Update TensorLayer (#1148)

Author: Laicheng0830

.circleci/config.yml

@@ -482,13 +482,11 @@ jobs:
           command: |
             echo "start tag workflow"
 
-###################################################################################
 ###################################################################################
 ###################################################################################
 #                               CircleCI WORKFLOWS                                #
 ###################################################################################
 ###################################################################################
-###################################################################################
 
 workflows:
   version: 2

.codacy.yaml

@@ -2,7 +2,7 @@
 ---
 engines:
   bandit:
-    enabled: false # FIXME: make it works
+    enabled: false
 exclude_paths:
 - scripts/*
 - setup.py

.dockerignore

@@ -1 +1,2 @@
+# dockerignore
 .git

.github/changelog.yml

@@ -18,6 +18,8 @@ include:
   - requirements_dev.txt
   - requirements_tf_cpu.txt
   - requirements_tf_gpu.txt
+  - requirements_doc.txt
+  - requirements_test.txt
 
   # Configuration Files
   - .travis.yml

.gitignore

@@ -26,7 +26,6 @@ wheels/
 MANIFEST
 *~
 
-# PyInstaller
 #  Usually these files are written by a python script from a template
 #  before PyInstaller builds the exe, so as to inject date/other infos into it.
 *.manifest

.pyup.yml

@@ -57,11 +57,6 @@ requirements:
     # Requirements for running unittests
     - requirements/requirements_test.txt
 
-# add a label to pull requests, default is not set
-# requires private repo permissions, even on public repos
-# default: empty
-#label_prs: update
-
 # configure the branch prefix the bot is using
 # default: pyup-
 branch_prefix: pyup-

.readthedocs.yml

@@ -1,7 +1,7 @@
 # https://docs.readthedocs.io/en/latest/yaml-config.html
 
 build:
-  image: latest  # For python 3.6
+  image: latest
 
 formats:
     - epub

CONTRIBUTING.md

@@ -140,17 +140,17 @@ make html
 # An example of a static model
 # A static model has inputs and outputs with fixed shape.
 inputs = tl.layers.Input([32, 784])
-dense1 = tl.layers.Dense(n_units=800, act=tf.nn.relu, in_channels=784, name='dense1')(inputs)
-dense2 = tl.layers.Dense(n_units=10,  act=tf.nn.relu, in_channels=800, name='dense2')(dense1)
+dense1 = tl.layers.Dense(n_units=800, act='relu', in_channels=784, name='dense1')(inputs)
+dense2 = tl.layers.Dense(n_units=10,  act='relu', in_channels=800, name='dense2')(dense1)
 model = tl.models.Model(inputs=inputs, outputs=dense2)
 
 # An example of a dynamic model
 # A dynamic model has more flexibility. The inputs and outputs may be different in different runs.
 class CustomizeModel(tl.models.Model):
     def __init__(self):
         super(CustomizeModel, self).__init__()
-        self.dense1 = tl.layers.Dense(n_units=800, act=tf.nn.relu, in_channels=784, name='dense1')
-        self.dense2 = tl.layers.Dense(n_units=10,  act=tf.nn.relu, in_channels=800, name='dense2')
+        self.dense1 = tl.layers.Dense(n_units=800, act='relu', in_channels=784, name='dense1')
+        self.dense2 = tl.layers.Dense(n_units=10,  act='relu', in_channels=800, name='dense2')
 
     # a dynamic model allows more flexibility by customising forwarding.
     def forward(self, x, bar=None):

README.md

@@ -22,7 +22,7 @@
 <!---  [![PyUP Updates](https://pyup.io/repos/github/tensorlayer/tensorlayer/shield.svg)](https://pyup.io/repos/github/tensorlayer/tensorlayer/) --->
 
 [TensorLayer](https://tensorlayer.readthedocs.io) is a novel TensorFlow-based deep learning and reinforcement learning library designed for researchers and engineers. It provides an extensive collection of customizable neural layers to build advanced AI models quickly, based on this, the community open-sourced mass [tutorials](https://github.com/tensorlayer/tensorlayer/blob/master/examples/reinforcement_learning/README.md) and [applications](https://github.com/tensorlayer). TensorLayer is awarded the 2017 Best Open Source Software by the [ACM Multimedia Society](https://twitter.com/ImperialDSI/status/923928895325442049). 
-This project can also be found at [iHub](https://code.ihub.org.cn/projects/328) and [Gitee](https://gitee.com/organizations/TensorLayer).
+This project can also be found at [OpenI](https://git.openi.org.cn/TensorLayer/tensorlayer3.0) and [Gitee](https://gitee.com/organizations/TensorLayer).
 
 # News
 

README.rst

@@ -107,7 +107,7 @@ Cite
 ====
 
 If you find this project useful, we would be grateful if you cite the
-TensorLayer paper：
+TensorLayer papers.
 
 ::
 
@@ -119,6 +119,14 @@ TensorLayer paper：
         year    = {2017}
     }
 
+    @inproceedings{tensorlayer2021,
+      title={Tensorlayer 3.0: A Deep Learning Library Compatible With Multiple Backends},
+      author={Lai, Cheng and Han, Jiarong and Dong, Hao},
+      booktitle={2021 IEEE International Conference on Multimedia \& Expo Workshops (ICMEW)},
+      pages={1--3},
+      year={2021},
+      organization={IEEE}
+
 License
 =======
 

docker/pypi_list.py

@@ -14,7 +14,6 @@
 
     args = parser.parse_args()
 
-    # create logger
     logger = logging.getLogger("PyPI_CLI")
 
     formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s')

docker/version_prefix.py

@@ -13,7 +13,6 @@
 
     args = parser.parse_args()
 
-    # create logger
     logger = logging.getLogger("VERSION_PREFIX_CLI")
 
     formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s')

docs/conf.py

@@ -34,12 +34,6 @@
 # extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
 # ones.
 
-# extensions = [
-#    'sphinx.ext.coverage',
-#    'sphinx.ext.githubpages',
-#    'numpydoc',
-# ]
-
 extensions = [
     'sphinx.ext.autodoc',
     'sphinx.ext.autosummary',

examples/basic_tutorials/tutorial_cifar10_cnn_static.py

@@ -136,21 +136,21 @@ def _map_fn_test(img, target):
 train_ds = tf.data.Dataset.from_generator(
     generator_train, output_types=(tf.float32, tf.int32)
 )  # , output_shapes=((24, 24, 3), (1)))
-train_ds = train_ds.map(_map_fn_train, num_parallel_calls=multiprocessing.cpu_count())
 # train_ds = train_ds.repeat(n_epoch)
 train_ds = train_ds.shuffle(shuffle_buffer_size)
 train_ds = train_ds.prefetch(buffer_size=4096)
 train_ds = train_ds.batch(batch_size)
+train_ds = train_ds.map(_map_fn_train, num_parallel_calls=multiprocessing.cpu_count())
 # value = train_ds.make_one_shot_iterator().get_next()
 
 test_ds = tf.data.Dataset.from_generator(
     generator_test, output_types=(tf.float32, tf.int32)
 )  # , output_shapes=((24, 24, 3), (1)))
 # test_ds = test_ds.shuffle(shuffle_buffer_size)
-test_ds = test_ds.map(_map_fn_test, num_parallel_calls=multiprocessing.cpu_count())
 # test_ds = test_ds.repeat(n_epoch)
 test_ds = test_ds.prefetch(buffer_size=4096)
 test_ds = test_ds.batch(batch_size)
+test_ds = test_ds.map(_map_fn_test, num_parallel_calls=multiprocessing.cpu_count())
 # value_test = test_ds.make_one_shot_iterator().get_next()
 
 for epoch in range(n_epoch):

examples/data_process/tutorial_fast_affine_transform.py

@@ -98,8 +98,8 @@ def _map_fn(image_path, target):
     dataset = tf.data.Dataset.from_generator(generator, output_types=(tf.string, tf.int64))
     dataset = dataset.shuffle(buffer_size=4096)  # shuffle before loading images
     dataset = dataset.repeat(n_epoch)
-    dataset = dataset.map(_map_fn, num_parallel_calls=multiprocessing.cpu_count())
     dataset = dataset.batch(batch_size)  # TODO: consider using tf.contrib.map_and_batch
+    dataset = dataset.map(_map_fn, num_parallel_calls=multiprocessing.cpu_count())
     dataset = dataset.prefetch(1)  # prefetch 1 batch
 
     n_step = 0

examples/data_process/tutorial_tf_dataset_voc.py

@@ -89,10 +89,10 @@ def _map_fn(filename, annotation):
 
 ds = tf.data.Dataset.from_generator(generator, output_types=(tf.string, tf.string))
 ds = ds.shuffle(shuffle_buffer_size)
-ds = ds.map(_map_fn, num_parallel_calls=multiprocessing.cpu_count())
 ds = ds.repeat(n_epoch)
 ds = ds.prefetch(buffer_size=2048)
 ds = ds.batch(batch_size)
+ds = ds.map(_map_fn, num_parallel_calls=multiprocessing.cpu_count())
 
 st = time.time()
 im, annbyte = next(iter(ds))

examples/data_process/tutorial_tfrecord.py

@@ -77,13 +77,11 @@
 def read_and_decode(filename):
     # generate a queue with a given file name
     raw_dataset = tf.data.TFRecordDataset([filename]).shuffle(1000).batch(4)
+    features = {}
     for serialized_example in raw_dataset:
-        features = tf.io.parse_example(
-            serialized_example, features={
-                'label': tf.io.FixedLenFeature([], tf.int64),
-                'img_raw': tf.io.FixedLenFeature([], tf.string),
-            }
-        )
+        features['label'] = tf.io.FixedLenFeature([], tf.int64)
+        features['img_raw'] = tf.io.FixedLenFeature([], tf.string)
+        features = tf.io.parse_example(serialized_example, features)
         # You can do more image distortion here for training data
         img_batch = tf.io.decode_raw(features['img_raw'], tf.uint8)
         img_batch = tf.reshape(img_batch, [4, 224, 224, 3])

examples/data_process/tutorial_tfrecord2.py

@@ -63,13 +63,11 @@
 def read_and_decode(filename):
     batchsize = 4
     raw_dataset = tf.data.TFRecordDataset([filename]).shuffle(1000).batch(batchsize)
+    features = {}
     for serialized_example in raw_dataset:
-        features = tf.io.parse_example(
-            serialized_example, features={
-                'label': tf.io.FixedLenFeature([], tf.int64),
-                'img_raw': tf.io.FixedLenFeature([], tf.string),
-            }
-        )
+        features['label'] = tf.io.FixedLenFeature([], tf.int64)
+        features['img_raw'] = tf.io.FixedLenFeature([], tf.string)
+        features = tf.io.parse_example(serialized_example, features)
         # You can do more image distortion here for training data
         img_batch = tf.io.decode_raw(features['img_raw'], tf.uint8)
         img_batch = tf.reshape(img_batch, [-1, 32, 32, 3])

examples/quantized_net/tutorial_binarynet_cifar10_tfrecord.py

@@ -161,21 +161,21 @@ def accuracy(_logits, y_batch):
 train_ds = tf.data.Dataset.from_generator(
     generator_train, output_types=(tf.float32, tf.int32)
 )  # , output_shapes=((24, 24, 3), (1)))
-train_ds = train_ds.map(_map_fn_train, num_parallel_calls=multiprocessing.cpu_count())
 # train_ds = train_ds.repeat(n_epoch)
 train_ds = train_ds.shuffle(shuffle_buffer_size)
 train_ds = train_ds.prefetch(buffer_size=4096)
 train_ds = train_ds.batch(batch_size)
+train_ds = train_ds.map(_map_fn_train, num_parallel_calls=multiprocessing.cpu_count())
 # value = train_ds.make_one_shot_iterator().get_next()
 
 test_ds = tf.data.Dataset.from_generator(
     generator_test, output_types=(tf.float32, tf.int32)
 )  # , output_shapes=((24, 24, 3), (1)))
 # test_ds = test_ds.shuffle(shuffle_buffer_size)
-test_ds = test_ds.map(_map_fn_test, num_parallel_calls=multiprocessing.cpu_count())
 # test_ds = test_ds.repeat(n_epoch)
 test_ds = test_ds.prefetch(buffer_size=4096)
 test_ds = test_ds.batch(batch_size)
+test_ds = test_ds.map(_map_fn_test, num_parallel_calls=multiprocessing.cpu_count())
 # value_test = test_ds.make_one_shot_iterator().get_next()
 
 for epoch in range(n_epoch):

examples/quantized_net/tutorial_dorefanet_cifar10_tfrecord.py

@@ -154,21 +154,21 @@ def accuracy(_logits, y_batch):
 train_ds = tf.data.Dataset.from_generator(
     generator_train, output_types=(tf.float32, tf.int32)
 )  # , output_shapes=((24, 24, 3), (1)))
-train_ds = train_ds.map(_map_fn_train, num_parallel_calls=multiprocessing.cpu_count())
 # train_ds = train_ds.repeat(n_epoch)
 train_ds = train_ds.shuffle(shuffle_buffer_size)
 train_ds = train_ds.prefetch(buffer_size=4096)
 train_ds = train_ds.batch(batch_size)
+train_ds = train_ds.map(_map_fn_train, num_parallel_calls=multiprocessing.cpu_count())
 # value = train_ds.make_one_shot_iterator().get_next()
 
 test_ds = tf.data.Dataset.from_generator(
     generator_test, output_types=(tf.float32, tf.int32)
 )  # , output_shapes=((24, 24, 3), (1)))
 # test_ds = test_ds.shuffle(shuffle_buffer_size)
-test_ds = test_ds.map(_map_fn_test, num_parallel_calls=multiprocessing.cpu_count())
 # test_ds = test_ds.repeat(n_epoch)
 test_ds = test_ds.prefetch(buffer_size=4096)
 test_ds = test_ds.batch(batch_size)
+test_ds = test_ds.map(_map_fn_test, num_parallel_calls=multiprocessing.cpu_count())
 # value_test = test_ds.make_one_shot_iterator().get_next()
 
 for epoch in range(n_epoch):

examples/quantized_net/tutorial_quanconv_cifar10.py

@@ -151,21 +151,21 @@ def accuracy(_logits, y_batch):
 train_ds = tf.data.Dataset.from_generator(
     generator_train, output_types=(tf.float32, tf.int32)
 )  # , output_shapes=((24, 24, 3), (1)))
-train_ds = train_ds.map(_map_fn_train, num_parallel_calls=multiprocessing.cpu_count())
 # train_ds = train_ds.repeat(n_epoch)
 train_ds = train_ds.shuffle(shuffle_buffer_size)
 train_ds = train_ds.prefetch(buffer_size=4096)
 train_ds = train_ds.batch(batch_size)
+train_ds = train_ds.map(_map_fn_train, num_parallel_calls=multiprocessing.cpu_count())
 # value = train_ds.make_one_shot_iterator().get_next()
 
 test_ds = tf.data.Dataset.from_generator(
     generator_test, output_types=(tf.float32, tf.int32)
 )  # , output_shapes=((24, 24, 3), (1)))
 # test_ds = test_ds.shuffle(shuffle_buffer_size)
-test_ds = test_ds.map(_map_fn_test, num_parallel_calls=multiprocessing.cpu_count())
 # test_ds = test_ds.repeat(n_epoch)
 test_ds = test_ds.prefetch(buffer_size=4096)
 test_ds = test_ds.batch(batch_size)
+test_ds = test_ds.map(_map_fn_test, num_parallel_calls=multiprocessing.cpu_count())
 # value_test = test_ds.make_one_shot_iterator().get_next()
 
 for epoch in range(n_epoch):

examples/quantized_net/tutorial_ternaryweight_cifar10_tfrecord.py

@@ -160,21 +160,21 @@ def accuracy(_logits, y_batch):
 train_ds = tf.data.Dataset.from_generator(
     generator_train, output_types=(tf.float32, tf.int32)
 )  # , output_shapes=((24, 24, 3), (1)))
-train_ds = train_ds.map(_map_fn_train, num_parallel_calls=multiprocessing.cpu_count())
 # train_ds = train_ds.repeat(n_epoch)
 train_ds = train_ds.shuffle(shuffle_buffer_size)
 train_ds = train_ds.prefetch(buffer_size=4096)
 train_ds = train_ds.batch(batch_size)
+train_ds = train_ds.map(_map_fn_train, num_parallel_calls=multiprocessing.cpu_count())
 # value = train_ds.make_one_shot_iterator().get_next()
 
 test_ds = tf.data.Dataset.from_generator(
     generator_test, output_types=(tf.float32, tf.int32)
 )  # , output_shapes=((24, 24, 3), (1)))
 # test_ds = test_ds.shuffle(shuffle_buffer_size)
-test_ds = test_ds.map(_map_fn_test, num_parallel_calls=multiprocessing.cpu_count())
 # test_ds = test_ds.repeat(n_epoch)
 test_ds = test_ds.prefetch(buffer_size=4096)
 test_ds = test_ds.batch(batch_size)
+test_ds = test_ds.map(_map_fn_test, num_parallel_calls=multiprocessing.cpu_count())
 # value_test = test_ds.make_one_shot_iterator().get_next()
 
 for epoch in range(n_epoch):

examples/reinforcement_learning/README.md

@@ -50,6 +50,19 @@ A corresponding [Springer textbook](https://deepreinforcementlearningbook.org) i
 
 `pip install --upgrade tf-nightly-2.0-preview tfp-nightly`
 
+## Quick Start
+```
+conda create --name tl python=3.6.4  
+conda activate tl
+pip install tensorflow-gpu==2.0.0-rc1 # if no GPU, use pip install tensorflow==2.0.0
+pip install tensorlayer
+pip install tensorflow-probability==0.9.0
+pip install gym
+pip install gym[atari] # for others, use pip instal gym[all]
+
+python tutorial_DDPG.py --train
+```
+
 ## Status: Beta
 
 We are currently open to any suggestions or pull requests from you to make the reinforcement learning tutorial with TensorLayer2.0 a better code repository for both new learners and senior researchers. Some of the algorithms mentioned in the this markdown may be not yet available, since we are still trying to implement more RL algorithms and optimize their performances. However, those algorithms listed above will come out in a few weeks, and the repository will keep updating more advanced RL algorithms in the future.

img/medium/Readme.md

@@ -15,5 +15,4 @@ It is 72px tall and can have a maximum width of 600px.
 
 ## Publication homepage images
 
-Under Homepage and settings >  Layout, you can select a header size, upload a logo and add a background image (large header size only).
-
+Under Homepage and settings >  Layout, you can select a header size, upload a logo and add a background image (large header size only).

requirements/requirements_tf_gpu.txt

@@ -1 +1 @@
-tensorflow-gpu>=2.0.0-alpha0
+tensorflow-gpu>=2.0.0-rc1

scripts/download_and_install_openmpi3_ubuntu.sh

@@ -23,7 +23,7 @@ URL=https://download.open-mpi.org/release/open-mpi/v${MPI_MAJOR}.${MPI_MINOR}/${
 tar -xf ${FILENAME}
 cd ${FOLDER}
 
-# will take about 8 min or longer depends on your machine
+# will take about 8 min or longer depends on your machine.
 ./configure --prefix=$HOME/local/openmpi
 make -j ${NPROC} all
 make install

scripts/install-requirements-for-rtd.sh

@@ -1,4 +1,5 @@
 # This script is for installing horovod on readthedocs only!
+
 set -e
 
 pwd

setup.cfg

@@ -30,21 +30,6 @@ column_limit=120
 each_dict_entry_on_separate_line = True
 
 # Put closing brackets on a separate line, dedented, if the bracketed
-# expression can't fit in a single line. Applies to all kinds of brackets,
-# including function definitions and calls. For example:
-#
-#   config = {
-#       'key1': 'value1',
-#       'key2': 'value2',
-#   }        # <--- this bracket is dedented and on a separate line
-#
-#   time_series = self.remote_client.query_entity_counters(
-#       entity='dev3246.region1',
-#       key='dns.query_latency_tcp',
-#       transform=Transformation.AVERAGE(window=timedelta(seconds=60)),
-#       start_ts=now()-timedelta(days=3),
-#       end_ts=now(),
-#   )        # <--- this bracket is dedented and on a separate line
 dedent_closing_brackets=True
 
 # Do not split consecutive brackets. Only relevant when DEDENT_CLOSING_BRACKETS is set