FLAML
A fast library for AutoML and tuning. Join our Discord: https://discord.gg/Cppx2vSPVP.
MIT License
Bot releases are hidden (Show)
Published by sonichi over 3 years ago
- micro/macro f1 metrics added for multi-classification
- simple preprocessing of datetime columns
- support lightgbm with customized objective
- bug fix in FLOW^2
Published by sonichi over 3 years ago
- Search space improvement
- Separation of initial points and low cost partial config in HPO input
Published by sonichi over 3 years ago
A number of new features and improvements:
- pickle the AutoML object
- get best model per estimator
- bug fix
- add verbose argument in AutoML.fit()
- nni support
- stateless HPO API
Published by sonichi over 3 years ago
perf improvement for parallel tuning and bug fix
Published by sonichi over 3 years ago
added support for nested search space
added experimental support for nni
minor bug fix
Published by sonichi over 3 years ago
make sample_weight right for ensemble; reduced dependency
Published by sonichi over 3 years ago
minor bug fix for rf
Published by sonichi over 3 years ago
Enhanced algorithms and implementations of hyperparameter tuning with support for parallel tuning and ray tune;
Support keyword arguments like sample_weight;
Enable mlflow logging.
Published by sonichi almost 4 years ago
Set default logging level to INFO
Published by sonichi almost 4 years ago
Hyperparameter + sample size: Cost Frugal Optimization (CFO) + Estimated Cost of Improvement (ECI)
Learner selection: ECI
Resampling strategy: threshold rule
Published by sonichi almost 4 years ago
First release of FLAML.
Hyperparameter + sample size: Cost Frugal Optimization (CFO) + Estimated Cost of Improvement (ECI)
Learner selection: ECI
Resampling strategy: threshold rule
