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

cuda : fix memset without set_device

Browse source
This commit is contained in:
slaren 2024-03-16 18:37:57 +01:00
parent 8e717e8cb8
commit 9cba8a183d
1 changed files with 1 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

1
ggml-cuda.cu
View file

@ -10618,6 +10618,7 @@ GGML_CALL static void ggml_backend_cuda_buffer_init_tensor(ggml_backend_buffer_t
size_t padded_size = ggml_backend_buft_get_alloc_size(buffer->buft, tensor); size_t padded_size = ggml_backend_buft_get_alloc_size(buffer->buft, tensor);
if (padded_size > original_size && tensor->view_src == nullptr) { if (padded_size > original_size && tensor->view_src == nullptr) {
ggml_cuda_set_device(ctx->device);
CUDA_CHECK(cudaMemset((char *)tensor->data + original_size, 0, padded_size - original_size)); CUDA_CHECK(cudaMemset((char *)tensor->data + original_size, 0, padded_size - original_size));
} }
} }