NitroML is a modular, portable, and scalable model-quality benchmarking framework for Machine Learning and Automated Machine Learning (AutoML) pipelines.
Add two-stage tuning in the custom Tuner with a budget of k trials. The tuner first does warmup tuning for p trials with the hyperparameters recommended by the metalearner and then does random tuning for k-p trials.
Add two-stage tuning in the custom Tuner with a budget of
ktrials. The tuner first does warmup tuning forptrials with the hyperparameters recommended by the metalearner and then does random tuning fork-ptrials.