Resources & References

Duration: 5 min

For Mac Users (Apple Silicon)

Official Documentation

Apple Silicon Resources

For Everyone Else (Google Colab)

Google Colab Setup

Google Colab provides free GPU access in the cloud. Perfect for learning PyTorch without buying hardware.

Getting Started:

Go to colab.research.google.com
Click "New notebook"
Enable GPU: Runtime → Change runtime type → GPU

Install PyTorch:

!pip install torch torchvision torchaudio

Example Colab Notebook:

import torch

# Check GPU
print(f"GPU available: {torch.cuda.is_available()}")
print(f"GPU name: {torch.cuda.get_device_name(0)}")

# Create tensor on GPU
x = torch.randn(1000, 1000, device='cuda')
y = torch.randn(1000, 1000, device='cuda')

# Matrix multiplication on GPU
z = torch.matmul(x, y)
print(f"Result shape: {z.shape}")

Colab Limitations:

Free tier: ~12 GPU hours/month
Session timeout: 12 hours
Limited to 1 GPU
Perfect for learning, not production

Colab Resources:

General PyTorch Resources

Learning

PyTorch Official Tutorials
Deep Learning with PyTorch
PyTorch Lightning - Simplified training
Hugging Face Transformers - Pre-trained models

Datasets

CIFAR-10 - Image classification
ImageNet - Large-scale image dataset
MNIST - Handwritten digits
Hugging Face Datasets - Thousands of datasets

Tools

Jupyter Notebook - Interactive development
VS Code - Code editor
Conda - Package management
Git - Version control

Community

Choosing Your Setup

Setup	Cost	Speed	Setup Time	Best For
Google Colab	Free (limited)	Fast GPU	2 min	Learning, experimentation
MacBook Pro M1/M2	$1,200-2,500	5-6x CPU	5 min	Local development, privacy
AWS/GCP GPU	$0.30-1.00/hr	Very fast	15 min	Production training
Local CPU	$0	Slow	5 min	Fundamentals, no budget

Next Steps

After completing this course:

Try in Colab: Run the code examples in Google Colab
Build a Project: Apply PyTorch to your own dataset
Explore Transfer Learning: Fine-tune pre-trained models
Learn MLOps: Deploy models to production
Join the Community: Share your work and learn from others

Troubleshooting

MPS not available?

Ensure macOS 12.3+
Update PyTorch: pip install --upgrade torch
Check with: torch.backends.mps.is_available()

Out of memory?

Reduce batch size
Use mixed precision (float16)
Use gradient accumulation
Monitor with torch.mps.current_allocated_memory()

Slow training?

Ensure model is on GPU/MPS device
Check with: next(model.parameters()).device
Profile with PyTorch Profiler
Use mixed precision for 1.5-2x speedup

No GPU access?

Use Google Colab (free)
Use CPU (slower but works)
Rent cloud GPU (AWS/GCP/Azure)

Feedback

Have questions or suggestions? Join our community:

Happy learning! 🚀