| { | |
| "adaptive_routing": true, | |
| "architectures": [ | |
| "SlimMoEForCausalLM" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_slim_moe.SlimMoEConfig", | |
| "AutoModelForCausalLM": "modeling_slim_moe.SlimMoEForCausalLM" | |
| }, | |
| "bos_token_id": 2, | |
| "dim": 768, | |
| "dropout": 0.1, | |
| "dtype": "float32", | |
| "eos_token_id": 3, | |
| "hidden_dim": 1536, | |
| "max_seq_len": 2048, | |
| "model_type": "slim_moe", | |
| "num_experts": 4, | |
| "num_heads": 12, | |
| "num_hidden_layers": 16, | |
| "pad_token_id": 0, | |
| "transformers_version": "4.57.1", | |
| "vocab_size": 50257 | |
| } | |