[https://nvbugs/6482297][fix] Parse boolean stage-config fields in test_to_stage_mapping#16671
Closed
Wanli-Jiang wants to merge 1 commit into
Closed
Conversation
…st_to_stage_mapping The stage-parsing regex in scripts/test_to_stage_mapping.py only accepted trailing numeric fields in Jenkins stage-config arrays, so the 47 stage entries carrying the runWithSbatch boolean (added by the Slurm test refactor NVIDIA#7176) were silently dropped from the test<->stage mapping, including all DGX_B200-PyTorch-* stages. It also matched commented-out stage lines. As a result, tests listed in l0_b200.yml resolved to no stage and test_cli_functionality failed whenever the seeded sample picked one of them. - Accept true/false fields in the stage-config regex and skip //-commented lines when parsing L0_Test.groovy. - Fix test_backend_filtering_consistency to tolerate tests legitimately listed under multiple backends (e.g. unittest/kv_cache_manager_v2_tests under both tensorrt and pytorch), which the parser fix exposed. Signed-off-by: Wanli Jiang <35160485+Wanli-Jiang@users.noreply.github.com>
Wanli-Jiang
requested review from
brnguyen2,
schetlur-nv,
yiqingy0 and
yuanjingx87
July 21, 2026 09:50
Contributor
📝 WalkthroughWalkthroughThe stage-mapping parser now handles boolean values in Jenkins configuration arrays and ignores comment lines. Backend consistency tests derive excluded backends from each test’s declared mappings. ChangesStage Mapping Consistency
Estimated code review effort: 2 (Simple) | ~10 minutes Possibly related PRs
Suggested reviewers: 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
🧹 Nitpick comments (2)
scripts/test_to_stage_mapping.py (1)
50-64: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick winAdd focused regression coverage for both parser changes.
Coverage is currently insufficient: the existing downstream checks do not directly prove that mixed integer/boolean arrays are parsed or that commented-out stages are excluded. Add fixtures in
tests/unittest/tools/test_test_to_stage_mapping.pyasserting bothstage_to_yamlandyaml_to_stagesfor these cases.Also applies to: 90-91
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@scripts/test_to_stage_mapping.py` around lines 50 - 64, Add focused fixtures in tests/unittest/tools/test_test_to_stage_mapping.py covering mixed integer/boolean array entries parsed by _STAGE_RE and commented-out stage entries being excluded. Assert both stage_to_yaml and yaml_to_stages outputs for each case, including expected stage names and YAML mappings.tests/unittest/tools/test_test_to_stage_mapping.py (1)
265-273: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick winMake the multi-backend regression explicit.
The set-subtraction logic is correct, but the capped
sample_testsselection can pass without exercisingunittest/kv_cache_manager_v2_tests. Add a focused test or deterministic fixture intests/unittest/tools/test_test_to_stage_mapping.pyasserting that a test declared under both TensorRT and PyTorch maps to both backend stage families.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/unittest/tools/test_test_to_stage_mapping.py` around lines 265 - 273, Add a focused regression test in the test-to-stage mapping suite that deterministically defines a test under both TensorRT and PyTorch, then asserts its mapping includes both corresponding backend stage families. Ensure this case is exercised independently of the capped sample_tests selection, while preserving the existing declared_backends and other_backends checks.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In `@scripts/test_to_stage_mapping.py`:
- Around line 50-64: Add focused fixtures in
tests/unittest/tools/test_test_to_stage_mapping.py covering mixed
integer/boolean array entries parsed by _STAGE_RE and commented-out stage
entries being excluded. Assert both stage_to_yaml and yaml_to_stages outputs for
each case, including expected stage names and YAML mappings.
In `@tests/unittest/tools/test_test_to_stage_mapping.py`:
- Around line 265-273: Add a focused regression test in the test-to-stage
mapping suite that deterministically defines a test under both TensorRT and
PyTorch, then asserts its mapping includes both corresponding backend stage
families. Ensure this case is exercised independently of the capped sample_tests
selection, while preserving the existing declared_backends and other_backends
checks.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: aeefd221-9988-4880-86d8-ea386ac6d77c
📒 Files selected for processing (2)
scripts/test_to_stage_mapping.pytests/unittest/tools/test_test_to_stage_mapping.py
Collaborator
Author
|
/bot run --disable-fail-fast |
Collaborator
|
PR_Github #60684 [ run ] triggered by Bot. Commit: |
Collaborator
|
PR_Github #60684 [ run ] completed with state
|
Wanli-Jiang
added a commit
to Wanli-Jiang/TensorRT-LLM
that referenced
this pull request
Jul 22, 2026
… and CI plumbing Trailing test/example/CI cleanup after the legacy TensorRT backend removal (TRTLLM-14026, PR NVIDIA#16369). CI schedules / jenkins: - Move backend-neutral tests off the retired -TensorRT- stages and delete all dead backend=tensorrt test-db buckets; flip remaining neutral buckets to pytorch so parked coverage runs again. - NEW stages A100X-PyTorch-Post-Merge-1 and L40S-PyTorch-Post-Merge-1 schedule the restored a100/l40s post-merge coverage (CI-owner review). - Retire the -TensorRT- plumbing in L0_Test.groovy (commented stage rows, backend=tensorrt mako branch, TEST_BACKEND changeMap entry, excludedBackends) and the -TensorRT- pattern in jenkins/scripts/cbts/blocks.py. - waives.txt: drop 42 waives for deleted/unscheduled tests (the validator requires waived tests to be in an active list). Legacy engine-flow example tests: - Delete the defs/examples wrappers whose every test drove the removed convert->trtllm-build->run flow: test_{bert,chatglm,commandr, draft_target_model,eagle,enc_dec,gptj,granite,internlm,llama,mamba, medusa,mistral,mixtral,multimodal,nemotron,nemotron_nas,ngram, openai,qwen,qwen2audio,qwenvl,redrafter,whisper,bindings,flux}.py, the whisper validators, and run_llm_fp8_quant_llama_70b.py. - test_gpt.py rewritten keeping only the live test_gpt_oss_20b_lora_torch; test_phi.py drops its quantization engine test; test_e2e.py drops test_gpt3_175b_1layers_build_only. - Remove the dead helper cascade in defs/common.py (convert_weights, quantize_data, prune_checkpoint, refit_model, find_tensorrt), defs/perf/build.py and test_perf.py's trtllm-build branches, and 19 orphaned example_root fixtures across defs/conftest.py and triton_server/conftest.py (verified: no dynamic getfixturevalue resolution exists for any of them). Dormant example scripts (imported deleted ModelRunner*/builder surfaces or drove engine flows; unreachable from CI): - examples/{run,summarize,utils,mmlu,eval_long_context}.py, examples/openai_triton/**, examples/python_plugin/**, examples/models/contrib/sdxl/**, examples/llm-eval/lm-eval-harness, the engine-flow multimodal/qwenvl/qwen2audio scripts, and examples/ngram/run_dtm_ngram.py (READMEs rewritten for the PyTorch flow where a live section remains). examples/ reference sweep (checked against tests, published docs, jenkins, and code): - contrib: delete arctic, blip2, chatglm-6b/2-6b/3-6b-32k, internlm, jais, skywork, smaug (README/requirements-only legacy-flow dirs; hyperclovax stays - live PyTorch README linked from release notes). - core: delete granite and mixtral; multimodal keeps the live Qwen-Image-Bench stub + eval script but drops utils.py, __init__.py and the per-model requirements-*.txt; nemotron drops the Nemo-flow README_nemotron-3.md + requirements.txt (PyTorch READMEs stay). - top level: delete generate_checkpoint_config.py, generate_xgrammar_tokenizer_info.py, hf_lora_convert.py, sample_weight_stripping/, cpp_library/, dora/, language_adapter/. - Kept deliberately: infinitebench (dependency of the retained CLI-flow suite) and longbench/opentelemetry/sparse_attention (unreferenced but document current features). Accuracy suite: - accuracy/test_cli_flow.py and CliFlowAccuracyTestHarness are KEPT (currently unscheduled; retained for possible future reuse). - Port TestQwen2_7BInstruct::test_tp2 to the PyTorch accuracy suite (skipped: TP2 hangs in the AUTO custom allreduce on PCIe-only nodes; the LMHead AllReduce ignores allreduce_strategy, so the strategy knob cannot work around it). Misc: - automodel.py isMedusa/isEagle branch and _utils.py::supports_inflight_batching (mapped to deleted TRT model classes); test-db README example refreshed to a real test id; docs fixes in ci-overview.md and the AutoDeploy testing strategy. Rider (squashed in): [https://nvbugs/6482297][fix] Parse boolean stage-config fields in test_to_stage_mapping (from PR NVIDIA#16671, same author). The stage regex only accepted numeric trailing fields, so the 47 runWithSbatch stage entries (all DGX_B200-PyTorch-*) were silently dropped and commented-out stage rows were matched; test_backend_filtering_consistency now tolerates tests legitimately listed under multiple backends. Carried here because this PR retires stages and reshuffles test-db buckets - the exact surface those consistency tests validate. legacy-files.txt/ruff baseline entries for the deleted files are left to a mechanical 'legacy_utils.py prune-files' pass after this and the python-cleanup PR merge, to avoid cross-PR conflicts. Verified: defs collection passes (4206 tests), pre-commit hooks incl. the test-list AST validator pass, no broken example links. Signed-off-by: Wanli Jiang <35160485+Wanli-Jiang@users.noreply.github.com>
1 task
Wanli-Jiang
added a commit
to Wanli-Jiang/TensorRT-LLM
that referenced
this pull request
Jul 22, 2026
… and CI plumbing Trailing test/example/CI cleanup after the legacy TensorRT backend removal (TRTLLM-14026, PR NVIDIA#16369). CI schedules / jenkins: - Move backend-neutral tests off the retired -TensorRT- stages and delete all dead backend=tensorrt test-db buckets; flip remaining neutral buckets to pytorch so parked coverage runs again. - NEW stages A100X-PyTorch-Post-Merge-1 and L40S-PyTorch-Post-Merge-1 schedule the restored a100/l40s post-merge coverage (CI-owner review). - Retire the -TensorRT- plumbing in L0_Test.groovy (commented stage rows, backend=tensorrt mako branch, TEST_BACKEND changeMap entry, excludedBackends) and the -TensorRT- pattern in jenkins/scripts/cbts/blocks.py. - waives.txt: drop 42 waives for deleted/unscheduled tests (the validator requires waived tests to be in an active list). Legacy engine-flow example tests: - Delete the defs/examples wrappers whose every test drove the removed convert->trtllm-build->run flow: test_{bert,chatglm,commandr, draft_target_model,eagle,enc_dec,gptj,granite,internlm,llama,mamba, medusa,mistral,mixtral,multimodal,nemotron,nemotron_nas,ngram, openai,qwen,qwen2audio,qwenvl,redrafter,whisper,bindings,flux}.py, the whisper validators, and run_llm_fp8_quant_llama_70b.py. - test_gpt.py rewritten keeping only the live test_gpt_oss_20b_lora_torch; test_phi.py drops its quantization engine test; test_e2e.py drops test_gpt3_175b_1layers_build_only. - Remove the dead helper cascade in defs/common.py (convert_weights, quantize_data, prune_checkpoint, refit_model, find_tensorrt), defs/perf/build.py and test_perf.py's trtllm-build branches, and 19 orphaned example_root fixtures across defs/conftest.py and triton_server/conftest.py (verified: no dynamic getfixturevalue resolution exists for any of them). Dormant example scripts (imported deleted ModelRunner*/builder surfaces or drove engine flows; unreachable from CI): - examples/{run,summarize,utils,mmlu,eval_long_context}.py, examples/openai_triton/**, examples/python_plugin/**, examples/models/contrib/sdxl/**, examples/llm-eval/lm-eval-harness, the engine-flow multimodal/qwenvl/qwen2audio scripts, and examples/ngram/run_dtm_ngram.py (READMEs rewritten for the PyTorch flow where a live section remains). examples/ reference sweep (checked against tests, published docs, jenkins, and code): - contrib: delete arctic, blip2, chatglm-6b/2-6b/3-6b-32k, internlm, jais, skywork, smaug (README/requirements-only legacy-flow dirs; hyperclovax stays - live PyTorch README linked from release notes). - core: delete granite and mixtral; multimodal keeps the live Qwen-Image-Bench stub + eval script but drops utils.py, __init__.py and the per-model requirements-*.txt; nemotron drops the Nemo-flow README_nemotron-3.md + requirements.txt (PyTorch READMEs stay). - top level: delete generate_checkpoint_config.py, generate_xgrammar_tokenizer_info.py, hf_lora_convert.py, sample_weight_stripping/, cpp_library/, dora/, language_adapter/. - Kept deliberately: infinitebench (dependency of the retained CLI-flow suite) and longbench/opentelemetry/sparse_attention (unreferenced but document current features). Accuracy suite: - accuracy/test_cli_flow.py and CliFlowAccuracyTestHarness are KEPT (currently unscheduled; retained for possible future reuse). - Port TestQwen2_7BInstruct::test_tp2 to the PyTorch accuracy suite (skipped: TP2 hangs in the AUTO custom allreduce on PCIe-only nodes; the LMHead AllReduce ignores allreduce_strategy, so the strategy knob cannot work around it). Misc: - automodel.py isMedusa/isEagle branch and _utils.py::supports_inflight_batching (mapped to deleted TRT model classes); test-db README example refreshed to a real test id; docs fixes in ci-overview.md and the AutoDeploy testing strategy. Rider (squashed in): [https://nvbugs/6482297][fix] Parse boolean stage-config fields in test_to_stage_mapping (from PR NVIDIA#16671, same author). The stage regex only accepted numeric trailing fields, so the 47 runWithSbatch stage entries (all DGX_B200-PyTorch-*) were silently dropped and commented-out stage rows were matched; test_backend_filtering_consistency now tolerates tests legitimately listed under multiple backends. Carried here because this PR retires stages and reshuffles test-db buckets - the exact surface those consistency tests validate. legacy-files.txt/ruff baseline entries for the deleted files are left to a mechanical 'legacy_utils.py prune-files' pass after this and the python-cleanup PR merge, to avoid cross-PR conflicts. Verified: defs collection passes (4206 tests), pre-commit hooks incl. the test-list AST validator pass, no broken example links. Signed-off-by: Wanli Jiang <35160485+Wanli-Jiang@users.noreply.github.com>
Collaborator
Author
|
The fix is merged into #16610 |
Wanli-Jiang
added a commit
to Wanli-Jiang/TensorRT-LLM
that referenced
this pull request
Jul 22, 2026
… and CI plumbing Trailing test/example/CI cleanup after the legacy TensorRT backend removal (TRTLLM-14026, PR NVIDIA#16369). CI schedules / jenkins: - Move backend-neutral tests off the retired -TensorRT- stages and delete all dead backend=tensorrt test-db buckets; flip remaining neutral buckets to pytorch so parked coverage runs again. - NEW stages A100X-PyTorch-Post-Merge-1 and L40S-PyTorch-Post-Merge-1 schedule the restored a100/l40s post-merge coverage (CI-owner review). - Retire the -TensorRT- plumbing in L0_Test.groovy (commented stage rows, backend=tensorrt mako branch, TEST_BACKEND changeMap entry, excludedBackends) and the -TensorRT- pattern in jenkins/scripts/cbts/blocks.py. - waives.txt: drop 42 waives for deleted/unscheduled tests (the validator requires waived tests to be in an active list). Legacy engine-flow example tests: - Delete the defs/examples wrappers whose every test drove the removed convert->trtllm-build->run flow: test_{bert,chatglm,commandr, draft_target_model,eagle,enc_dec,gptj,granite,internlm,llama,mamba, medusa,mistral,mixtral,multimodal,nemotron,nemotron_nas,ngram, openai,qwen,qwen2audio,qwenvl,redrafter,whisper,bindings,flux}.py, the whisper validators, and run_llm_fp8_quant_llama_70b.py. - test_gpt.py rewritten keeping only the live test_gpt_oss_20b_lora_torch; test_phi.py drops its quantization engine test; test_e2e.py drops test_gpt3_175b_1layers_build_only. - Remove the dead helper cascade in defs/common.py (convert_weights, quantize_data, prune_checkpoint, refit_model, find_tensorrt), defs/perf/build.py and test_perf.py's trtllm-build branches, and 19 orphaned example_root fixtures across defs/conftest.py and triton_server/conftest.py (verified: no dynamic getfixturevalue resolution exists for any of them). Dormant example scripts (imported deleted ModelRunner*/builder surfaces or drove engine flows; unreachable from CI): - examples/{run,summarize,utils,mmlu,eval_long_context}.py, examples/openai_triton/**, examples/python_plugin/**, examples/models/contrib/sdxl/**, examples/llm-eval/lm-eval-harness, the engine-flow multimodal/qwenvl/qwen2audio scripts, and examples/ngram/run_dtm_ngram.py (READMEs rewritten for the PyTorch flow where a live section remains). examples/ reference sweep (checked against tests, published docs, jenkins, and code): - contrib: delete arctic, blip2, chatglm-6b/2-6b/3-6b-32k, internlm, jais, skywork, smaug (README/requirements-only legacy-flow dirs; hyperclovax stays - live PyTorch README linked from release notes). - core: delete granite and mixtral; multimodal keeps the live Qwen-Image-Bench stub + eval script but drops utils.py, __init__.py and the per-model requirements-*.txt; nemotron drops the Nemo-flow README_nemotron-3.md + requirements.txt (PyTorch READMEs stay). - top level: delete generate_checkpoint_config.py, generate_xgrammar_tokenizer_info.py, hf_lora_convert.py, sample_weight_stripping/, cpp_library/, dora/, language_adapter/. - Kept deliberately: infinitebench (dependency of the retained CLI-flow suite) and longbench/opentelemetry/sparse_attention (unreferenced but document current features). Accuracy suite: - accuracy/test_cli_flow.py and CliFlowAccuracyTestHarness are KEPT (currently unscheduled; retained for possible future reuse). - Port TestQwen2_7BInstruct::test_tp2 to the PyTorch accuracy suite (skipped: TP2 hangs in the AUTO custom allreduce on PCIe-only nodes; the LMHead AllReduce ignores allreduce_strategy, so the strategy knob cannot work around it). Misc: - automodel.py isMedusa/isEagle branch and _utils.py::supports_inflight_batching (mapped to deleted TRT model classes); test-db README example refreshed to a real test id; docs fixes in ci-overview.md and the AutoDeploy testing strategy. Rider (squashed in): [https://nvbugs/6482297][fix] Parse boolean stage-config fields in test_to_stage_mapping (from PR NVIDIA#16671, same author). The stage regex only accepted numeric trailing fields, so the 47 runWithSbatch stage entries (all DGX_B200-PyTorch-*) were silently dropped and commented-out stage rows were matched; test_backend_filtering_consistency now tolerates tests legitimately listed under multiple backends. Carried here because this PR retires stages and reshuffles test-db buckets - the exact surface those consistency tests validate. legacy-files.txt/ruff baseline entries for the deleted files are left to a mechanical 'legacy_utils.py prune-files' pass after this and the python-cleanup PR merge, to avoid cross-PR conflicts. Verified: defs collection passes (4206 tests), pre-commit hooks incl. the test-list AST validator pass, no broken example links. Signed-off-by: Wanli Jiang <35160485+Wanli-Jiang@users.noreply.github.com>
Wanli-Jiang
added a commit
to Wanli-Jiang/TensorRT-LLM
that referenced
this pull request
Jul 22, 2026
… and CI plumbing Trailing test/example/CI cleanup after the legacy TensorRT backend removal (TRTLLM-14026, PR NVIDIA#16369). CI schedules / jenkins: - Move backend-neutral tests off the retired -TensorRT- stages and delete all dead backend=tensorrt test-db buckets; flip remaining neutral buckets to pytorch so parked coverage runs again. - NEW stages A100X-PyTorch-Post-Merge-1 and L40S-PyTorch-Post-Merge-1 schedule the restored a100/l40s post-merge coverage (CI-owner review). - Retire the -TensorRT- plumbing in L0_Test.groovy (commented stage rows, backend=tensorrt mako branch, TEST_BACKEND changeMap entry, excludedBackends) and the -TensorRT- pattern in jenkins/scripts/cbts/blocks.py. - waives.txt: drop 42 waives for deleted/unscheduled tests (the validator requires waived tests to be in an active list). Legacy engine-flow example tests: - Delete the defs/examples wrappers whose every test drove the removed convert->trtllm-build->run flow: test_{bert,chatglm,commandr, draft_target_model,eagle,enc_dec,gptj,granite,internlm,llama,mamba, medusa,mistral,mixtral,multimodal,nemotron,nemotron_nas,ngram, openai,qwen,qwen2audio,qwenvl,redrafter,whisper,bindings,flux}.py, the whisper validators, and run_llm_fp8_quant_llama_70b.py. - test_gpt.py rewritten keeping only the live test_gpt_oss_20b_lora_torch; test_phi.py drops its quantization engine test; test_e2e.py drops test_gpt3_175b_1layers_build_only. - Remove the dead helper cascade in defs/common.py (convert_weights, quantize_data, prune_checkpoint, refit_model, find_tensorrt), defs/perf/build.py and test_perf.py's trtllm-build branches, and 19 orphaned example_root fixtures across defs/conftest.py and triton_server/conftest.py (verified: no dynamic getfixturevalue resolution exists for any of them). Dormant example scripts (imported deleted ModelRunner*/builder surfaces or drove engine flows; unreachable from CI): - examples/{run,summarize,utils,mmlu,eval_long_context}.py, examples/openai_triton/**, examples/python_plugin/**, examples/models/contrib/sdxl/**, examples/llm-eval/lm-eval-harness, the engine-flow multimodal/qwenvl/qwen2audio scripts, and examples/ngram/run_dtm_ngram.py (READMEs rewritten for the PyTorch flow where a live section remains). examples/ reference sweep (checked against tests, published docs, jenkins, and code): - contrib: delete arctic, blip2, chatglm-6b/2-6b/3-6b-32k, internlm, jais, skywork, smaug (README/requirements-only legacy-flow dirs; hyperclovax stays - live PyTorch README linked from release notes). - core: delete granite and mixtral; multimodal keeps the live Qwen-Image-Bench stub + eval script but drops utils.py, __init__.py and the per-model requirements-*.txt; nemotron drops the Nemo-flow README_nemotron-3.md + requirements.txt (PyTorch READMEs stay). - top level: delete generate_checkpoint_config.py, generate_xgrammar_tokenizer_info.py, hf_lora_convert.py, sample_weight_stripping/, cpp_library/, dora/, language_adapter/. - Kept deliberately: infinitebench (dependency of the retained CLI-flow suite) and longbench/opentelemetry/sparse_attention (unreferenced but document current features). Accuracy suite: - accuracy/test_cli_flow.py and CliFlowAccuracyTestHarness are KEPT (currently unscheduled; retained for possible future reuse). - Port TestQwen2_7BInstruct::test_tp2 to the PyTorch accuracy suite (skipped: TP2 hangs in the AUTO custom allreduce on PCIe-only nodes; the LMHead AllReduce ignores allreduce_strategy, so the strategy knob cannot work around it). Misc: - automodel.py isMedusa/isEagle branch and _utils.py::supports_inflight_batching (mapped to deleted TRT model classes); test-db README example refreshed to a real test id; docs fixes in ci-overview.md and the AutoDeploy testing strategy. Rider (squashed in): [https://nvbugs/6482297][fix] Parse boolean stage-config fields in test_to_stage_mapping (from PR NVIDIA#16671, same author). The stage regex only accepted numeric trailing fields, so the 47 runWithSbatch stage entries (all DGX_B200-PyTorch-*) were silently dropped and commented-out stage rows were matched; test_backend_filtering_consistency now tolerates tests legitimately listed under multiple backends. Carried here because this PR retires stages and reshuffles test-db buckets - the exact surface those consistency tests validate. legacy-files.txt/ruff baseline entries for the deleted files are left to a mechanical 'legacy_utils.py prune-files' pass after this and the python-cleanup PR merge, to avoid cross-PR conflicts. Verified: defs collection passes (4206 tests), pre-commit hooks incl. the test-list AST validator pass, no broken example links. Signed-off-by: Wanli Jiang <35160485+Wanli-Jiang@users.noreply.github.com>
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
2 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.
The stage-parsing regex in scripts/test_to_stage_mapping.py only accepted trailing numeric fields in Jenkins stage-config arrays, so the 47 stage entries carrying the runWithSbatch boolean (added by the Slurm test refactor #7176) were silently dropped from the test<->stage mapping, including all DGX_B200-PyTorch-* stages. It also matched commented-out stage lines. As a result, tests listed in l0_b200.yml resolved to no stage and test_cli_functionality failed whenever the seeded sample picked one of them.
Summary by CodeRabbit
Description
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)
If PR introduces API changes, an appropriate PR label is added - either
api-compatibleorapi-breaking. Forapi-breaking, includeBREAKINGin the PR title.Any new dependencies have been scanned for license and vulnerabilities
CODEOWNERS updated if ownership changes
Documentation updated as needed
Update tava architecture diagram if there is a significant design change in PR.
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
To see a list of available CI bot commands, please comment
/bot help.