{"library":"triton-ascend","type":"library","category":null,"description":"Triton-Ascend is a compilation framework built for the Ascend platform, designed to enable Triton code to run efficiently on Ascend NPUs. It aims to reduce operator development difficulty by automatically handling memory allocation, data transfer, and computation based on hardware characteristics. The project is actively developed, with version 3.2.0 released, and plans to align with Triton 3.5.x in 2026.","language":"python","status":"active","version":"3.2.0","tags":["deep learning","compiler","hardware acceleration","NPU","Ascend","Triton","Huawei"],"install":[{"cmd":"pip install triton-ascend","imports":["import triton","import triton.language as tl"]}],"homepage":"https://gitcode.com/Ascend/triton-ascend/","github":null,"docs":null,"changelog":null,"pypi":"https://pypi.org/project/triton-ascend/","npm":null,"openapi_spec":null,"status_page":null,"smithery":null,"compatibility":{"summary":{"python_range":"3.10–3.9","success_rate":30,"avg_install_s":8.3,"avg_import_s":null,"wheel_type":"wheel"},"url":"https://checklist.day/v1/registry/triton-ascend/compatibility"},"provenance":{"verified_status":"install_fail","verified_at":"Sun Jun 28","last_verified":"Sun Jun 28","next_check":"Sun Jul 05","install_tag":null}}