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

Fix CudaMemcpy direction

Browse source
This commit is contained in:
Henrik Forstén 2023-12-22 14:44:04 +02:00
parent 48b7ff193e
commit a7a4205340
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml-cuda.cu
View file

@ -8843,7 +8843,7 @@ static void ggml_cuda_mul_mat_id(const ggml_tensor * src0, const ggml_tensor * s
const cudaMemcpyKind src1_kind = src1->backend == GGML_BACKEND_CPU ?
cudaMemcpyHostToDevice : cudaMemcpyDeviceToDevice;
const cudaMemcpyKind dst_kind = dst->backend == GGML_BACKEND_CPU ?
cudaMemcpyHostToDevice : cudaMemcpyDeviceToDevice;
cudaMemcpyDeviceToHost : cudaMemcpyDeviceToDevice;
for (int32_t row_id = 0; row_id < n_as; ++row_id) {
const struct ggml_tensor * src0_row = dst->src[row_id + 2];