Clarifai provides a full stack AI platform helping developers and teams to quickly and collaboratively get vision, language and audio AI into production.
Missed our webinar on Digital Asset Management and Personalization with AI? Don’t worry! Catch the recording here and learn how to drive more leads by using AI-based content organization and personalization.
#AI
#ContentOrganization
#DigitalAssetManagement
Google has released 7 free courses that helps you learn AI from scratch!
- Generative AI & LLMs fundamentals
- Diffusion Models
- Transformer and BERT Models
- Attention Mechanism
- Image Captioning Models
Thread:🧵 ⬇️
Stanford CS224N: NLP with Deep Learning course from Stanford!
The course covers topics ranging from basics of NLP to advanced concepts like Attention and Transformer networks.
Inarguably, one of the best NLP courses on the web!
Check this out:
LLMs Practical Guide!🔥
Here is the evolution Tree of Large Language Models from 2018 to 2023!
This Includes guide and resources to GPT-style Language Models:
→ GPT
→ GPT-2
→ GPT-3
→ GPT-4
→ OPT
→ PaLM
→ BLOOM
→ LaMDA
→ LLaMA 2
→ BloombergGPT
→ GPT-NeoX-20B
Also to
Google has released 7 free courses on Generative AI.
This includes:
- Generative AI & LLMs fundamentals
- Responsible AI
- Diffusion Models
- Transformer and BERT Models
- Attention Mechanism
- Image Captioning Models
Read more: ⬇️
Mathematics for Machine Learning from University of Berkeley.
Learn:
• Linear Algebra
• Calculus and Optimization
• Probabilities
100% Free.
Download it here:
Mathematics for Machine Learning from University of Berkeley!
100% Free.
Learn:
• Linear Algebra
• Calculus and Optimization
• Probabilities
Download it here:
Arguably, the best Linear Algebra course out there for Machine Learning.
Taught by MIT's Professor Gilbert Strang.
And it's FREE!
Build a strong foundation in mathematics for machine learning:
Deep Learning course from Carnegie Mellon University.
One of the best you'll find out there to learn Deep Learning fundamentals.
100% FREE
Check this out:
“A Survey of Large Language Models” is the most comprehensive survey paper on LLMs! 🔥
The paper provides an up-to-date review of the literature on LLMs, details about the training mechanisms like pre-training approaches along with instruction tuning techniques & further
Object Detection is vital for many Computer Vision Applications!
Introducing YOLO-NAS a new open-sourced model that outperforms the existing SOTA models.
Here is everything you need to know about the model: 🧵👇
If you want to get started with machine learning and want a hands-on tutorial:
Intro to Machine Learning from Kaggle will help you build your first model!
• 8 Lessons.
• 100% Free.
Check this out:
With a single line of text you can output the entire process of a software company along with carefully orchestrated SOPs 🔥
Introducing MetaGPT a Multi-Agent Framework that can act as an engineer, product manager, architect and project manager.
The speed of light through water is 46.2k miles per second slower than through air & underwater photos have always looked poor; color adjustment isn't enough to fix them.
@dakkaynak
uses multiple images without any model training; the results are amazing!
Once you train a machine learning model, the best way to showcase it to the world is to make a demo for others to try it.
Introducing Modules from Clarifai using
@streamlit
that let's you create a demo of your Machine Learning Project! 🚀
Thread🧵
@LangChainAI
is a framework for developing applications powered by language models.
In this example, let's build a simple Chat Application with Langchain and create a web app using UI Modules and deploy it to the cloud.
Let's look at it step-by-step
1/8 🧵👇
Multistage RAG Pipeline With DSPy!
Leverage Clarifai and DSPy to build a multi-stage RAG module.
The module extracts keywords to broaden the search, retrieves information using both the query and keywords, and employs a separate language model to evaluate and rank the retrieved
Another interesting use for computer vision: China's national diving team uses an AI system in their training. The system provides guidance on the athletes' posture and movements.
#training
#olympics
Source:
The recent LLM from
@huggingface
zephyr-7b-alpha outperforms Llama 2 70B Chat on MT Bench 🤯
You can now run zephyr-7b-alpha with just a few lines of code:
Read more: ⬇️
Vector databases are getting a lot of attention recently with the rise in large language models.
If you are not sure what exactly it is, here's a very simple introduction to Vector Databases: 🧵 ⬇️
1/8
Although the above code is super slow and doesn't include any training code, backpropagation, GPU/TPU support, batching, etc
This is a good way to get an idea of the implementation of GPT.
Checkout the complete code here:
Applications of Transformers!
A comprehensive survey paper on applications of Transformers for deep learning tasks.
This also includes list of transformer models for NLP, Computer Vision, Multi-Modal and Audio.
Cheatsheet and a quick reference guide that covers:
- LLM basics
- Word embeddings and vector representations
- Transformer, GPT etc architecture
- Training and fine-tuning LLMs
- Evaluation metrics for LLMs
Check this out:
Introducing Nougat from
@Meta
!
A Visual Transformer model that converts document images including complex math equations, into a structured text!
Thread ⬇️
Evaluating Large Language Models: A Comprehensive Survey on evaluating LLMs! 🔥
LLMs have attracted significant attention in recent past but they also suffer from potential risks like private data leaks or yield inappropriate, harmful, or misleading content.
This is a great
A Free Machine Learning course from Stanford taught by Andrew Ng.
You'll learn:
• Linear/Logistic regression
• Neural Networks
• Decision Trees
• Support Vector Machines
• Naive Bayes
• KNN
• Gradient Descent
And much more.
Want to understand how Large Language Models work?
The article helps you understand the core concepts in an interactive way!
One of the most beautiful visualizations on how LLMs work.
Check this out:
Open Source Large Language Models are taking over places.
Introducing the new Falcon-40B that outperformed all the other existing LLMs, including LLaMa-65B from Meta and MPT-7B.
Falcon-40B is ranked first on the OpenLLM leaderboard
Read more🧵👇
.
@AnthropicAI
released Claude 3, outperforming the GPT-4 model from OpenAI! ⚡️
Here are some key highlights: 🔥
- Three different models: Haiku, Sonnet, Opus.
- Opus model outperforms GPT-4 on every benchmark.
- 200K token context window.
- Sonnet is 2x faster than the
3D Pose and Tracking for Human Action Recognition from Meta AI 🔥
You can now do motion capture and action recognition from a single video.
Here is everything you need to know about the model:🧵👇
Mistral 7B Instruct is available now!
Mistral 7B is the latest model from
@MistralAI
that outperforms Llama 2 13B on every benchmark.
Mistral 7B is finetuned on publicly available instruction datasets resulting in Mistral 7B Instruct
Read more: ⬇️
With the rise in Generative AI Applications Vector Databases became really key to store embeddings
Introducing our new Vector Database Integration with
@LangChainAI
🎉
Now you can use Clarifai to store Text and Document data into embedding vectors!
Here are more details:🧵
Introducing the Clarifai's New Python SDK! 🔥
You can now:
• Import datasets along with annotations.
• Create apps, inputs, datasets, and do model predictions.
• Access models, workflows & other LLMs from the Clarifai community with just a few lines of code.
Read More:🧵👇
ChatGPT vision but OpenSource 🔥
Introducing LLaVa, a 13B Multimodal LLM with vision assistance providing impressive chat capabilities similar to multimodal ChatGPT!
DSPy is the framework that lets you program language models, not just prompting! 🔥
Our latest tutorial will walk you through DSPy and building a simple RAG system with DSPy and Clarifai! ⚡️
Watch the full video here:
Introducing StreamingLLM 🔥
A framework that enable Large Language Models(LLMs) to handle infinite sequence lengths without the need for fine-tuning.
Thread🧵 ⬇️
XGen-7B, a new open-source LLM by Salesforce 🤩
The model has a context window of 8k tokens!
This is fine-tuned on instructional data, enabling robust long sequence modeling and various NLP applications.
Try the demo here:👇
Clarifai turned 10 🎉
To Celebrate this moment, on our 10 year anniversary we are announcing the Full Stack AI Platform.
The platform offers end-to-end capabilities, including serverless inference, prompt engineering, model comparison and routing, low-rank fine-tuning,
Adding the New Generative AI Capability to our platform by integrating the latest Open Source and 3rd Party LLMs 🎉
Along with Open Source Models you can now access the models from
@OpenAI
@CohereAI
&
@AI21Labs
right within the Clarifai Platform!
Here are more details:🧵👇
"If someone did a 2-week online course in medicine and told you they are ready to perform a complex brain surgery on you, would you take their offer?"
@ChrisXtg
closes the gap of "real versus fake understanding" of Artificial Intelligence
Tesla had a surprise announcement with the promise of a Bot prototype coming next year. This will be quite the challenging timeline, but perhaps in line with Elon Musk's Mars aspirations – who better to be the first explorers of the red planet?
Fast Segment Anything Model: FastSAM🔥
SAM is an image segmentation model released by Meta which is capable of doing zero-shot segmentation.
Now Introducing FastSAM that achieves a comparable performance with the SAM at 50× higher run-time speed.
Read more:🧵👇
Llama 2 (7B) is now on Clarifai 🔥
The latest open-source model from
@Meta
will now supports a context window of up to 4k tokens!
Fine-tuned Llama-2 models are optimized especially for conversational AI applications.
Try it out now:👇
Introducing StarCoder a Large Language Model for Code!
This is trained on 80+ programming languages, Git commits, GitHub issues, and Jupyter notebooks.
Here is everything you need to know about the Model:🧵
Introducing Orca the new 13B Large Language Model from Microsoft 🔥
The best thing about this LLM is it learns by imitating Large Foundational Models like GPT4 & ChatGPT!
Orca crushes the state-of-the-art instruction-tuned models such as Vicuna-13B:🧵👇
We're delighted to welcome three YouTubers, Tina Huang, Andrea from
@aicoding_
, and
@KenJee_DS
who runs the movement
#66DaysOfData
. We'll have an informal panel chat where we discuss trends in DS/AI education, ethics, career paths, and challenges.
Advanced NLP from Carnegie Mellon University!
The course covers current state-of-the-art NLP techniques & algorithms.
One of the best NLP courses on the web!
Check this out:
Claude-2:
@AnthropicAI
's ChatGPT competitor was released and it's incredible🔥
This is Faster and 5x cheaper than GPT-4 and can also supports longer conversations.
Now you can access the Claude-2 and can also compare it with GPT-4 right from the Clarifai Platform
Thread🧵
Facebookresearch released Audiocraft!
A PyTorch library for deep learning research on audio generation.
This Contains the code for MusicGen, which is the state-of-the-art controllable text-to-music model.
Check this out:
We're thrilled to bring you the Clarifai - Llama 2 Hackathon! 🥳
Prizes up to 7000 $ 💰
Spend a fun (virtual) weekend together and build the best AI-powered projects with Llama 2!
Retweet and tag your friends to help us spread the word.
Register here:
Here are the Top 2 Projects in the recent LLM Hackathon by
@streamlit
built with Clarifai! 🎉
1. AI Research Buddy which is a Q&A and summarization bot using the lastest Nougat transformer model for Academic papers.
Test and compare LLMs!
We have created an LLM Battleground Module designed to help you compare different LLMs.
It's as easy as selecting the LLMs 🚀
Check this out:👇
Introducing GPT-Engineer an AI agent that can write an entire codebase with a single prompt!
It can also learns how you want your code to look.
Read more: ⬇️
New open-source LLM from Salesforce AI Research team🔥
Introducing XGen a 7 Billion Parameter Model trained on 8K Input sequence Length which uses the same architecture as LLaMa!
Here are more details of the Model:🧵
Introducing SSD-1B: Fastest text-to-image model 🔥
Segmind Stable Diffusion 1B is the latest open-source text-to-image model which is 50% smaller and 60% faster than SDXL 1.0!
Python Code to try out SSD-1B: ⬇️
Want to test and compare different LLMs on same prompt and see which works best for you?
We have created an LLM Battleground Module 🥊
Just select the LLMs and start comparing 🚀
Below is the Claude-2 vs GPT-4 Demo.
Check this out: 👇
StarCoder a Large Language Model for Code from
@BigCodeProject
!
This is trained on 80+ programming languages, Git commits, GitHub issues, and Jupyter notebooks.
Here are more details about StarCoder:🧵
Learn how to use
@CohereAI
Embedding Model, to train a classifier!
In this video, We deep dive into:
1. How to use Cohere Embedding Model that takes text as input and generate Embeddings
2. Training the classifier using transfer learning.
Check this:👇
Introducing the new Clarifai +
@llama_index
integration!
You can now access all the LLMs and embedding models from Clarifai Community through LlamaIndex!
Read more:🧵 ⬇️
Access the latest Claude 2.1 from
@AnthropicAI
with an API!
1. Claude 2.1 provides 200K token context window, doubling the amount of information it can process.
2. 2x Decrease in Hallucination Rates
Read more: ⬇️
A Gentle Introduction to Retrieval Augmented Generation (RAG) for the Intelligence Community!
LLMs are trained on large public datasets, but they don't have access to *your* data.
RAG has been revolutionizing how we access and utilize data, offering real-time information