
Guides an agent to write, debug and validate cuTile GPU kernels in Python
- Open source
- 4K GitHub stars
- Developer tools


Guides an agent to write, debug and validate cuTile GPU kernels in Python

Adds autotuning with cached config selection to CuTile GPU kernels

Guides an agent through adding a new cuTile GPU kernel operator to TileGym

Guides an agent in converting cuTile GPU kernels to Triton and debugging the result

Guides an agent in porting cuTile Python GPU kernels to Julia cuTile.jl

Guides an agent to swap TileGym cuTile kernels into Hugging Face transformers models

Guides an agent to find CUDA memory, shared-memory race and sync bugs with Compute

Guides an agent through evidence-gated MoE performance tuning in Megatron Bridge

Guides an agent through GPU DataFrame work with NVIDIA cuDF and dask-cuDF

Guides measuring and speeding up a working Newton physics simulation without changing

Diagnoses and cuts cold-start compile time in Warp GPU code without changing results

Guides an agent through drafting, testing and tuning new agent skills with measured evals
The closest alternatives to Tilegym improve cutile kernel perf are Tilegym cutile python, Tilegym cutile autotuning, Tilegym adding cutile kernel, Tilegym converting cutile to triton and Tilegym converting cutile to julia.
Yes. Tilegym cutile python, Tilegym cutile autotuning, Tilegym adding cutile kernel have a free or open-source option.