PyTorch is a popular open-source framework for machine learning that has been updated to version 1.2.0. The new release includes over 1900 fixes covering JIT, ONNX, distributed training modes, and performance improvements.
Some changes:
- New TorchScript API , rather than taking focus. easily convert nn.Module (including submodules and methods called in forward()) to ScriptModule.
- Full support for ONNX Opset versions 7(v1.2), 8(v1.3), 9(v1.4), and 10 (v1.5) has been added in collaboration with Microsoft. Additionally, users can now register their own symbols for exporting custom operations and specify dynamic input sizes during export.
- Support for tensorboard is no longer experimental.
- A nn.Transformer module has been added, based on the paper Attention Is All You Need.
- Numerous improvements in the C++ API.
Source: linux.org.ru
