2019-08-28 12:31:11 +00:00
|
|
|
{
|
2020-04-24 15:58:01 +00:00
|
|
|
"activation": "gelu",
|
2020-01-31 23:00:24 +00:00
|
|
|
"architectures": [
|
|
|
|
"DistilBertForMaskedLM"
|
|
|
|
],
|
2019-09-19 15:58:21 +00:00
|
|
|
"attention_dropout": 0.1,
|
|
|
|
"dim": 768,
|
|
|
|
"dropout": 0.1,
|
|
|
|
"hidden_dim": 3072,
|
|
|
|
"initializer_range": 0.02,
|
|
|
|
"max_position_embeddings": 512,
|
2020-04-24 15:58:01 +00:00
|
|
|
"model_type": "distilbert",
|
2019-09-19 15:58:21 +00:00
|
|
|
"n_heads": 12,
|
|
|
|
"n_layers": 6,
|
2020-04-24 15:58:01 +00:00
|
|
|
"pad_token_id": 0,
|
2019-09-19 15:58:21 +00:00
|
|
|
"qa_dropout": 0.1,
|
|
|
|
"seq_classif_dropout": 0.2,
|
2019-09-24 08:13:06 +00:00
|
|
|
"sinusoidal_pos_embds": false,
|
2019-09-19 15:58:21 +00:00
|
|
|
"tie_weights_": true,
|
2021-08-29 21:44:39 +00:00
|
|
|
"transformers_version": "4.10.0.dev0",
|
2019-09-19 15:58:21 +00:00
|
|
|
"vocab_size": 30522
|
|
|
|
}
|