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

ggml : sync ggml (add GPT-NeoX RoPE implementation)

Browse source
This commit is contained in:
Georgi Gerganov 2023-04-20 23:32:59 +03:00
parent 9ff334f3c9
commit 12b5900dbc
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
3 changed files with 49 additions and 17 deletions
Download patch file Download diff file Expand all files Collapse all files

5
llama.cpp
View file

@ -1618,6 +1618,11 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
// quantize only 2D tensors
quantize &= (tensor.ne.size() == 2);
// GG: uncomment this to keep the output layer in FP16
//if (tensor.name.rfind("output")) {
// quantize = false;
//}
enum ggml_type new_type;
void * new_data;
size_t new_size;