Less of an issue and more of a question, but I am wondering if it would make sense to move some of the existing codebase to PyTorch. Many modern "embedding" frameworks are already using the library and pre-trained models and embeddings are widely available. Would this be a useful undertaking?
Less of an issue and more of a question, but I am wondering if it would make sense to move some of the existing codebase to PyTorch. Many modern "embedding" frameworks are already using the library and pre-trained models and embeddings are widely available. Would this be a useful undertaking?