Fast T5 tokenization fails for models with additional special tokens not prefixed by extra_id_

[LoicDagnas]

System Info

• transformers version: 4.34.1
• Platform: Windows-10-10.0.22621-SP0
• Python version: 3.9.12
• Huggingface_hub version: 0.16.4
• Safetensors version: 0.3.2
• Accelerate version: not installed
• Accelerate config: not found
• PyTorch version (GPU?): 1.13.1+cpu (False)
• Tensorflow version (GPU?): 2.13.0 (False)
• Flax version (CPU?/GPU?/TPU?): not installed (NA)
• Jax version: not installed
• JaxLib version: not installed
• Using GPU in script?: no
• Using distributed or parallel set-up in script?: no

Who can help?

@ArthurZucker @amyeroberts

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

As described here, when I try to load this tokenizer

from transformers import AutoTokenizer

AutoTokenizer.from_pretrained("iarfmoose/t5-base-question-generator")

This exception is raised.

Expected behavior

The tokenizer should be loaded without error as it was for version < 4.34

[ArthurZucker]

Hey, the PR was merged and should have been part of the latest release. I no longer have this issue on main, nor on transformers == 4.35.0. It was not included in the patch as it was reported after 😉

[LoicDagnas]

It works with the latest release, thank you