Early-exit models accelerate inference by attaching internal classifiers to intermediate layers of the network, allowing computation to halt once a prediction meets a predefined exit criterion. Most early-exit methods rely on confidence-based exit strategies, which has motivated prior work to calibrate intermediate classifiers in pursuit of improved performance-efficiency trade-offs. In this paper, we argue that calibration metrics can be misleading indicators of multi-exit model performance. Specifically, we present empirical evidence showing that miscalibrated networks can outperform calibrated ones. As an alternative, we propose using failure prediction as a more informative proxy for early-exit model performance. Unlike calibration, failure prediction captures changes in sample rankings and correlates strongly with efficiency gains, offering a more reliable framework for designing and evaluating early-exit models.

As the development of lightweight deep learning algorithms, various deep neural network (DNN) models have been proposed for the remote sensing scene classification (RSSC) application. However, it is still challenging for these RSSC models to achieve optimal performance among model accuracy, inference latency, and energy consumption on resource-constrained edge devices. In this paper, we propose a lightweight RSSC framework, which includes a distilled global filter network (GFNet) model and an early-exit mechanism designed for edge devices to achieve state-of-the-art performance. Specifically, we first apply frequency domain distillation on the GFNet model to reduce model size. Then we design a dynamic early-exit model tailored for DNN models on edge devices to further improve model inference efficiency. We evaluate our E3C model on three edge devices across four datasets. Extensive experimental results show that it achieves an average of 1.3x speedup on model inference and over 40% improvement on energy efficiency, while maintaining high classification accuracy.

Building efficient inference framework has gained increasing interests for research community. Early-exit models, a variant of LLMs, improves the inference efficiency of LLMs by skipping rest layers and directly generate output tokens when they are confident enough. However, there is no work of LLM inference framework that takes early-exit models into consideration. This is non-trivial as prior art on LLM inference cannot be directly applied to early-exit models. In this work, we solves two key challenges in building efficient inference framework for early-exit models: (1) batch inference at iteration-level granularity; and (2) KV cache management. For the former, we propose to process the batch until all sequences surpass the early-exit confidence threshold. For the latter, we propose to fill the KV cache of rest layers before the iteration terminates. Our evaluation shows that, compared with the original vLLM operating at full layers, our solution achieves up to 1.25x speed up.

Early exits are an important efficiency mechanism integrated into deep neural networks that allows for the termination of the network's forward pass before processing through all its layers. By allowing early halting of the inference process for less complex inputs that reached high confidence, early exits significantly reduce the amount of computation required. Early exit methods add trainable internal classifiers which leads to more intricacy in the training process. However, there is no consistent verification of the approaches of training of early exit methods, and no unified scheme of training such models. Most early exit methods employ a training strategy that either simultaneously trains the backbone network and the exit heads or trains the exit heads separately. We propose a training approach where the backbone is initially trained on its own, followed by a phase where both the backbone and the exit heads are trained together. Thus, we advocate for organizing early-exit training strategies into three distinct categories, and then validate them for their performance and efficiency. In this benchmark, we perform both theoretical and empirical analysis of early-exit training regimes. We study the methods in terms of information flow, loss landscape and numerical rank of activations and gauge the suitability of regimes for various architectures and datasets.

Adaptive inference is a simple method for reducing inference costs. The method works by maintaining multiple classifiers of different capacities, and allocating resources to each test instance according to its difficulty. In this work, we compare the two main approaches for adaptive inference, Early-Exit and Multi-Model, when training data is limited. First, we observe that for models with the same architecture and size, individual Multi-Model classifiers outperform their Early-Exit counterparts by an average of 2.3%. We show that this gap is caused by Early-Exit classifiers sharing model parameters during training, resulting in conflicting gradient updates of model weights. We find that despite this gap, Early-Exit still provides a better speed-accuracy trade-off due to the overhead of the Multi-Model approach. To address these issues, we propose SWEET (Separating Weights in Early Exit Transformers), an Early-Exit fine-tuning method that assigns each classifier its own set of unique model weights, not updated by other classifiers. We compare SWEET's speed-accuracy curve to standard Early-Exit and Multi-Model baselines and find that it outperforms both methods at fast speeds while maintaining comparable scores to Early-Exit at slow speeds. Moreover, SWEET individual classifiers outperform Early-Exit ones by 1.1% on average. SWEET enjoys the benefits of both methods, paving the way for further reduction of inference costs in NLP.