In repoting columns is shuffle

[xeonvs]

I have an error in the report output and in the best parameters

r=ta.Reporting(h)
r.data
r.best_params('val_loss', [],ascending=True)[0]

some columns are shuffle together, of couse when scan is processed, params printing are ok. {'kernel_initializer': 'he_uniform', 'dropout': 0.1, 'second_neuron': 92, 'batch_size': 96, 'epochs': 30, 'lr': 0.001, 'first_neuron': 92, 'kernel_regularizer': 0.01, 'optimizer': 'Nadam'

My code:

p = {'first_neuron': [64, 92, 128],
     'second_neuron': [64, 92, 128],
     'batch_size': [64, 96, 128],
     'dropout': [0.0, 0.1, 0.2],
     'lr': [0.01, 0.02, 0.03, 0.001],
     'epochs': [30],
     'kernel_initializer': ['he_uniform', 'he_normal'],
     'kernel_regularizer': [0.1, 0.01],
     'optimizer': ['Adam', 'Nadam']
     }

h = ta.Scan(x=X_train, y=y_train, x_val=X_test, y_val=y_test,
            model=t_model_func,
            experiment_name='lstm_bidir',
            fraction_limit=0.002,
            params=p,
            print_params=True,
            seed=GLOBAL_SEED,
            reduction_method='correlation',
            reduction_interval=1,
            reduction_window=3,
            reduction_threshold=0.2,
            reduction_metric='val_loss',
            minimize_loss=True
            )

• [x] I'm up-to-date with the latest release:

  pip install -U git+git://github.com/autonomio/talos.git

• [x] I've confirmed that my Keras model works outside of Talos.

---

[mikkokotila]

What is your python version?

[xeonvs]

In colab.google Python 2.7.15+

[mikkokotila]

Talos does not support Python 2.7 as there seems to be no good reason to do it anymore. We used to do it but few months back it was given up. Do you have a particular reason for why you want to use it instead of python3?

Which Talos version you are using?

Up until Talos v0.5 the tests still incorporate Python2.7 and it passes (but all the new development is focused on v0.6 onwards and v0.5 will not be developed further, but there is support to fix bugs for at least some time still.

[xeonvs]

Do you have a particular reason for why you want to use it instead of python3?

Too many legacy code :(

Which Talos version you are using?

Latest from git.

Ok, I understand you, a reason to make refactoring to python3

[xeonvs]

In python3 all ok and you can close issue. Thsnks.

[mikkokotila]

That's great :)