| activations: Tanh | |
| batch_size: 32 | |
| class_identifier: earlyexitconfmulti_extra | |
| confidence_target: last | |
| dropout: 0.1 | |
| encoder_learning_rate: 1.0e-06 | |
| encoder_model: XLM-RoBERTa | |
| final_activation: null | |
| hidden_sizes: | |
| - 2048 | |
| - 1024 | |
| keep_embeddings_frozen: true | |
| layer: mix | |
| layer_norm: false | |
| layer_transformation: sparsemax | |
| layerwise_decay: 0.95 | |
| learning_rate: 1.5e-05 | |
| load_pretrained_weights: true | |
| local_files_only: false | |
| loss: mse | |
| nr_frozen_epochs: 0.3 | |
| optimizer: AdamW | |
| pool: cls | |
| pretrained_model: xlm-roberta-large | |
| train_data: | |
| - data/csv/train_da.csv | |
| validation_data: | |
| - data/csv/dev_da.csv | |
| warmup_steps: 0 | |