| { | |
| "architectures": [ | |
| "MCGPTForCausalLM" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_mcgpt.MCGPTConfig", | |
| "AutoModelForCausalLM": "modeling_mcgpt.MCGPTForCausalLM" | |
| }, | |
| "model_type": "mcgpt", | |
| "vocab_size": 33152, | |
| "hidden_size": 256, | |
| "num_layers": 4, | |
| "num_hidden_layers": 4, | |
| "nhead": 8, | |
| "num_experts": 4, | |
| "max_position_embeddings": 512, | |
| "dropout": 0.1, | |
| "dtype": "float32", | |
| "transformers_version": "4.48.0", | |
| "torch_dtype": "float32" | |
| } | |