Sockeye

Added

- Added kNN-MT model from [Khandelwal et al., 2021](https://arxiv.org/abs/2010.00710).
- Installation: see [faiss document](https://github.com/facebookresearch/faiss/blob/main/INSTALL.md) -- installation via conda is recommended.
- Building a faiss index from a sockeye model takes two steps:
- Generate decoder states: `sockeye-generate-decoder-states -m [model] --source [src] --target [tgt] --output-dir [output dir]`
- Build index: `sockeye-knn -i [input_dir] -o [output_dir] -t [faiss_index_signature]` where `input_dir` is the same as `output_dir` from the `sockeye-generate-decoder-states` command.
- Faiss index signature reference: [see here](https://github.com/facebookresearch/faiss/wiki/The-index-factory)
- Running inference using the built index: `sockeye-translate ... --knn-index [index_dir] --knn-lambda [interpolation_weight]` where `index_dir` is the same as `output_dir` from the `sockeye-knn` command.

Changed

- allow torch 1.13 in requirements.txt
- Replaced deprecated `torch.testing.assert_allclose` with `torch.testing.close` for PyTorch 1.14 compatibility.

Added

- `--tf32 0|1` bool device (`torch.backends.cuda.matmul.allow_tf32`)
enabling 10-bit precision (19 bit total) transparent float32
acceleration. default true for backward compat with torch < 1.12.
allow different `--tf32` training continuation

Changed

- `device.init_device()` called by train, translate, and score
- allow torch 1.12 in requirements.txt

Changed
- Updated to sacrebleu==2.3.1. Changed default BLEU floor smoothing offset from 0.01 to 0.1.

Fixed

- Updated DeepSpeed checkpoint conversion to support newer versions of DeepSpeed.

Changed

- Change decoder softmax size logging level from info to debug.