[CI] Pass the scale-out endpoint flag in EC E2E (#56819)

Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com>
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
This commit is contained in:
Kevin H. Luu
2026-09-15 12:25:30 +08:00
committed by GitHub
co-authored by OpenAI Codex Tyler Michael Smith
parent a491b6039d
commit caf56ce9da
2 changed files with 5 additions and 3 deletions
@@ -181,8 +181,9 @@ run_scale_out_ec() {
# Encode-only EC producer instance. It runs the vision encoder and
# publishes embeddings through the ECExampleConnector shared storage.
echo "Starting encode instance on GPU $GPU_E, port $ENCODE_PORT"
env CUDA_VISIBLE_DEVICES="$GPU_E" VLLM_ENABLE_SCALE_OUT_ENDPOINTS=1 \
env CUDA_VISIBLE_DEVICES="$GPU_E" \
vllm serve "$MODEL" \
--enable-scale-out \
--port "$ENCODE_PORT" \
--max-model-len "$MAX_MODEL_LEN" \
--enforce-eager \
@@ -204,8 +205,9 @@ run_scale_out_ec() {
# features plus ec_transfer_params and loads the embeddings that the
# encode instance published.
echo "Starting prefill instance on GPU $GPU_PD, port $PREFILL_PORT"
env CUDA_VISIBLE_DEVICES="$GPU_PD" VLLM_ENABLE_SCALE_OUT_ENDPOINTS=1 \
env CUDA_VISIBLE_DEVICES="$GPU_PD" \
vllm serve "$MODEL" \
--enable-scale-out \
--port "$PREFILL_PORT" \
--max-model-len "$MAX_MODEL_LEN" \
--enforce-eager \