Python numpy.type() Examples

def get_bool_nodes(nodes):   # type: (Tuple[NgraphNode, ...]) -> Tuple[NgraphNode, ...]
    """Convert each input node to bool data type if necessary.

    :param nodes: Input nodes to be converted.
    :return: Converted nodes.
    """
    bool_nodes = []
    for node in nodes:
        if not node.get_element_type() == NgraphType.boolean:
            bool_nodes.append(ng.convert(node, bool))
            logger.warning('Converting node of type: <{}> to bool.'.format(get_dtype(
                node.get_element_type())))
        else:
            bool_nodes.append(node)

    return tuple(bool_nodes) 

def quantize(arr, min_val, max_val, levels, dtype=np.int64):
    """Quantize an array of (-inf, inf) to [0, levels-1].

    Args:
        arr (ndarray): Input array.
        min_val (scalar): Minimum value to be clipped.
        max_val (scalar): Maximum value to be clipped.
        levels (int): Quantization levels.
        dtype (np.type): The type of the quantized array.

    Returns:
        tuple: Quantized array.
    """
    if not (isinstance(levels, int) and levels > 1):
        raise ValueError(
            f'levels must be a positive integer, but got {levels}')
    if min_val >= max_val:
        raise ValueError(
            f'min_val ({min_val}) must be smaller than max_val ({max_val})')

    arr = np.clip(arr, min_val, max_val) - min_val
    quantized_arr = np.minimum(
        np.floor(levels * arr / (max_val - min_val)).astype(dtype), levels - 1)

    return quantized_arr 

def dequantize(arr, min_val, max_val, levels, dtype=np.float64):
    """Dequantize an array.

    Args:
        arr (ndarray): Input array.
        min_val (scalar): Minimum value to be clipped.
        max_val (scalar): Maximum value to be clipped.
        levels (int): Quantization levels.
        dtype (np.type): The type of the dequantized array.

    Returns:
        tuple: Dequantized array.
    """
    if not (isinstance(levels, int) and levels > 1):
        raise ValueError(
            f'levels must be a positive integer, but got {levels}')
    if min_val >= max_val:
        raise ValueError(
            f'min_val ({min_val}) must be smaller than max_val ({max_val})')

    dequantized_arr = (arr + 0.5).astype(dtype) * (max_val -
                                                   min_val) / levels + min_val

    return dequantized_arr 

def charge_sector(self, value):
        if type(value) == int and value == 0:
            value = self.leg.chinfo.make_valid()  # zero charges
        elif value is not None:
            value = self.leg.chinfo.make_valid(value)
        self._charge_sector = value
        if value is not None:
            self._mask = np.all(self.leg.to_qflat() == value[np.newaxis, :], axis=1)
            self.shape = tuple([np.sum(self._mask)] * 2)
        else:
            chi2 = self.leg.ind_len
            self.shape = (chi2, chi2)
            self._mask = np.ones([chi2], dtype=np.bool) 

def onnx_tensor_type_to_numpy_type(data_type):  # type: (Any) -> np.dtype
    """Return ONNX TensorProto type mapped into numpy dtype.

    :param data_type: The type we want to convert from.
    :return: Converted numpy dtype.
    """
    if type(data_type) is int:
        return TENSOR_TYPE_TO_NP_TYPE[data_type]
    elif type(data_type) is str:
        return TENSOR_TYPE_TO_NP_TYPE[TensorProto.DataType.Value(data_type)]
    else:
        raise ValueError('Unsupported data type representation (%s).', str(type(data_type))) 

def np_dtype_to_tensor_type_name(data_type):  # type: (np.dtype) -> str
    """Return TensorProto type name respective to provided numpy dtype.

    :param data_type: Numpy dtype we want to convert.
    :return: String representation of TensorProto type name.
    """
    return TensorProto.DataType.Name(NP_TYPE_TO_TENSOR_TYPE[data_type]) 

def np_dtype_to_tensor_type(data_type):  # type: (np.type) -> int
    """Return TensorProto type for provided numpy dtype.

    :param data_type: Numpy data type object.
    :return: TensorProto.DataType enum value for corresponding type.
    """
    return NP_TYPE_TO_TENSOR_TYPE[data_type] 

def __repr__(self):
        return "{}({} {} {}{})".format(
            type(self).__name__.title(), self.shape, str(self.dtype), "; +batch" if self.has_batch_rank else
            "", "; +time" if self.has_time_rank else ""
        ) 

def load_mnist_image(shape=(28, 28), dtype=np.float32,
            bounds=(0, 1), data_format='channels_last',
            fname='mnist0.png', normalize=False):
    """Return the sample mnist image for testing

    Parameters
    ----------
    shape : list of integers
        The shape of the returned image.
    dype : np.type
        The type for loading the image
    bounds : float tuple
        the range of loaded image before normalization
    data_format : str
        "channels_first" or "channels_last"
    fname : str
        The name of sample image
    normalize : Bool
        Whether the image is needed to be normalized.
    """
    from PIL import Image

    path = os.path.join(os.path.dirname(__file__), 'images/%s' % fname)
    image = Image.open(path)
    image = np.asarray(image, dtype=dtype)
    if(data_format == 'channels_first'):
        image = image.reshape([1]+list(shape))
    else:
        image = image.reshape(list(shape)+[1])

    if bounds != (0, 255):
        image /= 255.

    return image 

def from_guess_with_pipe(cls, npc_matvec, v0_guess, labels_split=None, dtype=None):
        """Create a `FlatLinearOperator`` from a `matvec` function acting on multiple legs.

        This function creates a wrapper `matvec` function to allow acting on a "vector" with
        multiple legs. The wrapper combines the legs into a :class:`~tenpy.linalg.charges.LegPipe`
        before calling the actual `matvec` function, and splits them again in the end.

        Parameters
        ----------
        npc_matvec : function
            Function to calculate the action of the linear operator on an npc vector
            with the given split labels `labels_split`.
            Has to return an npc vector with the same legs.
        v0_guess : :class:`~tenpy.linalg.np_conserved.Array`
            Initial guess/starting vector which can be applied to `npc_matvec`.
        labels_split : None | list of str
            Labels of v0_guess in the order in which they are to be combined into a
            :class:`~tenpy.linalg.charges.LegPipe`. ``None`` defaults to
            ``v0_guess.get_leg_labels()``.
        dtype : np.dtype | None
            The data type of the arrays. ``None`` defaults to dtype of `v0_guess` (!).

        Returns
        -------
        lin_op : cls
            Instance of the class to be used as linear operator
        guess_flat : np.ndarray
            Numpy vector representing the guess `v0_guess`.
        """
        if dtype is None:
            dtype = v0_guess.dtype
        if labels_split is None:
            labels_split = v0_guess.get_leg_labels()
        v0_combined = v0_guess.combine_legs(labels_split, qconj=+1)
        if v0_combined.rank != 1:
            raise ValueError("`labels_split` must contain all the legs of `v0_guess`")
        pipe = v0_combined.legs[0]
        pipe_label = v0_combined.get_leg_labels()[0]
        res = cls(npc_matvec, pipe, dtype, v0_combined.qtotal, pipe_label)
        res._labels_split = labels_split
        res._npc_matvec_multileg = npc_matvec
        res.npc_matvec = res._npc_matvec_wrapper  # activate the wrapper
        guess_flat = res.npc_to_flat(v0_combined)
        return res, guess_flat 

def __init__(self, low, high, shape=None, add_batch_rank=False, add_time_rank=False, time_major=False,
                 dtype=np.float32):
        """
        Args:
            low (any): The lower bound (see Valid Inputs for more information).
            high (any): The upper bound (see Valid Inputs for more information).
            shape (tuple): The shape of this space.
            dtype (np.type): The data type (as numpy type) for this Space.
                Allowed are: np.int8,16,32,64, np.float16,32,64 and np.bool_.

        Valid inputs:
            BoxSpace(0.0, 1.0) # low and high are given as scalars and shape is assumed to be ()
                -> single scalar between low and high.
            BoxSpace(-1.0, 1.0, (3,4)) # low and high are scalars, and shape is provided -> nD array
                where all(!) elements are between low and high.
            BoxSpace(np.array([-1.0,-2.0]), np.array([2.0,4.0])) # low and high are arrays of the same shape
                (no shape given!) -> nD array where each dimension has different bounds.
        """
        super(BoxSpace, self).__init__(add_batch_rank=add_batch_rank, add_time_rank=add_time_rank,
                                       time_major=time_major)

        self.dtype = dtype

        # Determine the shape.
        if shape is None:
            if isinstance(low, (int, float, bool)):
                self._shape = ()
            else:
                self._shape = np.shape(low)
        else:
            assert isinstance(shape, (tuple, list)), "ERROR: `shape` must be None or a tuple/list."
            self._shape = tuple(shape)

        # Determine the bounds.
        # False if bounds are individualized (each dimension has its own lower and upper bounds and we can get
        # the single values from self.low and self.high), or a tuple of the globally valid low/high values that apply
        # to all values in all dimensions.
        # 0D Space.
        if self._shape == ():
            if isinstance(low, np.ndarray):
                assert low.shape == (), "ERROR: If shape == (), `low` must be scalar!"
                low = np.asscalar(low)
            if isinstance(high, np.ndarray):
                assert high.shape == (), "ERROR: If shape == (), `high` must be scalar!"
                high = np.asscalar(high)
            self.global_bounds = (low, high)
        # nD Space (n > 0). Bounds can be single number or individual bounds.
        else:
            # Low/high values are given individually per item.
            if isinstance(low, (list, tuple, np.ndarray)):
                self.global_bounds = False
            # Only one low/high value. Use these as generic bounds for all values.
            else:
                assert np.isscalar(low) and np.isscalar(high)
                self.global_bounds = (low, high)

        self.low = np.array(low)
        self.high = np.array(high)
        assert self.low.shape == self.high.shape 

def load_cifar_image(shape=(32, 32), dtype=np.float32,
            bounds=(0, 1), data_format='channels_last',
            fname='cifar0.png', normalize=True):
    """Return the sample mnist image for testing

    Parameters
    ----------
    shape : list of integers
        The shape of the returned image.
    dype : np.type
        The type for loading the image
    bounds : float tuple
        the range of loaded image before normalization
    data_format : str
        "channels_first" or "channels_last"
    fname : str
        The name of sample image
    normalize : Bool
        Whether the image is needed to be normalized.
    """
    from PIL import Image

    path = os.path.join(os.path.dirname(__file__), 'images/%s' % fname)
    image = Image.open(path)
    image = np.asarray(image, dtype=dtype)
    if(data_format == 'channels_first'):
        image = image.reshape([3]+list(shape))
    else:
        image = image.reshape(list(shape)+[3])

    if bounds != (0, 255):
        image /= 255.

    if(normalize):
        mean = np.array([0.485, 0.456, 0.406]).reshape(3,1,1)
        std = np.array([0.225, 0.225, 0.225]).reshape(3,1,1)
        image = image - mean
        image = image / std

    image = np.asarray(image, dtype=dtype)

    return image