byt5-small-japanese / config.json
sonoisa's picture
Add first version of ByT5 model and tokenizer pretrained on Japanese corpus
1e367fa
raw
history blame
625 Bytes
{
"_name_or_path": "/content/pretrained_model",
"architectures": [
"MT5ForConditionGeneration"
],
"d_ff": 3584,
"d_kv": 64,
"d_model": 1472,
"decoder_start_token_id": 0,
"dropout_rate": 0.0,
"eos_token_id": 1,
"feed_forward_proj": "gated-gelu",
"initializer_factor": 1.0,
"is_encoder_decoder": true,
"layer_norm_epsilon": 1e-06,
"model_type": "mt5",
"num_decoder_layers": 4,
"num_heads": 6,
"num_layers": 12,
"output_past": true,
"pad_token_id": 0,
"relative_attention_num_buckets": 32,
"tie_word_embeddings": false,
"tokenizer_class": "ByT5Tokenizer",
"vocab_size": 384
}