At Facebook’s @Scale engineering conference, the company announced an artificial intelligence tool called “SapFix.” The tool is a boon for Facebook engineers since it debugs code automatically.
The hybrid AI tool automatically generates fixes for specific bugs in the Facebook code, and then proposes it to the engineers for approval and deployment to production.
According to the company, SapFix is already in use for sending robust, stable code updates to millions of Android devices using Facebook.
SapFix run with or without Sapienz, which is Facebook’s intelligent automated software testing tool to do the heavy lifting. Sapienz pinpoints a specific portion of code that was involved in the code crash, which is then passed on to SapFix. The AI tool comes up with a few tactics to generate a patch fix.
“To our knowledge, this marks the first time that a machine-generated fix — with automated end-to-end testing and repair — has been deployed into a codebase of Facebook’s scale,” the blog post mentions.
Facebook is intending to make SapFix open-source for the with the engineering community in the future.
Facebook adds hardware partner support for Glow compiler
Apparently, SapFix is not the only advancement Facebook announced in its machine learning journey.
Facebook is officially adding partner support for its Glow Compiler used for hardware acceleration in machine learning. Cadence, Esperanto, Intel, Marvell, and Qualcomm Technologies Inc, a subsidiary of Qualcomm Incorporated, all are now part of Glow project which drastically improves silicon products.
Glow is a compiler software which helps in designing AI and ML-ready silicon products much faster than the previous methodologies.
While tech giants like Apple and Google are building their own silicon chips, it seems like Facebook is going to outsource high performing chips while using their own ML code.