From 4d452dbc1068e88979197077903a105d6aa60b74 Mon Sep 17 00:00:00 2001
From: Andrew Godfrey <andrew@gfreys.com>
Date: Sat, 21 Oct 2023 22:40:40 -0700
Subject: [PATCH] Add fprintf in ggml_cuda_op_add

When I tried CUDA offloading during finetuning following the readme, I got an assert here.
This probably isn't an important case because inference later gives a warning saying you should use f16 or f32 instead when using lora
---
 ggml-cuda.cu | 1 +
 1 file changed, 1 insertion(+)

diff --git a/ggml-cuda.cu b/ggml-cuda.cu
index db053e3b8..63685100d 100644
--- a/ggml-cuda.cu
+++ b/ggml-cuda.cu
@@ -5910,6 +5910,7 @@ inline void ggml_cuda_op_add(
     } else if (src0->type == GGML_TYPE_F16 && dst->type == GGML_TYPE_F16) {
         add_f16_f32_f16_cuda((const half *) src0_dd, src1_dd, (half *) dst_dd, ggml_nelements(src0), main_stream);
     } else {
+        fprintf(stderr, "%d, %d\n", src0->type, dst->type);
         GGML_ASSERT(false);
     }