| activations: Tanh |
| batch_size: 16 |
| class_identifier: unified_metric |
| cross_entropy_weights: null |
| dropout: 0.1 |
| encoder_learning_rate: 1.0e-06 |
| encoder_model: XLM-RoBERTa |
| error_labels: |
| - minor |
| - major |
| final_activation: null |
| hidden_sizes: |
| - 3072 |
| - 1024 |
| input_segments: |
| - mt |
| - src |
| - ref |
| keep_embeddings_frozen: true |
| layer: mix |
| layer_norm: true |
| layer_transformation: sparsemax |
| layerwise_decay: 0.95 |
| learning_rate: 1.5e-05 |
| load_pretrained_weights: true |
| local_files_only: false |
| loss: mse |
| loss_lambda: 0.65 |
| nr_frozen_epochs: 0.3 |
| optimizer: AdamW |
| pool: avg |
| pretrained_model: microsoft/infoxlm-large |
| sent_layer: mix |
| train_data: |
| - data/model_rotation_ref_ablation_iteration/combined_training_data.csv |
| validation_data: |
| - data/model_rotation_ref_ablation_iteration/xcomet_Bengali-Hindi_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_English-Czech_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_English-German_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_English-Finnish_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_English-Latvian_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_English-Turkish_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_English-Russian_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_English-Chinese_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Czech-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_German-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Finnish-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Turkish-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Russian-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Chinese-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_English-Estonian_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Estonian-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_English-Lithuanian_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Lithuanian-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_French-German_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Gujarati-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Kazakh-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_German-Czech_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_German-French_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Pashto-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_English-Japanese_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Japanese-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Central Khmer-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Polish-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Tamil-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_English-Hausa_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Hausa-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Icelandic-English_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Bengali-Hindi_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Hindi-Bengali_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Xhosa-Zulu_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Zulu-Xhosa_test.csv |
| - data/model_rotation_ref_ablation_iteration/xcomet_Ukrainian-English_test.csv |
| warmup_steps: 0 |
| word_layer: 24 |
| word_level_training: false |
|
|
