| { | |
| "architectures": ["Qwen3OmniForCausalLM"], | |
| "model_type": "qwen3-omni", | |
| "torch_dtype": "bfloat16", | |
| "_name_or_path": "Qwen/Qwen3-Omni-30B-A3B-Instruct", | |
| "hidden_size": 4096, | |
| "num_hidden_layers": 48, | |
| "num_attention_heads": 32, | |
| "intermediate_size": 16384, | |
| "moe": { | |
| "num_experts": 128, | |
| "num_experts_per_tok": 8 | |
| }, | |
| "quantization_config": { | |
| "quant_method": "custom_fp4", | |
| "bits": 4 | |
| }, | |
| "library_name": "fireecho" | |
| } |