Closed ccmaymay closed 1 year ago
Difficult to tell how to merge this with fairscale, which manages its own parallelization using torch.distributed.
Does not have clear support for model parallelism of large models, and other software exists that appears better (and more explicitly) suited to the task.
https://pytorch.org/serve/ https://github.com/pytorch/serve/blob/master/examples/image_classifier/mnist/Docker.md https://github.com/pytorch/serve/tree/master/examples/Huggingface_Transformers https://github.com/pytorch/serve/issues/1579