#nanogpt — Public Fediverse posts

Onde investi: Nano GPT + DeepSec

Quer saber onde eu tô colocando grana pra usar modelos de IA hoje? 🤔

- O que eu investi:
• US$ 8 no Nano GPT
• US$ 5 no DeepSec

- Por que eu tô usando:
• É o que estou testando agora ✅
• Prático e direto para uso diário 🚀

Isso resume o setup que estou testando — foco em custo-benefício e performance. Quer um comparativo entre Nano GPT e DeepSec? Comenta aí! 👇

Curta 👍 | Salve 💾...

#NanoGPT #DeepSec #InteligenciaArtificial #IA #Modelos #MorningCrypto

Modelos baratos e grátis pra testar IA

Quer testar modelos de IA sem gastar uma fortuna? 🤯

- NanoGPT: vários modelos com preço acessível — perfeito pra começar e rodar localmente ⚙️
- NVIDIA: oferece diversos modelos gratuitos e agentivos pra você testar na prática 🚀
- Ideal pra prototipar, aprender e fazer experiências sem depender de APIs pagas 💡

Quer que eu traga um mini-tutorial usando algum desses modelos? Comenta qual você prefere! 👇

Salva esse...

#NanoGPT #NVIDIA #IA #MorningCrypto

I use NanoGPT for API access to Qwen/GLM/Kimi/MiniMax, here's my link if you want 5% off

https://nano-gpt.com/subscription/k7AXDwBb

#ai #nanogpt

I use NanoGPT for API access to Qwen/GLM/Kimi/MiniMax, here's my link if you want 5% off

https://nano-gpt.com/subscription/k7AXDwBb

#ai #nanogpt

🎉 Behold, the #NanoGPT Slowrun: a marvel of 10x data efficiency that nobody asked for, achieved by throwing infinite compute at the problem like a toddler with a tantrum. 🤦‍♂️ Who knew the solution to our data bottleneck woes was to just ignore them and hope they go away with more compute power? 🚀 True #innovation at its finest!
https://qlabs.sh/10x #Slowrun #DataEfficiency #ComputePower #TechHumor #HackerNews #ngated

🎉 Behold, the #NanoGPT Slowrun: a marvel of 10x data efficiency that nobody asked for, achieved by throwing infinite compute at the problem like a toddler with a tantrum. 🤦‍♂️ Who knew the solution to our data bottleneck woes was to just ignore them and hope they go away with more compute power? 🚀 True #innovation at its finest!
https://qlabs.sh/10x #Slowrun #DataEfficiency #ComputePower #TechHumor #HackerNews #ngated

NanoGPT Slowrun: 10x Data Efficiency with Infinite Compute

https://qlabs.sh/10x

#HackerNews #NanoGPT #Slowrun #DataEfficiency #InfiniteCompute #AI

NanoGPT Slowrun: 10x Data Efficiency with Infinite Compute

https://qlabs.sh/10x

#HackerNews #NanoGPT #Slowrun #DataEfficiency #InfiniteCompute #AI

Ah, the 2026 #visionaries have graced us with #NanoGPT Slowrun! 🚀 An open letter to the future, where our data droughts are hilariously overshadowed by compute floods. It's like trying to fill a swimming pool with a teaspoon and wondering why you're not Olympic-ready yet. 🏊‍♂️🤡
https://qlabs.sh/slowrun #Slowrun #DataDroughts #ComputeFloods #FutureTech #HackerNews #ngated

Ah, the 2026 #visionaries have graced us with #NanoGPT Slowrun! 🚀 An open letter to the future, where our data droughts are hilariously overshadowed by compute floods. It's like trying to fill a swimming pool with a teaspoon and wondering why you're not Olympic-ready yet. 🏊‍♂️🤡
https://qlabs.sh/slowrun #Slowrun #DataDroughts #ComputeFloods #FutureTech #HackerNews #ngated

NanoGPT Slowrun: Language Modeling with Limited Data, Infinite Compute

https://qlabs.sh/slowrun

#HackerNews #NanoGPT #Slowrun #LanguageModeling #LimitedData #InfiniteCompute

NanoGPT Slowrun: Language Modeling with Limited Data, Infinite Compute

https://qlabs.sh/slowrun

#HackerNews #NanoGPT #Slowrun #LanguageModeling #LimitedData #InfiniteCompute

Today I tried out #AMD #Instinct #MI300a for my existing Deep Learning pipeline. Good news: It worked out of the box. Bad news: For some reason it could not beat my local #Nvidia #1080ti...
After trying all sorts of #ROCM installation methods via prebuild wheels, #apptainer images etc I tried #nanogpt by @karpathy and sure enought: The gpt code ran approx 2x faster than on a #a100 ... I hope that this is due to my programming skills. Not AMD prefering #transformers over #CNNs ...

That is a nice visualization to get an intuition of what goes on under the hood of LLMs:

https://bbycroft.net/llm

Since it is build on nano-gpt, tokenization is splitting by character, but the other phases should be pretty realistic.

"Let's build GPT: from scratch, in code, spelled out" (from Karpathy's, the author of nano-gpt) could be good additional material to understand the details:

* https://github.com/karpathy/nanoGPT
* Video: https://www.youtube.com/watch?v=kCc8FmEb1nY

#llm #nanogpt #ai

That is a nice visualization to get an intuition of what goes on under the hood of LLMs:

https://bbycroft.net/llm

Since it is build on nano-gpt, tokenization is splitting by character, but the other phases should be pretty realistic.

"Let's build GPT: from scratch, in code, spelled out" (from Karpathy's, the author of nano-gpt) could be good additional material to understand the details:

* https://github.com/karpathy/nanoGPT
* Video: https://www.youtube.com/watch?v=kCc8FmEb1nY

#llm #nanogpt #ai

1️⃣ Wie funktioniert eigentlich ChatGPT?
Der Titel sagt eigentlich alles: Eine Erklärung der Funktionsweise von #NanoGPT, Andrej Karpathys kleiner Schwester von ChatGPT. Und zwar so, dass man keine Informatik studiert haben muss.

Vermutlich das deutsche Standardwerk, wenn es um das Verständnis von grossen Sprachmodellen wie #ChatGPT geht. Es hat auch den Grundstein für die vielen daraus entstandenen Vorträge gelegt.
#LLM
📰 https://dnip.ch/2023/01/30/wie-funktioniert-eigentlich-chatgpt/
Mehr zu #KI: marcel-waldvogel.ch/ki

1️⃣ Wie funktioniert eigentlich ChatGPT?
Der Titel sagt eigentlich alles: Eine Erklärung der Funktionsweise von #NanoGPT, Andrej Karpathys kleiner Schwester von ChatGPT. Und zwar so, dass man keine Informatik studiert haben muss.

Vermutlich das deutsche Standardwerk, wenn es um das Verständnis von grossen Sprachmodellen wie #ChatGPT geht. Es hat auch den Grundstein für die vielen daraus entstandenen Vorträge gelegt.
#LLM
📰 https://dnip.ch/2023/01/30/wie-funktioniert-eigentlich-chatgpt/
Mehr zu #KI: marcel-waldvogel.ch/ki

Aaaaand the winner iiiis: 😊

1️⃣ How does ChatGPT work, actually? (2023-01)
A relatively easy-to-grasp explanation how #ChatGPT works, yet still accurate on a high level. Starting with an explanation of ChatGPT's little cousin, #NanoGPT, and then describing the differences.

Essentially it is a dual translation of #AndrejKarpathy's #NanoGPT video: Format from video to text; target audience from programmer to general public.

Enjoy!
https://netfuture.ch/2023/01/how-does-chatgpt-work-actually/

Aaaaand the winner iiiis: 😊

1️⃣ How does ChatGPT work, actually? (2023-01)
A relatively easy-to-grasp explanation how #ChatGPT works, yet still accurate on a high level. Starting with an explanation of ChatGPT's little cousin, #NanoGPT, and then describing the differences.

Essentially it is a dual translation of #AndrejKarpathy's #NanoGPT video: Format from video to text; target audience from programmer to general public.

Enjoy!
https://netfuture.ch/2023/01/how-does-chatgpt-work-actually/

Cerebras unveils gigaGPT, a simplified large-scale model implementation

https://stackdiary.com/gigagpt/

#nanoGPT #AI #LLM #PyTorch

I got my first bit of code merged into an upstream #PyTorch #Lightning project yesterday. It's for a Lightning variant of #minGPT / #nanoGPT. No functional changes, but it's still exciting to contribute something.

https://github.com/Lightning-Universe/lightning-GPT/pull/39

I got my first bit of code merged into an upstream #PyTorch #Lightning project yesterday. It's for a Lightning variant of #minGPT / #nanoGPT. No functional changes, but it's still exciting to contribute something.

https://github.com/Lightning-Universe/lightning-GPT/pull/39

Ich wollte schon immer mal wissen, ob ich die Resultate von #Karpathy mit #nanoGPT und #Shakespeare auch auf deutsch reproduzieren könne. Gestern Abend habe ich den Versuch in die Tat umgesetzt und der Rechner hat die Nacht durch #Goethe gebüffelt, bis er seinen Stil nachahmen konnte. (Es ist ein sehr kleines Modell, deshalb ist "nachahmen" sehr grosszügig zu interpretieren…)

#ChatGPT am Beispiel von nanoGPT erklärt:
https://dnip.ch/2023/01/30/wie-funktioniert-eigentlich-chatgpt/

Ich wollte schon immer mal wissen, ob ich die Resultate von #Karpathy mit #nanoGPT und #Shakespeare auch auf deutsch reproduzieren könne. Gestern Abend habe ich den Versuch in die Tat umgesetzt und der Rechner hat die Nacht durch #Goethe gebüffelt, bis er seinen Stil nachahmen konnte. (Es ist ein sehr kleines Modell, deshalb ist "nachahmen" sehr grosszügig zu interpretieren…)

#ChatGPT am Beispiel von nanoGPT erklärt:
https://dnip.ch/2023/01/30/wie-funktioniert-eigentlich-chatgpt/

I heard you've always wanted to be a trainer. Follow your dream, go train #nanoGPT.

https://github.com/karpathy/nanoGPT

#nanoGPT reproduces #GPT-2 (124M) on OpenWebText, running on a single 8XA100 40GB node in 38 hours of training.

https://github.com/karpathy/nanoGPT

#chatGPT #machinelearning #artificialintelligence

#nanoGPT reproduces #GPT-2 (124M) on OpenWebText, running on a single 8XA100 40GB node in 38 hours of training.

https://github.com/karpathy/nanoGPT

#chatGPT #machinelearning #artificialintelligence