[TRTLLM-7410][feat] Enable KV cache reuse and chunked prefill for mistral3.1 #7628
Conversation
📝 WalkthroughWalkthroughAdds multimodal support enhancements to Mistral3 input processing, including updating multimodal embeddings before fusion and exposing vocab and multimodal token IDs. Cleans up a debug print in the model engine. Introduces parameterized tests validating image-to-token calculations via mocked processor behavior and sets a shared patch size constant. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
actor U as User
participant P as Mistral3InputProcessor
participant V as Vision/AutoProcessor
participant M as Model
U->>P: provide input_ids, images (optional)
alt multimodal context present
P->>P: find_uncached_mm_embeds(images)
P->>P: fuse_input_embeds(input_ids, mm_embeds)
else
P->>P: prepare text-only embeds
end
P->>M: forward(embeds, mm_token_indices)
note over P,V: For token counts per image, P queries V._get_num_multimodal_tokens(...)
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Suggested reviewers
Tip 👮 Agentic pre-merge checks are now available in preview!Pro plan users can now enable pre-merge checks in their settings to enforce checklists before merging PRs. ✨ 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 Pre-merge checks❌ Failed checks (3 warnings)
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 0
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
tensorrt_llm/_torch/models/modeling_mistral.py (1)
1-1: Add required NVIDIA Apache-2.0 header (2025).Per repo guidelines, prepend the NVIDIA Apache-2.0 header to all Python sources.
Apply:
+# Copyright (c) 2025, NVIDIA CORPORATION. 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.tests/unittest/_torch/modeling/test_modeling_mistral.py (1)
1-1: Add required NVIDIA Apache-2.0 header (2025).Tests are Python sources too; add the standard header.
Apply:
+# Copyright (c) 2025, NVIDIA CORPORATION. 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 (3)
tensorrt_llm/_torch/models/modeling_mistral.py (2)
216-216: Constructor/MRO fine; consider cleaning unused arg.trust_remote_code is accepted but unused; either thread it through or drop it to avoid confusion. Also consider documenting the new public methods in the class docstring.
286-291: Return a stable dtype for token IDs.Downstream often compares against int32 input_ids; returning an explicit int32 CPU tensor avoids dtype surprises.
Apply:
- def get_mm_token_ids(self) -> torch.Tensor: - return torch.tensor([ - self._processor.image_token_id, - self._processor.image_break_token_id, - self._processor.image_end_token_id, - ]) + def get_mm_token_ids(self) -> torch.Tensor: + # Keep on CPU; callers can .to(...) if needed. + return torch.tensor( + [ + self._processor.image_token_id, + self._processor.image_break_token_id, + self._processor.image_end_token_id, + ], + dtype=torch.int32, + )tests/unittest/_torch/modeling/test_modeling_mistral.py (1)
507-541: Strengthen the new unit test by asserting the return value.Also stub the processor method’s return to a concrete value to ensure end-to-end behavior.
Apply:
@pytest.mark.parametrize("height, width", [(37, 91), (128, 256), (512, 512)]) @torch.no_grad() def test_processor_get_num_tokens_per_image( @@ - with mock.patch( + with mock.patch( "tensorrt_llm._torch.models.modeling_mistral.AutoProcessor" ) as mocked_auto_processor: + # Provide a deterministic return for token counting + mocked_auto_processor.from_pretrained.return_value._get_num_multimodal_tokens.return_value = { + "num_image_tokens": [123] + } input_processor = modeling_mistral.Mistral3InputProcessor( model_path=str(tmp_path), model_config=mistral_3_config, tokenizer=mock.MagicMock(), ) - input_processor.get_num_tokens_per_image( + ret = input_processor.get_num_tokens_per_image( image=Image.new("RGB", (width, height), color=(255, 128, 0)) ) mocked_auto_processor.from_pretrained.return_value._get_num_multimodal_tokens.assert_called_once_with( [(height, width)] ) + assert ret == 123
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (3)
tensorrt_llm/_torch/models/modeling_mistral.py(5 hunks)tensorrt_llm/_torch/pyexecutor/model_engine.py(0 hunks)tests/unittest/_torch/modeling/test_modeling_mistral.py(4 hunks)
💤 Files with no reviewable changes (1)
- tensorrt_llm/_torch/pyexecutor/model_engine.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:
tests/unittest/_torch/modeling/test_modeling_mistral.pytensorrt_llm/_torch/models/modeling_mistral.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:
tests/unittest/_torch/modeling/test_modeling_mistral.pytensorrt_llm/_torch/models/modeling_mistral.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:
tests/unittest/_torch/modeling/test_modeling_mistral.pytensorrt_llm/_torch/models/modeling_mistral.py
🧠 Learnings (1)
📓 Common learnings
Learnt from: dbari
PR: NVIDIA/TensorRT-LLM#6714
File: tests/integration/defs/triton_server/build_model.sh:724-726
Timestamp: 2025-08-13T08:21:13.006Z
Learning: Mistral Small 3.1 multimodal (pixtral model type) supports any batch size for the multimodal encoder, not just batch size 1. The max_batch_size parameter can be set to values like 2 without conflicts with the runtime batching logic.
🧬 Code graph analysis (2)
tests/unittest/_torch/modeling/test_modeling_mistral.py (3)
tests/unittest/_torch/helpers.py (1)
create_mock_engine(181-194)tensorrt_llm/_torch/models/modeling_mistral.py (1)
Mistral3InputProcessor(216-291)tensorrt_llm/inputs/registry.py (1)
get_num_tokens_per_image(115-133)
tensorrt_llm/_torch/models/modeling_mistral.py (2)
tensorrt_llm/_torch/models/modeling_multimodal_utils.py (2)
find_uncached_mm_embeds(33-105)fuse_input_embeds(148-202)tensorrt_llm/inputs/registry.py (4)
BaseMultimodalInputProcessor(45-163)InputProcessor(22-42)get_vocab_size(64-84)get_mm_token_ids(86-98)
⏰ 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 (7)
tensorrt_llm/_torch/models/modeling_mistral.py (4)
17-18: Good: fuse helpers imported where used.Brings in find_uncached_mm_embeds before fusion; aligns with mixed batches and KV reuse handling.
32-33: Good: adopt BaseMultimodalInputProcessor.This connects Mistral3 to shared multimodal utilities (token counting, processor access).
280-285: LGTM: vocab size override matches Mistral3 config layout.
418-419: No action needed:multimodal_params_listis ordered context-first. model_engine appends all context-phaseMultimodalParams(fromscheduled_requests.context_requests) before any generation-phase entries, so slicing by[:num_context_requests]correctly isolates context embeds.Likely an incorrect or invalid review comment.
tests/unittest/_torch/modeling/test_modeling_mistral.py (3)
10-13: LGTM: imports for PIL and Mistral3 module.
26-27: LGTM: shared _PATCH_SIZE for tests.
71-71: LGTM: vision_config.patch_size uses constant.
2ez4bz
force-pushed
the
dev-mistral3-num-tokens
branch
2 times, most recently
from
d8bceca to
80590ef
Compare
|
/bot run |
|
PR_Github #18103 [ run ] triggered by Bot |
|
PR_Github #18103 [ run ] completed with state |
2ez4bz
force-pushed
the
dev-mistral3-num-tokens
branch
from
80590ef to
4ada633
Compare
2ez4bz
force-pushed
the
dev-mistral3-num-tokens
branch
from
4ada633 to
5655c89
Compare
|
Also please update the doc: https://github.com/NVIDIA/TensorRT-LLM/blob/main/docs/source/models/supported-models.md and PR title |
2ez4bz
force-pushed
the
dev-mistral3-num-tokens
branch
from
5655c89 to
fb92867
Compare
2ez4bz
force-pushed
the
dev-mistral3-num-tokens
branch
from
fb92867 to
4a60897
Compare
|
/bot run --extra-stage "H100_PCIe-PyTorch-Post-Merge-1" |
|
PR_Github #18832 [ run ] triggered by Bot |
|
PR_Github #18832 [ run ] completed with state |
Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>
2ez4bz
force-pushed
the
dev-mistral3-num-tokens
branch
from
4a60897 to
ee5db78
Compare
2ez4bz
changed the title
|
/bot run --extra-stage "H100_PCIe-PyTorch-Post-Merge-1" |
|
PR_Github #18935 [ run ] triggered by Bot |
|
PR_Github #18935 [ run ] completed with state |
…tral3.1 (NVIDIA#7628) Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>
…tral3.1 (NVIDIA#7628) Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>
7 participants
Summary by CodeRabbit
New Features
Chores
Tests
Description
[TRTLLM-7410][feat] Implement image token counting for Mistral3 VLM
Test Coverage
Unit test added for new code.
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.