{ "architectures": [ "Qwen2ForEnsemblePRM" ], "auto_map": { "AutoConfig": "configuration_qwen2.QwenEnPRMConfig", "AutoModel": "modeling_qwen2.Qwen2ForEnsemblePRM" }, "attention_dropout": 0.0, "bos_token_id": 151643, "eos_token_id": 151645, "hidden_act": "silu", "hidden_size": 3584, "initializer_range": 0.02, "intermediate_size": 18944, "learning_probability": 1.0, "max_position_embeddings": 4096, "max_window_layers": 28, "model_type": "qwen2", "num_attention_heads": 28, "num_ensemble": 32, "num_hidden_layers": 28, "num_key_value_heads": 4, "pad_token_id": 151645, "problem_type": "single_label_classification", "regularization_lambda": 0.5, "rms_norm_eps": 1e-05, "rope_theta": 10000.0, "rr_token": "", "rr_token_id": 151651, "sliding_window": null, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "transformers_version": "4.48.3", "use_cache": false, "use_mrope": false, "use_sliding_window": false, "vocab_size": 152064 }