| { | |
| "_name_or_path": "mingru-shakespeare", | |
| "architectures": [ | |
| "MinGRUForCausalLM" | |
| ], | |
| "bos_token_id": 0, | |
| "classifier_dropout": 0.1, | |
| "dropout": 0.1, | |
| "eos_token_id": 0, | |
| "g_epsilon": 0.5, | |
| "hidden_size": 256, | |
| "initializer_range": 0.02, | |
| "layer_norm_epsilon": 1e-12, | |
| "model_type": "mingru", | |
| "num_hidden_layers": 3, | |
| "pad_token_id": 0, | |
| "pooling_activation": "silu", | |
| "pooling_type": "last", | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.46.0.dev0", | |
| "use_bias": false, | |
| "use_cache": true, | |
| "vocab_size": 256 | |
| } | |