Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
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
Original file line number Diff line number Diff line change
Expand Up @@ -156,6 +156,9 @@ override_tp4ep1_c32:
override_dep4_hicache_c64:
frontend:
type: sglang-router
# SGLang images after b947a615 (#50671) no longer ship sgl-model-gateway, so run the
# frozen sglang-router 0.3.2 from its official image (v0.3.2 index, pinned by digest).
container_image: registry-1.docker.io#lmsysorg/sgl-model-gateway:sha256:83711812e5e7dc706a1b71ba67734be4abb332ab44327544c4af9ab0a20f5d00
args:
policy: cache_aware
request-id-headers: x-correlation-id
Expand Down Expand Up @@ -207,6 +210,9 @@ override_dep4_hicache_c64:
override_dep4_hicache_c96:
frontend:
type: sglang-router
# SGLang images after b947a615 (#50671) no longer ship sgl-model-gateway, so run the
# frozen sglang-router 0.3.2 from its official image (v0.3.2 index, pinned by digest).
container_image: registry-1.docker.io#lmsysorg/sgl-model-gateway:sha256:83711812e5e7dc706a1b71ba67734be4abb332ab44327544c4af9ab0a20f5d00
args:
policy: cache_aware
request-id-headers: x-correlation-id
Expand Down Expand Up @@ -258,6 +264,9 @@ override_dep4_hicache_c96:
override_dep4_hicache_c128:
frontend:
type: sglang-router
# SGLang images after b947a615 (#50671) no longer ship sgl-model-gateway, so run the
# frozen sglang-router 0.3.2 from its official image (v0.3.2 index, pinned by digest).
container_image: registry-1.docker.io#lmsysorg/sgl-model-gateway:sha256:83711812e5e7dc706a1b71ba67734be4abb332ab44327544c4af9ab0a20f5d00
args:
policy: cache_aware
request-id-headers: x-correlation-id
Expand Down Expand Up @@ -309,6 +318,9 @@ override_dep4_hicache_c128:
override_dep4_hicache_c160:
frontend:
type: sglang-router
# SGLang images after b947a615 (#50671) no longer ship sgl-model-gateway, so run the
# frozen sglang-router 0.3.2 from its official image (v0.3.2 index, pinned by digest).
container_image: registry-1.docker.io#lmsysorg/sgl-model-gateway:sha256:83711812e5e7dc706a1b71ba67734be4abb332ab44327544c4af9ab0a20f5d00
args:
policy: cache_aware
request-id-headers: x-correlation-id
Expand Down Expand Up @@ -360,6 +372,9 @@ override_dep4_hicache_c160:
override_dep4_hicache_c256:
frontend:
type: sglang-router
# SGLang images after b947a615 (#50671) no longer ship sgl-model-gateway, so run the
# frozen sglang-router 0.3.2 from its official image (v0.3.2 index, pinned by digest).
container_image: registry-1.docker.io#lmsysorg/sgl-model-gateway:sha256:83711812e5e7dc706a1b71ba67734be4abb332ab44327544c4af9ab0a20f5d00
args:
policy: cache_aware
request-id-headers: x-correlation-id
Expand Down
2 changes: 1 addition & 1 deletion inferencex-e2e/configs/nvidia-master.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -5831,7 +5831,7 @@ dsv41flash-fp4-gb200-vllm-agentic-dspark:
# --speculative-dspark-block-size 5), thinking on, golden AL 3.51 for
# throughput, real verification for evals.
dsv41flash-fp4-gb200-sglang-agentic-dspark:
image: lmsysorg/sglang:nightly-dev-cu13-20261027-5884e4fe@sha256:61b9083bdcb7cf3741366f487cb996c05ea33ad2ddddc820d10cbc1c0cfb0a86
image: lmsysorg/sglang:nightly-dev-cu13-20261118-ecd3d75a@sha256:2a7e6d43ada9311fcb7d7dd5781bda10aee1a24c40186a956c223f2036212514
model: deepseek-ai/DeepSeek-V4.1-Flash
model-prefix: dsv41flash
runner: cluster:gb200-nv
Expand Down
7 changes: 7 additions & 0 deletions inferencex-e2e/perf-changelog.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -13275,3 +13275,10 @@
- "Move every key of the DeepSeek-V4.1-Flash MI355X vLLM AgentX curve to vllm/vllm-openai-rocm:nightly-9c1d45c21836370104d4eb51d3b46eda20719c47 (2026-11-19, digest sha256:694108707f333787b1041caa0e948dd18b86f2450365a94e38819f026275ee32) and rerun all 13 points, so the whole curve is on one image: TP4 c1-c2, TP2 c1-c64 and TP8 c1 (previously nightly-4fab5a07), DP2 c128 without KV offloading and DP2 c256/c384 with UMBP KV offloading (previously nightly-60bd6dd)."
- "Recipes, topology, concurrencies, speculative decoding and KV offloading are unchanged. Relative to nightly-60bd6dd, the image keeps the same ROCm 10.0, PyTorch 2.13.0, Triton, AITER v0.1.29 and MORI v1.2.3.post1 layers; only vLLM moves forward on main."
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/5151

- config-keys:
- dsv41flash-fp4-gb200-sglang-agentic-dspark
scenario-type: [agentic-coding]
description:
- "Update SGLang image to nightly-dev-cu13-20261118-ecd3d75a; run sglang-router 0.3.2 from its gateway image."
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/5170