| { | |
| "architectures": [ | |
| "Qwen3ForCausalLM" | |
| ], | |
| "model_type": "qwen3", | |
| "hidden_size": 1024, | |
| "intermediate_size": 3072, | |
| "num_hidden_layers": 28, | |
| "num_attention_heads": 16, | |
| "num_key_value_heads": 8, | |
| "head_dim": 128, | |
| "max_position_embeddings": 32768, | |
| "vocab_size": 151936, | |
| "torch_dtype": "bfloat16", | |
| "bos_token_id": 151643, | |
| "eos_token_id": 151645, | |
| "tie_word_embeddings": false | |
| } |