app.py

import torch
import gradio as gr
from transformers import AutoTokenizer, AutoModelForSequenceClassification, AutoModelForTokenClassification
import os

# Load the tokenizer and models for the first pipeline
tokenizer_ext = AutoTokenizer.from_pretrained("AlGe/deberta-v3-large_token")
model_ext = AutoModelForTokenClassification.from_pretrained("AlGe/deberta-v3-large_token")
tokenizer_ext.model_max_length = 512
pipe_ext = gr.pipeline("ner", model=model_ext, tokenizer=tokenizer_ext)

# Load the tokenizer and models for the second pipeline
tokenizer_ais = AutoTokenizer.from_pretrained("AlGe/deberta-v3-large_AIS-token")
model_ais = AutoModelForTokenClassification.from_pretrained("AlGe/deberta-v3-large_AIS-token")
tokenizer_ais.model_max_length = 512
pipe_ais = gr.pipeline("ner", model=model_ais, tokenizer=tokenizer_ais)

# Load the tokenizer and models for the third pipeline
auth_token = os.environ['HF_TOKEN']
model1 = AutoModelForSequenceClassification.from_pretrained("AlGe/deberta-v3-large_Int_segment", num_labels=1, use_auth_token=auth_token)
tokenizer1 = AutoTokenizer.from_pretrained("AlGe/deberta-v3-large_Int_segment", use_auth_token=auth_token)

model2 = AutoModelForSequenceClassification.from_pretrained("AlGe/deberta-v3-large_seq_ext", num_labels=1, use_auth_token=auth_token)

# Define functions to process inputs
def process_ner(text, pipeline):
    output = pipeline(text)
    entities = []
    current_entity = None

    for token in output:
        entity_type = token['entity'][2:]
        entity_prefix = token['entity'][:1]

        if current_entity is None or entity_type != current_entity['entity'] or (entity_prefix == 'B' and entity_type == current_entity['entity']):
            if current_entity is not None:
                entities.append(current_entity)
            current_entity = {
                "entity": entity_type,
                "start": token['start'],
                "end": token['end'],
                "score": token['score']
            }
        else:
            current_entity['end'] = token['end']
            current_entity['score'] = max(current_entity['score'], token['score'])

    if current_entity is not None:
        entities.append(current_entity)

    return {"text": text, "entities": entities}

def process_classification(text, model1, model2, tokenizer1):
    inputs1 = tokenizer1(text, max_length=512, return_tensors='pt', truncation=True, padding=True)
    
    with torch.no_grad():
        outputs1 = model1(**inputs1)
        outputs2 = model2(**inputs1)
    
    prediction1 = outputs1[0].item()
    prediction2 = outputs2[0].item()
    score = prediction1 / (prediction2 + prediction1)

    return f"{round(prediction1, 1)}", f"{round(prediction2, 1)}", f"{round(score, 2)}"

# Define Gradio interface
iface = gr.Interface(
    fn={
        "NER - Extended Sequence Classification": lambda text: process_ner(text, pipe_ext),
        "NER - Autobiographical Interview Scoring": lambda text: process_ner(text, pipe_ais),
        "Internal Detail Count": lambda text: process_classification(text, model1, model2, tokenizer1)[0],
        "External Detail Count": lambda text: process_classification(text, model1, model2, tokenizer1)[1],
        "Approximated Internal Detail Ratio": lambda text: process_classification(text, model1, model2, tokenizer1)[2]
    },
    inputs=gr.Textbox(placeholder="Enter sentence here..."),
    outputs=[
        gr.HighlightedText(label="NER - Extended Sequence Classification"),
        gr.HighlightedText(label="NER - Autobiographical Interview Scoring"),
        gr.Label(label="Internal Detail Count"),
        gr.Label(label="External Detail Count"),
        gr.Label(label="Approximated Internal Detail Ratio")
    ],
    title="Combined Demo",
    description="This demo combines two different NER models and two different sequence classification models. Enter a sentence to see the results.",
    theme="monochrome"
)

# Launch the combined interface
iface.launch()