The Python Machine Learning Ecosystem

Whereas Scikit-learn is a great library for machine learning problems based on tabular data, it is not so well suited to handling the massive scale of data required for natural language or vision-based use cases. For these applications, deep learning is required. PyTorch provides functionality largely centred around building and training neural networks -- the backbone of deep learning. PyTorch offers scalable distributed training of models across single or multiple CPUs and GPUs. It also has an ecosystem of its own which provides integrations with Scikit-learn (skorch), model serving (TorchServe) and scheduling (AdaptDL).