From ab33f7a338593f6cf1ae98b10b6f8684f63bd72c Mon Sep 17 00:00:00 2001
From: slaren <slarengh@gmail.com>
Date: Sun, 19 May 2024 14:19:37 +0200
Subject: [PATCH] cuda : clear error after buffer allocation failure (#7376)

---
 ggml-cuda.cu | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/ggml-cuda.cu b/ggml-cuda.cu
index 754611bf3..b82167cbf 100644
--- a/ggml-cuda.cu
+++ b/ggml-cuda.cu
@@ -539,6 +539,8 @@ GGML_CALL static ggml_backend_buffer_t ggml_backend_cuda_buffer_type_alloc_buffe
     void * dev_ptr;
     cudaError_t err = cudaMalloc(&dev_ptr, size);
     if (err != cudaSuccess) {
+        // clear the error
+        cudaGetLastError();
         GGML_CUDA_LOG_ERROR("%s: allocating %.2f MiB on device %d: cudaMalloc failed: %s\n", __func__, size / 1024.0 / 1024.0, buft_ctx->device, cudaGetErrorString(err));
         return nullptr;
     }