controlmt-v2.3 / tokenizer_config.json
anandkaman's picture
Fix HF auto_map (add tokenizer_config.json + list form); honest style framing + add curated eval scores
37b4c07 verified
Raw
History Blame Contribute Delete
322 Bytes
{
"tokenizer_class": "ControlMTTokenizer",
"auto_map": {
"AutoTokenizer": ["tokenization_controlmt.ControlMTTokenizer", null]
},
"model_max_length": 320,
"bos_token": "<s>",
"eos_token": "</s>",
"unk_token": "<unk>",
"pad_token": "<pad>",
"clean_up_tokenization_spaces": false,
"use_fast": false
}