Skip to content

[nv] add H200 SGLang disagg configs from srtslurm#582

Merged
ishandhanani merged 9 commits intomainfrom
nv/h200-sglang-disagg
Feb 5, 2026
Merged

[nv] add H200 SGLang disagg configs from srtslurm#582
ishandhanani merged 9 commits intomainfrom
nv/h200-sglang-disagg

Conversation

@ishandhanani
Copy link
Collaborator

Summary

Add H200 SGLang disaggregated multinode configurations, sourced from srtslurm recipes.

Depends on #570

Changes

  • Add dsr1-fp8-h200-dynamo-sglang config to nvidia-master.yaml
  • 1k1k: aggregated, low-latency (1P9D), high-throughput TEP/DEP (1P6D)
  • 8k1k: aggregated, TEP variants (1P7D, 1P6D, 1P3D, 2P3D), DEP (1P1D)
  • Add perf-changelog entry
  • Document recipe registration from srtslurm in AGENT.md

Config Details

ISL/OSL Mode Workers Concurrencies
1k1k Aggregated 1P 1-512
1k1k Low latency TEP 1P9D 1-256
1k1k High throughput TEP 1P6D 512-2048
1k1k High throughput DEP 1P6D 128-2048
8k1k Aggregated 1P 1-256
8k1k TEP variants 1P7D/1P6D/1P3D/2P3D 1-128
8k1k DEP 1P1D 64-256

@cquil11
Copy link
Collaborator

cquil11 commented Jan 27, 2026

great thx! sorry bout that

@ishandhanani
Copy link
Collaborator Author

All good friend

@ishandhanani ishandhanani changed the base branch from main to nv/dsr1-fp8-h200-dynamo-trtllm-260126 January 27, 2026 19:57
@ishandhanani ishandhanani changed the title Add H200 SGLang disagg configs from srtslurm [DO NOT MERGE] Add H200 SGLang disagg configs from srtslurm Jan 27, 2026
@ishandhanani ishandhanani force-pushed the nv/h200-sglang-disagg branch from 1baaf6d to 614d356 Compare January 27, 2026 20:01
Base automatically changed from nv/dsr1-fp8-h200-dynamo-trtllm-260126 to main January 29, 2026 15:57
@ishandhanani ishandhanani changed the title [DO NOT MERGE] Add H200 SGLang disagg configs from srtslurm [nv] add H200 SGLang disagg configs from srtslurm Feb 2, 2026
- Add dsr1-fp8-h200-dynamo-sglang config to nvidia-master.yaml
- Include 1k1k configs: aggregated, low-latency (1P9D), high-throughput TEP/DEP (1P6D)
- Include 8k1k configs: aggregated, TEP variants (1P7D, 1P6D, 1P3D, 2P3D), DEP (1P1D)
- Add perf-changelog entry for new configuration
- Document recipe registration process in AGENT.md
@ishandhanani ishandhanani force-pushed the nv/h200-sglang-disagg branch from aaf533d to a1f83d7 Compare February 2, 2026 18:19
ishandhanani and others added 2 commits February 2, 2026 10:20
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
@ishandhanani
Copy link
Collaborator Author

@cquil11
Copy link
Collaborator

cquil11 commented Feb 2, 2026

https://github.com/InferenceMAX/InferenceMAX/actions/runs/21603489954/job/62255208721
some pesky errors wrt files not being cleaned up

@cquil11
Copy link
Collaborator

cquil11 commented Feb 2, 2026

othweriwise looks good to me

@ishandhanani
Copy link
Collaborator Author

othweriwise looks good to me

Yep - this was just a config issue on my end. Fixed it just now

Copy link
Collaborator

@cquil11 cquil11 left a comment

Choose a reason for hiding this comment

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

please merge at your convenience. ty ishan!

@ishandhanani
Copy link
Collaborator Author

# Conflicts:
#	perf-changelog.yaml
…enceMAX into nv/h200-sglang-disagg

# Conflicts:
#	perf-changelog.yaml
@ishandhanani ishandhanani merged commit f0e131f into main Feb 5, 2026
13 checks passed
@ishandhanani ishandhanani deleted the nv/h200-sglang-disagg branch February 5, 2026 20:32
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

Development

Successfully merging this pull request may close these issues.

3 participants