Foundry 是 AI Skill Hub 本期精选AI工具之一。综合评分 7.5 分,整体质量较高。我们推荐使用将其纳入你的 AI 工具库,帮助提升工作效率。
高效AI工具:将CUDA图及其执行上下文保存到磁盘
Foundry 是一款基于 C++ 开发的开源工具,专注于 cuda-graph、gpu、inference 等核心功能。作为 GitHub 开源项目,它拥有活跃的社区支持和持续的版本迭代,代码完全透明可审计,支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流,都能提供稳定可靠的解决方案。
高效AI工具:将CUDA图及其执行上下文保存到磁盘
Foundry 是一款基于 C++ 开发的开源工具,专注于 cuda-graph、gpu、inference 等核心功能。作为 GitHub 开源项目,它拥有活跃的社区支持和持续的版本迭代,代码完全透明可审计,支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流,都能提供稳定可靠的解决方案。
# 克隆仓库 git clone https://github.com/foundry-org/foundry cd foundry # 查看安装说明 cat README.md # 按 README 完成环境依赖安装后即可使用
# 查看帮助 foundry --help # 基本运行 foundry [options] <input> # 详细使用说明请查阅文档 # https://github.com/foundry-org/foundry
# foundry 配置说明 # 查看配置选项 foundry --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量(覆盖配置文件) export FOUNDRY_CONFIG="/path/to/config.yml"
Instantaneous CUDA graph restoration via execution context materialization.
<p> <a href="https://www.python.org/"><img alt="Python" src="https://img.shields.io/badge/Python-3.12-blue"></a> <a href="https://arxiv.org/abs/2604.06664"><img alt="arXiv" src="https://img.shields.io/badge/arXiv-2604.06664-b31b1b?logo=arxiv&logoColor=white&labelColor=555555"></a> <a href="LICENSE"><img alt="License" src="https://img.shields.io/badge/License-Apache_2.0-green.svg"></a> </p> </div>
Foundry is a system that persists CUDA graph states through template-based context materialization. It materializes both the structure and execution context of captured CUDA graphs, making graph restoration kernel-agnostic and eliminating the need for hand-crafted patching rules. By intercepting CUDA driver calls, Foundry enforces a deterministic memory layout and automatically detects and serializes the binaries of kernels used in the CUDA graphs.
With Foundry, LLM serving engines can directly reload CUDA states from disk and skip the warmup process to start in a few seconds.
If you are using a conda environment, you can install the requirements with the following command:
conda install -c conda-forge boost-cpp boost
```bash pip install cmake # make sure cmake 4.0.0 +
pending = fdry.CUDAGraph.start_graph_builds( ["graph_0.json", "graph_1.json", ...], num_threads=24 )
conda deactivate conda activate xxx
conda install -c conda-forge libstdcxx-ng libgcc-ng
conda install -c conda-forge bear
bear -- python setup.py build_ext --inplace
Foundry ships engine integrations under foundry/python/foundry/integration/. Per-engine setup instructions live under recipe/.
| Engine | Integration code | Documentation | Setup Instructions |
|---|---|---|---|
| vLLM | [integration/vllm/](python/foundry/integration/vllm/) | [docs/vllm/overview.md](docs/vllm/overview.md) | [recipe/vllm/README.md](recipe/vllm/README.md) |
| SGLang | [integration/sglang/](python/foundry/integration/sglang/) | [docs/sglang/overview.md](docs/sglang/overview.md) | [recipe/sglang/README.md](recipe/sglang/README.md) |
| TensorRT-LLM | [integration/trtllm/](python/foundry/integration/trtllm/) | [docs/trtllm/overview.md](docs/trtllm/overview.md) | [recipe/trtllm/README.md](recipe/trtllm/README.md) |
fdry.load_cuda_modules_and_libraries('hook_archive')
高效的AI计算工具,具有较高的性能和可扩展性
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ Apache 2.0 — 宽松开源协议,可商用,需保留版权声明和 NOTICE 文件,含专利授权条款。
经综合评估,Foundry 在AI工具赛道中表现稳健,质量良好。如果你已有明确的使用需求,可以直接上手体验;如果还在评估阶段,建议对比同类工具后再做决策。
| 原始名称 | foundry |
| 原始描述 | 开源AI工具:Foundry materializes CUDA graphs along with its execution context to disk to sup。⭐19 · C++ |
| Topics | cuda-graphgpuinferencellmc++ |
| GitHub | https://github.com/foundry-org/foundry |
| License | Apache-2.0 |
| 语言 | C++ |
收录时间:2026-05-26 · 更新时间:2026-05-30 · License:Apache-2.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。