Awesome LLM Compression

Awesome LLM compression research papers and tools to accelerate LLM training and inference.

Contents

• 📑 Papers
  • Survey
  • Quantization
  • Pruning and Sparsity
  • Distillation
  • Efficient Prompting
  • Other
• 🔧 Tools
• 🙌 Contributing
• 🌟 Star History

Papers

Survey

• A Survey on Model Compression for Large Language Models
  Arxiv 2023 [Paper]

• The Efficiency Spectrum of Large Language Models: An Algorithmic Survey
  Arxiv 2023 [Paper]

• Efficient Large Language Models: A Survey
  Arxiv 2023 [Paper] [GitHub Page]

• Towards Efficient Generative Large Language Model Serving: A Survey from Algorithms to Systems
  Arxiv 2023 [Paper]

• Understanding LLMs: A Comprehensive Overview from Training to Inference
  Arxiv 2024 [Paper]

• A Survey of Resource-efficient LLM and Multimodal Foundation Models
  Arxiv 2024 [Paper]

• A Survey on Hardware Accelerators for Large Language Models
  Arxiv 2024 [Paper]

• A Comprehensive Survey of Compression Algorithms for Language Models
  Arxiv 2024 [Paper]

• Model Compression and Efficient Inference for Large Language Models: A Survey
  Arxiv 2024 [Paper]

• A Survey on Knowledge Distillation of Large Language Models
  Arxiv 2024 [Paper] [GitHub Page]

• Efficient Prompting Methods for Large Language Models: A Survey
  Arxiv 2024 [Paper]

Quantization

• ZeroQuant: Efficient and Affordable Post-Training Quantization for Large-Scale Transformers
  NeurIPS 2022 [Paper] [Code (DeepSpeed)]

• LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale
  NeurIPS 2022 [Paper] [Code]

• Outlier Suppression: Pushing the Limit of Low-bit Transformer Language Models
  NeurIPS 2022 [Paper] [Code]

• LUT-GEMM: Quantized Matrix Multiplication based on LUTs for Efficient Inference in Large-Scale Generative Language Models
  Arxiv 2022 [Paper]

• SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models
  ICML 2023 [Paper] [Code]

• FlexRound: Learnable Rounding based on Element-wise Division for Post-Training Quantization
  ICML 2023 [Paper] [Code (DeepSpeed)]

• Understanding INT4 Quantization for Transformer Models: Latency Speedup, Composability, and Failure Cases
  ICML 2023 [Paper] [Code]

• The case for 4-bit precision: k-bit Inference Scaling Laws
  ICML 2023 [Paper]

• GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers
  ICLR 2023 [Paper] [Code]

• PreQuant: A Task-agnostic Quantization Approach for Pre-trained Language Models
  ACL 2023 [Paper]

• Boost Transformer-based Language Models with GPU-Friendly Sparsity and Quantization
  ACL 2023 [Paper]

• QLoRA: Efficient Finetuning of Quantized LLMs
  NeurIPS 2023 [Paper] [Code]

• The Quantization Model of Neural Scaling
  NeurIPS 2023 [Paper]

• Quantized Distributed Training of Large Models with Convergence Guarantees
  Arxiv 2023 [Paper]

• RPTQ: Reorder-based Post-training Quantization for Large Language Models
  Arxiv 2023 [Paper] [Code]

• ZeroQuant-V2: Exploring Post-training Quantization in LLMs from Comprehensive Study to Low Rank Compensation
  Arxiv 2023 [Paper] [Code]

• Integer or Floating Point? New Outlooks for Low-Bit Quantization on Large Language Models
  Arxiv 2023 [Paper]

• Memory-Efficient Fine-Tuning of Compressed Large Language Models via sub-4-bit Integer Quantization
  NeurIPS 2023 [Paper]

• Compress, Then Prompt: Improving Accuracy-Efficiency Trade-off of LLM Inference with Transferable Prompt
  Arxiv 2023 [Paper]

• AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
  Arxiv 2023 [Paper] [Code]

• LLM-QAT: Data-Free Quantization Aware Training for Large Language Models
  Arxiv 2023 [Paper] [Code]

• SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression
  Arxiv 2023 [Paper] [Code]

• OWQ: Lessons learned from activation outliers for weight quantization in large language models
  Arxiv 2023 [Paper]

• SqueezeLLM: Dense-and-Sparse Quantization
  Arxiv 2023 [Paper] [Code]

• INT2.1: Towards Fine-Tunable Quantized Large Language Models with Error Correction through Low-Rank Adaptation
  Arxiv 2023 [Paper]

• LQ-LoRA: Low-rank Plus Quantized Matrix Decomposition for Efficient Language Model Finetuning
  Arxiv 2023 [Paper]

• INT-FP-QSim: Mixed Precision and Formats For Large Language Models and Vision Transformers
  Arxiv 2023 [Paper] [Code]

• QIGen: Generating Efficient Kernels for Quantized Inference on Large Language Models
  Arxiv 2023 [Paper] [Code]

• Do Emergent Abilities Exist in Quantized Large Language Models: An Empirical Study
  Arxiv 2023 [Paper]

• ZeroQuant-FP: A Leap Forward in LLMs Post-Training W4A8 Quantization Using Floating-Point Formats
  Arxiv 2023 [Paper] [Code (DeepSpeed)]

• OliVe: Accelerating Large Language Models via Hardware-friendly Outlier-Victim Pair Quantization
  ISCA 2023 [Paper]

• NUPES : Non-Uniform Post-Training Quantization via Power Exponent Search
  Arxiv 2023 [Paper]

• GPT-Zip: Deep Compression of Finetuned Large Language Models
  ICML 2023 Workshop ES-FoMO [Paper]

• Generating Efficient Kernels for Quantized Inference on Large Language Models
  ICML 2023 Workshop ES-FoMO [Paper]

• Gradient-Based Post-Training Quantization: Challenging the Status Quo
  Arxiv 2023 [Paper]

• FineQuant: Unlocking Efficiency with Fine-Grained Weight-Only Quantization for LLMs
  Arxiv 2023 [Paper]

• OmniQuant: Omnidirectionally Calibrated Quantization for Large Language Models
  ICLR 2024 [Paper] [Code]

• FPTQ: Fine-grained Post-Training Quantization for Large Language Models
  Arxiv 2023 [Paper]

• eDKM: An Efficient and Accurate Train-time Weight Clustering for Large Language Models
  Arxiv 2023 [Paper]

• QuantEase: Optimization-based Quantization for Language Models -- An Efficient and Intuitive Algorithm
  Arxiv 2023 [Paper]

• Norm Tweaking: High-performance Low-bit Quantization of Large Language Models
  Arxiv 2023 [Paper]

• Understanding the Impact of Post-Training Quantization on Large-scale Language Models
  Arxiv 2023 [Paper]

• MEMORY-VQ: Compression for Tractable Internet-Scale Memory
  Arxiv 2023 [Paper]

• Optimize Weight Rounding via Signed Gradient Descent for the Quantization of LLMs
  Arxiv 2023 [Paper] [Code]

• Efficient Post-training Quantization with FP8 Formats
  Arxiv 2023 [Paper] [Code (Intel® Neural Compressor)]

• QA-LoRA: Quantization-Aware Low-Rank Adaptation of Large Language Models
  Arxiv 2023 [Paper] [Code]

• Rethinking Channel Dimensions to Isolate Outliers for Low-bit Weight Quantization of Large Language Models
  Arxiv 2023 [Paper]

• ModuLoRA: Finetuning 3-Bit LLMs on Consumer GPUs by Integrating with Modular Quantizers
  Arxiv 2023 [Paper]

• PB-LLM: Partially Binarized Large Language Models
  Arxiv 2023 [Paper] [Code]

• Dual Grained Quantization: Efficient Fine-Grained Quantization for LLM
  Arxiv 2023 [Paper]

• Rethinking Channel Dimensions to Isolate Outliers for Low-bit Weight Quantization of Large Language Models
  Arxiv 2023 [Paper]

• QLLM: Accurate and Efficient Low-Bitwidth Quantization for Large Language Models
  Arxiv 2023 [Paper]

• LoftQ: LoRA-Fine-Tuning-Aware Quantization for Large Language Models
  Arxiv 2023 [Paper]

• QFT: Quantized Full-parameter Tuning of LLMs with Affordable Resources
  Arxiv 2023 [Paper]

• TEQ: Trainable Equivalent Transformation for Quantization of LLMs
  Arxiv 2023 [Paper] [Code (Intel® Neural Compressor)]

• BitNet: Scaling 1-bit Transformers for Large Language Models
  Arxiv 2023 [Paper] [Code]

• FP8-LM: Training FP8 Large Language Models
  Arxiv 2023 [Paper] [Code]

• QUIK: Towards End-to-End 4-Bit Inference on Generative Large Language Models
  Arxiv 2023 [Paper] [Code]

• AFPQ: Asymmetric Floating Point Quantization for LLMs
  Arxiv 2023 [Paper] [Code]

• AWEQ: Post-Training Quantization with Activation-Weight Equalization for Large Language Models
  Arxiv 2023 [Paper]

• Atom: Low-bit Quantization for Efficient and Accurate LLM Serving
  Arxiv 2023 [Paper]

• QMoE: Practical Sub-1-Bit Compression of Trillion-Parameter Models
  Arxiv 2023 [Paper]

• Dissecting the Runtime Performance of the Training, Fine-tuning, and Inference of Large Language Models
  Arxiv 2023 [Paper]

• How Does Calibration Data Affect the Post-training Pruning and Quantization of Large Language Models?
  Arxiv 2023 [Paper]

• A Speed Odyssey for Deployable Quantization of LLMs
  Arxiv 2023 [Paper]

• Enabling Fast 2-bit LLM on GPUs: Memory Alignment, Sparse Outlier, and Asynchronous Dequantization
  Arxiv 2023 [Paper]

• Quantizable Transformers: Removing Outliers by Helping Attention Heads Do Nothing
  NeurIPS 2023 [Paper] [Code]

• Efficient LLM Inference on CPUs
  NeurIPS 2023 on Efficient Natural Language and Speech Processing [Paper] [Code]

• The Cost of Compression: Investigating the Impact of Compression on Parametric Knowledge in Language Models
  EMNLP Findings 2023 [Paper]

• Zero-Shot Sharpness-Aware Quantization for Pre-trained Language Models
  EMNLP 2023 [Paper]

• Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference?
  EMNLP 2023 [Paper] [Code]

• Outlier Suppression+: Accurate quantization of large language models by equivalent and optimal shifting and scaling
  EMNLP 2023 [Paper]

• Watermarking LLMs with Weight Quantization
  EMNLP 2023 [Paper] [Code]

• Enhancing Computation Efficiency in Large Language Models through Weight and Activation Quantization
  EMNLP 2023 [Paper]

• LLM-FP4: 4-Bit Floating-Point Quantized Transformers
  EMNLP 2023 [Paper] [Code]

• Agile-Quant: Activation-Guided Quantization for Faster Inference of LLMs on the Edge
  AAAI 2024 [Paper]

• SmoothQuant+: Accurate and Efficient 4-bit Post-Training WeightQuantization for LLM
  Arxiv 2023 [Paper]

• CBQ: Cross-Block Quantization for Large Language Models
  Arxiv 2023 [Paper]

• ZeroQuant(4+2): Redefining LLMs Quantization with a New FP6-Centric Strategy for Diverse Generative Tasks
  Arxiv 2023 [Paper]

• QuIP: 2-Bit Quantization of Large Language Models With Guarantees
  NeurIPS 2023 [Paper] [Code]

• A Performance Evaluation of a Quantized Large Language Model on Various Smartphones
  Arxiv 2023 [Paper]

• DeltaZip: Multi-Tenant Language Model Serving via Delta Compression
  Arxiv 2023 [Paper] [Code]

• FlightLLM: Efficient Large Language Model Inference with a Complete Mapping Flow on FPGA
  Arxiv 2024 [Paper]

• Extreme Compression of Large Language Models via Additive Quantization
  Arxiv 2024 [Paper]

• Quantized Side Tuning: Fast and Memory-Efficient Tuning of Quantized Large Language Models
  Arxiv 2024 [Paper]

• Inferflow: an Efficient and Highly Configurable Inference Engine for Large Language Models
  Arxiv 2024 [Paper]

• FP6-LLM: Efficiently Serving Large Language Models Through FP6-Centric Algorithm-System Co-Design
  Arxiv 2024 [Paper]

• KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization
  Arxiv 2024 [Paper]

• Can Large Language Models Understand Context?
  Arxiv 2024 [Paper]

• EdgeQAT: Entropy and Distribution Guided Quantization-Aware Training for the Acceleration of Lightweight LLMs on the Edge
  Arxiv 2024 [Paper] [Code]

• Any-Precision LLM: Low-Cost Deployment of Multiple, Different-Sized LLMs
  Arxiv 2024 [Paper]

• LQER: Low-Rank Quantization Error Reconstruction for LLMs
  Arxiv 2024 [Paper]

• KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache
  Arxiv 2024 [Paper] [Code]

• BiLLM: Pushing the Limit of Post-Training Quantization for LLMs
  Arxiv 2024 [Paper] [Code]

• QuIP#: Even Better LLM Quantization with Hadamard Incoherence and Lattice Codebooks
  Arxiv 2024 [Paper] [Code]

• L4Q: Parameter Efficient Quantization-Aware Training on Large Language Models via LoRA-wise LSQ
  Arxiv 2024 [Paper]

• TP-Aware Dequantization
  Arxiv 2024 [Paper]

• ApiQ: Finetuning of 2-Bit Quantized Large Language Model
  Arxiv 2024 [Paper]

• Accurate LoRA-Finetuning Quantization of LLMs via Information Retention
  Arxiv 2024 [Paper] [Code]

• BitDelta: Your Fine-Tune May Only Be Worth One Bit
  Arxiv 2024 [Paper] [Code]

• QDyLoRA: Quantized Dynamic Low-Rank Adaptation for Efficient Large Language Model Tuning
  AAAI EIW Workshop 2024 [Paper]

• Any-Precision LLM: Low-Cost Deployment of Multiple, Different-Sized LLMs
  Arxiv 2024 [Paper]

• BitDistiller: Unleashing the Potential of Sub-4-Bit LLMs via Self-Distillation
  Arxiv 2024 [Paper] [Code]

• OneBit: Towards Extremely Low-bit Large Language Models
  Arxiv 2024 [Paper]

• DB-LLM: Accurate Dual-Binarization for Efficient LLMs
  Arxiv 2024 [Paper]

• WKVQuant: Quantizing Weight and Key/Value Cache for Large Language Models Gains More
  Arxiv 2024 [Paper]

• GPTVQ: The Blessing of Dimensionality for LLM Quantization
  Arxiv 2024 [Paper] [Code]

• APTQ: Attention-aware Post-Training Mixed-Precision Quantization for Large Language Models
  DAC 2024 [Paper]

• A Comprehensive Evaluation of Quantization Strategies for Large Language Models
  DAC 2024 [Paper]

• No Token Left Behind: Reliable KV Cache Compression via Importance-Aware Mixed Precision Quantization
  Arxiv 2024 [Paper]

• Evaluating Quantized Large Language Models
  Arxiv 2024 [Paper]

• FlattenQuant: Breaking Through the Inference Compute-bound for Large Language Models with Per-tensor Quantization
  Arxiv 2024 [Paper]

• LLM-PQ: Serving LLM on Heterogeneous Clusters with Phase-Aware Partition and Adaptive Quantization
  Arxiv 2024 [Paper]

• IntactKV: Improving Large Language Model Quantization by Keeping Pivot Tokens Intact
  Arxiv 2024 [Paper]

• On the Compressibility of Quantized Large Language Models
  Arxiv 2024 [Paper]

• EasyQuant: An Efficient Data-free Quantization Algorithm for LLMs
  Arxiv 2024 [Paper]

• QAQ: Quality Adaptive Quantization for LLM KV Cache
  Arxiv 2024 [Paper] [Code]

• GEAR: An Efficient KV Cache Compression Recipefor Near-Lossless Generative Inference of LLM
  Arxiv 2024 [Paper]

• What Makes Quantization for Large Language Models Hard? An Empirical Study from the Lens of Perturbation
  Arxiv 2024 [Paper]

• SVD-LLM: Truncation-aware Singular Value Decomposition for Large Language Model Compression
  Arxiv 2024 [Paper] [Code]

• AffineQuant: Affine Transformation Quantization for Large Language Models
  ICLR 2024 [Paper] [Code]

• Oh! We Freeze: Improving Quantized Knowledge Distillation via Signal Propagation Analysis for Large Language Models
  ICLR Practical ML for Low Resource Settings Workshop 2024 [Paper]

• Accurate Block Quantization in LLMs with Outliers
  Arxiv 2024 [Paper]

• QuaRot: Outlier-Free 4-Bit Inference in Rotated LLMs
  Arxiv 2024 [Paper] [Code]

• Minimize Quantization Output Error with Bias Compensation
  Arxiv 2024 [Paper] [Code]

• Cherry on Top: Parameter Heterogeneity and Quantization in Large Language Models
  Arxiv 2024 [Paper]

• Increased LLM Vulnerabilities from Fine-tuning and Quantization
  Arxiv 2024 [Paper]

• Quantization of Large Language Models with an Overdetermined Basis
  Arxiv 2024 [Paper]

• How Good Are Low-bit Quantized LLaMA3 Models? An Empirical Study
  Arxiv 2024 [Paper] [Code] [Model]

• How to Parameterize Asymmetric Quantization Ranges for Quantization-Aware Training
  Arxiv 2024 [Paper]

• When Quantization Affects Confidence of Large Language Models?
  NAACL 2024 [Paper]

Pruning and Sparsity

• The Lazy Neuron Phenomenon: On Emergence of Activation Sparsity in Transformers
  ICLR 2023 [Paper]

• Deja Vu: Contextual Sparsity for Efficient LLMs at Inference Time
  ICML 2023 [Paper] [Code]

• LoSparse: Structured Compression of Large Language Models based on Low-Rank and Sparse Approximation
  ICML 2023 [Paper] [Code]

• LLM-Pruner: On the Structural Pruning of Large Language Models
  NeurIPS 2023 [Paper] [Code]

• ZipLM: Inference-Aware Structured Pruning of Language Models
  NeurIPS 2023 [Paper] [Code]

• H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models
  NeurIPS 2023 [Paper] [Code]

• Scissorhands: Exploiting the Persistence of Importance Hypothesis for LLM KV Cache Compression at Test Time
  NeurIPS 2023 [Paper]

• The Emergence of Essential Sparsity in Large Pre-trained Models: The Weights that Matter
  NeurIPS 2023 [Paper] [Code]

• Learning to Compress Prompts with Gist Tokens
  NeurIPS 2023 [Paper]

• Dynamic Context Pruning for Efficient and Interpretable Autoregressive Transformers
  NeurIPS 2023 [Paper]

• Prune and Tune: Improving Efficient Pruning Techniques for Massive Language Models
  ICLR 2023 TinyPapers [Paper]

• SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot
  Arxiv 2023 [Paper] [Code]

• AdaLoRA: Adaptive Budget Allocation for Parameter-Efficient Fine-Tuning
  Arxiv 2023 [Paper]

• Unlocking Context Constraints of LLMs: Enhancing Context Efficiency of LLMs with Self-Information-Based Content Filtering
  Arxiv 2023 [Paper] [Code]

• Rethinking the Role of Scale for In-Context Learning: An Interpretability-based Case Study at 66 Billion Scale
  ACL 2023 [Paper] [Code]

• Structured Pruning for Efficient Generative Pre-trained Language Models
  ACL 2023 [Paper]

• A Simple and Effective Pruning Approach for Large Language Models
  Arxiv 2023 [Paper] [Code]

• Pruning Meets Low-Rank Parameter-Efficient Fine-Tuning
  Arxiv 2023 [Paper]

• Structural pruning of large language models via neural architecture search
  AutoML 2023 [Paper]

• Pruning Large Language Models via Accuracy Predictor
  ICASSP 2024 [Paper]

• Flash-LLM: Enabling Cost-Effective and Highly-Efficient Large Generative Model Inference with Unstructured Sparsity
  VLDB 2024 [Paper] [Cde]

• Compressing LLMs: The Truth is Rarely Pure and Never Simple
  Arxiv 2023 [Paper]

• Junk DNA Hypothesis: A Task-Centric Angle of LLM Pre-trained Weights through Sparsity
  Arxiv 2023 [Paper] [Code]

• Model Tells You What to Discard: Adaptive KV Cache Compression for LLMs
  Arxiv 2023 [Paper]

• Compresso: Structured Pruning with Collaborative Prompting Learns Compact Large Language Models
  Arxiv 2023 [Paper] [Code]

• Outlier Weighed Layerwise Sparsity (OWL): A Missing Secret Sauce for Pruning LLMs to High Sparsity
  Arxiv 2023 [Paper] [Code]

• Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning
  Arxiv 2023 [Paper] [Code]

• Dynamic Sparse No Training: Training-Free Fine-tuning for Sparse LLMs
  Arxiv 2023 [Paper] [Code]

• One-Shot Sensitivity-Aware Mixed Sparsity Pruning for Large Language Models
  ICASSP 2024 [Paper]

• Survival of the Most Influential Prompts: Efficient Black-Box Prompt Search via Clustering and Pruning
  EMNLP 2023 Findings [Paper]

• The Cost of Compression: Investigating the Impact of Compression on Parametric Knowledge in Language Models
  EMNLP Findings 2023 [Paper]

• Divergent Token Metrics: Measuring degradation to prune away LLM components -- and optimize quantization
  Arxiv 2023 [Paper]

• LoRAShear: Efficient Large Language Model Structured Pruning and Knowledge Recovery
  Arxiv 2023 [Paper]

• ReLU Strikes Back: Exploiting Activation Sparsity in Large Language Models
  Arxiv 2023 [Paper]

• E-Sparse: Boosting the Large Language Model Inference through Entropy-based N:M Sparsity
  Arxiv 2023 [Paper]

• Beyond Size: How Gradients Shape Pruning Decisions in Large Language Models
  Arxiv 2023 [Paper] [Code]

• How Does Calibration Data Affect the Post-training Pruning and Quantization of Large Language Models?
  Arxiv 2023 [Paper]

• BESA: Pruning Large Language Models with Blockwise Parameter-Efficient Sparsity Allocation
  OpenReview [Paper] [Code]

• PUSHING GRADIENT TOWARDS ZERO: A NOVEL PRUNING METHOD FOR LARGE LANGUAGE MODELS
  OpenReview 2023 [Paper]

• An Efficient Plug-and-Play Post-Training Pruning Strategy in Large Language Models
  Preprints 2023 [Paper]

• Lighter, yet More Faithful: Investigating Hallucinations in Pruned Large Language Models for Abstractive Summarization
  Arxiv 2023 [Paper] [Code]

• LORAPRUNE: PRUNING MEETS LOW-RANK PARAMETER-EFFICIENT FINE-TUNING
  Arxiv 2023 [Paper]

• Mini-GPTs: Efficient Large Language Models through Contextual Pruning
  Arxiv 2023 [Paper] [Code]

• The LLM Surgeon
  Arxiv 2023 [Paper]

• Fluctuation-based Adaptive Structured Pruning for Large Language Models
  AAAI 2024 [Paper]

• How to Prune Your Language Model: Recovering Accuracy on the "Sparsity May Cry'' Benchmark
  CPAL 2024 [Paper]

• PERP: Rethinking the Prune-Retrain Paradigm in the Era of LLMs
  Arxiv 2023 [Paper]

• Fast and Optimal Weight Update for Pruned Large Language Models
  Arxiv 2024 [Paper]

• APT: Adaptive Pruning and Tuning Pretrained Language Models for Efficient Training and Inference
  Arxiv 2024 [Paper]

• Scaling Sparse Fine-Tuning to Large Language Models
  Arxiv 2024 [Paper]

• SliceGPT: Compress Large Language Models by Deleting Rows and Columns
  ICLR 2024 [Paper] [Code]

• Shortened LLaMA: A Simple Depth Pruning for Large Language Models
  Arxiv 2024 [Paper]

• Everybody Prune Now: Structured Pruning of LLMs with only Forward Passes
  Arxiv 2024 [Paper] [Code]

• NutePrune: Efficient Progressive Pruning with Numerous Teachers for Large Language Models
  Arxiv 2024 [Paper]

• LaCo: Large Language Model Pruning via Layer Collapse
  Arxiv 2024 [Paper]

• Why Lift so Heavy? Slimming Large Language Models by Cutting Off the Layers
  Arxiv 2024 [Paper]

• EBFT: Effective and Block-Wise Fine-Tuning for Sparse LLMs
  Arxiv 2024 [Paper] [Code]

• Data-free Weight Compress and Denoise for Large Language Models
  Arxiv 2024 [Paper]

• Gradient-Free Adaptive Global Pruning for Pre-trained Language Models
  Arxiv 2024 [Paper]

• ShortGPT: Layers in Large Language Models are More Redundant Than You Expect
  Arxiv 2024 [Paper]

• LLaVA-PruMerge: Adaptive Token Reduction for Efficient Large Multimodal Models
  Arxiv 2024 [Paper] [Code]

• Compressing Large Language Models by Streamlining the Unimportant Layer
  Arxiv 2024 [Paper]

• LoRAP: Transformer Sub-Layers Deserve Differentiated Structured Compression for Large Language Models
  Arxiv 2024 [Paper]

• Shears: Unstructured Sparsity with Neural Low-rank Adapter Search
  NAACL 2024 [Paper]

• Eigenpruning
  NAACL 2024 Abstract [Paper]

Distillation

• Lifting the Curse of Capacity Gap in Distilling Language Models
  ACL 2023 [Paper] [Code]

• Symbolic Chain-of-Thought Distillation: Small Models Can Also "Think" Step-by-Step
  ACL 2023 [Paper]

• Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes
  ACL 2023 [Paper]

• SCOTT: Self-Consistent Chain-of-Thought Distillation
  ACL 2023 [Paper]

• DISCO: Distilling Counterfactuals with Large Language Models
  ACL 2023 [Paper] [Code]

• LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions
  Arxiv 2023 [Paper] [Code]

• How To Train Your (Compressed) Large Language Model
  Arxiv 2023 [Paper]

• The False Promise of Imitating Proprietary LLMs
  Arxiv 2023 [Paper]

• GPT4All: Training an Assistant-style Chatbot with Large Scale Data Distillation from GPT-3.5-Turbo
  Arxiv 2023 [Paper] [Code]

• PaD: Program-aided Distillation Specializes Large Models in Reasoning
  Arxiv 2023 [Paper]

• Knowledge Distillation of Large Language Models
  Arxiv 2023 [Paper] [Code]

• GKD: Generalized Knowledge Distillation for Auto-regressive Sequence Models
  Arxiv 2023 [Paper]

• Chain-of-Thought Prompt Distillation for Multimodal Named Entity and Multimodal Relation Extraction
  Arxiv 2023 [Paper]

• Task-agnostic Distillation of Encoder-Decoder Language Models
  Arxiv 2023 [Paper]

• Sci-CoT: Leveraging Large Language Models for Enhanced Knowledge Distillation in Small Models for Scientific QA
  Arxiv 2023 [Paper]

• Can a student Large Language Model perform as well as it's teacher?
  Arxiv 2023 [Paper]

• Multistage Collaborative Knowledge Distillation from Large Language Models
  Arxiv 2023 [Paper]

• Lion: Adversarial Distillation of Closed-Source Large Language Model
  EMNLP 2023 [Paper] [Code]

• MCC-KD: Multi-CoT Consistent Knowledge Distillation
  EMNLP 2023 [Paper]

• PromptMix: A Class Boundary Augmentation Method for Large Language Model Distillation
  EMNLP 2023 [Paper]

• YODA: Teacher-Student Progressive Learning for Language Models
  Arxiv 2023 [Paper]

• Knowledge Fusion of Large Language Models
  ICLR 2024 [Paper] [Code]

• Knowledge Distillation for Closed-Source Language Models
  Arxiv 2024 [Paper]

• TinyLLM: Learning a Small Student from Multiple Large Language Models
  Arxiv 2024 [Paper]

• Towards Cross-Tokenizer Distillation: the Universal Logit Distillation Loss for LLMs
  Arxiv 2024 [Paper]

• Revisiting Knowledge Distillation for Autoregressive Language Models
  Arxiv 2024 [Paper]

• Sinkhorn Distance Minimization for Knowledge Distillation
  COLING 2024 [Paper]

• Learning to Maximize Mutual Information for Chain-of-Thought Distillation
  Arxiv 2024 [Paper]

• DistiLLM: Towards Streamlined Distillation for Large Language Models
  Arxiv 2024 [Paper] [Code]

• Efficiently Distilling LLMs for Edge Applications
  NAACL 2024 [Paper]

• Rethinking Kullback-Leibler Divergence in Knowledge Distillation for Large Language Models
  Arxiv 2024 [Paper]

• Distilling Algorithmic Reasoning from LLMs via Explaining Solution Programs
  Arxiv 2024 [Paper]

Efficient Prompting

• Did You Read the Instructions? Rethinking the Effectiveness of Task Definitions in Instruction Learning
  ACL 2023 [Paper] [Code]

• Batch Prompting: Efficient Inference with Large Language Model APIs
  EMNLP 2023 [Paper] [Code]

• Adapting Language Models to Compress Contexts
  EMNLP 2023 [Paper] [Code]

• Compressing Context to Enhance Inference Efficiency of Large Language Models
  EMNLP 2023 [Paper] [Code]

• LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models
  EMNLP 2023 [Paper] [Code]

• Vector-Quantized Prompt Learning for Paraphrase Generation
  EMNLP 2023 Findings [Paper]

• Efficient Prompting via Dynamic In-Context Learning
  Arxiv 2023 [Paper]

• Learning to Compress Prompts with Gist Tokens
  Arxiv 2023 [Paper] [Code]

• In-context Autoencoder for Context Compression in a Large Language Model
  Arxiv 2023 [Paper]

• Discrete Prompt Compression with Reinforcement Learning
  Arxiv 2023 [Paper]

• BatchPrompt: Accomplish more with less
  Arxiv 2023 [Paper]

• (Dynamic) Prompting might be all you need to repair Compressed LLMs
  Arxiv 2023 [Paper]

• RECOMP: Improving Retrieval-Augmented LMs with Compression and Selective Augmentation
  Arxiv 2023 [Paper] [Code]

• LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compression
  Arxiv 2023 [Paper] [Code]

• Extending Context Window of Large Language Models via Semantic Compression
  Arxiv 2023 [Paper]

• Boosting LLM Reasoning: Push the Limits of Few-shot Learning with Reinforced In-Context Pruning
  Arxiv 2023 [Paper]

• The Impact of Reasoning Step Length on Large Language Models
  Arxiv 2024 [Paper]

• Compressed Context Memory For Online Language Model Interaction
  ICLR 2024 [Paper] [Code]

• Learning to Compress Prompt in Natural Language Formats
  Arxiv 2024 [Paper]

• Say More with Less: Understanding Prompt Learning Behaviors through Gist Compression
  Arxiv 2024 [Paper] [Code]

• StreamingDialogue: Prolonged Dialogue Learning via Long Context Compression with Minimal Losses
  Arxiv 2024 [Paper]

• LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compression
  Arxiv 2024 [Paper] [Code]

• PCToolkit: A Unified Plug-and-Play Prompt Compression Toolkit of Large Language Models
  Arxiv 2024 [Paper] [Code]

• PROMPT-SAW: Leveraging Relation-Aware Graphs for Textual Prompt Compression
  Arxiv 2024 [Paper]

• Prompts As Programs: A Structure-Aware Approach to Efficient Compile-Time Prompt Optimization
  Arxiv 2024 [Paper] [Code]

• Adapting LLMs for Efficient Context Processing through Soft Prompt Compression
  Arxiv 2024 [Paper]

Other

• FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
  Arxiv 2022 [Paper]

• TensorGPT: Efficient Compression of the Embedding Layer in LLMs based on the Tensor-Train Decomposition
  Arxiv 2023 [Paper]

• Dynamic Context Pruning for Efficient and Interpretable Autoregressive Transformers
  Arxiv 2023 [Paper]

• SkipDecode: Autoregressive Skip Decoding with Batching and Caching for Efficient LLM Inference
  Arxiv 2023 [Paper]

• Scaling In-Context Demonstrations with Structured Attention
  Arxiv 2023 [Paper]

• Response Length Perception and Sequence Scheduling: An LLM-Empowered LLM Inference Pipeline
  Arxiv 2023 [Paper] [Code]

• CPET: Effective Parameter-Efficient Tuning for Compressed Large Language Models
  Arxiv 2023 [Paper]

• Ternary Singular Value Decomposition as a Better Parameterized Form in Linear Mapping
  Arxiv 2023 [Paper]

• LLMCad: Fast and Scalable On-device Large Language Model Inference
  Arxiv 2023 [Paper]

• vLLM: Efficient Memory Management for Large Language Model Serving with PagedAttention
  Arxiv 2023 [Paper]

• LongLoRA: Efficient Fine-tuning of Long-Context Large Language Models
  Arxiv 2023 [Paper] [Code]

• LORD: Low Rank Decomposition Of Monolingual Code LLMs For One-Shot Compression
  Arxiv 2023 [Paper] [Code]

• Mixture of Tokens: Efficient LLMs through Cross-Example Aggregation
  Arxiv 2023 [Paper]

• Efficient Streaming Language Models with Attention Sinks
  Arxiv 2023 [Paper] [Code]

• Efficient Large Language Models Fine-Tuning On Graphs
  Arxiv 2023 [Paper]

• SparQ Attention: Bandwidth-Efficient LLM Inference
  Arxiv 2023 [Paper]

• Rethinking Compression: Reduced Order Modelling of Latent Features in Large Language Models
  Arxiv 2023 [Paper]

• PowerInfer: Fast Large Language Model Serving with a Consumer-grade GPU
  Arxiv 2023 [Paper] [Code]

• Dataset Quantization
  ICCV 2023 [Paper] [Code]

• Text Alignment Is An Efficient Unified Model for Massive NLP Tasks
  NeurIPS 2023 [Paper] [Code]

• Context Compression for Auto-regressive Transformers with Sentinel Tokens
  EMNLP 2023 [Paper] [Code]

• TCRA-LLM: Token Compression Retrieval Augmented Large Language Model for Inference Cost Reduction
  EMNLP 2023 Findings [Paper]

• Retrieval-based Knowledge Transfer: An Effective Approach for Extreme Large Language Model Compression
  EMNLP 2023 Findings [Paper]

• FFSplit: Split Feed-Forward Network For Optimizing Accuracy-Efficiency Trade-off in Language Model Inference
  Arxiv 2024 [Paper]

• LoMA: Lossless Compressed Memory Attention
  Arxiv 2024 [Paper]

• Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads
  Arxiv 2024 [Paper] [Code]

• BiTA: Bi-Directional Tuning for Lossless Acceleration in Large Language Models
  Arxiv 2024 [Paper] [Code]

• CompactifAI: Extreme Compression of Large Language Models using Quantum-Inspired Tensor Networks
  Arxiv 2024 [Paper]

• NoMAD-Attention: Efficient LLM Inference on CPUs Through Multiply-add-free Attention
  Arxiv 2024 [Paper]

• Not all Layers of LLMs are Necessary during Inference
  Arxiv 2024 [Paper]

• GaLore: Memory-Efficient LLM Training by Gradient Low-Rank Projection
  Arxiv 2024 [Paper]

• Dynamic Memory Compression: Retrofitting LLMs for Accelerated Inference
  Arxiv 2024 [Paper]

• Smart-Infinity: Fast Large Language Model Training using Near-Storage Processing on a Real System
  HPCA 2024 [Paper]

• Keyformer: KV Cache Reduction through Key Tokens Selection for Efficient Generative Inference
  MLSys 2024 [Paper]

• ALoRA: Allocating Low-Rank Adaptation for Fine-tuning Large Language Models
  Arxiv 2024 [Paper]

• Training LLMs over Neurally Compressed Text
  Arxiv 2024 [Paper]

• TriForce: Lossless Acceleration of Long Sequence Generation with Hierarchical Speculative Decoding
  Arxiv 2024 [Paper] [Code]

• SnapKV: LLM Knows What You are Looking for Before Generation
  Arxiv 2024 [Paper] [Code]

Tools

• BMCook: Model Compression for Big Models [Code]

• llama.cpp: Inference of LLaMA model in pure C/C++ [Code]

• LangChain: Building applications with LLMs through composability [Code]

• GPTQ-for-LLaMA: 4 bits quantization of LLaMA using GPTQ [Code]

• Alpaca-CoT: An Instruction Fine-Tuning Platform with Instruction Data Collection and Unified Large Language Models Interface [Code]

• vllm: A high-throughput and memory-efficient inference and serving engine for LLMs [Code]

• LLaMA Efficient Tuning: Fine-tuning LLaMA with PEFT (PT+SFT+RLHF with QLoRA) [Code]

• gpt-fast: Simple and efficient pytorch-native transformer text generation in <1000 LOC of python. [Code]

• Efficient-Tuning-LLMs: (Efficient Finetuning of QLoRA LLMs). QLoRA, LLama, bloom, baichuan-7B, GLM [Code]

• bitsandbytes: 8-bit CUDA functions for PyTorch [Code]

• ExLlama: A more memory-efficient rewrite of the HF transformers implementation of Llama for use with quantized weights. [Code]

• lit-gpt: Hackable implementation of state-of-the-art open-source LLMs based on nanoGPT. Supports flash attention, 4-bit and 8-bit quantization, LoRA and LLaMA-Adapter fine-tuning, pre-training. [Code]

• Lit-LLaMA: Implementation of the LLaMA language model based on nanoGPT. Supports flash attention, Int8 and GPTQ 4bit quantization, LoRA and LLaMA-Adapter fine-tuning, pre-training. [Code]

• lama.onnx: LLaMa/RWKV onnx models, quantization and testcase [Code]

• fastLLaMa: An experimental high-performance framework for running Decoder-only LLMs with 4-bit quantization in Python using a C/C++ backend. [Code]

• Sparsebit: A model compression and acceleration toolbox based on pytorch. [Code]

• llama2.c: Inference Llama 2 in one file of pure C [Code]

• Megatron-LM: Ongoing research training transformer models at scale [Code]

• ggml: Tensor library for machine learning [Code]

• LLamaSharp: C#/.NET binding of llama.cpp, including LLaMa/GPT model inference and quantization, ASP.NET core integration and UI [Code]

• rwkv.cpp: NT4/INT5/INT8 and FP16 inference on CPU for RWKV language model [Code]

• Can my GPU run this LLM?: Calculate GPU memory requirement & breakdown for training/inference of LLM models. Supports ggml/bnb quantization [Code]

• TinyChatEngine: On-Device LLM Inference Library [Code]

• TensorRT-LLM: TensorRT-LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and build TensorRT engines that contain state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. [Code]

• IntLLaMA: A fast and light quantization solution for LLaMA [Code]

• EasyLLM: Built upon Megatron-Deepspeed and HuggingFace Trainer, EasyLLM has reorganized the code logic with a focus on usability. While enhancing usability, it also ensures training efficiency [Code]

• GreenBit LLaMA: Advanced Ultra-Low Bitrate Compression Techniques for the LLaMA Family of LLMs [Code]

• Intel® Neural Compressor: An open-source Python library supporting popular model compression techniques on all mainstream deep learning frameworks (TensorFlow, PyTorch, ONNX Runtime, and MXNet) [Code]

• LLM-Viewer: Analyze the inference of Large Language Models (LLMs). Analyze aspects like computation, storage, transmission, and hardware roofline model in a user-friendly interface. [Code]

• LLaMA3-Quantization: A repository dedicated to evaluating the performance of quantizied LLaMA3 using various quantization methods. [Code]

• LLamaSharp: A C#/.NET library to run LLM models (🦙LLaMA/LLaVA) on your local device efficiently. [Code]

• Green-bit-LLM: A toolkit for fine-tuning, inferencing, and evaluating GreenBitAI's LLMs. [Code] [Model]

• Bitorch Engine: Streamlining AI with Open-Source Low-Bit Quantization. [Code]

Contributing

This is an active repository and your contributions are always welcome! Before you add papers/tools into the awesome list, please make sure that:

• The paper or tools is related to Large Language Models (LLMs). If the compression algorithms or tools are only evaluated on small-scale language models (e.g., BERT), they should not be included in the list.
• The paper should be inserted in the correct position in chronological order (publication/arxiv release time).
• The link to [Paper] should be the arxiv page, not the pdf page if this is a paper posted on arxiv.

Star History