https://cloud.google.com/tpu/docs/tutorials/kubernetes-engine-resnet https://cloud.google.com/kubernetes-engine
Minju:
- Edit
create_cluster.sh
CLUSTER_NAME - Edit
kub_job.yaml
TPU type + experiment name - Create a new experiment (experiment dims are hardcoded in prepare_experiments.py for now -- feel free to make this dynamic).
- Change out_dir in
jobs/pretrain_ilsvrc.sh