{ | |
"num_labels": 3, | |
"model_type": "deberta", | |
"attention_probs_dropout_prob": 0.1, | |
"hidden_act": "gelu", | |
"hidden_dropout_prob": 0.1, | |
"hidden_size": 1024, | |
"initializer_range": 0.02, | |
"intermediate_size": 4096, | |
"max_position_embeddings": 512, | |
"relative_attention": true, | |
"pos_att_type": "c2p|p2c", | |
"layer_norm_eps": 1e-7, | |
"max_relative_positions": -1, | |
"position_biased_input": false, | |
"num_attention_heads": 16, | |
"num_hidden_layers": 48, | |
"type_vocab_size": 0, | |
"pooling": { | |
"dropout": 0, | |
"hidden_act": "gelu" | |
}, | |
"vocab_size": 50265 | |
} | |