update cluster experiments, dataloaders and pruning strategies for vision nets#128

Merged

houman1359 merged 35 commits into

Feb 18, 2026

Commits on Jan 18, 2026

Update Wanda and SparseGPT to use unstructured pruning
audrey-tan
committed

Commits on Jan 20, 2026

Commits on Jan 21, 2026

Commits on Jan 22, 2026

Commits on Jan 23, 2026

Paper reproducibility: deterministic calibration subset, run metadata, configurable per-layer sparsity cap
houman1359
committed
Slurm: add generic vision unified runner (single seed)
houman1359
committed
implement and add new metrics and analysis for vision and llm works
houman1359
committed

Commits on Jan 26, 2026

update cluster pruning
houman1359
committed

Commits on Jan 27, 2026

Commits on Jan 28, 2026

Commits on Feb 9, 2026

Commits on Feb 11, 2026

add fine-tuning cluster
houman1359
committed

Commits on Feb 13, 2026

Make MI CAP variants consistent across all scheduling paths
houman1359
committed

Commits on Feb 15, 2026

Enforce exact CAP prune budgets and log achieved sparsity in pruning outputs
houman1359
committed

Commits on Feb 16, 2026

Enforce exact global channel sparsity and fix pruning sparsity logging
houman1359
committed

Commits on Feb 17, 2026

Add cluster-fix CAP variants and expand pruning pipeline tests
houman1359
committed

Commits on Feb 18, 2026