| { | |
| "architectures": [ | |
| "PreeForCausalLM" | |
| ], | |
| "bos_token_id": 0, | |
| "dtype": "float32", | |
| "eos_token_id": 2, | |
| "hidden_size": 1536, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 6144, | |
| "max_position_embeddings": 2048, | |
| "model_type": "pree", | |
| "num_attention_heads": 24, | |
| "num_hidden_layers": 24, | |
| "num_key_value_heads": 6, | |
| "pad_token_id": 1, | |
| "rms_norm_eps": 1e-06, | |
| "tie_word_embeddings": false, | |
| "transformers_version": "4.56.0", | |
| "use_cache": true, | |
| "vocab_size": 65536 | |
| } | |