{"model_type": "bert", "hidden_activation": "ELU", "output_activation": "Softmax", "attention_dim": 128, "num_labels": 10, "loss_function": "CrossEntropyLoss", "optimizer": "AdamW", "learning_rate": 2e-05, "weight_decay": 0.01} |
{"model_type": "bert", "hidden_activation": "ELU", "output_activation": "Softmax", "attention_dim": 128, "num_labels": 10, "loss_function": "CrossEntropyLoss", "optimizer": "AdamW", "learning_rate": 2e-05, "weight_decay": 0.01} |