Fix q8 weights (use uint8 for q8; int8 produces poor results)

#18

by Xenova HF staff - opened 1 day ago

base: refs/heads/main

←

from: refs/pr/18

Discussion Files changed

-2

Upload fixed q8 ONNX models (reduce_range=True, per_channel=True)06633a3e

Xenova

Hugging Face TB Research org 1 day ago

•

edited 1 day ago

Slightly better, but not great. Will play around with other settings

Upload folder using huggingface_hub0919b6ca

Fix q8 weights (use uint8 for q8; int8 produces poor results)4f13109a

Xenova changed pull request title from Upload fixed q8 ONNX models (reduce_range=True, per_channel=True) to Fix q8 weights (use uint8 for q8; int8 produces poor results) 1 day ago

Xenova changed pull request status to merged 1 day ago

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment