| { | |
| "architectures": [ | |
| "GIVT" | |
| ], | |
| "block_size": 60, | |
| "eps": 1e-08, | |
| "head_dim": 192, | |
| "hidden_dim": 1536, | |
| "input_dim": 1, | |
| "intermediate_dim": 4096, | |
| "model_type": "givt", | |
| "num_heads": 8, | |
| "num_layers": 16, | |
| "rope_base": 1000000, | |
| "rope_condense_ratio": 4.0, | |
| "rope_n_elem": 192, | |
| "rope_rotary_percentage": 1.0, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.36.2" | |
| } | |