Add BlenderBot small tokenizer to the init

[LysandreJik]

This class was forgotten, adding it to the init and the documentation.

[patil-suraj]

Thanks a lot for adding this !

[sgugger]

LGTM!

[LysandreJik]

Actually this tokenizer seems a bit broken:

tokenizer = AutoTokenizer.from_pretrained("facebook/blenderbot_small-90M")

Traceback (most recent call last):
  File "<input>", line 1, in <module>
  File "/home/lysandre/Workspaces/Python/transformers/src/transformers/models/auto/tokenization_auto.py", line 469, in from_pretrained
    return tokenizer_class_fast.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs)
  File "/home/lysandre/Workspaces/Python/transformers/src/transformers/tokenization_utils_base.py", line 1741, in from_pretrained
    return cls._from_pretrained(
  File "/home/lysandre/Workspaces/Python/transformers/src/transformers/tokenization_utils_base.py", line 1858, in _from_pretrained
    tokenizer = cls(*init_inputs, **init_kwargs)
  File "/home/lysandre/Workspaces/Python/transformers/src/transformers/models/blenderbot_small/tokenization_blenderbot_small_fast.py", line 76, in __init__
    ByteLevelBPETokenizer(
  File "/home/lysandre/transformers/.env/lib/python3.8/site-packages/tokenizers/implementations/byte_level_bpe.py", line 36, in __init__
    BPE(
Exception: Error while initializing BPE: Token `_</w>` out of vocabular

[LysandreJik]

Which only appears after fixing with 6d90d5a

cc @patil-suraj if I recall correctly you implemented this tokenizer, do you remember what might have gone wrong?