| { | |
| "model_type": "tiny-llama", | |
| "hidden_dim": 768, | |
| "n_heads": 12, | |
| "n_layer": 12, | |
| "intermediate_dim": 1024, | |
| "max_position_embedding": 256, | |
| "n_kv_heads": 4, | |
| "vocab_size": 32000, | |
| "padding_idx": 2, | |
| "rms_norm_eps": 1e-5, | |
| "attention_dropout": 0.1, | |
| "rope_theta": 10000 | |
| } | |