Ecosyste.ms: OpenCollective

An open API service for software projects hosted on Open Collective.

github.com/mozilla/translations

The code, training pipeline, and models that power Firefox Translations
https://github.com/mozilla/translations

Currency translation for English to German is incorrect

nordzilla opened this issue 3 months ago
Unify the `browsermt-marian-dev` git submodule dependencies

nordzilla opened this issue 3 months ago
Multiply comet score by 100 in online mode

vrigal opened this pull request 3 months ago
Fork bergamot-translator into this repostiory as inference

nordzilla opened this pull request 3 months ago
Firefox BroFirefox browser spell checker does not workwser

OpenGreenStreet opened this issue 3 months ago
Output all of test spew

gregtatum opened this pull request 3 months ago
Student trains too long

eu9ene opened this issue 3 months ago
Upload GCP experiments

eu9ene opened this issue 3 months ago
COMET is still logged in [0,1] interval

eu9ene opened this issue 3 months ago
Actualize documentation about experiment tracking on Weight & Biases

vrigal opened this pull request 3 months ago
Investigate switching to ICU segmenter

eu9ene opened this issue 4 months ago
fix: pass 'permission' to train action

bhearsum opened this pull request 4 months ago
feat: bump to taskgraph 11.1.0

bhearsum opened this pull request 4 months ago
We should audit our storage used in a full pipeline run

gregtatum opened this issue 4 months ago
Use the config.ci.yml for the training defaults

gregtatum opened this pull request 4 months ago
CI run is slowed down by failing to download fetches

gregtatum opened this issue 4 months ago
Add a pre-release test for models

gregtatum opened this issue 4 months ago
Update the Statistics class to use data attributes and use it in merge-mono.py

gregtatum opened this pull request 4 months ago
Add environment variables to all artifacts and fetches

gregtatum opened this pull request 4 months ago
Merge corpus rewrite to python

gregtatum opened this pull request 4 months ago
Only retain the best metric model, and delete the others

gregtatum opened this issue 4 months ago
Do not upload temporary directory to artifacts (#841)

eu9ene opened this pull request 4 months ago
New CI tests with opustrainer and dataset truncation changes

gregtatum opened this pull request 4 months ago
Speed up CI by changing model params

gregtatum opened this pull request 4 months ago
Stalled metrics are not reported for backward model

eu9ene opened this issue 4 months ago
Actualize documentation on experiment tracking

eu9ene opened this issue 4 months ago
Consider rebalancing datasets with clustering

eu9ene opened this issue 4 months ago
Made-up words in translations

eu9ene opened this issue 4 months ago
Rewrite train.sh to train.py

gregtatum opened this pull request 4 months ago
Do not upload temporary directory to artifacts

eu9ene opened this pull request 4 months ago
The train.py utility should check for a proper branch name

gregtatum opened this issue 4 months ago
[Experiment] CJK

eu9ene opened this pull request 4 months ago
Update Sacrebleu in config-generator

eu9ene opened this issue 4 months ago
Add an HPLT data importer

gregtatum opened this pull request 4 months ago
Splitter compression

gregtatum opened this pull request 4 months ago
Alignment fails on en-el tedx/valid dataset

eu9ene opened this issue 4 months ago
chore: update to taskgraph 11

bhearsum opened this pull request 4 months ago
Make the CI model training even slimmer

gregtatum opened this pull request 4 months ago
tests/test_data_importer.py is the slowest test in CI, due to installing dependencies

gregtatum opened this issue 4 months ago
train-student failed with UnicodeDecodeError

eu9ene opened this issue 4 months ago
evaluate-* is the slowest step in CI, as it's having to download nvidia cudnn and torch

gregtatum opened this issue 4 months ago
Add a test-fast command for faster local testing

gregtatum opened this pull request 4 months ago
Fix missing dependency in task train

gregtatum opened this pull request 4 months ago
Add known bad datasets and failures to the config generator

gregtatum opened this pull request 4 months ago
Ensure we always export the VCS_PATH into the PYTHONPATH

gregtatum opened this pull request 4 months ago
Rename all commands to use $TASK_WORKDIR/artifacts rather than an absolute path

gregtatum opened this pull request 4 months ago
Remove the compression command configuration, and only use zstandard

gregtatum opened this pull request 4 months ago
backport get_ancestors fix to ignore 404 errors

bhearsum opened this pull request 4 months ago
Docs deploy failed

eu9ene opened this issue 4 months ago
Add a memory logger

gregtatum opened this pull request 4 months ago
Add group ID suffix to group_logs metrics published from online evaluation tasks

vrigal opened this pull request 4 months ago
Investigate using LLMs for evaluation

eu9ene opened this issue 4 months ago
Pass W&B suffix to publish_group_logs (offline experiments)

vrigal opened this pull request 4 months ago
Publish experiments to W&B from the CI

vrigal opened this pull request 4 months ago
en-tr translation quality feedback

selimsum opened this issue 4 months ago
Experiment with one stage training

eu9ene opened this issue 4 months ago
Experiment with data cleaning

eu9ene opened this issue 4 months ago
Release cherry picks 2

eu9ene opened this pull request 4 months ago
Add a train task

gregtatum opened this pull request 4 months ago
en-hr model does not always correctly distinguish between Indian and Native American

nordzilla opened this issue 4 months ago
Taskcluster train actions fail

eu9ene opened this issue 5 months ago
Support uploading GCP Taskcluster artifacts to W&B

eu9ene opened this issue 5 months ago
Taskcluster evaluation artifacts on GCP are missing an importer

eu9ene opened this issue 5 months ago
Bump disk for cefilter

eu9ene opened this pull request 5 months ago
[taskcluster:error] Error uploading artifact: S3 returned status code 400 which could be an intermittent issue

eu9ene opened this issue 5 months ago
Suppress the ruff import sorting behavior

gregtatum opened this pull request 5 months ago
Explore uploading models to Hugging Face

eu9ene opened this issue 5 months ago
[meta] Train low resource languages

gregtatum opened this issue 5 months ago
Create a Python package to use translation models

eu9ene opened this issue 5 months ago
WIP Patch stack with nllb and hplt importer work

gregtatum opened this pull request 5 months ago
Make sure square brackets instead of parentheses don't lead to wrong translations

marco-c opened this issue 5 months ago
Suffix W&B runs with task group ID for offline Taskcluster publication from GCP

vrigal opened this pull request 5 months ago
Add an action to rebuild pipeline toolchains and docker images

gabrielBusta opened this pull request 5 months ago
Change base image to trigger toolchain rebuilds

gabrielBusta opened this pull request 5 months ago
Change base image to trigger toolchain rebuilds

gabrielBusta opened this pull request 5 months ago
production and staging repositories share caches

bhearsum opened this issue 5 months ago
Change base image to trigger toolchain rebuilds

gabrielBusta opened this pull request 5 months ago
Perform a comprehensive testing before the final reuploading

eu9ene opened this issue 5 months ago
Add publishing to CI

eu9ene opened this issue 5 months ago
Change base image to trigger toolchain rebuilds

gabrielBusta opened this pull request 5 months ago
Filter monolingual synthesized distillation data with a fluency score

gregtatum opened this issue 5 months ago
Filter monolingual data based on fluency scores

gregtatum opened this issue 5 months ago
Clean up after bicleaner downloader

gregtatum opened this pull request 5 months ago
Rewrite merge mono and add support for an OPUS monolingual importer

gregtatum opened this pull request 5 months ago
Bump disk for student

eu9ene opened this pull request 5 months ago
Improve GPU utilization for "translate" tasks

eu9ene opened this issue 5 months ago
Fix restarting downloads

gregtatum opened this pull request 5 months ago
Improve GPU utilization in student training

eu9ene opened this issue 5 months ago
add 2tb gpu workers

bhearsum opened this pull request 5 months ago
fix: don't run evaluate tasks on pretrained models

bhearsum opened this pull request 5 months ago
Add a mono nllb build script

gregtatum opened this pull request 5 months ago
Expand out marian command arguments

gregtatum opened this pull request 5 months ago
Investigate removing teacher ensemble training

gregtatum opened this issue 5 months ago
restrict github-push taskcluster events to `main`

bhearsum opened this pull request 5 months ago
feat: add scaffolding and basic tests for taskgraph generation

bhearsum opened this pull request 5 months ago
temp: use cpu worker pool that uses gpu image to see if it works

bhearsum opened this pull request 5 months ago
train-student OSError: [Errno 28] No space left on device

eu9ene opened this issue 5 months ago
Figure out the behavior of OpusTrainer augmentation on student distillation gap

gregtatum opened this issue 5 months ago
Investigate improving en-lt student distillation by adding more data

gregtatum opened this issue 5 months ago
Reduce monolingual data for da-en to investigate distillation performance

gregtatum opened this issue 5 months ago