This repository aims to record my process of learning the DL and also I hope this notebook can help others who are interested in DL to know about the it.
author: Rui Wu
| NVIDIA-SMI | Driver | CUDA | GPU | Memory | Pwr |
|---|---|---|---|---|---|
| 552.22 | 552.22 | 12.4 | GeForce RTX 3050Ti | 4096MiB | 95W |
This notebook is structured into these parts: P1 Introduction to Deep Learning, P2 Introduction to Large Language Model
-
In
P1 Introduction to Deep Learning, the classic models/networks like MLP, CNN, ResNet, ViT will be introduced. And train them on the MNIST, FashionMNIST, CIFAR10 and CIFAR100. There are the results we will get in P1:model dataset accuracy epoch augmentation pre-train MLP MNIST 98.1% 20 baseline no MLP FashionMNIST 81.6% 20 baseline no AlexNet FashionMNIST 91.0% 30 baseline no ResNet CIFAR10 94.5% 200 AutoAugment, RandomErasing no ResNet50 CIFAR100 79.1% 200 AutoAugment, RandomErasing yes -
In
P2 Introduction to Large Language Model, I will give the code to understand how to handle text data (NLP knowledge), how to use Prompt to make LLM's inference better and other skills for reasoning. -
In
P3 Introduction to AI interpreter, I will introduce some tools to interprete kinds of phenomenon in Deep Learning.
The notebook is provided in ipynb format, compatible with Jupyter Notebook or Google Colab. Each file contains a comprehensive topic and a corresponding small project for practical understanding.
To replicate the environment, follow these simple steps:
conda create -n ldl python==3.8
conda activate ldl
pip install torch==2.1.1 torchvision==0.16.1 torchaudio==2.1.1 --index-url https://download.pytorch.org/whl/cu118
pip install -r requirements.txt- For any questions or recommendations regarding this notebook, please create an issue.
- Feel free to reach out to me via email for further discussion or feedback.
- If you'd like to connect, you can find me at Rutgers University.