File size: 3,375 Bytes
df6828f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
import os
import streamlit as st
import torch
import string
from transformers import BertTokenizer, BertForMaskedLM

st.set_page_config(page_title='Next Word Prediction Model', page_icon=None, layout='centered', initial_sidebar_state='auto')

@st.cache()
def load_model(model_name):
  try:
    if model_name.lower() == "bert":
      bert_tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
      bert_model = BertForMaskedLM.from_pretrained('bert-base-uncased').eval()
      return bert_tokenizer,bert_model
  except Exception as e:
    pass

#use joblib to fast your function

def decode(tokenizer, pred_idx, top_clean):
  ignore_tokens = string.punctuation + '[PAD]'
  tokens = []
  for w in pred_idx:
    token = ''.join(tokenizer.decode(w).split())
    if token not in ignore_tokens:
      tokens.append(token.replace('##', ''))
  return '\n'.join(tokens[:top_clean])

def encode(tokenizer, text_sentence, add_special_tokens=True):
  text_sentence = text_sentence.replace('<mask>', tokenizer.mask_token)
    # if <mask> is the last token, append a "." so that models dont predict punctuation.
  if tokenizer.mask_token == text_sentence.split()[-1]:
    text_sentence += ' .'

    input_ids = torch.tensor([tokenizer.encode(text_sentence, add_special_tokens=add_special_tokens)])
    mask_idx = torch.where(input_ids == tokenizer.mask_token_id)[1].tolist()[0]
  return input_ids, mask_idx

def get_all_predictions(text_sentence, top_clean=5):
    # ========================= BERT =================================
  input_ids, mask_idx = encode(bert_tokenizer, text_sentence)
  with torch.no_grad():
    predict = bert_model(input_ids)[0]
  bert = decode(bert_tokenizer, predict[0, mask_idx, :].topk(top_k).indices.tolist(), top_clean)
  return {'bert': bert}

def get_prediction_eos(input_text):
  try:
    input_text += ' <mask>'
    res = get_all_predictions(input_text, top_clean=int(top_k))
    return res
  except Exception as error:
    pass

try:

  st.markdown("<h1 style='text-align: center;'>Next Word Prediction</h1>", unsafe_allow_html=True)
  st.markdown("<h4 style='text-align: center; color: #B2BEB5;'><i>Keywords  : BertTokenizer, BertForMaskedLM, Pytorch</i></h4>", unsafe_allow_html=True)

  st.sidebar.text("Next Word Prediction Model")
  top_k = st.sidebar.slider("Select How many words do you need", 1 , 25, 1) #some times it is possible to have less words
  print(top_k)
  model_name = st.sidebar.selectbox(label='Select Model to Apply',  options=['BERT', 'XLNET'], index=0,  key = "model_name")

  bert_tokenizer, bert_model  = load_model(model_name) 
  input_text = st.text_area("Enter your text here")

  #click outside box of input text to get result
  res = get_prediction_eos(input_text)

  answer = []
  print(res['bert'].split("\n"))
  for i in res['bert'].split("\n"):
  	answer.append(i)
  answer_as_string = "    ".join(answer)
  st.text_area("Predicted List is Here",answer_as_string,key="predicted_list") 
  st.image('https://freepngimg.com/download/keyboard/6-2-keyboard-png-file.png',use_column_width=True)
  st.markdown("<h6 style='text-align: center; color: #808080;'>Created By <a href='https://github.com/7Vivek'>Vivek</a> - Checkout complete project <a href='https://github.com/7Vivek/Next-Word-Prediction-Streamlit'>here</a></h6>", unsafe_allow_html=True)

except Exception as e:
  print("SOME PROBLEM OCCURED")