DLND

Deep Learning Nanodegree from Udacity

Contents

Assignments

Part 3: Convolutional Neural Networks

• Lecture 3: Transfer Learning
  • Transfer Learning
• Lecture 4: Weight Initialization
  • Weight Initialization
• Lecture 5: Autoencoder
  • Simple Autoencoder
  • Convolutional Autoencoder: Only for CPU and CUDA. XPU has memory leak issue.
  • Denoising Autoencoder: Only for CPU and CUDA. XPU has memory leak issue.

Part 4: Recurrent Neural Networks

• Lecture 5: Embeddings & Word2Vec
  • Skip Grams
• Lecture 8: Attention
  • Attention Basics

Part 5: Generative Adversarial Networks

• Lecture 1: Basic GAN
  • MNIST GAN
• Lecture 2: Deep Convolutional GANs
  • Batch Normalization
  • Generating Street View Home Number (SVHN): Only for CPU and CUDA. XPU has memory leak issue.
• Lecture 4: CycleGAN
  • Summer to Winter Image Transformation: Only for CPU and CUDA. XPU has memory leak issue.

Part 6: Deploying Model

Inference on CPU only (not using XPU or CUDA).

• Lecture 2: Building a Model using SageMaker
  • Predicting Boston Housing - XGBoost - Batch Transform - High Level
  • Sentiment Analysis - XGBoost - Batch Transform
• Lecture 3: Deploying and Using a Model
  • Predicting Boston Housing - XGBoost - Deploy - High Level
  • Sentiment Analysis - XGBoost - Web App

Projects

• Project 1: Predicting Bike-Sharing Patterns
• Project 2: Classifying Dog's Breed
• Project 4: Generating TV Script: Only for CPU and CUDA. 'aten::_thnn_fused_lstm_cell' is not currently implemented for the XPU.
• Project 5: Generating Face
• Project 7: Deploying Sentiment Analysis Model

Setup

Intel GPU

Please follow PyTorch 2.6 Prerequisites for Intel GPUs article to install Intel GPU driver and deep learning essentials.

Tested on the following hardware specification and software version.

Hardware Specification

• CPU: Intel® Core™ Ultra 9 Processor 285K
• CPU Cores: 24 (8 Performance-cores and 16 Efficient-cores)
• CPU Threads: 24
• Memory: 32 GiB
• GPU: Intel® Arc™ A770 Graphics
• GPU Memory: 16 GiB

Software Version

• Ubuntu 24.04.1 LTS
• Intel Deep Learning Essentials 2025.0.1-25
• Python 3.12.3
• PyTorch 2.6.0+xpu
• TorchVision 0.21.0+xpu
• opencv-python 4.10.0.84
• NumPy 2.1.2
• Matplotlib 3.10.0
• Pandas 2.2.3
• scikit-learn 1.6.0

Install Requirements

Create virtual environment.

$ python3 -m venv pytorch_arc_env
$ source pytorch_arc_env/bin/activate
$ python -m pip install --upgrade pip

Install PyTorch and other required packages.

$ pip install torch==2.6 torchvision torchaudio --index-url https://download.pytorch.org/whl/test/xpu
$ cd DLND
$ pip install --upgrade -r requirements.txt

Test Run

Activate virtual environment and setup variables.

$ source pytorch_arc_env/bin/activate

$ source /opt/intel/oneapi/compiler/2025.0/env/vars.sh
$ source /opt/intel/oneapi/umf/0.9/env/vars.sh
$ source /opt/intel/oneapi/pti/0.10/env/vars.sh

Detect GPU.

$ python -c "import torch; print(torch.xpu.is_available())"

True

Run notebooks.

$ cd DLND
$ jupyter lab