Deep learning has gained substantial popularity in recent years. Developers mainly rely on libraries and tools to add deep learning capabilities to their software. What kinds of bugs are frequently found in such software? What are the root causes of such bugs? What impacts do such bugs have? Which stages of deep learning pipeline are more bug prone? Are there any antipatterns? Understanding such characteristics of bugs in deep learning software has the potential to foster the development of better deep learning platforms, debugging mechanisms, development practices, and encourage the development of analysis and verification frameworks. Therefore, we study 2716 high-quality posts from Stack Overflow and 500 bug fix commits from Github about five popular deep learning libraries Caffe, Keras, Tensorflow, Theano, and Torch to understand the types of bugs, root causes of bugs, impacts of bugs, bug-prone stage of deep learning pipeline as well as whether there are some common antipatterns found in this buggy software. The key findings of our study include: data bug and logic bug are the most severe bug types in deep learning software appearing more than 48% of the times, major root causes of these bugs are Incorrect Model Parameter (IPS) and Structural Inefficiency (SI) showing up more than 43% of the times.We have also found that the bugs in the usage of deep learning libraries have some common antipatterns.
more »
« less
DeepLocalize: Fault Localization for Deep Neural Networks
Deep Neural Networks (DNNs) are becoming an integral part of most software systems. Previous work has shown that DNNs have bugs. Unfortunately, existing debugging techniques don't support localizing DNN bugs because of the lack of understanding of model behaviors. The entire DNN model appears as a black box. To address these problems, we propose an approach and a tool that automatically determines whether the model is buggy or not, and identifies the root causes for DNN errors. Our key insight is that historic trends in values propagated between layers can be analyzed to identify faults, and also localize faults. To that end, we first enable dynamic analysis of deep learning applications: by converting it into an imperative representation and alternatively using a callback mechanism. Both mechanisms allows us to insert probes that enable dynamic analysis over the traces produced by the DNN while it is being trained on the training data. We then conduct dynamic analysis over the traces to identify the faulty layer or hyperparameter that causes the error. We propose an algorithm for identifying root causes by capturing any numerical error and monitoring the model during training and finding the relevance of every layer/parameter on the DNN outcome. We have collected a benchmark containing 40 buggy models and patches that contain real errors in deep learning applications from Stack Overflow and GitHub. Our benchmark can be used to evaluate automated debugging tools and repair techniques. We have evaluated our approach using this DNN bug-and-patch benchmark, and the results showed that our approach is much more effective than the existing debugging approach used in the state-of-the-practice Keras library. For 34/40 cases, our approach was able to detect faults whereas the best debugging approach provided by Keras detected 32/40 faults. Our approach was able to localize 21/40 bugs whereas Keras did not localize any faults.
more »
« less
- NSF-PAR ID:
- 10287422
- Date Published:
- Journal Name:
- 2021 IEEE/ACM 43rd International Conference on Software Engineering (ICSE)
- Page Range / eLocation ID:
- 251 to 262
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Significant interest in applying Deep Neural Network (DNN) has fueled the need to support engineering of software that uses DNNs. Repairing software that uses DNNs is one such unmistakable SE need where automated tools could be very helpful; however, we do not fully understand challenges to repairing and patterns that are utilized when manually repairing them. What challenges should automated repair tools address? What are the repair patterns whose automation could help developers? Which repair patterns should be assigned a higher priority for automation? This work presents a comprehensive study of bug fix patterns to address these questions. We have studied 415 repairs from Stack Overflow and 555 repairs from GitHub for five popular deep learning libraries Caffe, Keras, Tensorflow, Theano, and Torch to understand challenges in repairs and bug repair patterns. Our key findings reveal that DNN bug fix patterns are distinctive compared to traditional bug fix patterns; the most common bug fix patterns are fixing data dimension and neural network connectivity; DNN bug fixes have the potential to introduce adversarial vulnerabilities; DNN bug fixes frequently introduce new bugs; and DNN bug localization, reuse of trained model, and coping with frequent releases are major challenges faced by developers when fixing bugs. We also contribute a benchmark of 667 DNN (bug, repair) instances.more » « less
-
Continuous Integration (CI) practices encourage developers to frequently integrate code into a shared repository. Each integration is validated by automatic build and testing such that errors are revealed as early as possible. When CI failures or integration errors are reported, existing techniques are insufficient to automatically locate the root causes for two reasons. First, a CI failure may be triggered by faults in source code and/or build scripts, while current approaches consider only source code. Second, a tentative integration can fail because of build failures and/or test failures, while existing tools focus on test failures only. This paper presents UniLoc, the first unified technique to localize faults in both source code and build scripts given a CI failure log, without assuming the failure’s location (source code or build scripts) and nature (a test failure or not). Adopting the information retrieval (IR) strategy, UniLoc locates buggy files by treating source code and build scripts as documents to search and by considering build logs as search queries. However, instead of naïvely applying an off-the-shelf IR technique to these software artifacts, for more accurate fault localization, UniLoc applies various domain-specific heuristics to optimize the search queries, search space, and ranking formulas. To evaluate UniLoc, we gathered 700 CI failure fixes in 72 open-source projects that are built with Gradle. UniLoc could effectively locate bugs with the average MRR (Mean Reciprocal Rank) value as 0.49, MAP (Mean Average Precision) value as 0.36, and NDCG (Normalized Discounted Cumulative Gain) value as 0.54. UniLoc outperformed the state-of-the-art IR-based tool BLUiR and Locus. UniLoc has the potential to help developers diagnose root causes for CI failures more accurately and efficiently.more » « less
-
null (Ed.)Automated debugging techniques, including fault localization and program repair, have been studied for over a decade. However, the only existing connection between fault localization and program repair is that fault localization computes the potential buggy elements for program repair to patch. Recently, a pioneering work, ProFL, explored the idea of unified debugging to unify fault localization and program repair in the other direction for the first time to boost both areas. More specifically, ProFL utilizes the patch execution results from one state-of-the-art repair system, PraPR, to help improve state-of-the-art fault localization. In this way, ProFL not only improves fault localization for manual repair, but also extends the application scope of automated repair to all possible bugs (not only the small ratio of bugs that can be automatically fixed). However, ProFL only considers one APR system (i.e., PraPR), and it is not clear how other existing APR systems based on different designs contribute to unified debugging. In this work, we perform an extensive study of the unified-debugging approach on 16 state-of-the-art program repair systems for the first time. Our experimental results on the widely studied Defects4J benchmark suite reveal various practical guidelines for unified debugging, such as (1) nearly all the studied 16 repair systems can positively contribute to unified debugging despite their varying repairing capabilities, (2) repair systems targeting multi-edit patches can bring extraneous noise into unified debugging, (3) repair systems with more executed/plausible patches tend to perform better for unified debugging, and (4) unified debugging effectiveness does not rely on the availability of correct patches in automated repair. Based on our results, we further propose an advanced unified debugging technique, UniDebug++, which can localize over 20% more bugs within Top-1 positions than state-of-the-art unified debugging technique, ProFL.more » « less
-
Dynamic memory managers are a crucial component of almost every modern software system. In addition to implementing efficient allocation and reclamation, memory managers provide the essential abstraction of memory as distinct objects, which underpins the properties of memory safety and type safety. Bugs in memory managers, while not common, are extremely hard to diagnose and fix. One reason is that their implementations often involve tricky pointer calculations, raw memory manipulation, and complex memory state invariants. While these properties are often documented, they are not specified in any precise, machine-checkable form. A second reason is that memory manager bugs can break the client application in bizarre ways that do not immediately implicate the memory manager at all. A third reason is that existing tools for debugging memory errors, such as Memcheck, cannot help because they rely on correct allocation and deallocation information to work. In this paper we present Permchecker, a tool designed specifically to detect and diagnose bugs in memory managers. The key idea in Permchecker is to make the expected structure of the heap explicit by associating typestates with each piece of memory. Typestate captures elements of both type (e.g., page, block, or cell) and state (e.g., allocated, free, or forwarded). Memory manager developers annotate their implementation with information about the expected typestates of memory and how heap operations change those typestates. At runtime, our system tracks the typestates and ensures that each memory access is consistent with the expected typestates. This technique detects errors quickly, before they corrupt the application or the memory manager itself, and it often provides accurate information about the reason for the error. The implementation of Permchecker uses a combination of compile-time annotation and instrumentation, and dynamic binary instrumentation (DBI). Because the overhead of DBI is fairly high, Permchecker is suitable for a testing and debugging setting and not for deployment. It works on a wide variety of existing systems, including explicit malloc/free memory managers and garbage collectors, such as those found in JikesRVM and OpenJDK. Since bugs in these systems are not numerous, we developed a testing methodology in which we automatically inject bugs into the code using bug patterns derived from real bugs. This technique allows us to test Permchecker on hundreds or thousands of buggy variants of the code. We find that Permchecker effectively detects and localizes errors in the vast majority of cases; without it, these bugs result in strange, incorrect behaviors usually long after the actual error occurs.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/ICSE43902.2021.00034
