Hi recently upgraded all the clearml, clearml-server, clearml-agent.
Now running k8s glue with clearml-agent=1.0.1rc1.
python3 k8s_glue_example.py --queue 1bgpu --template-yaml custom.yml --namespace defaultcustom.yml looks like this.
` apiVersion: v1
kind: Pod
spec:
containers:

• image: quay.io/jax79sg/clearml-agent:latest
  resources:
  nvidia.com/gpu: 1
  cpu: "1"
  memory: "4Gi"
  restartPolicy: Never I run my code (clearml=1.0.3) with extract of following task.set_base_docker("quay.io/jax79sg/detectron2:v4 )
  task.execute_remotely(queue_name="1gpu", exit_process=True) When the clearml runs, the CLearml Server UI shows http://quay.io/jax79sg/detectron2:v4 in "execution->container". But when i performkubectl describe pod [podname]` spawned by k8s glue, the pod is pulling the default image stated in the clearml.conf of the agent. Why is this so? This has broken most of our runs.