Ryan Leary
Ryan Leary
It would be nice to add in a third script `score.lua` which would take a trained model and an example (or many examples) and provide log probabilities of the character...
Labels can be more than single characters/unicode codepoints in practice.
Throw an exception when files do not exist rather than exiting the Python interpreter.
Should provide a link to more information on KenLM language models and brief writeup on how to build.
Aim is to be backwards compatible with API < 1.40 while only requiring gpus kwargs be used. Thanks to @sigurdkb for original implementation.
This simply allows '.' to exist in arguments. Some libraries use `.` to denote nested configurations which is not currently possible with cxxopts. - - - This change is [](https://reviewable.io/reviews/jarro2783/cxxopts/358)
There may be a bug including more than 2 layers.
Librosa's implementation seems to cause issues despite being faster.
Allow specifying multiple manifests for trainer.
Current implementation is extremely inefficient. Use grouped convolutions.