v0.4.1

0.4.1 (2019-10-01)

New:

• N/A

Changed:

• N/A

Fixed:

• Fix problem with text_classifier incorrectly using uncased_L-12_H-768_A-12 to build BERT model
  instead of multi_cased_L-12_H-768_A-12 when non-English language was detected.

v0.4.0

0.4.0 (2019-09-30)

New:

• Added multilingual support for text classification.
• Added experimental support for tf.keras. By default, ktrain will use standalone Keras.
  If os.environ['TF_KERAS'] is set, ktrian will attempt to use tf.keras.
  Some capabilities (e.g., predictor.explain for images) are not yet supported for tf.keras

Changed:

• When BERT is selected, check to make sure dataset is correctly preprocessed for BERT

Fixed:

• Fixed utils.bert_data_type and ensures it does more checks to validate BERT-style data

v0.3.1

0.3.1 (2019-09-19)

New:

• N/A

Changed:

• globally import tensorflow
• suppress tensorflow deprecation warnings from TF 1.14.0

Fixed:

• Resolved issue with text_classifier failing when BERT is selected and Preprocessor is supplied.

v0.3.0

0.3.0 (2019-09-17)

New:

• Support for sequence tagging with Bidirectional LSTM-CRF. Word embeddings can currently be either
  random or word2vec(cbow). If latter chosen, word vectors will be downloaded automaticlaly from Facebook fasttext site.
• Added ktra.text.texts_from_df function

Changed:

• Added FutureWarning in text.text_classifier, that preproc will be required argument in future.
• In text.text_classifier, when preproc=None, use the maximum feature ID to populate max_features.

Fixed:

• Fixed construction of custom_objects dictionary for BERT to ensure load_model works for
  custom BERT models
• Resolved issue with pretrained bigru models failing when max_features >= than total word count.

v0.2.5

0.2.5 (2019-08-27)

New:

• explain methods have been added to TextPredictor and ImagePredictor objects.
• TextPredictor.predict_proba and ImagePredictor.predict_proba_* convenience
  methods have been added.
• Added utils.is_classifier utility function

Changed:

• TextPredictor.predict method can now accept a single document as input instead of
  always requiring a list.
• Output of core.view_top_losses now includes the ground truth label of examples

Fixed:

• Fixed test of data loading

v0.2.4

0.2.4 (2019-08-20)

New:

• added additional tests of ktrain

Changed:

• Added classes argument to vision.images_from_folder. Only classes/subfolders
  matching a name in the classes list will be considered.

Fixed:

• Resolved issue with using learner.view_top_losses with BERT models.

v0.2.3

0.2.3 (2019-08-18)

New:

• N/A

Changed:

Fixed:

• Fixed issue with learner.validate and learner.predict failing when validation data is in
  the form of an Iterator (e.g., DirectoryIterator).

v0.2.2

0.2.2 (2019-08-16)

New:

• N/A

Changed:

• Added check in ktrain.lroptimize.lrfinder to stop training if learning rate exceeds a fixed maximum,
  which may happen when bad/dysfunctional model is supplied to learning rate finder.

Fixed:

• In ktrain.text.data.texts_from_folder function, only subfolders specified in classes argument
  are read in as training and validation data.

v0.2.1

0.2.1 (2019-08-15)

New:

• N/A

Changed:

• N/A

Fixed:

• Fixed error related to validation_steps=None in call to fit_generator in ktrain.core on Google Colab.

v0.2.0

0.2.0 (2019-08-12)

New:

• Support for pretrained BERT Text Classification

Changed:

• For Learner.lr_find, added optional max_epochs argument.
• Changed Learner.confusion_matrix to Learner.validate and added optional val_data argument.
  The use_valid argument has been removed.
• Removed pretrained_fpath argument to text.text_classifier. Pretrained word vectors are
  now downloaded automatically when 'bigru' is selected as model.

Fixed:

• Further cleanup of utils.is_iter function to use type check.