Omead Pooladzandi
Omead Pooladzandi
I mostly see fine-tuning code in those scripts. But there seems to be a general lack of pytorch xla examples which are full from scratch solutions.
I'm getting the same thing. pt-xla-profiler: Op(s) not lowered: aten::randperm.generator_out, Please open a GitHub issue with the above op lowering requests.
I am running into the same issue. Looks to be setting to PJRT mode by default. Not sure how to ask it to use the XRT mode.
I am also interested in extending this for use on a CNN. @sshkhr have you had any luck? I want to run it on a CNN with this architecture for...
Thank you for your reply @fmeirinhos . Yes, I actually followed that file and tried to extend it for CNNs to use with the Hv hessian method and fisher diag...
havent been able to get this working with docker if you can take a stab at it that would be great. as of now udocker is starting but cant see...
I was not able to get docker working properly on Google colab. I adjusted the notebook above. I am not too familiar with docker. If ur able to get it...
yea getting docker to work on colab seems to not be as straight forward. some recommend udocker but I havent been able to get either to work... Here is another...
can u post ur code for it? Also trying to re-produce rn. did you implement the updated version of BN?
Nope zero relation.