Merge to Main

CheatSheet.md

@@ -0,0 +1,2 @@
+# Cheat Sheet for PyTorch
+

README.md

@@ -1,8 +1,9 @@
 <div align="center">
-  <img src="torch.png" alt="Robot Image">
+  <img src="torchleet-llm.png" alt="Robot Image">
   <!-- <h1>TorchLeet</h1> -->
   <p align="center">
     🐦 <a href="https://twitter.com/charoori_ai">Follow me on Twitter</a> •
+    ➡️ <a href="https://github.com/Exorust/TorchLeet/tree/new-llm?tab=readme-ov-file#llm-set">Jump to LLMs!</a>
     📧 <a href="mailto:chandrahas.aroori@gmail.com?subject=LLM%20Cookbook">Feedback</a>
   </p>
   <p>
@@ -11,55 +12,118 @@
 </div>
 <br/>
 
-TorchLeet is a curated set of PyTorch practice problems, inspired by LeetCode-style challenges, designed to enhance your skills in deep learning and PyTorch.
+TorchLeet is broken into two sets of questions:
+1. **Question Set**: A collection of PyTorch practice problems, ranging from basic to hard, designed to enhance your skills in deep learning and PyTorch.
+2. **LLM Set**: A new set of questions focused on understanding and implementing Large Language Models (LLMs) from scratch, including attention mechanisms, embeddings, and more.
+3. 
+> [!NOTE]
+> Avoid using GPT. Try to solve these problems on your own. The goal is to learn and understand PyTorch concepts deeply.
 
 ## Table of Contents
-- [TorchLeet](#torchleet)
-  - [Table of Contents](#table-of-contents)
-  - [Question Set](#question-set)
-    - [🟢Easy](#easy)
-    - [🟡Medium](#medium)
-    - [🔴Hard](#hard)
-  - [Getting Started](#getting-started)
-    - [1. Install Dependencies](#1-install-dependencies)
-    - [2. Structure](#2-structure)
-    - [3. How to Use](#3-how-to-use)
+- [Question Set](#question-set)
+   - [🔵Basic](#basic)
+   - [🟢Easy](#easy)
+   - [🟡Medium](#medium)
+   - [🔴Hard](#hard)
+- [LLM Set](#llm-set)
+- [Getting Started](#getting-started)
+   - [1. Install Dependencies](#1-install-dependencies)
+   - [2. Structure](#2-structure)
+   - [3. How to Use](#3-how-to-use)
 - [Contribution](#contribution)
-- [Authors:](#authors)
 
 
 ## Question Set
 
-### 🟢Easy
-1. [Implement linear regression](https://github.com/Exorust/TorchLeet/blob/main/e1/lin-regression.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/e1/lin-regression_SOLN.ipynb)
-2. [Write a custom Dataset and Dataloader to load from a CSV file](https://github.com/Exorust/TorchLeet/blob/main/e2/custom-dataset.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/e2/custom-dataset_SOLN.ipynb) 
-3. [Write a custom activation function (Simple)](https://github.com/Exorust/TorchLeet/blob/main/e3/custom-activation.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/e3/custom-activation_SOLN.ipynb)
-4. [Implement Custom Loss Function (Huber Loss)](https://github.com/Exorust/TorchLeet/blob/main/e4/custom-loss.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/e4/custom-loss_SOLN.ipynb)  
-5. [Implement a Deep Neural Network](https://github.com/Exorust/TorchLeet/blob/main/e5/custon-DNN.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/e5/custon-DNN_SOLN.ipynb)  
-6. [Visualize Training Progress with TensorBoard in PyTorch](https://github.com/Exorust/TorchLeet/blob/main/e6/tensorboard.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/e6/tensorboard_SOLN.ipynb)  
-7. [Save and Load Your PyTorch Model](https://github.com/Exorust/TorchLeet/blob/main/e7/save_model.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/e7/save_model_SOLN.ipynb)  
+### 🔵Basic
+Mostly for beginners to get started with PyTorch.
+
+1. [Implement linear regression](torch/basic/lin-regression/lin-regression.ipynb) [(Solution)](torch/basic/lin-regression/lin-regression_SOLN.ipynb)
+2. [Write a custom Dataset and Dataloader to load from a CSV file](torch/basic/custom-dataset/custom-dataset.ipynb) [(Solution)](torch/basic/custom-dataset/custom-dataset_SOLN.ipynb) 
+3. [Write a custom activation function (Simple)](torch/easy/e3/custom-activation.ipynb) [(Solution)](torch/easy/e3/custom-activation_SOLN.ipynb)
+4. [Implement Custom Loss Function (Huber Loss)](torch/easy/e4/custom-loss.ipynb) [(Solution)](torch/easy/e4/custom-loss_SOLN.ipynb)  
+5. [Implement a Deep Neural Network](torch/easy/e5/custon-DNN.ipynb) [(Solution)](torch/easy/e5/custon-DNN_SOLN.ipynb)  
+6. [Visualize Training Progress with TensorBoard in PyTorch](torch/easy/e6/tensorboard.ipynb) [(Solution)](torch/easy/e6/tensorboard_SOLN.ipynb)  
+7. [Save and Load Your PyTorch Model](torch/easy/e7/save_model.ipynb) [(Solution)](torch/easy/e7/save_model_SOLN.ipynb) 
+10. Implement Softmax function from scratch
 
+---
+
+### 🟢Easy
+Recommended for those who have a basic understanding of PyTorch and want to practice their skills.
+1. [Implement a CNN on CIFAR-10](torch/medium/m2/CNN.ipynb) [(Solution)](torch/medium/m2/CNN_SOLN.ipynb)  
+2. [Implement an RNN from Scratch](torch/medium/m3/RNN.ipynb) [(Solution)](torch/medium/m3/RNN_SOLN.ipynb)  
+3. [Use `torchvision.transforms` to apply data augmentation](torch/medium/m4/augmentation.ipynb) [(Solution)](torch/medium/m4/augmentation_SOLN.ipynb)  
+4. [Add a benchmark to your PyTorch code](torch/medium/m5/bench.ipynb) [(Solution)](torch/medium/m5/bench_SOLN.ipynb)  
+5. [Train an autoencoder for anomaly detection](torch/medium/m6/autoencoder.ipynb) [(Solution)](torch/medium/m6/autoencoder_SOLN.ipynb)
+6. [Quantize your language model](torch/hard/h6/quantize-language-model.ipynb) [(Solution)](torch/hard/h6/quantize-language-model_SOLN.ipynb)
+7. [Implement Mixed Precision Training using torch.cuda.amp](torch/hard/h9/cuda-amp.ipynb) [(Solution)](torch/hard/h9/cuda-amp_SOLN.ipynb)
+
+---
 
 ### 🟡Medium 
-1. [Implement an LSTM](https://github.com/Exorust/TorchLeet/blob/main/m1/LSTM.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/m1/LSTM_SOLN.ipynb)  
-2. [Implement a CNN on CIFAR-10](https://github.com/Exorust/TorchLeet/blob/main/m2/CNN.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/m2/CNN_SOLN.ipynb)  
-3. [Implement parameter initialization for a CNN]() [(Solution)]()
-4. [Implement an RNN](https://github.com/Exorust/TorchLeet/blob/main/m3/RNN.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/m3/RNN_SOLN.ipynb)  
-5. [Use `torchvision.transforms` to apply data augmentation](https://github.com/Exorust/TorchLeet/blob/main/m4/augmentation.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/m4/augmentation_SOLN.ipynb)  
-6. [Add a benchmark to your PyTorch code](https://github.com/Exorust/TorchLeet/blob/main/m5/bench.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/m5/bench_SOLN.ipynb)  
-7. [Train an autoencoder for anomaly detection](https://github.com/Exorust/TorchLeet/blob/main/m6/autoencoder.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/m6/autoencoder_SOLN.ipynb)  
+These problems are designed to challenge your understanding of PyTorch and deep learning concepts. They require you to implement things from scratch or apply advanced techniques.
+1. [Implement parameter initialization for a CNN]() [(Solution)]()
+2. [Implement a CNN from Scratch]() 
+3. [Implement an LSTM from Scratch](torch/medium/m1/LSTM.ipynb) [(Solution)](torch/medium/m1/LSTM_SOLN.ipynb)  
+4. Implement AlexNet from scratch 
+5. Build a Dense Retrieval System using PyTorch
+6.  Implement KNN from scratch in PyTorch
+
+---
 
 ### 🔴Hard
-1. [Write a custom Autograd function for activation (SILU)](https://github.com/Exorust/TorchLeet/blob/main/h1/custom-autgrad-function.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/h1/custom-autgrad-function_SOLN.ipynb)
+These problems are for advanced users who want to push their PyTorch skills to the limit. They involve complex architectures, custom layers, and advanced techniques.
+1. [Write a custom Autograd function for activation (SILU)](torch/hard/h1/custom-autgrad-function.ipynb) [(Solution)](torch/hard/h1/custom-autgrad-function_SOLN.ipynb)
 2. Write a Neural Style Transfer  
-3. [Write a Transformer](https://github.com/Exorust/TorchLeet/blob/main/h3/transformer.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/h3/transformer_SOLN.ipynb)  
-4. [Write a GAN](https://github.com/Exorust/TorchLeet/blob/main/h4/GAN.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/h4/GAN_SOLN.ipynb)  
-5. [Write Sequence-to-Sequence with Attention](https://github.com/Exorust/TorchLeet/blob/main/h5/seq-to-seq-with-Attention.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/h5/seq-to-seq-with-Attention_SOLN.ipynb)  
-6. [Quantize your language model](https://github.com/Exorust/TorchLeet/blob/main/h6/quantize-language-model.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/h6/quantize-language-model_SOLN.ipynb)
-7. [Enable distributed training in pytorch (DistributedDataParallel)]
-8. [Work with Sparse Tensors]
-9. [Implement Mixed Precision Training using torch.cuda.amp](https://github.com/Exorust/TorchLeet/blob/main/h9/cuda-amp.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/h9/cuda-amp_SOLN.ipynb)
-10. [Add GradCam/SHAP to explain the model.](https://github.com/Exorust/TorchLeet/blob/main/h10/xai.ipynb) [(Solution)](https://github.com/Exorust/TorchLeet/blob/main/h10/xai_SOLN.ipynb)
+3. Build a Graph Neural Network (GNN) from scratch
+4. Build a Graph Convolutional Network (GCN) from scratch
+5. [Write a Transformer](torch/hard/h3/transformer.ipynb) [(Solution)](torch/hard/h3/transformer_SOLN.ipynb)  
+6. [Write a GAN](torch/hard/h4/GAN.ipynb) [(Solution)](torch/hard/h4/GAN_SOLN.ipynb)  
+7. [Write Sequence-to-Sequence with Attention](torch/hard/h5/seq-to-seq-with-Attention.ipynb) [(Solution)](torch/hard/h5/seq-to-seq-with-Attention_SOLN.ipynb)  
+8. [Enable distributed training in pytorch (DistributedDataParallel)]
+9. [Work with Sparse Tensors]
+10. [Add GradCam/SHAP to explain the model.](torch/hard/h10/xai.ipynb) [(Solution)](torch/hard/h10/xai_SOLN.ipynb)
+11. Linear Probe on CLIP Features
+12. Add Cross Modal Embedding Visualization to CLIP (t-SNE/UMAP)
+13. Implement a Vision Transformer
+14. Implement a Variational Autoencoder
+
+---
+
+## LLM Set
+
+**An all new set of questions to help you understand and implement Large Language Models from scratch.**
+
+Each question is designed to take you one step closer to building your own LLM.
+
+1. Implement KL Divergence Loss
+2. Implement RMS Norm
+3. [Implement Byte Pair Encoding from Scratch](llm/Byte-Pair-Encoding/BPE-q3-Question.ipynb) [(Solution)](llm/Byte-Pair-Encoding/BPE-q3.ipynb)
+4. Create a RAG Search of Embeddings from a set of Reviews
+5. Implement Predictive Prefill with Speculative Decoding
+6. [Implement Attention from Scratch](llm/Implement-Attention-from-Scratch/attention-q4-Question.ipynb) [(Solution)](llm/Implement-Attention-from-Scratch/attention-q4.ipynb)
+7. [Implement Multi-Head Attention from Scratch](llm/Multi-Head-Attention/multi-head-attention-q5-Question.ipynb) [(Solution)](llm/Multi-Head-Attention/multi-head-attention-q5.ipynb)
+8. [Implement Grouped Query Attention from Scratch](llm/Grouped-Query-Attention/grouped-query-attention-Question.ipynb) [(Solution)](llm/Grouped-Query-Attention/grouped-query-attention.ipynb)
+9. Implement KV Cache in Multi-Head Attention from Scratch
+10. [Implement Sinusoidal Embeddings](llm/Sinusoidal-Positional-Embedding/sinusoidal-q7-Question.ipynb) [(Solution)](llm/Sinusoidal-Positional-Embedding/sinusoidal-q7.ipynb)
+11. [Implement ROPE Embeddings](llm/Rotary-Positional-Embedding/rope-q8-Question.ipynb) [(Solution)](llm/Rotary-Positional-Embedding/rope-q8.ipynb)
+12. [Implement SmolLM from Scratch](llm/SmolLM/smollm-q12-Question.ipynb) [(Solution)](llm/SmolLM/smollm-q12.ipynb)
+13. Implement Quantization of Models
+    1.  GPTQ
+14. Implement Beam Search atop LLM for decoding
+15. Implement Top K Sampling atop LLM for decoding
+16. Implement Top p Sampling atop LLM for decoding
+17. Implement Temperature Sampling atop LLM for decoding
+18. Implement LoRA on a layer of an LLM
+    1.  QLoRA
+19. Mix two models to create a mixture of Experts
+20. Apply SFT on SmolLM 
+21. Apply RLHF on SmolLM
+22. Implement DPO based RLHF
+23. Add continous batching to your LLM
+24. Chunk Textual Data for Dense Passage Retreival
+25. Implement Lage scale Training => 5D Parallelism
 
 **What's cool? 🚀**
 - **Diverse Questions**: Covers beginner to advanced PyTorch concepts (e.g., tensors, autograd, CNNs, GANs, and more).
@@ -84,10 +148,8 @@ TorchLeet is a curated set of PyTorch practice problems, inspired by LeetCode-st
 **Happy Learning! 🚀**
 
 
-
-
 # Contribution
-Feel free to contribute by adding new questions or improving existing ones. Ensure that new problems are well-documented and follow the project structure.
+Feel free to contribute by adding new questions or improving existing ones. Ensure that new problems are well-documented and follow the project structure. Submit a PR and tag the authors.
 
 # Authors
 

Tricks.md

@@ -0,0 +1,9 @@
+# Tricks
+
+List of tricks in pytorch
+
+
+## TorchScript
+@torch.jit.scriptorch.jit.script
+
+## Unit Tests in PyTorch