[ROCm][CI] Fix entrypoints tests and Python-only installation test on ROCm (#28979)

Signed-off-by: Andreas Karatzas <akaratza@amd.com>

tests/entrypoints/pooling/embed/test_online_long_text.py

@@ -18,11 +18,6 @@ from tests.utils import RemoteOpenAIServer
 from vllm.entrypoints.pooling.embed.protocol import EmbeddingResponse
 from vllm.platforms import current_platform
 
-if current_platform.is_rocm():
-    pytest.skip(
-        "Encoder self-attention is not implemented on ROCm.", allow_module_level=True
-    )
-
 
 def _generate_random_text(word_count: int) -> str:
     """Generate random text with approximately the specified word count."""
@@ -228,6 +223,10 @@ def server_with_chunked_processing():
         "0.8",
     ]
 
+    # ROCm: Use Flex Attention to support encoder-only self-attention.
+    if current_platform.is_rocm():
+        args.extend(["--attention-backend", "FLEX_ATTENTION"])
+
     with RemoteOpenAIServer(MODEL_NAME, args) as remote_server:
         yield remote_server