| from transformers import DistilBertTokenizerFast, DistilBertForSequenceClassification | |
| import torch | |
| import pandas as pd | |
| import numpy as np | |
| import streamlit as st | |
| from transformers import pipeline | |
| from transformers_interpret import SequenceClassificationExplainer | |
| from transformers import AutoModelForSequenceClassification, AutoTokenizer | |
| st.title('Ethics Classifier') | |
| st.write('This app uses a pre-trained Distilbert model fine-tuned on the Commonsense Ethics dataset from the Aligning AI With Shared Human Values project (https://github.com/hendrycks/ethics). It judges whether a given action of scenario is wrong or not wrong and shows how the words in the scenario affected the judgment.') | |
| loaded_model = DistilBertForSequenceClassification.from_pretrained('commonsense_ethics') | |
| model_name = 'distilbert-base-uncased' | |
| tokenizer = DistilBertTokenizerFast.from_pretrained(model_name) | |
| cls_explainer = SequenceClassificationExplainer(loaded_model, tokenizer) | |
| clf = pipeline("text-classification", model = loaded_model, tokenizer = tokenizer) | |
| text = st.text_input('Enter a scenario or action.') | |
| if text: | |
| answer = clf(text) | |
| label = 'wrong' if answer[0]['label'] == 'LABEL_1' else 'not wrong' | |
| st.write(f'This action is {label} (confidence level {answer[0]["score"]*100:.2f}%).') | |
| attributions = cls_explainer(text) | |
| df = pd.DataFrame(attributions[1:-1]) | |
| df.rename(columns = {0: 'Token', 1: 'Contribution'}, inplace = True) | |
| st.write(df.style.hide(axis = 'index')) | |
| st.write(cls_explainer.visualize()) |