[AMD][AgentX] Port MI355X DeepSeek V4 MTP disagg to srt-slurm - #2555

Open

cquil11 wants to merge 21 commits into

agent/srt-slurm-amd-integrationfrom

agent/port-dsv4-mi355x-agentx-srt-slurm

Commits on Aug 11, 2026

feat(amd): port MI355X DSV4 AgentX to srt-slurm
cquil11
committed
c1e0cce
docs: register MI355X AgentX srt-slurm port
cquil11
committed
2c09431
refactor(amd): remove superseded DSV4 launcher path
cquil11
committed
a3c0ad1
fix(amd): respect Slurm CPU affinity for SGLang workers
cquil11
committed
2f6d590
fix(amd): allocate full MI355X CPU topology for DSV4
cquil11
committed
4e77ff0
chore(amd): pin resilient srt-slurm diagnostics
cquil11
committed
7fd1a27
fix(amd): restore DSV4 ROCm kernel routing
cquil11
committed
8851bb6
fix(amd): pin DSV4 runtime identity
cquil11
committed
7d53ce2
test(amd): drop redundant launcher contract suite
cquil11
committed
ab17f87
fix(ci): restore e2e eval matrix output
cquil11
committed
3ad7e2d
fix(router): gate SGLang startup on backend readiness
cquil11
committed
d0376da
Emit per-role average power on multinode disaggregated runs / 多节点分离式运行输出每角色平均功率 (#2553)
edwingao28
authored
4ef1368
Merge remote-tracking branch 'origin/main' into agent/port-dsv4-mi355x-agentx-srt-slurm
cquil11
committed
b1289ce
Retune DSV4 B300 AgentX MTP sweep / 调优 DSV4 B300 AgentX MTP 扫描 (#2536)

ivanium
and
cquil11
authored
c62f6e3
Merge remote-tracking branch 'origin/main' into agent/port-dsv4-mi355x-agentx-srt-slurm
cquil11
committed
73e8ada
docs(amd): record MI355X MoRI fabric contract
cquil11
committed
66c9caa
docs(amd): include MoRI fabric contract in changelog
cquil11
committed
2ec7a69
docs(amd): record exact MI355X srt runtime pin
cquil11
committed
ee00360
[AMD] [AGENTX] GLM-5.2 FP4 MI355X SGLang Agentic MTP (#2488)

authored
a5ec626
Pin MI355X runtime frontend fix
cquil11
committed
5cb444d
Merge remote-tracking branch 'origin/main' into agent/port-dsv4-mi355x-agentx-srt-slurm
cquil11
committed
a502712

Navigation Menu

Uh oh!

[AMD][AgentX] Port MI355X DeepSeek V4 MTP disagg to srt-slurm - #2555

[AMD][AgentX] Port MI355X DeepSeek V4 MTP disagg to srt-slurm#2555
cquil11 wants to merge 21 commits into
agent/srt-slurm-amd-integrationfrom
agent/port-dsv4-mi355x-agentx-srt-slurm

Commits on Aug 11, 2026

feat(amd): port MI355X DSV4 AgentX to srt-slurm

docs: register MI355X AgentX srt-slurm port

refactor(amd): remove superseded DSV4 launcher path

fix(amd): respect Slurm CPU affinity for SGLang workers

fix(amd): allocate full MI355X CPU topology for DSV4

chore(amd): pin resilient srt-slurm diagnostics

fix(amd): restore DSV4 ROCm kernel routing

fix(amd): pin DSV4 runtime identity

test(amd): drop redundant launcher contract suite

fix(ci): restore e2e eval matrix output

fix(router): gate SGLang startup on backend readiness

Emit per-role average power on multinode disaggregated runs / 多节点分离式运行输出每角色平均功率 (#2553)

Merge remote-tracking branch 'origin/main' into agent/port-dsv4-mi355x-agentx-srt-slurm

Retune DSV4 B300 AgentX MTP sweep / 调优 DSV4 B300 AgentX MTP 扫描 (#2536)

Merge remote-tracking branch 'origin/main' into agent/port-dsv4-mi355x-agentx-srt-slurm

docs(amd): record MI355X MoRI fabric contract

docs(amd): include MoRI fabric contract in changelog

docs(amd): record exact MI355X srt runtime pin

[AMD] [AGENTX] GLM-5.2 FP4 MI355X SGLang Agentic MTP (#2488)

Pin MI355X runtime frontend fix

Merge remote-tracking branch 'origin/main' into agent/port-dsv4-mi355x-agentx-srt-slurm