Llama 3 is Here! #ai #llms #llama3

Elvis Saravia · Intermediate ·📰 AI News & Updates ·2y ago

Key Takeaways

The video discusses the release of Llama 3, a new AI model by Meta, which includes 8B and 70B pretrained and instruction-tuned models, and provides an overview of its technical details and performance.

Full Transcript

hey everyone Elvis here today meta decided to announce latry this is exciting news there is a 8B and a 70b pre-train and instruction tune mold so we have different Ms of different sizes there is some performance also that they reported very impressive performance here uh this model 8B of performs Gemma 7B and michell 7B very strong pain models as well so this is great news for developers there are some technical details this is a decoder only Transformer it's a 128 K Tokyo 8K tokens sequence length for this is your context window and also it was pre-trained on 15 trillion tokens that's amazing and post training would include the standard supervised SP tuni rejection sampling Po and much more and there's also a 400 billion parameter mode that is still training and coming soon you can see how impressive those results are if you want to know more about this I have recorded a longer overview with some First Impressions and thoughts over on my YouTube you will see the link in the descript

Original Description

Meta just released Llama 3 which includes 8B and 70B pretrained and instruction-tuned models. Llama 3 announcement: https://llama.meta.com/llama3/ Blog: https://ai.meta.com/blog/meta-llama-3/ Model card: https://github.com/meta-llama/llama3/blob/main/MODEL_CARD.md
Watch on YouTube ↗ (saves to browser)
Sign in to unlock AI tutor explanation · ⚡30

Playlist

Uploads from Elvis Saravia · Elvis Saravia · 39 of 60

← Previous Next →
1 101 ways to solve search (by Pratik Bhavsar)
101 ways to solve search (by Pratik Bhavsar)
Elvis Saravia
 2 TLDR Generation of Scientific Documents | ML Interview #1 with Isabel Cachola
TLDR Generation of Scientific Documents | ML Interview #1 with Isabel Cachola
Elvis Saravia
 3 Sentiment Analysis: Key Milestones, Challenges and New Directions
Sentiment Analysis: Key Milestones, Challenges and New Directions
Elvis Saravia
 4 Discriminative Adversarial Search for Abstractive Summarization (by Thomas Scialom)
Discriminative Adversarial Search for Abstractive Summarization (by Thomas Scialom)
Elvis Saravia
 5 Question Understanding: COVID-Q: 1,600+ Questions about COVID-19
Question Understanding: COVID-Q: 1,600+ Questions about COVID-19
Elvis Saravia
 6 Getting Started with NLP
Getting Started with NLP
Elvis Saravia
 7 Building tools and frameworks for large-scale social media mining (by Dr. Juan M. Banda)
Building tools and frameworks for large-scale social media mining (by Dr. Juan M. Banda)
Elvis Saravia
 8 TextAttack: A Framework for Data Augmentation and Adversarial Training in NLP
TextAttack: A Framework for Data Augmentation and Adversarial Training in NLP
Elvis Saravia
 9 Dive into Deep Learning (Study Group): Introduction to Deep Learning | Session 1
Dive into Deep Learning (Study Group): Introduction to Deep Learning | Session 1
Elvis Saravia
 10 Dive into Deep Learning (Study Group): Multilayer Perceptrons | Session 4
Dive into Deep Learning (Study Group): Multilayer Perceptrons | Session 4
Elvis Saravia
 11 How I read and annotate ML papers
How I read and annotate ML papers
Elvis Saravia
 12 Keep Learning ML (Session 1) | DSV, CompLex, Modern tools for emotions
Keep Learning ML (Session 1) | DSV, CompLex, Modern tools for emotions
Elvis Saravia
 13 Dive into Deep Learning (Study Group): Preliminaries | Session 2
Dive into Deep Learning (Study Group): Preliminaries | Session 2
Elvis Saravia
 14 Keep Learning ML #2 | Language-conditioned policy learning, Effective ML Testing, EagerPy
Keep Learning ML #2 | Language-conditioned policy learning, Effective ML Testing, EagerPy
Elvis Saravia
 15 Dive into Deep Learning (Study Group): Linear Neural Networks | Session 3
Dive into Deep Learning (Study Group): Linear Neural Networks | Session 3
Elvis Saravia
 16 Dive into Deep Learning (Study Group): Multilayer Perceptrons | Session 4
Dive into Deep Learning (Study Group): Multilayer Perceptrons | Session 4
Elvis Saravia
 17 Keep Learning ML #3 | Contrastively Trained Structured World Models
Keep Learning ML #3 | Contrastively Trained Structured World Models
Elvis Saravia
 18 Dive into Deep Learning (Study Group): Deep Learning Computation with PyTorch | Session 5
Dive into Deep Learning (Study Group): Deep Learning Computation with PyTorch | Session 5
Elvis Saravia
 19 Dive into Deep Learning (Study Group): Convolutional Neural Networks | Session 6
Dive into Deep Learning (Study Group): Convolutional Neural Networks | Session 6
Elvis Saravia
 20 Dive into Deep Learning (Study Group): Modern CNNs | Session 7
Dive into Deep Learning (Study Group): Modern CNNs | Session 7
Elvis Saravia
 21 101 ways to solve neural search with Jina
101 ways to solve neural search with Jina
Elvis Saravia
 22 (Hopefully-Reusable) Life Lessons for PhD Students in NLP
(Hopefully-Reusable) Life Lessons for PhD Students in NLP
Elvis Saravia
 23 How to save the world and forward your career in 5 easy steps | Women in NLP Talks
How to save the world and forward your career in 5 easy steps | Women in NLP Talks
Elvis Saravia
 24 Prompt Engineering Overview
Prompt Engineering Overview
Elvis Saravia
 25 Getting Started with the OpenAI Playground
Getting Started with the OpenAI Playground
Elvis Saravia
 26 LM-Guided Chain of Thought
LM-Guided Chain of Thought
Elvis Saravia
 27 Elements of a Prompt
Elements of a Prompt
Elvis Saravia
 28 Reasoning with Intermediate Revision and Search with LLMs #chatgpt #ai #llms #science #programming
Reasoning with Intermediate Revision and Search with LLMs #chatgpt #ai #llms #science #programming
Elvis Saravia
 29 General Tips for Designing Prompts
General Tips for Designing Prompts
Elvis Saravia
 30 Efficient Infinite Context Transformers #ai #machinelearning #research #llms #science
Efficient Infinite Context Transformers #ai #machinelearning #research #llms #science
Elvis Saravia
 31 Best Practices and Lessons Learned on Synthetic Data for Language Models #ai #machinelearning #genai
Best Practices and Lessons Learned on Synthetic Data for Language Models #ai #machinelearning #genai
Elvis Saravia
 32 Reducing Hallucinations in Structured Outputs via RAG #chatgpt #ai #llms #programming
Reducing Hallucinations in Structured Outputs via RAG #chatgpt #ai #llms #programming
Elvis Saravia
 33 Basic Prompt Examples for LLMs
Basic Prompt Examples for LLMs
Elvis Saravia
 34 LLM In Context Recall is Prompt Dependent #llms #ai #chatgpt #machinelearning
LLM In Context Recall is Prompt Dependent #llms #ai #chatgpt #machinelearning
Elvis Saravia
 35 Zero-shot Prompting Explained
Zero-shot Prompting Explained
Elvis Saravia
 36 RAG Faithfulness #llms #ai #gpt4
RAG Faithfulness #llms #ai #gpt4
Elvis Saravia
 37 Understanding LLM Settings
Understanding LLM Settings
Elvis Saravia
 38 Llama 3 is here! | First impressions and thoughts
Llama 3 is here! | First impressions and thoughts
Elvis Saravia
 ▶ Llama 3 is Here! #ai #llms #llama3
Llama 3 is Here! #ai #llms #llama3
Elvis Saravia
 40 Microsoft introduces Phi-3 | The most capable small language model?
Microsoft introduces Phi-3 | The most capable small language model?
Elvis Saravia
 41 Microsoft introduces Phi-3! #ai #llms #microsoft
Microsoft introduces Phi-3! #ai #llms #microsoft
Elvis Saravia
 42 Make Your LLM Fully Utilize the Context #ai #llms #machinelearning
Make Your LLM Fully Utilize the Context #ai #llms #machinelearning
Elvis Saravia
 43 When to Retrieve? #ai #llms #machinelearning
When to Retrieve? #ai #llms #machinelearning
Elvis Saravia
 44 Training an LLM to effectively use information retrieval
Training an LLM to effectively use information retrieval
Elvis Saravia
 45 State-of-the-art open-source LLM judges #ai #machinelearning #gpt4
State-of-the-art open-source LLM judges #ai #machinelearning #gpt4
Elvis Saravia
 46 Better and Faster LLMs via Multi-token Prediction
Better and Faster LLMs via Multi-token Prediction
Elvis Saravia
 47 AlphaMath Almost Zero #ai #science #machinelearning
AlphaMath Almost Zero #ai #science #machinelearning
Elvis Saravia
 48 SWE-Agent | An LLM-based Software Engineering Agent
SWE-Agent | An LLM-based Software Engineering Agent
Elvis Saravia
 49 [LLM NEWS] AlphaFold 3, xLSTM, OpenAI's Model Spec, DeepSeek-V2, OpenDevin CodeAct 1.0
[LLM NEWS] AlphaFold 3, xLSTM, OpenAI's Model Spec, DeepSeek-V2, OpenDevin CodeAct 1.0
Elvis Saravia
 50 LLM-powered tool for web scraping #ai #chatgpt #engineering
LLM-powered tool for web scraping #ai #chatgpt #engineering
Elvis Saravia
 51 Learn about LLMs in this NEW course #ai #chatgpt #engineering
Learn about LLMs in this NEW course #ai #chatgpt #engineering
Elvis Saravia
 52 [LLM NEWS] KANs, Gemma 10M Context, OpenAI Updates?, Automatic Prompt Engineering, Tokenizer Arena
[LLM NEWS] KANs, Gemma 10M Context, OpenAI Updates?, Automatic Prompt Engineering, Tokenizer Arena
Elvis Saravia
 53 [LLM News] GPT4-o, Project Astra, Veo, Copilot+ PCs, Gemini 1.5 Flash, Chameleon
[LLM News] GPT4-o, Project Astra, Veo, Copilot+ PCs, Gemini 1.5 Flash, Chameleon
Elvis Saravia
 54 Enhancing Answer Selection in LLMs #ai #machinelearning #engineering
Enhancing Answer Selection in LLMs #ai #machinelearning #engineering
Elvis Saravia
 55 On exploring LLMs #ai #promptengineering #chatgpt
On exploring LLMs #ai #promptengineering #chatgpt
Elvis Saravia
 56 Transformers Can Do Arithmetic with the Right Embeddings #ai #machinelearning #engineering
Transformers Can Do Arithmetic with the Right Embeddings #ai #machinelearning #engineering
Elvis Saravia
 57 [LLM News] xAI Series B, Codestral, LLM Guide, AutoGen Course, Symbolic Chain-of-Thought
[LLM News] xAI Series B, Codestral, LLM Guide, AutoGen Course, Symbolic Chain-of-Thought
Elvis Saravia
 58 PR-Agent #ai #gpt4 #software
PR-Agent #ai #gpt4 #software
Elvis Saravia
 59 Extracting features from Claude 3 Sonnet
Extracting features from Claude 3 Sonnet
Elvis Saravia
 60 Has prompt engineering been solved?
Has prompt engineering been solved?
Elvis Saravia

Llama 3 is a new AI model released by Meta, which includes 8B and 70B pretrained and instruction-tuned models. The model is a decoder-only Transformer with a sequence length of 128K and was pre-trained on 15 trillion tokens. The video provides an overview of the model's technical details and performance.

Key Takeaways
  1. Read the Llama 3 announcement
  2. Explore the model card on GitHub
  3. Watch the longer overview video on YouTube
  4. Compare Llama 3 with other LLMs
  5. Design instruction-tuned models using Llama 3
💡 Llama 3's impressive performance is due to its large-scale pretraining and instruction tuning, making it a strong competitor to other LLMs.

Related AI Lessons

⚡
The AI Moat Paradox: The Better Models Become, the Less Models Matter
The AI moat paradox suggests that as AI models improve, their importance may decrease, and understanding this concept is crucial for AI professionals and businesses.
Medium · AI
 ⚡
170,927 AI Papers Reveal the Biggest Research Shifts of the First Half of 2026
Discover the biggest AI research shifts of 2026 based on 170,927 papers, and learn how to apply these trends to your work
Medium · Machine Learning
 ⚡
170,927 AI Papers Reveal the Biggest Research Shifts of the First Half of 2026
Discover the major research shifts in AI from 170,927 papers published in the first half of 2026, and learn how to analyze trends in AI research
Medium · Data Science
 ⚡
[PoV] When Everyone Is Smart, No One Is
In a world where AI makes everyone smart, the value of intelligence decreases, and new challenges arise
Medium · AI
Up next
‘ENOUGH IS ENOUGH’: Lebanon is STANDING UP to Iran, expert says
Fox Business
Watch →