| { | |
| "_name_or_path": "/workspace/store/outputs/step3/B2048E3LR3e-05_big-prefix/B2048E3LR3e-05_mir_mr-_jqa_bao_qui_qui_mqa-prefix/93", | |
| "architectures": [ | |
| "RetrievaBertModel" | |
| ], | |
| "attention_probs_dropout_prob": 0.1, | |
| "auto_map": { | |
| "AutoConfig": "configuration_retrieva_bert.RetrievaBertConfig", | |
| "AutoModel": "retrieva_modeling.RetrievaBertModel" | |
| }, | |
| "hidden_act": "silu", | |
| "hidden_dropout_prob": 0.1, | |
| "hidden_size": 768, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 3072, | |
| "layer_norm_eps": 1e-12, | |
| "lm_head_hidden_act": "gelu", | |
| "max_position_embeddings": 1024, | |
| "mlp_bias": false, | |
| "model_type": "retrieva-bert", | |
| "num_attention_heads": 12, | |
| "num_hidden_layers": 12, | |
| "num_key_value_heads": 12, | |
| "pad_token_id": 0, | |
| "position_embedding_type": "rope", | |
| "rope_theta": 10000.0, | |
| "rotary_percent": 1.0, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.44.0", | |
| "type_vocab_size": 0, | |
| "use_cache": true, | |
| "vocab_size": 99584 | |
| } | |