Tung Vu

[AI Day Ho Chi Minh City] Build Vietnamese LLMs that truly reason - not just respond. Don’t miss this must-attend session at #NVIDIAAIDay, where GreenNode will walk you through how to train Vietnamese language models that think before they answer. 💡 A practical, full-stack workflow for Vietnamese LLMs 💡 How to design domain-specific datasets that capture real nuance 💡 Techniques to fine-tune models that reason transparently and verifiably 🗓️ September 23 | 🕓 2:00 p.m. - 2:40 p.m. Vietnam time Attending the event? Add to your schedule: nvda.ws/46fd90t

Do you remember when you joined X? I do! #MyXAnniversary @greennode23

Looking for hosted DeepSeek or Llama 3.3 70b system on an Nvidia Cloud Partner in Asia Pacific that’s ISO 27001, PCI-DSS compliant to fine-tune with your own data? GreenNode.ai got you covered. SOC2-type 2 compliant is coming this year too.

@deepseek_ai we’ve hosted DeepSeek-R1 671b on our @greennode23 AI Platform and other open source models like llama-3.3 70b. Backed by H100 SXM5, Infiniband for multi-node training, high perf storage,we enable self-service via Stripe integration. We’re Nvidia Partner. marketplace.nvidia.com/en-us/enterpri…

🚀 DeepSeek-R1 is here! ⚡ Performance on par with OpenAI-o1 📖 Fully open-source model & technical report 🏆 MIT licensed: Distill & commercialize freely! 🌐 Website & API are live now! Try DeepThink at chat.deepseek.com today! 🐋 1/n

Live January 18 x.com/i/broadcasts/1…

Livestreaming Road to top 1B LLM by @qnguyen3 on GreenNode AI Platform.

Our 5-month journey conducting independent analysis & benchmarking of AMD MI300X vs Nvidia H100 + H200 Detailed, open source low-level benchmarks performance vs TCO Comprehensive public recommendations It’s not just immature software, they need to change how they do development

@karpathy The GPU billboards in SF are eye-opening. Saw increasing number of such ads during Nvidia GTC in March and recently @SaaStrAnnual in Sept. I was living in Austin for 4 years and didn’t see such things :D Serving US clients with H100/4090/L40S from APAC now with @greennode23

Driving around SF. Omg this is crazy I can't believe there's billboards advertising cloud GPUs on the streets of SF, the hype is totally out of control. That said, actually I would like some more GPU and I haven't heard of this company yet this looks interesting.

@jeseem @karpathy Ping me at tung.vu@greennode.ai. GreenNode is official Nvidia Cloud Partner, I would love to learn more specifically about your wandb private metrics. Other than that we have all other things you mentioned. Essentially a GPU cloud coupled with some key CPU-bases cloud offerings

@karpathy I initially asked all the traditional cloud companies for H100 and almost all of them denied it as they were all overbooked. The GPU only cloud is way to go. If they can add object store, wandb kind of private metrics and a private dataset store, would pick that over any cloud.

@divgarg @MultiOn_AI All the best Div for your next adventure. Thanks for the partnership with GreenNode.ai since our early days @greennode23

Chapter II. Update: I will be moving on from @MultiOn_AI to explore new things! I am enthusiastic about the road ahead and remain as an advisor to the team. Excited for what's NeXT.

@WenhuChen Our pleasure to support you and your team with @greennode23 AI GPU Cloud.

Very excited to share MEGA-Bench. The first benchmark to scale up foundation model evaluation to 500 (novel) real-world tasks. All the tasks were contributed and reviewed by our expert annotators. In the future, you don't need to need to evaluate on 27 datasets like Qwen2-VL! You only need to evaluate on our MEGA-Bench to see all the task performance! What's new about MEGA-Bench? - It contains 505 diverse tasks evaluating multimodal models across 8 grand application types, 7 input visual formats, 6 output formats, and 10 general multimodal skills, covering single-image, multi-image, and videos. - It moves beyond multiple-choice questions, offering diverse output formats like numbers, code, LATEX, phrases, free-form responses, and more. We developed 45 customized metrics to accurately evaluate these diverse outputs - It focuses on task diversity rather than repetitive examples, ensuring cost-efficient evaluation - It provides fine-grained capability reports across application type, input/output formats, and required skills. Website: tiger-ai-lab.github.io/MEGA-Bench/ Leaderboard: huggingface.co/spaces/TIGER-L… Dataset: huggingface.co/datasets/TIGER… Arxiv: arxiv.org/abs/2410.10563

@andrewchen We already applied on 09/05. We are @greennode23 an Nvidia Cloud Partner offering GPU Cloud but also a range of AI and Cybersecurity products and services. Look forward to hearing from you.

🚨 7 days left to apply a16z SPEEDRUN 004! 🚨 come work with me 1:1 -- I'll be coaching companies personally on everything from growth/marketing, launching, fundraising, hiring your initial team, and more. Would love to meet you. over the next week, we'll be investing $20M into founders and teams building in AI/infra, 3D tools, AR/VR, Web3 games, gamified apps and platforms, and much more. the way it works: - apply to the program (link below) - we interview you over the next week - offer within 48 hours - we wire soon after - then we hang in SF in Jan 2025!

@evanjconrad @sfcompute Congrats from @greennode23 team!! It’s great meeting you and SFC crew recently!

Howdy friends, we're gradually rolling out the new @sfcompute to the world. As we've been hinting at for a year, we built a compute market. Currently, it has the best H100 prices in the world, though we'd expect these to increase over time.

Selfie with the legendary @jasonlk. Recommend his book “From Impossible to Inevitable” a.co/d/6phhn9t

Greennode.ai is here at @SaaStrAnnual. As an Nvidia Cloud Partner with security guardrails (thanks to our CEO @redragonvn) we are serving AI startups and enterprises in the US with affordable&flexible term, large-scale Nvidia H100 GPU Cloud for fine-tuning workloads.

@aryaman2020 @greennode23 let's cook a welcome gift program for new PhD with our H100 SXM5 80G container instances 🤣

8x NVIDIA H100 80GB

@HaktanSuren @awscloud @runpod You can also try out greennode.ai, Nvidia Cloud Partner with H100 SXM5 GPU container instances.

Why @awscloud does not have 80 GB VRAM GPU? there is a huge gap either you get 24 GB or 320 G. Well, let me give it a shot to @runpod

@YiTayML @JeffDean @quocleix @lmthang @Diyi_Yang Thanks for sharing your insights and lessons learnt during the summit. Can’t imagine you tried 6 different GPU providers to train Reka models :D

Had a great weekend being part of the GenAI summit in Vietnam. 😄 Really enjoyed the impressive talks by @jeffdean, @quocleix @lmthang and @Diyi_Yang, along with the nice panel discussion that we had. 😄 Meanwhile, one major unexpected highlight of this trip for me personally was meeting @PoShenLoh and watching his talk on math, education, game theory and changing the world. This man is built different and cracked. Really such a breath of fresh air admist all the AI hype. 😄 The production quality of the summit was super serious and high quality 👌. The energy of the local crowd really makes me feel that Vietnam is really serious about AI. 🫡 This is my first time in Vietnam and it was amazing to be hosted by @lmthang. On the second day we visited some universities and left me a strong impression about the culture there. People there seem hungry for learning and science/math and that is amazing. Enjoyed the trip and the great hospitality and thoughtfulness by the hosts and organizers. 👍 Nit: Vietnam needs more ACs though🥵🥵🥵