Skip to content

[NVIDIA][AgentX] Add GB300 DSV4.1-Flash Dynamo + TensorRT-LLM disagg / [NVIDIA][AgentX] 新增 GB300 DSV4.1-Flash Dynamo + TensorRT-LLM 分离式配置 - #4991

Open
jh-nv wants to merge 5 commits into
SemiAnalysisAI:mainfrom
jh-nv:jihao/dsv41flash-gb300-dynamo-trt-agentx
Open

jh-nv wants to merge 5 commits into
SemiAnalysisAI:mainfrom
jh-nv:jihao/dsv41flash-gb300-dynamo-trt-agentx

Conversation

@jh-nv

@jh-nv jh-nv commented Nov 10, 2026 •

Copy link
Copy Markdown

Summary

Add dsv41flash-fp4-gb300-dynamo-trt-agentic-dspark-disagg: DeepSeek-V4.1-Flash NVFP4 (nvidia/DeepSeek-V4.1-Flash-NVFP4) AgentX on GB300 with Dynamo + TensorRT-LLM disaggregated serving. Seven points reuse the topologies, engine settings and concurrencies the TensorRT-LLM team tuned for trtllm-serve AgentX on this model. Five more points (c11718, c2286, c1283, c180, c114) fill the regions that Pareto leaves empty.

  • Image: nvcr.io/nvidia/tensorrt-llm/release:1.4.0rc4.dev202611160007 (pinned by digest), a TensorRT-LLM main nightly built from commit ab2dfeba8e69. It includes the DeepSeek-V4.1-Flash optimizations of NVIDIA/TensorRT-LLM#21438 (merged 2026-11-14), which these recipes use: MegaMoE shared-expert fusion, gen_prompt_tail_tokens, EPLB warmup_placement, host_cache_bind_to_gpu_numa_node, strict conversation affinity, max_gen_init_admissions_per_pass and enable_client_io_during_gpu_wait.
  • Dynamo: the 1.7.0.dev20261110 wheel, installed at job start. It is the first nightly with ai-dynamo/dynamo#18170.
Point Prefill Decode Target MoE backend Decode batch DSpark draft GPUs
c11718 10 × DEP2 1 × DEP16 MegaMoE CuTe DSL 256 3 36
c7000 5 × DEP2 1 × DEP8 MegaMoE CuTe DSL 320 3 18
c7900 7 × DEP2 1 × DEP16 MegaMoE CuTe DSL 128 3 30
c5100 5 × DEP2 1 × DEP16 MegaMoE CuTe DSL 64 3 26
c7600 7 × DEP2 1 × DEP32 MegaMoE CuTe DSL 25 5 46
c2286 3 × DEP2 1 × DEP32 MegaMoE CuTe DSL 8 5 38
c1283 2 × DEP2 1 × DEP32 MegaMoE CuTe DSL 4 5 36
c240 1 × DEP2 4 × TEP8 TRTLLM 4 3 34
c180 1 × DEP2 6 × TP4 TRTLLM 2 5 26
c140 1 × DEP2 4 × TEP8 TRTLLM 2 5 34
c114 1 × DEP2 6 × TP4 TRTLLM 1 5 26
c84 1 × DEP2 4 × TEP8 TRTLLM 1 5 34

Each DEP2 prefill rank keeps a native host KV tier on its GPU's NUMA node: 125 GiB at the DEP8/DEP16 points, 111 GiB at the DEP32 points and 80 GiB at the TEP8/TP4 points. Every decode worker streams every 20 iterations (stream_interval: 20); prefill keeps TensorRT-LLM's default of 1. Both roles use an FP8 KV cache (kv_cache_config.dtype: fp8). The MoE combine stays BF16 (TensorRT-LLM's default; use_low_precision_moe_combine is not set).

The DSpark draft runs as it ships. Its routed experts are MXFP4 in the checkpoint and run with MXFP8 activations (W4A8_MXFP4_MXFP8, the image default). At the DEP points the draft's MoE runs on CUTLASS (speculative_config.moe_backend: CUTLASS): left unset, the draft would inherit the target's MegaMoE backend, which serves NVFP4 experts only, and setting MegaMoE for the draft explicitly re-encodes them as NVFP4. At the TEP8/TP4 points the draft inherits the target's TRTLLM backend, which keeps the same MXFP4 path.

Dynamo settings

  • DEP points (c11718, c7000, c7900, c5100, c7600, c2286, c1283): KV router with session affinity. Prefill workers start a new conversation on the attention-DP rank the router chose (DYN_ENGINE_CONV_AFFINITY_DP_RANK_SOURCE=dynamo). This puts a subagent's first turn on the rank that holds its parent's prefix; in our runs it raised the first-turn prefix hit at c7600 from 0.29 to 0.42 and output throughput at c7000 by 4.9%. The AgentX client runs on a dedicated node.
  • TEP8 and TP4 points (c240, c180, c140, c114, c84): least-loaded decode routing without session affinity. Session affinity pinned each session to one small-batch decode worker and left others idle.
  • All points: the frontend and workers pass token IDs as packed bytes (DYN_TOKEN_IDS_AS_BYTES=1). Workers are bound to their GPU's NUMA node for CPUs (numa_cpu_bind) and prefer it for memory (numa_memory_bind: preferred).

Metrics and power

Every worker publishes TensorRT-LLM iteration and KV-cache metrics (engine.publish_metrics: true, no KV events). iter_perf_stats_interval and kv_cache_config.iteration_stats_interval are 40 on prefill and TEP8/TP4 decode and 400 on DEP decode. The client requires trtllm_kv_cache_utilization (AIPERF_REQUIRED_SERVER_METRIC_PREFIX). Every point sets power: true.

Per-request perf metrics stay off, which is the default since ai-dynamo/dynamo#18170 (--publish-per-request-metrics). On earlier Dynamo builds, --publish-metrics also turned them on. TensorRT-LLM's record_stats then rebuilt each finished ~130K-token prompt as a Python list in the decode handler and saturated it: total tok/s/GPU dropped 39.5% at c7900, 30.9% at c7600, 18.8% at c5100 and about 13% at c7000. NVIDIA/TensorRT-LLM#21674 makes those metrics cheap for anyone who turns them on. This config does not turn them on.

Validation. All twelve points ran end to end on an internal NVIDIA GB300 NVL72 cluster (no public run links) with metrics and power on, on the feat/ds-v4.1-bringup release image that #21438 was merged from (TensorRT-LLM 054a88929a1). In those runs the DEP points' draft MoE was still on MegaMoE. c7900, c5100 and c7600 were within 1% of the same recipes with metrics and power off. c7000 was 3.3% lower over the part of its window it measured; that cause is not isolated. The TEP8 points were 1–3% lower, which is what the shared DCGM telemetry block (--collect-interval=100) costs these small-batch points; the cost goes away at 1000 ms. Those runs also used an FP8 MoE combine on decode. Moving the DEP points' draft to CUTLASS, as in this PR, costs another 1–2% output throughput, and returning the combine to BF16 costs at most 0.5% (short runs). Short runs of this PR's exact config on the main nightly match the earlier image: output throughput +0.7% at c7900, +0.1% at c7600, +1.2% at c84 and +2.5% at c180.

Validation details

The runs used srt-slurm at this repository's pin with a 60-minute AgentX window and 3 warmup requests per lane; cluster scheduler settings were added at submission. Dynamo was 1.7.0.dev20261109 with #18170's Python change applied (the two files it touches), because no nightly carried #18170 yet; 1.7.0.dev20261110 ships that same change. Apart from that, the image, the draft MoE backend and the MoE combine, the bound recipes are identical to this PR; the short runs below cover those three differences.

Median tok/s/user is 1000 / p50 ITL. The baseline is the same recipe with metrics and power off; in those runs the seven reused points were within 1% of the TensorRT-LLM team's own trtllm-serve measurements in output tok/s/GPU.

Point Median tok/s/user Output tok/s/GPU Total tok/s/GPU vs metrics and power off
c11718 130.3 10,593 1,447,284 new
c7000 87.7 11,009 1,492,133 −3.3% (first 3,068 s)
c7900 161.3 10,156 1,404,065 −0.3%
c5100 230.4 8,334 1,172,727 +0.1%
c7600 375.1 6,074 827,879 −0.6%
c2286 406.0 2,563 356,859 new
c1283 488.3 1,623 230,308 new
c240 656.1 269.4 37,181 −1.2%
c180 706.8 298.3 42,005 new
c140 885.0 183.4 23,631 −1.4%
c114 858.4 183.0 23,786 new
c84 1,115.2 113.2 14,427 −3.1%
  • c11718 and c7000 reached the cluster's 2-hour job limit before AIPerf wrote its summary, so their rows come from the per-request export: c11718 covers the full 3,600 s, c7000 its first 3,068 s.
  • The DEP rows above used the MegaMoE draft. Short runs with this PR's CUTLASS draft (900 s window, same build and client), compared with the first 900 s of the runs above:
Point p10 tok/s/user Total tok/s/GPU Output tok/s/GPU
c7000 −1.7% −1.0% −1.1%
c7900 −1.5% −1.0% −1.2%
c7600 −2.8% −2.1% −1.9%
  • The rows above also used an FP8 MoE combine (use_low_precision_moe_combine: true). With this PR's BF16 combine, short runs of the pinned config (900 s window) against the same config with the FP8 combine:
Point p10 tok/s/user Total tok/s/GPU Output tok/s/GPU
c7900 −0.2% −0.3% −0.3%
c7600 −0.5% −0.4% −0.4%
c84 −0.3% −0.2% −0.5%
  • This PR's exact config (main nightly 1.4.0rc4.dev202611160007, Dynamo 1.7.0.dev20261110, CUTLASS draft, BF16 combine), short runs (900 s window) against the same config on the earlier image:
Point p10 tok/s/user Total tok/s/GPU Output tok/s/GPU
c7900 +0.9% +0.8% +0.7%
c7600 +0.2% +0.2% +0.1%
c84 +1.3% +0.2% +1.2%
c180 +3.1% +2.4% +2.5%

c180's earlier run also used the FP8 combine and the patched 1.7.0.dev20261109.

Metrics collection (Check 17)
  • Installed Dynamo: 1.7.0.dev20261110 (dynamo.source.wheel and router.version; not inferred from the container tag). Its dynamo/trtllm/backend_args.py has --publish-per-request-metrics, and workers/llm_worker.py sets return_perf_metrics = config.publish_per_request_metrics. It contains #12162, so --publish-metrics and KV events are independent controls.
  • engine.publish_metrics: true makes srtctl pass --publish-metrics to every prefill and decode worker. KV events stay off (event_buffer_max_size: 0). Iteration stats are on (enable_iter_perf_stats, iter_perf_stats_interval, kv_cache_config.iteration_stats_interval).
  • In the validation runs every worker's /metrics exposed trtllm_kv_cache_utilization. The client requires that metric, so a point without it fails.
AI model disclosure
  • Model/version: Claude Opus 5.5 (claude-opus-5-5), in Claude Code; one delegated agent on the same model.
  • Role: Claude Opus 5.5 wrote the recipe, master config and changelog entry, ran the validation runs and their analysis, and wrote this description. The delegated agent reviewed the diff independently against .github/codeowner-signoff-verify-prompt.md.
PR checklist and change type

Type of Change

  • Bug fix
  • New feature
  • Configuration change
  • Documentation update
  • Other (please describe)

Checklist

  • I have completed the AI model disclosure and kept it current
  • I have tested my changes locally
  • I have updated documentation if necessary
  • If python -m infx.matrix.changed --base origin/main lists affected config keys, I have appended one new entry covering them to the physical end of inferencex-e2e/perf-changelog.yaml; if it lists none, I have added none. I have not edited historical entries
  • If this PR appends an inferencex-e2e/perf-changelog.yaml entry, it carries exactly one primary sweep label (a maintainer applies it on fork PRs): full-sweep-fail-fast (recommended), full-sweep-enabled, or non-canary-full-sweep-enabled. Optional modifiers all-evals, evals-only, and agentx-fast require a primary label; the last two block reuse while applied.
  • Before merging via reuse, an authorized maintainer (OWNER/MEMBER/COLLABORATOR) has commented /use <run_id> (or the legacy /reuse-sweep-run) on this PR. Do this only once there is a final full sweep that is all green with evals passing, since after this comment the primary sweep label will no longer automatically kick off new sweeps. Remove and re-add the primary sweep label to force a new sweep.
中文

新增 dsv41flash-fp4-gb300-dynamo-trt-agentic-dspark-disagg:DeepSeek-V4.1-Flash NVFP4(nvidia/DeepSeek-V4.1-Flash-NVFP4)在 GB300 上使用 Dynamo + TensorRT-LLM 分离式部署运行 AgentX。其中七个点沿用 TensorRT-LLM 团队为 trtllm-serve AgentX 在该模型上调优的拓扑、引擎设置和并发;另外五个点(c11718、c2286、c1283、c180、c114)补齐该 Pareto 未覆盖的区域。

镜像为 nvcr.io/nvidia/tensorrt-llm/release:1.4.0rc4.dev202611160007(按 digest 固定),即基于提交 ab2dfeba8e69 构建的 TensorRT-LLM main nightly。它包含 NVIDIA/TensorRT-LLM#21438(2026-11-14 合入)中的 DeepSeek-V4.1-Flash 优化,这些配方用到其中的 MegaMoE shared expert 融合、gen_prompt_tail_tokens、EPLB warmup_placement、host_cache_bind_to_gpu_numa_node、严格会话亲和、max_gen_init_admissions_per_pass 与 enable_client_io_during_gpu_wait。Dynamo 在作业启动时安装 1.7.0.dev20261110 wheel,这是第一个包含 ai-dynamo/dynamo#18170 的 nightly。

配置点见上方表格:c11718(10×DEP2 + 1×DEP16,batch 256)、c7000(5×DEP2 + 1×DEP8,batch 320)、c7900(7×DEP2 + 1×DEP16,batch 128)、c5100(5×DEP2 + 1×DEP16,batch 64)、c7600(7×DEP2 + 1×DEP32,batch 25)、c2286(3×DEP2 + 1×DEP32,batch 8)、c1283(2×DEP2 + 1×DEP32,batch 4),目标模型 MoE 均使用 MegaMoE CuTe DSL;c240、c140、c84(1×DEP2 + 4×TEP8,batch 4/2/1)与 c180、c114(1×DEP2 + 6×TP4,batch 2/1)使用 TRTLLM MoE。DEP2 预填充 worker 的每个 rank 在其 GPU 所在 NUMA 节点上保留原生主机 KV 层:DEP8/DEP16 点为 125 GiB,DEP32 点为 111 GiB,TEP8/TP4 点为 80 GiB。所有解码 worker 每 20 次迭代输出一次(stream_interval: 20),预填充使用 TensorRT-LLM 默认值 1。两个角色均使用 FP8 KV cache(kv_cache_config.dtype: fp8)。MoE combine 保持 BF16(TensorRT-LLM 默认值,不设置 use_low_precision_moe_combine)。

DSpark 草稿模型按发布时的精度运行:其路由专家在检查点中为 MXFP4,并以 MXFP8 激活运行(W4A8_MXFP4_MXFP8,镜像默认路径)。DEP 点的草稿 MoE 使用 CUTLASS(speculative_config.moe_backend: CUTLASS):若不设置,草稿会继承目标模型的 MegaMoE 后端,而 MegaMoE 只支持 NVFP4 专家;若为草稿显式设置 MegaMoE,则会把这些专家重新编码为 NVFP4。TEP8/TP4 点的草稿继承目标模型的 TRTLLM 后端,保持同样的 MXFP4 路径。

Dynamo 设置:DEP 点使用带会话亲和的 KV 路由,预填充 worker 让新会话使用路由器选定的 attention-DP rank(DYN_ENGINE_CONV_AFFINITY_DP_RANK_SOURCE=dynamo),使子代理的首轮请求落在持有其父会话前缀的 rank 上,AgentX 客户端运行在独立节点上;TEP8 与 TP4 点的解码按最小负载路由,不使用会话亲和。所有点的 frontend 与 worker 以打包字节传递 token ID(DYN_TOKEN_IDS_AS_BYTES=1),worker 绑定 GPU 所在 NUMA 节点的 CPU,并优先使用该节点内存。

指标与功耗:所有 worker 发布 TensorRT-LLM 迭代与 KV cache 指标(engine.publish_metrics: true,不启用 KV 事件),预填充与 TEP8/TP4 解码的 iter_perf_stats_interval 和 kv_cache_config.iteration_stats_interval 为 40,DEP 解码为 400;客户端要求 trtllm_kv_cache_utilization 指标;所有点设置 power: true。逐请求性能指标保持关闭,这是 ai-dynamo/dynamo#18170(--publish-per-request-metrics)之后的默认行为。在更早的 Dynamo 中,--publish-metrics 会同时开启逐请求指标,TensorRT-LLM 的 record_stats 随后会在解码 handler 中把每个约 130K token 的 prompt 重建为 Python 列表并使其饱和:total tok/s/GPU 在 c7900 下降 39.5%、c7600 下降 30.9%、c5100 下降 18.8%、c7000 约下降 13%。NVIDIA/TensorRT-LLM#21674 让开启这些指标的用户开销很低;本配置不开启它们。

验证:全部十二个点在一个内部 NVIDIA GB300 NVL72 集群(无公开运行链接)上使用本 PR 绑定后的配方端到端运行(开启指标与功耗),当时 DEP 点的草稿 MoE 仍使用 MegaMoE。c7900、c5100、c7600 与关闭指标和功耗的同一配方相差不超过 1%;c7000 在其测得的窗口内低 3.3%,原因尚未定位;TEP8 点低 1–3%,即共享 DCGM 遥测配置(--collect-interval=100)对这些小 batch 点造成的代价,改为 1000 ms 后该代价消失。这些运行的解码还使用了 FP8 MoE combine。按本 PR 将 DEP 点的草稿改为 CUTLASS 后,输出吞吐再下降 1–2%(c7000、c7900、c7600 的短时运行,900 s 窗口,与上述运行的前 900 s 相比:output tok/s/GPU 分别 −1.1%、−1.2%、−1.9%);将 MoE combine 恢复为 BF16 的代价不超过 0.5%(c7900、c7600、c84 的短时运行:output tok/s/GPU 分别 −0.3%、−0.4%、−0.5%)。这些全量运行使用 #21438 所来自的 feat/ds-v4.1-bringup release 镜像(TensorRT-LLM 054a88929a1),Dynamo 为 1.7.0.dev20261109 并应用 #18170 的 Python 改动(当时尚无包含 #18170 的 nightly);1.7.0.dev20261110 包含同样的改动。在 main nightly 上以本 PR 的完整配置进行的短时运行与旧镜像一致:output tok/s/GPU 在 c7900 为 +0.7%、c7600 为 +0.1%、c84 为 +1.2%、c180 为 +2.5%。各点数据见上方 Validation details 表格。

AI 模型使用说明:使用 Claude Code 中的 Claude Opus 5.5(claude-opus-5-5)编写配方、master 配置与 changelog 条目,运行验证并分析结果,撰写本说明;另有一个同模型的委派 agent 按 .github/codeowner-signoff-verify-prompt.md 独立审查了改动。

改动类型:配置改动(Configuration change)。

🤖 Generated with Claude Code

jh-nv and others added 2 commits November 10, 2026 13:36
Twelve AgentX points for nvidia/DeepSeek-V4.1-Flash-NVFP4 on GB300 with Dynamo
1.7.0.dev20261110 (first nightly with ai-dynamo/dynamo#18170) over the TensorRT-LLM
feat/ds-v4.1-bringup release image: the seven points of the trtllm-serve AgentX
Pareto plus c11718, c2286, c1283, c180 and c114. Metrics and power on, decode
stream_interval 20.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The seven DEP points set the draft MoE backend to MEGAMOE_CUTEDSL, which
re-encodes the checkpoint's MXFP4 draft experts as NVFP4 with static
activation scales. CUTLASS keeps their default W4A8_MXFP4_MXFP8 path.
The changelog entry now also names the FP8 KV cache and the FP8 MoE
combine on decode, and a recipe comment no longer says workers publish
per-request metrics.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

@claude claude Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

…namo-trt-agentx-sa

# Conflicts:
#	inferencex-e2e/perf-changelog.yaml
@jh-nv
jh-nv marked this pull request as draft November 10, 2026 23:19

dsv41flash-fp4-gb300-dynamo-trt-agentic-dspark-disagg:
image: nvcr.io/nvidia/tensorrt-llm/release:feat-ds-v4.1-bringup-1.4.0a0.dev202610310238@sha256:960406744410b9601d134334d2b4f468f8fee5875f9194aaad1dccea5d7880e7
model: nvidia/DeepSeek-V4.1-Flash-NVFP4

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

do u have evals for this on terminalbench 4.0? showing that nvfp4 quant is just as good as mxfp4?

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

logs, manual reprod scripts, etc.

tokens_per_block: 128
max_seq_len: 996595
moe_config:
use_low_precision_moe_combine: true

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

do u have tb4.0 and deepswe evals (along with manual reprod scripts & logs) as discussed with @xinli-sw about lower precision reductions & nvfp4 dsv4.1 flash

jh-nv and others added 2 commits November 16, 2026 14:50
The image is now nvcr.io/nvidia/tensorrt-llm/release:1.4.0rc4.dev202611160007
(main commit ab2dfeba8e69), which includes NVIDIA/TensorRT-LLM#21438, so the
recipes no longer need the feat/ds-v4.1-bringup branch. Decode no longer sets
use_low_precision_moe_combine, so the MoE combine stays BF16.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…namo-trt-agentx-sa

# Conflicts:
#	inferencex-e2e/configs/nvidia-master.yaml
#	inferencex-e2e/perf-changelog.yaml
@jh-nv jh-nv changed the title [NVIDIA][AgentX] Add GB300 DSV4.1-Flash Dynamo + TensorRT-LLM disagg on the feat/ds-v4.1-bringup release / [NVIDIA][AgentX] 新增基于 feat/ds-v4.1-bringup release 的 GB300 DSV4.1-Flash Dynamo + TensorRT-LLM 分离式配置 [NVIDIA][AgentX] Add GB300 DSV4.1-Flash Dynamo + TensorRT-LLM disagg / [NVIDIA][AgentX] 新增 GB300 DSV4.1-Flash Dynamo + TensorRT-LLM 分离式配置 Nov 16, 2026
@jh-nv
jh-nv marked this pull request as ready for review November 16, 2026 20:14

@claude claude Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

@github-actions

Copy link
Copy Markdown
Contributor

Thanks for the contribution!

  • Review: If this PR changes files owned by someone other than a repository admin or @SemiAnalysisAI/core, ask one eligible CODEOWNER to complete the latest PR_REVIEW_CHECKLIST.md before contacting a core maintainer on Slack. Follow the template exactly, including As a PR reviewer and CODEOWNER, I have reviewed this and have, so sign-off verification triggers.
  • PR verification: Sweeps run only when the PR appends an inferencex-e2e/perf-changelog.yaml entry and carries exactly one primary label: full-sweep-fail-fast (strongly recommended; canary plus per-matrix fail-fast), full-sweep-enabled (canary; matrix jobs continue after a failure), or non-canary-full-sweep-enabled (no canary or fail-fast). The modifiers all-evals, evals-only, and agentx-fast require a primary label. On fork PRs, a maintainer applies the label. See sweep labels and reuse.
  • After merging: PR authors must ensure all GitHub Actions jobs pass. Transient failures often pass on rerun; see how to rerun failed jobs.
中文

感谢你的贡献!

  • **审阅:**如果 PR 修改的文件归属于仓库管理员及 @SemiAnalysisAI/core 之外的 CODEOWNER,请先联系一位有资格的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md,再通过 Slack 联系核心维护者。必须严格遵循模板,并保留 As a PR reviewer and CODEOWNER, I have reviewed this and have,才能触发签核验证。
  • **PR 验证:**只有当 PR 在 inferencex-e2e/perf-changelog.yaml 末尾追加了条目,并且恰好带有一个主标签时,才会运行扫描:full-sweep-fail-fast(强烈推荐;canary 加逐矩阵 fail-fast)、full-sweep-enabled(有 canary;矩阵任务在失败后继续运行)或 non-canary-full-sweep-enabled(无 canary,也无 fail-fast)。修饰标签 all-evals、evals-only 和 agentx-fast 必须与主标签一起使用。fork PR 的标签由维护者添加。参见扫描标签与复用。
  • **合并后:**PR 作者必须确保所有 GitHub Actions 任务通过。临时性失败通常可以通过重新运行恢复;参见重新运行失败任务的说明。

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Status: No status

Development

Successfully merging this pull request may close these issues.

2 participants