huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-18 11:37:11 +00:00
Code Issues Packages Projects Releases Wiki Activity

Cuda support for the mnist training. (#277)

Browse Source 
* Cuda support for the mnist training.

* min/max fix + testing.

* Add the argmin/argmax tests.

* More cuda support for argmin/argmax.

* Cuda kernels for argmin and argmax.
This commit is contained in:
Laurent Mazare
2023-07-29 19:48:04 +01:00
committed by GitHub
parent 16c33383eb
commit c950a5c6b1
6 changed files with 453 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
candle-kernels/src/cuda_utils.cuh
View File

@ -144,7 +144,8 @@ __device__ __forceinline__ double copysigng(double a, double b) { return copysig
__device__ __forceinline__ uint32_t ming(uint32_t a, uint32_t b) { return min(a, b); }
__device__ __forceinline__ uint32_t maxg(uint32_t a, uint32_t b) { return max(a, b); }
__device__ __forceinline__ uint8_t ming(uint8_t a, uint8_t b) { return min(a, b); }
__device__ __forceinline__ uint8_t maxg(uint8_t a, uint8_t b) { return max(a, b); }
#if __CUDA_ARCH__ >= 530
__device__ __forceinline__ __half powg(__half a, __half b) { return __float2half(powf(__half2float(a), __half2float(b))); }
__device__ __forceinline__ bool isnang(__half a) { return __hisnan(a); }