TensorLayerX

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 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 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 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 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 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 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 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.