Join Microsoft at 8 a.m. PT on Wednesday, June 21, to learn how Microsoft is accelerating scientific discovery with Azure Quantum. Accelerating scientific discovery with Azure Quantum https://
bit.ly/463HOMf
COMPUTING
-
Microsoft Azure Quantum Accelerates Scientific Discovery Event
By
–
-
Britain lacks cloud giant GPU infrastructure for AI training
By
–
'Although Oracle has a cluster of the advanced graphics processing units (GPUs) needed to train large models, none of the cloud-computing giants has yet chosen Britain as the base for what techies call the "compute".' https://
economist.com/leaders/2023/0
6/15/how-britain-can-become-an-ai-superpower
… -
Converting 2D iPhone pixels to 3D voxels explained
By
–
A lot of math. I got through calculus, but hated it, and this stuff is way over my head. That said, the math converts 2D pixels captured on an iPhone, say, to 3D voxels in 3D scenes. So, for most of us that's most of what we need to know.
-
OctoAI Simplifies Generative AI Production Deployment
By
–
Serving Generative AI just got a lot easier with OctoML's OctoAI Generative AI, because of its compute intensity, is especially hard to put into production. https://
zdnet.com/article/servin
g-generative-ai-just-got-a-lot-easier-with-octomls-octoai/
… @OctoML #AI #deeplearning #GenerativeAI -
Research SuperCluster Enables Breakthrough AI Research at Scale
By
–
The Research SuperCluster enables our research teams to do new, breakthrough AI research at an even greater scale and pace than ever before. Ning Dong and Maha Elbayad shared more about how it's enabling their work.
— AI at Meta (@AIatMeta) 15 juin 2023
More details ➡️ https://t.co/i1W6ld96Eb pic.twitter.com/ZgtcmrzF8OThe Research SuperCluster enables our research teams to do new, breakthrough AI research at an even greater scale and pace than ever before. Ning Dong and Maha Elbayad shared more about how it's enabling their work. More details https://
bit.ly/3N2qHCR -
Postdoc Position: Cloud Robotics and AWS/Google Cloud Systems
By
–
Seeking Postdoc with experience in Robotics and Cloud-Based Systems (AWS, Google Cloud, Azure) to work closely with @UCBerkeley Profs. Ken Goldberg and Joey Gonzales on research on cloud robotics architectures and platforms such as FogROS2: https://
docs.google.com/document/d/e/2
PACX-1vR5n3aHbRl0luk3acq2xmf3HzmS6gJvEX7LC24c503PIT8mlWybxJiWAb3ZRVT2XKJWU62ztMvpQrqr/pub
… -
Using 8 GPUs to reduce processing time to minutes
By
–
yes, you can of course use 8 GPUs to bring this down to a few minutes! 🙂
-
Memory reduction: Full finetuning vs Adapter vs LoRA comparison
By
–
Btw, it's also a big reduction in memory: 16 gigs instead of 6 x 40 GB. Due to the reduced parameter counts in the backward pass: Full finetuning: 7,217,189,760
Adapter: 1,365,330
Adapter v2: 3,839,186
LoRA: 3,506,176 (I used LoRA rank of 16 to match Adapter v2 above.) -
Efficient Falcon Finetuning with LoRA on Single GPU
By
–
Have been heads-down porting LoRA (low-rank-adaptation) to finetune Falcon more efficiently and ran some performance benchmarks. My longer write-up here: https://
lightning.ai/pages/communit
y/finetuning-falcon-efficiently/
… Long story short: you can finetune Falcon in 1 h on a 52k dataset using a single GPU with 16 GB RAM. -
MS Excel: Effective Tool for Data Silo Management
By
–
MS Excel has definitely presented itself as an effective computational tool for working with data silos. Want to get adept with various MS Excel resources? #ResourceWatch #data #datascience #algorithms