edit model hyperparameter configuration

[WidiaPrasetia]

Search before asking

• [X] I have searched the HUB issues and found no similar feature requests.

Description

After everything has been configured and then saved. maybe when the user wants to edit the train model configuration, it can be done. because there is no such menu feature. then in the custom features you might be able to add several other hyperparameters that can be customized, such as dropout for example and several other features that don't exist.

Use case

when you have finished configuring the train model and then save it. but there was something I forgot, then I wanted to edit it, for example patience which I should have configured to 10, but it was still 100 because I forgot to configure it. Well, this is where the configuration edit feature is needed

Additional

No response

[sergiuwaxmann]

@WidiaPrasetia Indeed, there isn't a way to update configuration after model creation but I added this request to our backlog. About extra parameters, isn't dropout available in the custom field under Advanced Model Configuration?

[WidiaPrasetia]

Thank you for responding to the issue I created.

Yes, there really isn't any dropout. does he have another name?

To make it clearer, I have attached the proof below.

Thank You Warm regards

[sergiuwaxmann]

@WidiaPrasetia Yes, I can confirm it is missing. Our team will introduce it in the next release. I will keep you updated.

[WidiaPrasetia]

thankyou sir for your response, good luck! I will always wait for the update

Oh ya, after the model training process is complete, after getting the visualization, can you also get the manual numbers? for example all history or train model details in excel form. thank you

[sergiuwaxmann]

@WidiaPrasetia dropout is available but only for the Classify task.

Dropout rate for regularization in classification tasks, preventing overfitting by randomly omitting units during training. (Read more here: https://docs.ultralytics.com/usage/cfg/?h=train+settings#train-settings)

Related to the Excel export - we don't have such option at the moment.

[WidiaPrasetia]

then for the detect task is there not yet?

[WidiaPrasetia]

Sorry, this deviates from the issue I created. can you help me in this case?

I have finished training the model with 100 epochs. but why does it still appear like this in the preview menu, which says model not trained.

and then the next problem is, when I refresh the browser on the train menu it appears "Training in progress. Checkpoint saved for epoch 27...". even though I have finished training on Google Colab.

please help, thank you

[sergiuwaxmann]

@WidiaPrasetia

No, the option is available only for the classification task as it does nothing when used with other tasks.

About your other issue, it looks like Colab didn't send the information properly to HUB... Unfortunately there is nothing you can do there (except starting the training again).

[WidiaPrasetia]

@sergiuwaxmann Sorry if my question deviated from the issue.

I have subscribed to Ultralytics Hub Pro, I think after subscribing I can immediately train the model without paying any more. Can I make a refund by canceling my subscription?

thankyou

[sergiuwaxmann]

@WidiaPrasetia I am not sure about a refund in this case as the information is clear in my opinion. For example, in our docs: https://docs.ultralytics.com/hub/pro https://docs.ultralytics.com/hub/cloud-training However, let me check internally and get back to you as soon as possible.

[WidiaPrasetia]

okay sir @sergiuwaxmann , I will wait for further information thank your for your response

[ruchitsolanki]

Hi @WidiaPrasetia, could you please share your HUB-associated email so we can look into the issue and take the necessary steps? We’ll review your case according to our internal policies and get back to you shortly.

[WidiaPrasetia]

@ruchitsolanki okay, this is my hub-associated email : citrabasegenep@gmail.com

[sergiuwaxmann]

@WidiaPrasetia Thank you!