Ethio-BBPE / config.json
Nexuss0781's picture
Production release: EthioBBPE tokenizer with perfect Amharic reconstruction
af2f42f verified
Raw
History Blame Contribute Delete
431 Bytes
{
"vocab_size": 16000,
"min_frequency": 2,
"show_progress": true,
"special_tokens": [
"<pad>",
"<unk>",
"<s>",
"</s>",
"<mask>"
],
"lowercase": false,
"dropout": null,
"data_dir": "./data",
"model_save_dir": "models",
"model_name": "EthioBBPE",
"use_checkpoint": true,
"checkpoint_dir": "./models/checkpoints",
"save_compressed": true,
"checkpoint_steps": null,
"num_threads": -1
}