AttributeError: Can't pickle local object 'main.<locals>.ConvertToMultiChannelBasedOnMartianClassesd'

[Harris91-lee]

Hi everyone, Hi @KumoLiu :)
Hope you are doing well?
For the first time im getting such issue (i didn't face that before) where it happens just before the training starts here :
for batch_data in train_loader:
step_start = time.time()
step += 1
inputs, labels = ( %% >> crushes here<<
batch_data["image"].to(device),
batch_data["label"].to(device),
)

the error comes as " AttributeError: Can't pickle local object 'main..ConvertToMultiChannelBasedOnMartianClassesd'"

where ConvertToMultiChannelBasedOnMartianClassesd is looking like this:

class ConvertToMultiChannelBasedOnMartianClassesd(MapTransform):
def call(self, data):
d = dict(data)
# d_ = dict(data)
for key in self.keys:
result = []

            # One-hot encoding for each labels
            for label_index in range(1, NUM_OF_LABELS + 1):
                result.append(d[key] == label_index)

            d[key] = np.stack(result, axis=0).astype(np.float32)
            # d[key] = np.stack(result, axis=0).astype(np.int8)
        return d

i wonder why this issue appears and how to solve it?