今日は学校の起業家工房Fab室にあるAIのワークステーションで、H100を使ってSV3Dを動かしてみました。デフォルトの設定しかまだ試せていませんが、背景のある画像でも背景を切り取ってうまく3D動画を生成してくれます。
←入力画像 出力動画→

Bugün h100 olduğumu öğrendim

H100のライト点いた姿がかっこよすぎ
雪煙上げて走る実車見たくなってきた

老雷获利了结英伟达，去年2月建仓至今2.5倍

1、当下能够继续让世界前行的只有AI（纳指的决定因素）

2、英伟达的财报，增长是降速的，原因是H100现在全世界该买的大客户基本上都已经买了，而小客户商业化还没完成，持观望态度

3、三季度大概率向下走，到770-850再上车

4、最终英伟达还会上1300

Elon Musk’s xAI just raised $6B to develop Grok AI and build a supercomputer with 100,000 $NVDA H100 chips.

Here was Jensen’s reaction (probably):

𝟱. Pre-Configured and Custom VMs:

InfraX offers pre-configured virtual machines for quick deployment tailored to specific use cases.

𝟲. H100 Access:

Users can access the high-performance H100 GPU for AI, data analytics, and computationally intensive workloads.

$AIGPU Proof of Compute Thread 📈

In this thread we will dive deep into the performance & optimization metrics of 9 RTX 4090’s achieving similar KPI’s as 1 H100 for half the cost.

Continue reading as we disclose the proof of compute & KPI’s⚡️

Breaking my silence - this banger also deserves a #1 on h100, Spotify and 1 billion

infraX | $INFRA Alright, let's dive into the fascinating features of infraX | $INFRA!

5️⃣ 𝗙𝗘𝗔𝗧𝗨𝗥𝗘𝗦 𝗢𝗙 𝗜𝗻𝗳𝗿𝗮𝗫:

- GPU Lending
- GPU Rental
- AI Node Rental
- Pay as you go Rental
- Custom VM
- Staking
- H100 access

Let's take a close look at how they function…

They're heeeeeeeere!

The $CETI team is busy setting up racks and getting the machines ready because the h100's are HERE.

Who is ready for the next evolution of decentralized AI?  Let's see those comments ⬇️

#ai #crypto #blockchain

Revolutionize your network management with AmpCon™! Effortlessly deploy and automate PicOS® switches remotely, while seamlessly managing your #InfiniBand H100 networks through a unified platform.
Learn More: bit.ly/4bFoAPa

#Networking #Tech #DataCenter

People underestimate how inefficient things are:

Tensorplex beat Meta's llama-model with a cluster 50x smaller on a 6 month old Bittensor market.

That's equivalent to five A100 -> H100 hardware upgrades in half a year.

and the difference came from allocation not compute.

旭川駅を発車するH100-86根室線ラッピング

Other 𝗦𝗲𝗿𝘃𝗶𝗰𝗲𝘀 include:

- Access to AI functionalities through API endpoints.
- Staking $INFRA tokens to earn rewards in Ethereum.
- Pre-configured or custom virtual machines for specific needs.
- Instant access to H100 computing power.

🎶Fine-tuning on Wikipedia Datasets🎶

I extract a dataset from wikipedia to fine-tuning Llama 3 on the Irish language.

Thanks to Daniel Han and Unsloth AI - I was able to fine-tune Llama 3 and get through 50k+ rows of data in just over an hour on an H100.

TIMESTAMPS:
0:00

夕暮れ時を行く、元上り普通終列車スジの4532Ｄ…
普通列車の終電が延びて約５年。夜については便利になりました。

そして、今日は、石北線ラッピングのデクモでした。
#H１００
#石北本線 　 #石北線
#みんなの石北本線
#上川町

🔎 xAI confirms raise of $6B at $24B post-money valuation from Valor, a16z, Sequoia, Fidelity, others

• 100K NVIDIA H100 “gigafactory of compute” planned for the fall of 2025, Elon told investors.

•4x larger than Meta’s current supercomputer (but not necessarily their 2025

✅💜
#LOST
#RM

I trained GPT-2 (124M) using Andrej Karpathy's llm.c in just 43 minutes with 8 x H100 GPUs.

This is 2.1x faster than the 90 minutes it took with 8 x A100 GPUs. Currently, the cost of renting an H100 GPU is around $2.50/hr (under 1-year commitment), which reduces the training cost for