Nick Harvey
@NicksHarvey
Followers
455
Following
1K
Media
163
Statuses
2K
#opensource and #machinelearning enthusiast who's always hacking around on something.
Cary, NC
Joined February 2009
What is your P(B200)?
Curious about Blackwell? Good. Now go test it 1-click NVIDIA HGX B200 clusters. Live on Lambda. No waitlists. No commitment. Just compute. https://t.co/hdGRc8gdw0
1
2
6
Llama 3 70B Pre-training B200 = 1.9x H100 Llama 3 70B Inference B200 FP8 = 1.8x H200 FP8 B200 FP4 = 2.4x H200 FP8 Llama 3.1 405B Inference B200 FP4 = 2.7x H200 FP8
Still looking for Blackwell clusters? NVIDIA HGX B200 clusters are live on Lambda Cloud. 1-click. Multi-node. Ready when you are. Try them →
0
2
12
AI agents aren’t the future. They’re here and you can build with them now. Next week in SF, we’re hosting the AI Agent Masterclass, a hands-on workshop for AI/ML engineers building real-world multi-agent systems with the @mlopscommunity. 🗓 April 24 📍 Microsoft Reactor, SF
0
10
28
Neural networks are going to replace most of the software in the world. https://t.co/uD4smGkqep
2
9
53
Today, we are announcing general availability of multi-node NVIDIA HGX B200-accelerated clusters on-demand through Lambda 1-Click Clusters. Time for AI teams to start innovating faster with the latest and greatest NVIDIA GPUs, without the overhead of long-term contracts or
8
12
50
Opening up Disney+ to watch It's always sunny just feels wrong
0
0
0
It's been released just a few days ago and already more than 500 derivative models of @deepseek_ai have been created all over the world on @huggingface with 2.5 million downloads (5x the original weights). The power of decentralized open-source AI!
51
223
1K
I love to watch open source continue to be a disruptive force for innovation! @deepseek_ai is amazing!
9
13
72
We’re excited to partner with @Forbes and @MeritechCapital to launch #ForbesAI50 2025! The list spotlights the most promising privately-held AI companies, with alumni including @OpenAI, @scale_AI, and @midjourney Should your company make the list? https://t.co/A7YgNkVC7D
10
9
27
9/ I've collected examples of effective PRDs for common features (login, search, etc.) in PRD World to help PMs get started. https://t.co/WpENJf2Bj3 Most importantly: Your PRD is a conversation starter, not a specification document.
prodmgmt.world
25+ plug-and-play Product Requirements Documents for common product features. Like Bootstrap but for PRDs.
0
1
0
Looking for AI products to try over the break? From the @a16z team + friends of the firm ⬇️
122
709
5K
Still think you need a GPU farm to be your own director 🎬? Mochi says otherwise! Fine-tune cinematic AI with just a single NVIDIA GPU. Here’s how:
How to Train Your Mochi: Introducing LoRA fine-tuning. Customize Mochi on a single GPU with just a few videos. Create any effect or create consistent characters. Make Mochi 1 truly yours.
1
3
12
How to Train Your Mochi: Introducing LoRA fine-tuning. Customize Mochi on a single GPU with just a few videos. Create any effect or create consistent characters. Make Mochi 1 truly yours.
92
82
591
Checkout this guide from @LambdaAPI about serving vLLM on GH200 Superchip!
Just Launched 🎉 - NVIDIA GH200 Superchip on Lambda On-Demand! Spin up your own instance today and enjoy up to 2x faster time-to-first-token (TTFT) with models like Llama3 70B. Try it out today 👉 https://t.co/fAt4A0lpmA
0
3
17
Sam Altman's spent years refining his note-taking process. Here's exactly what he does.
649
2K
20K
Today we are launching LoRA Land: 25 fine-tuned #mistral 7b #llm that outperform #gpt4 on task-specific applications ranging from sentiment detection to question answering. Try it: https://t.co/SE9Cykh7C5
#llm #opensource #openai Models are on @huggingface
https://t.co/gN4QkT5Hxm
15
55
236
@SnorkelAI 7B model is SOTA on AlpacaEval 2.0! - Method: RM rejection sampling + DPO - Tl/dr: Again, small model + better data = win Leaderboards only mean so much. The bigger idea here is programmatic alignment- custom-steerable LLMs w/out manual pref labeling w/ @SnorkelAI !
2
12
54