app.py

import gradio as gr
import torch
from minicons import cwe
from huggingface_hub import hf_hub_download
import os

from model import FFNModule, FeatureNormPredictor, FFNParams, TrainingParams


def predict (Word, Sentence, LM, Layer, Norm):
    models = {'bert': 'bert-base-uncased'}
    if Word not in Sentence: return "invalid input: word not in sentence"
    model_name = LM + str(Layer) + '_to_' + Norm
    lm = cwe.CWE(models[LM])
    if Layer not in range (lm.layers): return "invalid input: layer not in lm"

    model_path = hf_hub_download("jwalanthi/bert_layer8_to_binder", model_name+".ckpt", use_auth_token=os.environ['TOKEN'])
    label_path = hf_hub_download("jwalanthi/bert_layer8_to_binder", model_name+".txt", use_auth_token=os.environ['TOKEN'])

    # labels = "These are some fake features".split(" ")
    # vals = np.random.randint(-10,10,(5))
    # return model_name+" \n"+"\n".join([labels[i]+" "+str(vals[i]) for i in range(len(labels)) if vals[i]>0])
    model = FeatureNormPredictor.load_from_checkpoint(
        checkpoint_path=model_path,
        map_location=None
    )
    model.eval()

    with open (label_path, "r") as file:
        labels = [line.rstrip() for line in file.readlines()]

    data = (Sentence, Word)
    emb = lm.extract_representation(data, layer=8)
    pred = torch.nn.functional.relu(model(emb))
    pred = pred.squeeze(0)
    pred_list = pred.detach().numpy().tolist()
    
    output = [labels[i]+'\t'+str(pred_list[i]) for i in range(len(labels)) if pred_list[i] > 0.0]
    return "All Positive Predicted Values:\n"+"\n".join(output)

demo = gr.Interface(
    fn=predict,
    inputs=[
        "text", 
        "text", 
        gr.Radio(["bert", "albert"]),
        "number",
        gr.Radio(["binder", "mcrae", "buchanan"]),
    ],
    outputs=["text"],
)

demo.launch()

if __name__ == "__main__":
    demo.launch(share=True)