Update kepler-model-db link, serve model by spec, kepler-model-db integration CI, visualize power curve

[sunya-ch]

This PR introduces multiple significant changes to kepler-model-server to fill the model serving cycle and CI.

1. update Kepler-model-db link

For release v0.7, we expand power model ranges with SPECpower database and ec2 spot instance.
As specpower has only acpi (platform) energy source while ec2 has only intel_rapl (component) energy source, we need to refer to two pipelines for each energy source.

This PR also updates the config in manifest file from 0.6 to 0.7 (refer to #242)

2. serve model by spec

This PR provides a simplified version to serve a power model based on the number of CPU cores. If the power model is not available, the power model which is trained by machine with maximum number of cores will be selected as candidates.

3. kepler-model-db integration

To discover incompatibility between this repository and kepler-model-db, I added the CI with DB opt to determine whether the kepler-model-db is considered. (refer to #243)

This added CI can confirm the update on 1. should fix the deployment issue sustainable-computing-io/kepler#1306.

4. visualize power curve on export

To quickly validate whether the trained power model make sense or not, I provided a new function to plot power curve on export. (refer to #241)

---

This PR is WIP until the follow items are done:

• merge https://github.com/sustainable-computing-io/kepler-model-db/pulls
• update model/pipeline repository to official repository (currently use sunya-ch on specpower branch)

Signed-off-by: Sunyanan Choochotkaew [email protected]

[rootfs]

@KaiyiLiu1234 can you share your thoughts? thanks