[CI] Raise DSv4-Flash disaggregated engine readiness timeout to 1800s (#57334)

Signed-off-by: khluu <khluu000@gmail.com>
Co-authored-by: Kimi <noreply@moonshot.cn>
This commit is contained in:
Kevin H. Luu
2026-09-17 03:34:00 -07:00
committed by GitHub
co-authored by Kimi
parent 683f6c8edf
commit 99ea5f525d
+1 -1
View File
@@ -287,7 +287,7 @@ steps:
PREFILL_BLOCK_SIZE: "256"
DECODE_BLOCK_SIZE: "256"
MODEL_NAMES: "deepseek-ai/DeepSeek-V4-Flash"
VLLM_ENGINE_READY_TIMEOUT_S: "900"
VLLM_ENGINE_READY_TIMEOUT_S: "1800"
VLLM_SERVE_EXTRA_ARGS: "--trust-remote-code,--kv-cache-dtype,fp8"
# The V4 indexer only supports block-outermost layouts; let it pick.
PREFILLER_KV_LAYOUT: ""