[Cherry-pick 2.5][Zero-Dim] paddle.to_tensor support 0D (#53599)

* fix doc erros, test=allcase * conflict * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * test=allcase * fix doc erros, test=allcase * fix the to_tensor error
PaddlePaddle · May 9, 2023 · 2aefc45 · 2aefc45
1 parent 3ffe8f3
commit 2aefc45
Show file tree

Hide file tree

Showing 29 changed files with 214 additions and 174 deletions.
diff --git a/paddle/fluid/pybind/imperative.cc b/paddle/fluid/pybind/imperative.cc
@@ -1351,7 +1351,7 @@ void BindImperative(py::module *m_ptr) {
 
                 import paddle
 
-                x = paddle.to_tensor(1.0, stop_gradient=False)
+                x = paddle.to_tensor([1.0], stop_gradient=False)
                 detach_x = x.detach()
                 detach_x[:] = 10.0
                 print(x)  # Tensor(shape=[1], dtype=float32, place=CPUPlace, stop_gradient=False,

diff --git a/python/paddle/audio/functional/window.py b/python/paddle/audio/functional/window.py
@@ -13,6 +13,8 @@
 import math
 from typing import List, Tuple, Union
 
+import numpy as np
+
 import paddle
 from paddle import Tensor
 
@@ -38,7 +40,12 @@ def get(self, name):
 
 @window_function_register.register()
 def _cat(x: List[Tensor], data_type: str) -> Tensor:
-    l = [paddle.to_tensor(_, data_type) for _ in x]
+    l = []
+    for t in x:
+        if np.isscalar(t) and not isinstance(t, str):
+            l.append(paddle.to_tensor([t], data_type))
+        else:
+            l.append(paddle.to_tensor(t, data_type))
     return paddle.concat(l)
 
 

diff --git a/python/paddle/distributed/fleet/meta_parallel/pp_utils/p2p_communication.py b/python/paddle/distributed/fleet/meta_parallel/pp_utils/p2p_communication.py
@@ -117,7 +117,7 @@ def recv_meta(self, group):
 
     def _send_dims_shape_dtype(self, tensor, group):
         # send len(shape)
-        dims = paddle.to_tensor(len(tensor.shape))
+        dims = paddle.to_tensor([len(tensor.shape)])
         dst_rank = _hcg._get_p2p_next_rank()
 
         paddle.distributed.send(dims, dst=dst_rank, group=group)
@@ -127,11 +127,11 @@ def _send_dims_shape_dtype(self, tensor, group):
         paddle.distributed.send(shape, dst=dst_rank, group=group)
 
         # send dtype
-        dtype = paddle.to_tensor(paddle_2_number(tensor.dtype))
+        dtype = paddle.to_tensor([paddle_2_number(tensor.dtype)])
         paddle.distributed.send(dtype, dst=dst_rank, group=group)
 
         # send trainable
-        stop_grad = paddle.to_tensor(int(tensor.stop_gradient))
+        stop_grad = paddle.to_tensor([int(tensor.stop_gradient)])
         paddle.distributed.send(stop_grad, dst=dst_rank, group=group)
 
     def send_meta(self, tensor, group):
@@ -148,7 +148,7 @@ def send_meta(self, tensor, group):
             # send tensor type
             paddle.distributed.send(tensor_type, dst=dst_rank, group=group)
 
-            nums = paddle.to_tensor(len(tensor))
+            nums = paddle.to_tensor([len(tensor)])
             paddle.distributed.send(nums, dst=dst_rank, group=group)
 
             for d in tensor:

diff --git a/python/paddle/fluid/tests/unittests/auto_parallel/random_control_unittest.py b/python/paddle/fluid/tests/unittests/auto_parallel/random_control_unittest.py
@@ -82,7 +82,7 @@ def compare_mask_between_ranks(
     ):
 
         for np_mask in [mask_np_list[i] for i in comapre_idx]:
-            mask_tensor_local = paddle.to_tensor(np_mask.astype("float32"))
+            mask_tensor_local = paddle.to_tensor([np_mask.astype("float32")])
             if rank == 0:
                 mask_tensor_remote = paddle.ones_like(mask_tensor_local)
                 dy_broadcast_helper(mask_tensor_remote)

diff --git a/python/paddle/fluid/tests/unittests/auto_parallel_gpt_model.py b/python/paddle/fluid/tests/unittests/auto_parallel_gpt_model.py
@@ -233,7 +233,7 @@ def core_attn(self, q, k, v, attn_mask):
         product = paddle.matmul(x=q, y=k, transpose_y=True)
         product = paddle.multiply(
             product,
-            paddle.to_tensor(self.head_dim**-0.5, dtype=product.dtype),
+            paddle.to_tensor([self.head_dim**-0.5], dtype=product.dtype),
         )
         if attn_mask is not None:
             product = product + attn_mask

diff --git a/python/paddle/fluid/tests/unittests/distribution/test_distribution_bernoulli.py b/python/paddle/fluid/tests/unittests/distribution/test_distribution_bernoulli.py
@@ -184,10 +184,10 @@ def init_dynamic_data(self, probs, default_dtype, dtype):
         ('probs_00', 0.0, 'float64', 'float32'),
         ('probs_03', 0.3, 'float64', 'float32'),
         ('probs_10', 1.0, 'float64', 'float32'),
-        ('probs_tensor_03_32', paddle.to_tensor(0.3), 'float32', 'float32'),
+        ('probs_tensor_03_32', paddle.to_tensor([0.3]), 'float32', 'float32'),
         (
             'probs_tensor_03_64',
-            paddle.to_tensor(0.3, dtype='float64'),
+            paddle.to_tensor([0.3], dtype='float64'),
             'float64',
             'float64',
         ),
@@ -257,11 +257,11 @@ def test_variance(self):
             ),
             (
                 paddle.to_tensor(
-                    0.0,
+                    [0.0],
                 ),
             ),
-            (paddle.to_tensor(1.0),),
-            (paddle.to_tensor(0.0, dtype='float64'),),
+            (paddle.to_tensor([1.0]),),
+            (paddle.to_tensor([0.0], dtype='float64'),),
         ]
     )
     def test_log_prob(self, value):
@@ -291,9 +291,9 @@ def test_log_prob(self, value):
                     ]
                 ),
             ),
-            (paddle.to_tensor(0.0),),
-            (paddle.to_tensor(1.0),),
-            (paddle.to_tensor(0.0, dtype='float64'),),
+            (paddle.to_tensor([0.0]),),
+            (paddle.to_tensor([1.0]),),
+            (paddle.to_tensor([0.0], dtype='float64'),),
         ]
     )
     def test_prob(self, value):
@@ -323,11 +323,11 @@ def test_prob(self, value):
                     ]
                 ),
             ),
-            (paddle.to_tensor(0.0),),
-            (paddle.to_tensor(0.3),),
-            (paddle.to_tensor(0.7),),
-            (paddle.to_tensor(1.0),),
-            (paddle.to_tensor(0.0, dtype='float64'),),
+            (paddle.to_tensor([0.0]),),
+            (paddle.to_tensor([0.3]),),
+            (paddle.to_tensor([0.7]),),
+            (paddle.to_tensor([1.0]),),
+            (paddle.to_tensor([0.0], dtype='float64'),),
         ]
     )
     def test_cdf(self, value):
@@ -359,7 +359,7 @@ def test_entropy(self):
 
     def test_kl_divergence(self):
         with paddle.fluid.dygraph.guard(self.place):
-            other_probs = paddle.to_tensor(0.9, dtype=self.dtype)
+            other_probs = paddle.to_tensor([0.9], dtype=self.dtype)
 
             rv_paddle_other = Bernoulli(other_probs)
             rv_np_other = BernoulliNumpy(other_probs)
@@ -422,7 +422,7 @@ def test_kl_divergence(self):
         # 1-D probs
         (
             'probs_1d_1d_32',
-            paddle.to_tensor(0.3),
+            paddle.to_tensor([0.3]),
             'float32',
             'float32',
             [
@@ -432,7 +432,7 @@ def test_kl_divergence(self):
         ),
         (
             'probs_1d_1d_64',
-            paddle.to_tensor(0.3, dtype='float64'),
+            paddle.to_tensor([0.3], dtype='float64'),
             'float64',
             'float64',
             paddle.to_tensor(
@@ -444,15 +444,15 @@ def test_kl_divergence(self):
         ),
         (
             'probs_1d_2d',
-            paddle.to_tensor(0.3),
+            paddle.to_tensor([0.3]),
             'float32',
             'float32',
             [100, 2],
             [100, 2, 1],
         ),
         (
             'probs_1d_3d',
-            paddle.to_tensor(0.3),
+            paddle.to_tensor([0.3]),
             'float32',
             'float32',
             [100, 2, 3],

diff --git a/python/paddle/fluid/tests/unittests/distribution/test_distribution_transform.py b/python/paddle/fluid/tests/unittests/distribution/test_distribution_transform.py
@@ -523,7 +523,7 @@ def test_codomain(self, input, expected):
                 transform.ChainTransform(
                     (
                         transform.AffineTransform(
-                            paddle.to_tensor(0.0), paddle.to_tensor(1.0)
+                            paddle.to_tensor([0.0]), paddle.to_tensor([1.0])
                         ),
                         transform.ExpTransform(),
                     )
@@ -560,7 +560,7 @@ def test_forward(self, chain, input, expected):
                 transform.ChainTransform(
                     (
                         transform.AffineTransform(
-                            paddle.to_tensor(0.0), paddle.to_tensor(-1.0)
+                            paddle.to_tensor([0.0]), paddle.to_tensor([-1.0])
                         ),
                         transform.ExpTransform(),
                     )
@@ -595,9 +595,9 @@ def test_inverse(self, chain, input, expected):
                 transform.ChainTransform(
                     (
                         transform.AffineTransform(
-                            paddle.to_tensor(0.0), paddle.to_tensor(-1.0)
+                            paddle.to_tensor([0.0]), paddle.to_tensor([-1.0])
                         ),
-                        transform.PowerTransform(paddle.to_tensor(2.0)),
+                        transform.PowerTransform(paddle.to_tensor([2.0])),
                     )
                 ),
                 np.array([1.0, 2.0, 3.0]),
@@ -619,7 +619,7 @@ def test_forward_log_det_jacobian(self, chain, input, expected):
                 transform.ChainTransform(
                     (
                         transform.AffineTransform(
-                            paddle.to_tensor(0.0), paddle.to_tensor(-1.0)
+                            paddle.to_tensor([0.0]), paddle.to_tensor([-1.0])
                         ),
                         transform.ExpTransform(),
                     )
@@ -638,7 +638,7 @@ def test_forward_shape(self, chain, shape, expected_shape):
                 transform.ChainTransform(
                     (
                         transform.AffineTransform(
-                            paddle.to_tensor(0.0), paddle.to_tensor(-1.0)
+                            paddle.to_tensor([0.0]), paddle.to_tensor([-1.0])
                         ),
                         transform.ExpTransform(),
                     )
@@ -743,7 +743,7 @@ def test_inverse_shape(self, shape, expected_shape):
 @param.place(config.DEVICES)
 class TestPowerTransform(unittest.TestCase):
     def setUp(self):
-        self._t = transform.PowerTransform(paddle.to_tensor(2.0))
+        self._t = transform.PowerTransform(paddle.to_tensor([2.0]))
 
     def test_init(self):
         with self.assertRaises(TypeError):

diff --git a/...n/paddle/fluid/tests/unittests/distribution/test_distribution_transformed_distribution.py b/...n/paddle/fluid/tests/unittests/distribution/test_distribution_transformed_distribution.py
@@ -41,7 +41,7 @@ def _np_sum_rightmost(self, value, n):
         return np.sum(value, tuple(range(-n, 0))) if n > 0 else value
 
     def test_log_prob(self):
-        value = paddle.to_tensor(0.5)
+        value = paddle.to_tensor([0.5])
         np.testing.assert_allclose(
             self.simple_log_prob(value, self.base, self.transforms),
             self._t.log_prob(value),

diff --git a/python/paddle/fluid/tests/unittests/distribution/test_kl.py b/python/paddle/fluid/tests/unittests/distribution/test_kl.py
@@ -139,8 +139,8 @@ def test_dispatch_with_unregister(self):
         ),
         (
             'test-same-dist',
-            mock.Exponential(paddle.to_tensor(1.0)),
-            mock.Exponential(paddle.to_tensor(1.0)),
+            mock.Exponential(paddle.to_tensor([1.0])),
+            mock.Exponential(paddle.to_tensor([1.0])),
         ),
     ],
 )

diff --git a/python/paddle/fluid/tests/unittests/parallel_dygraph_none_var.py b/python/paddle/fluid/tests/unittests/parallel_dygraph_none_var.py
@@ -40,7 +40,7 @@ def __init__(self):
         self.step = 0
 
     def forward(self, x):
-        return paddle.to_tensor(0.0, dtype='float32')
+        return paddle.to_tensor([0.0], dtype='float32')
 
 
 def fake_sample_reader():

diff --git a/python/paddle/fluid/tests/unittests/test_cholesky_op.py b/python/paddle/fluid/tests/unittests/test_cholesky_op.py
@@ -104,7 +104,7 @@ def test_dygraph(self):
         a = np.random.rand(3, 3)
         a_t = np.transpose(a, [1, 0])
         x_data = np.matmul(a, a_t) + 1e-03
-        x = paddle.to_tensor(x_data)
+        x = paddle.to_tensor([x_data])
         out = paddle.cholesky(x, upper=False)
 
 

diff --git a/python/paddle/fluid/tests/unittests/test_deg2rad.py b/python/paddle/fluid/tests/unittests/test_deg2rad.py
@@ -74,7 +74,7 @@ def setUp(self):
     def test_dygraph(self):
         paddle.disable_static()
 
-        x2 = paddle.to_tensor(180)
+        x2 = paddle.to_tensor([180])
         result2 = paddle.deg2rad(x2)
         np.testing.assert_allclose(np.pi, result2.numpy(), rtol=1e-05)
 

diff --git a/python/paddle/fluid/tests/unittests/test_deprecated_decorator.py b/python/paddle/fluid/tests/unittests/test_deprecated_decorator.py
@@ -111,7 +111,7 @@ def test_ops_elementwise_mul(self):
     def test_tensor_gradient(self):
         paddle.__version__ = '2.1.0'
 
-        x = paddle.to_tensor(5.0, stop_gradient=False)
+        x = paddle.to_tensor([5.0], stop_gradient=False)
         y = paddle.pow(x, 4.0)
         y.backward()
 

diff --git a/python/paddle/fluid/tests/unittests/test_einsum_v2.py b/python/paddle/fluid/tests/unittests/test_einsum_v2.py
@@ -553,7 +553,7 @@ def test_shape(self):
             B = paddle.to_tensor(np.array([2.0, 3.0])).astype(paddle.bfloat16)
             B = B.cuda()
             C = paddle.einsum('i,i->', A, B)
-            D = paddle.to_tensor(8.0).astype(paddle.bfloat16)
+            D = paddle.to_tensor([8.0]).astype(paddle.bfloat16)
             self.assertEqual(C.item(), D.item())
 
 

diff --git a/python/paddle/fluid/tests/unittests/test_inplace.py b/python/paddle/fluid/tests/unittests/test_inplace.py
@@ -259,11 +259,11 @@ def inplace_api_processing(self, var):
 
 class TestDygraphInplaceReshapeTensor(TestDygraphInplace):
     def non_inplace_api_processing(self, var):
-        shape = paddle.to_tensor(-1)
+        shape = paddle.to_tensor([-1])
         return paddle.reshape(var, shape)
 
     def inplace_api_processing(self, var):
-        shape = paddle.to_tensor(-1)
+        shape = paddle.to_tensor([-1])
         return paddle.reshape_(var, shape)
 
 

diff --git a/python/paddle/fluid/tests/unittests/test_jit_save_load.py b/python/paddle/fluid/tests/unittests/test_jit_save_load.py
@@ -283,7 +283,7 @@ def __init__(self, in_size, out_size):
         super().__init__()
         self._linear_0 = Linear(in_size, out_size)
         self._linear_1 = Linear(in_size, out_size)
-        self._scale = paddle.to_tensor(9.9)
+        self._scale = paddle.to_tensor([9.9])
 
     @paddle.jit.to_static
     def forward(self, x):
@@ -1196,7 +1196,7 @@ def __init__(self, in_size, out_size):
         self._linear_1_0 = Linear(self.hidden, self.hidden)
         self._linear_1_1 = Linear(self.hidden, self.hidden)
         self._linear_2 = Linear(self.hidden, out_size)
-        self._scale = paddle.to_tensor(9.9)
+        self._scale = paddle.to_tensor([9.9])
 
     @paddle.jit.to_static
     def forward(self, x):
@@ -1319,7 +1319,7 @@ def __init__(self, in_size, out_size, load_path):
         self._linear_1_0 = Linear(out_size, in_size)
         self._linear_1_1 = Linear(out_size, in_size)
         self._linear_2 = Linear(out_size, out_size)
-        self._scale = paddle.to_tensor(9.9)
+        self._scale = paddle.to_tensor([9.9])
 
         # Load multiple times
         self._load_l1 = paddle.jit.load(load_path)