vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

llama_model_loader: fail if backend cannot allocate buffer

Browse source
This commit is contained in:
Pierrick HYMBERT 2024-03-21 21:05:15 +01:00
parent 1c931f3d4f
commit d8b567d254
1 changed files with 6 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

6
llama.cpp
View file

@ -5155,6 +5155,8 @@ static bool llm_load_tensors(
ggml_backend_buffer_get_size(buf));
}
#endif
} else {
throw std::runtime_error("failed to allocate cpu buffer");
}
}
}
@ -5168,6 +5170,8 @@ static bool llm_load_tensors(
ggml_backend_buffer_t buf = ggml_backend_metal_buffer_from_ptr((char *) addr + first, last - first, max_size);
if (buf != nullptr) {
bufs.push_back(buf);
} else {
throw std::runtime_error("failed to allocate metal buffer");
}
}
}
@ -5182,6 +5186,8 @@ static bool llm_load_tensors(
mlock_buf->grow_to(ggml_backend_buffer_get_size(buf));
}
bufs.push_back(buf);
} else {
throw std::runtime_error("failed to allocate backend buffer");
}
}
if (bufs.empty()) {