GoEmotions-PyTorch-TensorFlow

This project focuses on classifying emotions from text using the DistilBERT model, implemented in both PyTorch and TensorFlow. The dataset used is the "GoEmotions" dataset, which contains a variety of emotional labels.

Features:

1. Data Loading and Preprocessing:

   • Loads the "GoEmotions" dataset using the Hugging Face datasets library.
   • Filters the data to remove multi-label entries.
   • Maps labels to a string format for better visualization and understanding.

2. Data Visualization:

   • Visualizes class distribution using bar charts to observe any data imbalances.

3. Model Training:

   • Implements DistilBERT for sequence classification in both PyTorch and TensorFlow frameworks.
   • Trains the model on the preprocessed dataset, using accuracy as the evaluation metric.
   • Includes callbacks to save the model to the Hugging Face Hub and to monitor metrics during training.

4. Evaluation:

   • Evaluates the trained models on the validation set and displays a Confusion Matrix to visualize model performance across different emotions.

5. Emotion Prediction:

   • The model can classify user input into predefined emotion categories using a text classification pipeline.
   • A command-line interface allows users to interact with the model, input sentences, and receive emotion predictions.

You can use these models for emotion classification in your own projects.

PyTorch Model:

To use the PyTorch model for emotion classification, you can use the following code snippet:

from transformers import pipeline

pipe = pipeline("text-classification", model="tuhanasinan/go-emotions-distilbert-pytorch")

TensorFlow Model:

To use the TensorFlow model for emotion classification, you can use the following code snippet:

from transformers import pipeline

pipe = pipeline("text-classification", model="tuhanasinan/go_emotions-distilbert-tensorflow")

Links

Here are the links to the dataset and pre-trained and fine-tuned models used in this project:

• Dataset: GoEmotions Dataset
• Distilbert Model:DistilBERT
• PyTorch Model: go-emotions-distilbert-pytorch
• TensorFlow Model: go-emotions-distilbert-tensorflow