Deep Neural Networks (DNNs) have drawn attention because of their outstanding performance on various tasks. However, deploying full-fledged DNNs in resource-constrained devices (edge, mobile, IoT) is difficult due to their large size. To overcome the issue, various approaches are considered, like offloading part of the computation to the cloud for final inference (split computing) or performing the inference at an intermediary layer without passing through all layers (early exits). In this work, we propose combining both approaches by using early exits in split computing. In our approach, we decide up to what depth of DNNs computation to perform on the device (splitting layer) and whether a sample can exit from this layer or need to be offloaded. The decisions are based on a weighted combination of accuracy, computational, and communication costs. We develop an algorithm named SplitEE to learn an optimal policy. Since pre-trained DNNs are often deployed in new domains where the ground truths may be unavailable and samples arrive in a streaming fashion, SplitEE works in an online and unsupervised setup. We extensively perform experiments on five different datasets. SplitEE achieves a significant cost reduction ($>50\%$) with a slight drop in accuracy ($<2\%$) as compared to the case when all samples are inferred at the final layer. The anonymized source code is available at \url{https://anonymous.4open.science/r/SplitEE_M-B989/README.md}.

Transformer-based pre-trained language models (PLMs) mostly suffer from excessive overhead despite their advanced capacity. For resource-constrained devices, there is an urgent need for a spatially and temporally efficient model which retains the major capacity of PLMs. However, existing statically compressed models are unaware of the diverse complexities between input instances, potentially resulting in redundancy and inadequacy for simple and complex inputs. Also, miniature models with early exiting encounter challenges in the trade-off between making predictions and serving the deeper layers. Motivated by such considerations, we propose a collaborative optimization for PLMs that integrates static model compression and dynamic inference acceleration. Specifically, the PLM is slenderized in width while the depth remains intact, complementing layer-wise early exiting to speed up inference dynamically. To address the trade-off of early exiting, we propose a joint training approach that calibrates slenderization and preserves contributive structures to each exit instead of only the final layer. Experiments are conducted on GLUE benchmark and the results verify the Pareto optimality of our approach at high compression and acceleration rate with 1/8 parameters and 1/19 FLOPs of BERT.

Supersized pre-trained language models have pushed the accuracy of various NLP tasks to a new state-of-the-art (SOTA). Rather than pursuing the reachless SOTA accuracy, most works are pursuing improvement on other dimensions such as efficiency, leading to "Pareto SOTA". Different from accuracy, the metric for efficiency varies across different studies, making them hard to be fairly compared. To that end, this work presents ELUE (Efficient Language Understanding Evaluation), a standard evaluation, and a public leaderboard for efficient NLP models. ELUE is dedicated to depicting the Pareto Front for various language understanding tasks, such that it can tell whether and how much a method achieves Pareto improvement. Along with the benchmark, we also pre-train and release a strong baseline, ElasticBERT, whose elasticity is both static and dynamic. ElasticBERT is static in that it allows reducing model layers on demand. ElasticBERT is dynamic in that it selectively executes parts of model layers conditioned on the input. We demonstrate the ElasticBERT, despite its simplicity, outperforms or performs on par with SOTA compressed and early exiting models. The ELUE benchmark is publicly available at http://eluebenchmark.fastnlp.top/.