[Bugfix] Replace PoolingParams.normalize with use_activation (#32243)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

tests/entrypoints/pooling/embed/test_offline.py

@@ -53,7 +53,9 @@ def test_token_embed(llm: LLM):
 def test_pooling_params(llm: LLM):
     def get_outputs(normalize):
         outputs = llm.embed(
-            prompts, pooling_params=PoolingParams(normalize=normalize), use_tqdm=False
+            prompts,
+            pooling_params=PoolingParams(use_activation=normalize),
+            use_tqdm=False,
         )
         return torch.tensor([x.outputs.embedding for x in outputs])
 

tests/entrypoints/pooling/embed/test_online_long_text.py

@@ -216,7 +216,7 @@ def server_with_chunked_processing():
         "512",  # Set smaller max_model_len to trigger chunking mechanism
         "--pooler-config",
         (
-            '{"pooling_type": "MEAN", "normalize": true, '
+            '{"pooling_type": "MEAN", "use_activation": true, '
             '"enable_chunked_processing": true, "max_embed_len": 10000}'
         ),
         "--gpu-memory-utilization",

tests/entrypoints/pooling/score/test_online_score.py

@@ -236,17 +236,14 @@ class TestModel:
                     "use_activation": use_activation,
                 },
             )
-            if response.status_code != 200:
-                return response
-
             outputs = response.json()
             return torch.tensor([x["score"] for x in outputs["data"]])
 
-        if model["is_cross_encoder"]:
-            default = get_outputs(use_activation=None)
-            w_activation = get_outputs(use_activation=True)
-            wo_activation = get_outputs(use_activation=False)
+        default = get_outputs(use_activation=None)
+        w_activation = get_outputs(use_activation=True)
+        wo_activation = get_outputs(use_activation=False)
 
+        if model["is_cross_encoder"]:
             assert torch.allclose(default, w_activation, atol=1e-2), (
                 "Default should use activation."
             )
@@ -256,9 +253,3 @@ class TestModel:
             assert torch.allclose(F.sigmoid(wo_activation), w_activation, atol=1e-2), (
                 "w_activation should be close to activation(wo_activation)."
             )
-        else:
-            get_outputs(use_activation=None)
-
-            # The activation parameter only works for the is_cross_encoder model
-            response = get_outputs(use_activation=True)
-            assert response.status_code == 400