Hugging Face's logo

Spaces:
andreyunic23
/
beds_pipeline_beds
Sleeping

App Files Community
beds_pipeline_beds
File size: 5,191 Bytes
76d9a63
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
#bibliotecas
import pandas as pd
import numpy as np
import torch
from torch import cuda
from torch.nn import functional as F
#from sklearn.model_selection import train_test_split
import transformers
from transformers import (
    AutoTokenizer,
    AutoModelForSequenceClassification,
)
from sentence_transformers import SentenceTransformer


#classes e funcs

#parte 1 ###########################################################################################################
#parte 1 ###########################################################################################################

def convert_label(lista):
  for x in range(len(lista)):
    curr = lista[x]
    lista[x] = 0 if curr =='loss' else 1 if curr == 'hazard' else 2# if curr == 'constraint' else 3
  return lista

def df_with_pred(labels, predictions, data):
    lista = []
    cont = 0
    #predicted = np.argmax(results.logits.cpu(), axis=-1)
    for test,pred in zip(labels, predictions):
        lista.append([data.id.iloc[cont],data.req.iloc[cont],test,pred.item()])
        cont += 1
    return pd.DataFrame(lista, columns=['id','req', 'label', 'pred'])

#parte 2 ###########################################################################################################
#parte 2 ###########################################################################################################

# def organize_predictions_list(predicted, data):#data : ['id','req', 'label', 'pred']
#   list_loss = []
#   list_hazard = []
#   list_constraint = []
#   for x in range(len(predicted)):
#     if(predicted[x] == 0):
#       list_loss.append([data.id.iloc[x], data.req.iloc[x]])
#     elif(predicted[x] == 1):
#       list_hazard.append([data.id.iloc[x], data.req.iloc[x]])
#     elif(predicted[x] == 2):
#       list_constraint.append([data.id.iloc[x], data.req.iloc[x]])
#   return  pd.DataFrame(list_loss, columns=['id','req']),  pd.DataFrame(list_hazard, columns=['id','req']),  pd.DataFrame(list_constraint, columns=['id','req'])

def organize_step2_predictions(predictions, list_sentences):
  list_correct = []
  list_incorrect = []

  for prediction, sentence in zip(predictions, list_sentences):
    if prediction == 0:
      list_correct.append(sentence)
    else:
      list_incorrect.append(sentence)

  return list_correct, list_incorrect


def get_incorrect(predicted, data): #data : [id, req]
  list_incorrect = []
  for x in range(len(predicted)):
    if predicted[x] == 1:
      list_incorrect.append([data.id.iloc[x],data.req.iloc[x]])
  return pd.DataFrame(list_incorrect,columns=['id','req'])


#parte 3 ###########################################################################################################
#parte 3 ###########################################################################################################

def format_examples(df):
  examples = []
  for sentence in df:
    examples.append([sentence,sentence])
  return examples

def check_similarity_return(list_incorrect, list_correct, model):
  embeddings = model.encode(list_correct)
  for x in range(len(list_incorrect)):
    id = list_incorrect.id.iloc[x]
    sentence = list_incorrect.req.iloc[x]
    sentence = model.encode(sentence)
    similarity = model.similarity(sentence, embeddings)
    sim_pair = []

    for sim,correct in zip(similarity[0].tolist(), list_correct):
      sim_pair.append([id, sim, correct[0]])

    sim_pair.sort(key=lambda x: x[0])
    sim_pair.reverse()

  return sim_pair[:10]

def check_similarity_return2(list_incorrect, list_correct, model):
  sim_pair = []
  embeddings = model.encode(list_correct)

  for x in range(len(list_incorrect)):
    id = list_incorrect.id.iloc[x]
    sentence = list_incorrect.req.iloc[x]
    sentence = model.encode(sentence)
    similarity = model.similarity(sentence, embeddings)
    temp_list = []
    for sim,correct in zip(similarity[0].tolist(), list_correct):
      temp_list.append([id, sim, correct[0]])

    temp_list.sort(key=lambda x: x[1])
    temp_list.reverse()

    sim_pair.extend(temp_list[:10])
  # print(sim_pair)
  return sim_pair
#parte 4 ###########################################################################################################
#parte 4 ###########################################################################################################


def list_erro_with_pred(results, data, sub):
    diff_label = []
    cont = 0
    predicted = np.argmax(results.logits.cpu(), axis=-1)
    probabilidade = F.softmax(results.logits.cpu(), dim=-1)
    for id,req,pred,prob in zip(data.id, data.req, predicted, probabilidade):
        # print(pred)
        # print(sub[pred.item()])
        # print(prob.tolist())
        #diff_label.append([id,req,sub[pred.item()],prob.tolist()])
        diff_label.append([id,req,pred.item(),prob.tolist()])
        cont+=1
    return diff_label

########################################################################
########################################################################
########################################################################
########################################################################




###########################################