resnet_app.yaml

name: resnet-app

resources:
  cloud: aws
  accelerators:
    V100: 1

inputs: {
  gs://cloud-tpu-test-dataset/fake_imagenet: 70,
}

outputs: {
  resnet-model-dir: 0.1,
}

# file_mounts: {
#   /tmp/fake_imagenet: gs://cloud-tpu-test-datasets/fake_imagenet,
# }

setup: |
  git clone https://github.com/concretevitamin/tpu || true
  cd tpu
  git checkout 9459fee

  . $(conda info --base)/etc/profile.d/conda.sh
  pip install --upgrade pip

  conda activate resnet

  if [ $? -eq 0 ]; then
    echo "conda env exists"
  else
    conda create -n resnet python=3.7 -y
    conda activate resnet
    conda install cudatoolkit=11.0 -y
    pip install tensorflow==2.4.0 pyyaml
    pip install protobuf==3.20
    cd models
    pip install -e .
  fi

run: |
  cd tpu
  . $(conda info --base)/etc/profile.d/conda.sh
  conda activate resnet

  export XLA_FLAGS='--xla_gpu_cuda_data_dir=/usr/local/cuda/'
  python -u models/official/resnet/resnet_main.py --use_tpu=False \
      --mode=train --train_batch_size=256 --train_steps=250 \
      --iterations_per_loop=125 \
      --data_dir=gs://cloud-tpu-test-datasets/fake_imagenet \
      --model_dir=resnet-model-dir \
      --amp --xla --loss_scale=128