Introducing Universal-1, our most powerful speech recognition model to date.
Trained on over 12.5 million hours of multilingual audio data, Universal-1 achieves best-in-class speech-to-text accuracy across English, Spanish, French, and German.
Here are 10 common AI terms explained in an easily understandable way.
1. Classification
2. Regression
3. Underfitting
4. Overfitting
5. Cost function
6. Loss function
7. Validation data
8. Neural Network
9. Parameter
10. Hyperparameter
AI Thread🧵👇
5 FREE courses to learn Machine Learning for absolute beginners:
Kaggle Python
Kaggle Intro To ML
Kaggle Intermediate ML
Google ML Problem Framing
Google ML Crash Course
Links 👇
10 Machine Learning YouTube channels you should follow:
1. Data Professor
@thedataprof
2. Mısra Turp
@misraturp
3. Sentdex
4. AssemblyAI
5. Two Minute Papers
6. Tech with Tim
7. Machine Learning with Phil
8. Lex Fridman
9. Python Engineer
@python_engineer
10. Smitha Kolan
Introducing Conformer-1: our latest state-of-the-art speech recognition model.
Built on top of the Conformer architecture and trained on 650K hours of audio data, it achieves near-human-level performance, making up to 43% fewer errors on noisy data than other ASR models.
1/9
Linear and Logistic Regression from scratch in Python.
Both algorithms can be implemented using an Object Oriented Programming approach with only minimal code difference between the two classes:
Are you tired of googling how to work with pandas DataFrames?
Meet the Open Source package "sketch". It's an AI code-writing assistant that understands data content.
It helps you to analyze your data and to write code:
Let's see how to use it:
1/4
PCA from scratch in Python.
An unsupervised learning technique that works by projecting the data onto the first few principal components (= eigenvectors)
Python NumPy Tip:
If you need to count the number of occurrences of each value in an array, you can use numpy.bincount instead of collections.Counter.
bincount is much faster:
Do you know VS Code supports Jupyter Notebooks?
You just need a Python environment containing the Jupyter package and can then create a file with .ipynb ending.
Do you use this setup for your ML projects?
🎉ANNOUNCEMENT🎉
On Monday, we'll release a new free course:
Machine Learning From Scratch in Python
Learn how to implement:
KNN
Linear Regression
Logistic Regression
Decision Trees
Random Forest
Naive Bayes
PCA
Perceptron
SVM
KMeans
Set a reminder:
Perceptron from scratch in Python
A Perceptron is a single layer neural network with the unit step function as activation function, and a beautiful intuitive update rule that either increases or decreases the weights depending on the prediction
How to create an AI app with a free GPU using Flask ngrok, and Google Colab.
In this example, we build our own Stable Diffusion app.
Let's look at it step-by-step (The link to the code is at the end).
1/8🧵
12 ML and DL Libraries we love:
🤓NumPy
🤓Pandas
📊matplotlib
📊seaborn
🤖scikit-learn
🤖XGBoost
🧠PyTorch
🧠TensorFlow
🧠JAX
💬NLTK
👀OpenCV
🤗Transformers
What about you?
A few days ago we shared Logistic Regression.
Today we show you how to implement Linear Regression from scratch in Python with gradient descent.
Can you spot the difference?
5 amazing free courses to get started with AI and LLMs:
1. LLM University by
@CohereAI
:
2. LLM Bootcamp by
@full_stack_dl
:
3. Practical Deep Learning for Coders by
@FastDotAI
:
4. Deep Learning fundamentals…
As part of our end-of-the-year countdown collaboration with many amazing creators, we asked them to recommend must-read books for 2023.
Here is what they recommended: 1/4🧵
Found an awesome free Scikit-learn course:
It contains written guides, videos, and exercises!
Outline:
ML Concepts
Predictive Modeling Pipeline
Model Selection
Hyperparameter Tuning
Linear Models
Decision Trees
Model Ensembles
Evaluating Performance
Here are 3 of our favorite courses we created this year on our YouTube channel:
🟡Python Speech Recognition
🔵Machine Learning From Scratch
🟣Deep Learning Explained
Learn these topics for free with the links below🤗
A new paper that might be a game changer for AI image generation was released.
ControlNet - Adding Conditional Control to Text-to-Image Diffusion Models
We can now use additional input like sketches, outlines, or human poses to control diffusion models:
💡Tip:
Whenever you find a Jupyter notebook on GitHub and want to try it immediately in a Google Colab, just change the URL from
github. com/...
to
githubtocolab. com/...
Wanna learn about
@LangChainAI
?
Next to the very good official documentation, here are 3 resources to get you started:
- 3 minute Explainer video:
- 15 minute Crash Course:
- Google Colab Guide:
Does any of those roles sound interesting to you?
Data Scientist
Developer Educator
Front End Developer
API Support Engineer
Senior DevOps Engineer
Senior Backend Engineer, Python
Research Engineer, NLP Modeling
Deep Learning Researcher, Speech
Check out our careers page!👇
If you want to build voice-based LLM applications, you need many things:
- Record microphone
- Real-time speech recognition
- An LLM agent
- A voice synthesizer
Luckily, there is an open-source library from
@vocodehq
that makes all of this super simple.
Let's learn more👇
1/5
PyTorch 2.0 was announced!
Main new feature: torch.compile
A compiled mode that accelerates your model without needing to change your model code. It can speed up training by 38-76%, and 2.0 is fully backward-compatible🤗
🎉The first lesson of our Machine Learning From Scratch in Python course is out!!!🎉
Learn how to implement KNN using only built-in Python functions and Numpy!
In the next 9 days we'll release 9 more lessons with other popular algorithms, stay tuned!
Feeling overwhelmed by all the Machine Learning courses?
We created a minimal ML study guide with just 4 steps and only 2-3 necessary resources for each step:
1. Learn Python
2. Learn The ML Tech Stack
3. Machine Learning Courses
4. Practise!
Facebookresearch released audiocraft!
A PyTorch library for deep learning research on audio generation.
It contains the code for MusicGen, a state-of-the-art controllable music generation LM for text-to-music.
Pretty cool!
This repository gives you free cheatsheets for Stanford's CS 229 Machine Learning and all relevant fields:
- Supervised Learning
- Unsupervised Learning
- Deep Learning
- Tips and Tricks
- Probabilities and Statistics
- Algebra and Calculus
Enjoy!
Introducing Conformer-2, our latest AI model for automatic speech recognition trained on 1.1M hours of audio data that achieves state-of-the-art results for speech to text conversion.
1/6
A very simple example of an ML app with
@FastAPI
.
- Loads a stored model
- Has a `predict` endpoint that runs the model
- Specifies input and output fields & types for the endpoint
Python Tip⚡️
A really cool library to work with unevenly-spaced time series data is 'traces':
Let's look at this data from a light switch. We add the five measurements at 6:00am, 7:45am, etc.
Then we can get information about unknown time points, and also the distribution:
Reinforcement Learning is taking over the AI world.
Here are 4 RL courses from Stanford, UC Berkeley, DeepMind, and Hugging Face you can take for free to keep up with the trend:
Win a Raspberry Pi 400!
To enter:
1. Like this tweet
2. Follow
@AssemblyAI
on Twitter
US residents only. Ends September 29. Winner will be contacted via DM.
#raspberrypi
#raspberrypi400