Releases: tensorlayer/TensorLayer
Maintenance Release 1.5.4
Maintain release, recommended update !
This is an incomplete list among the many features added.
-
Updates
- Soft dice, hard dice and IoU by @zsdonghao
- DeConv2d automatically compute batch size by @zsdonghao
�
-
New Examples
-
Documentation by @zsdonghao
Release 1.5.3
Recommended Update !
This is an incomplete list among the many features added.
-
News
- Release Flickr dataset loader, see tl.files.load_flickr25k_dataset() and tl.files.load_flickr1M_dataset().
- Functions to real one image and read a batch of images, see tl.vis.read_image() and tl.vis.read_images().
- A bunch of file and folder manage functions in tl.files.*.
-
Updates
- Automatically compute output channels and batch size in SpatialTransformer2dAffineLayer.
- Fixed the L2 part of cifar10 example with
tl.layers.get_variables_with_name
by @yzwxx - tutorial_tfrecord2.py save images instead of plot images.
-
New Examples
- Release Sub-pixel Convolution for Super-resolution see SRGAN code
-
Documentation by @whfairy2007 @angerhang
Release 1.5.2
Recommended Update !
This is an incomplete list among the many features added.
-
Support
- SpatialTransformer2dAffineLayer for Spatial Transformer Networks see example code by @zsdonghao
- AtrousConv1dLayer for WaveNet by @akaraspt
- Save network into dictionary via tl.files.save_npz_dict and tl.files.load_npz_dict @craigleehi
Release 1.5.1
Recommended Update !
This is an incomplete list among the many features added.
-
Support
- TensorFlow 1.2
- normalized mean square error via
tl.cost. normalized_mean_square_error()
by @nebulaV - save and load model parameters into a dictionary via
tl.files.save_npz_dict()
andtl.files.load_npz_dict()
by @craigleehi - save one image via
tl.vis. save_image()
by @zsdonghao
-
Update
batch_size
option fortl.utils.predict()
by @xjx0524tl.utils.fit
by @Nandayang
Release 1.5.0
Recommended Update !
This is an incomplete list among the many features added.
-
Support
- Database management system (alpha version) by @fangde
⚠️ - New image preprocessing functions
pt2map
,find_contours
,dilation
andbinary_dilation
by @maxmo2009
- Database management system (alpha version) by @fangde
-
Update
- New alias
tl.visualize
==tl.vis
- New alias
-
Example
Maintenance Release 1.4.5
Recommended Update !
This is an incomplete list among the many features added.
- Update
DynamicRNNLayer
andBiDynamicRNNLayer
for TF1.1 by @CTTC- fix
MeanPool3d
string format bug by @boscotsang - tutorials of VGG16 and VGG19 with simplified CNN API models by @zsdonghao
Maintenance Release 1.4.4
Recommended Update !
This is an incomplete list among the many features added.
-
Update
Conv2dLayer
withuse_cudnn_on_gpu
anddata_format
tutorial_ptb_lstm_state_is_tuple.py
addedbatch_size
parameter inloss_fn
by @dakuang- Arg seed for
DropoutLayer
&GaussianNoiseLayer
by @boscotsang - modify the function
binary_cross_entropy
in cost package by @DallasBuyer
-
Official Example
-
Maintain Documentation
Maintenance Release 1.4.3
Recommended Update !
This is an incomplete list among the many features added.
- Update
- update
DynamicRNNLayer
- example for
tl.utils.class_balancing_oversample
- windows user installation guide by @michuanhaohao
SubpixelLayer
withact
tl.rein.discount_episode_rewards
withmode=0
or1
- update
Release 1.4.2
Recommended Update ! Work well with TF 1.0.1 .
This is an incomplete list among the many features added.
We are also implementing a database managment system in tl.db.
-
New features
- Super-resolution
tl.layers. SubpixelConv2d
by @zsdonghao - save multi-images
tl.visualize.save_images
by @zsdonghao - get random integer
tl.utils.get_random_int
by @zsdonghao - support global flat and dict
tl.global_flat
,tl.global_dict
by @zsdonghao
- Super-resolution
-
Update
- MSE for 2D, 4D tensor and new input argument
is_mean
by @zsdonghao - fix
tl.cost.cross_entropy_seq
by @zsdonghao
- MSE for 2D, 4D tensor and new input argument
-
Documentation
- Installation Guide for Windows User by @michuanhaohao
- prepro.rst by @wangg12
Release 1.4.1
Implementing Database Management System.
This is an incomplete list among the many features added.
- New Features
- Implementing database management system by @fangde @zsdonghao
from tensorlayer.db import TensorDB
db = TensorDB(ip='localhost', port=27017, db_name='atari', user_name=None, password=None)
- Updates
- Basic Layer for name scope by @hanshengchiu @akaraspt
- PoolLayer documentation by @Kongsea