fix uses of GGML_USE_CUBLAS in tests and examples (#879)

* fix uses of GGML_USE_CUBLAS in tests and examples

* fix ci/run.sh

ggml-ci

diff --git a/ci/run.sh b/ci/run.sh
index e61f639aac4f1ca1854f57df56e154524faaeba7..e5990747f0ab119b5395cb4fabb384f901266824 100644 (file)
--- a/ci/run.sh
+++ b/ci/run.sh
@@ -33,7 +33,7 @@ SRC=`pwd`
 CMAKE_EXTRA=""
 
 if [ ! -z ${GG_BUILD_CUDA} ]; then
-    CMAKE_EXTRA="${CMAKE_EXTRA} -DGGML_CUBLAS=ON"
+    CMAKE_EXTRA="${CMAKE_EXTRA} -DGGML_CUDA=ON"
 fi
 
 if [ ! -z ${GG_BUILD_METAL} ]; then

diff --git a/examples/gpt-2/CMakeLists.txt b/examples/gpt-2/CMakeLists.txt
index 12ba3ede369d56b11cdb0d4fd66809c7eeac75b9..686a9924a43b938e7a1102ceb023d669e6f8461b 100644 (file)
--- a/examples/gpt-2/CMakeLists.txt
+++ b/examples/gpt-2/CMakeLists.txt
@@ -30,15 +30,3 @@ target_link_libraries(${TEST_TARGET} PRIVATE ggml common common-ggml)
 set(TEST_TARGET gpt-2-batched)
 add_executable(${TEST_TARGET} main-batched.cpp)
 target_link_libraries(${TEST_TARGET} PRIVATE ggml common common-ggml)
-
-
-#
-# For GPU offloading
-
-if (GGML_CUBLAS)
-    add_compile_definitions(GGML_USE_CUBLAS)
-endif()
-
-if (GGML_METAL)
-    add_compile_definitions(GGML_USE_METAL)
-endif()

diff --git a/tests/test-arange.cpp b/tests/test-arange.cpp
index db63b6a8fb33bef51053518730e5de6eb63f8c97..39d7ad45fae50f2ab74e33624693c8266f8def9b 100644 (file)
--- a/tests/test-arange.cpp
+++ b/tests/test-arange.cpp
@@ -2,7 +2,7 @@
 #include "ggml-alloc.h"
 #include "ggml-backend.h"
 
-#ifdef GGML_USE_CUBLAS
+#ifdef GGML_USE_CUDA
 #include "ggml-cuda.h"
 #endif
 
@@ -22,7 +22,7 @@ int main(int /*argc*/, const char** /*argv*/) {
         ggml_backend_t backend = NULL;
         //ggml_backend_buffer_t buffer;
 
-        #ifdef GGML_USE_CUBLAS
+        #ifdef GGML_USE_CUDA
         if (use_gpu) {
             fprintf(stderr, "%s: using CUDA backend\n", __func__);
             backend = ggml_backend_cuda_init(0);

diff --git a/tests/test-conv1d.cpp b/tests/test-conv1d.cpp
index f2ea01b1ceb24ce6f90a570ee5cec869b690264b..53481470c509fb0ca3ddb15089ee773d95961648 100644 (file)
--- a/tests/test-conv1d.cpp
+++ b/tests/test-conv1d.cpp
@@ -2,9 +2,7 @@
 #include "ggml-alloc.h"
 #include "ggml-backend.h"
 
-// #define GGML_USE_CUBLAS
-
-#ifdef GGML_USE_CUBLAS
+#ifdef GGML_USE_CUDA
 #include "ggml-cuda.h"
 #endif
 
@@ -75,7 +73,7 @@ void load_model(test_model & model, bool use_gpu = false) {
     };
 
     // initialize the backend
-#ifdef GGML_USE_CUBLAS
+#ifdef GGML_USE_CUDA
     if (use_gpu) {
         fprintf(stderr, "%s: using CUDA backend\n", __func__);
         model.backend = ggml_backend_cuda_init(0);

diff --git a/tests/test-conv2d.cpp b/tests/test-conv2d.cpp
index 98005cd2d663ef4b46bfe51af53b406101452572..227f94c3fb7a5a7638975614f50c19062655a7ad 100644 (file)
--- a/tests/test-conv2d.cpp
+++ b/tests/test-conv2d.cpp
@@ -2,9 +2,7 @@
 #include "ggml-alloc.h"
 #include "ggml-backend.h"
 
-// #define GGML_USE_CUBLAS
-
-#ifdef GGML_USE_CUBLAS
+#ifdef GGML_USE_CUDA
 #include "ggml-cuda.h"
 #endif
 
@@ -75,7 +73,7 @@ void load_model(test_model & model, bool use_gpu = false) {
     };
 
     // initialize the backend
-#ifdef GGML_USE_CUBLAS
+#ifdef GGML_USE_CUDA
     if (use_gpu) {
         fprintf(stderr, "%s: using CUDA backend\n", __func__);
         model.backend = ggml_backend_cuda_init(0);

diff --git a/tests/test-mul-mat.cpp b/tests/test-mul-mat.cpp
index bf194ce1566a5be079a0e46c79da7f46f5f23396..cd218a2cb9a6409dd6b5975da5fa35da72ccccde 100644 (file)
--- a/tests/test-mul-mat.cpp
+++ b/tests/test-mul-mat.cpp
@@ -2,9 +2,7 @@
 #include "ggml-alloc.h"
 #include "ggml-backend.h"
 
-//#define GGML_USE_CUBLAS // uncomment this to use cuda backend, make sure build ggml lib with GGML_CUBLAS=ON
-
-#ifdef GGML_USE_CUBLAS
+#ifdef GGML_USE_CUDA
 #include "ggml-cuda.h"
 #endif
 
@@ -48,7 +46,7 @@ void load_model(test_model & model, float* a, float* b, int M, int N, int K, boo
     };
 
     // initialize the backend
-#ifdef GGML_USE_CUBLAS
+#ifdef GGML_USE_CUDA
     if (use_gpu) {
         fprintf(stderr, "%s: using CUDA backend\n", __func__);
         model.backend = ggml_backend_cuda_init(0);

diff --git a/tests/test-timestep_embedding.cpp b/tests/test-timestep_embedding.cpp
index 58bcd2f0d7a644ec6d9895fdcd02ea318befe140..9a87de31ef75369ae9f65ce37b47b862275779d3 100644 (file)
--- a/tests/test-timestep_embedding.cpp
+++ b/tests/test-timestep_embedding.cpp
@@ -2,7 +2,7 @@
 #include "ggml-alloc.h"
 #include "ggml-backend.h"
 
-#ifdef GGML_USE_CUBLAS
+#ifdef GGML_USE_CUDA
 #include "ggml-cuda.h"
 #endif
 
@@ -96,7 +96,7 @@ int main(int argc, const char** argv) {
         ggml_backend_t backend = NULL;
         ggml_backend_buffer_t params_buffer = NULL;
 
-        #ifdef GGML_USE_CUBLAS
+        #ifdef GGML_USE_CUDA
         if (use_gpu) {
             fprintf(stderr, "%s: using CUDA backend\n", __func__);
             backend = ggml_backend_cuda_init(0);