ilyada/web_accessibility_model

1from datasets import load_dataset2from transformers import BertTokenizer, BertForSequenceClassification, Trainer, TrainingArguments3import torch4 5# Load the dataset6dataset = load_dataset("ilyada/web_accessibility_dataset")7 8# Load pre-trained model and tokenizer9model_name = "bert-base-uncased"10tokenizer = BertTokenizer.from_pretrained(model_name)11model = BertForSequenceClassification.from_pretrained(model_name, num_labels=2)12 13# Tokenize the dataset14def tokenize_function(examples):15    return tokenizer(examples["text"], padding="max_length", truncation=True)16 17tokenized_datasets = dataset.map(tokenize_function, batched=True)18 19# Split the dataset into train and test20train_test_split = tokenized_datasets["train"].train_test_split(test_size=0.2)21train_dataset = train_test_split['train']22test_dataset = train_test_split['test']23 24# Define training arguments25training_args = TrainingArguments(26    output_dir="./results",27    evaluation_strategy="epoch",28    learning_rate=2e-5,29    per_device_train_batch_size=8,30    per_device_eval_batch_size=8,31    num_train_epochs=3,32    weight_decay=0.01,33    push_to_hub=True,  # This enables pushing the model to Hugging Face Hub34    hub_model_id="ilyada/web_accessibility_model",  # LLM generated dataset35    hub_strategy="end",36)37 38# Initialize the Trainer39trainer = Trainer(40    model=model,41    args=training_args,42    train_dataset=train_dataset,43    eval_dataset=test_dataset,44)45 46# Train the model47trainer.train()48 49# Evaluate the model50results = trainer.evaluate()51print(results)52 53# Push model to Hugging Face Hub54trainer.push_to_hub()