PicoGPT
A complete GPT — autograd, multi-head attention, AdamW optimizer — trained from scratch
in your browser.
Small enough to fit in a QR code.
Fetching training data…
Step 0/500
Loss —
LR —
A complete GPT — autograd, multi-head attention, AdamW optimizer — trained from scratch
in your browser.
Small enough to fit in a QR code.
Fetching training data…