[https://nvbugs/5478151][fix] Add missing spec for Llama-3.3 70B #7267
Conversation
📝 WalkthroughWalkthroughAdds a second FP8 quantization variant for meta-llama/Llama-3.3-70B-Instruct in Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes Possibly related PRs
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. ✨ 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
|
brb-nv
force-pushed
the
user/brb/missing-spec-for-llama-3.3
branch
from
4625c6c to
42f809f
Compare
brb-nv
changed the title
brb-nv
force-pushed
the
user/brb/missing-spec-for-llama-3.3
branch
from
42f809f to
6e54fb3
Compare
Signed-off-by: Balaram Buddharaju <[email protected]>
brb-nv
force-pushed
the
user/brb/missing-spec-for-llama-3.3
branch
from
6e54fb3 to
f306dea
Compare
![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
🧹 Nitpick comments (1)
tests/integration/defs/accuracy/references/gpqa_diamond.yaml (1)
8-10: Optional: add a clarifying inline comment to reduce ambiguityA brief YAML comment helps future readers quickly differentiate the two FP8 variants at a glance.
- - quant_algo: FP8 - kv_cache_quant_algo: FP8 - accuracy: 48.03 + - quant_algo: FP8 # weights FP8, KV cache FP8 + kv_cache_quant_algo: FP8 + accuracy: 48.03
📜 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)
tests/integration/defs/accuracy/references/gpqa_diamond.yaml(1 hunks)
⏰ 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 (2)
tests/integration/defs/accuracy/references/gpqa_diamond.yaml (2)
8-10: LGTM: Missing FP8+KV FP8 spec added and consistent with schemaThe added entry mirrors existing patterns elsewhere in this file: key ordering (quant_algo → kv_cache_quant_algo → accuracy), casing, and indentation all look correct. This should unblock consumers that expect the FP8+KV FP8 variant for Llama-3.3-70B-Instruct.
8-10: Two FP8 variants for Llama‑3.3‑70B — ensure reference selection keys on (quant_algo, kv_cache_quant_algo)gpqa_diamond.yaml contains two FP8 entries for meta-llama/Llama-3.3-70B-Instruct (one with kv_cache_quant_algo: FP8 and one without; both accuracy=48.03). Please verify the harness selects the intended variant deterministically.
Check these places:
- tests/integration/defs/accuracy/references/gpqa_diamond.yaml — the duplicate FP8 entries for that model.
- tests/integration/defs/accuracy/accuracy_core.py — reference loading (yaml.safe_load(...).get(model_name, []), ~lines 92–95) and the get_num_samples_and_threshold implementation (where a reference entry should be matched).
- tensorrt_llm/quantization/mode.py — QuantMode.from_quant_algo(quant_algo, kv_cache_quant_algo) shows kv_cache is a distinct dimension; reference matching should mirror this.
- Any callers/tests that construct QuantConfig (e.g., tests/integration/defs/accuracy/test_llm_api_pytorch.py) to ensure they pass the intended kv_cache_quant_algo when they require the KV variant.
Concrete ask / recommended check:
- Ensure the matching logic compares the tuple (entry.get("quant_algo"), entry.get("kv_cache_quant_algo")) to the requested (quant_algo, kv_cache_quant_algo) — treating absent fields as None — instead of matching only on quant_algo.
- If a fallback behavior is desired, make it explicit (prefer exact tuple match, then optionally fall back), and add a small unit test that asserts selection of FP8 vs FP8+KV when kv_cache_quant_algo is specified.
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (1)
tests/integration/defs/accuracy/references/gpqa_diamond.yaml (1)
8-10: Good addition; please add a brief comment and verify selection logic to avoid ambiguity.The new FP8 variant with kv_cache_quant_algo: FP8 looks correct and aligns with the PR intent. However, there are now two FP8 entries (one with and one without kv_cache quant), both reporting accuracy 48.03. If the reference resolver selects entries by first match on quant_algo alone, the more general FP8 entry could shadow the kv-cache FP8 case.
- Add a clarifying YAML comment so future readers know why both exist.
- Verify the resolver keys on both quant_algo and kv_cache_quant_algo (or equivalent) to ensure deterministic selection.
Apply this small doc tweak:
- quant_algo: FP8 accuracy: 48.03 + # FP8 with KV cache quantized to FP8 (added for Llama-3.3-70B; see nvbugs 5478151) - quant_algo: FP8 kv_cache_quant_algo: FP8 accuracy: 48.03If your resolver is first-match, consider ordering the more specific variant before the generic one to be extra-safe.
📜 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)
tests/integration/defs/accuracy/references/gpqa_diamond.yaml(1 hunks)
⏰ 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)
tests/integration/defs/accuracy/references/gpqa_diamond.yaml (1)
8-10: FP8+KV-FP8 golden confirmed and spec path noted
gpqa_diamond.yaml defines two FP8 entries for meta-llama/Llama-3.3-70B-Instruct:
- Plain FP8 → accuracy: 48.03
- FP8 + KV-cache FP8 → accuracy: 48.03
Both entries are present and distinct, confirming the 48.03 golden for KV-FP8 isn’t a stray copy but intentionally matches the FP8 baseline.Integration tests point to the TensorRT spec under
modelopt-hf-model-hub/Llama-3.3-70B-Instruct-fp8
for FP8 builds. By design, that spec is built with KV-cache FP8 support enabled (no separate “-fp8-kvfp8” directory is required).Please verify that the FP8 spec published on your model hub indeed includes KV-cache FP8 support so tests won’t skip or fallback.
|
/bot run |
|
PR_Github #16595 [ run ] triggered by Bot |
|
PR_Github #16595 [ run ] completed with state |
Signed-off-by: Balaram Buddharaju <[email protected]> Co-authored-by: Larry <[email protected]>
…DIA#7267) Signed-off-by: Balaram Buddharaju <[email protected]> Co-authored-by: Larry <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…DIA#7267) Signed-off-by: Balaram Buddharaju <[email protected]> Co-authored-by: Larry <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…DIA#7267) Signed-off-by: Balaram Buddharaju <[email protected]> Co-authored-by: Larry <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…DIA#7267) Signed-off-by: Balaram Buddharaju <[email protected]> Co-authored-by: Larry <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…DIA#7267) Signed-off-by: Balaram Buddharaju <[email protected]> Co-authored-by: Larry <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…DIA#7267) Signed-off-by: Balaram Buddharaju <[email protected]> Co-authored-by: Larry <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…DIA#7267) Signed-off-by: Balaram Buddharaju <[email protected]> Co-authored-by: Larry <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Description
This MR adds a missing spec for Llama 3.3 70B to fix https://nvbugspro.nvidia.com/bug/5478151.
Test Coverage
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.
Summary by CodeRabbit