From e02d3d0c29cd68f4d3ab7404047af1c4bc240fe3 Mon Sep 17 00:00:00 2001 From: "Y. Velkov" Date: Thu, 16 May 2024 09:57:32 +0300 Subject: [PATCH] post review --- ggml-cuda.cu | 15 +++++---------- llama.cpp | 2 ++ 2 files changed, 7 insertions(+), 10 deletions(-) diff --git a/ggml-cuda.cu b/ggml-cuda.cu index 8e206f6ff..6e4e7ff7e 100644 --- a/ggml-cuda.cu +++ b/ggml-cuda.cu @@ -51,14 +51,9 @@ static_assert(sizeof(half) == sizeof(ggml_fp16_t), "wrong fp16 size"); static void ggml_cuda_default_log_callback(enum ggml_log_level level, const char * msg, void * user_data) { + GGML_UNUSED(level); GGML_UNUSED(user_data); - if (level == GGML_LOG_LEVEL_WARN) { - fprintf(stderr, "warning: %s", msg); - } else if (level == GGML_LOG_LEVEL_ERROR) { - fprintf(stderr, "error: %s", msg); - } else { - fprintf(stderr, "%s", msg); - } + fprintf(stderr, "%s", msg); } ggml_log_callback ggml_cuda_log_callback = ggml_cuda_default_log_callback; @@ -98,9 +93,9 @@ void ggml_cuda_error(const char * stmt, const char * func, const char * file, in int id = -1; // in case cudaGetDevice fails cudaGetDevice(&id); - GGML_CUDA_LOG_INFO("CUDA error: %s\n", msg); - GGML_CUDA_LOG_INFO(" current device: %d, in function %s at %s:%d\n", id, func, file, line); - GGML_CUDA_LOG_INFO(" %s\n", stmt); + GGML_CUDA_LOG_ERROR("CUDA error: %s\n", msg); + GGML_CUDA_LOG_ERROR(" current device: %d, in function %s at %s:%d\n", id, func, file, line); + GGML_CUDA_LOG_ERROR(" %s\n", stmt); // abort with GGML_ASSERT to get a stack trace GGML_ASSERT(!"CUDA error"); } diff --git a/llama.cpp b/llama.cpp index b25b5494f..250bdd845 100644 --- a/llama.cpp +++ b/llama.cpp @@ -1778,6 +1778,8 @@ struct llama_state { llama_state() { #ifdef GGML_USE_METAL ggml_backend_metal_log_set_callback(log_callback, log_callback_user_data); +#elif GGML_USE_CUDA + ggml_backend_cuda_log_set_callback(log_callback, log_callback_user_data); #endif }