Analysis of Failures and Risks in Deep Learning Model Converters: A Case Study in the ONNX Ecosystem

TL;DR

This study investigates failure modes in deep learning model converters, especially in the ONNX ecosystem, revealing that most defects occur during node conversion and many lead to semantically incorrect models, highlighting areas for improvement.

Contribution

It provides a detailed failure analysis of DL model converters, especially ONNX, and formulates hypotheses about structural causes of these failures, guiding future research.

Findings

75% of defects occur during node conversion

33% of failures involve semantically incorrect models

Models with behavior inconsistencies share operator sequences

Abstract

Software engineers develop, fine-tune, and deploy deep learning (DL) models using a variety of development frameworks and runtime environments. DL model converters move models between frameworks and to runtime environments. Conversion errors compromise model quality and disrupt deployment. However, the failure characteristics of DL model converters are unknown, adding risk when using DL interoperability technologies. This paper analyzes failures in DL model converters. We survey software engineers about DL interoperability tools, use cases, and pain points (N=92). Then, we characterize failures in model converters associated with the main interoperability tool, ONNX (N=200 issues in PyTorch and TensorFlow). Finally, we formulate and test two hypotheses about structural causes for the failures we studied. We find that the node conversion stage of a model converter accounts for ~75% of the defects and 33% of reported failure are related to semantically incorrect models. The cause of semantically incorrect models is elusive, but models with behaviour inconsistencies share operator sequences. Our results motivate future research on making DL interoperability software simpler to maintain, extend, and validate. Research into behavioural tolerances and architectural coverage metrics could be fruitful.