stephenroller

Follow

🗨️

Stephen Roller stephenroller

🗨️

Follow

NLP researcher working on large language models.

535 followers · 44 following

Achievements

BetaSend feedback

Achievements

BetaSend feedback

Highlights

Pro

Block or Report

Block or report stephenroller

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse

Pinned

facebookresearch/metaseq Public

Repo for external large-scale work

Python 4.2k 432
facebookresearch/ParlAI Public

A framework for training and evaluating AI models on a variety of openly available dialogue datasets.

Python 9.7k 2k

955 contributions in the last year

Learn how we count contributions

Less More

Activity overview

Contributed to facebookresearch/metaseq, facebookresearch/ParlAI, huggingface/transformers and 4 other repositories

Contribution activity

December 2022

Created 2 commits in 1 repository

facebookresearch/metaseq 2 commits

Opened 4 pull requests in 1 repository

facebookresearch/metaseq 2 open 2 merged

Add Sharan to CODEOWNERS Dec 21
Fix config.yml dump in training runs. Dec 21
Bigbig update Dec 19
release intermediate checkpoints Dec 9

Reviewed 11 pull requests in 2 repositories

facebookresearch/metaseq 9 pull requests

Add optional type conversion to FSDP resharding script Dec 23
Updated paper Dec 22
Update paper link in OPTIML Readme Dec 22
add sample fetch time logging Dec 22
Automatically evaluate checkpoints after copying to NFS Dec 12
Scale LR instead Dec 7
Compute ratio to existing ewm Dec 6
barrierless rendezvous for faster init Dec 2
Automatically detect and skip loss spikes Dec 1

facebookresearch/ParlAI 2 pull requests

torch concat backward compatibility Dec 1
[FSDP] Zero 3 Optimization Support Dec 1

7 contributions in private repositories Dec 2 – Dec 22