github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-09-06 02:50:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix AVX support by removing direct linking to AVX2 libs (#1750)

Browse Source
This commit is contained in:
Jared Van Bortel
2023-12-13 12:11:09 -05:00
committed by GitHub
parent 0600f551b3
commit 3acbef14b7
6 changed files with 36 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
gpt4all-backend/llmodel.cpp
View File

@@ -82,7 +82,7 @@ const std::vector<LLModel::Implementation> &LLModel::Implementation::implementat
static auto* libs = new std::vector<Implementation>([] () {
std::vector<Implementation> fres;
std::string impl_name_re = "(bert|llama|gptj|llamamodel-mainline)";
std::string impl_name_re = "(bert|gptj|llamamodel-mainline)";
if (requires_avxonly()) {
impl_name_re += "-avxonly";
} else {
@@ -186,6 +186,27 @@ LLModel *LLModel::Implementation::construct(const std::string &modelPath, std::s
return fres;
}
LLModel *LLModel::Implementation::constructCpuLlama() {
const LLModel::Implementation *impl = nullptr;
for (const auto &i : implementationList()) {
if (i.m_buildVariant == "metal" || i.m_modelType != "LLaMA") continue;
impl = &i;
}
if (!impl) {
std::cerr << "LLModel ERROR: Could not find CPU LLaMA implementation\n";
return nullptr;
}
auto fres = impl->m_construct();
fres->m_implementation = impl;
return fres;
}
std::vector<LLModel::GPUDevice> LLModel::Implementation::availableGPUDevices() {
static LLModel *cpuLlama = LLModel::Implementation::constructCpuLlama(); // (memory leak)
if (cpuLlama) { return cpuLlama->availableGPUDevices(0); }
return {};
}
void LLModel::Implementation::setImplementationsSearchPath(const std::string& path) {
s_implementations_search_path = path;
}