"The war between ML frameworks has raged on since the rebirth of deep learning." That's a bit too dramatic though, no?https://twitter.com/gradientpub/status/1182327049307844613 …
-
Can you explain why you think its cherrypicked? I "cherry picked" the conference list, but I believe it should roughly match consensus. If you have other conferences you think I should add to the figures I'd be glad to do so.
As for why it's cherrypicked, you cherrypicked NLP a handful of conferences (where PT usage is higher), you miscounted mentions (Keras should be counted as part of TF), and you misstated the results. Also, very small sample size...
-
-
1. I showed results for NLP (NAACL, ACL, EMNLP), CV (CVPR, ICCV, ECCV), and ML (NeurIPS, ICML, ICLR). PyTorch usage is ~70% at both CVPR and the NLP conferences. It's >50% at ICML/ICLR.
-
2. I didn't include Keras because a. it doesn't change the results much, and b. some people do still use Keras with other backends. I have the figures in this ablation section for how including Keras affects the figures: https://thegradient.pub/p/cef6dd26-f952-4265-a2bc-f8bfb9eb1efb/ …
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.