| ▲ | TorchTPU: Running PyTorch Natively on TPUs at Google Scale(developers.googleblog.com) | |
| 31 points by mji 4 hours ago | 3 comments | ||
| ▲ | aldielshala1 a few seconds ago | parent | next [-] | |
Looks like a new backend, similar to MPS for Apple Silicon? | ||
| ▲ | in-silico an hour ago | parent | prev | next [-] | |
This is great to see. I did trained some research models using the existing PyTorch/XLA on TPUs, and it was a mess of undocumented behavior and bugs (silently hanging after 8 hours of training!). If anyone is trying to use PyTorch on TPU before TorchTPU is released, you can check out the training pipeline that I ended up building to support my research: https://github.com/aklein4/easy-torch-tpu | ||
| ▲ | Reubend 37 minutes ago | parent | prev [-] | |
Sounds good, but my main question is: is this a fork, or a new backend they're building in (like MPS)? | ||