github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-09-09 20:38:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

add Falcon 7B model

Browse Source 
Tested with https://huggingface.co/TheBloke/falcon-7b-instruct-GGML/blob/main/falcon7b-instruct.ggmlv3.q4_0.bin
This commit is contained in:
Aaron Miller
2023-06-26 19:53:57 -07:00
committed by AT
parent b8464073b8
commit 198b5e4832
7 changed files with 1085 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
gpt4all-backend/llamamodel.cpp
View File

@@ -282,14 +282,15 @@ DLL_EXPORT bool magic_match(std::istream& f) {
if (!(version LLAMA_VERSIONS)) {
return false;
}
llama_file_hparams hparams;
f.read(reinterpret_cast<char*>(&hparams), sizeof(hparams));
if (!(hparams.n_vocab >= 32000 && hparams.n_vocab <= 32100)) {
return false; // not a llama.
}
#ifdef GGML_USE_METAL
// Check quant supported on metal
// skip fields
off_t offset = sizeof(uint32_t) * 6; // n_vocab, n_embd, n_mult, n_head, n_layer, n_rot
f.seekg(offset, std::ios_base::cur);
uint32_t ftype;
f.read(reinterpret_cast<char*>(&ftype), sizeof(ftype)); // ftype
switch((enum llama_ftype) ftype) {
switch(hparams.ftype) {
// 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: