zhangyiss/mlx
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx.git synced 2025-12-16 01:49:05 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

5bit quants (#2226)

Browse Source 
* 5bit quants

* 5bit quants
This commit is contained in:
Awni Hannun
2025-05-30 12:12:10 -07:00
committed by GitHub
parent f76ee1ffd2
commit 6ef2f67e7f
7 changed files with 248 additions and 72 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
mlx/fast.cpp
View File

@@ -839,14 +839,14 @@ affine_quantize(const array& w, int group_size, int bits, StreamOrDevice s_) {
if (group_size != 32 && group_size != 64 && group_size != 128) {
std::ostringstream msg;
msg << "[quantize] The requested group size " << group_size
<< " is not supported. The supported group sizes are 64 and 128.";
<< " is not supported. The supported group sizes are 32, 64, and 128.";
throw std::invalid_argument(msg.str());
}
if (bits != 2 && bits != 3 && bits != 4 && bits != 6 && bits != 8) {
if (bits < 2 || bits > 8 || bits == 7) {
std::ostringstream msg;
msg << "[quantize] The requested number of bits " << bits
<< " is not supported. The supported bits are 2, 3, 4, 6 and 8.";
<< " is not supported. The supported bits are 2, 3, 4, 5, 6 and 8.";
throw std::invalid_argument(msg.str());
}