satya - 1/23/2024, 1:49:31 PM

Hugging face homepage

satya - 1/23/2024, 2:08:46 PM

Quick tour of the transformer library

satya - 1/23/2024, 2:12:05 PM

What is the difference between hugging face transformers library and the inference API?

satya - 1/23/2024, 3:47:31 PM

Scikit, Pytorch, Tensorflow

Scikit: A collection of ML algorithms to train, test, and deploy non traditional ML models
Pytorch, TensorFlow: Core ground level libraries to run Deeplearning models, establish neural networks, run through data, train, test etc.

satya - 1/23/2024, 3:51:07 PM

Transformers library from HuggingFace

For developing, training, or fine-tuning machine learning models. This library allows users to work with pre-trained models or to train their own models.
Uses other libraries like Pytorch and Tensorflow underneath
A large collection of models pretrained models
Provides an execution Environment if needed
It runs in the user's own environment, which means the user needs to manage the computational resources, dependencies, and everything else required to run the models.
It supports multiple deep learning frameworks like TensorFlow and PyTorch and provides interfaces in several programming languages, predominantly Python.

satya - 1/23/2024, 3:53:44 PM

The Hugging Face inference API

Models are managed in the cloud
API simplifies the interaction with the models
Used for using 'trained models', not to train them
The models are classified into the type of tasks they do
The API is also classified based on the tasks
The API definition (signature which encompasses the inputs, outputs based on task) also varies by task
Suitable for small to midsize companies

satya - 1/23/2024, 4:14:02 PM

Here is the HuggingFace TGI: Text Generation Inference Server

satya - 2/9/2024, 7:43:55 AM

Inference api