Archive - Thonk From First Principles

Why PyTorch is an amazing place to work... and Why I'm Joining Thinking Machines

In which I convince to you to join either PyTorch or Thinking Machines!

Mar 4 •

August 2024

FlexAttention: The Flexibility of PyTorch with the Performance of FlashAttention [external]

Freeing users from the software lottery tyranny of fused attention implementations.

Aug 7, 2024 •

April 2024

Strangely, Matrix Multiplications on GPUs Run Faster When Given "Predictable" Data! [short]

Great minds discuss flops per watt.

Apr 29, 2024 •

Solutions: What Shapes Do Matrix Multiplications Like?

Companion to https://www.thonking.ai/p/what-shapes-do-matrix-multiplications

Apr 8, 2024 •

What Shapes Do Matrix Multiplications Like? [medium]

Divining order from the chaos

Apr 1, 2024 •

February 2024

Supporting Mixtral in gpt-fast through torch.compile [short]

Long-form version of this tweet thread: https://twitter.com/cHHillee/status/1762269069351461196

Feb 26, 2024 •

and

#nojs-banner { position: fixed; bottom: 0; left: 0; padding: 16px 16px 16px 32px; width: 100%; box-sizing: border-box; background: red; color: white; font-family: -apple-system, "Segoe UI", Roboto, Helvetica, Arial, sans-serif, "Apple Color Emoji", "Segoe UI Emoji", "Segoe UI Symbol"; font-size: 13px; line-height: 13px; } #nojs-banner a { color: inherit; text-decoration: underline; } This site requires JavaScript to run correctly. Please turn on JavaScript or unblock scripts