Python cntk.reshape方法代碼示例

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def _remove_dims(x, axis, keepdims=False):
    if keepdims is False and isinstance(axis, list):
        # sequence axis is removed by default, so don't need reshape on it
        reduce_axes = []
        for a in axis:
            if isinstance(a, C.Axis) is False:
                reduce_axes.append(a)
        return _reshape_dummy_dim(x, reduce_axes)
    else:
        if isinstance(axis, list):
            has_seq = False
            for a in axis:
                if isinstance(a, C.Axis):
                    has_seq = True
                    break
            if has_seq:
                nones = _get_dynamic_axis_num(x)
                x = expand_dims(x, nones)
        return x 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def squeeze(x, axis):
    if isinstance(axis, tuple):
        axis = list(axis)
    if not isinstance(axis, list):
        axis = [axis]

    shape = list(int_shape(x))

    _axis = []
    for _ in axis:
        if isinstance(_, int):
            _axis.append(_ if _ >= 0 else _ + len(shape))

    if len(_axis) == 0:
        return x

    nones = _get_dynamic_axis_num(x)
    for _ in sorted(_axis, reverse=True):
        del shape[_]

    new_shape = shape[nones:]
    new_shape = tuple([C.InferredDimension if _ == C.FreeDimension else _ for _ in new_shape])
    return C.reshape(x, new_shape) 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def _reshape_dummy_dim(x, axis):
    shape = list(x.shape)

    _axis = [_ + len(shape) if _ < 0 else _ for _ in axis]

    if shape.count(C.InferredDimension) > 1 or shape.count(C.FreeDimension) > 1:
        result = x
        for index in sorted(_axis, reverse=True):
            result = C.reshape(result,
                               shape=(),
                               begin_axis=index,
                               end_axis=index + 1)
        return result
    else:
        for index in sorted(_axis, reverse=True):
            del shape[index]

        shape = [C.InferredDimension if _ == C.FreeDimension else _ for _ in shape]
        return C.reshape(x, shape) 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def repeat(x, n):
    # this is a workaround for recurrent layer
    # if n is inferred dimension,
    # we can't figure out how to repeat it in cntk now
    # return the same x to take cntk broadcast feature
    # to make the recurrent layer work.
    # need to be fixed in GA.
    if n is C.InferredDimension or n is C.FreeDimension:
        return x
    index = 1 - _get_dynamic_axis_num(x)
    if index < 0 or index > 1:
        raise NotImplementedError

    new_shape = list(x.shape)
    new_shape.insert(index, 1)
    new_shape = tuple(new_shape)
    x = C.reshape(x, new_shape)
    temp = [x] * n
    return C.splice(*temp, axis=index) 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def _layer_Conv(self):
        self.add_body(0, """
def convolution(input, is_transpose, name, **kwargs):
    dim = __weights_dict[name]['weights'].ndim

    if is_transpose:
        weight = np.transpose(__weights_dict[name]['weights'], [dim - 2, dim - 1] + list(range(0, dim - 2)))
        kwargs.pop('groups', None)
    else:
        weight = np.transpose(__weights_dict[name]['weights'], [dim - 1, dim - 2] + list(range(0, dim - 2)))
    w = cntk.Parameter(init=weight, name=name + '_weight')

    input = cntk.transpose(input, [dim - 2] + list(range(0, dim - 2)))

    if is_transpose:
        layer = ops.convolution_transpose(w, input, **kwargs)
    else:
        layer = ops.convolution(w, input, **kwargs)
    if 'bias' in __weights_dict[name]:
        bias = np.reshape(__weights_dict[name]['bias'], [-1] + [1] * (dim - 2))
        b = cntk.Parameter(init=bias, name=name + '_bias')
        layer = layer + b
    layer = cntk.transpose(layer, list(range(1, dim - 1)) + [0])
    return layer
""") 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def random_binomial(shape, p=0.0, dtype=None, seed=None):
    # use numpy workaround now
    if seed is None:
        # ensure that randomness is conditioned by the Numpy RNG
        seed = np.random.randint(10e7)
        np.random.seed(seed)
    if dtype is None:
        dtype = np.float32
    else:
        dtype = _convert_string_dtype(dtype)

    size = 1
    for _ in shape:
        if _ is None:
            raise ValueError('CNTK Backend: randomness op with '
                             'dynamic shape is not supported now. '
                             'Please provide fixed dimension '
                             'instead of `None`.')
        size *= _

    binomial = np.random.binomial(1, p, size).astype(dtype).reshape(shape)
    return variable(value=binomial, dtype=dtype) 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def local_conv1d(inputs, kernel, kernel_size, strides, data_format=None):
    if data_format is None:
        data_format = image_data_format()
    if data_format not in {'channels_first', 'channels_last'}:
        raise ValueError('Unknown data_format ' + str(data_format))

    stride = strides[0]
    kernel_shape = int_shape(kernel)
    output_length, feature_dim, filters = kernel_shape

    xs = []
    for i in range(output_length):
        slice_length = slice(i * stride,
                             i * stride + kernel_size[0])
        xs.append(reshape(inputs[:, slice_length, :],
                          (-1, 1, feature_dim)))
    x_aggregate = concatenate(xs, axis=1)
    # transpose kernel to output_filters first, to apply broadcast
    weight = permute_dimensions(kernel, (2, 0, 1))
    # Shape: (batch, filters, output_length, input_length * kernel_size)
    output = x_aggregate * weight
    # Shape: (batch, filters, output_length)
    output = sum(output, axis=3)
    # Shape: (batch, output_length, filters)
    return permute_dimensions(output, (0, 2, 1)) 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def expand_dims(x, axis=-1):
    shape = list(int_shape(x))
    nones = _get_dynamic_axis_num(x)
    index = axis if axis >= 0 else len(shape) + 1
    shape.insert(index, 1)
    new_shape = shape[nones:]
    new_shape = tuple(
        [C.InferredDimension if _ is None else _ for _ in new_shape])
    result = C.reshape(x, new_shape)
    if index < nones:
        result._keras_shape = shape
    return result 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def flatten(x):
    return reshape(x, (-1,)) 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def reshape(x, shape):
    shape = tuple([C.InferredDimension if _ == C.FreeDimension else _ for _ in shape])
    if isinstance(x, C.variables.Parameter):
        return C.reshape(x, shape)
    else:
        num_dynamic_axis = _get_dynamic_axis_num(x)

        if num_dynamic_axis == 1 and len(shape) > 0 and shape[0] == -1:
            # collapse axis with batch axis
            if b_any(_ == C.InferredDimension for _ in x.shape) or b_any(
                    _ == C.FreeDimension for _ in x.shape):
                warnings.warn(
                    'Warning: CNTK backend does not support '
                    'collapse of batch axis with inferred dimension. '
                    'The reshape did not take place.')
                return x
            return _reshape_batch(x, shape)
        else:
            # no collapse, then first need to padding the shape
            if num_dynamic_axis >= len(shape):
                i = 0
                while i < len(shape):
                    if shape[i] is None or shape[i] == -1:
                        i += 1
                    else:
                        break
                shape = tuple([-1 for _ in range(num_dynamic_axis - i)]) + shape

            new_shape = list(shape)
            new_shape = new_shape[num_dynamic_axis:]
            new_shape = [C.InferredDimension if _ is None else _ for _ in new_shape]
            return C.reshape(x, new_shape) 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def separable_conv2d(x, depthwise_kernel, pointwise_kernel, strides=(1, 1),
                     padding='valid', data_format=None, dilation_rate=(1, 1)):
    data_format = normalize_data_format(data_format)

    x = _preprocess_conv2d_input(x, data_format)
    depthwise_kernel = _preprocess_conv2d_kernel(depthwise_kernel, data_format)
    depthwise_kernel = C.reshape(C.transpose(depthwise_kernel, (1, 0, 2, 3)),
                                 (-1, 1) + depthwise_kernel.shape[2:])
    pointwise_kernel = _preprocess_conv2d_kernel(pointwise_kernel, data_format)
    padding = _preprocess_border_mode(padding)

    if dilation_rate == (1, 1):
        strides = (1,) + strides
        x = C.convolution(depthwise_kernel, x,
                          strides=strides,
                          auto_padding=[False, padding, padding],
                          groups=x.shape[0])
        x = C.convolution(pointwise_kernel, x,
                          strides=(1, 1, 1),
                          auto_padding=[False])
    else:
        if dilation_rate[0] != dilation_rate[1]:
            raise ValueError('CNTK Backend: non-square dilation_rate is '
                             'not supported.')
        if strides != (1, 1):
            raise ValueError('Invalid strides for dilated convolution')
        x = C.convolution(depthwise_kernel, x,
                          strides=dilation_rate[0],
                          auto_padding=[False, padding, padding])
        x = C.convolution(pointwise_kernel, x,
                          strides=(1, 1, 1),
                          auto_padding=[False])
    return _postprocess_conv2d_output(x, data_format) 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def batch_flatten(x):
    # cntk's batch axis is not in shape,
    # so just flatten all the dim in x.shape
    dim = np.prod(x.shape)
    x = C.reshape(x, (-1,))
    x._keras_shape = (None, dim)
    return x 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def categorical_crossentropy(target, output, from_logits=False, axis=-1):
    # Here, unlike other backends, the tensors lack a batch dimension:
    axis_without_batch = -1 if axis == -1 else axis - 1
    output_dimensions = list(range(len(output.shape)))
    if axis_without_batch != -1 and axis_without_batch not in output_dimensions:
        raise ValueError(
            '{}{}{}'.format(
                'Unexpected channels axis {}. '.format(axis_without_batch),
                'Expected to be -1 or one of the axes of `output`, ',
                'which has {} dimensions.'.format(len(output.shape))))
    # If the channels are not in the last axis, move them to be there:
    if axis_without_batch != -1 and axis_without_batch != output_dimensions[-1]:
        permutation = output_dimensions[:axis_without_batch]
        permutation += output_dimensions[axis_without_batch + 1:]
        permutation += [axis_without_batch]
        output = C.transpose(output, permutation)
        target = C.transpose(target, permutation)
    if from_logits:
        result = C.cross_entropy_with_softmax(output, target)
        # cntk's result shape is (batch, 1), while keras expect (batch, )
        return C.reshape(result, ())
    else:
        # scale preds so that the class probas of each sample sum to 1
        output /= C.reduce_sum(output, axis=-1)
        # avoid numerical instability with epsilon clipping
        output = C.clip(output, epsilon(), 1.0 - epsilon())
        return -sum(target * C.log(output), axis=-1) 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def sparse_categorical_crossentropy(target, output, from_logits=False, axis=-1):
    # Here, unlike other backends, the tensors lack a batch dimension:
    axis_without_batch = -1 if axis == -1 else axis - 1
    output_dimensions = list(range(len(output.shape)))
    if axis_without_batch != -1 and axis_without_batch not in output_dimensions:
        raise ValueError(
            '{}{}{}'.format(
                'Unexpected channels axis {}. '.format(axis_without_batch),
                'Expected to be -1 or one of the axes of `output`, ',
                'which has {} dimensions.'.format(len(output.shape))))
    target = C.one_hot(target, output.shape[axis_without_batch],
                       axis=axis_without_batch)
    target = C.reshape(target, output.shape)
    return categorical_crossentropy(target, output, from_logits, axis=axis) 

# 需要導入模塊: import cntk [as 別名]
# 或者: from cntk import reshape [as 別名]
def in_top_k(predictions, targets, k):
    _targets = C.one_hot(targets, predictions.shape[-1])
    result = C.classification_error(predictions, _targets, topN=k)
    return 1 - C.reshape(result, shape=())