Binary quantization?
Open
- Dominant language
- Python
- Stars
- 361
- Forks
- 33
- Avg merge
- 18h 9m
- Merged PRs (30d)
- 4
Description
The README mentions the [Binarized Neural Machine Translation](https://arxiv.org/abs/2302.04907) paper but does not really elaborate on how one can use AQT to implement one-bit weights and activations using AQT. For example, will the library take care of using LayerNorms as replacement for scaling factor, as mentioned in the paper?
Contributor guide
Assessment
This issue has not been assessed yet.