Combining Taints and Tolerates

Labels and taints can be used together for fine-grained control over workload placement.

Example

juice run --match-labels gpu-type=a100,env=pytorch-2.0 --tolerates hipaa-compliant=required,expensive-gpu=true my_secure_ml_app.py 

info

This command would run my_secure_ml_app.py on an agent that:

1. Has an A100 GPU

2. Has PyTorch 2.0 installed

3. Meets HIPAA compliance requirements

4. Is explicitly allowed to run expensive workloads

Juice users can leverage these features to ensure their workloads are run on appropriate hardware and in the correct environments, balancing performance needs with resource management and compliance requirements.