import spaces from pathlib import Path import gradio as gr import torch from transformers import AutoTokenizer, AutoModelForSequenceClassification # The directory containing the fine-tuned BERT model MODEL_DIR = Path(__file__).resolve().parent / "experiments_full" # Load the model once when the application starts tokenizer = AutoTokenizer.from_pretrained(str(MODEL_DIR)) model = AutoModelForSequenceClassification.from_pretrained( str(MODEL_DIR) ).to("cuda") model.eval() @spaces.GPU(duration=30) def predict(text): if not text or not text.strip(): raise gr.Error("Please enter an English tweet.") inputs = tokenizer( text, return_tensors="pt", truncation=True, padding="max_length", max_length=128, ) inputs = {k: v.to("cuda") for k, v in inputs.items()} with torch.inference_mode(): outputs = model(**inputs) probabilities = torch.softmax(outputs.logits, dim=-1)[0] return { "Not Disaster (0)": float(probabilities[0]), "Disaster (1)": float(probabilities[1]), } demo = gr.Interface( fn=predict, inputs=gr.Textbox( lines=4, label="Tweet", placeholder="Enter an English tweet...", ), outputs=gr.Label( num_top_classes=2, label="Prediction", ), title="BERT Disaster Tweet Classifier", description=( "Classify whether an English tweet refers to a real disaster. " "Fine-tuned BERT-base-uncased model." ), examples=[ ["Forest fire near La Ronge Sask. Canada"], ["I love spending time with my friends!"], ["A powerful earthquake has damaged several buildings."], ], ) if __name__ == "__main__": demo.launch()