Raj Dabre

Hi, I have implemented the fast version of attention and have tested it. I will push it tomorrow or so.

Hi @TimDettmers in my latest test, it turns out that saving the model is the source of this issue. Specifically the error pops up when I run this: optim_state =...

Im not a 100% sure but this might be taken care of in pytorch 2.0.

How did you get torchdistx to work with PT1.13? Would be really helpful. Thanks.

Hi, My toolkit does not support barthez but feel free to make some changes to the batching logic and send a PR!

I have some free time today so if you need barthez support then please tell me the input output format for barthez and I'll implement it if you haven't already.

Hi, This is because you also have to add the right if else conditions in the batching logic. I have made changes to the batching logic and some crucial change...

Since you are only fine tuning a Bart model there's no need to train your own vocab and thus sentencepiece binary installation is unnecessary.

However should you need to install sentencepiece then you may install it in your ~/.local folder. To do so, you can specify a cmake install directory as in https://confluence.ecmwf.int/plugins/servlet/mobile?contentId=38076656#content/view/38076656

Hi, The error you get is related to sentencepiece_train aka spm_train and not YANMTT. The python wrapper calling spm_train will need the core sentencepiece library installed and running. Since this...