Closed l-bat closed 1 week ago
Hybrid quantization does not improve PixArt's performance (CPU ticket: 141083); however, we have achieved memory reduction benefits. We are expecting a performance improvement after enabling per-token dynamic quantization (ref ticket: 143590).
Check out this pull request on
See visual diffs & provide feedback on Jupyter Notebooks.
Powered by ReviewNB
Hybrid quantization does not improve PixArt's performance (CPU ticket: 141083); however, we have achieved memory reduction benefits. We are expecting a performance improvement after enabling per-token dynamic quantization (ref ticket: 143590).