kosmos_processor sorted_length =sorted([(idx, len(x)) for idx, x in enumerate(text_encoding.input_ids)]) wrong?

[YamingZhang]

https://github.com/huggingface/transformers/blob/514de24abfd4416aeba6a6455ad5920f57f3567d/src/transformers/models/kosmos2/processing_kosmos2.py#L214C17-L214C17

The sorted code lacks a sorting function and should be changed to
sorted_length =sorted([(idx, len(x)) for idx, x in enumerate(text_encoding.input_ids)], key=lambda x: x[1])
, is that correct?

[ArthurZucker]

Mmmm yeah seems like it. Would you like to open a PR?
fyi @ydshieh

[YamingZhang]

Sure, I'd be happy to help out

[ydshieh]

Hi @YamingZhang

the latest version is

                sorted_length = sorted(
                    [(idx, len(x)) for idx, x in enumerate(text_encoding.input_ids)], key=lambda x: x[-1]
                )

see #27323.

[YamingZhang]

Hi @YamingZhang

the latest version is

                sorted_length = sorted(
                    [(idx, len(x)) for idx, x in enumerate(text_encoding.input_ids)], key=lambda x: x[-1]
                )

see #27323.

Thank you very much.