Training ShuffleNets with TPU
How to use this
I assume that you know how to use Google Cloud Platform.
- You need to prepare ImageNet data. Create a cheap compute instance and follow steps from here.
You will need to connect a disk with size ~400 GB to the instance. - Upload created data shards to a bucket.
- Create an instance with TPU and run
python train.py.
Credit
It is based on: https://github.com/tensorflow/tpu/tree/master/models/official/resnet