fix ut

Signed-off-by: MengqingCao <cmq0113@163.com>

Author: MengqingCao

requirements-dev.txt

@@ -13,12 +13,4 @@ types-psutil
 pytest-cov
 regex
 sentence_transformers
-# Use daliy release ray for pp, which includes https://github.com/ray-project/ray/commit/eae786c8674f6b1c29c3bda6fdce80d0ab4afcd8
-ray[default] @ https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-3.0.0.dev0-cp39-cp39-manylinux2014_aarch64.whl ; python_version == "3.9" and platform_system == "Linux" and platform_machine == "aarch64"
-ray[default] @ https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-3.0.0.dev0-cp310-cp310-manylinux2014_aarch64.whl ; python_version == "3.10" and platform_system == "Linux" and platform_machine ==  "aarch64"
-ray[default] @ https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-3.0.0.dev0-cp311-cp311-manylinux2014_aarch64.whl ; python_version == "3.11" and platform_system == "Linux" and platform_machine ==  "aarch64"
-ray[default] @ https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-3.0.0.dev0-cp312-cp312-manylinux2014_aarch64.whl ; python_version == "3.12" and platform_system == "Linux" and platform_machine ==  "aarch64"
-ray[default] @ https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-3.0.0.dev0-cp39-cp39-manylinux2014_x86_64.whl ; python_version == "3.9" and platform_system == "Linux" and platform_machine ==  "x86_64"
-ray[default] @ https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-3.0.0.dev0-cp310-cp310-manylinux2014_x86_64.whl ; python_version == "3.10" and platform_system == "Linux" and platform_machine ==  "x86_64"
-ray[default] @ https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-3.0.0.dev0-cp311-cp311-manylinux2014_x86_64.whl ; python_version == "3.11" and platform_system == "Linux" and platform_machine ==  "x86_64"
-ray[default] @ https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-3.0.0.dev0-cp312-cp312-manylinux2014_x86_64.whl ; python_version == "3.12" and platform_system == "Linux" and platform_machine ==  "x86_64"
+ray>=2.47.1

tests/ut/attention/test_attention_v1.py

@@ -400,19 +400,13 @@ def test_forward_head_size_192(self, mock_vanilla_prefill,
         layer = self.layer_no_quant
         mock_vanilla_prefill.return_value = MagicMock()
 
-        def mock_tensor(data, device=None, **kwargs):
-            if device == "npu":
-                return metadata.attn_mask
-            return torch.tensor(data, **kwargs)
-
-        with patch("torch.tensor", side_effect=mock_tensor):
-            output = self.impl_192.forward(layer,
-                                           query,
-                                           key,
-                                           value,
-                                           kv_cache,
-                                           metadata,
-                                           trace_flag=False)
+        output = self.impl_192.forward(layer,
+                                       query,
+                                       key,
+                                       value,
+                                       kv_cache,
+                                       metadata,
+                                       trace_flag=False)
 
         mock_vanilla_prefill.assert_called_once()
         assert output.shape == (10, 8 * 192)