| { | |
| "architectures": ["LlamaForCausalLM"], | |
| "auto_map": { | |
| "AutoTokenizer": "tokenization_kimi.TikTokenTokenizer" | |
| }, | |
| "bos_token_id": 1, | |
| "eos_token_id": 2, | |
| "hidden_size": 4096, | |
| "intermediate_size": 11008, | |
| "max_position_embeddings": 32768, | |
| "model_type": "llama", | |
| "num_attention_heads": 32, | |
| "num_hidden_layers": 32, | |
| "num_key_value_heads": 8, | |
| "pad_token_id": 0, | |
| "rope_theta": 10000.0, | |
| "rms_norm_eps": 1e-06, | |
| "tie_word_embeddings": false, | |
| "tokenizer_class": "TikTokenTokenizer", | |
| "torch_dtype": "bfloat16", | |
| "use_cache": true, | |
| "vocab_size": 32000 | |
| } | |