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

Simplify translate mono kind files

gregtatum opened this pull request 5 months ago
Normalize punctuation marks in parallel data

gregtatum opened this issue 5 months ago
Investigate using LLMs for data augmentation

marco-c opened this issue 5 months ago
Investigate using LLMs to generate training data

marco-c opened this issue 5 months ago
Integrate datasets used for LLM training as monolingual datasets

marco-c opened this issue 5 months ago
English to Serbian has low quality of the teacher models

eu9ene opened this issue 6 months ago
Improve automatic quality evaluation

eu9ene opened this issue 6 months ago
Process alignments in chunks

eu9ene opened this pull request 6 months ago
Increase disk size for merge-translated

eu9ene opened this pull request 6 months ago
fix: invalidate caches when fetch, docker, or toolchain tasks change

bhearsum opened this pull request 6 months ago
merge-translated-el-en keeps restarting

eu9ene opened this issue 6 months ago
Add a link to W&B dashboard

eu9ene opened this pull request 6 months ago
Add a cleaning rule for URL names, such as Amazon.com -> Amazon.it

gregtatum opened this issue 6 months ago
Retrain old models with robustness fixes

gregtatum opened this issue 6 months ago
English to Lithuanian did not meet our quality bar

gregtatum opened this issue 6 months ago
Delete temporary files after successfully generating alignments

bhearsum opened this pull request 6 months ago
Multiply comet metric by 100 before publication

La0 opened this pull request 6 months ago
Check shortlist for CJK

eu9ene opened this issue 6 months ago
Support data import for CJK

eu9ene opened this issue 6 months ago
Check alignments for CJK

eu9ene opened this issue 6 months ago
Investigate OpusTrainer compatibility for CJK

eu9ene opened this issue 6 months ago
Check Bicleaner-AI models for CJK

eu9ene opened this issue 6 months ago
Check decoding for CJK

eu9ene opened this issue 6 months ago
Check training for CJK

eu9ene opened this issue 6 months ago
Check evaluation procedure for CJK

eu9ene opened this issue 6 months ago
Investigate issues with SentencePiece vocabulary for CJK

eu9ene opened this issue 6 months ago
Implement corpus specific fixes for CJK

eu9ene opened this issue 6 months ago
Support dataset desegmentation for CJK

eu9ene opened this issue 6 months ago
Support CJK in OpusCleaner

eu9ene opened this issue 6 months ago
Implement convertion between Chinese Traditional and Simplified

eu9ene opened this issue 6 months ago
Support CJK in find_corpus and config generator

eu9ene opened this issue 6 months ago
Process alignments in chunks

eu9ene opened this issue 6 months ago
chore: bump taskgraph to 9.2.0

bhearsum opened this pull request 6 months ago
Improve translation of social posts

eu9ene opened this issue 6 months ago
Improve translation of URLs

eu9ene opened this issue 6 months ago
Create Sardinian config

gregtatum opened this pull request 6 months ago
W&B runs published from GCP experiments should be suffixed with the Task Group ID when possible

vrigal opened this issue 6 months ago
Evaluate other GPU types

eu9ene opened this issue 6 months ago
fix: pre-download fast text model in bicleaner.sh

gabrielBusta opened this pull request 6 months ago
COMET results are not visible on custom charts

eu9ene opened this issue 6 months ago
Consider using backward-forward translation for knowledge distillation

eu9ene opened this issue 6 months ago
Duplicate runs in W&B

eu9ene opened this issue 6 months ago
start_stage often reruns amost all "evaluate" tasks

eu9ene opened this issue 6 months ago
Use unique run names in Weight & Biases

vrigal opened this pull request 6 months ago
Support multilingual models

eu9ene opened this issue 6 months ago
debugging: use d2g on bhearsum's special worker type

bhearsum opened this pull request 6 months ago
GPUs stopped working on bicleaner-ai for id-en

eu9ene opened this issue 6 months ago
One of the teachers for el-en diverged

eu9ene opened this issue 6 months ago
Add a check that there are visible GPUs

gregtatum opened this pull request 6 months ago
Student alignments fail for en-uk

eu9ene opened this issue 6 months ago
Publish Marian/OpusTrainer configuration YAMLs and dataset statistics

vrigal opened this pull request 6 months ago
Improve usability of running selected tasks

eu9ene opened this issue 6 months ago
Do not use aggressive dash splitting in tokenization

eu9ene opened this pull request 6 months ago
Switch aligning student corpus to fast_align

eu9ene opened this pull request 6 months ago
Some metric tables are empty

eu9ene opened this issue 6 months ago
Process alignments in chunks

eu9ene opened this pull request 6 months ago
Logs don't load on OOM

eu9ene opened this issue 6 months ago
Exclude start stage tasks from existing tasks

eu9ene opened this pull request 6 months ago
fix: shorten non-URL dataset names that are more than 50 characters (fixes #654)

bhearsum opened this pull request 6 months ago
Can't restart the pipeline to run distillation from where it left

eu9ene opened this issue 6 months ago
switch to generic worker for all tasks

bhearsum opened this issue 6 months ago
Update ci config

eu9ene opened this pull request 6 months ago
Group logs online evals

vrigal opened this pull request 6 months ago
Lock dependencies for tracking together with other task requirements

eu9ene opened this issue 6 months ago
Fix poetry lock

eu9ene opened this pull request 6 months ago
Cherrypickin' ac1ee64 to release

gabrielBusta opened this pull request 6 months ago
Expose Takscluster task owner as author for Weight & Biases publication

La0 opened this pull request 6 months ago
Optimize alignments

eu9ene opened this pull request 7 months ago
Configure generic-worker (d2g) pools for translation alignment tasks

gabrielBusta opened this pull request 7 months ago
Cherry pick commits from release

gregtatum opened this pull request 7 months ago
WIP switch GPU workers to image that uses multi engine generic worker

bhearsum opened this pull request 7 months ago
Numbers are altered by Finnish -> English translation

Nezz opened this issue 7 months ago
Sync main with release

eu9ene opened this pull request 7 months ago
Publish group logs evals in online mode

eu9ene opened this issue 7 months ago
Publish OpusTrainer and Marian logs to W&B logs tab

vrigal opened this pull request 7 months ago
Use pip-compile for tracking dependencies

La0 opened this pull request 7 months ago
Add support for bs, id, sr, tr, vi from OpusCleaner

gregtatum opened this issue 7 months ago
Use a forked OpusCleaner in `release`

gregtatum opened this pull request 7 months ago
Fix task expiration issue

eu9ene opened this pull request 7 months ago
ERROR - `task.dependencies` references tasks that expires before `task.deadline`

eu9ene opened this issue 7 months ago
[Experiment] Optimize aln

eu9ene opened this pull request 7 months ago
tests fail with numpy 2.0

bhearsum opened this issue 7 months ago
Only 3 evaluation datasets are visible in custom charts

eu9ene opened this issue 7 months ago
Remove `find_upstreams` transform in favour of `from-deps`

bhearsum opened this pull request 7 months ago
Refactor b-cpu-xlargedisk worker pools to allow for experimentation w…

bhearsum opened this pull request 7 months ago
Bump memory for shortlist

eu9ene opened this pull request 7 months ago
Investigate multilingual models for similar language groups

gregtatum opened this issue 7 months ago
Override the existing_tasks explicitly provided in the action's input

gabrielBusta opened this pull request 7 months ago
allow for more than 99 training datasets (fixes #653)

bhearsum opened this pull request 7 months ago
Serbian is digraphic with both Latin and Cyrllic which will cause some issues for training

gregtatum opened this issue 7 months ago
collect_mono failed with Read error (39) : premature end

eu9ene opened this issue 7 months ago
Marian error in translate-corpus

eu9ene opened this issue 7 months ago
Configs stage2

eu9ene opened this pull request 7 months ago
Make mono shuffling deterministic to fix caching issues

eu9ene opened this pull request 7 months ago
Replace sorting with python

eu9ene opened this pull request 7 months ago
Make merging back-translations more reliable

eu9ene opened this issue 7 months ago
Refactor b-cpu-xlargedisk worker pools to allow for experimentation with different configurations

gabrielBusta opened this pull request 7 months ago
fix: Don't abort a training task entirely if a 4xx error is encountered when fetching an artifact from a previous run

bhearsum opened this pull request 7 months ago
Bump taskcluster-taskgraph to 8.2.0 in poetry

bhearsum opened this pull request 7 months ago
fix: Don't abort a training task entirely if a 4xx error is encountered when fetching an artifact from a previous run (fixes #667)

bhearsum opened this pull request 7 months ago