from transformers import AutoTokenizer, AutoModelForSequenceClassification, AutoConfig
import gradio as gr
from torch.nn import functional as F
import seaborn
import matplotlib
import platform
from transformers.file_utils import ModelOutput
if platform.system() == "Darwin":
    print("MacOS")
    matplotlib.use('Agg')
import matplotlib.pyplot as plt
import io
from PIL import Image
import matplotlib.font_manager as fm

# global var
MODEL_NAME = 'yseop/distilbert-base-financial-relation-extraction'
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForSequenceClassification.from_pretrained(MODEL_NAME)
config = AutoConfig.from_pretrained(MODEL_NAME)
MODEL_BUF = {
    "name": MODEL_NAME,
    "tokenizer": tokenizer,
    "model": model,
    "config": config
}
font_dir = ['./']
for font in fm.findSystemFonts(font_dir):
    print(font)
    fm.fontManager.addfont(font)
plt.rcParams["font.family"] = 'NanumGothicCoding'

def change_model_name(name):
    MODEL_BUF["name"] = name
    MODEL_BUF["tokenizer"] = AutoTokenizer.from_pretrained(name)
    MODEL_BUF["model"] = AutoModelForSequenceClassification.from_pretrained(name)
    MODEL_BUF["config"] = AutoConfig.from_pretrained(name)
def predict(model_name, text):
    if model_name != MODEL_NAME:
        change_model_name(model_name)
    
    tokenizer = MODEL_BUF["tokenizer"]
    model = MODEL_BUF["model"]
    config = MODEL_BUF["config"]
    tokenized_text = tokenizer([text], return_tensors='pt')
    model.eval()
    output, attention = model(**tokenized_text, output_attentions=True, return_dict=False)
    output = F.softmax(output, dim=-1)
    result = {}
    
    for idx, label in enumerate(output[0].detach().numpy()):
        result[config.id2label[idx]] = float(label)
    return result
if __name__ == '__main__':
    text = 'An A-B trust is a joint trust created by a married couple for the purpose of minimizing estate taxes.'
    model_name_list = [
        'yseop/distilbert-base-financial-relation-extraction'
    ]
    #Create a gradio app with a button that calls predict()
    app = gr.Interface(
        fn=predict,
        inputs=[gr.inputs.Dropdown(model_name_list, label="Model Name"), 'text'], outputs=['label'], 
        examples = [[MODEL_BUF["name"], text]],
        title="FReE",
        description="Financial relations classifier"
        )
    app.launch(inline=False)