Article : https://medium.com/@yash9439/introducing-falcon-llm-leading-the-way-in-language-generation-79a73a56dd5

Article : https://medium.com/@yash9439/unleashing-the-power-of-falcon-code-a-comparative-analysis-of-implementation-approaches-803048ce65dc

Installation

To install the required dependencies, run the following command:

!pip install -q transformers einops accelerate langchain bitsandbytes

Falcon Local

This repository contains three versions of the Falcon code, each designed for different use cases. These versions utilize different libraries and models for text generation. Below is a brief description of each version:

Version 1 : Using Transformers only.

File: falcon_local_withoutHuggingFacePipeline.py

This version uses the transformers library and the tiiuae/falcon-7b-instruct model. Its a text-generation model which gives answer for the question asked.

Version 2 : Using Transformers and HuggingFacePipeline only.

File: falcon_local_withHuggingFacePipeline.py

The second version of the Falcon code utilizes the langchain library and the tiiuae/falcon-7b-instruct model. It employs the HuggingFacePipeline class for text generation and generates a response to a given question.

Version 3 : Using HuggingFacePipeline Only.

File: falcon_local_HuggingFaceInferenceAPI.py

In this version, the code incorporates the dotenv, langchain, and textwrap libraries. It uses the HuggingFace model hub and the tiiuae/falcon-7b-instruct model. The code demonstrates a summarization chain for generating responses to questions.

Please refer to the individual code files for more details and instructions on running each version of the Falcon code.

License

This project is licensed under the Apache License 2.0.