facebookresearch/metaseq

Integrate LucidRain's RotaryEmbeddings

Open

#621 opened on Jan 27, 2023

View on GitHub
 (2 comments) (0 reactions) (0 assignees)Python (6,195 stars) (701 forks)batch import
enhancementgood first issue

Description

See https://github.com/lucidrains/rotary-embedding-torch/blob/main/rotary_embedding_torch/rotary_embedding_torch.py

And from PaLM paper:

We use RoPE embeddings (Su et al., 2021) rather than absolute or relative position embeddings, since RoPE embeddings have been shown to have better performance on long sequence lengths.

Contributor guide

Tech stack
pythonpytorch
Domain
machine learning
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.
1-2 days
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.
mostly clear
Prerequisites
PythonPyTorchrotary embeddings
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
30
Research direction
The issue requests integrating rotary embeddings from lucidrains's implementation into metaseq. Start by examining the existing positional encoding code in metaseq (likely in the model directory). Study the reference implementation and the PaLM paper to understand the integration points. The issue has no linked PRs or detailed discussion, so the first step is to propose a plan in the issue thread.
Integrate LucidRain's RotaryEmbeddings · facebookresearch/metaseq#621 | Good First Issue