Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning
read the original abstract
The popularity of LLaMA (Touvron et al., 2023a;b) and other recently emerged moderate-sized large language models (LLMs) highlights the potential of building smaller yet powerful LLMs. Regardless, the cost of training such models from scratch on trillions of tokens remains high. In this work, we study structured pruning as an effective means to develop smaller LLMs from pre-trained, larger models. Our approach employs two key techniques: (1) targeted structured pruning, which prunes a larger model to a specified target shape by removing layers, heads, and intermediate and hidden dimensions in an end-to-end manner, and (2) dynamic batch loading, which dynamically updates the composition of sampled data in each training batch based on varying losses across different domains. We demonstrate the efficacy of our approach by presenting the Sheared-LLaMA series, pruning the LLaMA2-7B model down to 1.3B and 2.7B parameters. Sheared-LLaMA models outperform state-of-the-art open-source models of equivalent sizes, such as Pythia, INCITE, OpenLLaMA and the concurrent TinyLlama models, on a wide range of downstream and instruction tuning evaluations, while requiring only 3% of compute compared to training such models from scratch. This work provides compelling evidence that leveraging existing LLMs with structured pruning is a far more cost-effective approach for building competitive small-scale LLMs
This paper has not been read by Pith yet.
Forward citations
Cited by 31 Pith papers
-
Distributionally Robust Multi-Task Reinforcement Learning via Adaptive Task Sampling
DRATS derives a minimax objective from a feasibility formulation of MTRL to adaptively sample tasks with the largest return gaps, leading to better worst-task performance on MetaWorld benchmarks.
-
Star Elastic: Many-in-One Reasoning LLMs with Efficient Budget Control
Star Elastic trains N nested submodels in a single post-training job on a parent reasoning LLM, supporting elastic budget control that matches or exceeds independent baselines while cutting training compute by up to 360x.
-
Fragile Knowledge, Robust Instruction-Following: The Width Pruning Dichotomy in Llama-3.2
Width pruning in Llama-3.2 models reduces parametric knowledge while enhancing instruction-following and preserving reasoning.
-
Data Mixing Agent: Learning to Re-weight Domains for Continual Pre-training
An RL agent learns domain re-weighting policies from evaluation feedback to improve balanced performance in continual pre-training of LLMs across source and target domains.
-
Federated Co-tuning Framework for Large and Small Language Models
FedCoLLM is a parameter-efficient federated co-tuning framework that improves client SLMs via server LLMs and enriches LLMs with client domain insights using adapters on NLP text generation tasks.
-
When AI Reviews Its Own Code: Recursive Self-Training Collapse in Code LLMs
Experiments across code LLMs show no-review collapses fastest, human-gated filters slow collapse, and AI self-gates lose effect over time, degenerating to ungated self-training under self-confirming acceptance as prov...
-
CascadeFormer: Depth-Tapered Transformers Motivated by Gradient Fan-in Asymmetry
CascadeFormer tapers Transformer width with depth based on gradient fan-in asymmetry to match uniform baselines in perplexity while cutting latency 8.6%.
-
DREAM-S: Speculative Decoding with Searchable Drafting and Target-Aware Refinement for Multimodal Generation
DREAM-S combines neural architecture search, target-aware supernet training, and attention-entropy-guided distillation to accelerate speculative decoding in VLMs, reporting up to 3.85x speedup over standard methods.
-
Pruning and Distilling Mixture-of-Experts into Dense Language Models
A systematic MoE-to-dense conversion via expert scoring, grouping, and distillation yields +6.3 pp average accuracy over dense-to-dense pruning at matched parameter count on tested models.
-
NITP: Next Implicit Token Prediction for LLM Pre-training
NITP augments standard next-token prediction with implicit semantic prediction in representation space using shallow-layer self-supervision, reporting consistent downstream gains on 0.5B-9B models including 5.7% on MM...
-
SAFE-SVD: Sensitivity-Aware Fidelity-Enforcing SVD for Physics Foundation Models
SAFE-SVD introduces a sensitivity-aware fidelity-enforcing SVD framework for compressing physics foundation models that maintains higher accuracy than standard methods at greater compression ratios.
-
Ghosted Layers: Unconstrained Activation Alignment for Recovering Layer-Pruned LLMs
Ghosted Layers recovers pruned LLM performance via an unconstrained closed-form linear operator for boundary activation alignment, outperforming prior constrained training-free baselines.
-
Compact SO(3) Equivariant Atomistic Foundation Models via Structural Pruning
Structural pruning of SO(3) equivariant atomistic models from large checkpoints yields 1.5-4x fewer parameters and 2.5-4x less pre-training compute than small models trained from scratch, while outperforming them on m...
-
SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training
Pruning pretrained MoE models outperforms training from scratch, different compression methods converge after continued pretraining, and combining KD with language modeling loss plus progressive schedules yields a com...
-
OmniZip: Audio-Guided Dynamic Token Compression for Fast Omnimodal Large Language Models
OmniZip introduces an audio-guided dynamic token compression framework that achieves 3.42X inference speedup and 1.4X memory reduction for omnimodal LLMs without any training.
-
VILA-U: a Unified Foundation Model Integrating Visual Understanding and Generation
VILA-U unifies visual understanding and generation inside one autoregressive next-token prediction model, removing separate diffusion components while claiming near state-of-the-art results.
-
MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies
MiniCPM 1.2B and 2.4B models reach parity with 7B-13B LLMs via model wind-tunnel scaling and a WSD scheduler that yields a higher optimal data-to-model ratio than Chinchilla scaling.
-
DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models
DoLa reduces hallucinations in LLMs by contrasting logits from later versus earlier layers during decoding, improving truthfulness on TruthfulQA by 12-17 absolute points without fine-tuning or retrieval.
-
A general tensor-structured compression scheme for efficient large language models
MixT compresses Transformer LLMs by substituting targeted linear projections with tensor-operator mixtures, preserving MMLU accuracy up to model-specific boundaries where parameter count drops 47.5% and inference memo...
-
NITP: Next Implicit Token Prediction for LLM Pre-training
NITP adds dense supervision from shallow model layers to predict implicit next-token semantics, yielding consistent downstream gains on 0.5B-9B models with ~2% extra training FLOPs.
-
Ghosted Layers: Unconstrained Activation Alignment for Recovering Layer-Pruned LLMs
Ghosted Layers recovers accuracy in layer-pruned LLMs via a closed-form unconstrained linear operator that aligns boundary activations using a small calibration set.
-
TAPIOCA: Why Task- Aware Pruning Improves OOD model Capability
Task-aware pruning improves OOD model performance by realigning distorted OOD layerwise norm and pairwise-distance profiles with the task-adapted geometry observed on ID inputs.
-
TAPIOCA: Why Task- Aware Pruning Improves OOD model Capability
Task-aware pruning improves OOD performance by removing layers that distort task-adapted representation profiles, realigning OOD inputs with the geometry observed on ID data.
-
SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training
Pruning pretrained MoE models outperforms training from scratch under fixed budget, different expert compression methods converge after continued training, and progressive pruning plus multi-token KD improves the fina...
-
Light-FMP: Lightweight Feature and Model Pruning for Enhanced Deep Recommender Systems
Light-FMP prunes features and model parameters in deep recommender systems by pretraining a hard-concrete masking layer on data subsets, then retraining the reduced model to improve both efficiency and accuracy over p...
-
Structural Pruning of Large Vision Language Models: A Comprehensive Study on Pruning Dynamics, Recovery, and Data Efficiency
Widthwise pruning of LVLM language backbones combined with supervised finetuning and hidden-state distillation recovers over 95% performance using just 5% of data across 3B-7B models.
-
On the Limits of Layer Pruning for Generative Reasoning in Large Language Models
Layer pruning preserves classification performance in LLMs but fundamentally limits recovery of generative reasoning capabilities even after extensive self-supervised finetuning.
-
TELL-TALE: Task Efficient LLMs with Task Aware Layer Elimination
TALE selectively prunes task-detrimental layers in LLMs at inference time to match or exceed baseline performance with lower computational cost across multiple models and tasks.
-
Theoretical Analysis of Sparse Optimization with Reparameterization, Weight Decay, and Adaptive Learning Rate
ReWA uses reparameterization with weight decay and adaptive rates to create a stable optimization landscape connected to ℓ_p regularization, yielding higher sparsity than ℓ1 on ResNets for CIFAR-10 and ImageNet while ...
-
A Survey on Efficient Inference for Large Language Models
The paper surveys techniques to speed up and reduce the resource needs of LLM inference, organized by data-level, model-level, and system-level changes, with comparative experiments on representative methods.
-
Small Language Models (SLMs) Can Still Pack a Punch: A survey (updated 2026)
A literature survey of Small Language Models (1-8B parameters) that can perform comparably or better than larger models, covering general-purpose and task-specific approaches plus creation techniques.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.