import pandas as pd
import numpy as np
import tensorflow as tf
from transformers.models.bert import BertTokenizer
from transformers import TFBertModel
import streamlit as st
import pandas as pd
from transformers import TFAutoModel
hist_loss= [0.1971,0.0732,0.0465,0.0319,0.0232,0.0167,0.0127,0.0094,0.0073,0.0058,0.0049,0.0042]
hist_acc = [0.9508,0.9811,0.9878,0.9914,0.9936,0.9954,0.9965,0.9973,0.9978,0.9983,0.9986,0.9988]
hist_val_acc = [0.9804,0.9891,0.9927,0.9956,0.9981,0.998,0.9991,0.9997,0.9991,0.9998,0.9998,0.9998]
hist_val_loss = [0.0759,0.0454,0.028,0.015,0.0063,0.0064,0.004,0.0011,0.0021,0.00064548,0.0010,0.00042896]
Epochs = [i for i in range(1,13)]
hist_loss[:] = [x * 100 for x in hist_loss]
hist_acc[:] = [x * 100 for x in hist_acc]
hist_val_acc[:] = [x * 100 for x in hist_val_acc]
hist_val_loss[:] = [x * 100 for x in hist_val_loss]
d = {'val_acc':hist_val_acc, 'acc':hist_acc,'loss':hist_loss, 'val_loss':hist_val_loss, 'Epochs': Epochs}
chart_data = pd.DataFrame(d)
chart_data.index = range(1,13)
@st.cache(suppress_st_warning=True, allow_output_mutation=True)
def load_model(show_spinner=True):
yorum_model = tf.keras.models.load_model('TC32_SavedModel')
tokenizer = BertTokenizer.from_pretrained('NimaKL/tc32_test')
return yorum_model, tokenizer
st.set_page_config(layout='wide', initial_sidebar_state='expanded')
st.markdown("TC32 Multi-Class Text Classification
Model Loss and Accuracy
", unsafe_allow_html=True)
#st.title("TC32 Multi-Class Text Classification")
#st.subheader('Model Loss and Accuracy')
st.markdown("
", unsafe_allow_html=True)
st.area_chart(chart_data, height=320)
yorum_model, tokenizer = load_model()
st.markdown("Sınıfı bulmak için bir şikayet girin. (Ctrl+Enter)
Enter complaint (in Turkish) to find the class.
", unsafe_allow_html=True)
#st.title("Sınıfı bulmak için bir şikayet girin. (Ctrl+Enter)")
#st.subheader("Enter complaint (in Turkish) to find the class.")
#st.subheader("Şikayet")
text = st.text_area("", "Bebeğim haftada bir kutu mama bitiriyor. Geçen hafta 135 tl'ye aldığım mama bugün 180 tl olmuş. Ben de artık aptamil almayacağım. Tüketici haklarına şikayet etmemiz gerekiyor. Yazıklar olsun.", height=285)
def prepare_data(input_text, tokenizer):
token = tokenizer.encode_plus(
input_text,
max_length=256,
truncation=True,
padding='max_length',
add_special_tokens=True,
return_tensors='tf'
)
return {
'input_ids': tf.cast(token.input_ids, tf.float64),
'attention_mask': tf.cast(token.attention_mask, tf.float64)
}
def make_prediction(model, processed_data, classes=['Alışveriş','Anne-Bebek','Beyaz Eşya','Bilgisayar','Cep Telefonu','Eğitim','Elektronik','Emlak ve İnşaat','Enerji','Etkinlik ve Organizasyon','Finans','Gıda','Giyim','Hizmet','İçecek','İnternet','Kamu','Kargo-Nakliyat','Kozmetik','Küçük Ev Aletleri','Medya','Mekan ve Eğlence','Mobilya - Ev Tekstili','Mücevher Saat Gözlük','Mutfak Araç Gereç','Otomotiv','Sağlık','Sigorta','Spor','Temizlik','Turizm','Ulaşım']):
probs = model.predict(processed_data)[0]
return classes[np.argmax(probs)]
if text:
with st.spinner('Wait for it...'):
processed_data = prepare_data(text, tokenizer)
result = make_prediction(yorum_model, processed_data=processed_data)
#st.markdown("
", unsafe_allow_html=True)
st.success(result)
# description = '|
'+result+'
|
'
# st.markdown(description, unsafe_allow_html=True)