SkyPilot: Run AI and batch jobs on any infra (Kubernetes or 12+ clouds). Get unified execution, cost savings, and high GPU availability via a simple interface.
It would be great to have the ability to define custom logs through SkyCallback for things OTHER than steps/training. For example, inference with different modes and params, warmup time, etc.
Something like sky_callback.report(key: string, value: string) that would then propagate as a column in the sky benchmark report from each VM.
It would be great to have the ability to define custom logs through SkyCallback for things OTHER than steps/training. For example, inference with different modes and params, warmup time, etc. Something like
sky_callback.report(key: string, value: string)
that would then propagate as a column in the sky benchmark report from each VM.Is there any way to currently achieve this?