Why Torch dominates Tensorflow?
retained text
I have been used Tensorflow since its version 1.X (1.8 was my first tensorflow as i remember). So i started from graph based building modeling, staying as much away from 'python' as possible. Keras keeps its philosophy, even under different interfaces. At that time, torch was more python-oriented, not optimized for graph building, so i never considered using it; running epochs on python loop seems crazy in negative sense Recently i had to use torch, since huggingface ecosystem is mostly built on torch. I have no idea why torch have been taken over tensorflow. I know tensorflow now is not even a competitor to torch. Please give your opinions :)
document fetched 2026-10-31 00:24 · canonical https://reddit.com/r/deeplearning/comments/1xj7m4d/why_torch_dominates_tensorflow