Yet people here praise Torch when Tensorflow equivalents are often faster in production. Tensorflow still has relevance and gets a bit too much hate here (and I personally prefer pytorch).