| { | |
| "architectures": [ | |
| "ViuAI", | |
| "SarusForCausalLM" | |
| ], | |
| "model_type": "sarus", | |
| "vocab_size": 64003, | |
| "hidden_size": 1280, | |
| "num_hidden_layers": 24, | |
| "num_attention_heads": 20, | |
| "num_key_value_heads": 4, | |
| "intermediate_size": 3456, | |
| "max_position_embeddings": 2048, | |
| "rope_theta": 10000.0, | |
| "rms_norm_eps": 1e-05, | |
| "bos_token_id": 1, | |
| "eos_token_id": 64002, | |
| "pad_token_id": 0, | |
| "torch_dtype": "bfloat16", | |
| "tie_word_embeddings": true | |
| } | |