File size: 1,597 Bytes
83dc660 ecb94b2 b5e004b 9ea2f77 55ccd60 9954112 9ea2f77 55ccd60 ec39271 55ccd60 9954112 923d55b 9ea2f77 55ccd60 9ea2f77 83dc660 9ea2f77 83dc660 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 |
import gradio as gr
from transformers import DistilBertTokenizer, DistilBertForSequenceClassification
import torch
from torch.nn import functional as F
# Load the trained intent classification model
model_name = "sonisatish119/PhysioMindAI-intent-classification-bert"
tokenizer = DistilBertTokenizer.from_pretrained(model_name)
model = DistilBertForSequenceClassification.from_pretrained(model_name)
# Define intent label mapping
intent_labels = {'reschedule_appointment': 0, 'appointment_location_details': 1, 'cancel_appointment': 2, 'emergency_booking': 3, 'check_appointment_status': 4, 'modify_appointment_details': 5, 'available_slots_inquiry': 6, 'appointment_requirements': 7, 'book_appointment': 8, 'appointment_reminder': 9}
# Function to predict intent
def predict_intent(text):
model.eval() # Set to evaluation mode
inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True, max_length=64)
with torch.no_grad():
outputs = model(**inputs)
logits = outputs.logits
predicted_class = torch.argmax(F.softmax(logits, dim=1)).item()
intent_name = [k for k, v in intent_labels.items() if v == predicted_class][0]
return intent_name
# Gradio Interface
demo = gr.Interface(
fn=predict_intent,
inputs=gr.Textbox(placeholder="Type your query here...", lines=2),
outputs=gr.Textbox(label="Predicted Intent"),
title="📝 Intent Classification Model",
description="This model classifies user queries into predefined appointment-related intents."
)
# Launch the app
if __name__ == "__main__":
demo.launch()
|