Kenneth Heafield
Kenneth Heafield
They come from moses. https://github.com/moses-smt/mosesdecoder/tree/master/scripts/share/nonbreaking_prefixes
Updated en-de models posted, thanks @kaleidoescape "checksum": "7f6bdcf60555fca479e014a6722729b34890e52ca8bfbffb5138f574ec91aec7", "url": "http://data.statmt.org/bergamot/models/deen/ende.student.base.tar.gz", "checksum": "5214a434a8b6d0562eb927ff5ffe42d4a60240370a0095e0c1369d960878254f", "url": "http://data.statmt.org/bergamot/models/deen/ende.student.tiny11.tar.gz", cc @andrenatal @lonnen
The deduper is designed to take the first input and remove the subsequent ones. I think you want the one with highest bicleaner score?
fixed-quant is dead since it's been merged into master and @afaji needs to pay attention to issue #24 to remove references to it. You can get a slower 8-bit model...
@XapaJIaMnu the 8-bit documentation is lacking.
1. Train an FP32 model as usual. 2. Optional: finetune the FP32 model with 8-bit damage. This step is mostly only useful if the model is particularly small (on the...
Oh my that is terrible. The whole file is Ale with various lengths. @eu9ene the pipeline shouldn't be continuing if quality is that terrible.
@mfomicheva Thanks for the repo https://github.com/browsermt/quality-estimation with QE training and soon the binarizer. @felipesantosk @mfomicheva Can you binarize the 3 JSON files and submit a PR to this repo (students)...
de-en models deployed in: http://data.statmt.org/bergamot/models/deen/deen.student.base.tar.gz http://data.statmt.org/bergamot/models/deen/deen.student.tiny11.tar.gz and mentioned in https://translatelocally.com/models.json cc @lonnen @andrenatal TODO: include them in this repo. Or, better, consolidate the translatelocally deployment with this students repo so...