Skip to content

Commit f733f3e

Browse files
DEKHTIARJonathanDEKHTIARJonathan
authored
tl.alphas and tl.alphas_like added following the tf.ones/zeros and tf.zeros_like/ones_like (tensorlayer#580)
* tl.alphas and tl.alphas_like added * YAPF Error Correct * Codacy Error Fix * Docstring Fixed * Codacy Error Fix * Documentation Added * Update CHANGELOG.md
1 parent b3bc15e commit f733f3e

File tree

6 files changed

+386
-7
lines changed

6 files changed

+386
-7
lines changed

‎CHANGELOG.md‎

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -69,10 +69,13 @@ To release a new version, please update the changelog as followed:
6969
## [Unreleased]
7070

7171
### Added
72+
- API:
73+
- `tl.alphas` and `tl.alphas_like` added following the tf.ones/zeros and tf.zeros_like/ones_like (by @DEKHTIARJonathan in #580)
7274
- Tutorials:
7375
- `tutorial_tfslim` has been introduced to show how to use `SlimNetsLayer` (by @2wins in #560).
7476
- Test:
7577
- `Layer_DeformableConvolution_Test` added to reproduce issue #572 with deformable convolution (by @DEKHTIARJonathan in #573)
78+
- `Array_Op_Alphas_Test` and `Array_Op_Alphas_Like_Test` added to test `tensorlayer/array_ops.py` file (by @DEKHTIARJonathan in #580)
7679
- CI Tool:
7780
- Danger CI has been added to enforce the update of the changelog (by @lgarithm and @DEKHTIARJonathan in #563)
7881
- https://github.com/apps/stale/ added to clean stale issues (by @DEKHTIARJonathan in #573)

‎docs/index.rst‎

Lines changed: 8 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -42,18 +42,19 @@ method, this part of the documentation is for you.
4242
.. toctree::
4343
:maxdepth: 2
4444

45-
modules/layers
45+
modules/activation
46+
modules/array_ops
4647
modules/cost
47-
modules/prepro
48+
modules/distributed
49+
modules/files
4850
modules/iterate
49-
modules/utils
51+
modules/models
5052
modules/nlp
53+
modules/layers
54+
modules/prepro
5155
modules/rein
52-
modules/files
56+
modules/utils
5357
modules/visualize
54-
modules/activation
55-
modules/models
56-
modules/distributed
5758
modules/db
5859

5960

‎docs/modules/array_ops.rst‎

Lines changed: 20 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,20 @@
1+
API - Array Operations
2+
======================
3+
4+
.. automodule:: tensorlayer.array_ops
5+
6+
.. autosummary::
7+
8+
alphas
9+
alphas_like
10+
11+
Tensorflow Tensor Operations
12+
----------------------------
13+
14+
tl.alphas
15+
^^^^^^^^^
16+
.. autofunction:: alphas
17+
18+
tl.alphas_like
19+
^^^^^^^^^^^^^^
20+
.. autofunction:: alphas_like

‎tensorlayer/__init__.py‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -7,6 +7,7 @@
77
import tensorflow
88

99
from . import activation
10+
from .array_ops import alphas, alphas_like
1011
from . import cost
1112
from . import files
1213
from . import iterate

‎tensorlayer/array_ops.py‎

Lines changed: 116 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,116 @@
1+
#! /usr/bin/python
2+
# -*- coding: utf-8 -*-
3+
"""A file containing functions related to array manipulation."""
4+
5+
from tensorflow.python.eager import context
6+
from tensorflow.python.framework import constant_op
7+
from tensorflow.python.framework import dtypes
8+
from tensorflow.python.framework import ops
9+
from tensorflow.python.framework import tensor_shape
10+
from tensorflow.python.framework.constant_op import constant
11+
from tensorflow.python.framework.ops import convert_to_tensor
12+
from tensorflow.python.ops.array_ops import shape_internal
13+
from tensorflow.python.ops.gen_array_ops import fill
14+
from tensorflow.python.ops.gen_array_ops import reshape
15+
16+
__all__ = ['alphas', 'alphas_like']
17+
18+
19+
def alphas(shape, alpha_value, name=None):
20+
"""Creates a tensor with all elements set to `alpha_value`.
21+
This operation returns a tensor of type `dtype` with shape `shape` and all
22+
elements set to alpha.
23+
24+
Parameters
25+
----------
26+
shape: A list of integers, a tuple of integers, or a 1-D `Tensor` of type `int32`.
27+
The shape of the desired tensor
28+
alpha_value: `float32`, `float64`, `int8`, `uint8`, `int16`, `uint16`, int32`, `int64`
29+
The value used to fill the resulting `Tensor`.
30+
name: str
31+
A name for the operation (optional).
32+
33+
Returns
34+
-------
35+
A `Tensor` with all elements set to alpha.
36+
37+
Examples
38+
--------
39+
>>> tl.alphas([2, 3], tf.int32) # [[alpha, alpha, alpha], [alpha, alpha, alpha]]
40+
"""
41+
42+
with ops.name_scope(name, "alphas", [shape]) as name:
43+
44+
alpha_tensor = convert_to_tensor(alpha_value)
45+
alpha_dtype = dtypes.as_dtype(alpha_tensor.dtype).base_dtype
46+
47+
if not isinstance(shape, ops.Tensor):
48+
try:
49+
shape = constant_op._tensor_shape_tensor_conversion_function(tensor_shape.TensorShape(shape))
50+
except (TypeError, ValueError):
51+
shape = ops.convert_to_tensor(shape, dtype=dtypes.int32)
52+
53+
if not shape._shape_tuple():
54+
shape = reshape(shape, [-1]) # Ensure it's a vector
55+
56+
try:
57+
output = constant(alpha_value, shape=shape, dtype=alpha_dtype, name=name)
58+
59+
except (TypeError, ValueError):
60+
output = fill(shape, constant(alpha_value, dtype=alpha_dtype), name=name)
61+
62+
if output.dtype.base_dtype != alpha_dtype:
63+
raise AssertionError("Dtypes do not corresponds: %s and %s" % (output.dtype.base_dtype, alpha_dtype))
64+
65+
return output
66+
67+
68+
def alphas_like(tensor, alpha_value, name=None, optimize=True):
69+
"""Creates a tensor with all elements set to `alpha_value`.
70+
Given a single tensor (`tensor`), this operation returns a tensor of the same
71+
type and shape as `tensor` with all elements set to `alpha_value`.
72+
73+
Parameters
74+
----------
75+
tensor: tf.Tensor
76+
The Tensorflow Tensor that will be used as a template.
77+
alpha_value: `float32`, `float64`, `int8`, `uint8`, `int16`, `uint16`, int32`, `int64`
78+
The value used to fill the resulting `Tensor`.
79+
name: str
80+
A name for the operation (optional).
81+
optimize: bool
82+
if true, attempt to statically determine the shape of 'tensor' and encode it as a constant.
83+
84+
Returns
85+
-------
86+
A `Tensor` with all elements set to `alpha_value`.
87+
88+
Examples
89+
--------
90+
>>> tensor = tf.constant([[1, 2, 3], [4, 5, 6]])
91+
>>> tl.alphas_like(tensor, 0.5) # [[0.5, 0.5, 0.5], [0.5, 0.5, 0.5]]
92+
"""
93+
94+
with ops.name_scope(name, "alphas_like", [tensor]) as name:
95+
tensor = ops.convert_to_tensor(tensor, name="tensor")
96+
97+
if context.in_eager_mode(): #and dtype is not None and dtype != tensor.dtype:
98+
ret = alphas(shape_internal(tensor, optimize=optimize), alpha_value=alpha_value, name=name)
99+
100+
else: # if context.in_graph_mode():
101+
102+
# For now, variant types must be created via zeros_like; as we need to
103+
# pass the input variant object to the proper zeros callback.
104+
105+
if (optimize and tensor.shape.is_fully_defined()):
106+
# We can produce a zeros tensor independent of the value of 'tensor',
107+
# since the shape is known statically.
108+
ret = alphas(tensor.shape, alpha_value=alpha_value, name=name)
109+
110+
# elif dtype is not None and dtype != tensor.dtype and dtype != dtypes.variant:
111+
else:
112+
ret = alphas(shape_internal(tensor, optimize=optimize), alpha_value=alpha_value, name=name)
113+
114+
ret.set_shape(tensor.get_shape())
115+
116+
return ret

0 commit comments

Comments
 (0)