Keypoint-Guided Optimal Transport

Existing Optimal Transport (OT) methods mainly derive the optimal transport plan/matching under the criterion of transport cost/distance minimization, which may cause incorrect matching in some cases. In many applications, annotating a few matched keypoints across domains is reasonable or even effortless in annotation burden. It is valuable to investigate how to leverage the annotated keypoints to guide the correct matching in OT. In this paper, we propose a novel KeyPoint-Guided model by ReLation preservation (KPG-RL) that searches for the optimal matching (i.e., transport plan) guided by the keypoints in OT. To impose the keypoints in OT, first, we propose a mask-based constraint of the transport plan that preserves the matching of keypoint pairs. Second, we propose to preserve the relation of each data point to the keypoints to guide the matching. The proposed KPG-RL model can be solved by Sinkhorn's algorithm and is applicable even when distributions are supported in different spaces. We further utilize the relation preservation constraint in the Kantorovich Problem and Gromov-Wasserstein model to impose the guidance of keypoints in them. Meanwhile, the proposed KPG-RL model is extended to the partial OT setting. Moreover, we deduce the dual formulation of the KPG-RL model, which is solved using deep learning techniques. Based on the learned transport plan from dual KPG-RL, we propose a novel manifold barycentric projection to transport source data to the target domain. As applications, we apply the proposed KPG-RL model to the heterogeneous domain adaptation and image-to-image translation. Experiments verified the effectiveness of the proposed approach.

Graph Enhanced BERT for Query Understanding

Query understanding plays a key role in exploring users' search intents and facilitating users to locate their most desired information. However, it is inherently challenging since it needs to capture semantic information from short and ambiguous queries and often requires massive task-specific labeled data. In recent years, pre-trained language models (PLMs) have advanced various natural language processing tasks because they can extract general semantic information from large-scale corpora. Therefore, there are unprecedented opportunities to adopt PLMs for query understanding. However, there is a gap between the goal of query understanding and existing pre-training strategies -- the goal of query understanding is to boost search performance while existing strategies rarely consider this goal. Thus, directly applying them to query understanding is sub-optimal. On the other hand, search logs contain user clicks between queries and urls that provide rich users' search behavioral information on queries beyond their content. Therefore, in this paper, we aim to fill this gap by exploring search logs. In particular, to incorporate search logs into pre-training, we first construct a query graph where nodes are queries and two queries are connected if they lead to clicks on the same urls. Then we propose a novel graph-enhanced pre-training framework, GE-BERT, which can leverage both query content and the query graph. In other words, GE-BERT can capture both the semantic information and the users' search behavioral information of queries. Extensive experiments on various query understanding tasks have demonstrated the effectiveness of the proposed framework.

ERNIE 3.0 Titan: Exploring Larger-scale Knowledge Enhanced Pre-training for Language Understanding and Generation

Pre-trained language models have achieved state-of-the-art results in various Natural Language Processing (NLP) tasks. GPT-3 has shown that scaling up pre-trained language models can further exploit their enormous potential. A unified framework named ERNIE 3.0 was recently proposed for pre-training large-scale knowledge enhanced models and trained a model with 10 billion parameters. ERNIE 3.0 outperformed the state-of-the-art models on various NLP tasks. In order to explore the performance of scaling up ERNIE 3.0, we train a hundred-billion-parameter model called ERNIE 3.0 Titan with up to 260 billion parameters on the PaddlePaddle platform. Furthermore, we design a self-supervised adversarial loss and a controllable language modeling loss to make ERNIE 3.0 Titan generate credible and controllable texts. To reduce the computation overhead and carbon emission, we propose an online distillation framework for ERNIE 3.0 Titan, where the teacher model will teach students and train itself simultaneously. ERNIE 3.0 Titan is the largest Chinese dense pre-trained model so far. Empirical results show that the ERNIE 3.0 Titan outperforms the state-of-the-art models on 68 NLP datasets.

End-to-end Adaptive Distributed Training on PaddlePaddle

Distributed training has become a pervasive and effective approach for training a large neural network (NN) model with processing massive data. However, it is very challenging to satisfy requirements from various NN models, diverse computing resources, and their dynamic changes during a training job. In this study, we design our distributed training framework in a systematic end-to-end view to provide the built-in adaptive ability for different scenarios, especially for industrial applications and production environments, by fully considering resource allocation, model partition, task placement, and distributed execution. Based on the unified distributed graph and the unified cluster object, our adaptive framework is equipped with a global cost model and a global planner, which can enable arbitrary parallelism, resource-aware placement, multi-mode execution, fault-tolerant, and elastic distributed training. The experiments demonstrate that our framework can satisfy various requirements from the diversity of applications and the heterogeneity of resources with highly competitive performance. The ERNIE language model with 260 billion parameters is efficiently trained on thousands of AI processors with 91.7% weak scalability. The throughput of the model from the recommender system by employing the heterogeneous pipeline asynchronous execution can be increased up to 2.1 times and 3.3 times that of the GPU-only and CPU-only training respectively. Moreover, the fault-tolerant and elastic distributed training have been successfully applied to the online industrial applications, which give a reduction of 34.49% in the number of failed long-term training jobs and an increase of 33.91% for the global scheduling efficiency in the production environment.

Prioritized Subnet Sampling for Resource-Adaptive Supernet Training

A resource-adaptive supernet adjusts its subnets for inference to fit the dynamically available resources. In this paper, we propose Prioritized Subnet Sampling to train a resource-adaptive supernet, termed PSS-Net. We maintain multiple subnet pools, each of which stores the information of substantial subnets with similar resource consumption. Considering a resource constraint, subnets conditioned on this resource constraint are sampled from a pre-defined subnet structure space and high-quality ones will be inserted into the corresponding subnet pool. Then, the sampling will gradually be prone to sampling subnets from the subnet pools. Moreover, the one with a better performance metric is assigned with higher priority to train our PSS-Net, if sampling is from a subnet pool. At the end of training, our PSS-Net retains the best subnet in each pool to entitle a fast switch of high-quality subnets for inference when the available resources vary. Experiments on ImageNet using MobileNetV1/V2 show that our PSS-Net can well outperform state-of-the-art resource-adaptive supernets. Our project is at https://github.com/chenbong/PSS-Net.

DPT: Deformable Patch-based Transformer for Visual Recognition

Transformer has achieved great success in computer vision, while how to split patches in an image remains a problem. Existing methods usually use a fixed-size patch embedding which might destroy the semantics of objects. To address this problem, we propose a new Deformable Patch (DePatch) module which learns to adaptively split the images into patches with different positions and scales in a data-driven way rather than using predefined fixed patches. In this way, our method can well preserve the semantics in patches. The DePatch module can work as a plug-and-play module, which can easily be incorporated into different transformers to achieve an end-to-end training. We term this DePatch-embedded transformer as Deformable Patch-based Transformer (DPT) and conduct extensive evaluations of DPT on image classification and object detection. Results show DPT can achieve 81.9% top-1 accuracy on ImageNet classification, and 43.7% box mAP with RetinaNet, 44.3% with Mask R-CNN on MSCOCO object detection. Code has been made available at: https://github.com/CASIA-IVA-Lab/DPT .

Training Compact CNNs for Image Classification using Dynamic-coded Filter Fusion

The mainstream approach for filter pruning is usually either to force a hard-coded importance estimation upon a computation-heavy pretrained model to select "important" filters, or to impose a hyperparameter-sensitive sparse constraint on the loss objective to regularize the network training. In this paper, we present a novel filter pruning method, dubbed dynamic-coded filter fusion (DCFF), to derive compact CNNs in a computation-economical and regularization-free manner for efficient image classification. Each filter in our DCFF is firstly given an inter-similarity distribution with a temperature parameter as a filter proxy, on top of which, a fresh Kullback-Leibler divergence based dynamic-coded criterion is proposed to evaluate the filter importance. In contrast to simply keeping high-score filters in other methods, we propose the concept of filter fusion, i.e., the weighted averages using the assigned proxies, as our preserved filters. We obtain a one-hot inter-similarity distribution as the temperature parameter approaches infinity. Thus, the relative importance of each filter can vary along with the training of the compact CNN, leading to dynamically changeable fused filters without both the dependency on the pretrained model and the introduction of sparse constraints. Extensive experiments on classification benchmarks demonstrate the superiority of our DCFF over the compared counterparts. For example, our DCFF derives a compact VGGNet-16 with only 72.77M FLOPs and 1.06M parameters while reaching top-1 accuracy of 93.47% on CIFAR-10. A compact ResNet-50 is obtained with 63.8% FLOPs and 58.6% parameter reductions, retaining 75.60% top-1 accuracy on ILSVRC-2012. Our code, narrower models and training logs are available at https://github.com/lmbxmu/DCFF.

FloorLevel-Net: Recognizing Floor-Level Lines with Height-Attention-Guided Multi-task Learning

The ability to recognize the position and order of the floor-level lines that divide adjacent building floors can benefit many applications, for example, urban augmented reality (AR). This work tackles the problem of locating floor-level lines in street-view images, using a supervised deep learning approach. Unfortunately, very little data is available for training such a network $-$ current street-view datasets contain either semantic annotations that lack geometric attributes, or rectified facades without perspective priors. To address this issue, we first compile a new dataset and develop a new data augmentation scheme to synthesize training samples by harassing (i) the rich semantics of existing rectified facades and (ii) perspective priors of buildings in diverse street views. Next, we design FloorLevel-Net, a multi-task learning network that associates explicit features of building facades and implicit floor-level lines, along with a height-attention mechanism to help enforce a vertical ordering of floor-level lines. The generated segmentations are then passed to a second-stage geometry post-processing to exploit self-constrained geometric priors for plausible and consistent reconstruction of floor-level lines. Quantitative and qualitative evaluations conducted on assorted facades in existing datasets and street views from Google demonstrate the effectiveness of our approach. Also, we present context-aware image overlay results and show the potentials of our approach in enriching AR-related applications.

PanGu-$α$: Large-scale Autoregressive Pretrained Chinese Language Models with Auto-parallel Computation

Large-scale Pretrained Language Models (PLMs) have become the new paradigm for Natural Language Processing (NLP). PLMs with hundreds of billions parameters such as GPT-3 have demonstrated strong performances on natural language understanding and generation with \textit{few-shot in-context} learning. In this work, we present our practice on training large-scale autoregressive language models named PanGu-$\alpha$, with up to 200 billion parameters. PanGu-$\alpha$ is developed under the MindSpore and trained on a cluster of 2048 Ascend 910 AI processors. The training parallelism strategy is implemented based on MindSpore Auto-parallel, which composes five parallelism dimensions to scale the training task to 2048 processors efficiently, including data parallelism, op-level model parallelism, pipeline model parallelism, optimizer model parallelism and rematerialization. To enhance the generalization ability of PanGu-$\alpha$, we collect 1.1TB high-quality Chinese data from a wide range of domains to pretrain the model. We empirically test the generation ability of PanGu-$\alpha$ in various scenarios including text summarization, question answering, dialogue generation, etc. Moreover, we investigate the effect of model scales on the few-shot performances across a broad range of Chinese NLP tasks. The experimental results demonstrate the superior capabilities of PanGu-$\alpha$ in performing various tasks under few-shot or zero-shot settings.

Contrastive Neural Architecture Search with Neural Architecture Comparators

One of the key steps in Neural Architecture Search (NAS) is to estimate the performance of candidate architectures. Existing methods either directly use the validation performance or learn a predictor to estimate the performance. However, these methods can be either computationally expensive or very inaccurate, which may severely affect the search efficiency and performance. Moreover, as it is very difficult to annotate architectures with accurate performance on specific tasks, learning a promising performance predictor is often non-trivial due to the lack of labeled data. In this paper, we argue that it may not be necessary to estimate the absolute performance for NAS. On the contrary, we may need only to understand whether an architecture is better than a baseline one. However, how to exploit this comparison information as the reward and how to well use the limited labeled data remains two great challenges. In this paper, we propose a novel Contrastive Neural Architecture Search (CTNAS) method which performs architecture search by taking the comparison results between architectures as the reward. Specifically, we design and learn a Neural Architecture Comparator (NAC) to compute the probability of candidate architectures being better than a baseline one. Moreover, we present a baseline updating scheme to improve the baseline iteratively in a curriculum learning manner. More critically, we theoretically show that learning NAC is equivalent to optimizing the ranking over architectures. Extensive experiments in three search spaces demonstrate the superiority of our CTNAS over existing methods.