| { | |
| "activate_func": "silu", | |
| "architectures": [ | |
| "MiniMindModelForCausalLM" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "minimind_model.MiniMindConfig", | |
| "AutoModel": "minimind_model.MiniMindModelForCausalLM", | |
| "AutoModelForCausalLM": "minimind_model.MiniMindModelForCausalLM" | |
| }, | |
| "dropout": 0, | |
| "dtype": "float32", | |
| "flash_attn": true, | |
| "hidden_size": 512, | |
| "intermediate_size": null, | |
| "kv_head": 2, | |
| "max_position_embeddings": 2048, | |
| "model_type": "minimind", | |
| "num_attention_heads": 8, | |
| "num_hidden_layers": 8, | |
| "q_dim": 64, | |
| "rms_eps": 1e-05, | |
| "rope_scaling": { | |
| "factor": 4, | |
| "original_max_position_embeddings": 512 | |
| }, | |
| "rope_theta": 10000.0, | |
| "rope_type": "yarn", | |
| "transformers_version": "4.57.1", | |
| "v_dim": 64, | |
| "vocab_size": 6400 | |
| } | |