kermitt2 / delft

a Deep Learning Framework for Text https://delft.readthedocs.io/

kermitt2/delft Issues

Last tensorflow-addons
Updated 8 months ago1
Print additional information when training
Updated a year ago3
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
Closed a year ago2
Sub-tokenization with certain transformers
Updated a year ago22
Support for Pre-trained ELMo Representations for Many Languages
Closed a year ago3
Transformer configuration should be passed to the tokenizer?
Closed 2 years ago4
Hardcoded padding tokens
Closed 2 years ago3
Allow uniform modification of nb_workers
Updated 2 years ago
Implement feature channel in classification
Updated 2 years ago
Bad performance to produce ELMo embedding
Closed 2 years ago2
Comparison version 0.2.6 and 0.3.0 with scibert
Closed 2 years ago10
Fix --output option to dump the model in a specific directory
Updated 2 years ago3
add parameter to optionally output raw results in the evaluation
Updated 2 years ago8
Print minimal configuration information when running training or train/eval
Closed 2 years ago3
Classification and transformers
Closed 2 years ago3
improve directory management for training/nfold-training
Updated 2 years ago
[suggestion] add transformer name / embedding names in the model name
Closed 2 years ago3
Switch to TensorFlow 2.0 and tf.keras
Closed 2 years ago12
Load transformer config and tokenizer from disk when n>1 for nfold traninig
Closed 2 years ago
Better support SentencePiece tokenizer(s)
Updated 2 years ago
Training callbacks are ignored when using a BERT architecture for sequence labelling
Closed 2 years ago
Feature-based approach with BERT for seq. labelling is super slow
Closed 2 years ago
Automatically download embeddings
Closed 2 years ago1
Eval metrics per class
Closed 3 years ago1
Command for saving the metrics in a file
Closed 3 years ago2
Allow multiple tokens per feature data row
Updated 3 years ago3
Unable to train NER using custom ELMO embedding model
Closed 3 years ago1
sequence labelling, n-fold training should use separate preprocessors
Updated 4 years ago1
header training data - mismatching features columns?
Closed 4 years ago6
Implement sliding window
Updated 4 years ago6
Annoying warning due to joblib
Closed 4 years ago2
Cannot have 2 models with ELMo embeddings of 2 different languages at the same time
Closed 4 years ago
Suggestion: save preprocessor in a more transferable format
Closed 4 years ago7
<PAD> tags should be filtered out from the output of the Tagger
Updated 4 years ago17
Find a way to disable the ELMo/BERT caching mechanism in "production" mode
Closed 4 years ago1
Add a way to disable multiprocessing in Classifier training options as it is for Sequence
Closed 4 years ago
Some links are broken in the documentation
Closed 4 years ago1
grobidTagger: make model an optional argument
Updated 4 years ago
reader.py misses the the <EX_ENAMEX> annotated entities on LeMonde Corpus
Closed 5 years ago3
Sorting of fields in the sequence labelling evaluation report
Closed 5 years ago1
new multiprocessing parameter for Sequence and TrainingConfig
Closed 5 years ago5
sequenceLabelling.Trainer.train method fails if validation set is None
Closed 5 years ago3
Be able to pass an additional callbacks argument to the train method of Sequence/Classifier object
Closed 5 years ago3
Incompatible arrays dimension when using ELMo and input is of length 1 (only 1 word)
Closed 5 years ago2
Average precision/recall/f1 per label
Closed 5 years ago2
--use-BERT option is ignored by nerTagger eval
Closed 5 years ago1
'patience' training_config parameter is ignored by Trainer class
Closed 5 years ago1
bert-{lang} bert-base-{lang} discrepancy in Embeddings
Closed 5 years ago1
LMDB embeddings creation is very slow on spinning drive
Closed 5 years ago2
Error when model checkpointing model if f1 is not available (yet)
Closed 5 years ago1