alexbakers/transformers_ipfs

Transformers-IPFS Demo

This Streamlit app demonstrates how to use Hugging Face Transformers with IPFS models using the transformers-ipfs integration.

Features

• —Load models directly from IPFS using their CID
• —Analyze sentiment of text using an IPFS-hosted model
• —View confidence scores and probability distribution for each sentiment class
• —Try with example texts or input your own

How It Works

The transformers-ipfs package patches the Hugging Face Transformers library to:

1. 1.Recognize ipfs:// URIs as valid model identifiers
2. 2.Download model files from IPFS nodes or gateways
3. 3.Cache models locally for faster loading in subsequent runs

Model Used

This demo uses a sentiment analysis model hosted on IPFS with CID: bafybeichqdarufyutqc7yd43k77fkxbmeuhhetbihd3g32ghcqvijp6fxi

The original Hugging Face model is: riturajpandey739/gpt2-sentiment-analysis-tweets

Usage

1. 1.Enter your own text or select from the examples
2. 2.Click "Analyze Sentiment" to get the sentiment prediction
3. 3.View the results: predicted sentiment, confidence, and probability distribution

Note: The first run may take longer as the model is downloaded from IPFS.

Requirements

• —streamlit
• —transformers
• —torch
• —transformers-ipfs