SindhiLM-Tokenizer-v1 / tokenizer_config.json
aakashMeghwar01's picture
Upgraded architecture: Unigram + Metaspace fix + 32k Vocab limit + Morpheme splitting
f82968f verified
Raw
History Blame Contribute Delete
246 Bytes
{
"backend": "tokenizers",
"bos_token": "<s>",
"eos_token": "</s>",
"mask_token": "<mask>",
"model_max_length": 1000000000000000019884624838656,
"pad_token": "<pad>",
"tokenizer_class": "TokenizersBackend",
"unk_token": "<unk>"
}