This project uses Tensorflow to implement a Neural LSTM Network that can classify sentiments present in purchase reviews from the top marketplaces in Brazil.
You can get about the project by reading the article on my blog.
This project was built with the following frameworks and libraries
- Tensorflow
- MLFlow Attempt |
- Numpy
conda env create -f conda.yaml
conda activate olist_sentiment_analisis
python main.py <options>
Options | Description | Default |
---|---|---|
--path_dataset | dataset path | |
--lr | learning rate | 0.001 |
--train_split | number of divide dataset train | 0.8 |
--random_state | random state | 42 |
--vocab_size | vocabulary size | 10000 |
--embedding_dim | number of embeddind dimension | 16 |
--max_length | max word length in sentence | 120 |
--batch_size | batch size | 128 |
--num_epochs | number of training steps | 5 |
--early_stopping_criteria | early stop criteria | 2 |
--dropout | dropout percentage | 0.3 |
--model_storage | model_storange | model_storage/lstm |
mlflow ui
mlflow models serve -m runs:/<run_id>/model