How to use tpu strategy please help!

[h4ck4l1]

i came across this guide which mentions how to do hyperparameter tuning with distributed strategy.
when should i instantiate the tpu? in guide it was given to use different tuner id's for different workers is it the same for tpu?before running the bash file or inside the tuning file?. I tried just using tpu strategy without any bash file creation but it seems i get this

UnimplementedError: File system scheme '[local]' not implemented (file: '/content/mnist/mnist_tuner/trial_0002/checkpoint_temp/part-00000-of-00001')
	Encountered when executing an operation using EagerExecutor. This error cancels all future operations and poisons their output tensors.

I am also encountering a different problem, i get this warning

 WARNING:tensorflow:Callback method `on_train_batch_end` is slow compared to the batch time (batch time: 0.0016s vs `on_train_batch_end` time: 0.1302s). Check your callbacks