TensorFlow 2 Binary Text Classifier

This contains the scripts and data needed to run the text classifier demo. The aim is to train a model which can identify whether a given text is in Tagalog or English.

This was used for the Google Solutions Challenge UP Diliman 2021: TensorFlow Workshop.

Accompanying slides can be found here: Slides Link

Recorded talk can be viewed here (1:40:14 mark): Tensorflow/Machine Learning Workshop

I. Dataset

A dataset was already prepared in en_tl_data.csv

II. Training and Inference

Refer to binary_text_classifier_demo.ipynb for training and inference

III. TensorFlow Serving

1. Make sure you've downloaded the trained model and take note of the path where it was saved

2. Run the ff on yout local machine

sudo docker pull tensorflow/serving

sudo docker run -t --rm -p 8501:8501 \
	-v "[path_to_model]:/models/tf_nnlm" \
	-e MODEL_NAME=tf_nnlm \
	--name tf-serving tensorflow/serving

e.g.
sudo docker run -t --rm -p 8501:8501 \
	-v "/home/abbydc/tf_nnlm:/models/tf_nnlm" \
	-e MODEL_NAME=tf_nnlm \
	--name tf-serving tensorflow/serving

to stop the tf-serving container:

sudo docker stop tf-serving

3.You may change the example in text_classifier_tf_serving.py and then run it with this command

python3 text_classifier_tf_serving.py