Skip to content

[VideoGenX] sweep H3 latency on 1, 2, 4 and 8 B200 GPUs / [VideoGenX] 在 1、2、4、8 张 B200 上扫描 H3 延迟 - #4852

Merged
edwingao28 merged 5 commits into
mainfrom
feat/h3-b200-gpu-count-sweep
Nov 6, 2026
Merged

edwingao28 merged 5 commits into
mainfrom
feat/h3-b200-gpu-count-sweep

Conversation

@edwingao28

@edwingao28 edwingao28 commented Nov 6, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

B200 has one MiniMax-H3 C1 video point, on 4 GPUs, so it cannot show how video latency scales with GPU count.

Changes:

  • h3-bf16-b200-sglang-video adds C1 points on 1, 2 and 8 GPUs beside the existing 4-GPU point.
  • B200 recipe takes its Ulysses degree from the GPU count, as the H200 recipe does.
  • Docs (MODELS.md, video-generation.md) list the new points.

The 4-GPU point renders the same recipe as main: its fingerprint is 58f75909… before and after.

Results: MiniMax-H3 C1, h3-v2.yaml release-v1 (20 clips, 1344x768, 124 frames, 49 steps), SGLang v0.5.23, qualifying sweep 42155754705 attempt 2 at e95d6f970.

GPUs Job Valid clips Client-ready p50 (s) Speedup vs 1 GPU Total GPU power (W) GPU energy per clip (Wh)
1 job 20/20 134.0 1.00x 979 36.4
2 job 20/20 70.8 1.89x 1,957 38.5
4 job 20/20 36.2 3.70x 3,893 39.2
8 job 20/20 19.6 6.83x 7,630 41.9
  • The PR matrix plans exactly these 4 jobs and no evals. Attempt 1 hit a GitHub Actions internal error before any job ran.
  • validate-video-results accepted all 4 native deployments, and the result bundles carry the planned recipe fingerprints.
  • Every job has valid power telemetry.
  • The table reports each job's client-ready p50, from submit to downloaded video, which is the latency the dashboard plots. The job summary's latency_median_seconds also counts client-side media validation, which adds 0.54 to 0.59 s to each median.
  • The client polls job status every 0.5 s and sees a finished job only at the next poll, so each point includes 0 to 0.5 s of polling delay. [VideoGenX] poll H3 jobs every 0.1 s and rerun all four SKUs / [VideoGenX] H3 任务改为每 0.1 s 轮询并重跑四种硬件 #4904 polls every 0.1 s and reruns all four SKUs.

Related: #4818 (MI355X), #4856 (H200) and #4857 (H100) add the other SKUs. Whichever merges later reconciles the shared doc sentence and the changelog tail.

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.

@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.

I reviewed this PR and found no bugs. The NVIDIA master config, the B200 srt-slurm recipe and MODELS.md all have code owners, and the new 1-, 2- and 8-GPU points haven't run on hardware yet, so a human look is still worthwhile.

What was reviewed:

  • Recipe: the --ulysses-degree placeholder now matches --num-gpus, and with TP1 the 4-GPU point still renders as "4", the same as before.
  • nvidia-master.yaml: the gpus 1/2/8 entries follow the existing gpus 4 entry (same spec, set and recipe, conc-list [1]).
  • Changelog: I checked whether the entry should carry append-only: true. It should not, because infx/matrix/plan.py refuses append-only for video-generation and requires a regular sweep. The entry is one English block appended at the end of the file.
Extended reasoning...

This change adds 1-, 2- and 8-GPU C1 points to the MiniMax-H3 B200 SGLang video config. It also makes the B200 recipe take its Ulysses degree from the GPU count, updates two docs and appends one perf-changelog entry. It touches no security-sensitive code. I deferred for two reasons. First, CODEOWNERS assigns nvidia-master.yaml, the b200-* srt-slurm recipes and docs/MODELS.md to specific owners. Second, the PR says no GPU run has been done for the new topologies, including the 1-GPU point with encoder-parallel auto and no DiT CPU offload.

This review covers commit db05b15, which is no longer the latest commit on this pull request; later commits are not covered by it.

@github-actions

github-actions Bot commented Nov 6, 2026

Copy link
Copy Markdown
Contributor

@edwingao28

Copy link
Copy Markdown
Collaborator Author

/use 42155754705

@github-actions

github-actions Bot commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

@edwingao28 staged run 42155754705.

This run has no staged performance curves.

This run remains available across future /stage-results requests. Staging the same run ID again updates its staged data. Staging workflow

@edwingao28
edwingao28 merged commit c872d1a into main Nov 6, 2026
30 checks passed
@edwingao28
edwingao28 deleted the feat/h3-b200-gpu-count-sweep branch November 6, 2026 09:48
edwingao28 added a commit that referenced this pull request Nov 6, 2026
…X] 在 1、2、4、8 张 MI355X 上扫描 H3 延迟 (#4818)

* feat: run native H3 video sweeps from YAML recipes

* feat: verify GPU ownership and seal H3 execution evidence

* feat: define reproducible H3 workload sets

* feat: measure H3 requests and validate generated media

* refactor: delete unused OwnedProcess.owns helper

* refactor: delete unused H3 bundle JSON digest helper

* test: restore shared launch and workflow test cases

* refactor: fold H3 spec digest and case helpers

* test: table-drive H3 spec rejection cases

* chore: compact H3 spec YAML and drop stale references

* fix: use warmup case in GPU report slots

* refactor: drop unreachable GPU job guards

* test: drop dead parametrize branches in measurement tests

* refactor: share no-follow open and canonical JSON helpers

* test: inline single-case video parametrize and share reseal fixtures

* refactor: share deadline budget in GPU probes

* refactor: reuse runner slots and hashing in compare

* test: drop dead parametrize branches in GPU tests

* refactor: dedupe evaluator encoding, status, and quantile rules

* refactor: drop duplicated video digest, srun, schema and env helpers

* refactor: share video srt recipe defaults like agentic

* docs: link video pins and timeouts to their sources

* refactor: move offline compare and report tools out of measurement core

* refactor: move offline GPU job tools out of native H3 sweep

* refactor: move manual srt-yaml commands out of native H3 sweep

* style: format native H3 evaluator modules with ruff

* refactor: expose canonical_json_bytes as the one H3 hash encoding

* fix: check the video readiness port and drop the empty prepare description

* refactor: drop citations and per-engine notes from H3 specs

* test: drop the hand-built fairness hash test

* refactor: validate H3 v0.1 policy once in the spec loader by schema version

* ci: lint and format all of experimental/video-generation and drop measurement-only deps

* revert: keep run comparison and PSNR in the measurement core

* refactor: move SGLang server timing out of the measurement core

* chore: drop the unreferenced smoke plan and the nonexistent mvp extra hint

* style: ruff format the measurement core

* refactor: share the native video file contract from infx.bench.video

* refactor: hash run bundles with the spec's canonical JSON

* refactor: name each rejected field of the sealed v0.1 server launch

* refactor: trust loader-validated plans and drop the redundant run_plan model_revision

* refactor: derive engine encoding requests and media checks from the plan's encoding

* refactor: keep native video checks at the binder, executor and ingest boundaries

* chore: drop unconsumed prose and constant fields from run data

* fix: keep a bounded redacted HTTP error excerpt and refuse every paid MiniMax host

* refactor: default the single-choice reference dtype and attention backend

* refactor: record native video job status as JSON

* fix: count malformed power samples in the video bundle instead of skipping them

* fix: drop unread constant bundle fields and validate each power sample's GPU coverage

* refactor: resolve the video harness at one path beside the project

* fix: hash the measurement window run with the shared canonical JSON

* refactor: seal the image digest for every native video plan and check the launch once

* fix: keep the video workload specs in nested matrix snapshots

* fix: report the native video bundle outcome as its CI phase

* fix: count video sweep jobs in run stats after they finish

* fix: drop the static protocol source pin so run configuration names only the image revision

* refactor: rename generic H3 spec and workflow names to video

* refactor: rename generic H3 measurement-core naming to video

* refactor: rename generic H3 native-sweep naming to video

* fix: record approved head and video matrix in trusted sweeps

* fix: validate trusted video sweeps with the dispatching workflow's validator

* feat: run H3 on SGLang v0.5.23 for B200 and H100 (#4776)

* fix: record engine parallelism in native video spec.json (#4777)

* chore: fold the v0.5.23 changelog entry into the H3 sweep entry

* chore: drop the H200 vLLM-Omni cell from the H3 sweep

* chore: run H3 H200 on SGLang v0.5.23

* chore: drop the H200 SGLang cell from the H3 sweep

* refactor: move the video harness to top-level video-generation

* refactor: name the 20-prompt H3 spec by its schema version

* docs: drop the release label from the runner's policy note

* fix: let vgbench run record the server's numerical identity

* fix: check reference case IDs before selecting all cases

* docs: drop the release label from the H3 docs and changelog entry

* fix: accept the newest video bundle per deployment across rerun attempts

* fix: retry the video results upload like other result uploads

* fix: stop serving-mode video runs promptly on interrupt

* fix: name ROCm video GPUs by board serial

* feat: add MiniMax-H3 MI355X SGLang video sweep

* feat: serve MI355X H3 video from the shared HF cache

* docs: link the MI355X H3 changelog entry to its PR

* fix: select AITER attention in the MI355X H3 recipe

* feat: sweep H3 video latency across 1, 2, 4 and 8 GPUs

* fix: keep a single video bundle in its named directory

[skip-sweep]

* chore: link GPU-count changelog entry to #4818

[skip-sweep]

* ci: raise video job limit to 150 minutes for 1-GPU points

* refactor: share single-node srt steps between points and video

[skip-sweep]

* feat: sweep B200 H3 video across 1, 2, 4 and 8 GPUs

* chore: move B200 GPU-count points to #4852

* chore: narrow the GPU-count sweep to MI355X
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

full-sweep-fail-fast Full sweep with canary gate; first failure cancels the rest of that matrix (recommended)

Projects

Status: Done

Development

Successfully merging this pull request may close these issues.

1 participant