task:
  name: "matching"
  type: "semantic_matching"

data:
  source: "huggingface"
  dataset_name: "sentence-transformers/paraphrase-MiniLM-L3-v2"
  input_field: "sentence1"
  target_field: "sentence2"
  label_field: "label"
  max_length: 128
  train_split: 0.8
  validation_split: 0.1
  test_split: 0.1

model:
  name: "sentence-transformers/all-MiniLM-L6-v2"
  max_length: 128

training:
  num_epochs: 3
  batch_size: 32
  learning_rate: 2e-5
  weight_decay: 0.01
  warmup_ratio: 0.1
  lr_scheduler_type: "linear"

inference:
  batch_size: 64
  similarity_threshold: 0.5