Skip to content
 
 

Repository files navigation

TileFoundry

Hand the compiler to the agent.

Not an agent that becomes the compiler,
and not an agent plugged in as one of its passes.
The compiler stays a tool — the agent is simply the one holding it.

One prompt in  ·  612 tok/s out  ·  nobody in the loop


PyPI Coverage TileFoundry optimizations shipped to TileOPs License: MIT

Documentation · Quick Start · Examples

Latest News

  • 08/2026 🎉: TileFoundry 0.0.1 is on PyPI — the first public release.
  • 08/2026 📦: Four worked examples added — Qwen3-1.7B (tilelang), Qwen3.5-35B-A3B (tilelang), MiniCPM3-4B (CuTeDSL) and granite-4.0-h-small (CUDA C) — each one a real agent run kept whole, with the decode throughput it measured.

Quick Start

1 · Install

pip install tilefoundry    # needs Python 3.12 or newer
tilefoundry                # check the install: the commands an agent will ask

This run also needs one NVIDIA GPU, pip install tilelang, the published Qwen/Qwen3-1.7B checkpoint on disk (3.8 GB), and a coding agent started in an empty directory.

2 · Hand it the prompt

There is no API to learn first. Give your coding agent this, with a checkpoint directory of your own:

Get real tokens out of Qwen3-1.7B on TileFoundry, and make it fast.
Weights and config: <checkpoint directory>
Backend: tilelang.

Everything about TileFoundry is to be asked of the `tilefoundry` command -- do not
ask a person, do not go looking elsewhere. The model itself is yours to research.

Done when this runs from outside, prints the continuation, and reports a
tokens-per-second number measured over the whole generation:

    python run.py \
        --prompt "Write a detailed explanation of how a GPU executes a matrix multiplication." \
        --max-new-tokens 2048

Measure over a long generation -- 2048 new tokens, more than 2000 characters of
text. A 32-token sample is too short for the number to mean anything.

That is the whole input — nothing under it is written by hand.

3 · Come back in two hours

Claude Opus 5 at xhigh reasoning effort worked 2.1 hours and 177 tool calls without a single interaction, and left a run.py behind — it prints the continuation, and the number it measured: 612 tok/s on one H200.

python run.py --ckpt <checkpoint directory> \
    --prompt "Write a detailed explanation of how a GPU executes a matrix multiplication." \
    --max-new-tokens 2048

The first run compiles the kernels — once, a few minutes.

Where to go next

That run is kept whole in examples/qwen3_1_7b-tilelang/, with three more beside it. The specifications are meant to be argued with: open an issue, or start from docs/develop.md.

License

This project is licensed under the MIT License.

Releases

Packages

Contributors

Languages