is it possible to stop package dependencies update when experiments are executing via an agent

Can you elaborate on this? You mean package dependencies when experiments are executing via agent?

CostlyOstrich36 Yes

Thanks CostlyOstrich36 for you response. However, I am still getting errors after forcind add requirement via code Torch CUDA 116 download page found
clearml_agent: Warning: could not resolve python wheel replacement for torchvision!=0.13.0,>=0.8.1
clearml_agent: ERROR: Could not install task requirements!
Exception when trying to resolve python wheel: Could not find pytorch wheel URL for: torchvision!=0.13.0,>=0.8.1 with cuda 117 support

You can specify specific package versions yourself via code
https://clear.ml/docs/latest/docs/references/sdk/task#taskadd_requirements

Hi BroadSeaturtle49 , this would be an option in the agent very soon

CostlyOstrich36 No

What are the packages that are detected when you run the experiment locally?

Are you running inside a docker?

BroadSeaturtle49 agent RC is out with a fix:
pip3 install clearml-agent==1.5.0rc0Let me know if it solved the issue