2019-06-18 17:06:51 +08:00
|
|
|
{
|
2020-02-01 07:00:18 +08:00
|
|
|
"architectures": [
|
|
|
|
|
"BertForMaskedLM"
|
|
|
|
|
],
|
2019-06-18 17:06:51 +08:00
|
|
|
"attention_probs_dropout_prob": 0.1,
|
2021-04-23 17:35:35 +08:00
|
|
|
"gradient_checkpointing": false,
|
2019-06-18 17:06:51 +08:00
|
|
|
"hidden_act": "gelu",
|
|
|
|
|
"hidden_dropout_prob": 0.1,
|
|
|
|
|
"hidden_size": 768,
|
|
|
|
|
"initializer_range": 0.02,
|
|
|
|
|
"intermediate_size": 3072,
|
2020-04-24 23:55:13 +08:00
|
|
|
"layer_norm_eps": 1e-12,
|
2019-06-18 17:06:51 +08:00
|
|
|
"max_position_embeddings": 512,
|
2020-04-24 23:55:13 +08:00
|
|
|
"model_type": "bert",
|
2019-06-18 17:06:51 +08:00
|
|
|
"num_attention_heads": 12,
|
|
|
|
|
"num_hidden_layers": 12,
|
2020-04-24 23:55:13 +08:00
|
|
|
"pad_token_id": 0,
|
2021-04-23 17:35:35 +08:00
|
|
|
"position_embedding_type": "absolute",
|
|
|
|
|
"transformers_version": "4.6.0.dev0",
|
2019-06-18 17:06:51 +08:00
|
|
|
"type_vocab_size": 2,
|
2021-04-23 17:35:35 +08:00
|
|
|
"use_cache": true,
|
2019-06-18 17:06:51 +08:00
|
|
|
"vocab_size": 30522
|
|
|
|
|
}
|
