update cluster experiments, dataloaders and pruning strategies for vision nets#128
Merged
houman1359 merged 35 commits intodevelopfrom Feb 18, 2026
Merged
update cluster experiments, dataloaders and pruning strategies for vision nets#128houman1359 merged 35 commits intodevelopfrom
houman1359 merged 35 commits intodevelopfrom
Conversation
…, configurable per-layer sparsity cap
The previous commit added supernode protection logic that uses PruningConfig but missed the import statement. This fixes it.
Supernode (activation outlier) protection is an LLM-specific phenomenon in FFN layers and does not apply to vision models like ResNet18. These configs were added by mistake when extending LLM experiments. Removed: - configs/paper/resnet18_cifar100_supernode_actSN_50_90.yaml - configs/paper/resnet18_cifar100_supernode_lpSN_50_90.yaml - drafts/LLM_prune/paper/scripts/aggregate_cifar100_supernode_def_sweep_50_90.py
Added two new methods to LLMAlignmentExperiment for paper experiments: 1. compute_mean_replacement_control(): Tests supernode importance by replacing LP/activation supernodes with mean values and measuring loss impact. Includes random replacement controls. 2. compute_lp_activation_analysis(): Computes LP vs activation correlation by percentile and supernode Jaccard overlap. This replaces manual computation in paper scripts. These methods allow paper scripts to use the alignment codebase instead of reimplementing SCAR metric computations directly.
- Added taylor_act_samples config option (base.py) - Added _compute_taylor_act_channel_scores() for Molchanov-style activation-based Taylor pruning (cluster_experiments.py) - Added taylor_act_* method name support (metric_based.py) This extends the weight-based Taylor baseline with an activation-based variant (|a * dL/da|) which is more aligned with the original Taylor channel pruning paper.
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.
Description
Type of Change
Testing
Related Issues
Closes #