🚀 NLP Model Fine-Tuning, Text Classification, and Text Generation

This repository demonstrates multiple Natural Language Processing (NLP) tasks using Hugging Face Transformers. It includes:

• Fine-Tuning a sentiment analysis model (siebert/sentiment-roberta-large-English).
• Text Classification using (facebook/bart-large-mnli, nlptown/bert-base-multilingual-uncased-sentiment, cardiffnlp/twitter-roberta-base-sentiment).
• Text Generation using (facebook/bart-large-cnn). Each task has been implemented using Google Colab, and the respective notebook links are included.

🛠️ 1. Fine-Tuning Sentiment RoBERTa Large (English)

Model Used:

• siebert/sentiment-roberta-large-English – A RoBERTa-based sentiment analysis model. Dataset:
• nvidia/OpenCodeReasoning – Used for fine-tuning. Fine-Tuning Process:
• Model Loading:
• The sentiment-roberta-large-English model was loaded from Hugging Face.
• Dataset Preprocessing:
• Tokenization of the dataset using the model’s tokenizer.
• Formatting the dataset into Tensor format for training.
• Training:
• Fine-tuned using Google Colab’s GPU.
• Used AdamW optimizer and a suitable learning rate.
• Set training epochs and batch size for efficient training.
• Logged training loss and monitored model improvements.
• Evaluation:
• Validated the fine-tuned model on test data.
• Generated accuracy and loss metrics for evaluation. 🔗 Colab Notebook: https://colab.research.google.com/drive/11cZyqShYyryehUhWXbkjvNYU7MaTxWR4?usp=drive_link

🏷️ 2. Text Classification

Models Used:

• facebook/bart-large-mnli – Zero-shot classification based on natural language inference.
• nlptown/bert-base-multilingual-uncased-sentiment – Multilingual sentiment analysis.
• cardiffnlp/twitter-roberta-base-sentiment – Sentiment classification specialized for Twitter data. Implementation Steps:
• Model Selection:
• Used Hugging Face pipeline to load models.
• Preprocessing:
• Tokenized input text and prepared batches for prediction.
• Prediction:
• Performed zero-shot classification with facebook/bart-large-mnli.
• Conducted sentiment analysis with BERT-based models. Example Usage: from transformers import pipeline

classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli") result = classifier("I love this movie!", candidate_labels=["positive", "negative", "neutral"]) print(result)

 🔗 Colab Notebook: https://colab.research.google.com/drive/1erBBnPC3G1oErvunXLuB6bzMU3Ra9RGT?usp=drive_link

✍️ 3. Text Generation

Model Used:

• facebook/bart-large-cnn – Used for summarization. Implementation Steps:
• Model Loading:
• Loaded facebook/bart-large-cnn from Hugging Face.
• Text Preprocessing:
• Formatted input text properly for summarization.
• Text Generation:
• Generated a summarized output using the model. Example Usage: from transformers import pipeline

summarizer = pipeline("summarization", model="facebook/bart-large-cnn") summary = summarizer("Your long article or paragraph goes here.", max_length=50, min_length=20, do_sample=False) print(summary)

 🔗 Colab Notebook: https://colab.research.google.com/drive/1erBBnPC3G1oErvunXLuB6bzMU3Ra9RGT?usp=drive_link

🔧 Dependencies

Before running the notebooks, install the necessary dependencies: pip install transformers datasets torch hugging face

📌 Credits

• Base Models:
• siebert/sentiment-roberta-large-English
• facebook/bart-large-mnli
• nlptown/bert-base-multilingual-uncased-sentiment
• cardiffnlp/twitter-roberta-base-sentiment
• facebook/bart-large-cnn
• Dataset: Nvidia/OpenCodeReasoning
• Framework: Hugging Face Transformers