Skip to content

Update MI355X DSV4 image and C1 mono kernel / 更新 MI355X DSV4 镜像及 C1 mono kernel #4602

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

Open
Duyi-Wang wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Update MI355X DSV4 image and C1 mono kernel / 更新 MI355X DSV4 镜像及 C1 mono kernel #4602

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
c6bd0be
perf(amd): update DSV4 image and enable C1 decode mono kernel
Duyi-Wang Oct 28, 2026
f645539
perf(amd): align C1 decode TP8 with the mono experiment
Duyi-Wang Oct 28, 2026
0cb89d9
docs: describe final DSV4 C1 configuration
Duyi-Wang Oct 28, 2026
f0a97e6
fix(amd): pin DSV4 nightly image digest
Duyi-Wang Oct 28, 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
5 changes: 3 additions & 2 deletions ...hmarks/multi_node/srt-slurm-recipes/dsv4/sglang/mi355x-fp4/agentx/disagg-umbp-dspark.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -213,13 +213,13 @@ override_c1:
speculative-dspark-block-size: 6
speculative-num-draft-tokens: 7
decode:
gpus: 4
gpus: 8
env:
SGLANG_MLA_HCA_KV_SPLITS: "8"
MORI_MAX_DISPATCH_TOKENS_DECODE: "448"
SGLANG_DSV4_UNIFIED_KV_FP8: "1"
args:
tp-size: 4
tp-size: 8
mem-fraction-static: 0.86
max-running-requests: 8
cuda-graph-bs-decode:
Expand All @@ -233,6 +233,7 @@ override_c1:
- 8
speculative-dspark-block-size: 6
speculative-num-draft-tokens: 7
enable-amd-monokernel: true

override_c16:
roles:
Expand Down
4 changes: 2 additions & 2 deletions inferencex-e2e/configs/amd-master.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1197,7 +1197,7 @@ dsv41flash-fp4-mi325x-vllm-agentic-dspark:
- { tp: 2, kv-offloading: none, spec-decoding: mtp, conc-list: [1, 2, 4, 8, 16, 32], srt-recipe: agentic.yaml }

dsv4-fp4-mi355x-sglang-disagg-agentic-umbp-dspark:
image: lmsysorg/sglang-rocm:v0.5.22-rocm10-mi35x-20261026
image: lmsysorg/sglang-rocm:v0.5.22-rocm10-mi35x-20261028@sha256:44202f87c180a28a6b7570f56dfb37a2c0e90dbec901b16381b87a9d582b1f27
model: deepseek-ai/DeepSeek-V4-Pro-0813
model-prefix: dsv4
runner: cluster:mi355x-amds
Expand All @@ -1223,7 +1223,7 @@ dsv4-fp4-mi355x-sglang-disagg-agentic-umbp-dspark:
dp-attn: false
decode:
num-worker: 1
tp: 4
tp: 8
ep: 1
dp-attn: false
- spec-decoding: "draft_model"
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 @@ -11489,3 +11489,12 @@
- "Update the MI355X DeepSeek-V4-Pro-0813 SGLang MoRI disaggregated AgentX serving image from lmsysorg/sglang-rocm:v0.5.22-rocm720-mi35x-20261019 to lmsysorg/sglang-rocm:v0.5.22-rocm10-mi35x-20261026, the latest published MI35x ROCm 10 nightly (digest sha256:edabad83728ab04e3eae39f315eb16261eeb64dbdb051112bd22fc88560b60c4)."
- "Preserve all eight concurrency points (1, 16, 32, 48, 192, 256, 384, 512), prefill/decode topology, DSpark settings, UMBP offload, MoRI transfer, and benchmark client image. The workload binder supplies the serving image to every selected recipe variant."
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4571

- config-keys:
- dsv4-fp4-mi355x-sglang-disagg-agentic-umbp-dspark
scenario-type:
- agentic-coding
description:
- "Update the MI355X DeepSeek-V4-Pro-0813 SGLang MoRI disaggregated AgentX serving image to lmsysorg/sglang-rocm:v0.5.22-rocm10-mi35x-20261028@sha256:44202f87c180a28a6b7570f56dfb37a2c0e90dbec901b16381b87a9d582b1f27 and use P-TP4 / D-TP8 with the AMD mono kernel on decode only at concurrency 1."
- "Preserve DSpark, offload, all other concurrency settings, and the benchmark client image."
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4602
Loading