Optimize Cast op when source and dest types are the same, add u8/i8 cast tests

robertknight / rten

ONNX neural network inference engine

124 stars 9 forks source link

Closed robertknight closed 1 month ago

robertknight commented 1 month ago

Optimize casts where the source/dest types are the same by using optimized tensor copy operations instead of Tensor::map, and skipping the copy altogether if the Cast op is run in-place
Refactor tests to make adding new cases easier, and add some tests for i8/u8 conversions