fix diffusers pipelines s class ut (#2096)

Author: lvyufeng

mindnlp/core/_tensor.py

@@ -82,6 +82,14 @@ def tensor(data, *, dtype=None, device=None, requires_grad=False):
         UserWarning("To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than core.tensor(sourceTensor).")
         return Tensor(data)
 
+    if isinstance(data, list):
+        new_data = []
+        for d in data:
+            if isinstance(d, Tensor):
+                d = d.item()
+            new_data.append(d)
+        data = new_data
+
     if device is None:
         device = get_default_device()
 
@@ -730,6 +738,13 @@ def __iter__(self):
     Tensor.__iter__ = __iter__
     StubTensor.__iter__ = __iter__
 
+    def __float__(self):
+        out = self.item()
+        return round(float(out), 5)
+
+    Tensor.__float__ = __float__
+    StubTensor.__float__ = __float__
+
 def _rebuild_from_type_v2(func, new_type, args, state):
     ret = func(*args)
     return ret

mindnlp/core/linalg/__init__.py

@@ -1,4 +1,6 @@
 from collections import namedtuple
+import numpy as np
+
 from mindspore import ops, mint
 from mindspore.ops._primitive_cache import _get_cache_prim
 
@@ -25,4 +27,7 @@ def norm(A, ord=None, dim=None, keepdim=False, *, out=None, dtype=None):
     return mint.norm(A, 2 if ord is None else ord, dim, keepdim, dtype=dtype)
 
 def vector_norm(x, ord=2, dim=None, keepdim=False, *, dtype=None, out=None):
-    return mint.linalg.vector_norm(x, ord, dim, keepdim, dtype=dtype)
+    return mint.linalg.vector_norm(x, ord, dim, keepdim, dtype=dtype)
+
+def solve(A, B, *, left=True, out=None):
+    return core.tensor(np.linalg.solve(A.numpy(), B.numpy()))

mindnlp/core/nn/modules/__init__.py

@@ -7,7 +7,7 @@
 from .dropout import Dropout, Dropout2d
 from .activation import *
 from .conv import Conv3d, Conv2d, Conv1d, ConvTranspose2d, ConvTranspose1d, ConvTranspose3d
-from .padding import ZeroPad2d, ConstantPad2d, ConstantPad1d, ConstantPad3d
+from .padding import *
 from .batchnorm import BatchNorm2d, BatchNorm1d, SyncBatchNorm
 from .pooling import AdaptiveAvgPool2d, AvgPool1d, MaxPool2d, MaxPool1d, AdaptiveAvgPool1d, AvgPool2d
 from .flatten import Unflatten, Flatten

mindnlp/core/nn/modules/padding.py

@@ -251,3 +251,311 @@ def __init__(self, padding: _size_6_t) -> None:
 
     def extra_repr(self) -> str:
         return f'{self.padding}'
+
+class _ReflectionPadNd(Module):
+    __constants__ = ["padding"]
+    padding: Sequence[int]
+
+    def forward(self, input: Tensor) -> Tensor:
+        return F.pad(input, self.padding, "reflect")
+
+    def extra_repr(self) -> str:
+        return f"{self.padding}"
+
+
+class ReflectionPad1d(_ReflectionPadNd):
+    r"""Pads the input tensor using the reflection of the input boundary.
+
+    For `N`-dimensional padding, use :func:`torch.nn.functional.pad()`.
+
+    Args:
+        padding (int, tuple): the size of the padding. If is `int`, uses the same
+            padding in all boundaries. If a 2-`tuple`, uses
+            (:math:`\text{padding\_left}`, :math:`\text{padding\_right}`)
+            Note that padding size should be less than the corresponding input dimension.
+
+    Shape:
+        - Input: :math:`(C, W_{in})` or :math:`(N, C, W_{in})`.
+        - Output: :math:`(C, W_{out})` or :math:`(N, C, W_{out})`, where
+
+          :math:`W_{out} = W_{in} + \text{padding\_left} + \text{padding\_right}`
+
+    Examples::
+
+        >>> m = nn.ReflectionPad1d(2)
+        >>> # xdoctest: +IGNORE_WANT("other tests seem to modify printing styles")
+        >>> input = torch.arange(8, dtype=torch.float).reshape(1, 2, 4)
+        >>> input
+        tensor([[[0., 1., 2., 3.],
+                 [4., 5., 6., 7.]]])
+        >>> m(input)
+        tensor([[[2., 1., 0., 1., 2., 3., 2., 1.],
+                 [6., 5., 4., 5., 6., 7., 6., 5.]]])
+        >>> # using different paddings for different sides
+        >>> m = nn.ReflectionPad1d((3, 1))
+        >>> m(input)
+        tensor([[[3., 2., 1., 0., 1., 2., 3., 2.],
+                 [7., 6., 5., 4., 5., 6., 7., 6.]]])
+    """
+
+    padding: tuple[int, int]
+
+    def __init__(self, padding: _size_2_t) -> None:
+        super().__init__()
+        self.padding = _pair(padding)
+
+
+class ReflectionPad2d(_ReflectionPadNd):
+    r"""Pads the input tensor using the reflection of the input boundary.
+
+    For `N`-dimensional padding, use :func:`torch.nn.functional.pad()`.
+
+    Args:
+        padding (int, tuple): the size of the padding. If is `int`, uses the same
+            padding in all boundaries. If a 4-`tuple`, uses (:math:`\text{padding\_left}`,
+            :math:`\text{padding\_right}`, :math:`\text{padding\_top}`, :math:`\text{padding\_bottom}`)
+            Note that padding size should be less than the corresponding input dimension.
+
+    Shape:
+        - Input: :math:`(N, C, H_{in}, W_{in})` or :math:`(C, H_{in}, W_{in})`.
+        - Output: :math:`(N, C, H_{out}, W_{out})` or :math:`(C, H_{out}, W_{out})` where
+
+          :math:`H_{out} = H_{in} + \text{padding\_top} + \text{padding\_bottom}`
+
+          :math:`W_{out} = W_{in} + \text{padding\_left} + \text{padding\_right}`
+
+    Examples::
+
+        >>> # xdoctest: +IGNORE_WANT("not sure why xdoctest is choking on this")
+        >>> m = nn.ReflectionPad2d(2)
+        >>> input = torch.arange(9, dtype=torch.float).reshape(1, 1, 3, 3)
+        >>> input
+        tensor([[[[0., 1., 2.],
+                  [3., 4., 5.],
+                  [6., 7., 8.]]]])
+        >>> m(input)
+        tensor([[[[8., 7., 6., 7., 8., 7., 6.],
+                  [5., 4., 3., 4., 5., 4., 3.],
+                  [2., 1., 0., 1., 2., 1., 0.],
+                  [5., 4., 3., 4., 5., 4., 3.],
+                  [8., 7., 6., 7., 8., 7., 6.],
+                  [5., 4., 3., 4., 5., 4., 3.],
+                  [2., 1., 0., 1., 2., 1., 0.]]]])
+        >>> # using different paddings for different sides
+        >>> m = nn.ReflectionPad2d((1, 1, 2, 0))
+        >>> m(input)
+        tensor([[[[7., 6., 7., 8., 7.],
+                  [4., 3., 4., 5., 4.],
+                  [1., 0., 1., 2., 1.],
+                  [4., 3., 4., 5., 4.],
+                  [7., 6., 7., 8., 7.]]]])
+    """
+
+    padding: tuple[int, int, int, int]
+
+    def __init__(self, padding: _size_4_t) -> None:
+        super().__init__()
+        self.padding = _quadruple(padding)
+
+
+class ReflectionPad3d(_ReflectionPadNd):
+    r"""Pads the input tensor using the reflection of the input boundary.
+
+    For `N`-dimensional padding, use :func:`torch.nn.functional.pad()`.
+
+    Args:
+        padding (int, tuple): the size of the padding. If is `int`, uses the same
+            padding in all boundaries. If a 6-`tuple`, uses
+            (:math:`\text{padding\_left}`, :math:`\text{padding\_right}`,
+            :math:`\text{padding\_top}`, :math:`\text{padding\_bottom}`,
+            :math:`\text{padding\_front}`, :math:`\text{padding\_back}`)
+            Note that padding size should be less than the corresponding input dimension.
+
+    Shape:
+        - Input: :math:`(N, C, D_{in}, H_{in}, W_{in})` or :math:`(C, D_{in}, H_{in}, W_{in})`.
+        - Output: :math:`(N, C, D_{out}, H_{out}, W_{out})` or :math:`(C, D_{out}, H_{out}, W_{out})`,
+          where
+
+          :math:`D_{out} = D_{in} + \text{padding\_front} + \text{padding\_back}`
+
+          :math:`H_{out} = H_{in} + \text{padding\_top} + \text{padding\_bottom}`
+
+          :math:`W_{out} = W_{in} + \text{padding\_left} + \text{padding\_right}`
+
+    Examples::
+
+        >>> # xdoctest: +IGNORE_WANT("not sure why xdoctest is choking on this")
+        >>> m = nn.ReflectionPad3d(1)
+        >>> input = torch.arange(8, dtype=torch.float).reshape(1, 1, 2, 2, 2)
+        >>> m(input)
+        tensor([[[[[7., 6., 7., 6.],
+                   [5., 4., 5., 4.],
+                   [7., 6., 7., 6.],
+                   [5., 4., 5., 4.]],
+                  [[3., 2., 3., 2.],
+                   [1., 0., 1., 0.],
+                   [3., 2., 3., 2.],
+                   [1., 0., 1., 0.]],
+                  [[7., 6., 7., 6.],
+                   [5., 4., 5., 4.],
+                   [7., 6., 7., 6.],
+                   [5., 4., 5., 4.]],
+                  [[3., 2., 3., 2.],
+                   [1., 0., 1., 0.],
+                   [3., 2., 3., 2.],
+                   [1., 0., 1., 0.]]]]])
+    """
+
+    padding: tuple[int, int, int, int, int, int]
+
+    def __init__(self, padding: _size_6_t) -> None:
+        super().__init__()
+        self.padding = _ntuple(6)(padding)
+
+
+class _ReplicationPadNd(Module):
+    __constants__ = ["padding"]
+    padding: Sequence[int]
+
+    def forward(self, input: Tensor) -> Tensor:
+        return F.pad(input, self.padding, "replicate")
+
+    def extra_repr(self) -> str:
+        return f"{self.padding}"
+
+
+class ReplicationPad1d(_ReplicationPadNd):
+    r"""Pads the input tensor using replication of the input boundary.
+
+    For `N`-dimensional padding, use :func:`torch.nn.functional.pad()`.
+
+    Args:
+        padding (int, tuple): the size of the padding. If is `int`, uses the same
+            padding in all boundaries. If a 2-`tuple`, uses
+            (:math:`\text{padding\_left}`, :math:`\text{padding\_right}`)
+            Note that the output dimensions must remain positive.
+
+    Shape:
+        - Input: :math:`(C, W_{in})` or :math:`(N, C, W_{in})`.
+        - Output: :math:`(C, W_{out})` or :math:`(N, C, W_{out})`, where
+
+          :math:`W_{out} = W_{in} + \text{padding\_left} + \text{padding\_right}`
+
+    Examples::
+
+        >>> # xdoctest: +IGNORE_WANT("not sure why xdoctest is choking on this")
+        >>> m = nn.ReplicationPad1d(2)
+        >>> input = torch.arange(8, dtype=torch.float).reshape(1, 2, 4)
+        >>> input
+        tensor([[[0., 1., 2., 3.],
+                 [4., 5., 6., 7.]]])
+        >>> m(input)
+        tensor([[[0., 0., 0., 1., 2., 3., 3., 3.],
+                 [4., 4., 4., 5., 6., 7., 7., 7.]]])
+        >>> # using different paddings for different sides
+        >>> m = nn.ReplicationPad1d((3, 1))
+        >>> m(input)
+        tensor([[[0., 0., 0., 0., 1., 2., 3., 3.],
+                 [4., 4., 4., 4., 5., 6., 7., 7.]]])
+    """
+
+    padding: tuple[int, int]
+
+    def __init__(self, padding: _size_2_t) -> None:
+        super().__init__()
+        self.padding = _pair(padding)
+
+
+class ReplicationPad2d(_ReplicationPadNd):
+    r"""Pads the input tensor using replication of the input boundary.
+
+    For `N`-dimensional padding, use :func:`torch.nn.functional.pad()`.
+
+    Args:
+        padding (int, tuple): the size of the padding. If is `int`, uses the same
+            padding in all boundaries. If a 4-`tuple`, uses (:math:`\text{padding\_left}`,
+            :math:`\text{padding\_right}`, :math:`\text{padding\_top}`, :math:`\text{padding\_bottom}`)
+            Note that the output dimensions must remain positive.
+
+    Shape:
+        - Input: :math:`(N, C, H_{in}, W_{in})` or :math:`(C, H_{in}, W_{in})`.
+        - Output: :math:`(N, C, H_{out}, W_{out})` or :math:`(C, H_{out}, W_{out})`, where
+
+          :math:`H_{out} = H_{in} + \text{padding\_top} + \text{padding\_bottom}`
+
+          :math:`W_{out} = W_{in} + \text{padding\_left} + \text{padding\_right}`
+
+    Examples::
+
+        >>> m = nn.ReplicationPad2d(2)
+        >>> # xdoctest: +IGNORE_WANT("non-deterministic")
+        >>> input = torch.arange(9, dtype=torch.float).reshape(1, 1, 3, 3)
+        >>> input
+        tensor([[[[0., 1., 2.],
+                  [3., 4., 5.],
+                  [6., 7., 8.]]]])
+        >>> m(input)
+        tensor([[[[0., 0., 0., 1., 2., 2., 2.],
+                  [0., 0., 0., 1., 2., 2., 2.],
+                  [0., 0., 0., 1., 2., 2., 2.],
+                  [3., 3., 3., 4., 5., 5., 5.],
+                  [6., 6., 6., 7., 8., 8., 8.],
+                  [6., 6., 6., 7., 8., 8., 8.],
+                  [6., 6., 6., 7., 8., 8., 8.]]]])
+        >>> # using different paddings for different sides
+        >>> m = nn.ReplicationPad2d((1, 1, 2, 0))
+        >>> m(input)
+        tensor([[[[0., 0., 1., 2., 2.],
+                  [0., 0., 1., 2., 2.],
+                  [0., 0., 1., 2., 2.],
+                  [3., 3., 4., 5., 5.],
+                  [6., 6., 7., 8., 8.]]]])
+    """
+
+    padding: tuple[int, int, int, int]
+
+    def __init__(self, padding: _size_4_t) -> None:
+        super().__init__()
+        self.padding = _quadruple(padding)
+
+
+class ReplicationPad3d(_ReplicationPadNd):
+    r"""Pads the input tensor using replication of the input boundary.
+
+    For `N`-dimensional padding, use :func:`torch.nn.functional.pad()`.
+
+    Args:
+        padding (int, tuple): the size of the padding. If is `int`, uses the same
+            padding in all boundaries. If a 6-`tuple`, uses
+            (:math:`\text{padding\_left}`, :math:`\text{padding\_right}`,
+            :math:`\text{padding\_top}`, :math:`\text{padding\_bottom}`,
+            :math:`\text{padding\_front}`, :math:`\text{padding\_back}`)
+            Note that the output dimensions must remain positive.
+
+    Shape:
+        - Input: :math:`(N, C, D_{in}, H_{in}, W_{in})` or :math:`(C, D_{in}, H_{in}, W_{in})`.
+        - Output: :math:`(N, C, D_{out}, H_{out}, W_{out})` or :math:`(C, D_{out}, H_{out}, W_{out})`,
+          where
+
+          :math:`D_{out} = D_{in} + \text{padding\_front} + \text{padding\_back}`
+
+          :math:`H_{out} = H_{in} + \text{padding\_top} + \text{padding\_bottom}`
+
+          :math:`W_{out} = W_{in} + \text{padding\_left} + \text{padding\_right}`
+
+    Examples::
+
+        >>> # xdoctest: +IGNORE_WANT("non-deterministic")
+        >>> m = nn.ReplicationPad3d(3)
+        >>> input = torch.randn(16, 3, 8, 320, 480)
+        >>> output = m(input)
+        >>> # using different paddings for different sides
+        >>> m = nn.ReplicationPad3d((3, 3, 6, 6, 1, 1))
+        >>> output = m(input)
+    """
+
+    padding: tuple[int, int, int, int, int, int]
+
+    def __init__(self, padding: _size_6_t) -> None:
+        super().__init__()
+        self.padding = _ntuple(6)(padding)