blackrock_esg_classifier

Local Final ESG Model

This folder contains the final fine-tuned ESG classification model used in the project.

Purpose

• Base model: distilbert-base-uncased
• Task: 4-class ESG text classification
• Labels:
  • Environmental
  • Social
  • Governance
  • Non-ESG

This local folder is the reproducible training artifact that matches the deployed Hugging Face app model:

• johnsonzhangzzz/blackrock_esg_classifier

Main Files

• config.json: model configuration
• model.safetensors: trained model weights
• tokenizer.json: tokenizer definition
• tokenizer_config.json: tokenizer configuration
• special_tokens_map.json: tokenizer special tokens
• vocab.txt: vocabulary file
• label_mapping.json: project label mapping
• training_args.bin: saved training arguments

Usage

This folder can be:

• uploaded to Hugging Face as the deployed model package, or
• loaded locally with transformers

Example:

from transformers import pipeline

pipe = pipeline(
    "text-classification",
    model="04_models/local_final_esg_model",
    tokenizer="04_models/local_final_esg_model",
)

Notes

• The deployed Streamlit app currently loads the Hugging Face repo version of this model.
• The local folder is kept for reproducibility, evaluation, and re-upload if the Hugging Face model needs to be refreshed.