Spaces:
Running
Running
File size: 13,011 Bytes
632338f |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 |
from email.utils import parseaddr
from huggingface_hub import HfApi
import os
import datetime
import pandas as pd
import json
import evaluate as nlp_evaluate
import re
import sqlite3
import random
from tqdm import tqdm
import sys
import numpy as np
from get_exact_and_f1_score.ext_services.jsql_parser import JSQLParser
from get_exact_and_f1_score.metrics.partial_match_eval.evaluate import evaluate
random.seed(10001)
bleu = nlp_evaluate.load("bleu")
rouge = nlp_evaluate.load('rouge')
LEADERBOARD_PATH = "Exploration-Lab/BookSQL-Leaderboard"
RESULTS_PATH = "Exploration-Lab/BookSQL-Leaderboard-results"
api = HfApi()
TOKEN = os.environ.get("TOKEN", None)
YEAR_VERSION = "2024"
sqlite_path = "accounting/accounting_for_testing.sqlite"
_jsql_parser = JSQLParser.create()
def format_error(msg):
return f"<p style='color: red; font-size: 20px; text-align: center;'>{msg}</p>"
def format_warning(msg):
return f"<p style='color: orange; font-size: 20px; text-align: center;'>{msg}</p>"
def format_log(msg):
return f"<p style='color: green; font-size: 20px; text-align: center;'>{msg}</p>"
def model_hyperlink(link, model_name):
return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline;text-decoration-style: dotted;">{model_name}</a>'
def input_verification(method_name, url, path_to_file, organisation, mail):
for input in [method_name, url, path_to_file, organisation, mail]:
if input == "":
return format_warning("Please fill all the fields.")
# Very basic email parsing
_, parsed_mail = parseaddr(mail)
if not "@" in parsed_mail:
return format_warning("Please provide a valid email adress.")
if path_to_file is None:
return format_warning("Please attach a file.")
return parsed_mail
def replace_current_date_and_now(_sql, _date):
_sql = _sql.replace('current_date', "\'"+_date+"\'")
_sql = _sql.replace(', now', ", \'"+_date+"\'")
return _sql
def remove_gold_Non_exec(data,df1, sqlite_path):
con = sqlite3.connect(sqlite_path)
cur = con.cursor()
out, non_exec=[], []
new_df = df1.copy()
new_df.loc[:, 'Exec/Non-Exec'] = 0
for i,s in tqdm(enumerate(data)):
_sql = str(s).replace('"', "'").lower()
_sql = replace_current_date_and_now(_sql, '2022-06-01')
_sql = replace_percent_symbol_y(_sql)
try:
cur.execute(_sql)
res = cur.fetchall()
out.append(i)
except:
non_exec.append(i)
print("_sql: ", _sql)
new_df.loc[out, 'Exec/Non-Exec'] = 1
con.close()
return out, non_exec, new_df
def remove_data_from_index(data, ind_list):
new_data=[]
for i in ind_list:
new_data.append(data[i])
return new_data
def get_exec_match_acc(gold, pred):
assert len(gold)==len(pred)
count=0
goldd = [re.sub(' +', ' ', str(g).replace("'", '"').lower()) for g in gold]
predd = [re.sub(' +', ' ', str(p).replace("'", '"').lower()) for p in pred]
# for g, p in zip(gold, pred):
# #extra space, double quotes, lower_case
# gg = re.sub(' +', ' ', str(g).replace("'", '"').lower())
# gg = re.sub(' +', ' ', str(p).replace("'", '"').lower())
# if gold==pred:
# count+=1
goldd = _jsql_parser.translate_batch(goldd)
predd = _jsql_parser.translate_batch(predd)
pcm_f1_scores = evaluate(goldd, predd)
pcm_em_scores = evaluate(goldd, predd, exact_match=True)
_pcm_f1_scores, _pcm_em_scores=[], []
for f1, em in zip(pcm_f1_scores, pcm_em_scores):
if type(f1)==float and type(em)==float:
_pcm_f1_scores.append(f1)
_pcm_em_scores.append(em)
assert len(_pcm_f1_scores) == len(_pcm_em_scores)
jsql_error_count=0 ####JSQLError
for i, score in enumerate(pcm_f1_scores):
if type(score)==str:
jsql_error_count+=1
print("JSQLError in sql: ", jsql_error_count)
return sum(_pcm_em_scores) / len(_pcm_em_scores), sum(_pcm_f1_scores) / len(_pcm_f1_scores)
def replace_percent_symbol_y(_sql):
_sql = _sql.replace('%y', "%Y")
return _sql
def get_exec_results(sqlite_path, scores, df, flag, gold_sql_map_res={}):
con = sqlite3.connect(sqlite_path)
cur = con.cursor()
i,j,count=0,0,0
out,non_exec={},{}
new_df = df.copy()
responses=[]
for s in tqdm(scores):
_sql = str(s).replace('"', "'").lower()
_sql = replace_current_date_and_now(_sql, '2022-06-01')
_sql = replace_percent_symbol_y(_sql)
try:
cur.execute(_sql)
res = cur.fetchall()
out[i] = str(res)
except Exception as err:
non_exec[i]=err
i+=1
if flag=='g':
new_df.loc[list(out.keys()), 'GOLD_res'] = list(out.values())
# assert len(gold_sql_map_res)==count
if flag=='p':
new_df.loc[list(out.keys()), 'PRED_res'] = list(out.values())
if flag=='d':
new_df.loc[list(out.keys()), 'DEBUG_res'] = list(out.values())
con.close()
return out, non_exec, new_df
def get_scores(gold_dict, pred_dict):
exec_count, non_exec_count=0, 0
none_count=0
correct_sql, incorrect_sql = [], []
for k, res in pred_dict.items():
if k in gold_dict:
if gold_dict[k]==str(None) or str(None) in gold_dict[k]:
none_count+=1
continue
if res==gold_dict[k]:
exec_count+=1
correct_sql.append(k)
else:
non_exec_count+=1
incorrect_sql.append(k)
return exec_count, non_exec_count, none_count, correct_sql, incorrect_sql
def get_total_gold_none_count(gold_dict):
none_count, ok_count=0, 0
for k, res in gold_dict.items():
if res==str(None) or str(None) in res:
none_count+=1
else: ok_count+=1
return ok_count, none_count
def evaluate(df):
# df - [id, pred_sql]
pred_sql = df['pred_sql'].to_list()
ids = df['id'].to_list()
f = open(f"tests/test.json")
questions_and_ids = json.load(f)
ts = open(f"tests/test_sql.json")
gold_sql = json.load(ts)
gold_sql_list=[]
pred_sql_list=[]
questions_list=[]
for idx, pred in zip(ids, pred_sql):
ques = questions_and_ids[idx]['Query']
gd_sql = gold_sql[idx]['SQL']
gold_sql_list.append(gd_sql)
pred_sql_list.append(pred_sql_list)
questions_list.append(ques)
df = pd.DataFrame({'NLQ':questions_list, 'GOLD SQL':gold_sql_list, 'PREDICTED SQL':pred_sql_list})
test_size = len(df)
pred_score = df['PREDICTED SQL'].str.lower().values
# debug_score = df['DEBUGGED SQL'].str.lower().values
gold_score1 = df['GOLD SQL'].str.lower().values
print("Checking non-exec Gold sql query")
gold_exec, gold_not_exec, new_df = remove_gold_Non_exec(gold_score1, df, sqlite_path)
print("GOLD Total exec SQL query: {}/{}".format(len(gold_exec), test_size))
print("GOLD Total non-exec SQL query: {}/{}".format(len(gold_not_exec), test_size))
prev_non_exec_df = new_df[new_df['Exec/Non-Exec'] == 0]
new_df = new_df[new_df['Exec/Non-Exec']==1]
prev_non_exec_df.reset_index(inplace=True)
new_df.reset_index(inplace=True)
#Removing Non-exec sql from data
print(f"Removing {len(gold_not_exec)} non-exec sql query from all Gold/Pred/Debug")
gold_score1 = remove_data_from_index(gold_score1, gold_exec)
pred_score = remove_data_from_index(pred_score, gold_exec)
# debug_score = remove_data_from_index(debug_score, gold_exec)
gold_score = [[x] for x in gold_score1]
assert len(gold_score) == len(pred_score) #== len(debug_score)
pred_bleu_score = bleu.compute(predictions=pred_score, references=gold_score)
pred_rouge_score = rouge.compute(predictions=pred_score, references=gold_score)
pred_exact_match, pred_partial_f1_score = get_exec_match_acc(gold_score1, pred_score)
print("PREDICTED_vs_GOLD Final bleu_score: ", pred_bleu_score['bleu'])
print("PREDICTED_vs_GOLD Final rouge_score: ", pred_rouge_score['rougeL'])
print("PREDICTED_vs_GOLD Exact Match Accuracy: ", pred_exact_match)
print("PREDICTED_vs_GOLD Partial CM F1 score: ", pred_partial_f1_score)
print()
new_df.loc[:, 'GOLD_res'] = str(None)
new_df.loc[:, 'PRED_res'] = str(None)
# new_df.loc[:, 'DEBUG_res'] = str(None)
print("Getting Gold results")
# gout_res_dict, gnon_exec_err_dict, gold_sql_map_res = get_exec_results(cur, gold_score1, 'g')
gout_res_dict, gnon_exec_err_dict, new_df = get_exec_results(sqlite_path, gold_score1, new_df, 'g')
total_gold_ok_count, total_gold_none_count = get_total_gold_none_count(gout_res_dict)
print("Total Gold None count: ", total_gold_none_count)
print("Getting Pred results")
pout_res_dict, pnon_exec_err_dict, new_df = get_exec_results(sqlite_path, pred_score, new_df, 'p')
# print("Getting Debug results")
# dout_res_dict, dnon_exec_err_dict = get_exec_results(cur, debug_score, 'd')
print("GOLD Total exec SQL query: {}/{}".format(len(gold_exec), test_size))
print("GOLD Total non-exec SQL query: {}/{}".format(len(gold_not_exec), test_size))
print()
print("PRED Total exec SQL query: {}/{}".format(len(pout_res_dict), len(pred_score)))
print("PRED Total non-exec SQL query: {}/{}".format(len(pnon_exec_err_dict), len(pred_score)))
print()
# print("DEBUG Total exec SQL query: {}/{}".format(len(dout_res_dict), len(debug_score)))
# print("DEBUG Total non-exec SQL query: {}/{}".format(len(dnon_exec_err_dict), len(debug_score)))
# print()
pred_correct_exec_acc_count, pred_incorrect_exec_acc_count, pred_none_count, pred_correct_sql, pred_incorrect_sql = get_scores(gout_res_dict, pout_res_dict)
# debug_correct_exec_acc_count, debug_incorrect_exec_acc_count, debug_none_count, debug_correct_sql, debug_incorrect_sql = get_scores(gout_res_dict, dout_res_dict)
# print("PRED_vs_GOLD None_count: ", total_gold_none_count)
print("PRED_vs_GOLD Correct_Exec_count without None: ", pred_correct_exec_acc_count)
print("PRED_vs_GOLD Incorrect_Exec_count without None: ", pred_incorrect_exec_acc_count)
print("PRED_vs_GOLD Exec_Accuracy: ", pred_correct_exec_acc_count/total_gold_ok_count)
print()
return pred_exact_match, pred_correct_exec_acc_count/total_gold_ok_count, pred_partial_f1_score, pred_bleu_score['bleu'], pred_rouge_score['rougeL']
def add_new_eval(
method_name: str,
url: str,
path_to_file: str,
organisation: str,
mail: str,
):
parsed_mail = input_verification(
method_name,
url,
path_to_file,
organisation,
mail,
)
# load the file
df = pd.read_csv(path_to_file)
submission_df = pd.read_csv(path_to_file)
# modify the df to include metadata
df["Method"] = method_name
df["url"] = url
df["organisation"] = organisation
df["mail"] = parsed_mail
df["timestamp"] = datetime.datetime.now()
submission_df = pd.read_csv(path_to_file)
submission_df["Method"] = method_name
submission_df["Submitted By"] = organisation
# upload to spaces using the hf api at
path_in_repo = f"submissions/{method_name}"
file_name = f"{method_name}-{organisation}-{datetime.datetime.now().strftime('%Y-%m-%d')}.csv"
EM, EX, PCM_F1, BLEU, ROUGE = evaluate(submission_df)
submission_df['EM'] = EM
submission_df['EX'] = EX
# submission_df['PCM_F1'] = PCM_F1
submission_df['BLEU'] = BLEU
submission_df['ROUGE'] = ROUGE
# upload the df to spaces
import io
buffer = io.BytesIO()
df.to_csv(buffer, index=False) # Write the DataFrame to a buffer in CSV format
buffer.seek(0) # Rewind the buffer to the beginning
api.upload_file(
repo_id=RESULTS_PATH,
path_in_repo=f"{path_in_repo}/{file_name}",
path_or_fileobj=buffer,
token=TOKEN,
repo_type="dataset",
)
# read the leaderboard
leaderboard_df = pd.read_csv(f"submissions/baseline/baseline.csv")
# append the new submission_df csv to the leaderboard
# leaderboard_df = leaderboard_df._append(submission_df)
leaderboard_df = pd.concat([leaderboard_df, submission_df], ignore_index=True)
# save the new leaderboard
# leaderboard_df.to_csv(f"submissions/baseline/baseline.csv", index=False)
leaderboard_buffer = io.BytesIO()
leaderboard_df.to_csv(leaderboard_buffer, index=False)
leaderboard_buffer.seek(0)
api.upload_file(
repo_id=LEADERBOARD_PATH,
path_in_repo=f"submissions/baseline/baseline.csv",
path_or_fileobj=leaderboard_buffer,
token=TOKEN,
repo_type="space",
)
return format_log(
f"Method {method_name} submitted by {organisation} successfully. \nPlease refresh the leaderboard, and wait a bit to see the score displayed"
)
|