CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-28 08:31:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add a warning for special devices (#15563)

Browse Source 
* Add warning

* Print the devices names

* Add newlines

* Apply suggestions from code review

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>

* Fix vector names

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
This commit is contained in:
Yoshi_likes_e4
2025-08-26 13:15:33 +07:00
committed by GitHub
parent 34bdbbd7c2
commit 4c37636b3e
1 changed files with 21 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
ggml/src/ggml-cuda/ggml-cuda.cu
View File

@@ -204,6 +204,8 @@ static ggml_cuda_device_info ggml_cuda_init() {
GGML_LOG_INFO("%s: GGML_CUDA_FORCE_CUBLAS: no\n", __func__);
#endif // GGML_CUDA_FORCE_CUBLAS
GGML_LOG_INFO("%s: found %d " GGML_CUDA_NAME " devices:\n", __func__, info.device_count);
std::vector<std::pair<int, std::string>> turing_devices_without_mma;
for (int id = 0; id < info.device_count; ++id) {
int device_vmm = 0;
@@ -261,9 +263,27 @@ static ggml_cuda_device_info ggml_cuda_init() {
info.devices[id].cc = 100*prop.major + 10*prop.minor;
GGML_LOG_INFO(" Device %d: %s, compute capability %d.%d, VMM: %s\n",
id, prop.name, prop.major, prop.minor, device_vmm ? "yes" : "no");
std::string device_name(prop.name);
if (device_name == "NVIDIA GeForce MX450") {
turing_devices_without_mma.push_back({ id, device_name });
} else if (device_name == "NVIDIA GeForce MX550") {
turing_devices_without_mma.push_back({ id, device_name });
} else if (device_name.substr(0, 21) == "NVIDIA GeForce GTX 16") {
turing_devices_without_mma.push_back({ id, device_name });
}
#endif // defined(GGML_USE_HIP)
}
if (ggml_cuda_highest_compiled_arch(GGML_CUDA_CC_TURING) >= GGML_CUDA_CC_TURING && !turing_devices_without_mma.empty()) {
GGML_LOG_INFO("The following devices will have suboptimal performance due to a lack of tensor cores:\n");
for (size_t device_pos = 0; device_pos < turing_devices_without_mma.size(); device_pos++) {
GGML_LOG_INFO(
" Device %d: %s\n", turing_devices_without_mma[device_pos].first, turing_devices_without_mma[device_pos].second.c_str());
}
GGML_LOG_INFO(
"Consider compiling with CMAKE_CUDA_ARCHITECTURES=61-virtual;80-virtual and DGGML_CUDA_FORCE_MMQ to force the use of the Pascal code for Turing.\n");
}
for (int id = 0; id < info.device_count; ++id) {
info.default_tensor_split[id] /= total_vram;
}