Causal language models have demonstrated remarkable capabilities, but their size poses significant challenges for deployment in resource-constrained environments. Knowledge distillation, a widely-used technique for transferring knowledge from a large teacher model to a small student model, presents a promising approach for model compression. A significant remaining issue lies in the major differences between teacher and student models, namely the substantial capacity gap, mode averaging, and mode collapse, which pose barriers during distillation.s To address these issues, we introduce Temporally Adaptive Interpolated Distillation (TAID), a novel knowledge distillation approach that dynamically interpolates student and teacher distributions through an adaptive intermediate distribution, gradually shifting from the student's initial distribution towards the teacher's distribution. We provide a theoretical analysis demonstrating TAID's ability to prevent mode collapse and empirically show its effectiveness in addressing the capacity gap while balancing mode averaging and mode collapse. Our comprehensive experiments demonstrate TAID's superior performance across various model sizes and architectures in both instruction tuning and pre-training scenarios. These results demonstrate TAID's effectiveness in creating high-performing and efficient models, advancing the development of more accessible AI technologies. Large language models are too large. Causal language models (LMs) are increasingly becoming essential tools across various sectors (Malinka et al., 2023; Wu et al., 2023; Zhang et al., 2023a; He et al., 2024). Scaling data size, model size, and training steps has been the primary approach to improve LM performance (Kaplan et al., 2020; Hoffmann et al., 2022; OpenAI et al., 2024), leading to rapid advancements in both proprietary and open-source LMs (Touvron et al., 2023; Abdin et al., 2024; Yang et al., 2024). This paradox of scale hinders the widespread deployment and use of LMs despite their potential and high demand. Knowledge distillation offers a promising prescription. One promising approach to developing compact yet high-performing models is knowledge distillation (KD) (Hinton et al., 2015).

Vision Language Models (VLMs) excel in zero-shot image classification by pairing images with textual category names. The expanding variety of Pre-Trained VLMs enhances the likelihood of identifying a suitable VLM for specific tasks. Thus, a promising zero-shot image classification strategy is selecting the most appropriate Pre-Trained VLM from the VLM Zoo, relying solely on the text data of the target dataset without access to the dataset's images. In this paper, we analyze two inherent challenges in assessing the ability of a VLM in this Language-Only VLM selection: the "Modality Gap" -- the disparity in VLM's embeddings across two different modalities, making text a less reliable substitute for images; and the "Capability Gap" -- the discrepancy between the VLM's overall ranking and its ranking for target dataset, hindering direct prediction of a model's dataset-specific performance from its general performance. We propose VLM Selection With gAp Bridging (SWAB) to mitigate the negative impact of these two gaps. SWAB first adopts optimal transport to capture the relevance between open-source datasets and target dataset with a transportation matrix. It then uses this matrix to transfer useful statistics of VLMs from open-source datasets to the target dataset for bridging those two gaps and enhancing the VLM's capacity estimation for VLM selection. Experiments across various VLMs and image classification datasets validate SWAB's effectiveness.

The considerable size of Large Language Models (LLMs) presents notable deployment challenges, particularly on resource-constrained hardware. Structured pruning, offers an effective means to compress LLMs, thereby reducing storage costs and enhancing inference speed for more efficient utilization. In this work, we study data-efficient and resource-efficient structure pruning methods to obtain smaller yet still powerful models. Knowledge Distillation is well-suited for pruning, as the intact model can serve as an excellent teacher for pruned students. However, it becomes challenging in the context of LLMs due to memory constraints. To address this, we propose an efficient progressive Numerous-teacher pruning method (NutePrune). NutePrune mitigates excessive memory costs by loading only one intact model and integrating it with various masks and LoRA modules, enabling it to seamlessly switch between teacher and student roles. This approach allows us to leverage numerous teachers with varying capacities to progressively guide the pruned model, enhancing overall performance. Extensive experiments across various tasks demonstrate the effectiveness of NutePrune. In LLaMA-7B zero-shot experiments, NutePrune retains 97.17% of the performance of the original model at 20% sparsity and 95.07% at 25% sparsity.

In this paper we revisit the efficacy of knowledge distillation as a function matching and metric learning problem. In doing so we verify three important design decisions, namely the normalisation, soft maximum function, and projection layers as key ingredients. We theoretically show that the projector implicitly encodes information on past examples, enabling relational gradients for the student. We then show that the normalisation of representations is tightly coupled with the training dynamics of this projector, which can have a large impact on the students performance. Finally, we show that a simple soft maximum function can be used to address any significant capacity gap problems. Experimental results on various benchmark datasets demonstrate that using these insights can lead to superior or comparable performance to state-of-the-art knowledge distillation techniques, despite being much more computationally efficient. In particular, we obtain these results across image classification (CIFAR100 and ImageNet), object detection (COCO2017), and on more difficult distillation objectives, such as training data efficient transformers, whereby we attain a 77.2% top-1 accuracy with DeiT-Ti on ImageNet. Code and models are publicly available.

Language model (LM) distillation is a trending area that aims to distil the knowledge resided in a large teacher LM to a small student one. While various methods have been proposed to push the distillation to its limits, it is still a pain distilling LMs when a large capacity gap is exhibited between the teacher and the student LMs. The pain is mainly resulted by the curse of capacity gap, which describes that a larger teacher LM cannot always lead to a better student LM than one distilled from a smaller teacher LM due to the affect of capacity gap increment. That is, there is likely an optimal point yielding the best student LM along the scaling course of the teacher LM. Even worse, the curse of capacity gap can be only partly yet not fully lifted as indicated in previous studies. However, the tale is not ever one-sided. Although a larger teacher LM has better performance than a smaller teacher LM, it is much more resource-demanding especially in the context of recent large LMs (LLMs). Consequently, instead of sticking to lifting the curse, leaving the curse as is should be arguably fine. Even better, in this paper, we reveal that the optimal capacity gap is almost consistent across different student scales and architectures, fortunately turning the curse into the law of capacity gap. The law later guides us to distil a 3B student LM (termed MiniMA) from a 7B teacher LM (adapted LLaMA2-7B). MiniMA is demonstrated to yield a new compute-performance pareto frontier among existing 3B LMs on commonly used benchmarks, and its instruction-tuned version (termed MiniChat) outperforms a wide range of 3B competitors in GPT4 evaluation and could even compete with several 7B chat models.

Resource-constrained perception systems such as edge computing and vision-for-robotics require vision models to be both accurate and lightweight in computation and memory usage. While knowledge distillation is a proven strategy to enhance the performance of lightweight classification models, its application to structured outputs like object detection and instance segmentation remains a complicated task, due to the variability in outputs and complex internal network modules involved in the distillation process. In this paper, we propose a simple yet surprisingly effective sequential approach to knowledge distillation that progressively transfers the knowledge of a set of teacher detectors to a given lightweight student. To distill knowledge from a highly accurate but complex teacher model, we construct a sequence of teachers to help the student gradually adapt. Our progressive strategy can be easily combined with existing detection distillation mechanisms to consistently maximize student performance in various settings. To the best of our knowledge, we are the first to successfully distill knowledge from Transformer-based teacher detectors to convolution-based students, and unprecedentedly boost the performance of ResNet-50 based RetinaNet from 36.5% to 42.0% AP and Mask R-CNN from 38.2% to 42.5% AP on the MS COCO benchmark.

Pretrained language models (LMs) have shown compelling performance on various downstream tasks, but unfortunately they require a tremendous amount of inference compute. Knowledge distillation finds a path to compress LMs to small ones with a teacher-student paradigm. However, when the capacity gap between the teacher and the student is large, a curse of capacity gap appears, invoking a deficiency in distilling LMs. While a few studies have been carried out to fill the gap, the curse is not yet well tackled. In this paper, we aim at lifting the curse of capacity gap via enlarging the capacity of the student without notably increasing the inference compute. Largely motivated by sparse activation regime of mixture of experts (MoE), we propose a mixture of minimal experts (MiniMoE), which imposes extra parameters to the student but introduces almost no additional inference compute. Experimental results on GLUE and CoNLL demonstrate the curse of capacity gap is lifted by the magic of MiniMoE to a large extent. MiniMoE also achieves the state-of-the-art performance at small FLOPs compared with a range of competitive baselines. With a compression rate as much as $\sim$50$\times$, MiniMoE preserves $\sim$95\% GLUE score of the teacher.

The crux of knowledge distillation -- as a transfer-learning approach -- is to effectively train a resource-limited student model with the guide of a pre-trained larger teacher model. However, when there is a large difference between the model complexities of teacher and student (i.e., capacity gap), knowledge distillation loses its strength in transferring knowledge from the teacher to the student, thus training a weaker student. To mitigate the impact of the capacity gap, we introduce knowledge distillation via intermediate heads. By extending the intermediate layers of the teacher (at various depths) with classifier heads, we cheaply acquire a cohort of heterogeneous pre-trained teachers. The intermediate classifier heads can all together be efficiently learned while freezing the backbone of the pre-trained teacher. The cohort of teachers (including the original teacher) co-teach the student simultaneously. Our experiments on various teacher-student pairs and datasets have demonstrated that the proposed approach outperforms the canonical knowledge distillation approach and its extensions.