I know the feeling – you’re excited to dive into PyTorch, but you’re not sure where to start or how to structure your learning. You’ve got the basics of machine learning down, and you’ve even implemented some algorithms from scratch in plain Python. But when it comes to PyTorch, you’re feeling stuck.
You’re not alone. I’ve been there too, and I’ve learned that the key to making progress is to find a balance between theory, practice, and project-based learning.
## Don’t Just Follow Tutorials
Tutorials are great for getting started, but they can also hold you back if you’re not careful. If you only follow tutorials, you’ll never really understand what’s happening under the hood. You’ll just be copying and pasting code without understanding the why behind it.
## Understand the Concepts
On the other hand, if you spend too much time trying to understand every concept before touching bigger projects, you’ll never actually get to the projects. You’ll be stuck in theory land, and you’ll never get to apply what you’ve learned.
## The Right Balance
So what’s the right balance? For me, it’s about understanding the basics of PyTorch – tensors, autograd, optimizers, etc. – and then applying that knowledge to small projects. As you work on these projects, you’ll naturally encounter more advanced concepts, and you can learn them as you go.
## From Scratch
How much ‘from scratch’ coding should you do in PyTorch? I think it’s essential to do some coding from scratch, but not too much. You want to understand how PyTorch works, but you don’t want to reinvent the wheel. Start with simple projects, and gradually work your way up to more complex ones.
## Final Thought
Learning PyTorch takes time and practice. Don’t be discouraged if you don’t understand something at first. Keep at it, and you’ll get there. And remember, the goal is to build meaningful projects and understand what’s happening under the hood.
*Further reading: [PyTorch tutorials](https://pytorch.org/tutorials/)*