# AI-CUDA-Engineer-Archive by Sakana AI: AI Training Dataset
Source: https://savrn.com/datasets/ai-cuda-engineer-archive
Markdown alternate of the page above; the site index is https://savrn.com/llms.txt

---

## Dataset Card

By Sakana AI, published under cc-by-4.0, revision 4edbe8d6d0b4.

### The AI CUDA Engineer Archive: Agentic CUDA Kernel Discovery, Optimization & Composition

We release [The AI CUDA Engineer archive](https://sakana.ai/ai-cuda-engineer/), a dataset consisting of approximately 30,000 CUDA kernels generated by [The AI CUDA Engineer](https://pub.sakana.ai/ai-cuda-engineer/paper). It is released under the CC-By-4.0 license and can be accessed via HuggingFace and interactively visualized [here](https://pub.sakana.ai/ai-cuda-engineer). The dataset is based on the Kernel tasks provided in [KernelBench](https://scalingintelligence.stanford.edu/KernelBenchLeaderboard/) and includes a torch reference implementation, torch, NCU and Clang-tidy profiling data, multiple kernels per task, error messages and speedup scores against torch native and compile runtimes.

We envision that this dataset can enable post-training of open-source models to perform better CUDA-enabling modules. This includes offline Reinforcement Learning, preference optimization, and standard supervised fine-tuning.

### Using the AI CUDA Engineer Archive

[Read the full dataset card (246 words)](https://savrn.com/datasets/ai-cuda-engineer-archive/card)

## Structure

### default 30,615 rows

| Split | Rows | Size |
| --- | --- | --- |
| level_1 | 12,157 | 161.0 MB |
| level_2 | 12,938 | 181.3 MB |
| level_3 | 5,520 | 69.2 MB |

Op_NamestringLevel_IDint64Task_IDint64Kernel_NamestringCUDA_Runtimefloat64PyTorch_Native_Runtimefloat64PyTorch_Compile_Runtimefloat64CUDA_Speedup_Nativefloat64CUDA_Speedup_Compilefloat64CUDA_CodestringPyTorch_Code_ModulestringPyTorch_Code_FunctionalstringCorrectboolMax_Difffloat64ErrorstringNCU_ProfilestringTorch_ProfilestringClang_Tidystring__index_level_0__int64

## Details

Repository

SakanaAI/AI-CUDA-Engineer-Archive

Publisher

Sakana AI

Task category

Not stated by the source

Tags

code

Size category

Not stated by the source

Languages

Not stated by the source

Revision

4edbe8d6d0b417e05aaf8ec7e23f78aecdc5516b

Last updated

2025-02-20

## Files

9 files, 67.7 MB in total.

Data3 files · 64.7 MB

Documentation1 file · 2.7 KB

Other3 files · 3.0 MB

Repository2 files · 8.6 KB

Every file

| File | Type | Size | SHA-256 |
| --- | --- | --- | --- |
| level_1.parquet | Data | 24.7 MB | a74e399809f8 |
| level_2.parquet | Data | 28.3 MB | 17be1e247272 |
| level_3.parquet | Data | 11.8 MB | 076f6abcceb2 |
| README.md | Documentation | 2.7 KB | — |
| archive_table.png | Other | 321.2 KB | cf1d8a166527 |
| db_scan_kernels.png | Other | 2.2 MB | 1c9f66d6d04c |
| results_table.png | Other | 388.4 KB | 13f04d830d46 |
| .DS_Store | Repository | 6.1 KB | — |
| .gitattributes | Repository | 2.5 KB | — |

## License and Download

License

cc-by-4.0

Access

No access gate

[Download from Sakana AI](https://huggingface.co/datasets/SakanaAI/AI-CUDA-Engineer-Archive)

Released by Sakana AI through its official repository on Hugging Face. [Read the license](https://creativecommons.org/licenses/by/4.0/).

## Sakana AI

[All models and datasets](https://savrn.com/model-publishers/sakanaai)

## Explore More

- [The dataset directory](https://savrn.com/datasets)
- [The model directory](https://savrn.com/models)
- [Research papers](https://savrn.com/papers)

## Source

- Repository metadata, read 2026-10-19.
- [Hugging Face record](https://huggingface.co/datasets/SakanaAI/AI-CUDA-Engineer-Archive)
- [How the hub is built](https://savrn.com/model-hub/methodology)