In [1]: import transformers

In [2]: t0tt = transformers.AutoTokenizer.from_pretrained('bigscience/T0pp')
You are using the default legacy behaviour of the <class 'transformers.models.t5.tokenization_t5.T5Tokenizer'>. This is expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you. If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it means, and thouroughly read the reason why this was added as explained in https://github.com/huggingface/transformers/pull/24565

In [3]: t0tt.save_pretrained('saved-tokenizer')
Out[3]: 
('saved-tokenizer/tokenizer_config.json',
 'saved-tokenizer/special_tokens_map.json',
 'saved-tokenizer/spiece.model',
 'saved-tokenizer/added_tokens.json',
 'saved-tokenizer/tokenizer.json')

In [4]: loaded_t0tt = transformers.AutoTokenizer.from_pretrained('saved-tokenizer')
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.

In [6]: t0tt._eos_token
Out[6]: AddedToken("</s>", rstrip=True, lstrip=True, single_word=False, normalized=True, special=True)

In [7]: loaded_t0tt._eos_token
Out[7]: AddedToken("</s>", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True)

In [8]: t0tt.eos_token
Out[8]: '</s>'

In [9]: t0tt('hello </s>        goodbye')
Out[9]: {'input_ids': [21820, 1, 23281, 1], 'attention_mask': [1, 1, 1, 1]}

In [10]: loaded_t0tt('hello </s>        goodbye')
Out[10]: {'input_ids': [21820, 3, 1, 23281, 1], 'attention_mask': [1, 1, 1, 1, 1]}