| { | |
| "architectures": [ | |
| "PhiForCausalLM" | |
| ], | |
| "model_type": "phi", | |
| "hidden_size": 4096, | |
| "num_hidden_layers": 32, | |
| "num_attention_heads": 32, | |
| "intermediate_size": 16384, | |
| "max_position_embeddings": 2048, | |
| "vocab_size": 50288, | |
| "bos_token_id": 0, | |
| "eos_token_id": 1, | |
| "pad_token_id": 50256, | |
| "attention_dropout": 0.1, | |
| "hidden_dropout": 0.1, | |
| "initializer_range": 0.02 | |
| } | |