| { | |
| "_name_or_path": "cycloneboy/chinese_mobilebert_base_f2", | |
| "architectures": [ | |
| "MobileBertForPreTraining" | |
| ], | |
| "attention_probs_dropout_prob": 0.1, | |
| "hidden_act": "relu", | |
| "hidden_dropout_prob": 0.0, | |
| "hidden_size": 768, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 512, | |
| "key_query_shared_bottleneck": true, | |
| "layer_norm_eps": 1e-12, | |
| "max_position_embeddings": 512, | |
| "num_attention_heads": 12, | |
| "num_hidden_layers": 12, | |
| "embedding_size": 768, | |
| "trigram_input": true, | |
| "use_bottleneck": true, | |
| "intra_bottleneck_size": 144, | |
| "use_bottleneck_attention": true, | |
| "num_feedforward_networks": 2, | |
| "normalization_type": "no_norm", | |
| "pad_token_id": 0, | |
| "model_type": "mobilebert", | |
| "transformers_version": "4.6.0.dev0", | |
| "classifier_activation": false, | |
| "type_vocab_size": 2, | |
| "vocab_size": 21128 | |
| } | |