sillsdev/silnlp

Issues

• Update translate script to not retranslate existing books

  #396 opened 5 months ago
  0

• Evaluate the effectiveness of using chrF as the early stopping metric

  #395 opened 5 months ago
  0

• Adjust default learning rate for LoRA models

  #394 opened 5 months ago
  0

• Set up ClearML agents and queues for new 'Cheetah' server

  #393 opened 2 months ago
  1

• Set up ORCA clearml agent

  #392 opened 5 months ago
  1

• Set up GCP autoscaling in clearml

  #391 opened 5 months ago
  2

• Adafactor option is deprecated

  #390 opened 5 months ago
  2

• Translate script gives zero exit code on failure

  #389 opened 5 months ago
  3

• Fine tuning LLM for machine translation task

  #388 opened 5 months ago
  0

• Fix warning for gradient checkpointing by specifying value for use_reentrant

  #387 opened 5 months ago
  1

• Use non-CUDA base images for Docker images

  #386 opened a month ago
  0

• Switch to conda for virtual environment/dependency management

  #384 opened a month ago
  2

• Remove OpenNMT and related dependencies from SILNLP

  #382 opened 6 months ago
  0

• Orphan end markers are sometimes reported without a matching opening marker even though the opening marker exists.

  #379 opened 2 months ago
  3

• Explore alternative method for spinning off silnlp jobs on ClearML

  #378 opened 6 months ago
  0

• Back Translation language code management

  #377 opened 3 months ago
  0

• Use pre-calculated alignments efficiently.

  #373 opened 5 months ago
  1

• Try using auto_find_batch_size

  #371 opened 3 months ago
  2

• Upgrade huggingface transformers to 4.40.0 for FlashAttention2 support for NLLB

  #370 opened 6 months ago
  1

• Add ISO codes to pair alignment filenames, add stats about trainable and draftable verses

  #368 opened 3 months ago
  2

• Translate script - parsing error on cross-references (\xt ... \xt*) in section reference lines (\sr)

  #367 opened 2 months ago
  2

• Check whether data is being tokenized before calculating tokenization stats

  #365 opened 7 months ago
  0

• Error with preprocessing with the --stats option for certain files.

  #362 opened 7 months ago
  4

• Multiple Scores

  #360 opened 7 months ago
  2

• 12 Terrabytes of Experiments - clean it up

  #355 opened 6 months ago
  1

• collect_verse_counts: Count the number of non-empty books (rather than total books)

  #353 opened 7 months ago
  1

• Normalize text before checking for missing characters in the tokenizer

  #352 opened 7 months ago
  1

• Research default probability for adding missing characters to unigram tokenizers

  #348 opened 7 months ago
  0

• Upgrading to latest commit results in pyarrow conflict during poetry install

  #343 opened 7 months ago
  0

• Handling "low surrogate" characters in KBTFRA SFM files

  #340 opened 7 months ago
  1

• Handle BOM (Byte-Order Mark) on custom stylesheets

  #338 opened 2 months ago
  1

• Add support for chrF+, chrF++ as scoring metric (and early stopping criteria)

  #337 opened 7 months ago
  0

• Add option to use English key terms glosses

  #334 opened a month ago
  0

• When using multiple corpus pairs, key terms are pulled multiple times.

  #333 opened 2 months ago
  0

• When translating a passage with embedded paragraph markers ("\pm"), verses are output in the reverse order

  #330 opened 2 months ago
  1

• When translating on a USFM file with a table spanning multiple verses, the output USFM drops some verse markers and table formatting markers

  #329 opened 2 months ago
  5

• Translate as an optional part of the experiment workflow

  #321 opened 8 months ago
  0

• When calculating --stats with the preprocess function don't recalculate existing results and include all results in the final scores.

  #320 opened 8 months ago
  3

• Onboarding: add versification checks

  #448 opened 9 months ago
  0

• When calculating alignments preserve the isocode from the filename in the corpus-stats.csv file.

  #317 opened 6 months ago
  5

• Tokenizers parallelism warning

  #316 opened 4 months ago
  0

• More tokenization stats during preprocessing

  #311 opened 9 months ago
  0

• Clowder: Add ability to incorporate resources from GDrive folder

  #310 opened 8 months ago
  1

• French-to-vernacular drafting - KBT's are not included in training data

  #309 opened 4 months ago
  1

• Can't pull docker image

  #307 opened 9 months ago
  1

• Allow a configuration for inferencing of SFM files that may differ from project to project.

  #306 opened 19 days ago
  1

• "Poetry shell" failing to create a virtual environment on fresh install of SILNLP

  #305 opened 9 months ago
  2

• Include testing from the source text(s) where possible.

  #375 opened 10 months ago
  2

• Leverage Ethnologue data to get linguistic information on languages for on-boarding

  #303 opened 10 months ago
  0

• Wrap clowder in a GUI for non-technical use

  #302 opened 3 months ago
  0