docker-compose.yml

version: "3.8"

services:
  bert_classifier:
    container_name: bert_classifier
    image: heartexlabs/label-studio-ml-backend:bertclass-master
    build:
      context: .
      args:
        TEST_ENV: ${TEST_ENV}
    environment:
      # Essential Training Settings
      
      # Controls how much the model updates its weights in response to errors
      # - Lower (1e-5): More stable but slower learning
      # - Higher (3e-5): Faster learning but might be unstable
      # - Default (2e-5): Good balance for BERT fine-tuning
      - LEARNING_RATE=2e-5
      
      # Number of complete passes through the training data
      # - Lower (1): Faster training but might underfit
      # - Higher (3+): Better learning but might overfit
      # - Default (2): Good balance for small datasets
      - NUM_TRAIN_EPOCHS=3
      
      # Prevents model weights from growing too large (regularization)
      # - Lower (0.001): Less regularization, might overfit
      # - Higher (0.1): More regularization, might underfit
      # - Default (0.01): Standard value for BERT fine-tuning
      - WEIGHT_DECAY=0.01
      
      # Number of annotations before starting training
      # - Set to 1 for testing (train after each annotation)
      # - Recommended (5-10) for production (more stable training)
      - START_TRAINING_EACH_N_UPDATES=1
    ports:
      - "9090:9090"
    volumes:
      - "./data/server:/data"
      - "./data/.cache:/root/.cache"