README.md

Companion artifact for [_GPU Forecasters: Language Models as Selective Surrogates for Kernel Runtime Optimization_](https://arxiv.org/abs/2605.31464). Code: [codezakh/gpu-surrogates](https://github.com/codezakh/gpu-surrogates).

LoRA adapter for [`openai/gpt-oss-20b`](https://huggingface.co/openai/gpt-oss-20b)
fine-tuned with the correctness + CRPS reward to forecast kernel speedups.

## Loading

```python
from peft import AutoPeftModelForCausalLM
model = AutoPeftModelForCausalLM.from_pretrained("codezakh/gpu-forecasters-gpt-oss-20b-correctness-crps")
```

## Training data

[`codezakh/gpu-forecasters-rl-training-pool`](https://huggingface.co/datasets/codezakh/gpu-forecasters-rl-training-pool).

## Reproducing

See `runbook/02_train_surrogate.py` in the paper repo
([codezakh/gpu-surrogates](https://github.com/codezakh/gpu-surrogates)).

## Citation

```bibtex
@article{khan2026gpuforecasters,
  title={GPU Forecasters: Language Models as Selective Surrogates for Kernel Runtime Optimization},
  author={Khan, Zaid and Chen, Justin Chih-Yao and Cho, Jaemin and Stengel-Eskin, Elias and Bansal, Mohit},
  journal={arXiv preprint arXiv:2605.31464},
  year={2026}
}
```