Releases: LGE-ARC-AdvancedAI/auptimizer
Releases · LGE-ARC-AdvancedAI/auptimizer
Auptimizer v2.0 Release
Major Features
- Add Dashboard to provide insightful visualizations to help users track experiment progress, analyze and contrast jobs, experiments, and optimization approaches
- Add Compressor - a model compression tool that helps reduce memory complexity and inference time of neural networks
- Add 4 early stopping strategies to improve the efficiency of HPO experiments
- Add feature to track intermediate results of both HPO and compression experiments
- Add feature to save the best model with the best hyperparameter combination
- Add feature to track multiple results returned from the HPO experiment
Enhancements
- Refactor documentation
- Minor bug fixes
Auptimizer v1.4 Release
Major Features
- Add Converter to convert models in various formats (SavedModel/Protobuf/Checkpoint/Keras/Pytorch) into TensorFlow Lite or ONNX
- Re-establish SSH connection upon connection failure
- Resubmit failed jobs and resume experiment from job failure
Enhancements
- Refactor documentation
Auptimizer v1.3 Release
Major Features
- Adding Profiler - a simulator for profiling performance of Machine Learning model scripts.