PyTorch 2.0 smoke test requirements #94914
Description
🐛 Describe the bug
PyTorch-2.0 is a big release with a lots of new feature, so we need to make sure that:
- CUDA
- pypi binaries with slimmed dependencies are usable in standard AWS containers (amazonlinux:2 regression in 1.13)
- pypi binaries with slimmed dependencies are usable with stock Ubuntu-20.04: torch-2.0.0-rc1 and torch-1.13.1 can not be installed on Ubuntu 20.04 #91067 (i can still repro it with RC1)
- PyTorch+CUDA11.8 has a working FFT on 4090
torch.compile- Basic test works (for example see test mentioned in Search for
libdevicerelative to shared library triton-lang/triton#1176 ) in PyTorch docker container torch.compileproduces a binary which can be used on 3090torch.compileraises an error if used on Windows
- Basic test works (for example see test mentioned in Search for
- Python-3.11
- Linux, Windows and MacOS(x86+arm64) builds are available for core and 4 domains (vision, audio, text, data)
torch.compileraises works for basic workflows or raises an error
- MPS
- Resnet is usable out of the box (i.e. https://github.com/pytorch/builder/blob/main/test/smoke_test/smoke_test.py passes for MPS device)
Versions
2.0