[doc][ci][Qwen3][nvbugs 5374145] Add Qwen3 235B eagle3 CI #6477
Conversation
📝 WalkthroughWalkthroughThis change updates the Qwen3 model documentation to describe Eagle3 speculative decoding support, adds new accuracy reference entries for Eagle3 in GSM8K and MMLU benchmarks, parameterizes and extends the related integration test to cover Eagle3, updates test lists to include the new Eagle3 variant, and removes a waiver for an unrelated test. Changes
Sequence Diagram(s)sequenceDiagram
participant Tester as Integration Test
participant LLM as LLM Context Manager
participant Qwen3 as Qwen3 Model
Tester->>LLM: Initialize with Eagle3 speculative_config (if eagle3=True)
LLM->>Qwen3: Load model with Eagle3 config
Qwen3-->>LLM: Ready for inference
LLM-->>Tester: Run evaluation (GSM8K/MMLU) with Eagle3 decoding
Estimated code review effort🎯 2 (Simple) | ⏱️ ~8 minutes Possibly related PRs
Suggested labels
Suggested reviewers
Note ⚡️ Unit Test Generation is now available in beta!Learn more here, or try it out under "Finishing Touches" below. 📜 Recent review detailsConfiguration used: .coderabbit.yaml 📒 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
🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
Documentation and Community
|
Signed-off-by: bhsueh <[email protected]>
Signed-off-by: bhsueh <[email protected]>
Signed-off-by: bhsueh <[email protected]>
byshiue
force-pushed
the
qwen3_235b_eagle3_ci
branch
from
2d4187c to
49edb97
Compare
|
/bot run |
|
PR_Github #13513 [ 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: 1
🧹 Nitpick comments (3)
tests/integration/defs/accuracy/references/mmlu.yaml (1)
173-176: Duplicate accuracy record – consider merging to avoid ambiguityThis new item repeats the exact
(quant_algo=NVFP4, kv_cache_quant_algo=FP8, accuracy=86)tuple already declared on Lines 170-172, differing only by addingspec_dec_algo: Eagle.
If the accuracy-validation logic keys solely offmodel + quant_algo + kv_cache_quant_algo, the duplicate may be ignored or, worse, generate the wrong reference being picked at random.- - quant_algo: NVFP4 - kv_cache_quant_algo: FP8 - accuracy: 86 - - spec_dec_algo: Eagle - quant_algo: NVFP4 - kv_cache_quant_algo: FP8 - accuracy: 86 + - spec_dec_algo: Eagle # merge the new flag into one entry + quant_algo: NVFP4 + kv_cache_quant_algo: FP8 + accuracy: 86Please confirm the de-duplication rules of the harness and collapse the records if necessary.
tests/integration/defs/accuracy/references/gsm8k.yaml (1)
89-92: Same duplication issue as in MMLU referenceLines 83-88 already store the
(quant_algo=NVFP4, kv_cache_quant_algo=FP8, accuracy=85.78)tuple; Lines 89-92 re-insert it withspec_dec_algo: Eagle.
Ensure the accuracy loader will not treat this as two independent references for the same setting and merge if required (see previous diff example).examples/models/core/qwen/README.md (1)
909-923: Use a heredoc instead of multi-lineechofor the sample configUsing
echo "…"with embedded new-lines is brittle and often collapses whitespace,
especially when copy-pasted. A small switch tocat <<EOF > ${path_config}is more
robust and readable.-echo " -enable_attention_dp: false -speculative_config: - decoding_type: Eagle - max_draft_len: 3 - speculative_model_dir: <EAGLE3_DRAFT_MODEL_PATH> -kv_cache_config: - enable_block_reuse: false -" >> ${path_config} +cat >"${path_config}" <<'EOF' +enable_attention_dp: false +speculative_config: + decoding_type: Eagle + max_draft_len: 3 + speculative_model_dir: <EAGLE3_DRAFT_MODEL_PATH> +kv_cache_config: + enable_block_reuse: false +EOFMinor, but it prevents accidental formatting issues for users following the doc.
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (6)
examples/models/core/qwen/README.md(2 hunks)tests/integration/defs/accuracy/references/gsm8k.yaml(1 hunks)tests/integration/defs/accuracy/references/mmlu.yaml(1 hunks)tests/integration/defs/accuracy/test_llm_api_pytorch.py(1 hunks)tests/integration/test_lists/test-db/l0_gb200_multi_nodes.yml(1 hunks)tests/integration/test_lists/waives.txt(0 hunks)
💤 Files with no reviewable changes (1)
- tests/integration/test_lists/waives.txt
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.py: The code developed for TensorRT-LLM should conform to Python 3.8+.
Indent Python code with 4 spaces. Do not use tabs.
Always maintain the namespace when importing in Python, even if only one class or function from a module is used.
Python filenames should use snake_case (e.g., some_file.py).
Python classes should use PascalCase (e.g., class SomeClass).
Python functions and methods should use snake_case (e.g., def my_awesome_function():).
Python local variables should use snake_case. Prefix k for variable names that start with a number (e.g., k_99th_percentile).
Python global variables should use upper snake_case and prefix G (e.g., G_MY_GLOBAL).
Python constants should use upper snake_case (e.g., MY_CONSTANT).
Avoid shadowing variables declared in an outer scope in Python.
Initialize all externally visible members of a Python class in the constructor.
For interfaces that may be used outside a file, prefer docstrings over comments in Python.
Comments in Python should be reserved for code within a function, or interfaces that are local to a file.
Use Google style docstrings for Python classes and functions, which can be parsed by Sphinx.
Attributes and variables in Python can be documented inline; attribute docstrings will be rendered under the docstring for the class.
Avoid using reflection in Python when functionality can be easily achieved without reflection.
When using try-except blocks in Python, limit the except to the smallest set of errors possible.
When using try-except blocks to handle multiple possible variable types in Python, keep the body of the try as small as possible, using the else block to implement the logic.
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
**/*.{cpp,h,cu,py}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
All TensorRT-LLM Open Source Software code should contain an NVIDIA copyright header that includes the current year. This includes .cpp, .h, .cu, .py, and any other source files which are compiled or interpreted.
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
🧠 Learnings (2)
tests/integration/test_lists/test-db/l0_gb200_multi_nodes.yml (1)
Learnt from: moraxu
PR: #6303
File: tests/integration/test_lists/qa/examples_test_list.txt:494-494
Timestamp: 2025-07-28T17:06:08.621Z
Learning: In TensorRT-LLM testing, it's common to have both CLI flow tests (test_cli_flow.py) and PyTorch API tests (test_llm_api_pytorch.py) for the same model. These serve different purposes: CLI flow tests validate the traditional command-line workflow, while PyTorch API tests validate the newer LLM API backend. Both are legitimate and should coexist.
tests/integration/defs/accuracy/test_llm_api_pytorch.py (1)
Learnt from: moraxu
PR: #6303
File: tests/integration/test_lists/qa/examples_test_list.txt:494-494
Timestamp: 2025-07-28T17:06:08.621Z
Learning: In TensorRT-LLM testing, it's common to have both CLI flow tests (test_cli_flow.py) and PyTorch API tests (test_llm_api_pytorch.py) for the same model. These serve different purposes: CLI flow tests validate the traditional command-line workflow, while PyTorch API tests validate the newer LLM API backend. Both are legitimate and should coexist.
🧬 Code Graph Analysis (1)
tests/integration/defs/accuracy/test_llm_api_pytorch.py (2)
tensorrt_llm/llmapi/llm_args.py (3)
KvCacheConfig(818-897)EagleDecodingConfig(386-412)speculative_model_dir(1237-1238)tests/integration/defs/conftest.py (1)
llm_models_root(77-83)
🔇 Additional comments (4)
tests/integration/test_lists/test-db/l0_gb200_multi_nodes.yml (1)
21-21: Verified:test_nvfp4[latency_moe_trtllm_eagle3]ExistsConfirmed in
tests/integration/defs/accuracy/test_llm_api_pytorch.pyat line 1899:
- The parameter list includes
"latency_moe_trtllm_eagle3"under thetest_nvfp4decorator.No changes required.
tests/integration/defs/accuracy/test_llm_api_pytorch.py (3)
1891-1900: LGTM! Parameter addition for Eagle3 testing.The parametrization correctly adds the
eagle3parameter with appropriate test configurations. The Eagle3 test case properly disables attention_dp, cuda_graph, and overlap_scheduler, which aligns with the documented limitations of Eagle3 speculative decoding.
1902-1903: LGTM! Method signature correctly updated.The
eagle3parameter has been properly added to the method signature, maintaining consistency with the parametrization.
1926-1927: LGTM! Proper integration with LLM constructor.The
kv_cache_configandspeculative_configare correctly passed to the LLM constructor, enabling Eagle3 speculative decoding when the parameter is set.
Signed-off-by: bhsueh <[email protected]>
|
/bot run |
|
PR_Github #13517 [ run ] triggered by Bot |
|
PR_Github #13513 [ run ] completed with state |
|
PR_Github #13517 [ run ] completed with state |
Signed-off-by: bhsueh <[email protected]> Signed-off-by: Lanyu Liao <[email protected]>
Signed-off-by: bhsueh <[email protected]>
This PR has following works:
Summary by CodeRabbit