| { | |
| "architectures": [ | |
| "LlamaForCausalLM" | |
| ], | |
| "hidden_size": 4096, | |
| "intermediate_size": 11008, | |
| "num_attention_heads": 32, | |
| "num_hidden_layers": 32, | |
| "hidden_act": "silu", | |
| "initializer_range": 0.02, | |
| "layer_norm_eps": 1e-5, | |
| "max_position_embeddings": 2048, | |
| "model_type": "llama", | |
| "torch_dtype": "float16", | |
| "transformers_version": "4.30.0", | |
| "use_cache": true, | |
| "vocab_size": 32000 | |
| } | |