64 Star 289 Fork 140

GVPopenEuler / A-Tune

2021-11-15 17:28

the number of yaml files for application parameters is increased to 50
automatically set disk/network info in server.yaml
adapt to CentOS
add the automatic execution method of test cases
add some tuning examples
add tools to convert csv file to database value
add Traverse algorithm
support for Go 1.15
support to connect engine to database
enable restart tuning for gridsearch algorithm
remove the MPI/CPI module and adapt to the atune_collector code repository
add gridsearch tuning algorithm
seperate atune-web and atune-engine service
the parameter list is displayed by parameter importance
collection data and analysis page display on UI
detect misclassified data
set default fluctuation value is 0.001 for the selection of important parameters
add a new feature selector based on variance reduction
read correct historical optimization data for incremental optimization
add make search command to search the best parameters using grid search
add grid search method for training models
only require prefetch_tuning on aarch64
add atune model training metrics description
fix some bugs

Last committed message: !363 atune: upgrade A-Tune to v1.0.0
2020-09-04 15:51

Support more applications and knobs
multiple optimization algorithms are supported.
support incremental optimization
New workload classification model
sensitive parameter identification and automatic filtering
Supports running on VMs.
Support independent engine deployment
Support make models to update models

Last committed message: !169 prepare for v0.3 release
2020-03-19 12:54

Some bugfix and code optimization

Last committed message: !65 prepare for v0.2 release
2020-01-20 16:39

Support analysis and auto optimization for following applications on Kunpeng 920
Support automatic parameter optimization
Support user-defined workload types and service models

release first version:)



184635 d8eb8a04 1850385 161156 f1cf3f24 1850385