| { |
| "architectures": [ |
| "UltraBaseForCausalLM" |
| ], |
| "auto_map": { |
| "AutoConfig": "configuration_ultrabase.UltraBaseConfig", |
| "AutoModelForCausalLM": "modeling_ultrabase.UltraBaseForCausalLM" |
| }, |
| "bos_token_id": 0, |
| "bypass_rate": 0.375, |
| "d_ff": 256, |
| "d_model": 256, |
| "dtype": "float32", |
| "eos_token_id": 0, |
| "head_dim": 16, |
| "latent_dim": 64, |
| "model_type": "ultrabase", |
| "n_heads": 12, |
| "n_layers": 16, |
| "num_private_experts": 6, |
| "num_shared_experts": 1, |
| "tie_word_embeddings": true, |
| "transformers_version": "5.12.1", |
| "vocab_size": 49152 |
| } |
|
|