def testReadmeExample(self):
data = random_ops.random_uniform((128, 128), 0, 10, dtype=dtypes.int32)
histogram = math_ops.bincount(data, minlength=10, maxlength=10)
cdf = math_ops.cumsum(histogram, exclusive=False)
cdf = array_ops.pad(cdf, [[1, 0]])
cdf = array_ops.reshape(cdf, [1, 1, -1])
data = math_ops.cast(data, dtypes.int16)
encoded = coder_ops.range_encode(data, cdf, precision=14)
decoded = coder_ops.range_decode(
encoded, array_ops.shape(data), cdf, precision=14)
with self.test_session() as sess:
self.assertAllEqual(*sess.run((data, decoded)))
def compress(self, inputs, device='cpu'):
"""Compress inputs and store their binary representations into strings.
Arguments:
inputs: `Tensor` with values to be compressed. Must have shape
[points, channels] (torch)
Returns:
compressed: String vector containing the compressed
representation of each batch element of `inputs`. (numpy)
min_v & max_v (numpy).
"""
# get symbols
values = self._quantize(inputs, "symbols")
values = values.detach()
self._channels = values.shape[-1]
# get range[min_v, max_v]
min_v = torch.min(values.data)
min_v = torch.floor(min_v)
min_v = min_v.short()
#print('min:',min_v)
max_v = torch.max(values.data)
max_v = torch.ceil(max_v)
max_v = max_v.short()
#print('max:', max_v)
min_v_data = min_v.cpu().numpy()
max_v_data = max_v.cpu().numpy()
# early stop
if min_v_data == max_v_data:
strings = bytes(1)
return strings, min_v_data, max_v_data
# get cdf
cdf = self._get_cdf(min_v_data, max_v_data, device)
# To tensorflow
values_data = values.detach().cpu().numpy()
values_tf = tf.convert_to_tensor(values_data)# shape=[points, channels]
values_tf = tf.reshape(values_tf, [-1, self._channels])# TODO: delete.
values_tf = tf.cast(values_tf, tf.int16)
values_tf -= min_v_data
# range encode.
strings = coder_ops.range_encode(
values_tf, cdf, precision=self._range_coder_precision)
# TODO: check numpy, torch, tf.
return strings.numpy(), min_v_data, max_v_data
def compress(self, inputs, loc, scale):
"""Compress inputs and store their binary representations into strings.
Arguments:
inputs: `Tensor` with values to be compressed. Must have shape
[**batch size**, length, width, height, channels]
locs & scales: same shape like inputs.
Returns:
compressed: String `Tensor` vector containing the compressed
representation of each batch element of `inputs`.
"""
inputs = tf.convert_to_tensor(inputs, dtype=self.dtype)
loc = tf.convert_to_tensor(loc, dtype=self.dtype)
scale = tf.convert_to_tensor(scale, dtype=self.dtype)
datashape = tf.shape(inputs)
channels = datashape[-1]
# reshape.
loc = tf.reshape(loc, [-1, channels])
scale = tf.reshape(scale, [-1, channels])
inputs = tf.reshape(inputs, [-1, channels])
# quantize.
values = self._quantize(inputs, "symbols")
# get cdf
min_v = tf.cast(tf.floor(tf.reduce_min(values)), dtype=tf.int32)
max_v = tf.cast(tf.ceil(tf.reduce_max(values)), dtype=tf.int32)
cdf = self._get_cdf(loc, scale, min_v, max_v,
datashape) # [BatchSizexHxWxD, C, N]
# range encode.
values = tf.cast(values, "int32")
values -= min_v
values = tf.cast(values, "int16")
strings = coder_ops.range_encode(values,
cdf,
precision=self._range_coder_precision)
return strings, min_v, max_v
def compress(self, inputs):
"""Compress inputs and store their binary representations into strings.
Arguments:
inputs: `Tensor` with values to be compressed. Must have shape
[**batch size**, length, width, height, channels]
Returns:
compressed: String `Tensor` vector containing the compressed
representation of each batch element of `inputs`.
"""
with tf.name_scope(self._name_scope()):
inputs = tf.convert_to_tensor(inputs)
if not self.built:
if self.dtype is None:
self._dtype = inputs.dtype.base_dtype.name
self.build(inputs.shape)
ndim = self.input_spec.ndim
channel_axes = ndim - 1
channels = self.input_spec.axes[channel_axes]
# quantize.
values = self._quantize(inputs, "symbols")
# get cdf
min_v = tf.cast(tf.floor(tf.reduce_min(values)), dtype=tf.int32)
max_v = tf.cast(tf.ceil(tf.reduce_max(values)), dtype=tf.int32)
cdf = self._get_cdf(min_v, max_v)
# range encode.
values = tf.reshape(values, [-1, channels])
values = tf.cast(values, tf.int32)
values = values - min_v
values = tf.cast(values, tf.int16)
strings = coder_ops.range_encode(
values, cdf, precision=self._range_coder_precision)
return strings, min_v, max_v
def loop_body(tensor):
return coder_ops.range_encode(
tensor, cdf, precision=self.range_coder_precision)
def loop_body(tensor):
return coder_ops.range_encode(
tensor, cdf, precision=self.range_coder_precision)