happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Update ggml/src/ggml-openvino/ggml-openvino-extra.cpp

This commit is contained in:
Mustafa Cavus 2026-01-13 12:27:33 -08:00
parent a92eceecd9
commit 599335c633
1 changed files with 9 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
ggml/src/ggml-openvino/ggml-openvino-extra.cpp
View File

@ -268,34 +268,33 @@ ggml_openvino_extracted_layout ggml_openvino_get_extracted_layout(const ggml_ten
}
// Normal extraction (no requant) - determine format based on tensor type
layout.is_u4 = false;
layout.weights_per_block = 32;
layout.is_symmetric = false;
switch (tensor->type) {
case GGML_TYPE_Q4_0:
layout.is_u4 = true;
layout.weights_per_block = 32;
layout.is_symmetric = true;
break;
case GGML_TYPE_Q4_1:
layout.is_u4 = true;
layout.weights_per_block = 32;
break;
case GGML_TYPE_Q4_K:
layout.is_u4 = true;
layout.weights_per_block = 32;
break;
case GGML_TYPE_Q8_0:
layout.is_u4 = false;
layout.weights_per_block = 32;
layout.is_symmetric = true;
break;
case GGML_TYPE_Q6_K:
layout.is_u4 = false;
layout.weights_per_block = 16;
layout.is_symmetric = true;
break;
case GGML_TYPE_Q5_K:
layout.is_u4 = false;
layout.weights_per_block = 32;
break;
default:
// Unsupported quantization type
return layout;