Mike Lasby

Vithursan Thangarasa, Ganesh Venkatesh, Mike Lasby et al.

Large language models have driven significant progress in natural language processing, but their deployment requires substantial compute and memory resources. As models scale, compression techniques become essential for balancing model quality with computational efficiency. Structured pruning, which removes less critical components of the model, is a promising strategy for reducing complexity. However, one-shot pruning often results in significant quality degradation, particularly in tasks requiring multi-step reasoning. To recover lost quality, supervised fine-tuning (SFT) is commonly applied, but it can lead to catastrophic forgetting by shifting the model's learned data distribution. Therefore, addressing the degradation from both pruning and SFT is essential to preserve the original model's quality. In this work, we utilize self-data distilled fine-tuning to address these challenges. Our approach leverages the original, unpruned model to generate a distilled dataset that preserves semantic richness and mitigates catastrophic forgetting by maintaining alignment with the base model's knowledge. Empirically, we demonstrate that self-data distillation consistently outperforms standard SFT, improving average accuracy by up to 8% on the HuggingFace OpenLLM Leaderboard v1. Specifically, when pruning six decoder blocks on Llama3.1-8B Instruct (i.e., 32 to 26 layers, reducing the model size from 8.03B to 6.72B parameters), our method retains 91.2% of the original model's accuracy compared to 81.7% with SFT, while reducing real-world FLOPs by 16.3%. Furthermore, combining self-data distilled models through model merging yields enhanced quality retention. Additionally, leveraging these pruned models in speculative decoding increases token acceptance rates, thereby improving inference efficiency in applied settings.

Cem Üyük, Mike Lasby, Mohamed Yassin et al.

Large neural networks exhibit exceptional performance across numerous tasks, yet their considerable size often hinders deployment on resource-constrained systems. While various model compression strategies have been well studied, parameter sharing remains underexplored. In this paper, we introduce Fine-grained Parameter Sharing (FiPS), a novel algorithm that leverages parameter sharing, tensor decomposition, and sparsity to effectively compress large-scale Vision Transformers (ViTs) and Large Language Models (LLMs). FiPS employs a shared base and sparse factors to represent neurons across multi-layer perceptron (MLP) modules, where initialization is guided by Singular Value Decomposition (SVD) and subsequent optimization is conducted through block-wise reconstruction error minimization. Experimental results show that FiPS reduces the parameter budget of MLP modules by 50-75% for DeiT-B and Swin-L and by 40-50% for various Gemma-2 and Llama-3 models while maintaining ViT model accuracy within 1% pt. of the original and LLM perplexity with negligible degradation.

Nasib Ullah, Erik Schultheis, Mike Lasby et al.

In recent years, Dynamic Sparse Training (DST) has emerged as an alternative to post-training pruning for generating efficient models. In principle, DST allows for a more memory efficient training process, as it maintains sparsity throughout the entire training run. However, current DST implementations fail to capitalize on this in practice. Because sparse matrix multiplication is much less efficient than dense matrix multiplication on GPUs, most implementations simulate sparsity by masking weights. In this paper, we leverage recent advances in semi-structured sparse training to apply DST in the domain of classification with large output spaces, where memory-efficiency is paramount. With a label space of possibly millions of candidates, the classification layer alone will consume several gigabytes of memory. Switching from a dense to a fixed fan-in sparse layer updated with sparse evolutionary training (SET); however, severely hampers training convergence, especially at the largest label spaces. We find that poor gradient flow from the sparse classifier to the dense text encoder make it difficult to learn good input representations. By employing an intermediate layer or adding an auxiliary training objective, we recover most of the generalisation performance of the dense model. Overall, we demonstrate the applicability and practical benefits of DST in a challenging domain -- characterized by a highly skewed label distribution that differs substantially from typical DST benchmark datasets -- which enables end-to-end training with millions of labels on commodity hardware.

Mike Lasby, Nish Sinnadurai, Valavan Manohararajah et al.

Speculative decoding is a powerful technique for reducing the latency of Large Language Models (LLMs), offering a fault-tolerant framework that enables the use of highly compressed draft models. In this work, we introduce Self-Distilled Sparse Drafters (SD$^2$), a novel methodology that leverages self-data distillation and fine-grained weight sparsity to produce highly efficient and well-aligned draft models. SD$^2$ systematically enhances draft token acceptance rates while significantly reducing Multiply-Accumulate operations (MACs), even in the Universal Assisted Generation (UAG) setting, where draft and target models originate from different model families. On a Llama-3.1-70B target model, SD$^2$ provides a 1.59$\times$ higher Mean Accepted Length (MAL) compared to layer-pruned draft models and reduces MACs by over 43.87% with a 8.36% reduction in MAL compared to a dense draft models. Our 1.5B and 3B unstructured sparse drafters outperform both dense and layer-pruned models in terms of end-to-end latency improvements; highlighting the potential of sparsity-aware fine-tuning and compression strategies to improve LLM inference efficiency while maintaining alignment with target models.

Mike Lasby, Ivan Lazarevich, Nish Sinnadurai et al.

Sparsely-activated Mixture-of-Experts (SMoE) models offer efficient pre-training and low latency but their large parameter counts create significant memory overhead, motivating research into expert compression. Contrary to recent findings favouring expert merging on discriminative benchmarks, we demonstrate that expert pruning is a superior strategy for generative tasks. We prove that merging introduces an irreducible error by causing a "functional subspace collapse", due to the loss of the router's independent, input-dependent control over experts. Leveraging this insight, we propose Router-weighted Expert Activation Pruning (REAP), a novel pruning criterion that considers both router gate-values and expert activation norms. Across a diverse set of SMoE models ranging from 20B to 1T parameters, REAP consistently outperforms merging and other pruning methods on generative benchmarks, especially at 50% compression. Notably, our method achieves near-lossless compression on code generation and tool-calling tasks with Qwen3-Coder-480B and Kimi-K2, even after pruning 50% of experts.

Mike Lasby, Anna Golubeva, Utku Evci et al.

Dynamic Sparse Training (DST) methods achieve state-of-the-art results in sparse neural network training, matching the generalization of dense models while enabling sparse training and inference. Although the resulting models are highly sparse and theoretically less computationally expensive, achieving speedups with unstructured sparsity on real-world hardware is challenging. In this work, we propose a sparse-to-sparse DST method, Structured RigL (SRigL), to learn a variant of fine-grained structured N:M sparsity by imposing a constant fan-in constraint. Using our empirical analysis of existing DST methods at high sparsity, we additionally employ a neuron ablation method which enables SRigL to achieve state-of-the-art sparse-to-sparse structured DST performance on a variety of Neural Network (NN) architectures. Using a 90% sparse linear layer, we demonstrate a real-world acceleration of 3.4x/2.5x on CPU for online inference and 1.7x/13.0x on GPU for inference with a batch size of 256 when compared to equivalent dense/unstructured (CSR) sparse layers, respectively.

Youssef Beauferris, Jonas Teuwen, Dimitrios Karkalousos et al.

Deep-learning-based brain magnetic resonance imaging (MRI) reconstruction methods have the potential to accelerate the MRI acquisition process. Nevertheless, the scientific community lacks appropriate benchmarks to assess MRI reconstruction quality of high-resolution brain images, and evaluate how these proposed algorithms will behave in the presence of small, but expected data distribution shifts. The Multi-Coil Magnetic Resonance Image (MC-MRI) Reconstruction Challenge provides a benchmark that aims at addressing these issues, using a large dataset of high-resolution, three-dimensional, T1-weighted MRI scans. The challenge has two primary goals: 1) to compare different MRI reconstruction models on this dataset and 2) to assess the generalizability of these models to data acquired with a different number of receiver coils. In this paper, we describe the challenge experimental design, and summarize the results of a set of baseline and state of the art brain MRI reconstruction models. We provide relevant comparative information on the current MRI reconstruction state-of-the-art and highlight the challenges of obtaining generalizable models that are required prior to broader clinical adoption. The MC-MRI benchmark data, evaluation code and current challenge leaderboard are publicly available. They provide an objective performance assessment for future developments in the field of brain MRI reconstruction.