1.1 [PyTorch] Intro
S - While leveraging power of NVIDIA GPUs for DL, need a high-level interface T - Built on CUDA, python framework Interface Type Contributor Intro Used in PyTorch DL framework Meta dynamic computing graph, easy to debug research TensorFlow DL framework Google support deployment Production Keras Highlevel API Independent first, then merged into TF easy to use for prototyping, default run on TF Education R Ultimately call highly optimized CUDA routines abstracts away the complexity of direct CUDA programming, allowing you to write intuitive Python code Close look into PyTorch Core components ...