fix(quantization): align eager-mode op call counts between capture and optimization passes - #78
Merged
pkmandke merged 5 commits intoAug 26, 2026
Conversation
Signed-off-by: Prathamesh Mandke <46148373+pkmandke@users.noreply.github.com>
Signed-off-by: Prathamesh Mandke <46148373+pkmandke@users.noreply.github.com>
pkmandke
marked this pull request as ready for review
August 26, 2026 20:28
pkmandke
changed the title
crowbat
approved these changes
Aug 26, 2026
Signed-off-by: Prathamesh Mandke <46148373+pkmandke@users.noreply.github.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
An op call whose arguments hold no quantizable tensor is not counted during capture (modes.py:798) but is counted during the optimization pass (modes.py:1123).
Both passes run inside
Quantizer.prepare(): the capture forward at handler.py:63, the optimization forward at quantizer.py:207. The counts then disagree andRuntimeErroris raised (registered_optimizers_tracker.py:214-221)."No quantizable tensor" here means no argument is a float tensor with more than one element, per
is_optimizable_tensor(thetorch.numel(tensor) <= 1clause). This is what Qwen2 hits in the toy config in coreai-models (please see this PR), it has nohead_dim, so RoPE'sfreqs * scalemultiplies a single-element tensor by a Python float.