Shagun Sodhani
shagunsodhani.bsky.social
Shagun Sodhani
@shagunsodhani.bsky.social
FAIR Researcher @metaai.bsky.social Previously Mila-Quebec, Microsoft Research, Adobe Research, IIT Roorkee
Reposted by Shagun Sodhani
📢 2 weeks until #CoLLAs2025!

Here’s what you need to make the most of it

🗓 Program: lifelong-ml.cc/Conferences/...

✅ Papers: lifelong-ml.cc/Conferences/...

📍 Venue: lifelong-ml.cc/Conferences/...

🔗 Info & registration: lifelong-ml.cc/Conferences/...

#AI #ML #ContinualLearning #LifelongLearning
CoLLAs
Collas
lifelong-ml.cc
July 27, 2025 at 1:58 PM
We've been using fairseq2 internally for over a year now. It's stable, developer-friendly, and integrates seamlessly with frameworks like vLLM and accelerate. Highly recommend giving it a try!
👋 Hello world! We’re thrilled to announce the v0.4 release of fairseq2 — an open-source library from FAIR powering many projects at Meta. pip install fairseq2 and explore our trainer API, instruction & preference finetuning (up to 70B), and native vLLM integration.
February 13, 2025 at 1:49 PM
Reposted by Shagun Sodhani
Excited to announce that Dr. Martha White will be an invited speaker at #CoLLAs2025! 🎤

An Associate Professor at the U of Alberta, Canada CIFAR AI Chair, & member of RLAI & Amii, her research focuses on adaptive systems that #continually learn and improve.

Stay tuned for more details on her talk!
February 4, 2025 at 6:18 PM
A very detailed note from #PyTorch devs about DTensor (Distributed Tensors) - how they are designed, what is supported / not supported and whats planned for the future: dev-discuss.pytorch.org/t/dtensor-st...
DTensor - Status, Design and Looking Forward
This post outlines the current status, design principles, and future work items for PyTorch DTensor. Overall Design Objective and Principles DTensor is the PyTorch native tensor sharding primitive. F...
dev-discuss.pytorch.org
January 30, 2025 at 9:30 PM
I will be at @neuripsconf.bsky.social next week! If you're around, let's connect! Would love to discuss #agents, #reasoning, continual learning, careers in industrial research labs & more! We are also presenting a paper on efficient SSL for vision openreview.net/forum?id=Y5D...
Harnessing small projectors and multiple views for efficient vision...
Recent progress in self-supervised (SSL) visual representation learning has led to the development of several different proposed frameworks that rely on augmentations of images but use different...
openreview.net
December 4, 2024 at 5:07 PM
Created a starter pack for people working in continual / lifelong ML. This is obviously incomplete - please comment names (including your own) and I will add them

go.bsky.app/RB4ekFs
November 28, 2024 at 5:10 PM
Collas (@collasconf.bsky.social) is on 🦋!

CoLLAs is the International Conference on Lifelong Learning Agents. It was founded in 2021 to advance research in lifelong learning, meta-learning, multi-task learning, transfer learning & OOD generalization.

Learn more here: lifelong-ml.cc 1/
CoLLAs
Collas
lifelong-ml.cc
November 28, 2024 at 4:03 PM