TinyGrad Experiments
A series of small projects built with tinygrad to deepen my understanding of machine learning internals.
| Year | 2024 |
| Website | GitHub |
These experiments are all about learning by doing. Using tinygrad's concise codebase, I built gradient descent visualizations and toy networks to see how tensors, autodiff and optimizers work under the hood.
Challenge
Demystify deep learning mechanics without relying on heavyweight frameworks.
Action
Experimenting with tinygrad
Implemented small training loops and visualizations that expose each step of forward and backward passes.
Result
These hands-on exercises provided an intuitive grasp of gradients and model training, making complex concepts much more approachable.