[None] [doc] Update DeepSeek example doc #7358
Conversation
Signed-off-by: jiahanc <[email protected]>
📝 WalkthroughWalkthroughAdds B200 FP4 and FP8 serving configurations and guidance to the DeepSeek README, switches example commands and OpenAI/curl snippets from DeepSeek‑V3 to DeepSeek‑R1, updates TOC and wording, and includes OOM/memory adjustment notes. Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 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. ✨ 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
|
![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 (3)
examples/models/core/deepseek_v3/README.md (3)
236-236: Tighten wording and avoid “want”.Minor style nit; suggest clearer phrasing.
-Below are example B200 serving configurations for both min-latency and max-throughput in FP4 and FP8. If you want to explore configurations, see the [blog](https://github.com/NVIDIA/TensorRT-LLM/blob/main/docs/source/blogs/Best_perf_practice_on_DeepSeek-R1_in_TensorRT-LLM.md). **Treat these as starting points—tune for your model and workload to achieve the best performance.** +Below are example B200 serving configurations for both min-latency and max-throughput in FP4 and FP8. For additional configurations, see the [blog](https://github.com/NVIDIA/TensorRT-LLM/blob/main/docs/source/blogs/Best_perf_practice_on_DeepSeek-R1_in_TensorRT-LLM.md). **Treat these as starting points—tune for your model and workload.**
240-251: Clarify FP4 vs FP8 KV cache to avoid confusion.Section title says “FP4”, but YAML sets
kv_cache_config.dtype: fp8. That’s valid (FP4 weights + FP8 KV cache), but readers may misinterpret it.cat >./extra-llm-api-config.yml <<EOF +# FP4 weights with FP8 KV cache cuda_graph_config: enable_padding: true max_batch_size: 1024 enable_attention_dp: false kv_cache_config: dtype: fp8 stream_interval: 10 EOF
332-332: Model name in curl matches the server example—LGTM.Consider adding an alternate curl snippet for FP4 users (see naming note below).
📜 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)
examples/models/core/deepseek_v3/README.md(2 hunks)
🧰 Additional context used
🪛 LanguageTool
examples/models/core/deepseek_v3/README.md
[style] ~236-~236: Consider using an alternative to strengthen your wording.
Context: ...d max-throughput in FP4 and FP8. If you want to explore configurations, see the [blo...
(WANT_KEEN)
⏰ 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)
examples/models/core/deepseek_v3/README.md (2)
31-35: TOC additions look good.Anchors match the added headings below.
259-276: Batch size ladder looks sane; config LGTM.Descending ladder and enabling attention DP for throughput are consistent with the section intent.
jiahanc
changed the title
jiahanc
changed the title
Signed-off-by: jiahanc <[email protected]>
jiahanc
changed the title
There was a problem hiding this comment.
Actionable comments posted: 0
♻️ Duplicate comments (5)
examples/models/core/deepseek_v3/README.md (5)
31-34: Resolve FP8 support inconsistency for B200/GB200.TOC adds “B200 FP8 …” sections, but the Hardware Requirements table (Line 67) says FP8 on B200/GB200 is “Not supported yet, WIP.” Align by updating the table to “experimental” or gating FP8 sections with a clear disclaimer.
- | B200/GB200| Not supported yet, WIP | 4 (8 GPUs is recommended for best perf) | Not supported yet, WIP | + | B200/GB200| 8 (experimental; tune config) | 4 (8 GPUs is recommended for best perf) | Not supported yet, WIP |
339-340: Link model IDs to HF pages.Make model IDs clickable for quick access.
-For DeepSeek-R1 FP4, use the model name `nvidia/DeepSeek-R1-FP4-v2`. -For DeepSeek-V3, use the model name `deepseek-ai/DeepSeek-V3`. +For DeepSeek-R1 FP4, use the model name [`nvidia/DeepSeek-R1-FP4-v2`](https://huggingface.co/nvidia/DeepSeek-R1-FP4-v2). +For DeepSeek-V3, use the model name [`deepseek-ai/DeepSeek-V3`](https://huggingface.co/deepseek-ai/DeepSeek-V3).
278-309: Unify memory budget knobs and call out FP8’s experimental status.These FP8 blocks set
free_gpu_memory_fraction: 0.8while the launch command later uses--kv_cache_free_gpu_memory_fraction 0.9. Make them consistent and document precedence (CLI vs YAML).- free_gpu_memory_fraction: 0.8 + free_gpu_memory_fraction: 0.8 # Keep YAML as the source of truth; remove the CLI flag below or match 0.8
311-324: Make KV cache memory fraction consistent and define precedence.Either drop the CLI flag (prefer YAML), or match the YAML’s 0.8 and state which source wins.
--pp_size 1 \ - --kv_cache_free_gpu_memory_fraction 0.9 \ + # Prefer YAML for KV cache memory budget to avoid conflicting sources: + # --kv_cache_free_gpu_memory_fraction 0.8 \ --extra_llm_api_options ./extra-llm-api-config.yml
325-325: Fix flag name typo and tighten OOM guidance.Use the exact flag
--kv_cache_free_gpu_memory_fractionand align “max‑throughput” advice with the 8192 shown above.-It's possible seeing OOM issues on some configs. Considering reducing `kv_cache_free_gpu_mem_fraction` to a smaller value as a workaround. We're working on the investigation and addressing the problem. If you are using max-throughput config, reduce `max_num_tokens` to `3072` to avoid OOM issues. +You may encounter OOM on some configs. Reduce `--kv_cache_free_gpu_memory_fraction` as a workaround. We are investigating the issue. For max‑throughput configs, reducing `--max_num_tokens` (e.g., from 8192 to 3072) can also help.
🧹 Nitpick comments (4)
examples/models/core/deepseek_v3/README.md (4)
236-236: Tighten wording and add FP8 experimental disclaimer.Strengthen tone and preface FP8 on B200 as experimental to avoid contradicting the table.
-Below are example B200 serving configurations for both min-latency and max-throughput in FP4 and FP8. If you want to explore configurations, see the [blog](...). **Treat these as starting points—tune for your model and workload to achieve the best performance.** +Below are example B200 serving configurations for both min‑latency and max‑throughput in FP4 and FP8. For additional configurations, see the [blog](...). **On B200, FP8 is experimental—treat these as starting points and tune for your model/workload.**
240-251: Clarify “FP4 config” vs FP8 KV cache dtype in YAML.Readers may assume everything is FP4. Add a comment noting FP4 refers to weights, while KV cache uses FP8 for perf.
kv_cache_config: dtype: fp8 +# FP4 refers to model weights; using FP8 KV cache improves throughput/latency.
253-276: Batch sizes list: consider adding 384 for better CUDA graph coverage.You use 384 elsewhere in the doc; adding it can improve capture coverage for mid‑range throughputs.
batch_sizes: + - 384 - 1024 - 896 - 512
332-337: Align model name with recommended FP4 checkpoint.Use the recommended R1 FP4 ID for consistency with the Serving section.
- "model": "deepseek-ai/DeepSeek-R1", + "model": "nvidia/DeepSeek-R1-FP4-v2",
📜 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)
examples/models/core/deepseek_v3/README.md(2 hunks)
🧰 Additional context used
🪛 LanguageTool
examples/models/core/deepseek_v3/README.md
[style] ~236-~236: Consider using an alternative to strengthen your wording.
Context: ...d max-throughput in FP4 and FP8. If you want to explore configurations, see the [blo...
(WANT_KEEN)
⏰ 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
|
/bot skip |
GitHub Bot Help
Provide a user friendly way for developers to interact with a Jenkins server. Run See details below for each supported subcommand.
Launch build/test pipelines. All previously running jobs will be killed.
kill
Kill all running builds associated with pull request. skip
Skip testing for latest commit on pull request. reuse-pipeline
Reuse 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. |
|
/bot skip --comment "dos only" |
|
PR_Github #17249 [ skip ] triggered by Bot |
|
PR_Github #17249 [ skip ] completed with state |
4 participants
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.