AI/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2024-10-01 01:06:10 -04:00
Code Issues Packages Projects Releases Wiki Activity

llamamodel: metal supports all quantization types now

Browse Source
This commit is contained in:
Cebtenzzre 2023-09-25 14:49:00 -04:00 committed by Adam Treat
parent 507753a37c
commit 1d29e4696c
1 changed files with 1 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
gpt4all-backend/llamamodel.cpp
View File

@ -392,27 +392,7 @@ DLL_EXPORT bool magic_match(const char * fname) {
return false; return false;
bool isValid = gguf_get_version(ctx_gguf) <= 2; bool isValid = gguf_get_version(ctx_gguf) <= 2;
isValid = get_arch_name(ctx_gguf) != "llama" ? false : isValid; isValid = isValid && get_arch_name(ctx_gguf) == "llama";
#ifdef GGML_USE_METAL
const int n_tensors = gguf_get_n_tensors(ctx_gguf);
for (int i = 0; i < n_tensors; i++) {
const char * name = gguf_get_tensor_name(ctx_gguf, i);
struct ggml_tensor * meta = ggml_get_tensor(ctx_meta, name);
switch(meta->type) {
// currently supported on Metal https://github.com/ggerganov/llama.cpp/blob/ae9663f1887513e152839e91f61c513075a19422/ggml-metal.m#L51-L55
case LLAMA_FTYPE_MOSTLY_F16:
case LLAMA_FTYPE_MOSTLY_Q2_K:
case LLAMA_FTYPE_MOSTLY_Q4_0:
case LLAMA_FTYPE_MOSTLY_Q6_K:
case LLAMA_FTYPE_MOSTLY_Q4_K_S:
case LLAMA_FTYPE_MOSTLY_Q4_K_M:
break;
default: // unsupported quant-type for Metal
isValid = false;
}
}
#endif
gguf_free(ctx_gguf); gguf_free(ctx_gguf);
return isValid; return isValid;