diff options
| author | Kawrakow <48489457+ikawrakow@users.noreply.github.com> | 2024-07-27 07:55:01 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-07-27 07:55:01 +0200 |
| commit | 154e0d75fccf1784fe9ff6fd76a630b66563da3d (patch) | |
| tree | 81ce6dbb5b1900c1aa78a879f0593c694cab9d27 /gguf-py/gguf/quants.py | |
| parent | 0684c3e9c70d49323b4fc517128cbe222cab7f96 (diff) | |
Merge mainline llama.cpp (#3)
* Merging mainline - WIP
* Merging mainline - WIP
AVX2 and CUDA appear to work.
CUDA performance seems slightly (~1-2%) lower as it is so often
the case with llama.cpp/ggml after some "improvements" have been made.
* Merging mainline - fix Metal
* Remove check
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'gguf-py/gguf/quants.py')
| -rw-r--r-- | gguf-py/gguf/quants.py | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/gguf-py/gguf/quants.py b/gguf-py/gguf/quants.py index b22eec16..16e0a9aa 100644 --- a/gguf-py/gguf/quants.py +++ b/gguf-py/gguf/quants.py @@ -43,7 +43,7 @@ def __apply_over_grouped_rows(func: Callable[[np.ndarray], np.ndarray], arr: np. osize *= dim out = np.empty(shape=osize, dtype=otype) # compute over groups of 16 rows (arbitrary, but seems good for performance) - n_groups = rows.shape[0] // 16 + n_groups = (rows.shape[0] // 16) or 1 np.concatenate([func(group).ravel() for group in np.array_split(rows, n_groups)], axis=0, out=out) return out.reshape(oshape) |