From c6874f9759fe4c4839254a20e46d4f2ccac10720 Mon Sep 17 00:00:00 2001 From: Haohui Mai Date: Sat, 18 Nov 2023 20:51:09 -0800 Subject: [PATCH 1/3] Fix incorrect format strings and uninitialized variables. --- examples/server/server.cpp | 2 ++ ggml-cuda.cu | 2 +- ggml.c | 4 +++- 3 files changed, 6 insertions(+), 2 deletions(-) diff --git a/examples/server/server.cpp b/examples/server/server.cpp index bb87b532b2c18..01ed53e38d589 100644 --- a/examples/server/server.cpp +++ b/examples/server/server.cpp @@ -1095,6 +1095,7 @@ struct llama_server_context std::lock_guard lock(mutex_results); task_result res; res.id = id; + res.stop = false; res.error = true; res.result_json = { { "content", error } }; queue_results.push_back(res); @@ -1255,6 +1256,7 @@ struct llama_server_context std::lock_guard lock(mutex_tasks); task_server task; task.id = id_gen++; + task.target_id = 0; task.data = data; task.infill_mode = infill; task.embedding_mode = embedding; diff --git a/ggml-cuda.cu b/ggml-cuda.cu index 50e03de500747..8bd36a7de11c9 100644 --- a/ggml-cuda.cu +++ b/ggml-cuda.cu @@ -8057,7 +8057,7 @@ bool ggml_cuda_compute_forward(struct ggml_compute_params * params, struct ggml_ if (tensor->op == GGML_OP_MUL_MAT) { if (tensor->src[0]->ne[3] != tensor->src[1]->ne[3]) { #ifndef NDEBUG - fprintf(stderr, "%s: cannot compute %s: src0->ne[3] = %d, src1->ne[3] = %d - fallback to CPU\n", __func__, tensor->name, tensor->src[0]->ne[3], tensor->src[1]->ne[3]); + fprintf(stderr, "%s: cannot compute %s: src0->ne[3] = %ld, src1->ne[3] = %ld - fallback to CPU\n", __func__, tensor->name, tensor->src[0]->ne[3], tensor->src[1]->ne[3]); #endif return false; } diff --git a/ggml.c b/ggml.c index f92292b39c635..8b843b3413778 100644 --- a/ggml.c +++ b/ggml.c @@ -1586,7 +1586,7 @@ inline static void ggml_vec_argmax_f32(const int n, int * s, const float * x) { // data types // -static const char * GGML_OP_NAME[GGML_OP_COUNT] = { +static const char * GGML_OP_NAME[GGML_OP_COUNT + 2] = { "NONE", "DUP", @@ -1664,6 +1664,8 @@ static const char * GGML_OP_NAME[GGML_OP_COUNT] = { "CROSS_ENTROPY_LOSS", "CROSS_ENTROPY_LOSS_BACK", + "", + "", }; static_assert(GGML_OP_COUNT == 68, "GGML_OP_COUNT != 68"); From 566785f560cd82a52f3e70ebc6108d94fb73a217 Mon Sep 17 00:00:00 2001 From: Haohui Mai Date: Mon, 20 Nov 2023 22:04:21 -0800 Subject: [PATCH 2/3] Address comments --- ggml-cuda.cu | 2 +- ggml.c | 4 +--- 2 files changed, 2 insertions(+), 4 deletions(-) diff --git a/ggml-cuda.cu b/ggml-cuda.cu index 8bd36a7de11c9..c065d8ec63115 100644 --- a/ggml-cuda.cu +++ b/ggml-cuda.cu @@ -8057,7 +8057,7 @@ bool ggml_cuda_compute_forward(struct ggml_compute_params * params, struct ggml_ if (tensor->op == GGML_OP_MUL_MAT) { if (tensor->src[0]->ne[3] != tensor->src[1]->ne[3]) { #ifndef NDEBUG - fprintf(stderr, "%s: cannot compute %s: src0->ne[3] = %ld, src1->ne[3] = %ld - fallback to CPU\n", __func__, tensor->name, tensor->src[0]->ne[3], tensor->src[1]->ne[3]); + fprintf(stderr, "%s: cannot compute %s: src0->ne[3] = " PRId64 ", src1->ne[3] = " PRId64 " - fallback to CPU\n", __func__, tensor->name, tensor->src[0]->ne[3], tensor->src[1]->ne[3]); #endif return false; } diff --git a/ggml.c b/ggml.c index 8b843b3413778..f92292b39c635 100644 --- a/ggml.c +++ b/ggml.c @@ -1586,7 +1586,7 @@ inline static void ggml_vec_argmax_f32(const int n, int * s, const float * x) { // data types // -static const char * GGML_OP_NAME[GGML_OP_COUNT + 2] = { +static const char * GGML_OP_NAME[GGML_OP_COUNT] = { "NONE", "DUP", @@ -1664,8 +1664,6 @@ static const char * GGML_OP_NAME[GGML_OP_COUNT + 2] = { "CROSS_ENTROPY_LOSS", "CROSS_ENTROPY_LOSS_BACK", - "", - "", }; static_assert(GGML_OP_COUNT == 68, "GGML_OP_COUNT != 68"); From 9216e7bebaf329c4a87f185158bed6c209349a14 Mon Sep 17 00:00:00 2001 From: Haohui Mai Date: Wed, 22 Nov 2023 08:16:34 -0800 Subject: [PATCH 3/3] Add the missing include statement --- ggml-cuda.cu | 1 + 1 file changed, 1 insertion(+) diff --git a/ggml-cuda.cu b/ggml-cuda.cu index c065d8ec63115..f0db7ae357a2f 100644 --- a/ggml-cuda.cu +++ b/ggml-cuda.cu @@ -1,4 +1,5 @@ #include +#include #include #include #include