app.py

import os
import gradio as gr

import numpy as np
import torch
from transformers import AutoTokenizer, Pipeline

from inference_utils import prepare_stance_texts
from models import StanceEncoderModel

CLASS_DICT = {0: 'FAVOR', 1: 'AGAINST', 2: 'NEITHER'}

params = {'lang': 'pl',
          'masked_lm_prompt': 4, }


class StancePipeline(Pipeline):

    def _sanitize_parameters(self, **pipeline_parameters):
        return pipeline_parameters, {}, {}

    def preprocess(self, input):
        prompt_text, prompt_target = prepare_stance_texts([input['text'], ], [input['target'], ], params,
                                                          self.tokenizer)
        inputs = self.tokenizer(prompt_text, prompt_target, return_tensors="pt", padding=True, truncation='only_first')
        return {'input_ids': inputs['input_ids'], 'attention_mask': inputs['attention_mask'],
                'sequence_ids': torch.tensor((np.array(inputs.sequence_ids()) == 1).astype(int)).unsqueeze(0)}

    def _forward(self, model_inputs):
        outputs = self.model(**model_inputs)
        return outputs

    def postprocess(self, model_outputs):
        probas = model_outputs["logits"].softmax(-1)
        score = probas.max(-1)[0].item()
        return {'stance': CLASS_DICT[probas.argmax(-1).item()], 'score': score}


model = StanceEncoderModel.from_pretrained('clarin-knext/stance-pl-1',
                                           use_auth_token=os.environ['TOKEN'])
tokenizer = AutoTokenizer.from_pretrained('clarin-knext/stance-pl-1',
                                          use_auth_token=os.environ['TOKEN'])
pipeline = StancePipeline(model=model, tokenizer=tokenizer, batch_size=1)


def predict(text, target):
    predictions = pipeline({'text': text, 'target': target})
    return f'{predictions["stance"]} ({predictions["score"]:.3f})'


gradio_app = gr.Interface(
    predict,
    inputs=[gr.TextArea(label="Text", placeholder="Enter text here..."),
            gr.Textbox(label="Target", placeholder="Enter stance target here...")],
    outputs=[gr.Label(label="Stance class")],
    title="Polish stance detection",
)

if __name__ == "__main__":
    gradio_app.launch()