[GPU] Recognize parameters as valid inputs for compressed weights #32276
+144
−52
Conversation
This file contains hidden or 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 change enables use of quantized LoRA weights, passed as parameters during execution, to be recognized by the transformaions that produce FullyConnectedCompressed nodes for QGEMM execution.
github-actions
bot
added
category: GPU
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Outdated suggestions cannot be applied.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.
Details:
Description of the issue:
At present, the FC_COMPRESSED_WEIGHT_PATTERN macro contains a pattern for dequantization of a constant integer weight. This pattern is used to recognize and fold cases where fused weight dequantization can be used, replacing them with FullyConnectedCompressed nodes. Due to expecting a constant weight input, this pattern fails to recognize quantized LoRA weights, which are provided as parameters:
With the changes in this patch, these weights can be recognized, and the transformations can proceed and produce nodes that would then leverage oneDNN fused QGEMM for execution:
Tickets: