def test_squeeze(self):
self.assertEqual([3, 2], infer.squeeze([3, 2, 1], [-1]))
self.assertEqual([2, 3], infer.squeeze([1, 2, 3], [0]))
self.assertEqual([2, 2], infer.squeeze([2, 1, 2, 1], [1, -1]))
self.assertEqual([], infer.squeeze([1, 1], [0, -1]))
self.assertEqual([3, 2], infer.squeeze([1, 3, 1, 1, 2, 1, 1]))
def nnef_squeeze(input, axes):
return input.reshape(shape_inference.squeeze(input.shape, axes))
def propagate_squeeze(op, const_value_by_tensor):
# type: (TFOperation, _ConstValueByTensorT)->typing.Tuple[typing.List[typing.List[int]], typing.List[str]]
axes = op.attribs['squeeze_dims']
return [infer.squeeze(input=op.input.shape, axes=axes if axes else None)], [op.attribs['T']]
def propagate_squeeze(op):
# type: (ONNXOperation)->typing.Tuple[typing.List[typing.List[int]], typing.List[str]]
return [
infer.squeeze(input=op.input.shape, axes=op.attribs.get('axes', None))
], [op.input.dtype]
def squeeze_shape(op):
# type: (Caffe2Operation)->ShapeResult
return infer.squeeze(op.inputs[0].shape,
axes=op.attribs['dims']), op.inputs[0].dtype