app.py

import gradio as gr
from transformers import BertConfig, BertForSequenceClassification, AutoTokenizer
from safetensors import safe_open
import torch

config_path = "peterkros/cofogv1-bert/modelbert2/config.json"
config = BertConfig.from_json_file(config_path)
model = BertForSequenceClassification(config)


model_path = "modelbert2"
model = BertForSequenceClassification.from_pretrained(model_path)
tokenizer = AutoTokenizer.from_pretrained(model_path)

tokenizer = AutoTokenizer.from_pretrained("peterkros/cofogv1-bert/modelbert2/")

# Load the label encoder
import pickle
with open('peterkros/cofogv1-bert/label_encoder.pkl', 'rb') as file:
    label_encoder = pickle.load(file)

def predict(text):
    inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True, max_length=512)
    with torch.no_grad():
        outputs = model(**inputs)
    probs = torch.nn.functional.softmax(outputs.logits, dim=-1)
    predicted_class = torch.argmax(probs, dim=-1).item()
    predicted_label = label_encoder.inverse_transform([predicted_class])[0]
    return predicted_label


# Define the markdown text with bullet points
markdown_text = """
- Trainied with ~1500 rows of data on bert-base-uncased	110M, English
- Input one budget line per time. 
- Accuracy of the model is ~72%.
"""

# Define the interface
iface = gr.Interface(
    fn=predict, 
    inputs=gr.inputs.Textbox(lines=1, placeholder="Enter Budget line here..."), 
    outputs="text", 
    title="COFOG Level 1 Classification",
    description=markdown_text  # Add the markdown text to the description
)

# Run the interface
if __name__ == "__main__":
    iface.launch()