Andrés Villa

92 posts

Andrés Villa

Andrés Villa

@AndrsVi28

I'm a Postdoctoral Fellowship at KAUST, working on Multimodal Models, Continual Learning and AI4Sciences.

Jeddah Katılım Mart 2020
269 Takip Edilen119 Takipçiler
Andrés Villa
Andrés Villa@AndrsVi28·
🎉 Excited to share that I’ll be presenting my poster: 🪄 Behind the Magic: MERLIM A benchmark for evaluating Large Vision-Language Models (IT-LVLMs) on core vision tasks. 📅 June 11 from 10:05 to 10:45am | 📍@CVPR Exhall D board 46. 🔗 arxiv.org/abs/2312.02219 #AI #Multimodal
English
0
0
0
38
Andrés Villa retweetledi
Bernard Ghanem
Bernard Ghanem@BernardSGhanem·
For those attending @CVPR, there will be a vote for ICCV29 after the PAMI TC meeting tomorrow. Make sure to vote. There is a bid for Dubai and I am one of the PCs. Check out our website/bid for details: iccv2029.com. Let's bring ICCV to the Middle East!
English
0
12
46
5.1K
Andrés Villa retweetledi
Prof. Anima Anandkumar
Prof. Anima Anandkumar@AnimaAnandkumar·
For the first time, we show that the Llama 7B LLM can be trained on a single consumer-grade GPU (RTX 4090) with only 24GB memory. This represents more than 82.5% reduction in memory for storing optimizer states during training. Training LLMs from scratch currently requires huge computational resources with large memory GPUs. While there has been significant progress in reducing memory requirements during fine-tuning (e.g., LORA), they do not apply for pre-training LLMs. We design methods that overcome this obstacle and provide significant memory reduction throughout training LLMs. Training LLMs often requires the use of preconditioned optimization algorithms such as Adam to achieve rapid convergence. These algorithms accumulate extensive gradient statistics, proportional to the model's parameter size, making the storage of these optimizer states the primary memory constraint during training. Instead of focusing just on engineering and system efforts to reduce memory consumption, we went back to fundamentals. We looked at the slow-changing low-rank structure of the gradient matrix during training. We introduce a novel approach that leverages the low-rank nature of gradients via Gradient Low-Rank Projection (GaLore). So instead of expressing the weight matrix as low rank, which leads to a big performance degradation during pretraining, we instead express the gradient weight matrix as low rank without performance degradation, while significantly reducing memory requirements. @jiawzhao @BeidiChen @tydsh
AK@_akhaliq

GaLore Memory-Efficient LLM Training by Gradient Low-Rank Projection Training Large Language Models (LLMs) presents significant memory challenges, predominantly due to the growing size of weights and optimizer states. Common memory-reduction approaches, such as low-rank

English
48
363
2.2K
407.5K
Andrés Villa retweetledi
lxcv
lxcv@latinxincv·
🌞We are glad to announce a new venture: the @latinxincv coaching program! 🥳 💡Get to know the reseach done by LatinX researchers and become the new generation of researchers in Computer Vision and AI! Work along well-known professors and researchers around the world!
lxcv tweet media
English
1
7
8
1.1K
Andrés Villa retweetledi
Rodolfo Valiente
Rodolfo Valiente@rvalienter·
Don't miss this opportunity 📢 . The submission deadline for full papers for LXAI #CVPR2024 has been extended! 🌐 🌟There are travel grants for accepted papers. Share this opportunity with peers and encourage them to participate 🤝 @latinxincv @CVPR
lxcv@latinxincv

🚨 Important Update 🚨 The deadline to submit full papers to the LatinX in CV Workshop at @CVPR 2024 has been extended (March 10th, 2024)! Take advantage of this additional time to refine your contributions 🌐 #CVPR2024 #CallForPapers Details: buff.ly/3Syl29o

English
0
5
10
4.6K
Sara Hooker
Sara Hooker@sarahookr·
@3scorciav @_LXAI @latinxincv @SomosNLP_ @OmarUFlorez @LucianaBenotti For the Americas right now, we are supporting Portuguese, Spanish, Haitian Creole and Hawaiian. We are still looking for language ambassadors for Haitian Creole and Hawaiian. These two languages are currently at high risk. If you know of anyone -- please let us know.
English
2
0
3
213
Sara Hooker
Sara Hooker@sarahookr·
Are you a speaker of Xhosa, Zulu, Afrikaans or Malagasy? These are the four languages project AYA still doesn't have ambassadors for. Join us, and meet others who are working on multilingual representation across the world. We are open sourcing all data and models we build.
English
6
33
38
15.8K
Adel Bibi
Adel Bibi@Adel_Bibi·
@ferjadnaeem @CVPR Hey Ferjad, I wonder, what type of visa did you apply for? Business visa or the temporary visit visa? Was there any place in the application form where you can enter CVPR's event code sent in the invitation letter?
English
2
0
1
475
Ferjad Naeem
Ferjad Naeem@ferjadnaeem·
Updating on here since I got several DMs about my experience. It was a pleasant surprise once I submitted my application. I received my request for biometrics a day later, gave my biometrics a week after and received the approval and request to submit passport 3 days after that.
Ferjad Naeem@ferjadnaeem

I have faced this glass ceiling as a citizen of a developing country since the start of my Ph.D. I am applying early for a visa for #CVPR2023, apparently, there is a 197 days processing time from Switzerland. Does someone know if they speed it up for conference presenters? @CVPR

English
4
0
19
3.7K
Andrés Villa
Andrés Villa@AndrsVi28·
What platform do you recommend to apply for the Canadian Visiting Visa, the GCKey, or the new IRCC portal? #CVPR2023
English
0
0
6
1.1K
Victor Escorcia
Victor Escorcia@3scorciav·
Este año promete 😋😁 Ahora que mis compatriotas 🇨🇴 pueden entrar a UK 🇬🇧 sin visa hay que recibirlos como se merecen. Quien dijo yo?
Victor Escorcia tweet media
Español
2
0
3
450