🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Halo 2 Montage out now. Rts are appreciated ❤️ youtu.be/l__2ARoBjUo

Pracking with da boyz Joey Gargaro Vamp!r3 ^ω^ #WheresEvader twitch.tv/amishacorns mixer.com/Revampe twitch.tv/twin_savior

Great games today at UGC and made it to the pool play tomorrow with Argo Cycling Team! No time for FFA though =(

Jordan Dotzel STILL GOT IT!

Cute little snipes video made from a friend and fellow snipper snapper, Feccez youtu.be/d1g0yNsifmE

@TheMLGClutch Wasn't me this time...

What should I make my gt? (Please ignore if you hacked me)

New GT: Amish JD Add me if you were previously on my list and ill do my best to add everyone else

Find me or my students (Jordan Dotzel and @akhauriyash123) at ICML Conference next week to discuss our work on LLM quantization and NAS. DM me to get coffee or to hang out and talk about ML efficiency. Papers: - arxiv.org/abs/2405.03103 - arxiv.org/abs/2403.02484

Im back

Excited to see Jason Weston highlighting the importance of contextual behavior in transformers! In our #EMNLP2024 paper 'ShadowLLM', we show that a tiny neural network can contextually predict which heads and neurons to prune. arxiv.org/abs/2406.16635 (1/4)

Thanks Dr. Masahiro Tanaka from Microsoft Deepspeed for an impressive guest lecture in my Deep Learning Efficiency course. youtube.com/watch?v=9FogHL…

Why Many Token When Few Do Trick? Meet Attamba – Attamba replaces Key-Value projections with SSMs, unlocking multi-token compression before attention, improving perplexity by 24% over transformers of similar memory footprint! Paper: arxiv.org/abs/2411.17685 🧵

What if a small model could solve most of a task and call on a bigger, more powerful one only when it hits a really hard part? It's exactly an idea behind SplitReason, a new method from Cornell University researchers. To make SlipReason efficient on hardware, they also introduced a new

Video, meet audio. 🎥🤝🔊 With Veo 3, our new state-of-the-art generative video model, you can add soundtracks to clips you make. Create talking characters, include sound effects, and more while developing videos in a range of cinematic styles. 🧵

Not consulting AI models in 2025 is medical malpractice. No other way to put it. I follow the field closely to know how to use it, and that saved my life. Not every patient does. People are dying today for misdiagnoses that o3 would get right 10 out of 10 times

I'm backkk GT: Amish JD Not JD Amish

Randomly, I said I was a pro gamer, wrote my essay on Mason Cobb, and I got accepted to Cornell

I can now legally rent a port-a-potty #sickcoolomg

Found Amish Acorns