Use of standardScaler() and TPOT

It is good practice to first split the data into train and test, then scale by `fit_transform` on the training set, and `transform` on the test set.

So I create a train test split:
```python
training_features, testing_features, training_target, testing_target = \
train_test_split(data, labels, random_state=1,stratify=labels.values,test_size = 0.2)
```
Scale the data
```python
sc = StandardScaler()
training_features_scaled = sc.fit_transform(training_features)
testing_features_scaled = sc.transform(testing_features)
```

Now I feed `training_features_scaled` into `TPOT`.

However, TPOT then runs cross fold validation on this dataset, which bascially means that in this case, scaling has been done before the split and there's a potential for leakage between train and test during each fold. Is this something that we should avoid, and if so, how can we avoid this?