Feiyang Kang, Newsha Ardalani, Michael Kuchnik, Youssef Emad, Mostafa Elhoushi, Shubhabrata Sengupta, Shang-Wen Li, Ramya Raghavendra, Ruoxi Jia, Carole-Jean Wu. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

We present any4, a learned 4-bit weight quantization solution for large language models (LLMs) providing arbitrary numeric representations without requiring pre-processing of weights or activations. any4 yields higher accuracy compared to other related 4-bit numeric representation types: int4, fp4 and nf4, as evaluated on a range of model sizes, generations and families (Llama 2, Llama 3, Mistral and Mixtral). While any4 does not require preprocessing of weights or activations, it is also competitive with orthogonal techniques that require such preprocessing (e.g., AWQ and GPTQ). We also experiment with any3 and any2 and show competitiveness at lower bits. Additionally, we show that we can calibrate using a single curated diverse sample rather than hundreds of samples from a dataset as done in most quantization approaches. We also open source tinygemm, a latency optimized GPU matrix multiplication library for LLMs, that implements any4 using a GPU-efficient lookup table strategy along with other common quantization methods. We open source our code at https://github.com/facebookresearch/any4.

Machine learning solutions are rapidly adopted to enable a variety of key use cases, from conversational AI assistants to scientific discovery. As the adoption of machine learning models becomes increasingly prevalent, the associated lifecycle carbon footprint is expected to increase, including both *operational carbon* from training and inference and *embodied carbon* from AI hardware manufacturing. We introduce CATransformers, the first carbon-aware co-optimization framework for Transformer-based models and hardware accelerators. By integrating both operational and embodied carbon into early-stage design space exploration, CATransformers enables sustainability-driven model architecture and hardware accelerator co-design that reveals fundamentally different trade-offs than latency- or energy-centric approaches. Evaluated across a range of Transformer models, CATransformers consistently demonstrates the potential to reduce total carbon emissions --by up to 30\% -- while maintaining accuracy and latency. We further highlight its extensibility through a focused case study on multi-modal models. Our results emphasize the need for holistic optimization methods that prioritize carbon efficiency without compromising model capability and execution time performance. The source code of CATransformers is available at https://github.com/facebookresearch/CATransformers.

Mostafa Elhoushi, Akshat Shrivastava, Diana Liskovich, Basil Hosmer, Bram Wasti, Liangzhen Lai, Anas Mahmoud, Bilge Acun, Saurabh Agarwal, Ahmed Roman, Ahmed Aly, Beidi Chen, Carole-Jean Wu. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Vision-Language Models (VLMs) are pretrained on large, diverse, and noisy web-crawled datasets. This underscores the critical need for dataset pruning, as the quality of these datasets is strongly correlated with the performance of VLMs on downstream tasks. Using CLIPScore from a pretrained model to only train models using highly-aligned samples is one of the most successful methods for pruning. We argue that this approach suffers from multiple limitations including: false positives and negatives due to CLIP's pretraining on noisy labels. We propose a pruning signal, Sieve, that employs synthetic captions generated by image-captioning models pretrained on small, diverse, and well-aligned image-text pairs to evaluate the alignment of noisy image-text pairs. To bridge the gap between the limited diversity of generated captions and the high diversity of alternative text (alt-text), we estimate the semantic textual similarity in the embedding space of a language model pretrained on unlabeled text corpus. Using DataComp, a multimodal dataset filtering benchmark, when evaluating on 38 downstream tasks, our pruning approach, surpasses CLIPScore by 2.6% and 1.7% on medium and large scale respectively. In addition, on retrieval tasks, Sieve leads to a significant improvement of 2.7% and 4.5% on medium and large scale respectively.

Sparse Convolution (SC) is widely used for processing 3D point clouds that are inherently sparse. Different from dense convolution, SC preserves the sparsity of the input point cloud by only allowing outputs to specific locations. To efficiently compute SC, prior SC engines first use hash tables to build a kernel map that stores the necessary General Matrix Multiplication (GEMM) operations to be executed (Map step), and then use a Gather-GEMM-Scatter process to execute these GEMM operations (GMaS step). In this work, we analyze the shortcomings of prior state-of-the-art SC engines, and propose Minuet, a novel memory-efficient SC engine tailored for modern GPUs. Minuet proposes to (i) replace the hash tables used in the Map step with a novel segmented sorting double-traversed binary search algorithm that highly utilizes the on-chip memory hierarchy of GPUs, (ii) use a lightweight scheme to autotune the tile size in the Gather and Scatter operations of the GMaS step, such that to adapt the execution to the particular characteristics of each SC layer, dataset, and GPU architecture, and (iii) employ a padding-efficient GEMM grouping approach that reduces both memory padding and kernel launching overheads. Our evaluations show that Minuet significantly outperforms prior SC engines by on average 1.74× (up to 2.22×) for end-to-end point cloud network executions. Our novel segmented sorting double-traversed binary search algorithm achieves superior speedups by 15.8× on average (up to 26.8×) over prior SC engines in the Map step. The source code of Minuet is publicly available at https://github.com/UofT-EcoSystem/Minuet.

Large Language Models (LLMs) with hundreds of billions of parameters have transformed the field of machine learning. However, serving these models at inference time is both compute and memory intensive, where a single request can require multiple GPUs and tens of Gigabytes of memory. Multi-head attention is one of the key components of LLMs, which can for over 50% of LLMs memory and compute requirement. We observe that there is a high amount of redundancy across heads on which tokens they pay attention to. Based on this insight, we propose Clustered HeadAttention ( CHAI ). CHAI combines heads with a high amount of correlation for self-attention at runtime, thus reducing both memory and compute. In our experiments, we show that CHAI is able to reduce the memory requirements for storing K,V cache by up to 21.4% and inference time latency by up to 1.73× without any fine-tuning required. CHAI achieves this with a maximum 3.2% deviation in accuracy across 3 different models (i.e. OPT-66B, LLAMA-7B, LLAMA-33B) and 5 different evaluation datasets.

Large language models (LLMs) have made significant advancements in code-related tasks, yet many LLMs treat code as simple sequences, neglecting its structured nature. We introduce AST-T5, a novel pretraining paradigm that leverages the Abstract Syntax Tree (AST) for enhanced code generation, transpilation, and understanding. Using dynamic programming, our AST-Aware Segmentation retains code structure, while our AST-Aware Span Corruption objective equips the model to reconstruct various code structures. Unlike other models, AST-T5 avoids complex program analyses or architectural changes, so it integrates seamlessly with any encoder-decoder Transformer. Evaluations show that AST-T5 consistently outperforms similar-sized LMs across various code-related tasks including HumanEval and MBPP. Structure-awareness makes AST-T5 particularly powerful in code-to-code tasks, surpassing CodeT5 by 2 points in exact match score for the Bugs2Fix task and by 3 points in exact match score for Java-C# Transpilation in CodeXGLUE. Our code and model are publicly available at https://github.com/gonglinyuan/ast_t5.

We introduce **S**yntax-**A**ware **F**ill-**i**n-the-**M**iddle (SAFIM), a new benchmark for evaluating Large Language Models (LLMs) on the code Fill-in-the-Middle (FIM) task. This benchmark focuses on syntax-aware completions of program structures such as code blocks and conditional expressions, and includes 17,720 examples from multiple programming languages, sourced from recent code submissions after April 2022 to minimize data contamination. SAFIM provides a robust framework with various prompt designs and novel syntax-aware post-processing techniques, facilitating accurate and fair comparisons across LLMs. Our comprehensive evaluation of 15 LLMs shows that FIM pretraining not only enhances FIM proficiency but also improves Left-to-Right (L2R) inference using LLMs. Our findings challenge conventional beliefs and suggest that pretraining methods and data quality have more impact than model size. SAFIM thus serves as a foundational platform for future research in effective pretraining strategies for code LLMs. The evaluation toolkit and dataset are available at https://github.com/gonglinyuan/safim, and the leaderboard is available at https://safimbenchmark.com.

Finding the optimal pass sequence of compilation can lead to a significant reduction in program size. Prior works on compilation pass ordering have two major drawbacks. They either require an excessive budget (in terms of the number of compilation passes) at compile time or fail to generalize to unseen programs. In this work, instead of predicting passes sequentially, we directly learn a policy on the pass sequence space, which outperforms the default -Oz flag by an average of 4.5% over a large collection (4683) of unseen code repositories from diverse domains across 14 datasets. To achieve this, we first identify a small set (termed coreset) of pass sequences that generally optimize the size of most programs. Then, a policy is learned to pick the optimal sequences by predicting the normalized values of the pass sequences in the coreset. Our results demonstrate that existing human-designed compiler passes can be improved with a simple yet effective technique that leverages pass sequence space which contains dense rewards, while approaches operating on the individual pass space may suffer from issues of sparse reward, and do not generalize well to held-out programs from different domains. Website: https://rlcompopt.github.io.

The size of deep neural networks has grown exponentially in recent years. Unfortunately, hardware devices have not kept pace with the rapidly increasing memory requirements. To cope with this, researchers have proposed various techniques including spilling, rematerialization, reduced precision training, model pruning, and so on. However, these approaches suffer from various limitations, such as increasing training time, affecting model accuracy, or requiring extensive manual modifications to the neural networks. We present MODeL, an algorithm that optimizes the lifetime and memory location of the tensors used to train neural networks. Our method automatically reduces the memory usage of existing neural networks without any of the drawbacks of other techniques. We formulate the problem as a joint integer linear program (ILP). We present several techniques to simplify the encoding of the problem, and enable our approach to scale to the size of state-of-the-art neural networks using an off-the-shelf ILP solver. We experimentally demonstrate that MODeL only takes seconds to allow the training of neural networks using 30% less memory on average.

This paper presents MLGOPerf; the first end-to-end framework capable of optimizing performance using LLVM’s ML-Inliner. It employs a secondary ML model to generate rewards used for training a retargeted Reinforcement learning agent, previously used as the primary model by MLGO. It does so by predicting the post-inlining speedup of a function under analysis and it enables a fast training framework for the primary model which otherwise wouldn’t be practical. The experimental results show MLGOPerf is able to gain up to 1.8% with respect to LLVM’s optimization at O3 when trained for performance on SPEC CPU2006. Furthermore, the proposed approach provides up to 26% increased opportunities to autotune code regions for our benchmarks which can be translated into an additional 3.7% speedup value.

Dynamic model pruning is a recent direction that allows for the inference of a different sub-network for each input sample during deployment. However, current dynamic methods rely on learning a continuous channel gating through regularization by inducing sparsity loss. This formulation introduces complexity in balancing different losses (e.g task loss, regularization loss). In addition, regularization based methods lack transparent tradeoff hyper- parameter selection to realize a computational budget. Our contribution is two-fold: 1) decoupled task and pruning losses. 2) Simple hyperparameter selection that enables FLOPs reduction estimation before training. Inspired by the Hebbian theory in Neuroscience: “neurons that fire together wire together”, we propose to predict a mask to process k filters in a layer based on the activation of its previous layer. We pose the problem as a self-supervised binary classification problem. Each mask predictor module is trained to predict if the log-likelihood for each filter in the current layer belongs to the top-k activated filters. The value k is dynamically estimated for each input based on a novel criterion using the mass of heatmaps. We show experiments on several neural architectures, such as VGG, ResNet and MobileNet on CIFAR and ImageNet datasets. On CIFAR, we reach similar accuracy to SOTA methods with 15% and 24% higher FLOPs reduction. Similarly in ImageNet, we achieve lower drop in accuracy with up to 13% improvement in FLOPs reduction.

Recent advances in neural networks pruning have made it possible to remove a large number of filters without any perceptible drop in accuracy. However, the gain in speed depends on the number of filters per layer. In this paper, we propose a one-shot layer-wise proxy classifier to estimate layer importance that in turn allows us to prune a whole layer. In contrast to existing filter pruning methods which attempt to reduce the layer width of a dense model, our method reduces its depth and can thus guarantee inference speed up. In our proposed method, we first go through the training data once to construct proxy classifiers for each layer using imprinting. Next, we prune layers with smallest accuracy difference from their preceding layer till a latency budget is achieved. Finally, we fine-tune the newly pruned model to improve accuracy. Experimental results showed 43.70% latency reduction with 1.27% accuracy increase on CIFAR100 for the pruned VGG19. Further, we achieved 16% and 25% latency reduction with 0.58% increase and 0.01% decrease in accuracy respectively on ImageNet for ResNet-50. The major advantage of our proposed method is that these latency reductions cannot be achieved with existing filter pruning methods as they are bounded by the original model's depth. Code is available at https://github.com/selkerdawy/one-shot-layer-pruning.

Recognition of the mode of motion or mode of transit of the user or platform carrying a device is needed in portable navigation, as well as other technological domains. An extensive survey on motion mode recognition approaches is provided in this survey paper. The survey compares and describes motion mode recognition approaches from different viewpoints: usability and convenience, types of devices in terms of setup mounting and data acquisition, various types of sensors used, signal processing methods employed, features extracted, and classification techniques. This paper ends with a quantitative comparison of the performance of motion mode recognition modules developed by researchers in different domains.