-
Notifications
You must be signed in to change notification settings - Fork 49
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Support multiple weights attributes quantization in core (#964)
Adding support in multiple attributes quantization from the MCT core side. There are three main parts to this change: * Modifying the node candidate quantization configurations list to include per-attribute candidates: a new class WeightsAttrQuantizationConfig was added and a config of this type is created for each config that appears in the attribute's TPC configs mapping. * Enabling per-attribute operations. * Restricting several features for kernel-only: Mixed precision, GPTQ, QAT, and pruning currently continue to work only on the kernel attribute. Several features, like SNC, bias correction, etc. are also relevant only for nodes that have a quantized kernel. Currently, only kernel quantization is enabled by default.
- Loading branch information
Showing
81 changed files
with
1,505 additions
and
841 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.