KeyError: 'callback_type' when using hyperparameter search on a deep learning model

Good afternoon,

I got an error when trying to use the hyperparameter search tool on a deep learning model created by using the tools: Create a deep learning model architecture and Create deep learning model. In the error message it says: KeyError: ‘callback_type’ (see below).

However when I look at the parameters of the deep learning model there is a callback_type specified (see below).So I do not understand why it gives the KeyError.

Any tips on how to solve this problem are welcome!


Look closer at that value in your second screenshot. The value is defaulting to None. That can be a valid input but for only one use case down in the Help section. Maybe that is a clue about what is actually missing?

It isn’ clear if you are following a GTN tutorial or not. If this is some external tutorial, maybe the command-line settings were not translated over to the the tool form in full, or there is some format problem with input data that doesn’t match what Galaxy is expecting (format, content).

Please clarify if you need more help. Tutorial link + screenshots of full settings (the Tool Parameter listing is right above where job logs are) and expanded inputs (to expose the datatype/metadata, and first few lines of the data). Troubleshooting errors

Let’s start there :slight_smile:

Thankyou for the response.

Even when I try out different callback types such as CSV_logger it gives the same error message. I am not following any tutorial but doing my own research using Galaxy. I added a full screenshot of the parameter settings below:

Using the same input but with a different model such as SVM is able to perform hyperparameter search, so I think it is not the input but something with the deep learning model specifications.

Ok, thanks for providing more details. The tool author @bjoern.gruening is one of the administrators for the server where you are working. Please submit a bug report directly from a red error dataset :beetle: and include in the comments your observations. You can also include a link to this topic for more context.