PechaBridgeOCR / repro /text_normalization.json
TibetanCodexAITeam's picture
Upload DONUT checkpoint checkpoint-154000 via PechaBridge
c122017 verified
Raw
History Blame Contribute Delete
446 Bytes
{
"unicode_normalization": "NFC",
"newline_normalization": "\\r\\n and \\r -> \\n",
"special_token_cleanup": "tokenizer.batch_decode(skip_special_tokens=True) plus explicit string removal of tokenizer.all_special_tokens",
"zero_width_removed": [
"\\u200b",
"\\u200c",
"\\u200d",
"\\ufeff"
],
"whitespace_normalization": "collapse horizontal spaces/tabs, trim around newlines, strip",
"metric_newline_token": "<NL>"
}