MLproject

# Author: Luis Kuhn Cuellar
# Email: luis.kuhn@gmail.com
# Version: 1.0.0
# License: MIT


name: liver-ct-segmentation

# conda_env: environment.yml

docker_env:
  image: ghcr.io/mlf-core/liver_ct_segmentation:1.0.0
  volumes: ["${PWD}/data:/data"]
  environment: [["MLF_CORE_DOCKER_RUN", "TRUE"],["CUBLAS_WORKSPACE_CONFIG", ":4096:8"]]

entry_points:
  main:
    parameters:
      max_epochs: {type: int, default: 1000}
      gpus: {type: int, default: 2}
      accelerator: {type: string, default: "ddp"}
      general-seed: {type: int, default: 0}
      pytorch-seed: {type: int, default: 0}
      log-interval: {type: int, default: 3000}
      lr: {type: float, default: 0.0001}
      training-batch-size: {type: int, default: 1}
      test-batch-size: {type: int, default: 1}
      class-weights: {type: string, default: '0.2, 1.0, 2.5'}
      test-percent: {type: float, default: 0.15}
      test-epochs: {type: int, default: 10}
      dataset-path: {type: string, default: '/data/'}
      dataset-size: {type: int, default: 131}
      n-channels: {type: int, default: 1}
      n-class: {type: int, default: 3}
      num_workers: {type: int, default: 24}
      dropout-rate: {type: float, default: 0.25}
    command: |
          python liver_ct_segmentation/liver_ct_segmentation.py \
            --max_epochs {max_epochs} \
            --gpus {gpus} \
            --accelerator {accelerator} \
            --general-seed {general-seed} \
            --pytorch-seed {pytorch-seed} \
            --log-interval {log-interval} \
            --lr {lr} \
            --training-batch-size {training-batch-size} \
            --test-batch-size {test-batch-size} \
            --class-weights {class-weights} \
            --test-percent {test-percent} \
            --test-epochs {test-epochs} \
            --dataset-path {dataset-path} \
            --dataset-size {dataset-size} \
            --n-channels {n-channels} \
            --n-class {n-class} \
            --num_workers {num_workers} \
            --dropout-rate {dropout-rate}