-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathdemo_prepare.py
162 lines (133 loc) · 5.49 KB
/
demo_prepare.py
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
149
150
151
152
153
154
155
156
157
158
159
import openai
from utils import dispatch_openai_api_requests, return_args, save_json, read_json
from data_prepare import data_prepare, data_prepare_split
from prompts import role_str_demo_prepare, demos_descri, demos_descri_hint
from tqdm import tqdm
import json
import random
config = return_args()
target_dict = {'line':0, 'scatter':1, 'bar':2, 'box':3}
threshold = 0.05
summary_save_dict = read_json("output/final_summary_save_dict_"+str(config["seed"])+".json")
demo_data_list, test_data_list, selected_keys = data_prepare_split(config['seed'])
demo_data_list = demo_data_list[:]
dpi_dict = {}
demo_pool = []
for iter_ in range (6):
prompt_list = []
new_data_list = []
count = 0
total = 0
print (f'\n==Stage 1, iter:{iter_}==')
for d_i, data_elem in enumerate(demo_data_list[:]):
total += 1
feature_dict = {}
for k in selected_keys:
if k == 'trace_type':
continue
feature_dict[k] = data_elem[k]
target = data_elem['trace_type']
if target.lower() == 'line': #{'line chart': 0.5, 'scatter plot': 0, 'bar chart': 0, 'box plot': 1}
target_str = 'line chart'
elif target.lower() == 'scatter':
target_str = 'scatter plot'
elif target.lower() == 'bar':
target_str = 'bar chart'
elif target.lower() == 'box':
target_str = 'box plot'
fid = data_elem['fid']
target_idx = target_dict[target.lower()]
elem_fe_de = summary_save_dict[fid]
if fid not in dpi_dict:
dpi_dict[fid] = {'final_step':0, 'stop':0, 'iters':{}}
if dpi_dict[fid]['stop'] == 1:
dip_elem_x = dpi_dict[fid]
demos_descri.format(elem_fe_de) + "\n" +dip_elem_x['iters'][dip_elem_x['final_step']]['predctions'] + "\n\n\n"
demo_pool.append(demos_descri)
continue
dpi_dict[fid]['iters'][dpi_dict[fid]['final_step']] = {}
messages=[
{"role": "system", "content": role_str_demo_prepare},
]
if dpi_dict[fid]['final_step'] == 0:
data_f_prompt = demos_descri.format( elem_fe_de)
messages.append({"role": "user", "content": data_f_prompt})
else:
# pred_answer_x = dpi_dict[fid]['final_step']
pred_answer = dpi_dict[fid]['iters'][dpi_dict[fid]['final_step']-1]['pred_answer']
predctions_json = dpi_dict[fid]['iters'][dpi_dict[fid]['final_step']-1]['predctions_json']
data_f_prompt = demos_descri_hint.format(target_str, \
pred_answer[1][0].lower(), \
predctions_json, elem_fe_de)
if iter_ > 3:
random.seed(config['seed'])
random.shuffle(demo_pool)
demo_str_x = demo_pool[0]
data_f_prompt = demo_str_x+ data_f_prompt
messages.append({"role": "user", "content": data_f_prompt})
dpi_dict[fid]['iters'][dpi_dict[fid]['final_step']]['demos_descri'] = demos_descri
dpi_dict[fid]['iters'][dpi_dict[fid]['final_step']]['elem_fe_de'] = elem_fe_de
prompt_list.append(messages)
new_data_list.append(data_elem)
if len(prompt_list) == 0:
break
print (f"number of prompts: {len(prompt_list)}")
openai_responses = dispatch_openai_api_requests(prompt_list, len(prompt_list), api_batch=int(config['api_batch']), api_model_name = "gpt-3.5-turbo")
print (f'\n==Stage 2, iter:{iter_}==')
for d_i, data_elem in tqdm(enumerate(new_data_list[:])):
target = data_elem['trace_type']
if target.lower() == 'line':
target_str = 'line chart'
elif target.lower() == 'scatter':
target_str = 'scatter plot'
elif target.lower() == 'bar':
target_str = 'bar chart'
elif target.lower() == 'box':
target_str = 'box plot'
fid = data_elem['fid']
target_idx = target_dict[target.lower()]
openai_response = openai_responses[d_i]
response = openai_response['choices'][0]['message']['content']
demos_str_x = []
for xxx in response.split('.'):
if 'hint' in xxx.lower() or 'previous' in xxx.lower():
pass
else:
demos_str_x.append(xxx)
predctions = ".".join(demos_str_x)
start_index = predctions.find('{')
end_index = predctions.find('}') + 1
# Extract the JSON substring
json_str = predctions[start_index:end_index]
# Parse the JSON string into a dictionary
try:
predctions_json = json.loads(json_str)
predctions_json['line chart']
predctions_json['scatter plot']
predctions_json['bar chart']
predctions_json['box plot']
except:
predctions_json = {'line chart':0.25, 'scatter plot':0.25, 'bar chart':0.25, 'box plot':0.25}
print ('predctions_json:',predctions_json, target_str)
pred_answer = list(sorted(predctions_json.items(), key=lambda x: float(x[-1])))[-2:]
dpi_dict[fid]['iters'][dpi_dict[fid]['final_step']]['pred_answer'] = pred_answer
dpi_dict[fid]['iters'][dpi_dict[fid]['final_step']]['predctions_json'] = predctions_json
dpi_dict[fid]['iters'][dpi_dict[fid]['final_step']]['predctions'] = predctions
if target_str.lower() == pred_answer[1][0].lower() and pred_answer[1][1] - pred_answer[0][1]>threshold:
dpi_dict[fid]['stop'] = 1
# print ('stop')
else:
dpi_dict[fid]['final_step'] +=1
# print ('nonstop')
dip_save_dict = {}
for fid, dip_elem in dpi_dict.items():
if dip_elem['stop'] == 1:
xx = dip_elem['iters'][dip_elem['final_step']]
dip_save_dict[fid] = [xx['demos_descri'], xx['elem_fe_de'], xx['predctions']]
save_json(dpi_dict, "output/final_dip_hint_"+str(config['seed'])+"_all.json")
save_json(dip_save_dict, "output/final_dip_hint_"+str(config['seed'])+"_use.json")
print ("save done!")
# [demos_descri, elem_fe_de, predctions]
# print ('---',d_i)
# print (response)
# print ()