Version 3.1.11

• Updating celery
• Updating Readme
• Adjusting limit of tokens in BERT model training to 384 to avoid memory overflow
• Dynamically reducing number of epochs during training based on the number of sentences to avoid overfitting and improve training time

Version 3.1.9

• Fixing bert_multilang model

Version 3.1.8

• Adding interpreters automatic cache cleaning

Version v3.1.7

Updating bothub-nlp-celery package

• New queues pattern
• New pt_br spacy_suggestion model

Bump version v3.1.5

Merge pull request #295 from bothub-it/staging

Staging

New Version NLP Refactored

Merge pull request #167 from Ilhasoft/staging

Staging

Refactored NLP

Merge pull request #160 from Ilhasoft/feature/analyse_char

Added analyze char

Evaluation process

This version contains the running of the evaluation process by generating reports and saving logs to the database.

Use name entities

This version use name entities to switch over Spacy NER in the pipeline.

Bug fixes

Fixes

• Update the bothub-engine dependency to version 1.19.1;
• Fix on the labels tokenizations;