dawilliams-nvidia

Dask-XGBoost non-deterministic results

9

I have been trying to train XGBoost on multi-gpu via Dask. It is required that the training pipeline be made deterministic (for example for comparing Hyperparameter runs), exactly the same...

Attempting to use the Bayesian Optimization package to run many iterations of Dask XGBoost. The issue is that dask workers will disconnect during the optimization process, usually within the first...

convert_results.yaml instead of convert.yaml?

1

Following the quick start guide on a fresh install, receive the following error mod: ``` model-navigator run --model-name add_sub \ --model-path examples/quick-start/model.pt \ --inputs INPUT__0:-1,16:float32 INPUT__1:-1,16:float32 \ --outputs OUTPUT__0:-1,16:float32 OUTPUT__1:-1,16:float32...