Arthur

@sayakpaul thanks for the context. I am guessing you are loading the `transformers` model through the `from_pretrained` api of `PreTrainedModel` which does not really leaves the freedom for that so...

I don't mind having a "feature" for that in transformers if this simplifies the way we load models and add some values overall! 🤗

Hey! Sorry not yet, it's pretty tricky, but I hope I'll resolve it soon! 🤗

Closing in favor of #20106. Thanks for your contribution

Cool let's just ask for a final review from @sgugger ! 🤗

Hey @stancld ! Thanks for the addition! There are a few approaches we can take here. Sometimes the tolerance is a bit too high and part of the hidden states...

Hey! So after investigating in detail, it seems that we indeed have problem, but the good new is that it is not a major issue. First, we have to use...

Yep, I will have a look asap

So! Sorry for the late reply. My first answer would be that the `attention_mask` and the inputs are different. - In the first case, you are feeding `[ 64, 275,...

Okay this fixes it for me : ```python from transformers import AutoModelForCausalLM, AutoTokenizer import torch model = AutoModelForCausalLM.from_pretrained('gpt2') tokenizer = AutoTokenizer.from_pretrained('gpt2', padding_side="left") tokenizer.pad_token = tokenizer.eos_token s = ["a b c",...