DynaBERT_MNLI / config.json
{
"attention_probs_dropout_prob": 0.1,
"finetuning_task": "mnli",
"hidden_act": "gelu",
"hidden_dropout_prob": 0.1,
"hidden_size": 768,
"initializer_range": 0.02,
"intermediate_size": 3072,
"layer_norm_eps": 1e-12,
"max_position_embeddings": 512,
"num_attention_heads": 12,
"num_hidden_layers": 12,
"num_labels": 3,
"output_attentions": true,
"output_hidden_states": true,
"output_intermediate": true,
"output_past": true,
"pruned_heads": {},
"torchscript": false,
"type_vocab_size": 2,
"use_bfloat16": false,
"vocab_size": 30522,
"model_type": "bert"
}