README (1).md

metadata

library_name: transformers
tags: []

Model Card for Model ID

This is the baseline model for the news source classification project.

Please run the following evaluation pipeline code:

START

Imports

from huggingface_hub import hf_hub_download 
import joblib 
!huggingface-cli login 
import pandas as pd  
import torch  
from transformers import AutoTokenizer, AutoModel
import torchvision
from torchvision import transforms, utils 
import torch.nn as nn
import torch.optim as optim
import torchvision.transforms as transforms  
from PIL import Image
from skimage import io, transform 
from torchvision.io import read_image
from torch.utils.data import Dataset, DataLoader 
from sklearn.metrics import accuracy_score  
import numpy as np 
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import nltk
from nltk.corpus import stopwords
nltk.download('stopwords')
nltk.download('wordnet')

import re
from transformers import DistilBertTokenizer, DistilBertModel

Load model from Huggingface (Please load test data into test_df below)

repo_id='awngsz/nn_model'  
filename='nn_model_v3.joblib'

model_file_path=hf_hub_download(repo_id=repo_id, filename=filename)  
 
model=joblib.load(model_file_path) 
print(model)  
  
#Load test dataset (assuming the name is the same as the one in the Ed post) 
 
test_df = pd.read_csv(file_path) 

#Copying the naming convention from the sample dataset in the edpost 
 
X_test = test_df['title']  
y_test = test_df['labels']  

Clean the data

def clean_headlines(df, column_name):
    """
    Cleans a specified column in a DataFrame by:
    - Removing HTML tags
    - Removing ', '', regex=True)

    # Remove special characters
    df[column_name] = df[column_name].str.strip().str.replace(r'[&*|~`^=_+{}[\]<>\\]', ' ', regex=True)

    # Remove repeating special characters
    df[column_name] = df[column_name].str.strip().str.replace(r'([?!])\1+', r'\1', regex=True)

    # Remove tabs
    df[column_name] = df[column_name].str.replace(r'\t', ' ', regex=True)

    # Remove newline characters
    df[column_name] = df[column_name].str.replace(r'\n', ' ', regex=True)

    # Normalize all references to US as u.s.
    df[column_name] = df[column_name].str.replace(r'US', 'u.s.', regex=True)
    df[column_name] = df[column_name].str.replace(r'UN', 'u.n.', regex=True)

    # Remove extra spaces including leading/trailing whitespaces
    df[column_name] = df[column_name].str.strip().str.replace(r'\s+', ' ', regex=True)

    # get rid of these fox news patterns we see
    df[column_name] = df[column_name].str.replace(r'fox news poll:', '', regex=True)

    df[column_name] = df[column_name].str.replace(r'| fox news', '', regex=True)

    df[column_name] = df[column_name].str.replace(r'Fox News', '', regex=True)
    df[column_name] = df[column_name].str.replace(r'fox news', '', regex=True)

    df[column_name] = df[column_name].str.replace(r'news poll:', '', regex=True)

    df[column_name] = df[column_name].str.replace(r'opinion:', '', regex=True)

    df[column_name] = df[column_name].str.replace(r"reporter's notebook", '', regex=True)

    # Normalize double quotes to single quotes
    # df[column_name] = df[column_name].str.replace(r'"', "'", regex=True)

    # Punctuation
    # df[column_name] = df[column_name].str.replace(r'[.,()]', '', regex=True)

    return df 

def normalize_headlines(df, column_name):
  """
    Normalizes a given headline by:
    - converting it to lowercase
    - removing stopwords
    - applying stemming or lemmatization to reduce words to their base forms

    Args:
        df (pd.DataFrame): The DataFrame containing the column to clean
        column_name (str): The name of the column to clean

    Returns:
        pd.DataFrame: A DataFrame with the cleaned column
  """

  # Convert headlines to lowercase
  df[column_name] = df[column_name].str.lower()

  # Remove stopwords from headline
  stop_words = set(stopwords.words('english'))
  df[column_name] = df[column_name].apply(lambda x: ' '.join([word for word in x.split() if word not in (stop_words)]))

  # Lemmatize words to base form
  lemmatizer = nltk.stem.WordNetLemmatizer()
  df[column_name] = df[column_name].apply(lambda x: ' '.join([lemmatizer.lemmatize(word) for word in x.split()]))

  return df 

def handle_missing_data(df, column_name):
    """
    Handles missing or incomplete data in a given column of a DataFrame, including:

    - Replacing NULL values with "Unknown Headline"
    - Augmenting the data by creating headlines with synonyms of words in other headlines

    Args:
        df (pd.DataFrame): The DataFrame containing the column to clean
        column_name (str): The name of the column to clean

    Returns:
        pd.DataFrame: A DataFrame with the cleaned column
    """

    # Remove NULL headlines
    df = df.dropna(subset=[column_name])

    # Set a minimum word count threshold
    min_word_count = 3

    # Filter out titles with fewer words
    df = df[df[column_name].str.split().apply(len) >= min_word_count].reset_index(drop=True)


    return df 

def consistency_checks(df, column_name):
  """
    Ensures all headlines follow a consistent format by:
      - Removing duplicate headlines

    Args:
        df (pd.DataFrame): The DataFrame containing the column to clean
        column_name (str): The name of the column to clean

    Returns:
        pd.DataFrame: A DataFrame with the cleaned column

  """

  # Remove duplicate headlines
  df = df.drop_duplicates(subset=[column_name])

  # Filter headlines with too few or too many words
  #df = df[df['title'].str.split().apply(len).between(3, 20)]


  return df 

X_test = clean_headlines(X_test, 'title')
X_test = normalize_headlines(X_test, 'title')
X_test = X_test.dropna(subset = ['title'])
X_test = handle_missing_data(X_test, 'title')
X_test = consistency_checks(X_test, 'title') 

Load the embedding model from Huggingface. Transformer: DistilBERT

def get_embeddings(text_all, tokenizer, model, device, max_len=128):
    '''
    Generate embeddings using a transformer model on GPU if available.
    Args:
    - text_all: List of input texts
    - tokenizer: Tokenizer for the model
    - model: Transformer model
    - device: torch.device to run the computations
    - max_len: Maximum token length for the input
    Returns:
    - embeddings: List of embeddings for each input text
    '''
    embeddings = []
    
    count = 0
    print('Start embeddings:')

    for text in text_all:
        count += 1
        if count % (len(text_all) // 10) == 0:
            print(f'{count / len(text_all) * 100:.1f}% done ...')

        # Tokenize the input text
        model_input_token = tokenizer(
            text,
            add_special_tokens=True,
            max_length=max_len,
            padding='max_length',
            truncation=True,
            return_tensors='pt'
        ).to(device)  # Move input tensors to GPU

        # Generate embeddings without gradient computation
        with torch.no_grad():
            model_output = model(**model_input_token)
            cls_embedding = model_output.last_hidden_state[:, 0, :]  # Use CLS token embedding
            cls_embedding = cls_embedding.squeeze().cpu().numpy()  # Move back to CPU for numpy
            embeddings.append(cls_embedding)

  return embeddings 

Check for GPU availability

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(f'Using device: {device}')

# Load the tokenizer and model for 'all-mpnet-base-v2'
print("Loading model and tokenizer...")
# Load model and tokenizer
tokenizer_news = AutoTokenizer.from_pretrained('distilbert-base-uncased')
model_news = AutoModel.from_pretrained('distilbert-base-uncased').to(device)

# Set the model to evaluation mode
model_news.eval()

############################################# DBERT UNCASED Embedding #############################################
############################################# Embedding #############################################
print("Computing DBERT embeddings for training data...")

y_test = X_test['labels']
X_test = X_test['title']
  
X_test_embeddings_DBERT = get_embeddings(X_test, tokenizer_news, model_news, device, max_len=128)
print("DBERT embeddings for training data computed!")


prediction = model.predict(X_test_embeddings_DBERT)

Accuracy

label_map = {'NBC': 0, 'FoxNews': 1}

def compute_category_accuracy(y_true, y_pred, label):
  y_true = np.array(y_true)
  n_correct = np.sum((y_true == label) & (y_pred == label)) 
  n_total = np.sum(y_true == label) 
  cat_accuracy = n_correct / n_total 
  return cat_accuracy

#Print accuracy 
print(f'Test accuracy: {accuracy_score(y_test, prediction) * 100:.2f}%') 
print(f'Test accuracy for NBC: {compute_category_accuracy(y_test, prediction, label_map["NBC"]) * 100:.2f}%') 
print(f'Test accuracy for FoxNews: {compute_category_accuracy(y_test, prediction, label_map["FoxNews"]) * 100:.2f}%')

END

Model Details

Model Description

This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.

• Developed by: [More Information Needed]
• Funded by [optional]: [More Information Needed]
• Shared by [optional]: [More Information Needed]
• Model type: [More Information Needed]
• Language(s) (NLP): [More Information Needed]
• License: [More Information Needed]
• Finetuned from model [optional]: [More Information Needed]

Model Sources [optional]

• Repository: [More Information Needed]
• Paper [optional]: [More Information Needed]
• Demo [optional]: [More Information Needed]

Uses

Direct Use

[More Information Needed]

Downstream Use [optional]

[More Information Needed]

Out-of-Scope Use

[More Information Needed]

Bias, Risks, and Limitations

[More Information Needed]

Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

How to Get Started with the Model

Use the code below to get started with the model.

[More Information Needed]

Training Details

Training Data

[More Information Needed]

Training Procedure

Preprocessing [optional]

[More Information Needed]

Training Hyperparameters

• Training regime: [More Information Needed]

Speeds, Sizes, Times [optional]

[More Information Needed]

Evaluation

Testing Data, Factors & Metrics

Testing Data

[More Information Needed]

Factors

[More Information Needed]

Metrics

[More Information Needed]

Results

[More Information Needed]

Summary

Model Examination [optional]

[More Information Needed]

Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

• Hardware Type: [More Information Needed]
• Hours used: [More Information Needed]
• Cloud Provider: [More Information Needed]
• Compute Region: [More Information Needed]
• Carbon Emitted: [More Information Needed]

Technical Specifications [optional]

Model Architecture and Objective

[More Information Needed]

Compute Infrastructure

[More Information Needed]

Hardware

[More Information Needed]

Software

[More Information Needed]

Citation [optional]

BibTeX:

[More Information Needed]

APA:

[More Information Needed]

Glossary [optional]

[More Information Needed]

More Information [optional]

[More Information Needed]

Model Card Authors [optional]

[More Information Needed]

Model Card Contact

[More Information Needed]