tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-18 14:18:50 -04:00
Code Issues Packages Projects Releases Wiki Activity

ggml-quants : use ceiling division when quantizing q1_3

Browse Source
This commit is contained in:
Francis Couture-Harpin
2024-06-26 15:31:48 -04:00
parent 9465ec6e12
commit 89dc3b254c
4 changed files with 12 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
gguf-py/gguf/quants.py
View File

@@ -149,8 +149,7 @@ def __quantize_q1_3_rows(n: np.ndarray) -> np.ndarray:
q4 = np.sum(q4 * pow3.reshape((1, 4)), axis=1, keepdims=True)
q48 = q48 + (q12 * 81)
q = np.concatenate([q48, q4], axis=1)
q = ((q.astype(np.uint16) * 256) // 243).astype(np.uint8)
q = np.where(q != 0, q + 1, 0)
q = (((q.astype(np.uint16) * 256) + (243 - 1)) // 243).astype(np.uint8)
return q.reshape(__quantize_q1_3_shape_change(shape))