Timo Schick

I don't know how efficient RTX 3090's are, but with a single Nvidia Geforce 1080Ti, training PET (not iPET) with the default parameters is a matter of a few hours....

Hi @FlorisFok, do you have suggestions as to how this should be fixed?

Hi @luffycodes, the accuracy reported in the PET paper is exactly what you obtain using this library. You can check out details about the "How many data points is a...

Hi @Dxy-cpu, could you provide some further information? What command are you using to run PET? What dataset are you training on? The error seems to occur when the model...

Hi @sherlcok314159, could you specify which experiments exactly you want to reproduce (i.e., which paper and which datasets are you referring to)?

Ah, my bad - I thought we had included zero-shot in this script. Do you need to reproduce the exact results from our paper? Otherwise, I'd recommend to just use...

Hi @jackfeinmann5, the library currently doesn't support evaluation during training, but you can easily implement that yourself by calling the evaluation logic [around here](https://github.com/timoschick/pet/blob/c33d2089f16d52483a4359ba6d6482b874e08853/pet/wrapper.py#L311). If you do not want to...

Hi @super-buster, unfortunately, there aren't any arguments for that. However, iPET basically works by independently training many different models with each model having its own folder, and it detects when...

Hi @rajicon, sorry for the late reply! I recall that I added the concept of `OverwriteableEmbedding`s at the very end of this research project and didn't retrain a form-only model...

Hi @pribadihcr, yes, we definitely plan to release the example code. However, the code is still kind of messy and I am really busy right now so it might take...