Created by: kbinias
This PR improves the performance of sequence_pad operator. From 3x (for small tensors) to 50x (for large tensors e.g. 1118x128) speedup.
Resolves #17271 (closed)
Created by: kbinias
This PR improves the performance of sequence_pad operator. From 3x (for small tensors) to 50x (for large tensors e.g. 1118x128) speedup.
Resolves #17271 (closed)