PyTorch, Facebook’s open-source deep-learning framework, announced the release of version 1.6 which includes new APIs and performance improvements. Along with the release, Microsoft announced it will take over development and maintenance of the Windows version of the framework.
In a recent blog post, the PyTorch team highlighted the major features of the release. Native automatic mixed precision (AMP) training can produce memory savings of up to 50% in Tensor Core GPUs and a beta-version of new memory profiler assists in debugging memory issues.
