kermitt2/delft

Issues

• Last tensorflow-addons

  #162 opened 10 months ago by lfoppiano
  1

• Print additional information when training

  #160 opened a year ago by lfoppiano
  3

• I am trying to make a model that reads problems and solutions and then it will generate text (when I give it a problem it will give the solutions) but i am stuck

  #159 opened a year ago by raneem199
  2

• Sub-tokenization with certain transformers

  #150 opened a year ago by lfoppiano
  22

• Support for Pre-trained ELMo Representations for Many Languages

  #155 opened a year ago by kermitt2
  3

• Transformer configuration should be passed to the tokenizer?

  #144 opened a year ago by lfoppiano
  4

• Hardcoded padding tokens

  #151 opened a year ago by lfoppiano
  3

• Allow uniform modification of nb_workers

  #139 opened 2 years ago by lfoppiano
  0

• Implement feature channel in classification

  #152 opened a year ago by lfoppiano
  0

• Bad performance to produce ELMo embedding

  #145 opened 2 years ago by SeanLee97
  2

• Comparison version 0.2.6 and 0.3.0 with scibert

  #132 opened 2 years ago by lfoppiano
  10

• Fix --output option to dump the model in a specific directory

  #124 opened 2 years ago by lfoppiano
  3

• add parameter to optionally output raw results in the evaluation

  #117 opened 3 years ago by lfoppiano
  8

• Print minimal configuration information when running training or train/eval

  #129 opened 2 years ago by lfoppiano
  3

• Classification and transformers

  #136 opened 2 years ago by lfoppiano
  3

• improve directory management for training/nfold-training

  #126 opened 2 years ago by lfoppiano
  0

• [suggestion] add transformer name / embedding names in the model name

  #130 opened 2 years ago by lfoppiano
  3

• Switch to TensorFlow 2.0 and tf.keras

  #87 opened 2 years ago by de-code
  12

• Load transformer config and tokenizer from disk when n>1 for nfold traninig

  #125 opened 2 years ago by kermitt2
  0

• Better support SentencePiece tokenizer(s)

  #128 opened 2 years ago by kermitt2
  0

• Training callbacks are ignored when using a BERT architecture for sequence labelling

  #96 opened 2 years ago by oterrier
  0

• Feature-based approach with BERT for seq. labelling is super slow

  #80 opened 2 years ago by kermitt2
  0

• Automatically download embeddings

  #79 opened 2 years ago by kermitt2
  1

• Eval metrics per class

  #119 opened 3 years ago by MahmoudAliEng
  1

• Command for saving the metrics in a file

  #118 opened 3 years ago by MahmoudAliEng
  2

• Allow multiple tokens per feature data row

  #99 opened 4 years ago by de-code
  3

• Unable to train NER using custom ELMO embedding model

  #113 opened 3 years ago by sathiyabalu89
  1

• sequence labelling, n-fold training should use separate preprocessors

  #102 opened 4 years ago by de-code
  1

• header training data - mismatching features columns?

  #81 opened 4 years ago by lfoppiano
  6

• Implement sliding window

  #90 opened 4 years ago by de-code
  6

• Annoying warning due to joblib

  #103 opened 4 years ago by oterrier
  2

• Cannot have 2 models with ELMo embeddings of 2 different languages at the same time

  #100 opened 4 years ago by oterrier
  0

• Suggestion: save preprocessor in a more transferable format

  #91 opened 4 years ago by de-code
  7

• <PAD> tags should be filtered out from the output of the Tagger

  #64 opened 5 years ago by oterrier
  17

• Find a way to disable the ELMo/BERT caching mechanism in "production" mode

  #66 opened 4 years ago by oterrier
  1

• Add a way to disable multiprocessing in Classifier training options as it is for Sequence

  #94 opened 4 years ago by oterrier
  0

• Some links are broken in the documentation

  #93 opened 4 years ago by lfoppiano
  1

• grobidTagger: make model an optional argument

  #88 opened 4 years ago by de-code
  0

• reader.py misses the the <EX_ENAMEX> annotated entities on LeMonde Corpus

  #61 opened 4 years ago by pjox
  3

• Sorting of fields in the sequence labelling evaluation report

  #60 opened 4 years ago by kermitt2
  1

• new multiprocessing parameter for Sequence and TrainingConfig

  #71 opened 4 years ago by bfreuden
  5

• sequenceLabelling.Trainer.train method fails if validation set is None

  #69 opened 4 years ago by bfreuden
  3

• Be able to pass an additional callbacks argument to the train method of Sequence/Classifier object

  #67 opened 4 years ago by oterrier
  3

• Incompatible arrays dimension when using ELMo and input is of length 1 (only 1 word)

  #63 opened 4 years ago by oterrier
  2

• Average precision/recall/f1 per label

  #47 opened 5 years ago by lfoppiano
  2

• --use-BERT option is ignored by nerTagger eval

  #48 opened 5 years ago by bfreuden
  1

• 'patience' training_config parameter is ignored by Trainer class

  #50 opened 5 years ago by bfreuden
  1

• bert-{lang} bert-base-{lang} discrepancy in Embeddings

  #52 opened 5 years ago by bfreuden
  1

• LMDB embeddings creation is very slow on spinning drive

  #54 opened 5 years ago by bfreuden
  2

• Error when model checkpointing model if f1 is not available (yet)

  #56 opened 5 years ago by bfreuden
  1