[None][fix] Revert "[None][feat] Return topk logprobs in torch backend (#7756)" #7969
Conversation
…)" This reverts commit 2f8dc6f. Signed-off-by: Iman Tabrizian <10105175+tabrizian@users.noreply.github.com>
|
/bot skip --comment "reverting PR" |
Tabrizian
changed the title
|
PR_Github #19838 [ skip ] triggered by Bot |
📝 WalkthroughWalkthroughRemoved per-request num_logprobs plumbing from request construction and workers; reworked sampler to optionally consume/propagate host-provided log_probs tensors, updating method signatures and device/host synchronization. Added validation to reject logprobs > 1 for PyTorch backend. Adjusted scaffolding to stop passing logprobs. Removed a PyTorch logprobs test. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant Client
participant LLM API
participant Scheduler
participant Sampler (Torch)
participant GPU
participant HostState
Client->>LLM API: submit request (may include logprobs<=1)
LLM API-->>Client: accepted
LLM API->>Scheduler: enqueue request
alt logprobs requested
Scheduler->>Sampler: step(logits, log_probs_host tensor)
Sampler->>GPU: allocate logprob buffer (per-step)
Sampler->>HostState: read log_probs_host (CPU)
Sampler->>GPU: scatter per-request logprobs to GPU buffer
Sampler->>HostState: mirror/scatter to host tracking
else no logprobs
Scheduler->>Sampler: step(logits, log_probs_host=None)
Sampler-->>GPU: skip logprob computation
end
Sampler-->>Scheduler: new_tokens (+optional logprobs host mirror)
Scheduler-->>LLM API: outputs
LLM API-->>Client: tokens (+optional logprobs)
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Pre-merge checks and finishing touches❌ Failed checks (2 warnings)
✅ Passed checks (1 passed)
✨ Finishing touches
🧪 Generate unit tests
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (3)
tensorrt_llm/_torch/pyexecutor/sampler.py (1)
1-1: Missing NVIDIA Apache-2.0 header (compliance).All source files must start with the NVIDIA Apache-2.0 copyright header for the current year (2025).
Apply this diff:
+# Copyright (c) 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License.tensorrt_llm/llmapi/llm.py (1)
1-1: Missing NVIDIA Apache-2.0 header (compliance).Add the standard header at the top of the file.
Apply this diff:
+# Copyright (c) 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License.tensorrt_llm/scaffolding/worker.py (1)
1-1: Missing NVIDIA Apache-2.0 header (compliance).Add the standard header at the top of the file.
Apply this diff:
+# Copyright (c) 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License.
🧹 Nitpick comments (2)
tensorrt_llm/_torch/pyexecutor/sampler.py (1)
1163-1177: Avoid computing softmax for requests that don't need it (perf).When any request enables logprobs, this computes softmax for all grouped requests. Limit computation to the union of:
- requests needing draft probs (speculative decoding), and
- requests with py_return_log_probs.
Apply this diff:
- if log_probs_host is not None: - softmax_req_indices = group_req_indices - softmax_grp_indices = torch.arange(len(group_req_indices), - dtype=torch.int32) - speculation_softmax_indices = torch.tensor( - speculation_group_indices, dtype=torch.int32) + if log_probs_host is not None: + # Requests in this group that require logprobs + lp_grp_indices = torch.tensor( + [i for i, idx in enumerate(group_req_indices.tolist()) + if requests[idx].py_return_log_probs], + dtype=torch.int32) + # Union: speculative + logprobs + if lp_grp_indices.numel() > 0: + softmax_grp_indices = torch.unique( + torch.cat((lp_grp_indices, + torch.tensor(speculation_group_indices, + dtype=torch.int32)))) + else: + softmax_grp_indices = torch.tensor( + speculation_group_indices, dtype=torch.int32) + softmax_req_indices = group_req_indices[softmax_grp_indices] + # Map speculation indices into compact softmax indices space + speculation_softmax_indices = torch.arange( + (softmax_grp_indices == torch.tensor( + speculation_group_indices, dtype=torch.int32).unsqueeze(1) + ).any(dim=0).sum().item(), + dtype=torch.int32)Note: If you prefer avoiding tolist(), compute lp_grp_indices via boolean masks on tensors; the above keeps the change localized.
tensorrt_llm/llmapi/llm.py (1)
601-604: Guard to reject top-k logprobs on PyTorch backend is correct.This aligns with the revert: PyTorch supports only logprobs=1. Consider shortening the message to satisfy TRY003 lint.
Apply this diff:
- raise ValueError( - f"PyTorch backend currently only supports `logprobs=1`. Received `logprobs={sampling_params.logprobs}` (Top{sampling_params.logprobs} logprobs). Please set `logprobs=1` in `sampling_params` instead." - ) + raise ValueError( + f"PyTorch backend supports only logprobs=1 (got {sampling_params.logprobs})." + )
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (6)
tensorrt_llm/_torch/pyexecutor/llm_request.py(0 hunks)tensorrt_llm/_torch/pyexecutor/sampler.py(9 hunks)tensorrt_llm/executor/base_worker.py(0 hunks)tensorrt_llm/llmapi/llm.py(1 hunks)tensorrt_llm/scaffolding/worker.py(1 hunks)tests/unittest/llmapi/test_llm_pytorch.py(0 hunks)
💤 Files with no reviewable changes (3)
- tensorrt_llm/executor/base_worker.py
- tests/unittest/llmapi/test_llm_pytorch.py
- tensorrt_llm/_torch/pyexecutor/llm_request.py
🧰 Additional context used
📓 Path-based instructions (3)
**/*.{h,hpp,hh,hxx,cpp,cxx,cc,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Use only spaces, no tabs; indent with 4 spaces.
Files:
tensorrt_llm/scaffolding/worker.pytensorrt_llm/llmapi/llm.pytensorrt_llm/_torch/pyexecutor/sampler.py
**/*.py
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+.
Indent Python code with 4 spaces; do not use tabs.
Maintain module namespace when importing; prefer 'from package.subpackage import foo' then 'foo.SomeClass()' instead of importing the class directly.
Python filenames should be snake_case (e.g., some_file.py).
Python classes use PascalCase names.
Functions and methods use snake_case names.
Local variables use snake_case; prefix 'k' for variables that start with a number (e.g., k_99th_percentile).
Global variables use upper SNAKE_CASE prefixed with 'G' (e.g., G_MY_GLOBAL).
Constants use upper SNAKE_CASE (e.g., MY_CONSTANT).
Avoid shadowing variables from an outer scope.
Initialize all externally visible members of a class in the constructor.
Prefer docstrings for interfaces that may be used outside a file; comments for in-function or file-local interfaces.
Use Google-style docstrings for classes and functions (Sphinx-parsable).
Document attributes and variables inline so they render under the class/function docstring.
Avoid reflection when a simpler, explicit approach suffices (e.g., avoid dict(**locals()) patterns).
In try/except, catch the most specific exceptions possible.
For duck-typing try/except, keep the try body minimal and use else for the main logic.
Files:
tensorrt_llm/scaffolding/worker.pytensorrt_llm/llmapi/llm.pytensorrt_llm/_torch/pyexecutor/sampler.py
**/*.{cpp,cxx,cc,h,hpp,hh,hxx,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Prepend the NVIDIA Apache-2.0 copyright header with current year to the top of all source files (e.g., .cpp, .h, .cu, .py).
Files:
tensorrt_llm/scaffolding/worker.pytensorrt_llm/llmapi/llm.pytensorrt_llm/_torch/pyexecutor/sampler.py
🧠 Learnings (1)
📚 Learning: 2025-08-28T10:25:22.370Z
Learnt from: ixlmar
PR: NVIDIA/TensorRT-LLM#7294
File: tensorrt_llm/_torch/pyexecutor/sampler.py:887-891
Timestamp: 2025-08-28T10:25:22.370Z
Learning: In tensorrt_llm/_torch/pyexecutor/sampler.py, the draft_probs and target_probs tensors have shapes [1, steps] not [steps, vocab_size] as might be expected, making the .squeeze(0) operations appropriate for removing the batch dimension of size 1.
Applied to files:
tensorrt_llm/_torch/pyexecutor/sampler.py
🧬 Code graph analysis (3)
tensorrt_llm/scaffolding/worker.py (1)
tests/unittest/llmapi/test_llm.py (6)
task(481-488)task(528-533)task(1869-1878)task(1988-2001)task(2388-2389)task(2479-2498)
tensorrt_llm/llmapi/llm.py (1)
tensorrt_llm/scaffolding/task.py (1)
logprobs(99-100)
tensorrt_llm/_torch/pyexecutor/sampler.py (3)
tensorrt_llm/_torch/pyexecutor/llm_request.py (1)
log_probs(226-227)tensorrt_llm/executor/result.py (1)
Logprob(37-40)tensorrt_llm/_torch/pyexecutor/scheduler.py (1)
all_requests(38-39)
🪛 Ruff (0.13.1)
tensorrt_llm/llmapi/llm.py
602-604: Avoid specifying long messages outside the exception class
(TRY003)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: Pre-commit Check
🔇 Additional comments (3)
tensorrt_llm/_torch/pyexecutor/sampler.py (2)
855-866: Logprobs emission path looks correct for single-beam, per-iteration buffers.The indexing matches the per-iteration layout populated in _unbatch_sampling_results, and result payloads align with Logprob(dataclass). Rank=1 is acceptable for “selected token only”.
Please confirm downstream consumers don’t rely on top-k ranks (now always 1) for PyTorch backend.
1316-1342: Advanced indexing dtypes must be long (int64).Indexing with int32 can error on some PyTorch builds. Convert row/column indices to int64.
Apply this diff:
- batch_dest_probs_indices_cuda = batch_dest_probs_cuda_indexer[:].to( - batch_softmax_cuda.device, non_blocking=True) + batch_dest_probs_indices_cuda = batch_dest_probs_cuda_indexer[:].to( + batch_softmax_cuda.device, non_blocking=True) # NB: torch.arange is needed to enable "advanced indexing", # cf. https://numpy.org/devdocs/user/basics.indexing.html#integer-array-indexing - batch_token_probs = batch_softmax_cuda[ - torch.arange(batch_softmax_cuda.size(0), - device=batch_softmax_cuda.device, - dtype=torch.int32), batch_next_tokens_cuda_int] + idx_rows = torch.arange(batch_softmax_cuda.size(0), + device=batch_softmax_cuda.device, + dtype=torch.int64) + idx_cols = batch_next_tokens_cuda_int.to(torch.long) + batch_token_probs = batch_softmax_cuda[idx_rows, idx_cols]Optionally clamp before log to avoid -inf for extreme underflow:
- torch.log(batch_token_probs)) + torch.log(batch_token_probs.clamp_min(1e-45)))⛔ Skipped due to learnings
Learnt from: ixlmar PR: NVIDIA/TensorRT-LLM#7294 File: tensorrt_llm/_torch/pyexecutor/sampler.py:368-392 Timestamp: 2025-08-27T15:03:57.149Z Learning: In TensorRT-LLM's sampler.py, int32 usage for softmax_indices and related tensor indexing is intentional and should not be changed to int64. The torch.IntTensor type hint is correct for the sample() function's softmax_indices parameter.Learnt from: ixlmar PR: NVIDIA/TensorRT-LLM#7294 File: tensorrt_llm/_torch/pyexecutor/sampler.py:1068-1085 Timestamp: 2025-08-28T10:21:46.652Z Learning: torch.index_select works with int32 indices in practice despite documentation stating LongTensor requirement. In TensorRT-LLM codebase, int32 indices are used intentionally and work correctly.tensorrt_llm/scaffolding/worker.py (1)
178-184: LGTM: removed logprobs from SamplingParams construction.Matches the revert to disallow top-k logprobs for the PyTorch path.
|
PR_Github #19838 [ skip ] completed with state |
5 participants
This reverts commit 2f8dc6f.
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)
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.