Reference library
SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot
Paper summary
SparseGPT prunes large pretrained Transformers in one pass using efficient reconstruction of remaining weights. Evaluations include unstructured and hardware-oriented sparsity patterns, with limited quality loss without retraining.
Relevance to the project
A cost-aware pruning reference for optional LLMs and a comparator for later connection-removal experiments. Sparse storage must be distinguished from measured execution speed.
Post-training pruning of fixed models is not local runtime graph adaptation.
Source record
- Reference ID
- sparsegpt
- Archived version
- 2301.00774v3
- Summary provenance
- reference-library/papers/sparsegpt/metadata.json · summary
- Review status
- initial orientation summary; not a critical full-paper review
- Reading scope
- Selected abstract/opening and available closing discussion/conclusion passages in the preserved text extract; broader project notes provide context. No reproduction of experiments, exhaustive methods review, or equation/figure audit.