Releases: rwth-i6/returnn
Releases · rwth-i6/returnn
Transformer architecture support beta
Transformer architecture support beta
Support for batching for NMT in on-line mode
- Support for translating batches of input in on-line (deployment mode)
- SwitchLayer
TF-based pre-release with a function for inference with a single string as input
Pre-release
- Support for model training and inference with TensorFlow
- Supports inference (translation) of single sentences (intended for server mode)