-
Notifications
You must be signed in to change notification settings - Fork 135
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
* [Feature] enhance advanced ptq with multi-type input. Previous advanced ptq only support `torch.tensor`, but sometimes `dict` or `list` are alse needed. * [Fix] getitem should not be quantized twice. * [Feature] Add multi args cache. Note that the code find the placeholder rather than the input module now. So cache the output of placeholder but the input of module. * [Feature] support multiple inputs to a graph * [Fix] prune extra node in a block * [Fix] fix `keep_gpu` flag for non-tensor input * [Feature] assign node prefix in config to exclude some certain nodes. Sometimes, there is no need to quantize all nodes in the network. Ignore these nodes and keep them float. Co-authored-by: fanyunqian <fanyunqian@sensetime.com>
- Loading branch information
1 parent
05c915e
commit 72eebeb
Showing
4 changed files
with
318 additions
and
57 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
Oops, something went wrong.