mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2026-08-26 15:16:23 +02:00
+2
-2
@@ -8419,8 +8419,8 @@ static void ggml_cuda_mul_mat_mat_batched_cublas(const ggml_tensor * src0, const
|
||||
cu_compute_type = CUBLAS_COMPUTE_32F;
|
||||
cu_data_type = CUDA_R_32F;
|
||||
|
||||
alpha = (const char *) &alpha_f32;
|
||||
beta = (const char *) &beta_f32;
|
||||
alpha = &alpha_f32;
|
||||
beta = &beta_f32;
|
||||
}
|
||||
|
||||
GGML_ASSERT(ne12 % ne02 == 0);
|
||||
|
||||
Reference in New Issue
Block a user