Update llama_cpp.py

matthoffner · web-flow · commit f97ff3c5bbc2 · 2023-05-01T20:40:06.000-07:00
diff --git a/llama_cpp/llama_cpp.py b/llama_cpp/llama_cpp.py
@@ -136,9 +136,9 @@ class llama_context_params(Structure):
 )  # 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
-LLAMA_FTYPE_MOSTYL_Q5_0 = ctypes.c_int(8)  # except 1d tensors
-LLAMA_FTYPE_MOSTYL_Q5_1 = ctypes.c_int(9)  # except 1d tensors
+LLAMA_FTYPE_MOSTLY_Q8_0 = ctypes.c_int(7)  # except 1d tensors
+LLAMA_FTYPE_MOSTLY_Q5_0 = ctypes.c_int(8)  # except 1d tensors
+LLAMA_FTYPE_MOSTLY_Q5_1 = ctypes.c_int(9)  # except 1d tensors
 
 # Functions
 
