TypeError: string indices must be integers

I am using a custom batch function

def custom_prepare_batch(batch, device, non_blocking):
    x, y = batch["img"], batch["lab"]
    return (
        convert_tensor(x, device=device, non_blocking=non_blocking),
        convert_tensor(y, device=device, non_blocking=non_blocking),
    )

and my processing function is as follows:

def train_function(unused_engine, batch):
        images, targets = custom_prepare_batch(batch, device=device, non_blocking=True)
#         images, targets = batch["img"], batch["lab"]
    
        model.train()
        optimizer.zero_grad()

        y_preds = model(images)

        if agreement_threshold > 0.0:
            # The "batch_size" in this function refers to the batch size per env
            # Since we treat every example as one env, we should set the parameter
            # n_agreement_envs equal to batch size
            mean_loss, masks = and_mask_utils.get_grads(
                agreement_threshold=agreement_threshold,
                batch_size=1,
                loss_fn=criterion,
                n_agreement_envs=batch_size,
                params=optimizer.param_groups[0]['params'],
                output=y_preds,
                target=targets,
                method="and_mask",
                scale_grad_inverse_sparsity=scale_grad_inverse_sparsity,
            )
        else:
            mean_loss = criterion(y_pred, y)
            mean_loss.backward()

        optimizer.step()

        return y_preds, targets

and I am also passing the custom batch function as follows train_evaluator = create_supervised_evaluator(model, device=device, non_blocking=True, prepare_batch=custom_prepare_batch)

when I run my code

trainer_rng = np.random.RandomState()
trainer.run(next(iter(train_loader)), max_epochs=2,
                seed=trainer_rng.randint(2 ** 32))

I get the error

Current run is terminating due to exception: string indices must be integers.
Engine run is terminating due to exception: string indices must be integers.
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-16-099d5afd435f> in <module>
     32 
     33 trainer_rng = np.random.RandomState()
---> 34 trainer.run(next(iter(train_loader)), max_epochs=2,
     35                 seed=trainer_rng.randint(2 ** 32))

~/.conda/envs/mila/lib/python3.8/site-packages/ignite/engine/engine.py in run(self, data, max_epochs, epoch_length, seed)
    689 
    690         self.state.dataloader = data
--> 691         return self._internal_run()
    692 
    693     @staticmethod

~/.conda/envs/mila/lib/python3.8/site-packages/ignite/engine/engine.py in _internal_run(self)
    760             self._dataloader_iter = None
    761             self.logger.error("Engine run is terminating due to exception: %s.", str(e))
--> 762             self._handle_exception(e)
    763 
    764         self._dataloader_iter = None

~/.conda/envs/mila/lib/python3.8/site-packages/ignite/engine/engine.py in _handle_exception(self, e)
    465             self._fire_event(Events.EXCEPTION_RAISED, e)
    466         else:
--> 467             raise e
    468 
    469     @property

~/.conda/envs/mila/lib/python3.8/site-packages/ignite/engine/engine.py in _internal_run(self)
    728                     self._setup_engine()
    729 
--> 730                 time_taken = self._run_once_on_dataset()
    731                 # time is available for handlers but must be update after fire
    732                 self.state.times[Events.EPOCH_COMPLETED.name] = time_taken

~/.conda/envs/mila/lib/python3.8/site-packages/ignite/engine/engine.py in _run_once_on_dataset(self)
    826         except Exception as e:
    827             self.logger.error("Current run is terminating due to exception: %s.", str(e))
--> 828             self._handle_exception(e)
    829 
    830         return time.time() - start_time

~/.conda/envs/mila/lib/python3.8/site-packages/ignite/engine/engine.py in _handle_exception(self, e)
    465             self._fire_event(Events.EXCEPTION_RAISED, e)
    466         else:
--> 467             raise e
    468 
    469     @property

~/.conda/envs/mila/lib/python3.8/site-packages/ignite/engine/engine.py in _run_once_on_dataset(self)
    809                 self.state.iteration += 1
    810                 self._fire_event(Events.ITERATION_STARTED)
--> 811                 self.state.output = self._process_function(self, self.state.batch)
    812                 self._fire_event(Events.ITERATION_COMPLETED)
    813 

<ipython-input-12-0c14fe7efe33> in train_function(unused_engine, batch)
      1 def train_function(unused_engine, batch):
----> 2         images, targets = custom_prepare_batch(batch, device=device, non_blocking=True)
      3 #         images, targets = batch["img"], batch["lab"]
      4 
      5         model.train()

<ipython-input-11-206b9292729e> in custom_prepare_batch(batch, device, non_blocking)
      1 def custom_prepare_batch(batch, device, non_blocking):
----> 2     x, y = batch["img"], batch["lab"]
      3     return (
      4         convert_tensor(x, device=device, non_blocking=non_blocking),
      5         convert_tensor(y, device=device, non_blocking=non_blocking),

TypeError: string indices must be integers

Kindly, assist me. Thank you.

[etetteh]

Thank you. The code ran successfully after implementing your suggestions.

[vfdev-5]

I think the issue is related to the way you pass the data: next(iter(train_loader)) which looks a bit strange. Maybe, you can pass directly train_loader ?

Btw: looks like you are using v0.3.0, please think to migrate to v0.4.2