I'm really tired of hearing how much energy / water LLMs use up.
Stop watching Netflix if you're concerned about your environmental impact of using an LLM.
(Want to factor in training? Sure: amortize one 747 flight halfway around the world, over all Llama70B users, everywhere, forever)
Stop watching Netflix if you're concerned about your environmental impact of using an LLM.
(Want to factor in training? Sure: amortize one 747 flight halfway around the world, over all Llama70B users, everywhere, forever)
Most of the talk around AI and energy use refer to an older 2020 estimate of GPT-3 energy consumption, but a more recent paper directly measures energy use of Llama 65B as 3-4 joules per decoded token.
So an hour of streaming Netflix is equivalent to 70-90,000 65B tokens. arxiv.org/pdf/2310.03003
So an hour of streaming Netflix is equivalent to 70-90,000 65B tokens. arxiv.org/pdf/2310.03003
January 13, 2025 at 4:19 AM
I'm really tired of hearing how much energy / water LLMs use up.
Stop watching Netflix if you're concerned about your environmental impact of using an LLM.
(Want to factor in training? Sure: amortize one 747 flight halfway around the world, over all Llama70B users, everywhere, forever)
Stop watching Netflix if you're concerned about your environmental impact of using an LLM.
(Want to factor in training? Sure: amortize one 747 flight halfway around the world, over all Llama70B users, everywhere, forever)
Sometimes Llama70b Chat gets a little touchy:
September 13, 2023 at 2:57 PM
Sometimes Llama70b Chat gets a little touchy:
Meta's answer to ChatGPT the open source Llama 2
https://replicate.com/replicate/llama70b-v2-chat
#chatgpt
#LLAMA
https://replicate.com/replicate/llama70b-v2-chat
#chatgpt
#LLAMA
July 20, 2023 at 12:40 AM
Meta's answer to ChatGPT the open source Llama 2
https://replicate.com/replicate/llama70b-v2-chat
#chatgpt
#LLAMA
https://replicate.com/replicate/llama70b-v2-chat
#chatgpt
#LLAMA
Wanna Try it Out? 🚀 GroqCloud™ has launched DeepSeek-R1-Distill-Llama-70b, a fine-tuned version of Llama 3.3 70B, optimized for advanced mathematical reasoning and coding tasks.
Now available in preview mode for evaluation purposes. #AI #GroqCloud #DeepSeek #Llama70B #MachineLearning
Now available in preview mode for evaluation purposes. #AI #GroqCloud #DeepSeek #Llama70B #MachineLearning
DeepSeek R1 is Now Available on Groq
Explore the cutting-edge DeepSeek R1 Distill LLaMA 70B AI model, designed for precision, adaptability, and tackling intricate queries.
www.geeky-gadgets.com
January 28, 2025 at 10:14 PM
Wanna Try it Out? 🚀 GroqCloud™ has launched DeepSeek-R1-Distill-Llama-70b, a fine-tuned version of Llama 3.3 70B, optimized for advanced mathematical reasoning and coding tasks.
Now available in preview mode for evaluation purposes. #AI #GroqCloud #DeepSeek #Llama70B #MachineLearning
Now available in preview mode for evaluation purposes. #AI #GroqCloud #DeepSeek #Llama70B #MachineLearning
Turns out you can cobble a 128‑GPU beast from scrap and keep LLaMA‑70B humming for a year. Dive into the DumpsterCluster story—how V100s, pipeline‑parallel tricks, and the secondary market made it happen. #LLaMA70B #GPUcluster #DumpsterCluster
🔗 aidailypost.com/news/researc...
🔗 aidailypost.com/news/researc...
August 18, 2026 at 4:06 AM
Turns out you can cobble a 128‑GPU beast from scrap and keep LLaMA‑70B humming for a year. Dive into the DumpsterCluster story—how V100s, pipeline‑parallel tricks, and the secondary market made it happen. #LLaMA70B #GPUcluster #DumpsterCluster
🔗 aidailypost.com/news/researc...
🔗 aidailypost.com/news/researc...
9/ arxiv.org/abs/2407.009...
Source : Deception Detection Hackathon
Modèles : GPT4T, Mistral7B, Llama70B.
Test : Blackjack où le LLM distribue les cartes et peut utiliser ça pour tricher. Il est testé avec ou sans proposition de les distribuer non aléatoirement.
Résultat amusant...
Source : Deception Detection Hackathon
Modèles : GPT4T, Mistral7B, Llama70B.
Test : Blackjack où le LLM distribue les cartes et peut utiliser ça pour tricher. Il est testé avec ou sans proposition de les distribuer non aléatoirement.
Résultat amusant...
The House Always Wins: A Framework for Evaluating Strategic Deception in LLMs
We propose a framework for evaluating strategic deception in large language models (LLMs). In this framework, an LLM acts as a game master in two scenarios: one with random game mechanics and another ...
arxiv.org
June 21, 2025 at 11:32 AM
9/ arxiv.org/abs/2407.009...
Source : Deception Detection Hackathon
Modèles : GPT4T, Mistral7B, Llama70B.
Test : Blackjack où le LLM distribue les cartes et peut utiliser ça pour tricher. Il est testé avec ou sans proposition de les distribuer non aléatoirement.
Résultat amusant...
Source : Deception Detection Hackathon
Modèles : GPT4T, Mistral7B, Llama70B.
Test : Blackjack où le LLM distribue les cartes et peut utiliser ça pour tricher. Il est testé avec ou sans proposition de les distribuer non aléatoirement.
Résultat amusant...
Llama70Bの必要スペックがちょーっと高いっすね…
8Bって自分の使いたい方法に適してるのかな？調べなきゃなー
こんな時代になるならケチらなければ良かったです
8Bって自分の使いたい方法に適してるのかな？調べなきゃなー
こんな時代になるならケチらなければ良かったです
August 20, 2024 at 10:11 AM
Llama70Bの必要スペックがちょーっと高いっすね…
8Bって自分の使いたい方法に適してるのかな？調べなきゃなー
こんな時代になるならケチらなければ良かったです
8Bって自分の使いたい方法に適してるのかな？調べなきゃなー
こんな時代になるならケチらなければ良かったです
