# SPDF: Sparse Pre-training and Dense Fine-tuning for Large Language Models

arXiv: https://arxiv.org/abs/2303.10464 · HTML: https://mlsanity.com/paper/2303.10464/

## Key facts

- Cited by 28 papers in the map (data through 14 Oct 2026).
- Builds on 58 of its 87 references that are in the map.
- First posted on arXiv 18 Mar 2023.

## Builds on (most cited first, 30 of 58)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Layer Normalization](https://mlsanity.com/paper/1607.06450/) (2016)
- [PaLM: Scaling Language Modeling with Pathways](https://mlsanity.com/paper/2204.02311/) (2022)
- [Parameter-Efficient Transfer Learning for NLP](https://mlsanity.com/paper/1902.00751/) (2019)
- [Prefix-Tuning: Optimizing Continuous Prompts for Generation](https://mlsanity.com/paper/2101.00190/) (2021)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [Emergent Abilities of Large Language Models](https://mlsanity.com/paper/2206.07682/) (2022)
- [The Pile: An 800GB Dataset of Diverse Text for Language Modeling](https://mlsanity.com/paper/2101.00027/) (2020)
- [The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks](https://mlsanity.com/paper/1803.03635/) (2018)
- [CIDEr: Consensus-based Image Description Evaluation](https://mlsanity.com/paper/1411.5726/) (2014)
- [BitFit: Simple Parameter-efficient Fine-tuning for Transformer-based Masked Language-models](https://mlsanity.com/paper/2106.10199/) (2021)
- [LaMDA: Language Models for Dialog Applications](https://mlsanity.com/paper/2201.08239/) (2022)
- [Scaling Language Models: Methods, Analysis & Insights from Training Gopher](https://mlsanity.com/paper/2112.11446/) (2021)
- [SNIP: Single-shot Network Pruning based on Connection Sensitivity](https://mlsanity.com/paper/1810.02340/) (2018)
- [A Convergence Theory for Deep Learning via Over-Parameterization](https://mlsanity.com/paper/1811.03962/) (2018)
- [Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning](https://mlsanity.com/paper/1709.00103/) (2017)
- [Intrinsic Dimensionality Explains the Effectiveness of Language Model Fine-Tuning](https://mlsanity.com/paper/2012.13255/) (2020)
- [It's Not Just Size That Matters: Small Language Models Are Also Few-Shot Learners](https://mlsanity.com/paper/2009.07118/) (2020)
- [To prune, or not to prune: exploring the efficacy of pruning for model compression](https://mlsanity.com/paper/1710.01878/) (2017)
- [Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model](https://mlsanity.com/paper/2201.11990/) (2022)
- [Sparsity in Deep Learning: Pruning and growth for efficient inference and training in neural networks](https://mlsanity.com/paper/2102.00554/) (2021)
- [Compositional Semantic Parsing on Semi-Structured Tables](https://mlsanity.com/paper/1508.00305/) (2015)
- [Rigging the Lottery: Making All Tickets Winners](https://mlsanity.com/paper/1911.11134/) (2019)
- [Pruning neural networks without any data by iteratively conserving synaptic flow](https://mlsanity.com/paper/2006.05467/) (2020)
- [The State of Sparsity in Deep Neural Networks](https://mlsanity.com/paper/1902.09574/) (2019)

## Built on by (most cited first, 28 of 28)

- [APOLLO: SGD-like Memory, AdamW-level Performance](https://mlsanity.com/paper/2412.05270/) (2024)
- [SLTrain: a sparse plus low-rank approach for parameter and memory efficient pretraining](https://mlsanity.com/paper/2406.02214/) (2024)
- [Q-GaLore: Quantized GaLore with INT4 Projection and Layer-Adaptive Low-Rank Gradients](https://mlsanity.com/paper/2407.08296/) (2024)
- [SLoPe: Double-Pruned Sparse Plus Lazy Low-Rank Adapter Pretraining of LLMs](https://mlsanity.com/paper/2405.16325/) (2024)
- [Sparse maximal update parameterization: A holistic approach to sparse training dynamics](https://mlsanity.com/paper/2405.15743/) (2024)
- [ProMoAI: Process Modeling with Generative AI](https://mlsanity.com/paper/2403.04327/) (2024)
- [Reparameterized LLM Training via Orthogonal Equivalence Transformation](https://mlsanity.com/paper/2506.08001/) (2025)
- [SparseLoRA: Accelerating LLM Fine-Tuning with Contextual Sparsity](https://mlsanity.com/paper/2506.16500/) (2025)
- [Sparse-IFT: Sparse Iso-FLOP Transformations for Maximizing Training Efficiency](https://mlsanity.com/paper/2303.11525/) (2023)
- [Joint Localization and Activation Editing for Low-Resource Fine-Tuning](https://mlsanity.com/paper/2502.01179/) (2025)
- [A Comprehensive Performance Study of Large Language Models on Novel AI Accelerators](https://mlsanity.com/paper/2310.04607/) (2023)
- [CR-Net: Scaling Parameter-Efficient Training with Cross-Layer Low-Rank Structure](https://mlsanity.com/paper/2509.18993/) (2025)
- [Scalable Efficient Training of Large Language Models with Low-dimensional Projected Attention](https://mlsanity.com/paper/2411.02063/) (2024)
- [POET-X: Memory-efficient LLM Training by Scaling Orthogonal Transformation](https://mlsanity.com/paper/2603.05500/) (2026)
- [Brain network science modelling of sparse neural networks enables Transformers and LLMs to perform as fully connected](https://mlsanity.com/paper/2501.19107/) (2025)
- [An Efficient Subspace Algorithm for Federated Learning on Heterogeneous Data](https://mlsanity.com/paper/2509.05213/) (2025)
- [Enhancing Zero-Shot Crypto Sentiment with Fine-tuned Language Model and Prompt Engineering](https://mlsanity.com/paper/2310.13226/) (2023)
- [Communication-Efficient and Personalized Federated Foundation Model Fine-Tuning via Tri-Matrix Adaptation](https://mlsanity.com/paper/2503.23869/) (2025)
- [Memory-Efficient Fine-Tuning via Low-Rank Activation Compression](https://mlsanity.com/paper/2509.23472/) (2025)
- [SSFL: Discovering Sparse Unified Subnetworks at Initialization for Efficient Federated Learning](https://mlsanity.com/paper/2405.09037/) (2024)
- [Enhancing One-shot Pruned Pre-trained Language Models through Sparse-Dense-Sparse Mechanism](https://mlsanity.com/paper/2408.10473/) (2024)
- [SD2: Self-Distilled Sparse Drafters](https://mlsanity.com/paper/2504.08838/) (2025)
- [Sparsifying Spiking Networks through Local Rhythms](https://mlsanity.com/paper/2305.10191/) (2023)
- [The impact of allocation strategies in subset learning on the expressive power of neural networks](https://mlsanity.com/paper/2502.06300/) (2025)
- [Generalizing Scaling Laws for Dense and Sparse Large Language Models](https://mlsanity.com/paper/2508.06617/) (2025)
- [An Empirical Study of Sample Selection Strategies for Large Language Model Repair](https://mlsanity.com/paper/2510.20428/) (2025)
- [Compression Trinity: Exploring Sparsity, Quantization, and Low-Rank Approximations for LLM Compression](https://mlsanity.com/paper/2608.24070/) (2026)
- [Beyond Masked Sparsity: SNACK Enables Truly Sparse Neural Networks on GPU](https://mlsanity.com/paper/2610.04093/) (2026)

## Abstract (arXiv)

The pre-training and fine-tuning paradigm has contributed to a number of breakthroughs in Natural Language Processing (NLP). Instead of directly training on a downstream task, language models are first pre-trained on large datasets with cross-domain knowledge (e.g., Pile, MassiveText, etc.) and then fine-tuned on task-specific data (e.g., natural language generation, text summarization, etc.). Scaling the model and dataset size has helped improve the performance of LLMs, but unfortunately, this also lead to highly prohibitive computational costs. Pre-training LLMs often require orders of magnitude more FLOPs than fine-tuning and the model capacity often remains the same between the two phases. To achieve training efficiency w.r.t training FLOPs, we propose to decouple the model capacity between the two phases and introduce Sparse Pre-training and Dense Fine-tuning (SPDF). In this work, we show the benefits of using unstructured weight sparsity to train only a subset of weights during pre-training (Sparse Pre-training) and then recover the representational capacity by allowing the zeroed weights to learn (Dense Fine-tuning). We demonstrate that we can induce up to 75% sparsity into a 1.3B parameter GPT-3 XL model resulting in a 2.5x reduction in pre-training FLOPs, without a significant loss in accuracy on the downstream tasks relative to the dense baseline. By rigorously evaluating multiple downstream tasks, we also establish a relationship between sparsity, task complexity and dataset size. Our work presents a promising direction to train large GPT models at a fraction of the training FLOPs using weight sparsity, while retaining the benefits of pre-trained textual representations for downstream tasks.