README

Content locked

The catalog is public. Reading this content requires approved access.

hw7

Deep Learning with CNNs and Language Models

Overview

1. Environment Setup

1.1 Install Conda

On CSL

On Your Machine

1.2 Create Environment for HW7

2. Dataset: CIFAR100

2.1 Helper Code

Backup Setup

3. Program Specification

3.1 Creating LeNet-5 (20 points)

3.2 Count the Number of Trainable Parameters of LeNet-5 (15 points)

3.3 Training LeNet-5 under Different Configurations (35 points)

Evaluation

Required Configurations

4. Training nanoGPT on Shakespeare Data and Text Generation (15 points)

4.1 Overview

4.2 Setup

4.3 Model Training

Option 1: Modify the Config File

Option 2: Specifying All Hyperparameters

4.4 Text Generation (Inference)

5. DistilGPT2 Supervised Fine-Tuning and Text Generation (15 points)

5.1 Overview

5.2 Learning Goals

5.3 Dataset and Setup

5.4 What You Will Do

5.5 How to Run

6. Help with Training

6.1 Training on CSL

7. Submission Notes