[Misc] Split up pooling tasks (#10820)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
2024-12-11 17:28:00 +08:00
parent 40766ca1b8
commit 8f10d5e393
27 changed files with 527 additions and 168 deletions
--- a/tests/models/embedding/vision_language/test_dse_qwen2_vl.py
+++ b/tests/models/embedding/vision_language/test_dse_qwen2_vl.py
@@ -93,7 +93,7 @@ def _run_test(
    # if we run HF first, the cuda initialization will be done and it
    # will hurt multiprocessing backend with fork method (the default method).
    with vllm_runner(model,
-                     task="embedding",
+                     task="embed",
                     dtype=dtype,
                     enforce_eager=True,
                     max_model_len=8192) as vllm_model: