Skip to content

[AgentX] Update Muse Spark Open B200 SGLang image and add c64 / [AgentX] 更新 Muse Spark Open B200 SGLang 镜像并新增 c64 #4400

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
Oseltamivir merged 2 commits into from
Oct 23, 2026
Merged

[AgentX] Update Muse Spark Open B200 SGLang image and add c64 / [AgentX] 更新 Muse Spark Open B200 SGLang 镜像并新增 c64 #4400

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
11358f4
[AgentX] Update Muse Spark Open B200 SGLang image and add c64 / [Agen…
shyeh25 Oct 22, 2026
8ec8761
Merge branch 'main' into musespark-fp4-b200-sglang-agentic-c1c287a4-c64
Oseltamivir Oct 23, 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
34 changes: 31 additions & 3 deletions ...x-e2e/benchmarks/single_node/srt-slurm-recipes/musespark/sglang/b200-fp4-mtp/agentic.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,7 @@ base:
SGLANG_TIMEOUT_KEEP_ALIVE: '1800'
OMP_NUM_THREADS: '8'
SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN: '1'
SGLANG_FLASHINFER_NVFP4_PER_TOKEN_ACTIVATION: '0'
SGLANG_UNIFIED_RADIX_TREE_CORE_BACKEND: python
HF_HUB_OFFLINE: '1'
health_check:
Expand Down Expand Up @@ -179,8 +180,10 @@ override_tp8_c56:
args:
max-running-requests: 56
cuda-graph-max-bs-decode: 56
chunked-prefill-size: 32768
max-prefill-tokens: 32768
schedule-policy: shortest-prefill-first
prefill-decode-interval: 48
prefill-decode-interval: 64
speculative-draft-attention-backend: null
swa-full-tokens-ratio: 0.25
enable-layernorm-sp: true
Expand All @@ -191,13 +194,38 @@ override_tp8_c56:
SGLANG_OPT_DEEPGEMM_MEGA_MOE_NUM_MAX_TOKENS_PER_RANK: '8192'
SGLANG_OPT_UNIFIED_CACHE_FREE_OUT_OF_WINDOW_SLOTS: '1'
SGLANG_TRTLLM_MHA_DECODE_SEQ_LEN_SPLITS: '4'
SGLANG_UNIFIED_RADIX_TREE_CORE_BACKEND: null
benchmark:
env:
CONC: '56'
KV_OFFLOADING: dram


override_tp8_c64:
roles:
agg:
args:
max-running-requests: 64
cuda-graph-max-bs-decode: 64
chunked-prefill-size: 32768
max-prefill-tokens: 32768
schedule-policy: shortest-prefill-first
prefill-decode-interval: 48
speculative-draft-attention-backend: null
swa-full-tokens-ratio: 0.25
enable-layernorm-sp: true
enable-megamoe-for-layernorm-sp: true
enable-hierarchical-cache: true
hicache-size: 160
env:
SGLANG_OPT_DEEPGEMM_MEGA_MOE_NUM_MAX_TOKENS_PER_RANK: '8192'
SGLANG_OPT_UNIFIED_CACHE_FREE_OUT_OF_WINDOW_SLOTS: '1'
SGLANG_TRTLLM_MHA_DECODE_SEQ_LEN_SPLITS: '4'
benchmark:
env:
CONC: '64'
KV_OFFLOADING: dram


override_tp4_c8:
roles:
agg:
Expand Down Expand Up @@ -294,7 +322,7 @@ override_tp4_c24:
expert-parallel-size: 4
max-running-requests: 48
cuda-graph-max-bs-decode: 48
mem-fraction-static: 0.95
mem-fraction-static: 0.98
chunked-prefill-size: 8192
max-prefill-tokens: 8192
schedule-policy: shortest-prefill-first
Expand Down
4 changes: 2 additions & 2 deletions inferencex-e2e/configs/nvidia-master.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8701,7 +8701,7 @@ musespark-fp4-gb200-sglang-agentic-dflash:
- { tp: 4, ep: 4, kv-offloading: dram, kv-offload-backend: { name: hicache }, spec-decoding: draft_model, conc-list: [2, 4, 6, 8, 10, 16, 20, 24, 28, 32], srt-recipe: agentic.yaml }

musespark-fp4-b200-sglang-agentic-dflash:
image: lmsysorg/sglang:nightly-dev-20261019-3c560089
image: lmsysorg/sglang:nightly-dev-20261021-c1c287a4
model: meta-models/Muse-Spark-Open-816B-A42B-NVFP4
model-prefix: musespark
runner: cluster:b200-nscale
Expand All @@ -8714,7 +8714,7 @@ musespark-fp4-b200-sglang-agentic-dflash:
- dram-utilization: 0.90
search-space:
- { tp: 8, ep: 8, kv-offloading: none, spec-decoding: draft_model, conc-list: [1, 2, 4, 6, 8, 10], srt-recipe: agentic.yaml }
- { tp: 8, ep: 8, kv-offloading: dram, kv-offload-backend: { name: hicache }, spec-decoding: draft_model, conc-list: [56], srt-recipe: agentic.yaml }
- { tp: 8, ep: 8, kv-offloading: dram, kv-offload-backend: { name: hicache }, spec-decoding: draft_model, conc-list: [56, 64], srt-recipe: agentic.yaml }
- { tp: 4, ep: 4, kv-offloading: dram, kv-offload-backend: { name: hicache }, spec-decoding: draft_model, conc-list: [8, 10, 16, 24], srt-recipe: agentic.yaml }

musespark-fp4-h200-sglang-agentic-dflash:
Expand Down
9 changes: 9 additions & 0 deletions inferencex-e2e/perf-changelog.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10779,6 +10779,15 @@
- "Enable SGLANG_HIP_FP4_INDEXER_ROWGROUP=1 on all points. Enable SGLANG_HIP_DSV41_INDEX_K_PAGE8=1 only on the non-monokernel points (TP2 c32, DP2 c32 and c64); the monokernel refuses that layout."
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4367

- config-keys:
- musespark-fp4-b200-sglang-agentic-dflash
scenario-type:
- agentic-coding
description:
- "Update the Muse Spark Open B200 curve to lmsysorg/sglang:nightly-dev-20261021-c1c287a4 and set SGLANG_FLASHINFER_NVFP4_PER_TOKEN_ACTIVATION=0 at every concurrency."
- "Use 32768-token prefill chunks and the Python radix tree core at TP8 concurrency 56 (prefill-decode-interval 64) and new concurrency 64, and raise mem-fraction-static to 0.98 at TP4 concurrency 24."
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4400

- config-keys:
- dsv4-fp4-gb200-llmd-vllm-agentic-mtp-disagg
- dsv4-fp4-gb200-llmd-vllm-agentic-mtp-agg
Expand Down
Loading