[PIR]Open unittest for PIR

python/paddle/nn/functional/loss.py

@@ -984,7 +984,7 @@ def hsigmoid_loss(
     if num_classes < 2:
         raise ValueError(f'Expected num_classes >= 2 (got {num_classes})')
 
-    if in_dynamic_or_pir_mode():
+    if in_dynamic_mode():
         out, _, _ = _C_ops.hsigmoid_loss(
             input,
             label,
@@ -997,27 +997,41 @@ def hsigmoid_loss(
             is_sparse,
         )
         return out
-    else:
+
+    check_variable_and_dtype(
+        input, 'input', ['float32', 'float64'], 'hsigmoid_loss'
+    )
+    check_variable_and_dtype(label, 'label', ['int64'], 'hsigmoid_loss')
+    check_variable_and_dtype(
+        weight, 'weight', ['float32', 'float64'], 'hsigmoid_loss'
+    )
+    if bias is not None:
         check_variable_and_dtype(
-            input, 'input', ['float32', 'float64'], 'hsigmoid_loss'
+            bias, 'bias', ['float32', 'float64'], 'hsigmoid_loss'
         )
-        check_variable_and_dtype(label, 'label', ['int64'], 'hsigmoid_loss')
+    if path_table is not None:
         check_variable_and_dtype(
-            weight, 'weight', ['float32', 'float64'], 'hsigmoid_loss'
+            path_table, 'path_table', ['int64'], 'hsigmoid_loss'
+        )
+    if path_code is not None:
+        check_variable_and_dtype(
+            path_code, 'path_code', ['int64'], 'hsigmoid_loss'
         )
-        if bias is not None:
-            check_variable_and_dtype(
-                bias, 'bias', ['float32', 'float64'], 'hsigmoid_loss'
-            )
-        if path_table is not None:
-            check_variable_and_dtype(
-                path_table, 'path_table', ['int64'], 'hsigmoid_loss'
-            )
-        if path_code is not None:
-            check_variable_and_dtype(
-                path_code, 'path_code', ['int64'], 'hsigmoid_loss'
-            )
 
+    if in_pir_mode():
+        out, _, _ = _C_ops.hsigmoid_loss(
+            input,
+            label,
+            weight,
+            bias,
+            path_table,
+            path_code,
+            num_classes,
+            is_sparse,
+            is_sparse,
+        )
+        return out
+    else:
         attrs = {
             "num_classes": num_classes,
             "is_sparse": is_sparse,

python/paddle/tensor/linalg.py

@@ -27,7 +27,12 @@
     convert_dtype,
 )
 from ..common_ops_import import Variable
-from ..framework import LayerHelper, in_dynamic_mode, in_dynamic_or_pir_mode
+from ..framework import (
+    LayerHelper,
+    in_dynamic_mode,
+    in_dynamic_or_pir_mode,
+    in_pir_mode,
+)
 from .creation import full
 from .manipulation import cast
 from .math import _get_reduce_axis
@@ -2830,26 +2835,30 @@ def eigh(x, UPLO='L', name=None):
              [ 0.3826833963394165j    , -0.9238795042037964j    ]])
 
     """
-    if in_dynamic_or_pir_mode():
+    if in_dynamic_mode():
         return _C_ops.eigh(x, UPLO)
-    else:
 
-        def __check_input(x, UPLO):
-            x_shape = list(x.shape)
-            if len(x.shape) < 2:
-                raise ValueError(
-                    "Input(input) only support >=2 tensor, but received "
-                    "length of Input(input) is %s." % len(x.shape)
-                )
-            if x_shape[-1] != x_shape[-2]:
-                raise ValueError(
-                    f"The input matrix must be batches of square matrices. But received x's dimention: {x_shape}"
-                )
-            if UPLO != 'L' and UPLO != 'U':
-                raise ValueError(
-                    f"UPLO must be L or U. But received UPLO is: {UPLO}"
-                )
+    def __check_input(x, UPLO):
+        x_shape = list(x.shape)
+        if len(x.shape) < 2:
+            raise ValueError(
+                "Input(input) only support >=2 tensor, but received "
+                "length of Input(input) is %s." % len(x.shape)
+            )
+        if x_shape[-1] != x_shape[-2]:
+            raise ValueError(
+                f"The input matrix must be batches of square matrices. But received x's dimention: {x_shape}"
+            )
+        if UPLO != 'L' and UPLO != 'U':
+            raise ValueError(
+                f"UPLO must be L or U. But received UPLO is: {UPLO}"
+            )
 
+    if in_pir_mode():
+        __check_input(x, UPLO)
+        return _C_ops.eigh(x, UPLO)
+
+    else:
         __check_input(x, UPLO)
 
         helper = LayerHelper('eigh', **locals())
@@ -3342,27 +3351,32 @@ def eigvalsh(x, UPLO='L', name=None):
             Tensor(shape=[2], dtype=float32, place=Place(cpu), stop_gradient=True,
             [0.17157286, 5.82842731])
     """
-    if in_dynamic_or_pir_mode():
+    if in_dynamic_mode():
         values, _ = _C_ops.eigvalsh(x, UPLO, x.stop_gradient)
         return values
-    else:
 
-        def __check_input(x, UPLO):
-            x_shape = list(x.shape)
-            if len(x.shape) < 2:
-                raise ValueError(
-                    "Input(input) only support >=2 tensor, but received "
-                    "length of Input(input) is %s." % len(x.shape)
-                )
-            if x_shape[-1] != x_shape[-2]:
-                raise ValueError(
-                    f"The input matrix must be batches of square matrices. But received x's dimention: {x_shape}"
-                )
-            if UPLO != 'L' and UPLO != 'U':
-                raise ValueError(
-                    f"UPLO must be L or U. But received UPLO is: {UPLO}"
-                )
+    def __check_input(x, UPLO):
+        x_shape = list(x.shape)
+        if len(x.shape) < 2:
+            raise ValueError(
+                "Input(input) only support >=2 tensor, but received "
+                "length of Input(input) is %s." % len(x.shape)
+            )
+        if x_shape[-1] != x_shape[-2]:
+            raise ValueError(
+                f"The input matrix must be batches of square matrices. But received x's dimention: {x_shape}"
+            )
+        if UPLO != 'L' and UPLO != 'U':
+            raise ValueError(
+                f"UPLO must be L or U. But received UPLO is: {UPLO}"
+            )
 
+    if in_pir_mode():
+        __check_input(x, UPLO)
+        values, _ = _C_ops.eigvalsh(x, UPLO, x.stop_gradient)
+        return values
+
+    else:
         __check_input(x, UPLO)
 
         helper = LayerHelper('eigvalsh', **locals())

python/paddle/tensor/logic.py

@@ -27,7 +27,12 @@
 from paddle.tensor.math import broadcast_shape
 from paddle.utils.inplace_utils import inplace_apis_in_dygraph_only
 
-from ..framework import LayerHelper, in_dynamic_mode, in_dynamic_or_pir_mode
+from ..framework import (
+    LayerHelper,
+    in_dynamic_mode,
+    in_dynamic_or_pir_mode,
+    in_pir_mode,
+)
 
 __all__ = []
 
@@ -363,14 +368,16 @@ def is_empty(x, name=None):
             False)
 
     """
-    if in_dynamic_or_pir_mode():
+    if in_dynamic_mode():
         return _C_ops.is_empty(x)
-    else:
-        check_variable_and_dtype(
-            x, 'x', ['float32', 'float64', 'int32', 'int64'], 'is_empty'
-        )
-        check_type(name, "name", (str, type(None)), "is_empty")
 
+    check_variable_and_dtype(
+        x, 'x', ['float32', 'float64', 'int32', 'int64'], 'is_empty'
+    )
+    check_type(name, "name", (str, type(None)), "is_empty")
+    if in_pir_mode():
+        return _C_ops.is_empty(x)
+    else:
         helper = LayerHelper("is_empty", **locals())
         cond = helper.create_variable_for_type_inference(dtype='bool')
         cond.stop_gradient = True

python/paddle/tensor/manipulation.py

@@ -818,13 +818,9 @@ def crop(x, shape=None, offsets=None, name=None):
     if shape is None:
         shape = x.shape
 
-    if in_dynamic_or_pir_mode():
+    if in_dynamic_mode():
         return _C_ops.crop(x, shape, offsets)
 
-    out = helper.create_variable_for_type_inference(x.dtype)
-    ipts = {'X': x}
-    attrs = {}
-
     def _attr_shape_check(shape_val):
         if not isinstance(shape_val, int):
             raise TypeError(
@@ -854,6 +850,48 @@ def _attr_offsets_check(offset_val):
                 % str(offset_val)
             )
 
+    if in_pir_mode():
+        if isinstance(offsets, paddle.pir.Value):
+            offsets.stop_gradient = True
+        elif paddle.utils._contain_var(offsets):
+            new_offsets_tensor = []
+            for dim in offsets:
+                if isinstance(dim, paddle.pir.Value):
+                    dim.stop_gradient = True
+                    new_offsets_tensor.append(dim)
+                else:
+                    _attr_offsets_check(dim)
+                    temp_out = fill_constant([1], 'int32', dim, force_cpu=True)
+                    new_offsets_tensor.append(temp_out)
+            offsets = new_offsets_tensor
+        else:
+            for offset in offsets:
+                _attr_offsets_check(offset)
+
+        if isinstance(shape, paddle.pir.Value):
+            shape.stop_gradient = True
+        elif paddle.utils._contain_var(shape):
+            new_shape_tensor = []
+            for dim_size in shape:
+                if isinstance(dim_size, paddle.pir.Value):
+                    dim_size.stop_gradient = True
+                    new_shape_tensor.append(dim_size)
+                else:
+                    _attr_shape_check(dim_size)
+                    temp_out = fill_constant(
+                        [1], 'int32', dim_size, force_cpu=True
+                    )
+                    new_shape_tensor.append(temp_out)
+            shape = new_shape_tensor
+        else:
+            for dim_size in shape:
+                _attr_shape_check(dim_size)
+        return _C_ops.crop(x, shape, offsets)
+
+    out = helper.create_variable_for_type_inference(x.dtype)
+    ipts = {'X': x}
+    attrs = {}
+
     if isinstance(offsets, Variable):
         offsets.stop_gradient = True
         ipts['Offsets'] = offsets

test/legacy_test/test_activation_op.py

@@ -1133,6 +1133,7 @@ def test_backward(self):
 
 
 class TestCoshOpError(unittest.TestCase):
+    @test_with_pir_api
     def test_errors(self):
         with static_guard():
             with program_guard(Program()):
@@ -1331,6 +1332,7 @@ def test_dygraph_api(self):
             for r in [out1, out2]:
                 np.testing.assert_allclose(out_ref, r.numpy(), rtol=1e-05)
 
+    @test_with_pir_api
     def test_errors(self):
         with static_guard():
             with paddle.static.program_guard(paddle.static.Program()):
@@ -1398,6 +1400,7 @@ def test_dygraph_api(self):
             for r in [out1, out2]:
                 np.testing.assert_allclose(out_ref, r.numpy(), rtol=1e-05)
 
+    @test_with_pir_api
     def test_errors(self):
         with static_guard():
             with paddle.static.program_guard(paddle.static.Program()):
@@ -3122,6 +3125,7 @@ def test_base_api(self):
             out = paddle.nn.functional.hardswish(x)
             np.testing.assert_allclose(out_ref, out.numpy(), rtol=1e-05)
 
+    @test_with_pir_api
     def test_errors(self):
         with static_guard():
             with paddle.static.program_guard(paddle.static.Program()):
@@ -4531,6 +4535,7 @@ def test_base_api(self):
         out = paddle.nn.functional.hardsigmoid(x, slope=0.2)
         np.testing.assert_allclose(out_ref, out.numpy(), rtol=1e-05)
 
+    @test_with_pir_api
     def test_errors(self):
         with static_guard():
             with paddle.static.program_guard(paddle.static.Program()):

test/legacy_test/test_crop_tensor_op.py

@@ -18,6 +18,7 @@
 from op_test import OpTest
 
 import paddle
+from paddle.pir_utils import test_with_pir_api
 
 
 def crop(data, offsets, crop_shape):
@@ -225,6 +226,7 @@ def initTestCase(self):
 
 
 class TestCropTensorException(unittest.TestCase):
+    @test_with_pir_api
     def test_exception(self):
         input1 = paddle.static.data(
             name="input1", shape=[2, 3, 6, 6], dtype="float32"

test/legacy_test/test_diagflat.py

@@ -21,6 +21,7 @@
 from paddle.static import Program, program_guard
 
 
+@test_with_pir_api
 class TestDiagFlatError(unittest.TestCase):
     def test_errors(self):
         paddle.enable_static()

test/legacy_test/test_eigh_op.py

@@ -223,6 +223,7 @@ def init_input_shape(self):
 
 
 class TestEighAPIError(unittest.TestCase):
+    @test_with_pir_api
     def test_error(self):
         main_prog = paddle.static.Program()
         startup_prog = paddle.static.Program()

test/legacy_test/test_eigvalsh_op.py

@@ -204,6 +204,7 @@ def init_input_shape(self):
 
 
 class TestEigvalshAPIError(unittest.TestCase):
+    @test_with_pir_api
     def test_error(self):
         main_prog = paddle.static.Program()
         startup_prog = paddle.static.Program()

test/legacy_test/test_fold_op.py

@@ -20,6 +20,7 @@
 import paddle
 from paddle import base
 from paddle.base import core
+from paddle.pir_utils import test_with_pir_api
 
 paddle.enable_static()
 
@@ -190,6 +191,7 @@ def test_info(self):
 
 
 class TestFoldOpError(unittest.TestCase):
+    @test_with_pir_api
     def test_errors(self):
         from paddle.base.framework import Program, program_guard
         from paddle.nn.functional import fold

test/legacy_test/test_histogram_op.py

@@ -104,6 +104,7 @@ def net_func():
         with self.assertRaises(ValueError):
             self.run_network(net_func)
 
+    @test_with_pir_api
     def test_min_max_range_error(self):
         """Test range of min, max is not finite"""
 
@@ -116,6 +117,7 @@ def net_func():
         with self.assertRaises(TypeError):
             self.run_network(net_func)
 
+    @test_with_pir_api
     def test_type_errors(self):
         with paddle.static.program_guard(paddle.static.Program()):
             # The input type must be Variable.