[TRTLLM-13783][test] Remove TensorRT-backend tests

[Wanli-Jiang]

PR3.1 test changes — per-file record

Commit 425ca67b0a · [TRTLLM-13783][test] Remove TensorRT-backend tests
Base: main @ e5a05b25be · 161 files, +559 / −40,887

Every change below is under tests/; no runtime source (tensorrt_llm/**, cpp/**)
is touched. Totals: 143 removed · 2 relocated · 2 added · 14 modified.

Legend: Removed = deleted; Relocated = git-renamed to a surviving location (kept
because it tests PyTorch code); Added = new file; Modified = edited in place.

---

Added (2)

File	Reason
tests/unittest/_torch/thop/serial/test_nvfp4_marlin_gemm.py	Rescued from the deleted trt/functional/test_fp4_gemm.py::test_nvfp4_marlin_gemm. Tests the live PyTorch op torch.ops.trtllm.marlin_nvfp4_gemm (+ gptq_marlin_repack, block_scale_interleave_reverse, marlin_utils) against a torch.matmul reference — no TRT engine. Had no surviving coverage elsewhere, so it is re-homed rather than lost. TRT-only imports dropped; reference kept as torch.matmul.
tests/integration/defs/triton_server/test_triton_llmapi.py	New PyTorch-path Triton coverage (test_llmapi_backend TP1/TP4, test_llmapi_lora, test_llmapi_backend_multi_instance) that replaces the removed TRT-engine Triton harness (test_triton_llm.py etc.).

Relocated (2) — PyTorch tests kept out of the cull

Both lived under trt/quantization/ but exercise surviving PyTorch code, so they are
git-renamed (not deleted). Reformatted on move because their new paths are not in
pyproject's ruff legacy-exclude (tests/unittest/trt/* was), so ruff-format now applies.

From → To	What it tests	Reason to keep
trt/quantization/test_fp8_quantization.py → _torch/thop/serial/test_fp8_quantization.py (R077)	PyTorch custom ops torch.ops.{tensorrt_llm,trtllm}.quantize_e4m3_* + _torch.autotuner	No TRT engine; op is live in the PyTorch backend
trt/quantization/test_mode.py → quantization/test_mode.py (R095)	QuantMode config enum	Config enum used by the PyTorch backend; backend-agnostic

Modified (14)

File	Change	Reason
tests/integration/test_lists/test-db/l0_a10.yml	Drop dangling TRT entries; add unittest/quantization/test_mode.py	Remove refs to deleted tests; register the relocated test_mode (it had no CI entry before or after the move — closes a pre-existing gap)
tests/integration/test_lists/test-db/l0_a100.yml	Drop dangling TRT entries	Referenced tests no longer exist on disk
tests/integration/test_lists/test-db/l0_a30.yml	Drop dangling TRT entries	Referenced tests no longer exist on disk
tests/integration/test_lists/test-db/l0_b200.yml	Drop dangling TRT entries	Referenced tests no longer exist on disk
tests/integration/test_lists/test-db/l0_dgx_h200.yml	Drop dangling TRT entries	Referenced tests no longer exist on disk
tests/integration/test_lists/test-db/l0_gb203.yml	Drop dangling TRT entries	Referenced tests no longer exist on disk
tests/integration/test_lists/test-db/l0_gh200.yml	Drop dangling TRT entries	Referenced tests no longer exist on disk
tests/integration/test_lists/test-db/l0_h100.yml	Drop dangling TRT entries	Referenced tests no longer exist on disk
tests/integration/test_lists/test-db/l0_l40s.yml	Drop dangling TRT entries	Referenced tests no longer exist on disk
tests/integration/test_lists/test-db/l0_sanity_check.yml	Drop dangling TRT entries	Referenced tests no longer exist on disk
tests/integration/test_lists/waives.txt	Drop stale TRT waives	Waives pointed at now-deleted TRT tests
tests/integration/test_lists/qa/llm_triton_integration.txt	Replace TRT-engine Triton cases with test_triton_llmapi.py cases	TRT-engine Triton tests removed; list the PyTorch-path replacement
tests/integration/defs/llmapi/test_llm_examples.py	Remove test_llmapi_tensorrt_engine	Drove a TRT-engine build via an already-deleted example
tests/integration/defs/triton_server/test_triton.py	Drop refs to removed TRT-engine Triton tests	Those test files are deleted

---

Removed (143)

Shared rationale: each targets TensorRT-backend code only — engine build, the
tensorrt_llm.functional graph layer, TRT plugins, or the TRT runtime. None test the
PyTorch backend. Where the behavior also matters for PyTorch, it is already covered by a
surviving _torch/** suite (noted per group).

trt/functional/ — 70 (69 op tests + __init__.py)

Reason: each builds a TRT INetworkDefinition (net_guard/create_session) around a
tensorrt_llm.functional op and checks the engine output. The op exists only for TRT
graph construction — there is no PyTorch equivalent to re-cover. (The one live-op method,
test_nvfp4_marlin_gemm, was rescued to _torch/thop/serial/ — see Added.)

__init__ · test_alibi · test_allreduce_norm · test_allreduce_prepost_residual_norm ·
test_arange · test_argmax · test_assertion · test_avg_pool2d · test_cast · test_conv2d ·
test_conv3d · test_cos · test_cumsum · test_dora · test_einsum · test_embedding_single_gpu ·
test_exp · test_expand · test_flatten · test_flip · test_fp4_gemm · test_fp4_gemm_ootb ·
test_gather · test_gather_nd · test_geglu · test_gelu · test_gemm_swiglu · test_group_norm ·
test_identity · test_index_select · test_interpolate · test_logsoftmax · test_lora ·
test_low_latency_gemm · test_mamba_conv1d · test_masked_scatter · test_masked_select ·
test_matmul · test_meshgrid2d · test_moe · test_nccl · test_nonzero · test_outer · test_pad ·
test_permute · test_pp_reduce_scatter · test_quant · test_rearrange · test_repeat ·
test_repeat_interleave · test_rg_lru · test_sample · test_scatter · test_scatter_nd ·
test_select · test_selective_scan · test_sigmoid · test_silu · test_sin · test_slice ·
test_softplus · test_split · test_squeeze · test_swiglu · test_topk · test_transpose ·
test_unbind · test_unsqueeze · test_view · test_where

trt/model/ + trt/model_api/ — 26

Reason: build a tensorrt_llm.models PretrainedModel → TRT engine → run. Covered by
tests/unittest/_torch/modeling/. The eagle/redrafter files test TRT plugins for
those speculative paths (_torch/speculative/ covers the PyTorch flow).

File	Note
trt/model/__init__.py	package marker
trt/model/eagle/test_{decode_draft_tokens,prepare_drafter_inputs,sample_accept_draft_tokens}_plugin.py (3)	TRT Eagle plugins
trt/model/redrafter/test_{beams2tree,draft_token,draft_token_indices,gather_beams,mask,packed_position_ids,prefix_match_indices,prepare_input,process_logits,top1,unpack_gen_data,validate}.py (12)	TRT ReDrafter plugin/graph tests
trt/model/test_{gpt,gpt_e2e,llama,mamba,mistral,nemotron_nas,phi,unet}.py (8)	PretrainedModel → engine → run
trt/model_api/profile_utils.py, trt/model_api/test_model_api_multi_gpu.py (2)	TRT model-API multi-GPU engine build + helper

trt/quantization/ — 13

Reason: TRT quant GEMM / plugin engine tests + helpers. PyTorch quant is covered by
_torch/thop/ and _torch/modules/. (The two non-TRT files here — test_fp8_quantization
and test_mode — were relocated, not deleted.)

__init__ · _utils · test_fp8_rowwise_gemm · test_functional · test_moe_weight_only_quant_matmul ·
test_qserve_gemm · test_quant · test_quant_layer · test_smooth_quant_gemm ·
test_smooth_quant_layer_norm · test_smooth_quant_rms_norm ·
test_weight_only_groupwise_quant_matmul · test_weight_only_quant_matmul

trt/attention/ — 6

Reason: TRT gpt/bert attention plugin engine tests. PyTorch attention is covered by
_torch/attention/.

hf_dynamic_cache_compat · test_bert_attention · test_gpt_attention · test_gpt_attention_IFB ·
test_gpt_attention_no_cache · test_sage_attention

trt/python_plugin/ — 2 · trt/__init__.py — 1

Reason: TRT Python-plugin wrapper tests + helper (plugin_wrapper_utils.py,
test_plugin_wrapper.py) and the trt/ package marker. No PyTorch equivalent.

unittest/others/ — 10

Reason: exercise TRT-only machinery. No PyTorch equivalent except KV cache (covered by
kv_cache_manager_v2_tests/).

File	What it tested
test_builder.py	TRT Builder
test_debugging_api.py	TRT network debugging API
test_graph_rewriter.py	TRT graph rewriter
test_kv_cache_manager.py	old runtime.kv_cache_manager (TRT generation.py) — PyTorch KV cache is kv_cache_manager_v2_tests/
test_layer.py	TRT tensorrt_llm.layers graph construction
test_model_dtype.py	TRT PretrainedModel dtype handling
test_module.py	TRT Module base class
test_plugins.py	TRT plugin registration
test_precision_control.py	TRT per-layer precision control
test_session.py	TRT runtime.Session

unittest/llmapi/ — 3

File	Reason
test_build_cache.py	TRT engine-build cache
test_llm_models.py	_tensorrt_engine.LLM engine builds per model
test_llm_multi_gpu.py	BuildConfig-driven multi-GPU engine builds

unittest/tools/plugin_gen/ — 3

Reason: the TRT plugin-generator tool (test_core.py, test_plugin_gen.py,
test_shape_infer.py). No PyTorch equivalent.

unittest/_torch/modules/tests_lora_modules/ — 2

Reason: these validate surviving PyTorch LoRA code but use a built TRT engine as the
numeric reference, so they are structurally TRT-coupled and cannot run once the backend is
gone. Kept deleted; the numeric coverage should be re-added against a pure-torch reference
in a follow-up.

File	Validated	Reference
test_lora_plugin_vs_lora_op.py	torch.ops.trtllm.lora_grouped_gemm numerics	TRT lora_plugin
test_lora_attention_pytorch_flow_vs_trt.py	_torch LlamaAttention + LoRA + KVCacheManager flow	built TRT engine

integration/defs/triton_server/ — 5

Reason: TRT-engine Triton harness (trtllm-build + C++ Triton backend). Replaced by the
added test_triton_llmapi.py.

build_engines · test_triton_llm · test_triton_memleak · test_triton_multi_node · test_triton_rcca

Top-level — 2

File	Reason
unittest/test_model_runner_cpp.py	ModelRunnerCpp over a TRT engine — PyTorch path is _torch/pyexecutor
unittest/utils/test_medusa_utils.py	runtime.medusa_utils (TRT) — PyTorch Medusa is _torch/speculative/

---

Verification

• 75 relocated/added unit tests pass on an H200 (test_nvfp4_marlin_gemm,
  test_fp8_quantization, test_mode); test_triton_llmapi.py is collection-clean
  (executes in the Triton integration CI stage).
• scripts/check_test_list.py --validate passes — no test-list entry references a deleted
  test; the new test_mode entry resolves.
• CI coverage after this commit: marlin + fp8 via the unittest/_torch/thop/serial dir entry
  (H100), test_triton_llmapi via qa/llm_triton_integration.txt, test_mode via l0_a10.yml.

PR3.2 waived-examples cleanup — per-file record

Commit bdb9795a55 · [TRTLLM-13987][test] Remove waived legacy TensorRT examples tests
Base: f33293814e ([TRTLLM-13783][test] Remove TensorRT-backend tests)
Scope: 10 files, −303 lines, 0 additions. Every change is under
tests/integration/test_lists/; no runtime source and no test .py file is touched.

---

Why this commit exists

PR #15763 ([TRTLLM-13781] Remove legacy TensorRT examples) deleted the TensorRT
example scripts but could not delete the tests that drove them in the same PR, so it
waived every one of those tests with the marker:

SKIP (TRTLLM-13781: legacy TensorRT examples removed; tests to be removed in follow-up PR3)

That produced 140 dead (skipped) test entries. This commit is that follow-up: it removes
the waived entries from the CI schedule lists and their waive lines, so the dead
entries stop being carried. It was prompted by review feedback (xinhe-nv + an automated
reviewer) asking whether the remaining TRT-flow tests were tracked.

The core invariant

A test runs iff it appears in a test-db/*.yml (or qa/) schedule list; a waives.txt
line only forces SKIP. So fully removing a waived test requires editing both:

• Remove schedule entry only → nothing skips it → it runs against a deleted example → fails.
• Remove waive only → validate-test-lists pre-commit hook fails ("WAIVE NOT IN ACTIVE TEST LISTS") and, if still scheduled, it runs and fails.
• Remove both → clean. ✅

test_executor_bindings.py is the one case where "both" is impossible (see below), so it is
intentionally left waived.

---

Counts at a glance

Bucket	Count
Distinct waived tests targeted (TRTLLM-13781 batch)	140
— removed here (schedule + waive)	139
— intentionally kept waived	1 (test_executor_bindings.py)
Schedule-list entries removed (test-db), net	143
Waive lines removed from waives.txt	141 (139 TRTLLM-13781 + 2 orphaned platform waives)
Empty condition blocks pruned	1 (l0_a100.yml)
Shared C++ unit tests re-activated (un-waived; see correction below)	6

Schedule math: the 139 unique waived tests map to 149 list lines (ten tests are
scheduled on more than one GPU list — e.g. examples/test_whisper.py on a10/a100/h100/l40s).
Of those 149, 6 turned out to be example-independent shared C++ unit tests that were
over-waived by #15763; they are restored as active rather than deleted, leaving 143
net schedule-line removals.

---

Modified files (10)

Schedule lists — test-db/*.yml (9 files, −162 lines total)

Each removed line is a schedule entry whose test is in the TRTLLM-13781 waived batch. The
matching waive line is removed from waives.txt in the same commit. Grouped by the test
file each removed entry belongs to:

List file	−lines	Removed entries (by test file)
l0_a10.yml	5	cpp/test_e2e.py ×3, examples/test_bert.py ×1, examples/test_whisper.py ×1
l0_a100.yml	17	examples/test_whisper.py ×4 + 13-line empty tensorrt condition block (pruned)
l0_a30.yml	39	accuracy/test_cli_flow.py ×7, cpp/test_e2e.py ×2, examples/test_commandr.py ×2, examples/test_draft_target_model.py ×4, examples/test_enc_dec.py ×12, examples/test_gpt.py ×8, examples/test_medusa.py ×1, examples/test_multimodal.py ×2, examples/test_redrafter.py ×1
l0_b200.yml	7	accuracy/test_cli_flow.py ×7
l0_dgx_h100.yml	29	cpp/test_multi_gpu.py ×35 removed − 6 shared unit tests restored as active (net 29)
l0_dgx_h200.yml	12	accuracy/test_cli_flow.py ×7, examples/test_enc_dec.py ×2, examples/test_gpt.py ×1, examples/test_llama.py ×1, examples/test_mamba.py ×1
l0_gh200.yml	1	accuracy/test_cli_flow.py ×1
l0_h100.yml	32	accuracy/test_cli_flow.py ×4, cpp/test_e2e.py ×4, examples/test_eagle.py ×2, examples/test_enc_dec.py ×5, examples/test_granite.py ×1, examples/test_medusa.py ×2, examples/test_multimodal.py ×8, examples/test_phi.py ×1, examples/test_qwen.py ×1, examples/test_redrafter.py ×4
l0_l40s.yml	14	accuracy/test_cli_flow.py ×3, examples/test_gpt.py ×2, examples/test_granite.py ×1, examples/test_llama.py ×1, examples/test_medusa.py ×2, examples/test_qwen.py ×1, examples/test_redrafter.py ×4

Special case — l0_a100.yml empty-block prune. Removing the 1-GPU tensorrt block's
only surviving tests would have left a tests: key with no list items. YAML parses that as
null, and trt-test-db then crashes in spread() with
TypeError: 'NoneType' object is not iterable (the exact failure that motivated the
l0_dgx_h200.yml fix in the prior commit). So the whole now-empty - condition: block was
deleted, not just its test lines. This is why l0_a100.yml shows 17 deletions for only 4
test entries.

waives.txt (−141 lines)

Removed	Count	Rationale
... SKIP (TRTLLM-13781: legacy TensorRT examples removed; ...)	139	The batch this commit resolves. Grouped by test file: cpp/test_multi_gpu.py ×35, accuracy/test_cli_flow.py ×28, examples/test_enc_dec.py ×19, examples/test_gpt.py ×11, examples/test_multimodal.py ×10, cpp/test_e2e.py ×9, examples/test_whisper.py ×4, examples/test_redrafter.py ×4, examples/test_draft_target_model.py ×4, examples/test_medusa.py ×3, examples/test_qwen.py ×2, examples/test_llama.py ×2, examples/test_eagle.py ×2, examples/test_commandr.py ×2, examples/test_phi.py ×1, examples/test_mamba.py ×1, examples/test_granite.py ×1, examples/test_bert.py ×1
full:GH200/...video-neva... SKIP (https://nvbugs/4731514)	1	Orphaned platform waive. The video-neva multimodal test carried three waives — the TRTLLM-13781 one plus two full:<platform>/-scoped ones. Once its only schedule entry (l0_h100.yml) was removed, this GH200-scoped waive pointed at a test in no active list, so validate-test-lists failed. Removed.
full:sm100/...video-neva... SKIP (megatron-core 0.8 is not supported in python 3.12)	1	Same test, sm100-scoped waive; same orphaning. Removed.

The 2 orphaned platform waives were not in the original TRTLLM-13781 batch — they were
found only because the validate-test-lists pre-commit hook flagged them after the first
removal pass. A comprehensive re-scan (strip full:<platform>/ prefix, compare base test
id against the removed set) confirmed these were the only two.

---

Correction — 6 over-waived shared C++ unit tests restored as active

While reviewing tests/integration/defs/cpp/, six entries in the TRTLLM-13781 batch turned
out not to depend on examples at all — #15763 over-waived them by bulk-waiving the whole
cpp/test_multi_gpu.py file. They are restored as active in l0_dgx_h100.yml (un-waived),
not deleted:

cpp/test_multi_gpu.py::test_mpi_utils[90]
cpp/test_multi_gpu.py::test_fused_gemm_allreduce[4proc-90]
cpp/test_multi_gpu.py::test_cache_transceiver[2proc-ucx_kvcache-90] ISOLATION
cpp/test_multi_gpu.py::test_cache_transceiver[8proc-nixl_kvcache-90] ISOLATION
cpp/test_multi_gpu.py::test_cache_transceiver[8proc-ucx_kvcache-90] ISOLATION
cpp/test_multi_gpu.py::test_user_buffer[2proc-90]

Evidence they are example-independent:

• Their pytest signatures take only build_google_tests + build_dir (+ nprocs/kvcache_type) —
  not multi_gpu_model / prepare_models_disagg, the fixtures that build a TRT engine from
  example checkpoints (via cpp_common.prepare_model_tests → examples/…/build_*_engines.py).
• build_google_tests only compiles the C++ google-tests target; no engine, no example.
• They run pure C++ gtest binaries with synthetic inputs from build/tests/unit_tests/multi_gpu/:
  mpiUtilsTest, kernels/gemmAllReduceTest --m=2032 --n=8200 --k=1024, cacheTransceiverTest,
  userBufferTest. No model / engine / checkpoint reference.
• The only autouse fixture (keep_log_files) just backs up XML results.
• Git history: before [TRTLLM-13781][chore] Remove legacy TensorRT examples #15763 these were scheduled and running in l0_dgx_h100.yml; [TRTLLM-13781][chore] Remove legacy TensorRT examples #15763
  waived them under "legacy TensorRT examples removed", and this commit's first pass then removed
  the dead entries. Net path was running → waived → removed — an unintended coverage loss for
  non-deprecated multi-GPU C++ (MoE GEMM-allreduce, disagg KV-cache transceiver, user buffers,
  MPI utils). Restoring them reverts that regression.

test_cache_transceiver[8proc-mooncake_kvcache-90] keeps its real waive (nvbugs/5838199) —
that one is unrelated to examples. The engine-based test_multi_gpu.py cases
(test_llama_executor*, test_enc_dec, test_trt_gpt_real_decoder, TestDisagg::*) stay
removed/waived; they genuinely need the removed example engines and belong with the later C++
TRT-engine-path removal.

Intentionally NOT changed

Kept waived: unittest/bindings/test_executor_bindings.py

The one member of the TRTLLM-13781 batch left in waives.txt. Reasons:

• It is a whole-file waive, and the file is scheduled via the unittest/bindings
  directory entry (in l0_a10.yml, l0_gh200.yml, l0_h100.yml) — there is no
  per-test line to delete, so it can't be descheduled without either a waive or deleting the file.
• Every test in it constructs trtllm.Executor from a built rank0.engine (its
  model_files fixture generates engines via the removed example build path), so removing
  the waive would un-skip a test that cannot build its engine → CI failure.
• Its correct removal belongs with the C++ Executor-bindings removal (later PR), not a
  test-list cleanup. Keeping the waive is the lowest-risk holding state.

Test .py files untouched

No test source file was deleted. Files like examples/test_gpt.py, test_medusa.py,
test_multimodal.py, etc. still hold surviving PyTorch-path cases (e.g.
examples/test_gpt.py::test_gpt_oss_20b_lora_torch). Only the specific TRT-flow
parametrizations were unscheduled/unwaived. Deleting now-orphaned example test files (those
whose every case was TRT-only) is left as a possible later cleanup.

---

Verification

• YAML: all test-db/*.yml parse; no condition block has a null/empty tests:.
• No dangling schedule entries: none of the 139 removed test ids appear in any schedule
  list (test-db/*.yml, qa/*).
• No orphaned waives: no remaining waives.txt line (after stripping any
  full:<platform>/ prefix) references a removed test id.
• Pre-commit: full hook suite passes, including
  Validate test list entries exist in source files (AST) (validate-test-lists) and
  Checks for duplicated test items in waives.txt.
• Kept waive present: the single test_executor_bindings.py TRTLLM-13781 waive remains.

Summary by CodeRabbit

• New Features
  • Expanded end-to-end Triton LLM API backend integration coverage, including LoRA and multi-instance scenarios.
  • Added a new FP8/NVFP4 GEMM unit test.
• Bug Fixes / Testing
  • Removed or trimmed several integration and unit tests to align the suite with current validated coverage.
  • Updated integration test selections and waivers accordingly.
• Style
  • Applied minor formatting cleanups in a few existing unit tests.

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-compatible or api-breaking. For api-breaking, include BREAKING in 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.

[Wanli-Jiang]

/bot run --disable-fail-fast

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR removes legacy TensorRT-based integration and unit tests, adds a Triton llmapi integration module, updates test lists and waivers, and adds a new nvfp4 Marlin GEMM unit test with minor quantization formatting changes.

Changes

Triton LLM API Integration Migration

Layer / File(s)	Summary
New llmapi integration coverage tests/integration/defs/triton_server/test_triton_llmapi.py	Adds Triton llmapi backend, LoRA, and multi-instance integration tests with server setup and client-side validation.
Legacy Triton integration removals tests/integration/defs/llmapi/test_llm_examples.py, tests/integration/defs/triton_server/test_triton.py, tests/integration/test_lists/qa/llm_triton_integration.txt, tests/integration/test_lists/test-db/l0_sanity_check.yml, tests/integration/test_lists/waives.txt	Removes test_llmapi_tensorrt_engine, test_triton_extensive, and related list/waiver entries.
Integration test-db pruning tests/integration/test_lists/test-db/*.yml	Trims GPU-specific integration test lists across A10, A100, A30, B200, DGX, GB203, GH200, H100, L40S, and sanity-check configs.
Waiver list updates tests/integration/test_lists/waives.txt	Updates waiver entries for deleted tests and rewrites several cpp, example, sm100, triton, and llmapi blocks.

Unit Test Cleanup and Additions

Layer / File(s)	Summary
TRT functional test removals tests/unittest/trt/functional/*.py, tests/unittest/trt/attention/*.py, tests/unittest/trt/functional/test_*	Removes TRT functional coverage for attention, math, tensor ops, communication, quantization, and related helpers.
Other unit test removals tests/unittest/llmapi/*.py, tests/unittest/others/*.py, tests/unittest/tools/plugin_gen/*.py, tests/unittest/trt/attention/hf_dynamic_cache_compat.py	Deletes llmapi, tooling, plugin-gen, cache, model dtype, debugging, and other legacy unit tests/helpers.
nvfp4 Marlin and quant mode tests/unittest/_torch/thop/serial/test_nvfp4_marlin_gemm.py, tests/unittest/quantization/test_mode.py	Adds a new marlin_nvfp4_gemm unit test and adjusts quant mode assertions/formatting.

Estimated code review effort: 4 (Complex) | ~45 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#13262: Also prunes Triton integration QA manifest entries tied to the Triton test matrix.
• Suggested reviewers: Superjomn, brb-nv, QiJune, yuxianq, litaotju

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The Description and Test Coverage sections are left empty, so the PR narrative is incomplete.	Add a short description of the issue and solution, plus the relevant tests or validation run for these changes.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title is concise and accurately summarizes the main change: removing TensorRT-backend tests.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}

[QiJune]

LGTM

[xinhe-nv]

@coderabbitai review the pr, if there is missing tensorrt flow tests

[coderabbitai]

Tip

For best results, initiate chat on the files or code changes.

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
set -uo pipefail

echo "=== Remaining files under tests/unittest/trt/ ==="
fd . tests/unittest/trt 2>/dev/null

echo "=== Remaining trt-only test dirs ==="
fd -t d 'trt' tests 2>/dev/null

echo "=== References to build_engines module ==="
rg -n 'build_engines' tests --type=py -g '!*.pyc'

echo "=== References to ModelRunnerCpp in tests ==="
rg -n 'ModelRunnerCpp' tests

echo "=== References to medusa_utils ==="
rg -n 'medusa_utils' tests

echo "=== References to old runtime.kv_cache_manager ==="
rg -n 'kv_cache_manager' tests | grep -v '_torch'

echo "=== dangling references to deleted test names in test-lists ==="
rg -n 'test_gpt_attention|test_bert_attention|test_llm_models|test_llm_multi_gpu|test_model_runner_cpp|test_build_cache|plugin_gen' tests/integration/test_lists

echo "=== check waives.txt for dangling refs to trt tests ==="
rg -n 'unittest/trt/|triton_server/test_triton_rcca|triton_server/test_triton_memleak|triton_server/test_triton_multi_node|test_lora_attention_pytorch_flow_vs_trt|test_lora_plugin_vs_lora_op' tests/integration/test_lists/waives.txt

echo "=== check any remaining .yml referencing triton_server/test_triton_llm.py (old removed file?) ==="
rg -n 'test_triton_llm\.py' tests/integration

Length of output: 53737

---

🏁 Script executed:

#!/bin/bash
set -uo pipefail

echo "=== Context around ModelRunnerCpp usage in test_e2e.py ==="
sed -n '80,120p' tests/integration/defs/test_e2e.py

echo "=== function name using ModelRunnerCpp in test_e2e.py ==="
rg -n 'def test_' tests/integration/defs/test_e2e.py | awk -F: '{print $1":"$2}' | head -1
ast-grep run --pattern 'def $FUNC($$$) { $$$ }' --lang python tests/integration/defs/test_e2e.py 2>/dev/null | head -5

echo "=== Check whisper test wiring in test-lists ==="
rg -n 'test_whisper|validate_whisper' tests/integration/test_lists/test-db/*.yml

echo "=== medusa cpp session entries in l0_gh200.yml ==="
rg -n 'medusa' tests/integration/test_lists/test-db/l0_gh200.yml

echo "=== Check test_medusa.py for TRT engine building (kept file) ==="
fd test_medusa.py tests
rg -n 'ModelRunnerCpp|trtllm-build|engine_dir' tests/integration/defs/examples/test_medusa.py | head -30

Length of output: 5272

---

🏁 Script executed:

#!/bin/bash
set -uo pipefail

echo "=== check test_triton.py::test_whisper still present? ==="
rg -n 'def test_whisper' tests/integration/defs/triton_server/test_triton.py

echo "=== full test_triton.py current tests ==="
rg -n 'def test_' tests/integration/defs/triton_server/test_triton.py

echo "=== check test_e2e.py::test_gpt3_175b_1layers_build_only wired in test-lists ==="
rg -n 'test_gpt3_175b_1layers_build_only' tests/integration/test_lists -r

echo "=== check for other trtllm-build usages remaining under tests/integration/defs (examples) ==="
rg -l 'trtllm-build' tests/integration/defs/examples | sort

Length of output: 5115

---

@xinhe-nv Here's what I found after reviewing the PR for remaining TensorRT-flow test coverage:

Fully removed ✅

• All of tests/unittest/trt/** (attention/functional/model tests) are gone.
• tests/unittest/others/test_kv_cache_manager.py, test_layer.py, test_module.py, test_plugins.py, etc. are gone.
• tests/integration/defs/triton_server/build_engines.py (the TRT-engine-preparation helper module) is deleted, along with test_triton_rcca.py, test_triton_memleak.py, test_triton_multi_node.py.
• Test-lists and waives.txt don't have any dangling references to the deleted files — validated clean.

Still present (TensorRT-engine-based flow tests not removed by this PR) ⚠️

1. tests/integration/defs/triton_server/test_triton.py — only test_triton_extensive was removed. It still contains ~25 TRT-engine-building test functions that call build_engines.py-equivalent fixtures indirectly (e.g., test_gpt, test_mistral, test_gptj, test_medusa, test_eagle, test_whisper, test_t5_ib, test_bart_ib, test_blip2_opt, test_mllama, test_llava*, test_qwen2_vl) at Lines 112-427. These are still referenced from tests/integration/test_lists/test-db/l0_a30.yml (triton_server/test_triton.py::test_whisper[whisper]).
2. tests/integration/defs/examples/ — 22 files still call trtllm-build directly to construct TensorRT engines: test_bert.py, test_chatglm.py, test_commandr.py, test_draft_target_model.py, test_eagle.py, test_enc_dec.py, test_gpt.py, test_granite.py, test_llama.py, test_mamba.py, test_medusa.py, test_mistral.py, test_mixtral.py, test_multimodal.py, test_nemotron.py, test_nemotron_nas.py, test_ngram.py, test_phi.py, test_qwen.py, test_qwen2audio.py, test_redrafter.py, test_whisper.py.
3. tests/integration/defs/test_e2e.py::test_gpt3_175b_1layers_build_only still uses trtllm-build + the legacy ModelRunnerCpp runtime.
4. validate_whisper_log_probs_determinism.py and validate_whisper_beam_logits.py still import/use ModelRunnerCpp.
5. These are still actively wired into l0_gh200.yml (medusa cpp/py session), l0_a10.yml/l0_a100.yml/l0_h100.yml/l0_l40s.yml (test_whisper cpp/py runtime), and l0_a30.yml (triton_server/test_triton.py::test_whisper).

So while this PR removes the pure-unittest TRT coverage (tests/unittest/trt/**) and a portion of Triton-engine integration tests, there's a substantial remaining surface of TensorRT-engine-build integration tests under tests/integration/defs/examples/ and tests/integration/defs/triton_server/test_triton.py that weren't touched here. If the deprecation is meant to be comprehensive, these would need a follow-up PR; if they're intentionally deferred (e.g., because they still exercise model conversion/build paths not yet ported), it'd be good to confirm that scope explicitly in the PR description.

Would you like me to open a follow-up issue tracking the remaining TensorRT-engine-based integration tests (tests/integration/defs/examples/*.py, test_triton.py, test_e2e.py::test_gpt3_175b_1layers_build_only) for a future cleanup PR?

(๑•̀ㅁ•́๑)✧ investigating mode engaged

[tensorrt-cicd]

PR_Github #57192 [ run ] triggered by Bot. Commit: 425ca67 Link to invocation

[tensorrt-cicd]

PR_Github #57192 [ run ] completed with state SUCCESS. Commit: 425ca67
/LLM/main/L0_MergeRequest_PR pipeline #45964 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[Wanli-Jiang]

/bot run --disable-fail-fast

[coderabbitai]

Actionable comments posted: 1

🤖 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.

Inline comments:
In `@tests/integration/test_lists/waives.txt`:
- Around line 325-326: The waiver entries are being parsed as a single test id
per line, so this combined waiver in the waives list only records one malformed
skip instead of multiple tests. Split the combined entry into separate lines in
the waives file, with one test id or skip entry per line, so each affected test
is actually waived.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: b663f624-6317-4e56-bf62-7fefba8a3cf4

📥 Commits

Reviewing files that changed from the base of the PR and between fb0d68b and c955d12.

📒 Files selected for processing (162)

• tests/integration/defs/llmapi/test_llm_examples.py
• tests/integration/defs/triton_server/build_engines.py
• tests/integration/defs/triton_server/test_triton.py
• tests/integration/defs/triton_server/test_triton_llm.py
• tests/integration/defs/triton_server/test_triton_llmapi.py
• tests/integration/defs/triton_server/test_triton_memleak.py
• tests/integration/defs/triton_server/test_triton_multi_node.py
• tests/integration/defs/triton_server/test_triton_rcca.py
• tests/integration/test_lists/qa/llm_triton_integration.txt
• tests/integration/test_lists/test-db/l0_a10.yml
• tests/integration/test_lists/test-db/l0_a100.yml
• tests/integration/test_lists/test-db/l0_a30.yml
• tests/integration/test_lists/test-db/l0_b200.yml
• tests/integration/test_lists/test-db/l0_dgx_h100.yml
• tests/integration/test_lists/test-db/l0_dgx_h200.yml
• tests/integration/test_lists/test-db/l0_gb203.yml
• tests/integration/test_lists/test-db/l0_gh200.yml
• tests/integration/test_lists/test-db/l0_h100.yml
• tests/integration/test_lists/test-db/l0_l40s.yml
• tests/integration/test_lists/test-db/l0_sanity_check.yml
• tests/integration/test_lists/waives.txt
• tests/unittest/_torch/modules/tests_lora_modules/test_lora_attention_pytorch_flow_vs_trt.py
• tests/unittest/_torch/modules/tests_lora_modules/test_lora_plugin_vs_lora_op.py
• tests/unittest/_torch/thop/serial/test_fp8_quantization.py
• tests/unittest/_torch/thop/serial/test_nvfp4_marlin_gemm.py
• tests/unittest/llmapi/test_build_cache.py
• tests/unittest/llmapi/test_llm_models.py
• tests/unittest/llmapi/test_llm_multi_gpu.py
• tests/unittest/others/test_builder.py
• tests/unittest/others/test_debugging_api.py
• tests/unittest/others/test_graph_rewriter.py
• tests/unittest/others/test_kv_cache_manager.py
• tests/unittest/others/test_layer.py
• tests/unittest/others/test_model_dtype.py
• tests/unittest/others/test_module.py
• tests/unittest/others/test_plugins.py
• tests/unittest/others/test_precision_control.py
• tests/unittest/others/test_session.py
• tests/unittest/quantization/test_mode.py
• tests/unittest/test_model_runner_cpp.py
• tests/unittest/tools/plugin_gen/test_core.py
• tests/unittest/tools/plugin_gen/test_plugin_gen.py
• tests/unittest/tools/plugin_gen/test_shape_infer.py
• tests/unittest/trt/__init__.py
• tests/unittest/trt/attention/hf_dynamic_cache_compat.py
• tests/unittest/trt/attention/test_bert_attention.py
• tests/unittest/trt/attention/test_gpt_attention.py
• tests/unittest/trt/attention/test_gpt_attention_IFB.py
• tests/unittest/trt/attention/test_gpt_attention_no_cache.py
• tests/unittest/trt/attention/test_sage_attention.py
• tests/unittest/trt/functional/__init__.py
• tests/unittest/trt/functional/test_alibi.py
• tests/unittest/trt/functional/test_allreduce_norm.py
• tests/unittest/trt/functional/test_allreduce_prepost_residual_norm.py
• tests/unittest/trt/functional/test_arange.py
• tests/unittest/trt/functional/test_argmax.py
• tests/unittest/trt/functional/test_assertion.py
• tests/unittest/trt/functional/test_avg_pool2d.py
• tests/unittest/trt/functional/test_cast.py
• tests/unittest/trt/functional/test_conv2d.py
• tests/unittest/trt/functional/test_conv3d.py
• tests/unittest/trt/functional/test_cos.py
• tests/unittest/trt/functional/test_cumsum.py
• tests/unittest/trt/functional/test_dora.py
• tests/unittest/trt/functional/test_einsum.py
• tests/unittest/trt/functional/test_embedding_single_gpu.py
• tests/unittest/trt/functional/test_exp.py
• tests/unittest/trt/functional/test_expand.py
• tests/unittest/trt/functional/test_flatten.py
• tests/unittest/trt/functional/test_flip.py
• tests/unittest/trt/functional/test_fp4_gemm.py
• tests/unittest/trt/functional/test_fp4_gemm_ootb.py
• tests/unittest/trt/functional/test_gather.py
• tests/unittest/trt/functional/test_gather_nd.py
• tests/unittest/trt/functional/test_geglu.py
• tests/unittest/trt/functional/test_gelu.py
• tests/unittest/trt/functional/test_gemm_swiglu.py
• tests/unittest/trt/functional/test_group_norm.py
• tests/unittest/trt/functional/test_identity.py
• tests/unittest/trt/functional/test_index_select.py
• tests/unittest/trt/functional/test_interpolate.py
• tests/unittest/trt/functional/test_logsoftmax.py
• tests/unittest/trt/functional/test_lora.py
• tests/unittest/trt/functional/test_low_latency_gemm.py
• tests/unittest/trt/functional/test_mamba_conv1d.py
• tests/unittest/trt/functional/test_masked_scatter.py
• tests/unittest/trt/functional/test_masked_select.py
• tests/unittest/trt/functional/test_matmul.py
• tests/unittest/trt/functional/test_meshgrid2d.py
• tests/unittest/trt/functional/test_moe.py
• tests/unittest/trt/functional/test_nccl.py
• tests/unittest/trt/functional/test_nonzero.py
• tests/unittest/trt/functional/test_outer.py
• tests/unittest/trt/functional/test_pad.py
• tests/unittest/trt/functional/test_permute.py
• tests/unittest/trt/functional/test_pp_reduce_scatter.py
• tests/unittest/trt/functional/test_quant.py
• tests/unittest/trt/functional/test_rearrange.py
• tests/unittest/trt/functional/test_repeat.py
• tests/unittest/trt/functional/test_repeat_interleave.py
• tests/unittest/trt/functional/test_rg_lru.py
• tests/unittest/trt/functional/test_sample.py
• tests/unittest/trt/functional/test_scatter.py
• tests/unittest/trt/functional/test_scatter_nd.py
• tests/unittest/trt/functional/test_select.py
• tests/unittest/trt/functional/test_selective_scan.py
• tests/unittest/trt/functional/test_sigmoid.py
• tests/unittest/trt/functional/test_silu.py
• tests/unittest/trt/functional/test_sin.py
• tests/unittest/trt/functional/test_slice.py
• tests/unittest/trt/functional/test_softplus.py
• tests/unittest/trt/functional/test_split.py
• tests/unittest/trt/functional/test_squeeze.py
• tests/unittest/trt/functional/test_swiglu.py
• tests/unittest/trt/functional/test_topk.py
• tests/unittest/trt/functional/test_transpose.py
• tests/unittest/trt/functional/test_unbind.py
• tests/unittest/trt/functional/test_unsqueeze.py
• tests/unittest/trt/functional/test_view.py
• tests/unittest/trt/functional/test_where.py
• tests/unittest/trt/model/__init__.py
• tests/unittest/trt/model/eagle/test_decode_draft_tokens_plugin.py
• tests/unittest/trt/model/eagle/test_prepare_drafter_inputs_plugin.py
• tests/unittest/trt/model/eagle/test_sample_accept_draft_tokens_plugin.py
• tests/unittest/trt/model/redrafter/test_beams2tree.py
• tests/unittest/trt/model/redrafter/test_draft_token.py
• tests/unittest/trt/model/redrafter/test_draft_token_indices.py
• tests/unittest/trt/model/redrafter/test_gather_beams.py
• tests/unittest/trt/model/redrafter/test_mask.py
• tests/unittest/trt/model/redrafter/test_packed_position_ids.py
• tests/unittest/trt/model/redrafter/test_prefix_match_indices.py
• tests/unittest/trt/model/redrafter/test_prepare_input.py
• tests/unittest/trt/model/redrafter/test_process_logits.py
• tests/unittest/trt/model/redrafter/test_top1.py
• tests/unittest/trt/model/redrafter/test_unpack_gen_data.py
• tests/unittest/trt/model/redrafter/test_validate.py
• tests/unittest/trt/model/test_gpt.py
• tests/unittest/trt/model/test_gpt_e2e.py
• tests/unittest/trt/model/test_llama.py
• tests/unittest/trt/model/test_mamba.py
• tests/unittest/trt/model/test_mistral.py
• tests/unittest/trt/model/test_nemotron_nas.py
• tests/unittest/trt/model/test_phi.py
• tests/unittest/trt/model/test_unet.py
• tests/unittest/trt/model_api/profile_utils.py
• tests/unittest/trt/model_api/test_model_api_multi_gpu.py
• tests/unittest/trt/python_plugin/plugin_wrapper_utils.py
• tests/unittest/trt/python_plugin/test_plugin_wrapper.py
• tests/unittest/trt/quantization/__init__.py
• tests/unittest/trt/quantization/_utils.py
• tests/unittest/trt/quantization/test_fp8_rowwise_gemm.py
• tests/unittest/trt/quantization/test_functional.py
• tests/unittest/trt/quantization/test_moe_weight_only_quant_matmul.py
• tests/unittest/trt/quantization/test_qserve_gemm.py
• tests/unittest/trt/quantization/test_quant.py
• tests/unittest/trt/quantization/test_quant_layer.py
• tests/unittest/trt/quantization/test_smooth_quant_gemm.py
• tests/unittest/trt/quantization/test_smooth_quant_layer_norm.py
• tests/unittest/trt/quantization/test_smooth_quant_rms_norm.py
• tests/unittest/trt/quantization/test_weight_only_groupwise_quant_matmul.py
• tests/unittest/trt/quantization/test_weight_only_quant_matmul.py
• tests/unittest/utils/test_medusa_utils.py

💤 Files with no reviewable changes (85)

• tests/unittest/trt/functional/test_gelu.py
• tests/unittest/others/test_plugins.py
• tests/unittest/trt/functional/test_embedding_single_gpu.py
• tests/unittest/others/test_kv_cache_manager.py
• tests/unittest/trt/functional/test_outer.py
• tests/unittest/trt/functional/test_logsoftmax.py
• tests/unittest/trt/functional/test_cast.py
• tests/unittest/trt/functional/test_dora.py
• tests/unittest/trt/functional/test_matmul.py
• tests/unittest/trt/functional/test_pad.py
• tests/unittest/trt/functional/test_cos.py
• tests/unittest/others/test_builder.py
• tests/unittest/trt/functional/test_geglu.py
• tests/integration/test_lists/test-db/l0_gb203.yml
• tests/unittest/trt/functional/test_flip.py
• tests/unittest/trt/functional/test_index_select.py
• tests/integration/defs/triton_server/test_triton_memleak.py
• tests/unittest/trt/functional/test_expand.py
• tests/unittest/others/test_model_dtype.py
• tests/unittest/_torch/modules/tests_lora_modules/test_lora_plugin_vs_lora_op.py
• tests/unittest/others/test_debugging_api.py
• tests/unittest/trt/attention/hf_dynamic_cache_compat.py
• tests/integration/defs/llmapi/test_llm_examples.py
• tests/unittest/trt/functional/test_gather_nd.py
• tests/unittest/others/test_module.py
• tests/unittest/tools/plugin_gen/test_shape_infer.py
• tests/unittest/trt/functional/test_gather.py
• tests/unittest/trt/functional/test_masked_scatter.py
• tests/unittest/trt/functional/test_nonzero.py
• tests/unittest/trt/functional/test_interpolate.py
• tests/unittest/tools/plugin_gen/test_core.py
• tests/unittest/trt/functional/test_group_norm.py
• tests/unittest/trt/functional/test_identity.py
• tests/unittest/trt/functional/test_gemm_swiglu.py
• tests/integration/test_lists/test-db/l0_b200.yml
• tests/unittest/others/test_session.py
• tests/unittest/trt/functional/test_argmax.py
• tests/unittest/others/test_precision_control.py
• tests/unittest/trt/functional/test_cumsum.py
• tests/unittest/trt/functional/test_flatten.py
• tests/unittest/trt/attention/test_bert_attention.py
• tests/unittest/trt/functional/test_lora.py
• tests/unittest/trt/functional/test_avg_pool2d.py
• tests/unittest/trt/functional/test_meshgrid2d.py
• tests/unittest/trt/functional/test_exp.py
• tests/unittest/trt/functional/test_low_latency_gemm.py
• tests/unittest/trt/functional/test_moe.py
• tests/unittest/trt/functional/test_conv2d.py
• tests/unittest/trt/functional/test_arange.py
• tests/unittest/_torch/modules/tests_lora_modules/test_lora_attention_pytorch_flow_vs_trt.py
• tests/unittest/trt/functional/test_assertion.py
• tests/unittest/tools/plugin_gen/test_plugin_gen.py
• tests/unittest/trt/functional/test_fp4_gemm_ootb.py
• tests/unittest/trt/functional/test_allreduce_prepost_residual_norm.py
• tests/unittest/test_model_runner_cpp.py
• tests/unittest/trt/attention/test_gpt_attention_no_cache.py
• tests/unittest/trt/functional/test_conv3d.py
• tests/unittest/trt/functional/test_mamba_conv1d.py
• tests/integration/defs/triton_server/test_triton_multi_node.py
• tests/integration/test_lists/test-db/l0_gh200.yml
• tests/unittest/llmapi/test_build_cache.py
• tests/integration/test_lists/test-db/l0_sanity_check.yml
• tests/unittest/trt/functional/test_fp4_gemm.py
• tests/unittest/trt/functional/test_permute.py
• tests/unittest/trt/functional/test_alibi.py
• tests/integration/test_lists/test-db/l0_l40s.yml
• tests/unittest/llmapi/test_llm_models.py
• tests/integration/defs/triton_server/test_triton.py
• tests/integration/test_lists/test-db/l0_a30.yml
• tests/unittest/trt/functional/test_nccl.py
• tests/unittest/llmapi/test_llm_multi_gpu.py
• tests/unittest/trt/functional/test_einsum.py
• tests/unittest/trt/functional/test_allreduce_norm.py
• tests/unittest/trt/attention/test_gpt_attention_IFB.py
• tests/unittest/others/test_layer.py
• tests/integration/test_lists/test-db/l0_h100.yml
• tests/integration/defs/triton_server/test_triton_rcca.py
• tests/unittest/trt/attention/test_sage_attention.py
• tests/integration/test_lists/test-db/l0_dgx_h100.yml
• tests/integration/defs/triton_server/build_engines.py
• tests/unittest/trt/attention/test_gpt_attention.py
• tests/integration/test_lists/test-db/l0_a100.yml
• tests/integration/test_lists/test-db/l0_dgx_h200.yml
• tests/unittest/trt/functional/test_masked_select.py
• tests/unittest/others/test_graph_rewriter.py

[tensorrt-cicd]

PR_Github #57336 [ run ] triggered by Bot. Commit: 5954f17 Link to invocation

[Wanli-Jiang]

/bot run --add-multi-gpu-test --disable-fail-fast

[tensorrt-cicd]

PR_Github #57360 [ run ] triggered by Bot. Commit: fd7467e Link to invocation

[tensorrt-cicd]

PR_Github #57336 [ run ] completed with state ABORTED. Commit: 5954f17

Link to invocation

[Wanli-Jiang]

/bot run --add-multi-gpu-test --disable-fail-fast

[tensorrt-cicd]

PR_Github #57665 [ run ] triggered by Bot. Commit: 970f453 Link to invocation

[tensorrt-cicd]

PR_Github #57593 [ run ] completed with state ABORTED. Commit: 1c6c430

Link to invocation

[tensorrt-cicd]

PR_Github #57665 [ run ] completed with state SUCCESS. Commit: 970f453
/LLM/main/L0_MergeRequest_PR pipeline #46385 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[Wanli-Jiang]

/bot skip --comment "the last CI passed with two time-out runs, and these two time-out runs passed previously and not related with my changes"

[tensorrt-cicd]

PR_Github #57769 [ skip ] triggered by Bot. Commit: 76a74b7 Link to invocation

[tensorrt-cicd]

PR_Github #57769 [ skip ] completed with state SUCCESS. Commit: 76a74b7
Skipping testing for commit 76a74b7

Link to invocation