Fix (optimum): Fix gptq and ONNX export #810
Merged
Conversation
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
Giuseppe5
force-pushed
the
optimum_improv
branch
from
4b6b2d6
to
02e701a
Compare
Giuseppe5
force-pushed
the
optimum_improv
branch
from
95f97b7
to
5b340cc
Compare
Giuseppe5
force-pushed
the
optimum_improv
branch
2 times, most recently
from
49bfd35
to
28ffb88
Compare
Giuseppe5
force-pushed
the
optimum_improv
branch
from
28ffb88
to
278d30d
Compare
Giuseppe5
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
The user can now specify the suffix of the layer names that can be run in parallel in GPTQ, to speed-up that computation.
This PR also fixed the export of ONNX files >2gb. The idea is that exporting the weight as integer causes the appearances of multiple parameters (the original FP weights, and the integer ones) in the model which are not handled correctly by pytorch during export.
This PR restores the possibility of exporting the FP weights + QDQ (instead of integer weight + DQ). In this way, having to handle only the original parameters of the model, everything seems to work fine.