[AMD]: Enable NIXL PD disaggregation for ROCm(1/n) (#28348)
Co-authored-by: bingxche <bingxche@amd.com>
This commit is contained in:
@@ -1134,6 +1134,23 @@ jobs:
|
||||
- name: Install dependencies
|
||||
run: bash scripts/ci/amd/amd_ci_install_dependency.sh
|
||||
|
||||
- name: Verify NIXL in Container
|
||||
run: |
|
||||
docker exec ci_sglang bash -lc '
|
||||
set -eux
|
||||
export LD_LIBRARY_PATH=/opt/ucx/lib:/opt/rocm/lib:${LD_LIBRARY_PATH:-}
|
||||
python -c "import nixl; print(\"nixl OK:\", nixl.__file__)"
|
||||
test -d /opt/ucx/lib
|
||||
UCX_INFO=$(command -v ucx_info || true)
|
||||
if [ -z "$UCX_INFO" ]; then
|
||||
UCX_INFO=/opt/ucx/bin/ucx_info
|
||||
fi
|
||||
test -x "$UCX_INFO"
|
||||
"$UCX_INFO" -v
|
||||
"$UCX_INFO" -d | grep -iq rocm
|
||||
python -c "from nixl._api import nixl_agent, nixl_agent_config; nixl_agent(\"ci-smoke\", nixl_agent_config(backends=[\"UCX\"])); print(\"nixl UCX backend init OK\")"
|
||||
'
|
||||
|
||||
- name: Verify RDMA in Container
|
||||
run: |
|
||||
docker exec -u root ci_sglang bash -c '
|
||||
|
||||
Reference in New Issue
Block a user