-
-
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Is there any reason you guys are using pyTorch over other frameworks?
-
"The main reason we chose PyTorch is to increase our research productivity at scale on GPUs. It is very easy to try and execute new research ideas in PyTorch; e.g., switching to PyTorch decreased our iteration time on research ideas in generative modeling from weeks to days."
- Još 1 odgovor
Novi razgovor -
-
-
Cool! I had it ported around the time it was released https://github.com/kashif/firedup Happy to see the switch!
-
Indeed---it was noticed! Thank you for your contribution to the Spinning Up resource ecosystem!! :) https://spinningup.openai.com/en/latest/user/introduction.html#long-term-support-and-support-history …
Kraj razgovora
Novi razgovor -
-
-
Can you explain how, exactly, this change increases research productivity? Like, is it that PyTorch is easier to use at a high level? Or is it that PyTorch is easier to modify and extend at a low level?
-
I can see if it's the latter, but a lot of people are pointing to how "readable" PyTorch is as its biggest benefit. TF2 reads just as easily (imo) so I'm guessing there's a difference between what the average Twitter user sees as "easy to use" versus what a researcher sees.
- Još 1 odgovor
Novi razgovor -
-
-
What about your baselines repository, will this switch to PyTorch as well?
-
Or give Stable.Baselines some support -https://mobile.twitter.com/araffin2/status/1223310856471138306 …
Kraj razgovora
Novi razgovor -
-
-
which framework was used before
-
Mostly TensorFlow-based I presume
- Još 1 odgovor
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.