| { | |
| "architectures": ["NordicTranslatorForCausalLM"], | |
| "model_type": "nordic_translator", | |
| "auto_map": { | |
| "AutoConfig": "modeling_nordic.NordicTranslatorConfig", | |
| "AutoModelForCausalLM": "modeling_nordic.NordicTranslatorForCausalLM" | |
| }, | |
| "vocab_size": 65008, | |
| "hidden_dim": 2048, | |
| "num_layers": 18, | |
| "num_attention_heads": 16, | |
| "num_kv_heads": 4, | |
| "head_dim": 128, | |
| "ffn_intermediate": 6144, | |
| "rope_theta": 500000.0, | |
| "rope_partial_factor": 0.25, | |
| "rms_norm_eps": 1e-06, | |
| "max_position": 4096, | |
| "num_hidden_layers": 18, | |
| "hidden_size": 2048, | |
| "num_key_value_heads": 4, | |
| "max_position_embeddings": 4096, | |
| "tie_word_embeddings": false, | |
| "use_cache": false, | |
| "bos_token_id": 1, | |
| "eos_token_id": 2, | |
| "torch_dtype": "bfloat16" | |
| } | |