menu

Talos

Talos is a hyperparameter search and tuning solution for Keras users.

Channels
Team

Using talos on a Model with a list as training data

January 29, 2019 at 10:05am

Using talos on a Model with a list as training data

January 29, 2019 at 10:05am

Hello, first thank you for the great repo. I need to know if Talos accept a model with a training set as a list? Such as in this example

h=ta.Scan(

x=x_train,

y=y_train,

params=params,

model=NeuMf_model,

val_split=0.1,

experiment_no="1",

debug=True)

Where

x_train = [train.user_id, train.item_id]

An exceptions raises "only integer scalar arrays can be converted to a scalar index" no matter how I format the element the x_train list. Is there a particular way to tell Talos no to preprocess my input before feeding them to the model?


January 29, 2019 at 10:38am

You're most welcome :) Streamlining inputs of lists and dictionaries instead of 1d and 2d arrays is a high priority item at the moment, most likely there will be progress with this during this week still.

  • reply
  • like

Thanks for the prompt response, that's good to hear, looking forward to it.

  • reply
  • like

You can try now with @daily-dev. I've tested according to the scenario in #155 where the same x_train is provided as input in model.fit() in a list, and the case where Scan(x=[array, array]) and both work. The trick is that you have to declare Scan(....x_val, y_val...) as otherwise Talos will try to split for validation, and that will not work.

  • reply
  • like