Sunil Thaha
Sunil Thaha
CI now (#293 ) builds base image if dockerfiles are changed. It also gives us the opportunity to build and test model-server image as well when the image is rebuilt...
Running `hatch fmt --check` points a lot issues with the project. E.g. - [ ] E402 Module level import not at top of file - [ ] F401 [*] `train.node_info_column`...
### What happened? Running model-server latest image produces the following log ``` model-server-1 | * Serving Flask app 'model_server' (lazy loading) model-server-1 | * Environment: production model-server-1 | WARNING: This...
Allows selection of `kepler_**level**` and create panels for the following ---
Currently the trainer name is encoded in the name of the model's file while this allows for fast lookup, it prevents files to have different names (e.g. acpi_AbsPower.json) since the...
https://github.com/sustainable-computing-io/kepler/pull/1728 wasn't able to update the `intel_rapl_DynPower` model since the model is missing in model-db (see: https://github.com/sustainable-computing-io/kepler-model-db/issues/27). The task is to update the models when the linked bug is fixed.
See: https://github.com/sustainable-computing-io/kepler-metal-ci/blob/main/docs/train-validate-e2e/2024-09-09_22-41-38/AbsPower/BPFOnly/ExponentialRegressionTrainer/report-v0.7.11-212-gac5ee8b8.md#platform---idle Which reports platform power use as `160,000 W` (as if there is a divide / 1000 error) measured through ACPI
MAPE produced by the [analytics notebook](https://github.com/sustainable-computing-io/kepler-metal-ci/blob/main/docs/analytics/data_kepler_analytics.ipynb) is different to the the one produced by validator. https://github.com/sustainable-computing-io/kepler-metal-ci/blob/main/docs/analytics/example.jpg Lets investigate if the MAPE produced by the validator script is the right one.