Sentiment Analysis model is built using pre-trained BERT transformer large scale language learnings and analysed smile annotations dataset using PyTorch Framework. The architecture is for multi-class classification. Exploratory data analysis(EDA) is done with loading tokenizer and further encoding data. Data loaders are created to facilitate Batch processing followed up with setting up of Optimizer and Scheduler to control the training of model.
The Performance metrics were designed for model followed by creating a Training loop to control PyTorch fine tuning of BERT acceleration. The pre-trained fine tuned model was loaded and it's performance is evaluated achieving good accuracy.