nr_frozen_epochs: 0.3
class_identifier: referenceless_regression_metric
keep_embeddings_frozen: True
optimizer: AdamW
encoder_learning_rate: 1.0e-06
learning_rate: 1.5e-05
layerwise_decay: 0.95
encoder_model: XLM-RoBERTa
pretrained_model: Davlan/afro-xlmr-large
pool: avg
layer: mix
layer_transformation: sparsemax
layer_norm: False
loss: mse
dropout: 0.1
batch_size: 4
train_data: 
  - /SAN/intelsys/llm/jiaywang/COMET-2.0.2/data/annotation/wmt/train/wmt_final_previous.csv
validation_data:
  - /SAN/intelsys/llm/jiaywang/COMET-2.0.2/data/annotation/final/annotation_1027_dev_devtest_prep_ade/en-kik-ade_dev.csv
hidden_sizes:
  - 3072
  - 1024
activations: Tanh