-
Notifications
You must be signed in to change notification settings - Fork 152
[Wanda Refactor] TerminalModuleCompressor
contract
#1885
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM, I like the inclusion of the contract class. My one nit is I'd prefer all the "GPT" classes live in different files, one base, sparseGPT and Wanda.
Done! |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM pending comment and naming as discussed
torch.backends.cudnn.allow_tf32 = False | ||
|
||
|
||
class TerminalModuleCompressor(ABC): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
would be good to get rid of this since we just have 1 implementor class and don't expect any in the near future
86bbaed
to
c6f9ab0
Compare
TerminalModuleCompressor
contract
Closingg cause requested updates are now a part of #1887 |
* Add wanda base * Initial implementation * Update Wanda Base * Refactor to use WandaLayerCompressor Update WrappedGPT * Rename WrappedGPT to WandaGPT * [Wanda Refactor] Wanda/OBCQ Modifier Refactor (#1887) * Define GPT contract * rename tmp -> batch_size * Define LayerCompressor Contract * Rename gpt_helpers to gpts Fix some docstrings * add named argument to function call * Wanda/OBCQ refactor * propagate target-ids * Address review comments from * #1885 * #1886 * Fix typo * Update test * Fix regression
This PR removes some code duplicacy b/w SparseGPT and WandaGPT by defining a contract and Moving common functionality to a base class
Subsequent diffs will do similar refactor for LayerCompressor and the actual WandaModifier