The BigCode project is an open-scientific collaboration working on the responsible development of large language models for code. This tech report describes the progress of the collaboration until December 2022, outlining the current state of the Personally Identifiable Information (PII) redaction pipeline, the experiments conducted to de-risk the model architecture, and the experiments investigating better preprocessing methods for the training data. We train 1.1B parameter models on the Java, JavaScript, and Python subsets of The Stack and evaluate them on the MultiPL-E text-to-code benchmark. We find that more aggressive filtering of near-duplicates can further boost performance and, surprisingly, that selecting files from repositories with 5+ GitHub stars deteriorates performance significantly. Our best model outperforms previous open-source multilingual code generation models (InCoder-6.7B and CodeGen-Multi-2.7B) in both left-to-right generation and infilling on the Java, JavaScript, and Python portions of MultiPL-E, despite being a substantially smaller model. All models are released under an OpenRAIL license at https://hf.co/bigcode.
Diversity maximization aims to select a diverse and representative subset of items from a large dataset. It is a fundamental optimization task that finds applications in data summarization, feature selection, web search, recommender systems, and elsewhere. However, in a setting where data items are associated with different groups according to sensitive attributes like sex or race, it is possible that algorithmic solutions for this task, if left unchecked, will under- or over-represent some of the groups. Therefore, we are motivated to address the problem of \emph{max-min diversification with fairness constraints}, aiming to select $k$ items to maximize the minimum distance between any pair of selected items while ensuring that the number of items selected from each group falls within predefined lower and upper bounds. In this work, we propose an exact algorithm based on integer linear programming that is suitable for small datasets as well as a $\frac{1-\varepsilon}{5}$-approximation algorithm for any $\varepsilon \in (0, 1)$ that scales to large datasets. Extensive experiments on real-world datasets demonstrate the superior performance of our proposed algorithms over existing ones.
Fine-grained visual parsing, including fine-grained part segmentation and fine-grained object recognition, has attracted considerable critical attention due to its importance in many real-world applications, e.g., agriculture, remote sensing, and space technologies. Predominant research efforts tackle these fine-grained sub-tasks following different paradigms, while the inherent relations between these tasks are neglected. Moreover, given most of the research remains fragmented, we conduct an in-depth study of the advanced work from a new perspective of learning the part relationship. In this perspective, we first consolidate recent research and benchmark syntheses with new taxonomies. Based on this consolidation, we revisit the universal challenges in fine-grained part segmentation and recognition tasks and propose new solutions by part relationship learning for these important challenges. Furthermore, we conclude several promising lines of research in fine-grained visual parsing for future research.
Fine-grained visual recognition is to classify objects with visually similar appearances into subcategories, which has made great progress with the development of deep CNNs. However, handling subtle differences between different subcategories still remains a challenge. In this paper, we propose to solve this issue in one unified framework from two aspects, i.e., constructing feature-level interrelationships, and capturing part-level discriminative features. This framework, namely PArt-guided Relational Transformers (PART), is proposed to learn the discriminative part features with an automatic part discovery module, and to explore the intrinsic correlations with a feature transformation module by adapting the Transformer models from the field of natural language processing. The part discovery module efficiently discovers the discriminative regions which are highly-corresponded to the gradient descent procedure. Then the second feature transformation module builds correlations within the global embedding and multiple part embedding, enhancing spatial interactions among semantic pixels. Moreover, our proposed approach does not rely on additional part branches in the inference time and reaches state-of-the-art performance on 3 widely-used fine-grained object recognition benchmarks. Experimental results and explainable visualizations demonstrate the effectiveness of our proposed approach. The code can be found at https://github.com/iCVTEAM/PART.
Non-Pharmaceutical Interventions (NPIs), such as social gathering restrictions, have shown effectiveness to slow the transmission of COVID-19 by reducing the contact of people. To support policy-makers, multiple studies have first modeled human mobility via macro indicators (e.g., average daily travel distance) and then studied the effectiveness of NPIs. In this work, we focus on mobility modeling and, from a micro perspective, aim to predict locations that will be visited by COVID-19 cases. Since NPIs generally cause economic and societal loss, such a micro perspective prediction benefits governments when they design and evaluate them. However, in real-world situations, strict privacy data protection regulations result in severe data sparsity problems (i.e., limited case and location information). To address these challenges, we formulate the micro perspective mobility modeling into computing the relevance score between a diffusion and a location, conditional on a geometric graph. we propose a model named Deep Graph Diffusion Infomax (DGDI), which jointly models variables including a geometric graph, a set of diffusions and a set of locations.To facilitate the research of COVID-19 prediction, we present two benchmarks that contain geometric graphs and location histories of COVID-19 cases. Extensive experiments on the two benchmarks show that DGDI significantly outperforms other competing methods.
Graph neural networks (GNNs) are popular weapons for modeling relational data. Existing GNNs are not specified for attribute-incomplete graphs, making missing attribute imputation a burning issue. Until recently, many works notice that GNNs are coupled with spectral concentration, which means the spectrum obtained by GNNs concentrates on a local part in spectral domain, e.g., low-frequency due to oversmoothing issue. As a consequence, GNNs may be seriously flawed for reconstructing graph attributes as graph spectral concentration tends to cause a low imputation precision. In this work, we present a regularized graph autoencoder for graph attribute imputation, named MEGAE, which aims at mitigating spectral concentration problem by maximizing the graph spectral entropy. Notably, we first present the method for estimating graph spectral entropy without the eigen-decomposition of Laplacian matrix and provide the theoretical upper error bound. A maximum entropy regularization then acts in the latent space, which directly increases the graph spectral entropy. Extensive experiments show that MEGAE outperforms all the other state-of-the-art imputation methods on a variety of benchmark datasets.
Large Language Models (LLMs) play an ever-increasing role in the field of Artificial Intelligence (AI)--not only for natural language processing but also for code understanding and generation. To stimulate open and responsible research on LLMs for code, we introduce The Stack, a 3.1 TB dataset consisting of permissively licensed source code in 30 programming languages. We describe how we collect the full dataset, construct a permissively licensed subset, present a data governance plan, discuss limitations, and show promising results on text2code benchmarks by training 350M-parameter decoders on different Python subsets. We find that (1) near-deduplicating the data significantly boosts performance across all experiments, and (2) it is possible to match previously reported HumanEval and MBPP performance using only permissively licensed data. We make the dataset available at https://hf.co/BigCode, provide a tool called "Am I in The Stack" (https://hf.co/spaces/bigcode/in-the-stack) for developers to search The Stack for copies of their code, and provide a process for code to be removed from the dataset by following the instructions at https://www.bigcode-project.org/docs/about/the-stack/.
Applying Deep Learning (DL) models to graphical causal learning has brought outstanding effectiveness and efficiency but is still far from widespread use in domain sciences. In research of EHR (Electronic Healthcare Records), we realize that some confounding bias inherently exists in the causally formed data, which DL cannot automatically adjust. Trace to the source is because the Acyclic Causal Graph can be Multi-Dimensional, so the bias and causal learning happen in two subspaces, which makes it unobservable from the learning process. This paper initially raises the concept of Dimensionality for causal graphs. In our case, the 3-Dimensional DAG (Directed Acyclic Graph) space is defined by the axes of causal variables, the Absolute timeline, and Relative timelines; This is also the essential difference between Causality and Correlation problems. We propose a novel new framework Causal Representation Learning (CRL), to realize Graphical Causal Learning in latent space, which aims to provide general solutions for 1) the inherent bias adjustment and 2) the DL causal models generalization problem. We will also demonstrate the realization of CRL with originally designed architecture and experimentally confirm its feasibility.
We present HPD: Harry Potter Dialogue Dataset to facilitate the study of building dialogue agents for characters in a story. It differs from existing dialogue datasets in two aspects: 1) HPD provides rich background information about the novel Harry Potter, including scene, character attributes, and character relations; 2) All these background information will change as the story goes on. In other words, each dialogue session in HPD correlates to a different background, and the storyline determines how the background changes. We evaluate some baselines (e.g., GPT-2, BOB) on both automatic and human metrics to determine how well they can generate Harry Potter-like responses. Experimental results indicate that although the generated responses are fluent and relevant to the dialogue history, they are remained to sound out of character for Harry, indicating there is a large headroom for future studies. Our dataset is available.
For computational efficiency, surrogate models have been used to emulate mathematical simulators for physical or biological processes. High-speed simulation is crucial for conducting uncertainty quantification (UQ) when the simulation is repeated over many randomly sampled input points (aka, the Monte Carlo method). In some cases, UQ is only feasible with a surrogate model. Recently, Deep Neural Network (DNN) surrogate models have gained popularity for their hard-to-match emulation accuracy. However, it is well-known that DNN is prone to errors when input data are perturbed in particular ways, the very motivation for adversarial training. In the usage scenario of surrogate models, the concern is less of a deliberate attack but more of the high sensitivity of the DNN's accuracy to input directions, an issue largely ignored by researchers using emulation models. In this paper, we show the severity of this issue through empirical studies and hypothesis testing. Furthermore, we adopt methods in adversarial training to enhance the robustness of DNN surrogate models. Experiments demonstrate that our approaches significantly improve the robustness of the surrogate models without compromising emulation accuracy.