None values passed as strings when tuning YOlO8 model hyper-parameters

[yashtiwari96]

Describe the bug

A clear and concise description of what the bug is.
I have trained a simple ultralytics yolov8 model using coco128 model. I am trying to tune this hyper parameters using the script provided by clearml. Some of these hyper-parameters are None values and they need to stay like that. However, clearml is injecting this as string type "None" which is causing type error.

To reproduce

Exact steps to reproduce the bug. Provide example code if possible.

Expected behaviour

What is the expected behaviour? What should've happened but didn't?
The automated tuning process should pas None as None but not as a string type

Environment

• Server type (self hosted \ app.clear.ml)
• ClearML SDK Version
• ClearML Server Version (Only for self hosted). Can be found on the bottom right corner of the settings screen.
• Python Version
• OS (Windows \ Linux \ Macos)

Related Discussion

If this continues a slack thread, please provide a link to the original slack thread.

[eugen-ajechiloae-clearml]

Hi @yashtiwari96 ! What clearml version are you using? Also, how exactly did you obtain the base task? Could you also screenshot the configuration section of the base task and hover over the type of line_width?

[yashtiwari96]

I have updated my clearml server yesterday to check if the issue was resolved but there is its still the same. From my understanding the problem is in loading parameters when tuning the model. the None values are converted to empty strings '' which is causing this issue.
I have verified this by cloning a task and fetching the parameters by using cloned_task.get_parameters() function. On converting the empty string to None my program is working fine. However, running a manual hyper-parameter search has a lot of disadvantages and I am not able to utilize the auto hyper-parameter optimization features

[eugen-ajechiloae-clearml]

@yashtiwari96 Note that task.get_parameters will cast all parameters to string automatically. If you wish to cast them to the right type, use cast=True: task.get_parameters(cast=True)

[yashtiwari96]

I am using the hyperparameter optimization from clearml.automation where it automatically creates tasks and fetches parameters from the base task id provided. Thus, I can’t make this change