activations: Tanh
batch_size: 8
class_identifier: regression_metric
dropout: 0.1
encoder_learning_rate: 1.0e-05
encoder_model: MiniLM
final_activation: null
hidden_sizes:
- 384
keep_embeddings_frozen: true
layer: 12
layerwise_decay: 0.95
learning_rate: 3.1e-05
load_weights_from_checkpoint: lightning_logs/cometinho_part-i/checkpoints/epoch=0-step=899999.ckpt
nr_frozen_epochs: 0.0
optimizer: AdamW
pool: avg
pretrained_model: microsoft/Multilingual-MiniLM-L12-H384
train_data: data/euro-distil.da.part_ii.csv
validation_data: data/2019-da-dev.csv