[WIP][AMD][AgentX] Add Qwen3.8 FP8 MI355X two-node vLLM - #2724

Open

haic0 wants to merge 8 commits into

SemiAnalysisAI:mainfrom

haic0:rocm/qwen3.8-2.4T-FP8-agentic

Commits on Aug 14, 2026

feat(qwen3.8): add MI355X two-node Ray sweeps

haic0
and
cursoragent
committed
76d9812
fix(qwen3.8): launch AgentX client through bash

authored and
haichzha@amd.com
committed
869b6b5

Commits on Aug 25, 2026

Merge latest main into Qwen3.8 AgentX branch

haichzha@amd.com
and
cursoragent
committed
ffedab1
docs(qwen3.8): link support entry to PR 2724

haichzha@amd.com
and
cursoragent
committed
dda24bc
refactor(qwen3.8): keep only the AgentX route

haichzha@amd.com
and
cursoragent
committed
f079012

Commits on Aug 26, 2026

fix(qwen3.8): harden CI path for missing model staging and eval matrix

haic0
and
cursoragent
committed
b0f0729
Merge branch 'main' into rocm/qwen3.8-2.4T-FP8-agentic

haic0
and
cursoragent
committed
4b94621
fix(qwen3.8): stop inheriting /it-share/data as MODEL_PATH

haic0
and
cursoragent
committed
bcfcbe7