Hugging Face

For most of the ML problems, we use hugging face prediction models as these models give better performance than any other models. It helps in addressing the technological advancements in an organisation. Any organisation that wants to adopt to latest technologies should consider Hugging face. Hugging face has many open-source transformer models hosted. The scope of this product is to give better performance on NLP problems.

• It’s better accuracy over all other models.
• Easy to implement with little understanding on the models.
• Has access to hundreds of models and easy to change.

• Can attract more clients if using the latest and advanced technologies.
• Thus increases ROI of a company.
• Can have impact the business highly by delivering the best product to customer

TensorFlow and Keras

TensorFlow, Keras

I have use Hugging Face to develop Natural Language Processing applications for other amazon web services customers. Some of the common applications are intelligent document processing, call center support, machine translation, sentiment analysis and so on. These Hugging Face solutions are implemented on the cloud for easier manage and maintain as well.

• Open source
• Simple to use
• Rich api

• Increase use case number
• Support more features
• Increase ROI
• decreased development time

Amazon Comprehend

AWS CodeBuild, AWS Batch, Zoom

We use Hugging Face models and datasets to design, test a compare multiple approaches for ML projects and, and in general, for research purposes. Thanks to Hugging Face, we do not need extensive training, and our NLP models' fine-tuning is simpler and more cost efficient.

• Multiple models for multiple tasks to try
• Multiple dataset for multiple tasks to try

• Using Hugging Face is cost efficient vs other paid alternatives

OpenAI

OpenAI

We use Hugging Face APIs to import the models in our code (mostly language models with weights). This is very important use case as it makes the building part of model very easy. We don't have to spend much time refering to repositories, reading complex ReadMe's. Other than that, we deploy demo apps on the Hugging Face spaces using the gradio tool they provide. This helps in testing out the product very easily by not spending much time on making the UI and also not caring about the compute management.

• Model APIs
• Hugging Face spaces

• Reduced the time spent drastically in building complex transformer models
• Very quick deployment of demo apps, that reduces the time spent on making UIs

Weights & Biases, Overleaf, GitHub

In our organization, Hugging Face is used for a lot of text-processing and natural language processing tasks. Hugging face addresses our business problem of finding good NLP algorithms for running classification analysis by using open source API to keep our costs low. The website provides world's best systems for doing NLP and using this model, we are able to do advance NLP analyses and classification using text data.

• NLP libraries
• Text Preprocessing Libraries
• Tutorials

• Improved classification model
• cleaner text data

Treasure Data, LiveRamp, Amperity