import gradio as gr
from transformers import AutoTokenizer
import timm

title = "Finetuning [BERT] on A Financial News Sentiment Dataset"
description = """
The LLM was finetuned on a Financial News Tweet Sentiment Dataset. The documents have 3 different labels: 
 "LABEL_0": "Bearish", 
  "LABEL_1": "Bullish", 
  "LABEL_2": "Neutral"
<img src="https://huggingface.co/spaces/course-demos/Rick_and_Morty_QA/resolve/main/rick.png" width=200px>
"""

article = "Check out the dataset that [BERT cased]((https://huggingface.co/bert-base-cased?text=Paris+is+the+%5BMASK%5D+of+France.)) was [finetuned on](https://huggingface.co/datasets/zeroshot/twitter-financial-news-sentiment/viewer/zeroshot--twitter-financial-news-sentiment/train?row=9505)."


def sentiment_analyzer(tweet):
  model_reloaded = timm.create_model('hf_hub:at2507/zeroshot_finetuned_sentiment', pretrained=True)
  # model = model.load("models/at2507/zeroshot_finetuned_sentiment")
  # gr.Interface.load("models/at2507/zeroshot_finetuned_sentiment").launch()
  tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")
  zeroshotsent_model = pipeline("text-classification", model = model.to('cpu:0'), tokenizer=tokenizer)
  return zeroshotsent_model(tweet)

gr.Interface(
    fn=sentiment_analyzer,
    inputs="textbox",
    outputs="text",
    title=title,
    description=description,
    article=article,
    examples=[["CLNE, TRXC, TGE and ADMS among midday movers"], 
     ["CRISPR Therapeutics among healthcare gainers; Plus Therapeutics leads the losers"]],
).launch()