Sounds like a better solution. Less clicks required in order to get to my page. Thanks 🙂

AgitatedDove14 I was trying to get all tasks that hold a specific value/pattern in a hyperparameter. e.g. - after performing grid-search, get all training tasks with a specific learning rate

Sweet! Will check it out and try to adopt 🙂

Thanks for the prompt response SuccessfulKoala55 ! exactly what I was looking for 🙂

AgitatedDove14 - using your suggested solution looks a bit more generic, and a proper .edit method is a good idea for that. However - it doesn’t resolve the API version issue I had (that is, unless you also suggest to add a method to add execution.script , which isn’t initialized when I use Task.create )

I have a template which I populate on another process I wrote, which runs once my experiment is over.
A report generation functionality sounds great. add my vote for it 🙂

🤩

client version 0.15.1

Exactly what I was looking for!
So just report scalars as I would anyways and then set the dispay to wall-time, right?

0.15.1-366

Thanks SuccessfulKoala55 , dropping irrelevant fields did improve the response time.

AgitatedDove14 My use-case was a bit different. I only populate the repository url, entrypoint, and commit SHA in the Script object.
We wanted to have something informative enough in our task on one hand, but not to load with redundant data on the other - so a commit SHA made perfect sense.
I ended up using from trains.backend_api.services import tasks and then initialize with tasks.Script(…)

I think I’d expect to be able to view the page I’m uploading inline, or at least a thumbnail of it.

To clarify - the html I’m creating is an experiment summary which includes metrics/scalars from my experiment, as well as dataset statistics. I’m not sure it should go to the “debug images” tab - that’s just where it went when I used report_media

found it.

` fpr, tpr, thresholds = my_roc_calc(my_input)
t = Task.get_task(my_task_id)

t.get_logger().report_scatter2d("ROC", "MyModel",iteration=1, scatter=list(zip(fpr, tpr))) `

oh, it is indeed misleading.
what should I assign to internal_task_representation.execution.script then (per your example snippet)? a dictionary?

also - should the value passed to task be the task id?

found a reasonable solution. I had to specify http://execution.parameters.my _hyperparameter in the fields list.

Thanks AgitatedDove14
Using your snippet caused the following error:
TypeError: __init__() missing 1 required positional argument: 'metrics'
when I used client.events.debug_images(task='aabbcc', metrics=None) I got the following: ValueError: Unsupported keyword arguments: task