jrok seungrokjung
seungrokjung
/ quant_cuda_kernel.cu
Created
May 5, 2023 12:22
GPTQ quantization kernel of dequantization + fp16 gemm operation compatible with Hipify. Original cuda code is from "https://github.com/oobabooga/GPTQ-for-LLaMa/blob/cuda/quant_cuda_kernel.cu"
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| #include <torch/all.h> | |
| #include <torch/python.h> | |
| #include <cuda.h> | |
| #include <cuda_runtime.h> | |
| #include <cuda_fp16.h> | |
| // atomicAdd for double-precision floating-point numbers on hardware with | |
| // compute capability < 6.0 from: | |
| // https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#atomic-functions | |
| #if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 600 |