vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix warnings

Browse source
This commit is contained in:
slaren 2023-11-01 19:43:27 +01:00
parent a9ab02eb6f
commit 1354122c21
1 changed files with 8 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

14
ggml-cuda.cu
View file

@ -6627,8 +6627,10 @@ inline void ggml_cuda_op_clamp(
GGML_ASSERT(src0->type == GGML_TYPE_F32); GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT( dst->type == GGML_TYPE_F32); GGML_ASSERT( dst->type == GGML_TYPE_F32);
const float min = ((float *) dst->op_params)[0]; float min;
const float max = ((float *) dst->op_params)[1]; float max;
memcpy(&min, dst->op_params, sizeof(float));
memcpy(&max, (float *) dst->op_params + 1, sizeof(float));
clamp_f32_cuda(src0_dd, dst_dd, min, max, ggml_nelements(src0), main_stream); clamp_f32_cuda(src0_dd, dst_dd, min, max, ggml_nelements(src0), main_stream);
CUDA_CHECK(cudaGetLastError()); CUDA_CHECK(cudaGetLastError());
@ -7153,7 +7155,7 @@ static void ggml_cuda_mul_mat_vec_nc(const ggml_tensor * src0, const ggml_tensor
} }
__global__ void k_compute_batched_ptrs( __global__ void k_compute_batched_ptrs(
half * src0_as_f16, half * src1_as_f16, half * dst_f16, const half * src0_as_f16, const half * src1_as_f16, half * dst_f16,
void ** ptrs, void ** ptrs,
int ne12, int ne13, int ne12, int ne13,
int ne23, int ne23,
@ -7298,9 +7300,9 @@ static void ggml_cuda_mul_mat_mat_batched_cublas(const ggml_tensor * src0, const
CUBLAS_CHECK( CUBLAS_CHECK(
cublasGemmBatchedEx(g_cublas_handles[id], CUBLAS_OP_T, CUBLAS_OP_N, cublasGemmBatchedEx(g_cublas_handles[id], CUBLAS_OP_T, CUBLAS_OP_N,
ne01, ne11, ne10, ne01, ne11, ne10,
&alpha_f16, (const void **) (ptrs_as + 0*ne23), CUDA_R_16F, nb01/sizeof(half), &alpha_f16, (const void * const *) (ptrs_as + 0*ne23), CUDA_R_16F, nb01/sizeof(half),
(const void **) (ptrs_as + 1*ne23), CUDA_R_16F, nb11/sizeof(float), (const void * const *) (ptrs_as + 1*ne23), CUDA_R_16F, nb11/sizeof(float),
&beta_f16, ( void **) (ptrs_as + 2*ne23), CUDA_R_16F, ne01, &beta_f16, ( void ** ) (ptrs_as + 2*ne23), CUDA_R_16F, ne01,
ne23, ne23,
CUBLAS_COMPUTE_16F, CUBLAS_COMPUTE_16F,
CUBLAS_GEMM_DEFAULT_TENSOR_OP)); CUBLAS_GEMM_DEFAULT_TENSOR_OP));