Deep Learning Workshop Repository

Welcome to the repository for the 2-Day Deep Learning Workshop that I recently completed. This repository contains the resources, code, and materials provided during the workshop, along with some of my own modifications and additions. 🎉📚✨

---

Repository Structure 🎯📁🔍

1. Day-wise Workshop Folders

Each day’s folder contains Colab notebook files with different deep learning tasks: 🧠📓💻

Day 1

• Image Classification with CNN: Implementation of Convolutional Neural Networks for image classification.
• Image Classification with VGG16: Transfer learning using the pre-trained VGG16 model for image classification.
• Emotion Recognition with LSTM: Implementation of LSTM networks for emotion recognition tasks using: 1-hot encoding

Day 2

• Emotion Recognition with LSTM: Implementation of LSTM networks for emotion recognition tasks using: Word2Vec embedding
• Fine-tuning BERT for Emotion Recognition: Code to fine-tune a BERT model for emotion recognition tasks.

---

2. Materials Folder

This folder includes: 📄📂✨

• Workshop Slides: All the slides presented during the workshop.
• Workshop Contents List: A comprehensive list of the topics covered during the two days.

---

3. Modified Codes Folder

This folder contains: 🛠️🖋️🚀

• My custom modifications to the provided codes, reflecting improvements, enhancements, or experiments I conducted based on workshop learnings.

---

4. To-Do Requirements Files

Two separate files detailing requirements for: 📝✔️📌

1. Task-Specific Adjustments:

   • Freezing convolutional layers and ensuring non-trainable parameters are properly handled.
   • Adding precision and recall calculations alongside F1 scores.
   • Including train loss in plots when using Weights and Biases (WandB).

2. BERT Fine-Tuning:

   • Identifying the most attentive tokens for each class in the test dataset by leveraging BERT's attention scores.
   • Aggregating and ranking token attention scores to determine their importance for specific classes.
   • Implementing a new method in the BERT_FineTuning class to extract and process attention-based insights for tokens.

---

5. Python Lightning Command Line Instruction File

A file with command-line instructions for running PyTorch Lightning tasks effectively. ⚡🐍💻

---

How to Use This Repository 🧑‍💻🛠️📖

1. Clone the repository:

   git clone https://github.com/SamiINReciept/deep-learning-workshop.git

2. Navigate to the specific folders to explore notebooks, materials, or modified codes.

---

Key Learnings from the Workshop 🌟📚💡

• Advanced techniques for image classification using CNNs and pre-trained models.
• Emotion recognition with LSTM networks leveraging various embedding techniques.
• Fine-tuning transformer-based models like BERT for NLP tasks.
• Practical implementation of PyTorch Lightning for streamlined deep learning experiments.

---

Acknowledgments 🤝🎓👏

Special thanks to the workshop organizers at CCDS.AI for their detailed explanations and valuable insights into deep learning concepts.

---

Contributions and Feedback 🌟💻✨

Feel free to explore, use, and modify the resources in this repository. Contributions, suggestions, and feedback are always welcome! Your input helps improve and expand this repository further.

---

License 📜🔓💡

This repository is licensed under the MIT License.