how to generate a kernel with cuda c code implementation with triton

[TigerYang414]

need to run with c++ code with online infer

[ptillet]

Unfortunately this is not possible, but you can get the PTX -- though this would require more glue code to call it for inference

[yiakwy-xpu-ml-framework-team]

Yes you can. Triton will generate cubin, then you can use CUmodule to load them in c file then executed in cuLaunchKernel. That means now you have the c source for the kernel.

@ptillet @TigerYang414