facebookresearch/fairseq

Disparity with huggingface implementation when reproducing the fill_mask

Open

#1,399 opened on Nov 19, 2019

View on GitHub
 (8 comments) (0 reactions) (0 assignees)Python (29,107 stars) (6,224 forks)batch import
bughelp wanted

Description

I was not able to reproduce the fill mask provided here with the huggingface interface, wondering if they are using an outdated model (if a new one has been released) :)

Contributor guide

Tech stack
pythonpytorch
Domain
machine learningai
Issue type
bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
4
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
stale
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
needs investigation
Prerequisites
fairseq basicsHuggingFace transformersfill mask task understanding
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
20
Research direction
Investigate the fill mask reproduction steps in fairseq and compare with HuggingFace's implementation. Check if the model checkpoint used by fairseq is outdated or if there are differences in tokenization or masking logic. Look at the fairseq documentation and any linked examples, then identify the specific disparities and suggest a fix or clarification.