pytorch/pytorch v2.6.0: PyTorch 2.6.0 Release
This release features multiple improvements for PT2: torch.compile can now be used with Python 3.13; new performance-related knob torch.compiler.setstance; several AOTInductor enhancements.
Key points
- If you build PyTorch extensions with custom C++ or CUDA extensions, please update these builds to use CXXABI=1 as well and report any issues you are seeing.
- Also in this release as an important security improvement measure we have changed the default value for weightsonly parameter of torch.load.
- This release is composed of 3892 commits from 520 contributors since PyTorch 2.5.
- As always, we encourage you to try these out and report any issues as we improve PyTorch.
Sources (1)
- [1]pytorch/pytorch v2.6.0: PyTorch 2.6.0 ReleaseGitHub: pytorch/pytorch · Jan 29, 05:18 PM
This release features multiple improvements for PT2: `torch.compile` can now be used with Python 3.13; new performance-related knob `torch.compiler.set_stance`; several AOTInductor enhancements.
If you build PyTorch extensions with custom C++ or CUDA extensions, please update these builds to use CXX_ABI=1 as well and report any issues you are seeing.
Extractive summary: sentences quoted from the sources.