| { | |
| "_name_or_path": "models/fleonce/dnrti-t5-base", | |
| "activation_fn": "relu", | |
| "architectures": [ | |
| "ITER" | |
| ], | |
| "d_ff": 3072, | |
| "d_model": 768, | |
| "dataset": "custom", | |
| "dropout": 0.3, | |
| "features": 2048, | |
| "max_length": 512, | |
| "max_nest_depth": 1, | |
| "model_type": "iter", | |
| "num_links": 0, | |
| "num_types": 13, | |
| "threshold": 0.5, | |
| "torch_dtype": "float32", | |
| "transformer_config": { | |
| "_name_or_path": "google/t5-efficient-base", | |
| "architectures": [ | |
| "T5ForConditionalGeneration" | |
| ], | |
| "d_ff": 3072, | |
| "d_model": 768, | |
| "decoder_start_token_id": 0, | |
| "max_length": 512, | |
| "model_type": "t5", | |
| "n_positions": 512, | |
| "num_decoder_layers": 12, | |
| "num_heads": 12, | |
| "num_layers": 12, | |
| "task_specific_params": null, | |
| "torch_dtype": "float32" | |
| }, | |
| "transformers_version": "4.41.2", | |
| "use_bias": false, | |
| "use_gate": true, | |
| "use_mlp": true, | |
| "use_scale": true | |
| } | |