v0.7.12-release

What's Changed

• fix: missing run in run-collector-client make target by @sunya-ch in #389
• feat: get type in /best-models API and pass source and type from estimator by @sunya-ch in #392
• fix: remove remaining intel_rapl source in script by @sunya-ch in #400
• feat: add --machine-spec arg to estimator and get_machine_spec by @sunya-ch in #398
• feat: add compute similarity and loose selection by @sunya-ch in #401
• feat: add machine_spec to metadata by @sunya-ch in #405
  server/pull/416
• chore(deps): Bump scipy from 1.14.0 to 1.14.1 by @dependabot in #393
• chore(deps): Bump werkzeug from 3.0.3 to 3.0.4 by @dependabot in #403
• fix: pass true as string in pipeline param by @sunya-ch in #430
• fix(estimator): logging and machine-spec computation by @sthaha in #444
• feat: support --config-dir arg to point to the configuration directory by @sthaha in #456
• chore(deps): Bump scikit-learn from 1.5.1 to 1.5.2 by @dependabot in #446
• chore(deps): Bump boto3 from 1.34.155 to 1.35.39 by @dependabot in #497
• chore(deps): Bump pandas from 2.2.2 to 2.2.3 by @dependabot in #500
• chore(deps): Bump prometheus-client from 0.20.0 to 0.21.0 by @dependabot in #501
• chore(deps): Bump boto3 from 1.35.39 to 1.35.43 by @dependabot in #504
• chore: refactor and enhance e2e test script by @vprashar2929 in #483
• fix(workflow): enable tekton test and integration test on every PR by @vimalk78 in #477
• fix: disable expose idle power by @vimalk78 in #510
• chore(deps): Bump psutil from 6.0.0 to 6.1.0 by @dependabot in #512
• fix(config): use predictable paths and proper typings by @sthaha in #509
• fix(prom): remove unused PROM_ configs by @sthaha in #515
• feat(vm_metrics): Enabled vm metric use for local model training by @KaiyiLiu1234 in #464
• fix: check length of PROM_THIRDPARTY_METRICS by @vprashar2929 in #516

Full Changelog: v0.7.11-2-release...v0.7.12

v0.7.11-2-release

What's Changed

• fix: add boto3, spec pipeline version, version export path, and BREAKING CHANGE remove group feature by @sunya-ch in #341
• chore(deps): Bump xgboost from 2.1.0 to 2.1.1 by @dependabot in #338
• chore(deps): Bump protobuf from 5.27.2 to 5.27.3 by @dependabot in #340
• chore: update README to use v0.7.11 by @sthaha in #335
• chore(deps): Bump boto3 from 1.34.69 to 1.34.154 by @dependabot in #353
• chore(deps): Bump boto3 from 1.34.154 to 1.34.155 by @dependabot in #357
• fix: pd.unique warning and ignore file not found by @sunya-ch in #360
• bump up kepler-action to 0.0.8 by @SamYuan1990 in #313
• Revert "bump up kepler-action to 0.0.8" by @sthaha in #363
• fix: incompatability of specpower model on 0.7.11 by @sunya-ch in #367
• chore: Use top-level kepler_model python package by @sthaha in #323
• doc(tekton): fix tekton release URL by @ideaship in #372
• chore: add integration test with latest by @sunya-ch in #374
• fix(estimator): use click to handle log-level by @sthaha in #375
• fix(server): use logging instead of print by @sthaha in #376
• feat: update select logic with spec similarity computation by @sunya-ch in #370
• fix: apply kepler tag to deploy manifests by @sunya-ch in #377
• Revert "feat: update select logic with spec similarity computation" by @sthaha in #381
• fix(estimator): crash when logging by @sthaha in #380
• fix: ignore whitespaces in config files by @sthaha in #384
• fix: mismatch model request check (unexpectedly-repeated model request) by @sunya-ch in #383
• fix: ignore whitespaces in MODEL_CONFIG file by @sthaha in #386
• chore: hatch fmt to format all source to 120 char width by @sthaha in #388
• chore(compose): add compose for local development by @sthaha in #382
• chore: minor cleanup to add logging by @sthaha in #387

New Contributors

• @ideaship made their first contribution in #372

Full Changelog: v0.7.11-release...v0.7.11-2-release

v0.7.11-release

What's Changed

• fix logic to obtain new model by @sunya-ch in #246
• change kepler config path and export func name by @sunya-ch in #253
• update local_dev_cluster_version to v0.0.5 by @sunya-ch in #255
• fix(s3): use python 3.10 by @sthaha in #285
• fix(tekton): update the README by @vprashar2929 in #298
• Revert to latest tag instead of v0.7 by @KaiyiLiu1234 in #283
• change image version to latest by @sunya-ch in #310
• enrich ec2 train/plot and remove replacement of NodeAttribute.PROCESSOR with instance profile by @sunya-ch in #311
• Base image fix and try to add container in dependbot scan by @SamYuan1990 in #317
• add manual training with entrypoint instruction by @sunya-ch in #312
• bump kepler version 0.7.11 by @sunya-ch in #325
• update instance list and pipeline for ec2-0.7.11 by @sunya-ch in #322
• change default pipeline to ec2-0.7.11 by @sunya-ch in #327
• make filter pod by benchmark optional by @sunya-ch in #326
• remove condition to append energy source by @sunya-ch in #330
• change default pipeline name to std_v0.7.11 by @sunya-ch in #329
• apply energy source param on train-model flow by @sunya-ch in #331
• fix: enhance cluster setup script for model training by @vprashar2929 in #320

New Contributors

• @vprashar2929 made their first contribution in #298

Full Changelog: v0.7.7-release...v0.7.11-release

v0.7.7-release

Highlights

• Add page_cache_hit in BPFOnly feature group
• Remove KubeletOnly feature group
• Change bpf_cpu_time unit from us to ms
• Change component energy source from rapl to intel_rapl
• Support third party metrics
• Add XgboostTrainer and CurveFitModel trainer class
• Add new trainers: XgboostFitTrainer, ExponentialRegressionTrainer, LogisticRegressionTrainer, LogarithmicRegressionTrainer
• Tekton pipeline integration for power model training
• Self-hosted power data collection and model training on EC2 baremetal spot instance
• Support multiple node types in a pipeline
• CI updates
  • Add base-image build
  • Add local-model-db
  • Add Tekton test
  • Add kepler-model-db integration test
  • Enable dependabot
• New 317 node types of CPU Power Models from SPECPower database (for platform power - acpi energy source)
• New 5 node types of CPU power Models from Kepler metrics on EC2 instances with Stress-NG workload.

What's Changed

• add missing writer file by @sunya-ch in #177
• change python version 3.8+ to 3.8 by @sunya-ch in #180
• update base image for 0.7 by @sunya-ch in #182
• relabel with version 0.7 by @sunya-ch in #183
• migrate xgboost to trainer implementation by @sunya-ch in #184
• Support third party metrics in model server by @LeiZhou-97 in #185
• minor bug fixes to prepare a kind cluster by @knarayan in #186
• fix unittest (fix dep)/integration test (remove short option) by @sunya-ch in #188
• add page_cache_hit and MAPE by @sunya-ch in #192
• update weight format by @sunya-ch in #190
• changes to collect metrics from Prometheus with benchmark run outside of kepler-model-server by @knarayan in #191
• encode xgboost weight, fix MAPE, reorganize exported folder by @sunya-ch in #193
• add isolate_from_data and train_from_data, with refactor entrypoint (tekton prerequisite) by @sunya-ch in #195
• add Tekton pipelinerun by @sunya-ch in #196
• add to-csv option to query command and correct outdated description by @sunya-ch in #197
• [CI] enable dependabot for kepler model server by @SamYuan1990 in #198
• Bump docker/setup-qemu-action from 2 to 3 by @dependabot in #201
• Bump docker/login-action from 1 to 3 by @dependabot in #202
• Bump docker/setup-buildx-action from 1 to 3 by @dependabot in #203
• Bump sustainable-computing-io/kepler-action from 0.0.1 to 0.0.4 by @dependabot in #204
• Bump actions/checkout from 2 to 4 by @dependabot in #200
• [CI] image build logic enhancement by @SamYuan1990 in #199
• Bump actions/checkout from 3 to 4 by @dependabot in #207
• Integrate tekton to model_training and GitHub workflows by @sunya-ch in #209
• Bump sustainable-computing-io/aws_ec2_self_hosted_runner from 1 to 3 by @dependabot in #211
• Bump sustainable-computing-io/kepler-action from 0.0.4 to 0.0.5 by @dependabot in #208
• upgrade xgboost to 2.0.1 by @sunya-ch in #210
• Bump sustainable-computing-io/aws_ec2_self_hosted_runner from 3 to 4 by @dependabot in #217
• Add curvefit trainer (fix minor issue, update test data) by @sunya-ch in #215
• Update energy source rapl to intel_rapl by @Yanbo0101 in #218
• add local db, remove kubelet, update push-pr by @sunya-ch in #222
• update README, simplified fig by @sunya-ch in #223
• fix wrong image name by @sunya-ch in #224
• fix: minor doc typo corrections by @sthaha in #226
• add node type indexing by @sunya-ch in #225
• change cpu_time unit (us to ms) by @sunya-ch in #227
• fix pipeline bug and workflow by @sunya-ch in #228
• Add metadata plot and Fix bugs by @sunya-ch in #229
• feat: add uml diagrams by @vimalk78 in #230
• update curvefit (log func) and generate_spec by @sunya-ch in #232
• add MAPE to estimator result and update exported figure by @sunya-ch in #233
• specify kepler tag by @sunya-ch in #236
• fix env error in push-to-main ci by @sunya-ch in #237
• Bump sustainable-computing-io/kepler-action from 0.0.5 to 0.0.6 by @dependabot in #239
• [CI] adjust CI settings by @SamYuan1990 in #240
• Update kepler-model-db link, serve model by spec, kepler-model-db integration CI, visualize power curve by @sunya-ch in #244

New Contributors

• @LeiZhou-97 made their first contribution in #185
• @knarayan made their first contribution in #186
• @SamYuan1990 made their first contribution in #198
• @dependabot made their first contribution in #201
• @sthaha made their first contribution in #226
• @vimalk78 made their first contribution in #230

Full Changelog: v0.6.0-release...v0.7.7-release

v0.6.0

Power model information

Model Output Type

AbsPower, DynPower

Energy source

Energy/power source	Energy/power components
rapl	package, core, uncore, dram
acpi	platform

Feature group

Group Name	Features	Kepler Metric Source(s)
CounterOnly	COUNTER_FEATURES	Hardware Counter
CgroupOnly	CGROUP_FEATURES	cGroups
BPFOnly	BPF_FEATURES	BPF
KubeletOnly	KUBELET_FEATURES	Kubelet
IRQOnly	IRQ_FEATURES	IRQ
AcceleratorOnly	ACCELERATOR_FEATURES	Accelerator
CounterIRQCombined	COUNTER_FEATURES, IRQ_FEATURES	BPF and Hardware Counter
Basic	COUNTER_FEATURES, CGROUP_FEATURES, BPF_FEATURES, KUBELET_FEATURES	All except IRQ and node information
WorkloadOnly	COUNTER_FEATURES, CGROUP_FEATURES, BPF_FEATURES, IRQ_FEATURES, KUBELET_FEATURES, ACCELERATOR_FEATURES	All except node information
Full	WORKLOAD_FEATURES, SYSTEM_FEATURES	All

Train methods

• PolynomialRegression
• GradientBoostingRegressor
• SGDRegressor
• KNeighborsRegressor
• LinearRegression
• SVRRegressor

Power model accuracy report

version	machine ID	pipeline	feature group	component power source	total power source	Local LR MAE in watts (Node Components/Total)	Estimator Sidecar MAE in watts (Node Components/Total)	Reference Power Range in watts
0.6	nx12	std_v0.6	BPFOnly	rapl	acpi	66.32/93.57	34.40/49.52	505.79

PRs to other projects

• sustainable-computing-io/kepler#947
• sustainable-computing-io/kepler-model-db#16

Contributors

• @rootfs made their first contribution in #1
• @KaiyiLiu1234 made their first contribution in #3
• @husky-parul made their first contribution in #20
• @Shreyanand made their first contribution in #32
• @sunya-ch made their first contribution in #34
• @Yanbo0101 made their first contribution in #128
• @jiangphcn made their first contribution in #142
• @marceloamaral made their first contribution in #146
• @mcalman made their first contribution in #159
• @Saurabhkr952 made their first contribution in #168

Full Changelog: https://github.com/sustainable-computing-io/kepler-model-server/commits/v0.6.0-release