Skip to content

fix: recover PR 2619 production ingest / 修复 PR 2619 生产环境写入 #2636

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
adibarra merged 11 commits into from
Aug 18, 2026
Merged

fix: recover PR 2619 production ingest / 修复 PR 2619 生产环境写入 #2636

Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
668e916
[AgentX] Mirror GB300 MTP tuning (#2571) to GB200 with 4-GPU->8-GPU w…
xinli-sw Aug 15, 2026
445d98f
Fill in pr-link for GB200 mtp2 changelog entry
xinli-sw Aug 15, 2026
0f144af
Switch GB200 mtp2 recipes to stock nightly image + regular MegaMoE
xinli-sw Aug 15, 2026
ecbb141
Revert "Switch GB200 mtp2 recipes to stock nightly image + regular Me…
xinli-sw Aug 15, 2026
97644bd
GB200 mtp2 disagg: lower Mooncake segment to 160GB, halve conc points
xinli-sw Aug 16, 2026
0cf99d4
Fix decode CUDA-graph capture size on GB200 c128 disagg recipe
xinli-sw Aug 16, 2026
3914b27
Lower GB200 c128 prefill gpu-memory-utilization to leave OOM headroom
xinli-sw Aug 16, 2026
c082cd5
GB200 mtp2 disagg: 140GB Mooncake segment, add 1p1d c256 point
xinli-sw Aug 16, 2026
b975a49
chore: prepare pr 2619 ingest recovery
adibarra Aug 18, 2026
bccd1d3
fix: recover pr 2619 production ingest
adibarra Aug 18, 2026
8553c3a
chore: attach reusable sweep run 31965016666
adibarra Aug 18, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 9 additions & 0 deletions perf-changelog.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6067,3 +6067,12 @@
- "Add GB200 DeepSeek-V4-Pro FP4 Dynamo-vLLM AgentX mirroring the GB300 PR #2571 MTP tuning, with every GB300 4-GPU worker sized to 8 GPUs on GB200."
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2619

- config-keys:
- dsv4-fp4-gb200-dynamo-vllm-agentic-mtp2-agg
- dsv4-fp4-gb200-dynamo-vllm-agentic-mtp2-disagg
scenario-type:
- agentic-coding
description:
- "Add GB200 DeepSeek-V4-Pro FP4 Dynamo-vLLM AgentX mirroring the GB300 PR #2571 MTP tuning, with every GB300 4-GPU worker sized to 8 GPUs on GB200."
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2636