I deepdive into Ultralytics code and found that clearML is not logging the expereiment, but rather the model used in checking amp on the machine, specifically def check_amp(model): create a new model and get tracked, which is not the one for training.
Thanks! I will test it for a bit and see if I can utilized the model feature in clearml

hmmm it seems that it is not the case. Neverthless, the following code works:

input_model: InputModel = InputModel.import_model(name="test_input", weights_url="model/yolov8n.pt")
task.connect(input_model)

model = YOLO(input_model.get_local_copy())

just fyi i read Model SDK doc for several time and didn't expect the option is inside Task . Many Thanks!

This approach disable file server auth service. Another fix is to set the correct cookie domain.

Hi, I am getting the same error as well.

On my local setup, everything works fine with a local ip address.
When I request the image directly on my local setup, the browser use the following cookie:

clearml_token_basic: XXX
domian: 192.168.51.60

When I request the image on my k8s cluster, the browser cannot find the cookie and return a 401 respond.

The default cookie level is different as well:

clearml_token_basic: XXX
domain: app.xxx.com

I try plugging the cookie whe...

@<1523701070390366208:profile|CostlyOstrich36> There's still issue in the frontend. The frontend for plots does not include the cookie when requesting the images, but it does when showing the debug samples. (I am using the open source and self-hosting with k8s)

![image](https://clearml-web-assets.s3.amazonaws.com/scoold/images/TT9ATQXJ5-F07UFB3RPSP/screenshot_...

Also I tried follow None

but it have some SSL issues. The ssl cert only allow for 1 subdomain level (only 1 dot). Is there any work around? I am currently using http only.