VarCLR: Variable Semantic Representation Pre-training via Contrastive Learning

Last update: Oct 24, 2022

Overview

VarCLR: Variable Representation Pre-training via Contrastive Learning

New: Paper accepted by ICSE 2022. Preprint at arXiv!

This repository contains code and pre-trained models for VarCLR, a contrastive learning based approach for learning semantic representations of variable names that effectively captures variable similarity, with state-of-the-art results on [email protected].

VarCLR: Variable Representation Pre-training via Contrastive Learning

Step 0: Install

pip install -e .

Step 1: Load a Pre-trained VarCLR Model

from varclr.models import Encoder
model = Encoder.from_pretrained("varclr-codebert")

Step 2: VarCLR Variable Embeddings

Get embedding of one variable

emb = model.encode("squareslab")
print(emb.shape)
# torch.Size([1, 768])

Get embeddings of list of variables (supports batching)

emb = model.encode(["squareslab", "strudel"])
print(emb.shape)
# torch.Size([2, 768])

Step 2: Get VarCLR Similarity Scores

Get similarity scores of N variable pairs

print(model.score("squareslab", "strudel"))
# [0.42812108993530273]
print(model.score(["squareslab", "average", "max", "max"], ["strudel", "mean", "min", "maximum"]))
# [0.42812108993530273, 0.8849745988845825, 0.8035818338394165, 0.889922022819519]

Get pairwise (N * M) similarity scores from two lists of variables

variable_list = ["squareslab", "strudel", "neulab"]
print(model.cross_score("squareslab", variable_list))
# [[1.0000007152557373, 0.4281214475631714, 0.7207341194152832]]
print(model.cross_score(variable_list, variable_list))
# [[1.0000007152557373, 0.4281214475631714, 0.7207341194152832],
#  [0.4281214475631714, 1.0000004768371582, 0.549992561340332],
#  [0.7207341194152832, 0.549992561340332, 1.000000238418579]]

Step 3: Reproduce IdBench Benchmark Results

Load the IdBench benchmark

from varclr.benchmarks import Benchmark

# Similarity on IdBench-Medium
b1 = Benchmark.build("idbench", variant="medium", metric="similarity")
# Relatedness on IdBench-Large
b2 = Benchmark.build("idbench", variant="large", metric="relatedness")

Compute VarCLR scores and evaluate

id1_list, id2_list = b1.get_inputs()
predicted = model.score(id1_list, id2_list)
print(b1.evaluate(predicted))
# {'spearmanr': 0.5248567181503295, 'pearsonr': 0.5249843473193132}

print(b2.evaluate(model.score(*b2.get_inputs())))
# {'spearmanr': 0.8012168379981921, 'pearsonr': 0.8021791703187449}

Let's compare with the original CodeBERT

codebert = Encoder.from_pretrained("codebert")
print(b1.evaluate(codebert.score(*b1.get_inputs())))
# {'spearmanr': 0.2056582946575104, 'pearsonr': 0.1995058696927054}
print(b2.evaluate(codebert.score(*b2.get_inputs())))
# {'spearmanr': 0.3909218857993804, 'pearsonr': 0.3378219622284688}

Results on IdBench benchmarks

Similarity

Method	Small	Medium	Large
FT-SG	0.30	0.29	0.28
LV	0.32	0.30	0.30
FT-cbow	0.35	0.38	0.38
VarCLR-Avg	0.47	0.45	0.44
VarCLR-LSTM	0.50	0.49	0.49
VarCLR-CodeBERT	0.53	0.53	0.51

Combined-IdBench	0.48	0.59	0.57
Combined-VarCLR	0.66	0.65	0.62

Relatedness

Method	Small	Medium	Large
LV	0.48	0.47	0.48
FT-SG	0.70	0.71	0.68
FT-cbow	0.72	0.74	0.73
VarCLR-Avg	0.67	0.66	0.66
VarCLR-LSTM	0.71	0.70	0.69
VarCLR-CodeBERT	0.79	0.79	0.80

Combined-IdBench	0.71	0.78	0.79
Combined-VarCLR	0.79	0.81	0.85

Pre-train your own VarCLR models

Coming soon.

Cite

If you find VarCLR useful in your research, please cite our [email protected]:

@misc{chen2021varclr,
      title={VarCLR: Variable Semantic Representation Pre-training via Contrastive Learning},
      author={Qibin Chen and Jeremy Lacomis and Edward J. Schwartz and Graham Neubig and Bogdan Vasilescu and Claire Le Goues},
      year={2021},
      eprint={2112.02650},
      archivePrefix={arXiv},
      primaryClass={cs.SE}
}

VarCLR: Variable Semantic Representation Pre-training via Contrastive Learning

Related tags

Overview

VarCLR: Variable Representation Pre-training via Contrastive Learning

Step 0: Install

Step 1: Load a Pre-trained VarCLR Model

Step 2: VarCLR Variable Embeddings

Get embedding of one variable

Get embeddings of list of variables (supports batching)

Step 2: Get VarCLR Similarity Scores

Get similarity scores of N variable pairs

Get pairwise (N * M) similarity scores from two lists of variables

Step 3: Reproduce IdBench Benchmark Results

Load the IdBench benchmark

Compute VarCLR scores and evaluate

Let's compare with the original CodeBERT

Results on IdBench benchmarks

Similarity

Relatedness

Pre-train your own VarCLR models

Cite

Owner

squaresLab

State-to-Distribution (STD) Model

Official Repository for the ICCV 2021 paper "PixelSynth: Generating a 3D-Consistent Experience from a Single Image"

PyTorch implementation of our ICCV 2021 paper, Interpretation of Emergent Communication in Heterogeneous Collaborative Embodied Agents.

Implementation of popular SOTA self-supervised learning algorithms as Fastai Callbacks.

For medical image segmentation

LF-YOLO (Lighter and Faster YOLO) is used to detect defect of X-ray weld image.

Official implementation of the paper "Lightweight Deep CNN for Natural Image Matting via Similarity Preserving Knowledge Distillation"

[NeurIPS 2021] SSUL: Semantic Segmentation with Unknown Label for Exemplar-based Class-Incremental Learning

A minimalist tool to display a network graph.

BaseCls BaseCls 是一个基于 MegEngine 的预训练模型库，帮助大家挑选或训练出更适合自己科研或者业务的模型结构

Semantic Segmentation Architectures Implemented in PyTorch

Code accompanying the paper Shared Independent Component Analysis for Multi-subject Neuroimaging

Concept drift monitoring for HA model servers.

Python Library for learning (Structure and Parameter) and inference (Statistical and Causal) in Bayesian Networks.

Empower Sequence Labeling with Task-Aware Language Model

Metadata-Extractor - Metadata Extractor Script can be used to read in exif metadata

OMAMO: orthology-based model organism selection

Google Landmark Recogntion and Retrieval 2021 Solutions

code for paper -- "Seamless Satellite-image Synthesis"

Graph-based community clustering approach to extract protein domains from a predicted aligned error matrix