Verify Implementation Equivalence of Large Models

TL;DR

Emerge is a framework that automatically verifies the equivalence of large-model implementations across frameworks by synthesizing and validating rewrite rules, aiding debugging and ensuring correctness.

Contribution

Emerge introduces an automated, rule-synthesis based approach for implementation equivalence verification of large models, reducing manual effort and improving bug detection.

Findings

Successfully verifies equivalence in practical scenarios

Detects 10 of 13 known bugs and finds 8 new issues

Synthesizes effective block-level rewrite rules

Abstract

Verifying whether two implementations of the same large model are equivalent across frameworks is difficult in practice. Even when they realize the same computation, their graphs may differ substantially in operator decomposition, tensor layout, and the use of fused or opaque kernels, making manual rewrite rules hard to build and maintain. We present Emerge, a framework for checking Implementation Equivalence over computation graphs of large-model implementations. Instead of writing rules manually, Emerge represents the two implementations in an e-graph, infers candidate relations from execution values, and synthesizes rewrite rules on demand when existing rules are insufficient. Each synthesized rule is validated using the strongest applicable method, including SMT- based checking for symbolically tractable cases and constraint-aware randomized testing for opaque kernels, and then propagated through e-graph rebuilding to establish larger equivalences. Our current implementation targets inference computation graphs captured from HuggingFace Transformers and vLLM. Our evaluation shows that Emerge establishes equivalence for correct implementation pairs at practical cost, while also providing useful by-products for debugging: it detects 10 of 13 known implementation bugs and uncovers 8 previously unknown implementation issues that were later confirmed by developers. In addition, Emerge synthesizes block-level rules that compare favorably with manually authored ones.