Fix (GPxQ): unwrap QuantTensor when dealing with QuantLinear #915
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When running GPxQ and we use quantized activations, once a
QuantLinear
layer is processed,update_batch
tries to unsqueeze (GPFQ) or transpose (GPTQ) thatQuantTensor
, leading to an error.This PR solves that issue by unwrapping that tensor. It also adds a quant linear model in the fixtures for reproducing the issue.
Some of the tests still fail as it does not raise a
ValueError
because it actually has aquant_input
(here). @i-colbert maybe we need to change the conditions for an expected fail?