| { | |
| "architectures": ["LDT"], | |
| "model_type": "ldt", | |
| "vocab_size": 12288, | |
| "hidden_size": 320, | |
| "num_hidden_layers": 5, | |
| "num_attention_heads": 5, | |
| "intermediate_size": 896, | |
| "max_position_embeddings": 512, | |
| "tie_word_embeddings": true, | |
| "dtype": "float32", | |
| "step": 30000, | |
| "best_val": 3.6892 | |
| } |