-
Notifications
You must be signed in to change notification settings - Fork 316
Open
Description
https://docs.pytorch.org/ao/main/quantization.html
Our Quantization Overview docs page is a bit outdated and too developer focused. We should structure it around:
- Two main quantization APIs,
quantize_
and PT2E ones - Different quantization features supported, example user code snippets, link to different doc pages
- Move all developer concepts (e.g. AQT, layout, quant primitives) to a separate page
- Delete static quant and serialization sections. These have their own tutorials
- Refresh dtypes supported (still mentions adding
torch.uint2
), add newer mx dtypes - Replace ascii art with real image
Metadata
Metadata
Assignees
Labels
No labels