{ | |
"model_type": "deberta-v2", | |
"architectures": [ | |
"DebertaV2ForMaskedLM" | |
], | |
"attention_probs_dropout_prob": 0.1, | |
"attention_head_size": 64, | |
"hidden_act": "gelu", | |
"hidden_dropout_prob": 0.1, | |
"hidden_size": 1536, | |
"initializer_range": 0.02, | |
"intermediate_size": 6144, | |
"max_position_embeddings": 512, | |
"relative_attention": true, | |
"position_buckets": 256, | |
"norm_rel_ebd": "layer_norm", | |
"share_att_key": true, | |
"pos_att_type": [ | |
"p2c", | |
"c2p" | |
], | |
"conv_kernel_size": 3, | |
"pooler_dropout": 0, | |
"pooler_hidden_act": "gelu", | |
"pooler_hidden_size": 1536, | |
"conv_act": "gelu", | |
"layer_norm_eps": 1e-7, | |
"max_relative_positions": -1, | |
"position_biased_input": false, | |
"num_attention_heads": 24, | |
"num_hidden_layers": 24, | |
"type_vocab_size": 0, | |
"num_labels": 119, | |
"vocab_size": 12800 | |
} |