| { | |
| "architectures": [ | |
| "LLaMAModel" | |
| ], | |
| "model_type": "passguess", | |
| "vocab_size": 500, | |
| "hidden_size": 384, | |
| "num_hidden_layers": 6, | |
| "num_attention_heads": 6, | |
| "max_position_embeddings": 24, | |
| "intermediate_size": 1024, | |
| "hidden_act": "silu", | |
| "rms_norm_eps": 1e-06, | |
| "rope_theta": 10000.0, | |
| "torch_dtype": "float16" | |
| } |