| { | |
| "architectures": [ | |
| "GPTMiniForCausalLM" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_gptoss_mini.GPTMiniConfig", | |
| "AutoModelForCausalLM": "modeling_gptoss_mini.GPTMiniForCausalLM" | |
| }, | |
| "bos_token_id": 2, | |
| "dtype": "float32", | |
| "eos_token_id": 3, | |
| "hidden_size": 768, | |
| "intermediate_size": 3072, | |
| "max_position_embeddings": 512, | |
| "model_type": "gptoss-mini", | |
| "num_experts": 4, | |
| "num_heads": 8, | |
| "num_hidden_layers": 6, | |
| "num_layers": 6, | |
| "pad_token_id": 0, | |
| "top_k": 2, | |
| "transformers_version": "4.56.0", | |
| "use_cache": true, | |
| "vocab_size": 48000 | |
| } | |