PyTorch Releases Major Update, Now Officially Supports Windows

PyTorch, an open source machine learning library for Python, today announced the release of PyTorch 0.4.0 with Windows support.

PyTorch can now be installed on Windows OS via Conda or Pip command line. The new version also merges Tensor and Variable, which means torch.autograd.Variable and torch.Tensor are now in the same class; and unifies the return 0-dimensional vector of size, which makes it more similar to NumPy features. Also, a set of more flexible context managers has replaced the volatile flag.

As a Facebook-backed open source package released in October 2016, PyTorch has been very well-received in the developer community, and has more than 14.4k stars on GitHub (Google-backed TensorFlow has 97.4K stars, and Amazon-backed Apache MXNet has 13.7K stars on GitHub). It can leverage the capability of GPU, speed up computing for AI tasks, provide GPU-friendly NumPy functions, and robustly support Tensor.

Detailed update content:

• Major Core Changes
  • Tensor / Variable merged
  • Zero-dimensional Tensors
  • dtypes
  • migration guide
• New Features
  • Tensors
    • Full support for advanced indexing
    • Fast Fourier Transforms
  • Neural Networks
    • Trade-off memory for compute
    • bottleneck – a tool to identify hotspots in your code
  • torch.distributions
    • 24 basic probability distributions
    • Added cdf, variance, entropy, perplexity etc.
  • Distributed Training
    • Launcher utility for ease of use
    • NCCL2 backend
  • C++ Extensions
  • Windows Support
  • ONNX Improvements
    • RNN support
• Performance Improvements
• Bug Fixes