SEAM
Automated CUDA project migration to Chinese domestic GPUs
All Skills in This Repository (4)
Pure Emerald Level Indicators
Fudan-SMI-labtorch-npu-venv-setup-cpu-base
Pin CPU-base PyTorch and install matching torch-npu and torchvision.
Fudan-SMI-labtorch-npu-pyyaml-preinstall
Preinstall PyYAML before torch-npu to prevent transitive dependency failures.
Fudan-SMI-labapex-cuda-extension-removal
Replace NVIDIA Apex imports and APIs with NPU-native PyTorch equivalents.
Fudan-SMI-labe2e-v2-0dc9b40e0da8-promoted
Create virtual environments inheriting system torch-npu stacks on Ascend hosts.
Frequently Asked Questions
FAQPage SchemaHow to install SEAM?▼
Run `npx skills add Fudan-SMI-lab/SEAM --all -g -y` in your terminal to install all skills in this suite globally.
How to migrate CUDA code to Huawei Ascend NPU?▼
SEAM automatically converts your CUDA project for Ascend NPU, including dependency setup, code rewriting, and custom operator porting, then verifies the result with end-to-end tests.
Which Chinese GPUs does SEAM support?▼
SEAM supports Huawei Ascend NPU, MetaX, and Alibaba T-Head PPU across PyTorch, vLLM, and SGLang frameworks.
How does SEAM fix torch-npu dependency errors?▼
It includes proven skills that resolve common issues like CPU-base torch version conflicts, missing PyYAML, and NVIDIA Apex removal with NPU-native replacements.
Does SEAM verify migration results?▼
Yes. It runs end-to-end tests, generates migration reports with accuracy and performance evidence, and uses strict validation gates to prevent false success claims.
Related Repositories in Software Engineering
View All in Software Engineering→
openclawopenclaw
Run a personal AI assistant across your devices and chat apps
obrasuperpowers
Gives coding agents a disciplined workflow from idea to merged code
facebookreact
AI agent skills for building, testing, and porting React core