Release 1.4.0

Recommended Update !

This is an incomplete list among the many features added.

• New features

  • EstimatorLayer accepts model_fn that described the model by @zsdonghao
  • OneHotLayer accepts indices output one-hot format by @zsdonghao

• Update

  • Conv1d and Conv1dLayer by @zsdonghao
  • InputLayer by @zsdonghao

• Documentation

  • @luomai @zsdonghao

Maintenance Release 1.3.11

Maintain release for old TF versions.

This is an incomplete list among the many features added.

• New support
  • TimeDistributedLayer applies a function to every timestep of the input. @zsdonghao
• Update
  • fix PReluLayer typo by @geometrikal
  • update BatchNormLayer for TF 0.12.1 @zsdonghao @nebulaV
• Documentation
  @zsdonghao

Release 1.3.10

Compatible with Keras !

• NEWS
  • TensorLayer is now compatible with Keras, see tutorial_keras.py
  • tl.layers.get_layers_with_name : get a list of tensor outputs from a network
• Update
  • tl.cost.sigmoid_cross_entropy for TF0.11
  • tl.file.load_and_assign_npz
  • log of tl.file.save_npz
  • EmbeddingAttentionSeq2seqWrapper for TF1.0 by @pnpnpn
  • tl.nlpVocabulary with pad_id
• Deprecate
  • tl.prepro.distorted_images and tl.prepro.crop_central_whiten_images, please use tf.map_fn or tl.prepro.threading_data instead
• Documentation
  • @anxingle

This version works well under TF 0.10, 0.11, 0.12 and 1.0 backend, feel free to report if you found a bug.
Many thanks!

Maintenance Release 1.3.8-9

Compatible with TF 1.0 and old versions.

• Update
  • tl.cost.cross_entropy : name is not required by TF0.11~0.12
  • tl.cost.cross_entropy_seq : automatically compute the batch_size and num_steps, and support TF1.0
  • tl.rein.cross_entropy_reward_loss for TF1.0
  • shorten the terminal logging e.g. tensorlayer --> [TL]
• Example
  • A simple web service - TensorFlask by JoelKronander
• Documentation
  • @zsdonghao
• TODO
  • debug tutorial_dynamic_rnn.py, tutorial_tfrecord3.py and tutorial_translation.py(the wrapper will be deprecated)

Thanks to @wagamamaz for the useful tricks to use TensorLayer

This version works well under TF 0.10, 0.11, 0.12 and 1.0 backend, feel free to report if you found a bug.
Many thanks!

Release 1.3.6-7 for TensorFlow 1.0

Compatible with TF 1.0 and old versions ! All examples are runnable now !

• Added features for TF 1.0
  • rewrite APIs by @zsdonghao
  • cross-entropy by @YinongLong
  • RNN layer by @ssdf93
  • tutorial_minist.py for windows by @selay01
  • test all examples by @zsdonghao
• TODO
  • debug tutorial_dynamic_rnn.py and tutorial_tfrecord3.py

Thanks to @wagamamaz for the useful tricks to use TensorLayer

This version works well under TF 0.10, 0.11, 0.12 and 1.0 backend, feel free to report if you found a bug.
Many thanks!

Maintain Release 1.3.5

Dear all,

As TF r1.0 change a lot, feel free to report bug and pust request.
This update is for TF r1.0 mainly.

This is an incomplete list among the many features added.

• For r1.0
  • update ConcatLayer by @zsdonghao
  • update cross_entropyby qq user 胜势待发
  • update cross_entropy_seq_with_mask by qq user 胜势待发
• Others
  • load and assign a npz model load_and_assign_npz by @zsdonghao
  • update Layer by @zsdonghao
  • hint to use simplified convolutional APIs by @zsdonghao

This version works well under TF 0.10, 0.11 and 0.12 backend, it also works well with TF 1.0 in most of the case, feel free to report if you found a bug. Many thanks!

Release 1.3.4 for TensorFlow r1.0

Update it if you want to use TensorFlow r1.0.

This is an incomplete list among the many features added.

• Update
  • all RNNLayers for TF1.0
  • DropoutLayer and GaussianNoiseLayer with is_train

This version works well under TF 0.10, 0.11 and 0.12 backend, it also works well with TF 1.0 in most of the case, feel free to report if you found a bug. Many thanks!

Release 1.3.3 for TensorFlow r1.0

Update it if you want to use TensorFlow r1.0.

This is an incomplete list among the many features added.

• Support
  • TensorBoard visualization with utils.fit by @JoelKronander
  • TensorFlow r1.0 by @zsdonghao
  • DownSampling2dLayer by @zsdonghao
• Update
  • nlp/creat_vocab, force 0 to be padding index by @zsdonghao
  • AtrousConv2dLayer change W to filters by @zsdonghao

This version works well under TF 0.10, 0.11 and 0.12 backend, it also works well with TF 1.0 in most of the case, feel free to report if you found a bug. Many thanks!

Release 1.3.2

Recommended Update !

This is an incomplete list among the many features added.

• Support
  • new function for file: maybe_download_or_extract by @JoelKronander
  • new layers: ExpandDimsLayer and TileLayer by @zsdonghao
  • new layer: Seq2Seq by @zsdonghao
  • new prepro functions for NLP: precess_sequences, sequences_add_start_id and sequences_get_mask by @zsdonghao
  • new prepro function for image drop by @zsdonghao
• Update
  • tutorial_cifar10_tfrecord.py for TF12 by @JoelKronander
  • read_words for Python3.5 by @SmartAILM
  • Add path argument for tl.file.load* by @boscotsang
• Call for Contribution
  • Seq2Seq: PeekySeq2Seq and AttentionSeq2Seq, click here
  • Import Caffe model into TensorLayer, click here
  • End-to-End Memory Network tutorial, click here
• New Examples
  • U-Net: Convolutional Networks for Image Segmentation
  • Text-to-Image Synthesis with GAN

This version works well under TF 0.10 and 0.11 backend, it also works well with TF 0.12 in most of the case, feel free to report if you found a bug. Many thanks!

Maintain Release 1.3.1

Recommended Update !

This is an incomplete list among the many features added.

• Update
  • deconv2d_bilinear_initalizer for SegNet by @JoelKronander
  • fix BiDynamicRNNLayer to avoid error in gradient updating by @akaitsuki-ii
  • update cost function for TF12 by @zsdonghao
  • update BatchNormLayer for TF12 by @zsdonghao

This version works well under TF 0.10 and 0.11 backend, it also works well with TF 0.12 in most of the case, feel free to report if you found a bug. Many thanks!