B E R T Banking77

Introduction

The BERT-Banking77 model, developed by Hugging Face staff member Phil Schmid, is a text classification model tailored for the banking domain. It is trained using AutoTrain on the BANKING77 dataset and optimized for multi-class classification tasks.

Architecture

BERT-Banking77 utilizes the BERT architecture, a transformer-based model developed by Google. It is implemented in PyTorch and designed for text classification tasks within the banking sector.

Training

The model is trained with AutoTrain, a tool that automates the training process for machine learning models. It has been evaluated on the BANKING77 dataset with the following key metrics:

• Accuracy: 92.64%
• Macro F1: 92.64%
• Weighted F1: 92.6%

Guide: Running Locally

To run the BERT-Banking77 model locally, follow these steps:

1. Install Dependencies: Ensure you have Python and the transformers library installed. You can install it using pip:

   pip install transformers
   

2. Load the Model: Use the following Python script to load the model and tokenizer, and classify text inputs:

   from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
   
   model_id = 'philschmid/BERT-Banking77'
   tokenizer = AutoTokenizer.from_pretrained(model_id)
   model = AutoModelForSequenceClassification.from_pretrained(model_id)
   classifier = pipeline('text-classification', tokenizer=tokenizer, model=model)
   
   result = classifier('What is the base of the exchange rates?')
   print(result)
   

3. Inference: You can also perform inference using cURL with the Hugging Face API:

   curl -X POST -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" -d '{"inputs": "I love AutoTrain"}' https://api-inference.huggingface.co/models/philschmid/BERT-Banking77
   

Cloud GPUs: For large-scale deployment or faster inference, consider using cloud GPU services from providers like AWS, Google Cloud, or Azure.

License

The BERT-Banking77 model and its code are available under the terms specified by Hugging Face. Users should refer to the model card on the Hugging Face website for detailed licensing information.