|
activations: Tanh |
|
batch_size: 32 |
|
class_identifier: regression_metric |
|
dropout: 0.1 |
|
encoder_learning_rate: 1.0e-06 |
|
encoder_model: XLM-RoBERTa |
|
final_activation: null |
|
hidden_sizes: |
|
- 3072 |
|
- 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: avg |
|
pretrained_model: xlm-roberta-large |
|
train_data: |
|
- data/model_rotation_ref/combined_training_data.csv |
|
validation_data: |
|
- data/model_rotation_ref/xcomet_Bengali-Hindi_test.csv |
|
- data/model_rotation_ref/xcomet_English-Czech_test.csv |
|
- data/model_rotation_ref/xcomet_English-German_test.csv |
|
- data/model_rotation_ref/xcomet_English-Finnish_test.csv |
|
- data/model_rotation_ref/xcomet_English-Latvian_test.csv |
|
- data/model_rotation_ref/xcomet_English-Turkish_test.csv |
|
- data/model_rotation_ref/xcomet_English-Russian_test.csv |
|
- data/model_rotation_ref/xcomet_English-Chinese_test.csv |
|
- data/model_rotation_ref/xcomet_Czech-English_test.csv |
|
- data/model_rotation_ref/xcomet_German-English_test.csv |
|
- data/model_rotation_ref/xcomet_Finnish-English_test.csv |
|
- data/model_rotation_ref/xcomet_Turkish-English_test.csv |
|
- data/model_rotation_ref/xcomet_Russian-English_test.csv |
|
- data/model_rotation_ref/xcomet_Chinese-English_test.csv |
|
- data/model_rotation_ref/xcomet_English-Estonian_test.csv |
|
- data/model_rotation_ref/xcomet_Estonian-English_test.csv |
|
- data/model_rotation_ref/xcomet_English-Lithuanian_test.csv |
|
- data/model_rotation_ref/xcomet_Lithuanian-English_test.csv |
|
- data/model_rotation_ref/xcomet_French-German_test.csv |
|
- data/model_rotation_ref/xcomet_Gujarati-English_test.csv |
|
- data/model_rotation_ref/xcomet_Kazakh-English_test.csv |
|
- data/model_rotation_ref/xcomet_German-Czech_test.csv |
|
- data/model_rotation_ref/xcomet_German-French_test.csv |
|
- data/model_rotation_ref/xcomet_Pashto-English_test.csv |
|
- data/model_rotation_ref/xcomet_English-Japanese_test.csv |
|
- data/model_rotation_ref/xcomet_Japanese-English_test.csv |
|
- data/model_rotation_ref/xcomet_Central Khmer-English_test.csv |
|
- data/model_rotation_ref/xcomet_Polish-English_test.csv |
|
- data/model_rotation_ref/xcomet_Tamil-English_test.csv |
|
- data/model_rotation_ref/xcomet_English-Hausa_test.csv |
|
- data/model_rotation_ref/xcomet_Hausa-English_test.csv |
|
- data/model_rotation_ref/xcomet_Icelandic-English_test.csv |
|
- data/model_rotation_ref/xcomet_Bengali-Hindi_test.csv |
|
- data/model_rotation_ref/xcomet_Hindi-Bengali_test.csv |
|
- data/model_rotation_ref/xcomet_Xhosa-Zulu_test.csv |
|
- data/model_rotation_ref/xcomet_Zulu-Xhosa_test.csv |
|
- data/model_rotation_ref/xcomet_Ukrainian-English_test.csv |
|
warmup_steps: 0 |
|
|