[None][fix] DeepSeek-R1 W4A8 weight loading issue; fixes regression from #6200 #7123
Conversation
📝 WalkthroughWalkthroughAdds per-layer expert quantization resolution to Deepseek V3 MoE components: introduces a helper to fetch per-layer expert QuantConfig (with fallback), switches weight-loading decisions to use per-layer quant_mode, and adds an assertion forbidding MIXED_PRECISION during decoder/MoE initialization. Changes
Sequence Diagram(s)sequenceDiagram
participant App as Model init
participant Layer as DecoderLayer (MoE)
participant QC as _get_experts_quant_config
participant Weights as Weight Loader
App->>Layer: Initialize layer (layer_idx)
Layer->>QC: Fetch experts QuantConfig(layer_idx)
QC-->>Layer: QuantConfig (per-layer or fallback)
alt QuantConfig.quant_algo == MIXED_PRECISION
Layer->>Layer: Assert fail (disallow MIXED_PRECISION)
else
Layer->>Weights: Determine weight_loading_mode from per-layer quant_mode
Weights-->>Layer: Load expert weights
end
sequenceDiagram
participant App as MTP init
participant MTP as DeepseekV3MTP
participant QC as _get_experts_quant_config
participant Weights as Weight Loader
App->>MTP: Initialize MoE experts (layer_idx)
MTP->>QC: Get experts QuantConfig(layer_idx)
QC-->>MTP: QuantConfig
MTP->>Weights: Set weight_loading_mode based on per-layer quant_mode
Weights-->>MTP: Load expert weights
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Suggested labels
Suggested reviewers
Tip 🔌 Remote MCP (Model Context Protocol) integration is now available!Pro plan users can now connect to remote MCP servers from the Integrations page. Connect with popular remote MCPs such as Notion and Linear to add more context to your reviews and chats. 📜 Recent review detailsConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro 💡 Knowledge Base configuration:
You can enable these sources in your CodeRabbit configuration. 📒 Files selected for processing (1)
🚧 Files skipped from review as they are similar to previous changes (1)
⏰ 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)
✨ 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. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. CodeRabbit Commands (Invoked using PR/Issue comments)Type Other keywords and placeholders
Status, Documentation and Community
|
rosenrodt
force-pushed
the
ds-w4a8-load-fix
branch
from
f12ab8a to
99b382b
Compare
|
/bot run |
|
PR_Github #16307 [ run ] triggered by Bot |
|
PR_Github #16307 [ run ] completed with state |
rosenrodt
force-pushed
the
ds-w4a8-load-fix
branch
from
99b382b to
64ab562
Compare
|
/bot run |
|
PR_Github #16318 [ run ] triggered by Bot |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🧹 Nitpick comments (2)
tensorrt_llm/_torch/models/modeling_deepseekv3.py (2)
48-48: Drop QuantAlgo import if the global assert is removedIf you adopt the per-layer NVFP4 resolution and remove the MIXED_PRECISION assert (Lines 649-652), this import becomes unused and may fail lint in CI.
-from tensorrt_llm.quantization.mode import QuantAlgo
461-467: Add minimal unit coverage for W4A8 per-layer resolutionThis PR changes the selection of
MoEWeightLoadingModebased on per-layer config, but ships without tests. Add a focused test that constructs aModelConfigwith aquant_config_dictentry formodel.layers.{i}.mlp.expertsset to an int4-per-group mode and asserts thatDeepseekv3MoE(..., layer_idx=i)picksW4A8_CUSTOM; also verify the fallback toVANILLAwhen the entry is absent.I can draft a minimal test that stubs QuantConfig/QuantMode to assert the mode switch without needing GPU ops. Want me to open a follow-up PR with the test?
Also applies to: 532-538
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
💡 Knowledge Base configuration:
- MCP integration is disabled by default for public repositories
- Jira integration is disabled by default for public repositories
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
📒 Files selected for processing (1)
tensorrt_llm/_torch/models/modeling_deepseekv3.py(4 hunks)
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+
Python indentation: 4 spaces, no tabs
Maintain module namespace in imports (from package.subpackage import foo; then use foo.SomeClass())
Python file names use snake_case
Python class names use PascalCase
Python functions/methods and local variables use snake_case; variables starting with a number get k_ prefix (e.g., k_99th_percentile)
Global variables use G_ prefixed UPPER_SNAKE_CASE (e.g., G_MY_GLOBAL)
Constants use UPPER_SNAKE_CASE in Python
Avoid shadowing variables from outer scopes in Python
Initialize all externally visible members of a Python class in init
Prefer docstrings for interfaces used outside a file; comments for local code
Use Google-style docstrings for classes and functions (Sphinx-parsable)
Document attributes/variables inline with short docstrings
Avoid reflection when simple alternatives exist (e.g., prefer explicit parameters over dict(**locals()))
In try/except, catch the narrowest exceptions possible
For duck-typing with try/except, keep try body minimal and put logic in else
Files:
tensorrt_llm/_torch/models/modeling_deepseekv3.py
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Prepend NVIDIA copyright header (current year) to all source files
Files:
tensorrt_llm/_torch/models/modeling_deepseekv3.py
🧬 Code graph analysis (1)
tensorrt_llm/_torch/models/modeling_deepseekv3.py (4)
tensorrt_llm/quantization/mode.py (2)
QuantAlgo(23-46)is_int4_weight_only_per_group(133-134)tensorrt_llm/_torch/modules/fused_moe/interface.py (1)
MoEWeightLoadingMode(13-19)tensorrt_llm/models/modeling_utils.py (3)
layer_quant_mode(166-170)layer_quant_mode(306-312)quant_algo(547-548)tensorrt_llm/llmapi/llm_args.py (2)
quant_config(2175-2178)quant_config(2181-2182)
⏰ 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 (1)
tensorrt_llm/_torch/models/modeling_deepseekv3.py (1)
649-652: No action needed:layer_quant_modeAPI signatures confirmed•
layer_quant_modeexists as a zero-argument cached property (returns a uniformQuantMode) in
tensorrt_llm/models/modeling_utils.py at line 166.
• It also exists as a method accepting a singlelayer_nameargument (for per-layer modes) in
tensorrt_llm/models/modeling_utils.py at line 306.Since this change neither introduces any new calls to
layer_quant_modenor relies onQuantAlgo.MIXED_PRECISION, the existing API surface remains compatible.
|
PR_Github #16318 [ run ] completed with state |
|
/bot run |
|
PR_Github #16340 [ run ] triggered by Bot |
|
PR_Github #16340 [ run ] completed with state |
Signed-off-by: Anthony Chang <[email protected]>
rosenrodt
force-pushed
the
ds-w4a8-load-fix
branch
from
64ab562 to
cdda63e
Compare
|
/bot run |
|
PR_Github #16381 [ run ] triggered by Bot |
|
/bot run |
|
PR_Github #16398 [ run ] triggered by Bot |
|
PR_Github #16381 [ run ] completed with state |
|
PR_Github #16398 [ run ] completed with state |
|
/bot run |
|
PR_Github #16425 [ run ] triggered by Bot |
|
PR_Github #16425 [ run ] completed with state |
|
/bot run |
|
PR_Github #17726 [ run ] triggered by Bot |
|
PR_Github #17726 [ run ] completed with state |
|
/bot run |
|
PR_Github #17778 [ run ] triggered by Bot |
|
PR_Github #17778 [ run ] completed with state |
…rom NVIDIA#6200 (NVIDIA#7123) Signed-off-by: Anthony Chang <[email protected]>
3 participants
Summary by CodeRabbit
New Features
Bug Fixes
Refactor
Description
MoE quant mode should be determined by individual layers in the quant_config for DS-R1 MIXED_PRECISION quant_algo.
Test Coverage
No test coverage at this moment.
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.