| { |
| "architectures": [ |
| "AnubisMoeForCausalLM" |
| ], |
| "auto_map": { |
| "AutoConfig": "modeling_anubis_moe.AnubisMoeConfig", |
| "AutoModelForCausalLM": "modeling_anubis_moe.AnubisMoeForCausalLM" |
| }, |
| "model_type": "anubis_moe", |
| "vocab_size": 50257, |
| "context_length": 1024, |
| "embed_dim": 768, |
| "num_hidden_layers": 12, |
| "num_attention_heads": 12, |
| "drop_rate": 0.1, |
| "qkv_bias": false, |
| "num_experts": 8, |
| "top_k_experts": 2, |
| "expert_capacity": 0, |
| "torch_dtype": "float32", |
| "transformers_version": "4.56.1", |
| "is_decoder": true |
| } |
