Closed ASvyatkovskiy closed 7 years ago
The main purpose of this PR is to continue work started in https://github.com/PPPLDeepLearning/plasma-python/pull/15 to enforce specific precision, and fully enable float16 training.
This PR focuses mostly on precision in data, which affects preprocessing, cutting, resampling, normalization, and loading of shots. It introduces explicit dtype parameters
dtype
In addition, TBA
The main purpose of this PR is to continue work started in https://github.com/PPPLDeepLearning/plasma-python/pull/15 to enforce specific precision, and fully enable float16 training.
This PR focuses mostly on precision in data, which affects preprocessing, cutting, resampling, normalization, and loading of shots. It introduces explicit
dtype
parametersIn addition, TBA