julfr

Hi, I tried fine-tuning first with `"anonymous-german-nlp/german-gpt2"`, and it worked well. But now with `"dbmdz/german-gpt2"` (and btw the same with `"stefan-it//german-gpt2-larger"`) i get a CUDA-error: ``` Num examples = 104...

Hello, thank you for your great project! I tried to fine-tune the model with my german text and noticed, that generated sentences are mostly copied from my training dataset. Could...