From cbd26fdcc116dc692308f2d262083dfd1ddaa142 Mon Sep 17 00:00:00 2001 From: Andrei Betlen Date: Tue, 25 Apr 2023 19:03:41 -0400 Subject: [PATCH] Update llama.cpp --- llama_cpp/llama_cpp.py | 1 + vendor/llama.cpp | 2 +- 2 files changed, 2 insertions(+), 1 deletion(-) diff --git a/llama_cpp/llama_cpp.py b/llama_cpp/llama_cpp.py index 2b5af66..1097d74 100644 --- a/llama_cpp/llama_cpp.py +++ b/llama_cpp/llama_cpp.py @@ -119,6 +119,7 @@ LLAMA_FTYPE_MOSTLY_Q4_1_SOME_F16 = ctypes.c_int( ) # tok_embeddings.weight and output.weight are F16 LLAMA_FTYPE_MOSTLY_Q4_2 = ctypes.c_int(5) # except 1d tensors LLAMA_FTYPE_MOSTYL_Q4_3 = ctypes.c_int(6) # except 1d tensors +LLAMA_FTYPE_MOSTYL_Q8_0 = ctypes.c_int(7) # except 1d tensors # Functions diff --git a/vendor/llama.cpp b/vendor/llama.cpp index 54bb60e..4afcc37 160000 --- a/vendor/llama.cpp +++ b/vendor/llama.cpp @@ -1 +1 @@ -Subproject commit 54bb60e26858be251a0eb3cb70f80322aff804a0 +Subproject commit 4afcc378698e057fcde64e23eb664e5af8dd6956