/beam_search_test_robustness
branch
from
0a4c810 to
86520ad
Compare
|
/bot run |
|
PR_Github #22307 [ run ] triggered by Bot. Commit: |
|
PR_Github #22307 [ run ] completed with state |
3 participants
@coderabbitai summary
Description
Updated beam search tests to provide deterministic output, to prevent the choice of hardware to affect the results
Test Coverage
PR Checklist
Please review the following before submitting your PR:
PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.
PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.
Test cases are provided for new code paths (see test instructions)
Any new dependencies have been scanned for license and vulnerabilities
CODEOWNERS updated if ownership changes
Documentation updated as needed
The reviewers assigned automatically/manually are appropriate for the PR.
Please check this after reviewing the above items as appropriate for this PR.
GitHub Bot Help
/bot [-h] ['run', 'kill', 'skip', 'reuse-pipeline'] ...Provide a user friendly way for developers to interact with a Jenkins server.
Run
/bot [-h|--help]to print this help message.See details below for each supported subcommand.
run [--reuse-test (optional)pipeline-id --disable-fail-fast --skip-test --stage-list "A10-PyTorch-1, xxx" --gpu-type "A30, H100_PCIe" --test-backend "pytorch, cpp" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx" --detailed-log --debug(experimental)]Launch build/test pipelines. All previously running jobs will be killed.
--reuse-test (optional)pipeline-id(OPTIONAL) : Allow the new pipeline to reuse build artifacts and skip successful test stages from a specified pipeline or the last pipeline if no pipeline-id is indicated. If the Git commit ID has changed, this option will be always ignored. The DEFAULT behavior of the bot is to reuse build artifacts and successful test results from the last pipeline.--disable-reuse-test(OPTIONAL) : Explicitly prevent the pipeline from reusing build artifacts and skipping successful test stages from a previous pipeline. Ensure that all builds and tests are run regardless of previous successes.--disable-fail-fast(OPTIONAL) : Disable fail fast on build/tests/infra failures.--skip-test(OPTIONAL) : Skip all test stages, but still run build stages, package stages and sanity check stages. Note: Does NOT update GitHub check status.--stage-list "A10-PyTorch-1, xxx"(OPTIONAL) : Only run the specified test stages. Examples: "A10-PyTorch-1, xxx". Note: Does NOT update GitHub check status.--gpu-type "A30, H100_PCIe"(OPTIONAL) : Only run the test stages on the specified GPU types. Examples: "A30, H100_PCIe". Note: Does NOT update GitHub check status.--test-backend "pytorch, cpp"(OPTIONAL) : Skip test stages which don't match the specified backends. Only support [pytorch, cpp, tensorrt, triton]. Examples: "pytorch, cpp" (does not run test stages with tensorrt or triton backend). Note: Does NOT update GitHub pipeline status.--only-multi-gpu-test(OPTIONAL) : Only run the multi-GPU tests. Note: Does NOT update GitHub check status.--disable-multi-gpu-test(OPTIONAL) : Disable the multi-GPU tests. Note: Does NOT update GitHub check status.--add-multi-gpu-test(OPTIONAL) : Force run the multi-GPU tests in addition to running L0 pre-merge pipeline.--post-merge(OPTIONAL) : Run the L0 post-merge pipeline instead of the ordinary L0 pre-merge pipeline.--extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx"(OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx".--detailed-log(OPTIONAL) : Enable flushing out all logs to the Jenkins console. This will significantly increase the log volume and may slow down the job.--debug(OPTIONAL) : Experimental feature. Enable access to the CI container for debugging purpose. Note: Specify exactly one stage in thestage-listparameter to access the appropriate container environment. Note: Does NOT update GitHub check status.For guidance on mapping tests to stage names, see
docs/source/reference/ci-overview.mdand the
scripts/test_to_stage_mapping.pyhelper.kill
killKill all running builds associated with pull request.
skip
skip --comment COMMENTSkip testing for latest commit on pull request.
--comment "Reason for skipping build/test"is required. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.reuse-pipeline
reuse-pipelineReuse a previous pipeline to validate current commit. This action will also kill all currently running builds associated with the pull request. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.