TensorLayerX 0.5.8 Release

TensorLayerX 0.5.8 is a maintenance release .

It contains numerous bug fixes.

Feel free to use it and make suggestions!

TensorLayerX 0.5.7 Release

TensorLayerX 0.5.7 is a maintenance release . In this release , we have the following changes.

• Fix PyTorch back-end depthtospace operator.
• Fix where the training API could not accept multiple inputs.
• Add the example of importing trained models from PyTorch or Paddle to TensorLayerX.
• Add roll and logsoftmax operators.
• Update the model trained by any backend of TensorLayerX can be imported to any backend of TensorLayerX.

Feel free to use it and make suggestions!

TensorLayerX 0.5.6 Release

TensorLayerX 0.5.6 is a maintenance release . In this release , we have the following changes .

• Fixed Sequential mode ONNX node collection .
• Fixed bug with RNN LSTM GRU training parameters .
• Fixed the inconsistency of different backends parameters of DepthWiseConv2d.
• Fixed the bug of saving parameters to npz.
• Updated padding layers.

Feel free to use it and make suggestions!

TensorLayerX 0.5.5 Release

TensorLayerX 0.5.5 is a maintenance release.In this release, we have the following changes.

• Added get_device, to_device operator.
• Changed the parameter name of the average pooling layer to (AvgPool1d, GlobalAvgPool1d, AdaptiveAvgPool1d, AvgPool2d, GlobalAvgPool2d Etc.)
• Fixed LSTM RNN GRU.
• Fixed a bug where ParameterList and ParameterDict training parameters on the TensorFlow backend were not collected.
• Fixed support for MindSpore1.7.0 version.

Feel free to use it and make suggestions!

TensorLayerX 0.5.4 Release

TensorLayerX 0.5.4 is a maintenance release.In this release, we have the following changes.

• Added documentation for metric functions
• Add Einsum
• Fixed PyTorch back-end optimizers
• Fixed preprocessing when activation functions are used as parameters

Feel free to use it and make suggestions!

TensorLayerX 0.5.3 Release

TensorLayerX 0.5.3 is a maintenance release.In this release, we have the following changes.

• Added kernel_size, stride, dilation parameters can be int or tuple.
• Added padding mode can be int, tuple, or str. str is "SAME" or "VALID".
• Added TensorLayerX model topology for ONNX model export, can generate topology by model.build_graph(inputs).
• Fix the problem of slow training speed due to MindSpore optimizer wrapping.

Feel free to use it and make suggestions!

TensorLayerX 0.5.1 Release

TensorLayerX 0.5.1 is a maintenance release.In this release, we have updated a number of interface and parameter names to make them easier to use. Some examples have been improved.

Feel free to use it and make suggestions!

TensorLayerX 0.5.0 Release

TensorLayerX 0.5.0 is a maintenance release，it supports TensorFlow、MindSpore and PaddlePaddle backends, and supports some PyTorch operator backends, allowing users to run the code on different hardware like Nvidia-GPU and Huawei-Ascend. Feel free to use it and make suggestions.