Your own Copilot in VSCode lightning-fast thanks to Groq and Llama3

In this thread 🧵, I'll guide you through connecting Groq Inc with VSCode, unlocking the full potential of Llama 3 (@AIatMeta) as your Copilot👇

We need a new Quantization algo if we want to get more juices from these models trained on huge data, else quantization is degrading the great amount of information hidden in the full-precision tensors.

'LLAMA3 still suffers non-negligent degradation in these scenarios,

LLaMA3 is faaast 💨 shipping tomorrow!🤞

Fully local Video Summaries with llama3 on ollama. Breaks up longer videos into chunks and provides a summary. Lots of questions on the last vid so here i am fixing errors real time.

Open-source: git.new/local-vid-summ…

Passing the vibe check🤣😭✅
#qroqai Groq Inc #llama3

LobeChat + Llama3/Ph3 用起来真的爽

Okay, Yeah Llama3 is working locally. A bit slow. but it can write a valid Hello world in #golang

Microsoft Phi-3 Beats Llama3 8B: Did it Pass the Coding Test?

🚀 Phi-3 mini (3.8B)
🏆 Medium (14B) outperforms GPT-3.5
🏅 Small (7B) tops Llama 3
📱 Runs on Android, iPhones
🧬 Mini trained on 3.3T tokens

Subscribe: youtube.com/@MervinPraison
YT: youtube.com/watch?v=lavCDJ…

Super fast⚡️Video Summaries⚡️powered by Llama3 on Groq Inc. Breaks up longer videos into chunks to provide a full summary.

Here we are summarizing videos by Harrison Chase and Andrej Karpathy, wip but will keep improving.

Open-source: git.new/video-summary

プロンプト作成に、ローカルにインストールしたLLMでやってみた。
Ollama + Open WebUI に、SDプロンプト作成用のモデルの　llama3_ifai_sd_prompt_mkr_q4km で、プロンプト作成。
#VIRDEL_12
#AIウッドストック音楽祭

Nvidia has published a competitive llama3-70b question answering and retrieval-augumented generation (RAG) fine tune - ChatQA-1.5

Benches look good, but missing that 'llama 3' in the name as per the requirements of meta 😃

huggingface.co/nvidia/ChatQA-…

The LLaMA3 speed is ridiculous 🚀

Comparing Llama3 and ChatGPT Responses.

First I do Llama3 vs GPT3.5, then Llama3 vs GPT4.

Notice how SLOW GPT-4 is, oof!

Llama3 seems more comprehensive.

Looking forward to digging in more now that I have the right tools.

everyone: oh wow look at all the tokens llama3 supports

the tokens:

takribi bir hesaba göre llama3 modeli metanın tüm kaynaklarıyla (yaklaşık 25.000 adet h100 cluster) 2.6 milyon dolar civarına ve sadece üç günde eğitildi

Phi3 is better than Llama3. and we has deployed it on M4N(AX650N inside) #Phi3 #Llama3 #Sipeed

Did you know you can run Llama3 in YOUR browser?

This new range of small language models are revolutionary and compete on GPT3/4 on RAG, reasoning, data extraction, categorization etc.

And the best part is that you can try by yourself online, here's how:

Using these tools:

🤯 A GUI for fine-tuning the Llama-3 models on a free T4 GPU!

👉 𝐋𝐋𝐚𝐌𝐀 𝐅𝐚𝐜𝐭𝐨𝐫𝐲
🚀 Two Llama-3-fine-tunes using LLaMA Factory are already available at Hugging Face Hub -- check Llama3-8B-Chinese-Chat and Llama3-Chinese for details.

I found a new tool calling champion

Llama3 70b on Groq Inc

Challenge: given user query, extract financial quarters and years.

Example: 'How did revenue change between Q4 2023 and year before that?'

The 70b model:
• passed the task
• was very fast
• had best pricing

I

コントローラーが家庭用になっちゃったけど、やってみました。
プロンプトはALTで。llama3のプロンプト作成モデルに日本語で投入、出来たプロンプトをDalle3で生成しています
#dalle3