How to switch data provider during training

In this example, we will see how one can easily switch the data provider during the training usingset_data().

Basic Setup

Required Dependencies

!pip install pytorch-ignite

Import

from ignite.engineimport Engine, Events

Data Providers

data1= [1,2,3]data2= [11,12,13]

Create dummytrainer

Let’s create a dummytrain_step which will print the current iteration and batch of data.

deftrain_step(engine,batch):print(f"Iter[{engine.state.iteration}] Current datapoint = ", batch)trainer=Engine(train_step)

Attach handler to switch data

Now we have to decide when to switch the data provider. It can be after an epoch, iteration or something custom. Below, we are going to switch data after some specific iteration. And then we attach a handler totrainer that will be executed once afterswitch_iteration and useset_data() so that when:

• iteration <=switch_iteration, batch is fromdata1
• iteration >switch_iteration, batch is fromdata2

switch_iteration=5@trainer.on(Events.ITERATION_COMPLETED(once=switch_iteration))defswitch_dataloader():print("<------- Switch Data ------->")    trainer.set_data(data2)

And finally we run thetrainer for some epochs.

trainer.run(data1,max_epochs=5)

Iter[1] Current datapoint =  1Iter[2] Current datapoint =  2Iter[3] Current datapoint =  3Iter[4] Current datapoint =  1Iter[5] Current datapoint =  2<------- Switch Data ------->Iter[6] Current datapoint =  11Iter[7] Current datapoint =  12Iter[8] Current datapoint =  13Iter[9] Current datapoint =  11Iter[10] Current datapoint =  12Iter[11] Current datapoint =  13Iter[12] Current datapoint =  11Iter[13] Current datapoint =  12Iter[14] Current datapoint =  13Iter[15] Current datapoint =  11State:iteration: 15epoch: 5epoch_length: 3max_epochs: 5output: <class 'NoneType'>batch: 11metrics: <class 'dict'>dataloader: <class 'list'>seed: <class 'NoneType'>times: <class 'dict'>