HPLT

/

hplt_bert_base_ja

like 0

Follow

HPLT 36

And when using transformers>=4.40, two Metaspace's in tokenizer.json need "prepend_scheme" as follows:

      {
        "type": "Metaspace",
        "replacement": "▁",
        "add_prefix_space": false,
        "prepend_scheme": "never"
      },

Hi, thank you very much for reporting these issues! I will look more into it next week. We're still discussing what to do about the Metaspace pretokenizer, its new behavior might silently break more things: https://huggingface.co./HPLT/hplt_bert_base_en/discussions/1

Thank you @davda54 for new tokenizer.json with https://huggingface.co./HPLT/hplt_bert_base_ja/commit/3ba81b4d5b8885c06c3a0c8f4c7feb79fefee1cb , well, how about modeling_ltgbert.py?

Hi, I'm really sorry that it took me so long! Thank you once again for your fix, it's now applied to the Japanese BERT as well as to other HPLT-BERT models :)