github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-09-06 11:00:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama on Metal (#885)

Browse Source 
Support latest llama with Metal

---------

Co-authored-by: Adam Treat <adam@nomic.ai>
Co-authored-by: niansa/tuxifan <tuxifan@posteo.de>
This commit is contained in:
Aaron Miller
2023-06-09 11:58:12 -07:00
committed by GitHub
parent 14e9ccbc6a
commit c55f81b860
7 changed files with 138 additions and 64 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
gpt4all-backend/llmodel.cpp
View File

@@ -121,20 +121,30 @@ LLModel *LLModel::construct(const std::string &modelPath, std::string buildVaria
if (!has_at_least_minimal_hardware())
return nullptr;
//TODO: Auto-detect CUDA/OpenCL
if (buildVariant == "auto") {
if (requires_avxonly()) {
buildVariant = "avxonly";
} else {
buildVariant = "default";
}
}
// Read magic
std::ifstream f(modelPath, std::ios::binary);
if (!f) return nullptr;
// Get correct implementation
auto impl = implementation(f, buildVariant);
if (!impl) return nullptr;
const LLModel::Implementation* impl = nullptr;
#if defined(__APPLE__) && defined(__arm64__) // FIXME: See if metal works for intel macs
if (buildVariant == "auto") {
impl = implementation(f, "metal");
}
#endif
if (!impl) {
//TODO: Auto-detect CUDA/OpenCL
if (buildVariant == "auto") {
if (requires_avxonly()) {
buildVariant = "avxonly";
} else {
buildVariant = "default";
}
}
impl = implementation(f, buildVariant);
if (!impl) return nullptr;
}
f.close();
// Construct and return llmodel implementation
return impl->construct();