valohai.yaml

---

- step:
    name: tensorflow-check
    image: valohai/keras:2.1.3-tensorflow1.4.0-python3.5-cuda8.0-cudnn6-devel-ubuntu14.04
    command:
      - pwd
      - lsb_release -a
      - python --version
      - python -c "import keras; print(keras.__version__);"
      - python -c "import tensorflow; print(tensorflow.__version__);"
      - python -c "from tensorflow.python.client import device_lib; device_lib.list_local_devices();"
      - nvidia-smi
      - nvcc --version | grep release
      - cat /usr/include/x86_64-linux-gnu/cudnn_v*.h | grep CUDNN_MAJOR -A 2

- step:
    name: keras-dvc-cnn-simple
    image: valohai/keras:2.1.3-tensorflow1.4.0-python3.5-cuda8.0-cudnn6-devel-ubuntu14.04
    command:
      - tar -xf /valohai/inputs/dataset/*.tar -C /valohai/inputs/dataset/
      - python valohai/keras-dvc-cnn-simple.py {parameters}
    inputs:
      - name: dataset
    parameters:
      - name: epochs
        pass-as: --epochs={v}
        type: integer
        default: 20
      - name: inference
        pass-as: --inference={v}
        type: integer
        default: 0

- step:
    name: keras-20ng-cnn
    image: valohai/keras:2.1.3-tensorflow1.4.0-python3.5-cuda8.0-cudnn6-devel-ubuntu14.04
    command:
      - tar xf /valohai/inputs/dataset/glove6b100dtxt.tar.gz -C /valohai/inputs/dataset/
      - tar xf /valohai/inputs/dataset/news20.tar.gz -C /valohai/inputs/dataset/
      - python valohai/keras-20ng-cnn.py {parameters}
    inputs:
      - name: dataset
    parameters:
      - name: epochs
        pass-as: --epochs={v}
        type: integer
        default: 20
      - name: inference
        pass-as: --inference={v}
        type: integer
        default: 1

- step:
    name: keras-20ng-rnn
    image: valohai/keras:2.1.3-tensorflow1.4.0-python3.5-cuda8.0-cudnn6-devel-ubuntu14.04
    command:
      - tar xf /valohai/inputs/dataset/glove6b100dtxt.tar.gz -C /valohai/inputs/dataset/
      - tar xf /valohai/inputs/dataset/news20.tar.gz -C /valohai/inputs/dataset/
      - python valohai/keras-20ng-rnn.py {parameters}
    inputs:
      - name: dataset
    parameters:
      - name: epochs
        pass-as: --epochs={v}
        type: integer
        default: 20
      - name: inference
        pass-as: --inference={v}
        type: integer
        default: 1

- step:
    name: keras-dvc-cnn-pretrained
    image: valohai/keras:2.1.3-tensorflow1.4.0-python3.5-cuda8.0-cudnn6-devel-ubuntu14.04
    command:
      - tar xf /valohai/inputs/dataset/dogs-vs-cats.tar -C /valohai/inputs/dataset/
      - python valohai/keras-dvc-cnn-pretrained.py {parameters}
    inputs:
      - name: dataset
        default: https://object.pouta.csc.fi/swift/v1/AUTH_dac/mldata/dogs-vs-cats.tar
    parameters:
      - name: epochs
        pass-as: --epochs={v}
        type: integer
        default: 10

- step:
    name: keras-sfnet-cnn
    image: valohai/keras:2.1.3-tensorflow1.4.0-python3.5-cuda8.0-cudnn6-devel-ubuntu14.04
    command:
      - tar xf /valohai/inputs/dataset/*.tar.gz -C /valohai/inputs/dataset/
      - pip install tqdm
      - python valohai/keras-sfnet-cnn.py {parameters}
    inputs:
      - name: dataset
        default: https://object.pouta.csc.fi/swift/v1/AUTH_dac/mldata/sfnet2007-2008.tar.gz
      - name: embedding
        default: https://object.pouta.csc.fi/swift/v1/AUTH_dac/mldata/cc.fi.300.vec.gz

- step:
    name: keras-sfnet-lstm
    image: valohai/keras:2.1.3-tensorflow1.4.0-python3.5-cuda8.0-cudnn6-devel-ubuntu14.04
    command:
      - tar xf /valohai/inputs/dataset/*.tar.gz -C /valohai/inputs/dataset/
      - pip install tqdm
      - python valohai/keras-sfnet-lstm.py {parameters}
    inputs:
      - name: dataset
        default: https://object.pouta.csc.fi/swift/v1/AUTH_dac/mldata/sfnet2007-2008.tar.gz
      - name: embedding
        default: https://object.pouta.csc.fi/swift/v1/AUTH_dac/mldata/cc.fi.300.vec.gz

- step:
    name: pytorch_dvc_cnn_simple
    image: nvcr.io/nvidia/pytorch:19.01-py3
    command:
      - nvidia-smi
      - nvcc --version | grep release
      - cat /usr/include/x86_64-linux-gnu/cudnn_v*.h | grep CUDNN_MAJOR -A 2
      - pip install torchvision
      - tar xf /valohai/inputs/dataset/dogs-vs-cats.tar -C /valohai/inputs/dataset/
      - python valohai/pytorch_dvc_cnn_simple.py {parameters}
    inputs:
      - name: dataset
        default: https://object.pouta.csc.fi/swift/v1/AUTH_dac/mldata/dogs-vs-cats.tar
    parameters:
      - name: test
        pass-as: --test
        optional: True

- endpoint:
    name: predict-animal
    image: ufoym/deepo:all-py36-cpu
    wsgi: valohai.prediction_server:predict_wsgi
    files:
      - name: model
        path: model.h5

- endpoint:
    name: predict-nyyssi
    image: ufoym/deepo:all-py36-cpu
    wsgi: valohai.prediction_server_text:predict_wsgi
    files:
      - name: model
        path: model.h5
      - name: tokenizer
        path: tokenizer_sfnet.pkl
      - name: valohai-metadata.json
        path: valohai-metadata.json