Leaving Meta and PyTorch

How to train a model on 10k H100 GPUs?

Growing open source from Torch to PyTorch