facebookresearch/ParlAI

Support tied positional weights in transformers

Open

#2,184 opened on Nov 19, 2019

View on GitHub
 (1 comment) (0 reactions) (0 assignees)Python (2,089 forks)batch import
EnhancementH1 2020Help WantedMediumdonotreap

Repository metrics

Stars
 (10,371 stars)
PR merge metrics
 (No merged PRs in 30d)

Description

Add a flag to use tied positional embeddings in transformer/generator and transformer/retrieval, and implement the tied weights. Should be False by default for backwards compatibility, and upgrade_opt should be used to migrate old models.

Contributor guide

Research direction
Study the existing transformer implementation in ParlAI. Add a new flag (e.g., tied positional embeddings, default False) to both transformer/generator and transformer/retrieval. Implement weight tying by sharing the weight matrix for positional embeddings. Update upgrade opt to convert old models. Test with a unit test.
Tech stack
python
Domain
ai
Issue type
Feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Mostly clear
Prerequisites
PythonGit
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
65