Spaces:
Running
Running
File size: 66,307 Bytes
600117d aff96fc 50853e6 d53141f 11885ff 50853e6 d53141f 50853e6 d53141f 50853e6 d53141f 50853e6 d53141f 50853e6 d53141f 50853e6 d53141f 50853e6 d53141f 50853e6 d53141f 50853e6 600117d aff96fc 600117d aff96fc 50853e6 d53141f 11885ff d53141f 11885ff d53141f 11885ff d53141f 11885ff d53141f 11885ff d53141f aff96fc 11885ff e7a189a 11885ff aff96fc 11885ff d53141f 11885ff d53141f 11885ff d53141f 11885ff d53141f 11885ff 62c2281 11885ff 4c8b2a6 62c2281 4c8b2a6 62c2281 4c8b2a6 11885ff aff96fc d53141f 50853e6 d53141f 50853e6 d53141f 50853e6 d53141f 50853e6 d53141f 50853e6 e56d66b 50853e6 e5ac68d 4c8b2a6 50853e6 11885ff d53141f 11885ff 50853e6 11885ff aff96fc 11885ff aff96fc 11885ff f91d95b 8c5ff69 de0f30c 7079c78 f91d95b 11885ff e7a189a a638e77 e7a189a 1ded406 e7a189a 1ded406 e7a189a 1ded406 e7a189a 1ded406 11885ff 1ded406 e56d66b 1ded406 e56d66b 1ded406 aff96fc d53141f aff96fc e7a189a aff96fc db0280f 4c8b2a6 aff96fc e7a189a d53141f 1ded406 d53141f 1ded406 d53141f 1ded406 aff96fc 1ded406 aff96fc 1ded406 e56d66b 1ded406 d53141f 1ded406 aff96fc 1ded406 50853e6 1ded406 50853e6 1ded406 50853e6 1ded406 bf8bb9c 1ded406 d53141f 1ded406 11885ff 1ded406 11885ff 1ded406 11885ff 1ded406 e56d66b 1ded406 50853e6 1ded406 50853e6 1ded406 aff96fc 1ded406 aff96fc 1ded406 aff96fc 1ded406 aff96fc e56d66b 1ded406 e56d66b 1ded406 4c8b2a6 1ded406 e56d66b aff96fc e56d66b aff96fc f91d95b 57a1eca da299cd 57a1eca |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 |
import gradio as gr
import torch
from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
import pandas as pd
import json
import io
import csv
from typing import List, Dict
import threading
import time
import queue
from concurrent.futures import ThreadPoolExecutor, as_completed
import asyncio
# Global model cache and loading status
MODEL_CACHE = {}
MODEL_LOADING_STATUS = {}
MODEL_LOADING_LOCK = threading.Lock()
def check_model_loading_status(model_names: List[str]) -> Dict:
"""Check loading status of multiple models"""
with MODEL_LOADING_LOCK:
status = {}
for model_name in model_names:
if model_name in MODEL_CACHE:
status[model_name] = "ready"
elif model_name in MODEL_LOADING_STATUS:
status[model_name] = MODEL_LOADING_STATUS[model_name]
else:
status[model_name] = "not_loaded"
return status
def load_model_with_status_tracking(model_name: str):
"""Load model with status tracking"""
with MODEL_LOADING_LOCK:
if model_name in MODEL_CACHE:
return MODEL_CACHE[model_name], None
if model_name in MODEL_LOADING_STATUS:
return None, f"โมเดล {model_name} กำลังโหลดอยู่..."
MODEL_LOADING_STATUS[model_name] = "loading"
try:
print(f"🔄 เริ่มโหลดโมเดล {model_name}...")
# Update status
with MODEL_LOADING_LOCK:
MODEL_LOADING_STATUS[model_name] = "downloading"
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
with MODEL_LOADING_LOCK:
MODEL_LOADING_STATUS[model_name] = "loading_model"
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype=torch.float16,
device_map="auto",
trust_remote_code=True
)
with MODEL_LOADING_LOCK:
MODEL_LOADING_STATUS[model_name] = "creating_pipeline"
generator = pipeline("text-generation", model=model, tokenizer=tokenizer)
with MODEL_LOADING_LOCK:
MODEL_CACHE[model_name] = generator
MODEL_LOADING_STATUS[model_name] = "ready"
print(f"✅ โหลดโมเดล {model_name} สำเร็จ")
return generator, None
except Exception as e:
error_msg = f"❌ ไม่สามารถโหลดโมเดล {model_name}: {str(e)}"
print(error_msg)
with MODEL_LOADING_LOCK:
if model_name in MODEL_LOADING_STATUS:
del MODEL_LOADING_STATUS[model_name]
return None, error_msg
def preload_models_async(model_names: List[str], progress_callback=None):
"""Preload models asynchronously"""
def load_single_model(model_name):
generator, error = load_model_with_status_tracking(model_name)
if progress_callback:
progress_callback(model_name, "ready" if generator else "error", error)
return model_name, generator, error
results = {}
with ThreadPoolExecutor(max_workers=2) as executor: # Limit concurrent loading
futures = {executor.submit(load_single_model, model): model for model in model_names}
for future in as_completed(futures):
model_name, generator, error = future.result()
results[model_name] = {"generator": generator, "error": error}
return results
# Predefined task templates with Thai language support
TASK_TEMPLATES = {
"text_generation": {
"name": "การสร้างข้อความ (Text Generation)",
"template": "เขียนเรื่องราวสร้างสรรค์เกี่ยวกับ {topic}",
"description": "สร้างข้อความสร้างสรรค์ภาษาไทยจากหัวข้อที่กำหนด"
},
"question_answering": {
"name": "คำถาม-คำตอบ (Question Answering)",
"template": "คำถาม: {question}\nคำตอบ:",
"description": "สร้างคู่คำถาม-คำตอบภาษาไทย"
},
"summarization": {
"name": "การสรุปข้อความ (Text Summarization)",
"template": "สรุปข้อความต่อไปนี้: {text}",
"description": "สร้างตัวอย่างการสรุปข้อความภาษาไทย"
},
"translation": {
"name": "การแปลภาษา (Translation)",
"template": "แปลจาก {source_lang} เป็น {target_lang}: {text}",
"description": "สร้างคู่ข้อมูลสำหรับการแปลภาษา"
},
"classification": {
"name": "การจำแนกข้อความ (Text Classification)",
"template": "จำแนกอารมณ์ของข้อความนี้: {text}\nอารมณ์:",
"description": "สร้างตัวอย่างการจำแนกอารมณ์หรือหมวดหมู่ของข้อความ"
},
"conversation": {
"name": "บทสนทนา (Conversation)",
"template": "มนุษย์: {input}\nผู้ช่วย:",
"description": "สร้างข้อมูลบทสนทนาภาษาไทย"
},
"instruction_following": {
"name": "การทำตามคำสั่ง (Instruction Following)",
"template": "คำสั่ง: {instruction}\nการตอบสนอง:",
"description": "สร้างคู่คำสั่ง-การตอบสนองภาษาไทย"
},
"thai_poetry": {
"name": "กวีนิพนธ์ไทย (Thai Poetry)",
"template": "แต่งกวีนิพนธ์เกี่ยวกับ {topic} ในรูปแบบ {style}",
"description": "สร้างกวีนิพนธ์ไทยในรูปแบบต่างๆ"
},
"thai_news": {
"name": "ข่าวภาษาไทย (Thai News)",
"template": "เขียนข่าวภาษาไทยเกี่ยวกับ {topic} ในหัวข้อ {category}",
"description": "สร้างข้อความข่าวภาษาไทยในหมวดหมู่ต่างๆ"
}
}
# Thai language models from Hugging Face
THAI_MODELS = {
"typhoon-7b": {
"name": "🌪️ Typhoon-7B (SCB10X)",
"model_id": "scb10x/typhoon-7b",
"description": "โมเดลภาษาไทยขนาด 7B พารามิเตอร์ ประสิทธิภาพสูง"
},
"openthaigpt": {
"name": "🇹🇭 OpenThaiGPT 1.5-7B",
"model_id": "openthaigpt/openthaigpt1.5-7b-instruct",
"description": "โมเดลภาษาไทยรองรับคำสั่งและบทสนทนาหลายรอบ"
},
"wangchanlion": {
"name": "🦁 Gemma2-9B WangchanLION",
"model_id": "aisingapore/Gemma2-9b-WangchanLIONv2-instruct",
"description": "โมเดลขนาด 9B รองรับไทย-อังกฤษ พัฒนาโดย AI Singapore"
},
"sambalingo": {
"name": "🌍 SambaLingo-Thai-Base",
"model_id": "sambanovasystems/SambaLingo-Thai-Base",
"description": "โมเดลภาษาไทยพื้นฐาน รองรับทั้งไทยและอังกฤษ"
},
"other": {
"name": "🔧 โมเดลอื่นๆ (Custom)",
"model_id": "custom",
"description": "ระบุชื่อโมเดลที่ต้องการใช้งานเอง"
}
}
def load_file_data(file_path: str) -> List[Dict]:
"""Load data from uploaded file"""
try:
if file_path.endswith('.csv'):
df = pd.read_csv(file_path)
return df.to_dict('records')
elif file_path.endswith('.json'):
with open(file_path, 'r', encoding='utf-8') as f:
return json.load(f)
elif file_path.endswith('.txt'):
with open(file_path, 'r', encoding='utf-8') as f:
lines = f.readlines()
return [{'text': line.strip()} for line in lines if line.strip()]
else:
raise ValueError("Unsupported file format. Use CSV, JSON, or TXT files.")
except Exception as e:
raise Exception(f"Error reading file: {str(e)}")
def generate_from_template(template: str, data_row: Dict) -> str:
"""Generate prompt from template and data"""
try:
return template.format(**data_row)
except KeyError as e:
return f"Template error: Missing field {e}"
def load_model(model_name):
"""Load a Hugging Face model for text generation"""
try:
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)
generator = pipeline("text-generation", model=model, tokenizer=tokenizer)
return generator, None
except Exception as e:
return None, str(e)
def generate_dataset(model_name, prompt_template, num_samples, max_length, temperature, top_p):
"""Generate dataset using Hugging Face model"""
try:
generator, error = load_model(model_name)
if error:
return None, f"Error loading model: {error}"
dataset = []
for i in range(num_samples):
# Generate text
generated = generator(
prompt_template,
max_length=max_length,
temperature=temperature,
top_p=top_p,
num_return_sequences=1,
do_sample=True
)
generated_text = generated[0]['generated_text']
dataset.append({
'id': i + 1,
'prompt': prompt_template,
'generated_text': generated_text,
'full_text': generated_text
})
# Convert to DataFrame for display
df = pd.DataFrame(dataset)
# Create downloadable files
csv_data = df.to_csv(index=False)
json_data = json.dumps(dataset, indent=2, ensure_ascii=False)
return df, csv_data, json_data, None
except Exception as e:
return None, None, None, f"Error generating dataset: {str(e)}"
def generate_dataset_from_task(model_name, task_type, custom_template, file_data, num_samples, max_length, temperature, top_p):
"""Generate dataset using task templates or file input"""
try:
generator, error = load_model(model_name)
if error:
return None, f"Error loading model: {error}"
dataset = []
# Determine the template to use
if custom_template and custom_template.strip():
template = custom_template
elif task_type in TASK_TEMPLATES:
template = TASK_TEMPLATES[task_type]["template"]
else:
template = "Generate text: {input}"
# Generate samples
for i in range(num_samples):
if file_data and len(file_data) > 0:
# Use file data cyclically
data_row = file_data[i % len(file_data)]
prompt = generate_from_template(template, data_row)
else:
# Use template with placeholder values
prompt = template.replace("{topic}", "artificial intelligence") \
.replace("{question}", "What is machine learning?") \
.replace("{text}", "Sample text for processing") \
.replace("{input}", f"Sample input {i+1}") \
.replace("{instruction}", f"Complete this task {i+1}")
# Generate text
generated = generator(
prompt,
max_length=max_length,
temperature=temperature,
top_p=top_p,
num_return_sequences=1,
do_sample=True,
pad_token_id=generator.tokenizer.eos_token_id
)
generated_text = generated[0]['generated_text']
dataset.append({
'id': i + 1,
'task_type': task_type,
'prompt': prompt,
'generated_text': generated_text,
'original_data': data_row if file_data else None
})
# Convert to DataFrame for display
df = pd.DataFrame(dataset)
# Create downloadable files
csv_data = df.to_csv(index=False)
json_data = json.dumps(dataset, indent=2, ensure_ascii=False)
return df, csv_data, json_data, None
except Exception as e:
return None, None, None, f"Error generating dataset: {str(e)}"
# Multi-model generation status tracking
class ModelStatus:
def __init__(self):
self.models = {}
self.record_status = {} # record_id: {"status": "pending/processing/completed", "model": "model_name"}
self.completed_records = []
self.lock = threading.Lock()
def set_record_processing(self, record_id: int, model_name: str):
with self.lock:
self.record_status[record_id] = {"status": "processing", "model": model_name}
def set_record_completed(self, record_id: int, result: dict):
with self.lock:
self.record_status[record_id]["status"] = "completed"
self.completed_records.append(result)
def get_next_available_record(self, total_records: int, model_name: str) -> int:
with self.lock:
for i in range(total_records):
if i not in self.record_status or self.record_status[i]["status"] == "pending":
self.record_status[i] = {"status": "pending", "model": model_name}
return i
return -1 # No available records
def get_progress(self, total_records: int) -> dict:
with self.lock:
completed = len([r for r in self.record_status.values() if r["status"] == "completed"])
processing = len([r for r in self.record_status.values() if r["status"] == "processing"])
return {
"completed": completed,
"processing": processing,
"total": total_records,
"percentage": (completed / total_records * 100) if total_records > 0 else 0
}
def load_model_with_cache(model_name: str, cache: dict):
"""Load model with caching and progress feedback"""
if model_name in cache:
return cache[model_name], None
try:
print(f"🔄 กำลังโหลดโมเดล {model_name}...")
# Use smaller models or quantized versions for faster loading
if "typhoon" in model_name.lower():
# Load with optimizations
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype=torch.float16, # Use half precision
device_map="auto",
trust_remote_code=True
)
else:
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype=torch.float16,
device_map="auto"
)
generator = pipeline("text-generation", model=model, tokenizer=tokenizer)
cache[model_name] = generator
print(f"✅ โหลดโมเดล {model_name} สำเร็จ")
return generator, None
except Exception as e:
error_msg = f"❌ ไม่สามารถโหลดโมเดล {model_name}: {str(e)}"
print(error_msg)
return None, error_msg
def generate_single_record(generator, prompt: str, record_id: int, model_name: str,
max_length: int, temperature: float, top_p: float,
task_type: str, original_data: dict, status_tracker: ModelStatus):
"""Generate a single record with the given model"""
try:
# Mark record as processing
status_tracker.set_record_processing(record_id, model_name)
# Generate text
generated = generator(
prompt,
max_length=max_length,
temperature=temperature,
top_p=top_p,
num_return_sequences=1,
do_sample=True,
pad_token_id=generator.tokenizer.eos_token_id if hasattr(generator.tokenizer, 'eos_token_id') else generator.tokenizer.pad_token_id
)
generated_text = generated[0]['generated_text']
result = {
'id': record_id + 1,
'model_used': model_name,
'task_type': task_type,
'prompt': prompt,
'generated_text': generated_text,
'original_data': original_data,
'generation_time': time.time()
}
# Mark record as completed
status_tracker.set_record_completed(record_id, result)
return result
except Exception as e:
# If generation fails, mark as pending again for other models to try
with status_tracker.lock:
if record_id in status_tracker.record_status:
status_tracker.record_status[record_id]["status"] = "pending"
return None
def model_worker(model_name: str, model_cache: dict, prompts: List[str],
task_type: str, original_data_list: List[dict],
max_length: int, temperature: float, top_p: float,
status_tracker: ModelStatus, progress_callback=None):
"""Worker function for each model to process available records"""
# Load model
generator, error = load_model_with_cache(model_name, model_cache)
if error:
return f"Error loading {model_name}: {error}"
total_records = len(prompts)
processed_count = 0
while True:
# Get next available record
record_id = status_tracker.get_next_available_record(total_records, model_name)
if record_id == -1: # No more records available
break
# Generate record
prompt = prompts[record_id]
original_data = original_data_list[record_id] if original_data_list else None
result = generate_single_record(
generator, prompt, record_id, model_name,
max_length, temperature, top_p, task_type,
original_data, status_tracker
)
if result:
processed_count += 1
# Update progress
if progress_callback:
progress = status_tracker.get_progress(total_records)
progress_callback(progress, model_name, processed_count)
return f"{model_name}: Processed {processed_count} records"
def generate_dataset_multi_model(selected_models: List[str], task_type: str, custom_template: str,
file_data: List[dict], num_samples: int, max_length: int,
temperature: float, top_p: float, progress_callback=None):
"""Generate dataset using multiple models collaboratively"""
try:
# Prepare prompts
prompts = []
original_data_list = []
# Determine template
if custom_template and custom_template.strip():
template = custom_template
elif task_type in TASK_TEMPLATES:
template = TASK_TEMPLATES[task_type]["template"]
else:
template = "Generate text: {input}"
# Generate prompts for all records
for i in range(num_samples):
if file_data and len(file_data) > 0:
data_row = file_data[i % len(file_data)]
prompt = generate_from_template(template, data_row)
original_data_list.append(data_row)
else:
# Use template with placeholder values
prompt = template.replace("{topic}", f"หัวข้อที่ {i+1}") \
.replace("{question}", f"คำถามที่ {i+1} เกี่ยวกับการเรียนรู้ของเครื่อง") \
.replace("{text}", f"ข้อความตัวอย่างที่ {i+1} สำหรับการประมวลผล") \
.replace("{input}", f"ข้อมูลนำเข้าที่ {i+1}") \
.replace("{instruction}", f"คำสั่งที่ {i+1}: ให้ทำงานนี้") \
.replace("{category}", "เทคโนโลยี") \
.replace("{style}", "โคลงสี่สุภาพ")
original_data_list.append(None)
prompts.append(prompt)
# Initialize status tracker
status_tracker = ModelStatus()
model_cache = {}
# Start worker threads for each model
with ThreadPoolExecutor(max_workers=len(selected_models)) as executor:
futures = []
for model_name in selected_models:
future = executor.submit(
model_worker, model_name, model_cache, prompts,
task_type, original_data_list, max_length,
temperature, top_p, status_tracker, progress_callback
)
futures.append((future, model_name))
# Wait for all workers to complete
for future, model_name in futures:
try:
result = future.result(timeout=300) # 5 minute timeout per model
print(f"Model {model_name} completed: {result}")
except Exception as e:
print(f"Model {model_name} failed: {str(e)}")
# Collect results
dataset = sorted(status_tracker.completed_records, key=lambda x: x['id'])
if not dataset:
return None, None, None, "ไม่สามารถสร้างข้อมูลได้"
# Convert to DataFrame
df = pd.DataFrame(dataset)
# Create downloadable files
csv_data = df.to_csv(index=False)
json_data = json.dumps(dataset, indent=2, ensure_ascii=False)
return df, csv_data, json_data, None
except Exception as e:
return None, None, None, f"Error in multi-model generation: {str(e)}"
def create_interface():
with gr.Blocks(title="🇹🇭 Thai Dataset Generator", theme=gr.themes.Soft()) as demo:
gr.Markdown("# 🤗 เครื่องมือสร้างชุดข้อมูลภาษาไทยคุณภาพสูง")
gr.Markdown("⚡ **เคล็ดลับ**: ใช้โมเดลใดก็ได้จาก Hugging Face - เริ่มต้นด้วยโมเดลเล็กๆ เพื่อทดสอบก่อน")
with gr.Row():
with gr.Column():
# Flexible model input
gr.Markdown("### 🤖 เลือกโมเดลจาก Hugging Face")
gr.Markdown("💡 **คำแนะนำ**: ใส่ชื่อโมเดลจาก [Hugging Face](https://huggingface.co/models) เช่น `microsoft/DialoGPT-small`, `gpt2`, `scb10x/typhoon-7b`")
model_input_mode = gr.Radio(
choices=[
("📝 ใส่ชื่อโมเดลเอง", "manual"),
("📋 เลือกจากรายการแนะนำ", "suggested"),
("🔀 ใช้หลายโมเดลพร้อมกัน", "multiple")
],
value="manual",
label="วิธีการเลือกโมเดล"
)
# Manual model input
manual_model_group = gr.Group(visible=True)
with manual_model_group:
single_model_name = gr.Textbox(
label="ชื่อโมเดลจาก Hugging Face",
value="microsoft/DialoGPT-small",
placeholder="เช่น gpt2, microsoft/DialoGPT-medium, scb10x/typhoon-7b",
info="ใส่ชื่อโมเดลที่ต้องการใช้งาน"
)
model_verification = gr.Button("🔍 ตรวจสอบโมเดล", variant="secondary", size="sm")
model_download = gr.Button("⬇️ ดาวน์โหลดโมเดล", variant="secondary", size="sm")
model_status = gr.Textbox(
label="สถานะโมเดล",
value="ยังไม่ได้ตรวจสอบ",
interactive=False
)
# เชื่อมปุ่มตรวจสอบโมเดลกับฟังก์ชันตรวจสอบ
def verify_model(model_name):
from transformers import AutoTokenizer
try:
# ลองโหลด tokenizer (เร็วกว่าโหลด model)
AutoTokenizer.from_pretrained(model_name)
return gr.update(value=f"✅ พบโมเดล {model_name} ใน Hugging Face", interactive=False)
except Exception as e:
return gr.update(value=f"❌ ไม่พบโมเดลหรือโหลดไม่ได้: {str(e)}", interactive=False)
model_verification.click(
fn=verify_model,
inputs=[single_model_name],
outputs=[model_status]
)
# ปุ่มดาวน์โหลดโมเดล (preload)
def download_model(model_name):
import time
from transformers import AutoTokenizer, AutoModelForCausalLM
try:
t0 = time.time()
model_status_msg = f"⏳ กำลังดาวน์โหลดและโหลดโมเดล {model_name} ..."
yield gr.update(value=model_status_msg, interactive=False)
# โหลด tokenizer และ model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)
t1 = time.time()
msg = f"✅ โหลดโมเดล {model_name} สำเร็จใน {t1-t0:.1f} วินาที"
yield gr.update(value=msg, interactive=False)
except Exception as e:
yield gr.update(value=f"❌ ไม่สามารถโหลดโมเดล: {str(e)}", interactive=False)
model_download.click(
fn=download_model,
inputs=[single_model_name],
outputs=[model_status]
)
# Suggested models
suggested_model_group = gr.Group(visible=False)
with suggested_model_group:
gr.Markdown("#### โมเดลแนะนำ")
suggested_models = gr.Dropdown(
choices=[
# Small/Fast models
("⚡ DistilGPT2 (เล็ก, เร็ว)", "distilgpt2"),
("⚡ GPT2 (กลาง)", "gpt2"),
("⚡ DialoGPT-small (บทสนทนา)", "microsoft/DialoGPT-small"),
("⚡ DialoGPT-medium (บทสนทนา)", "microsoft/DialoGPT-medium"),
# Thai models
("🇹🇭 Typhoon-7B (ไทย, ใหญ่)", "scb10x/typhoon-7b"),
("🇹🇭 OpenThaiGPT-1.5-7B (ไทย)", "openthaigpt/openthaigpt1.5-7b-instruct"),
("🇹🇭 WangchanLION-7B (ไทย)", "aisingapore/llama2-7b-chat-thai"),
# Multilingual models
("🌍 mGPT (หลายภาษา)", "ai-forever/mGPT"),
("🌍 Bloom-560m (หลายภาษา, เล็ก)", "bigscience/bloom-560m"),
("🌍 Bloom-1b1 (หลายภาษา)", "bigscience/bloom-1b1"),
# Instruction-following
("🎯 Flan-T5-small (คำสั่ง)", "google/flan-t5-small"),
("🎯 Flan-T5-base (คำสั่ง)", "google/flan-t5-base"),
# Other popular models
("🔥 OPT-350m (Meta)", "facebook/opt-350m"),
("🔥 OPT-1.3b (Meta)", "facebook/opt-1.3b"),
],
value="distilgpt2",
label="เลือกโมเดลแนะนำ"
)
# Multiple models
multiple_model_group = gr.Group(visible=False)
with multiple_model_group:
multiple_model_names = gr.Textbox(
label="ชื่อโมเดลหลายตัว (แยกด้วยเครื่องหมายจุลภาค)",
value="distilgpt2, microsoft/DialoGPT-small",
placeholder="gpt2, microsoft/DialoGPT-medium, scb10x/typhoon-7b",
lines=3,
info="ใส่ชื่อโมเดลหลายตัวแยกด้วยเครื่องหมายจุลภาค"
)
model_distribution_mode = gr.Radio(
choices=[
("🔄 แบ่งงานกัน (Collaborative)", "collaborative"),
("🎲 สุ่มเลือก (Random)", "random"),
("📊 เท่าๆ กัน (Round-robin)", "round_robin")
],
value="collaborative",
label="วิธีการใช้โมเดลหลายตัว"
)
# Model info display
current_models_display = gr.Textbox(
label="โมเดลที่จะใช้",
value="microsoft/DialoGPT-small",
interactive=False
)
# Task selection with Thai tasks
gr.Markdown("### 📝 เลือกประเภทงาน")
task_dropdown = gr.Dropdown(
choices=[(v["name"], k) for k, v in TASK_TEMPLATES.items()],
value="text_generation",
label="ประเภทงานที่ต้องการ"
)
task_description = gr.Textbox(
label="คำอธิบายงาน",
value=TASK_TEMPLATES["text_generation"]["description"],
interactive=False
)
# File upload section
gr.Markdown("### 📁 อัปโหลดข้อมูลต้นฉบับ (ไม่บังคับ)")
gr.Markdown("อัปโหลดไฟล์ CSV, JSON หรือ TXT ที่มีข้อมูลต้นฉบับภาษาไทย")
file_upload = gr.File(
label="อัปโหลดไฟล์ข้อมูล",
file_types=[".csv", ".json", ".txt"]
)
file_preview = gr.Dataframe(
label="ตัวอย่างข้อมูลจากไฟล์ (5 แถวแรก)",
visible=False
)
# State สำหรับเก็บข้อมูลไฟล์ (ต้องอยู่ก่อนใช้งาน)
file_data_state = gr.State()
# ฟังก์ชัน handle file upload
def handle_file_upload(file):
import pandas as pd
import json
if file is None:
return gr.update(visible=False), None
try:
if file.name.endswith('.csv'):
df = pd.read_csv(file.name)
elif file.name.endswith('.json'):
with open(file.name, 'r', encoding='utf-8') as f:
data = json.load(f)
df = pd.DataFrame(data)
elif file.name.endswith('.txt'):
with open(file.name, 'r', encoding='utf-8') as f:
lines = f.readlines()
df = pd.DataFrame({'text': [line.strip() for line in lines if line.strip()]})
else:
return gr.update(visible=True, value="ไม่รองรับไฟล์นี้"), None
preview = df.head(5)
# คืน preview และข้อมูลทั้งหมด (list of dict)
return gr.update(visible=True, value=preview), df.to_dict('records')
except Exception as e:
return gr.update(visible=True, value=f"❌ อ่านไฟล์ผิดพลาด: {str(e)}"), None
file_upload.change(
fn=handle_file_upload,
inputs=[file_upload],
outputs=[file_preview, file_data_state]
)
# Template customization with multi-prompt support
gr.Markdown("### 🎯 ปรับแต่งเทมเพลตและ Prompt")
gr.Markdown("ใช้ {ชื่อฟิลด์} สำหรับตัวแปรในเทมเพลต")
prompt_mode = gr.Radio(
choices=[
("📝 Prompt เดียว (Single)", "single"),
("📋 หลาย Prompt (Multiple)", "multiple"),
("🎲 สุ่มจาก Template (Random)", "random")
],
value="single",
label="โหมดการใส่ Prompt"
)
# Single prompt mode
single_prompt_group = gr.Group(visible=True)
with single_prompt_group:
template_display = gr.Textbox(
label="เทมเพลตปัจจุบัน",
value=TASK_TEMPLATES["text_generation"]["template"],
interactive=False
)
custom_template = gr.Textbox(
label="เทมเพลตกำหนดเอง (ไม่บังคับ)",
lines=3,
placeholder="สร้างเทมเพลตของคุณเองที่นี่..."
)
# Multiple prompts mode
multi_prompt_group = gr.Group(visible=False)
with multi_prompt_group:
gr.Markdown("#### 📋 ใส่หลาย Prompt (แต่ละบรรทัดคือ prompt หนึ่งตัว)")
multi_prompts = gr.Textbox(
label="Prompts หลายตัว (แยกด้วยการขึ้นบรรทัดใหม่)",
lines=10,
placeholder="""เขียนเรื่องราวเกี่ยวกับการผจญภัยในป่า
สร้างบทสนทนาระหว่างครูกับนักเรียน
อธิบายวิธีการทำอาหารไทย
เขียนบทกวีเกี่ยวกับธรรมชาติ
สร้างเรื่องสั้นเกี่ยวกับมิตรภาพ"""
)
prompt_distribution = gr.Radio(
choices=[
("📊 กระจายเท่าๆ กัน", "equal"),
("🎯 ตามสัดส่วนที่กำหนด", "weighted"),
("🎲 สุ่ม", "random")
],
value="equal",
label="วิธีการกระจาย Prompt"
)
prompt_weights = gr.Textbox(
label="น้ำหนักของแต่ละ Prompt (เช่น 2,1,3,1,2)",
placeholder="2,1,3,1,2",
visible=False
)
# Random template mode
random_prompt_group = gr.Group(visible=False)
with random_prompt_group:
gr.Markdown("#### 🎲 สุ่ม Prompt จาก Template ที่เลือก")
random_templates = gr.CheckboxGroup(
choices=[(v["name"], k) for k, v in TASK_TEMPLATES.items()],
value=["text_generation", "conversation"],
label="เลือก Template ที่จะสุ่ม"
)
random_variables = gr.Textbox(
label="ตัวแปรสำหรับสุ่ม (JSON format)",
lines=5,
value="""{
"topic": ["การเดินทาง", "เทคโนโลยี", "อาหาร", "ธรรมชาติ", "ศิลปะ"],
"question": ["AI คืออะไร", "โลกร้อนคืออะไร", "การศึกษาสำคัญอย่างไร"],
"instruction": ["เขียนบทความ", "สรุปข้อมูล", "วิเคราะห์ปัญหา"]
}""",
placeholder="ใส่ตัวแปรในรูปแบบ JSON"
)
# Prompt preview and count
prompt_preview = gr.Textbox(
label="ตัวอย่าง Prompt ที่จะใช้",
lines=3,
interactive=False
)
prompt_count = gr.Textbox(
label="จำนวน Prompt ที่พร้อมใช้",
value="1 prompt",
interactive=False
)
# State สำหรับเก็บข้อมูลไฟล์
file_data_state = gr.State()
# ตัวเลือกจำนวนแถวข้อมูล (row_preset)
row_preset = gr.Dropdown(
choices=[
("10 แถว", 10),
("100 แถว", 100),
("500 แถว", 500),
("1000 แถว", 1000)
],
value=10,
label="จำนวนแถวข้อมูลที่ต้องการสร้าง"
)
# กำหนดจำนวนแถวเอง (custom_rows)
custom_rows = gr.Textbox(
label="จำนวนแถวกำหนดเอง (ถ้าเว้นว่างจะใช้ค่าจากด้านบน)",
placeholder="ใส่ตัวเลข เช่น 123"
)
# ตัวเลือกการตั้งค่าการสร้างข้อความ
max_length = gr.Slider(
minimum=16,
maximum=2048,
value=128,
step=1,
label="ความยาวสูงสุดของข้อความที่สร้าง (max_length)"
)
temperature = gr.Slider(
minimum=0.1,
maximum=2.0,
value=1.0,
step=0.05,
label="Temperature (ความสุ่ม)"
)
top_p = gr.Slider(
minimum=0.1,
maximum=1.0,
value=0.95,
step=0.01,
label="Top-p (nucleus sampling)"
)
batch_size = gr.Slider(
minimum=1,
maximum=32,
value=1,
step=1,
label="Batch size"
)
# ปุ่มสร้างข้อมูล
generate_btn = gr.Button("🚀 สร้างข้อมูล", variant="primary")
# Data Quality Settings
gr.Markdown("### 🧼 การจัดการคุณภาพข้อมูล")
enable_cleaning = gr.Checkbox(
label="เปิดใช้การทำความสะอาดข้อมูล",
value=True
)
remove_duplicates = gr.Checkbox(
label="ลบข้อมูลซ้ำซ้อน",
value=True
)
min_quality_score = gr.Slider(
minimum=0.0,
maximum=1.0,
value=0.5,
step=0.1,
label="คะแนนคุณภาพขั้นต่ำ (0-1)"
)
# ตัวเลือกแยกชุดข้อมูล (train/val/test split)
create_splits = gr.Checkbox(
label="แยกชุดข้อมูลเป็น train/val/test",
value=False
)
# Export Settings
gr.Markdown("### 📦 การส่งออกข้อมูล")
export_format = gr.CheckboxGroup(
choices=[
("📊 CSV (Excel, Spreadsheet)", "csv"),
("📋 JSON (Web APIs, General)", "json"),
("📄 JSONL (Fine-tuning, Streaming)", "jsonl"),
("🤗 Hugging Face Dataset (Complete Package)", "huggingface"),
("📝 TXT (Simple Text)", "txt"),
("🗃️ Parquet (Big Data, Analytics)", "parquet"),
("📋 TSV (Tab-separated)", "tsv"),
("🎯 Custom Format", "custom")
],
value=["csv", "json"],
label="เลือกรูปแบบไฟล์ที่ต้องการ (สามารถเลือกหลายแบบ)"
)
# Custom format settings
custom_format_group = gr.Group(visible=False)
with custom_format_group:
gr.Markdown("#### 🎯 การตั้งค่ารูปแบบกำหนดเอง")
custom_template_format = gr.Textbox(
label="Template สำหรับแต่ละ record",
value="Input: {input}\nOutput: {output}\n---",
lines=3,
placeholder="ใช้ {field_name} สำหรับข้อมูล"
)
custom_file_extension = gr.Textbox(
label="นามสกุลไฟล์",
value="txt",
placeholder="เช่น txt, md, xml"
)
# Advanced export options
with gr.Accordion("⚙️ ตัวเลือกขั้นสูง", open=False):
include_metadata = gr.Checkbox(
label="รวม Metadata (model_used, timestamp, etc.)",
value=True
)
include_quality_score = gr.Checkbox(
label="รวม Quality Score",
value=True
)
file_naming_pattern = gr.Textbox(
label="รูปแบบชื่อไฟล์",
value="thai_dataset_{task}_{timestamp}",
placeholder="ใช้ {task}, {timestamp}, {model}, {count}"
)
compression = gr.Radio(
choices=[
("ไม่บีบอัด", "none"),
("ZIP", "zip"),
("GZIP", "gzip")
],
value="none",
label="การบีบอัดไฟล์"
)
# ...existing code...
with gr.Column():
with gr.Tabs():
with gr.TabItem("📊 ตัวอย่างข้อมูล"):
dataset_preview = gr.Dataframe(
headers=["id", "task_type", "input", "output", "quality_score"],
interactive=False
)
status_message = gr.Markdown(
value="",
visible=True
)
# State สำหรับข้อมูลที่สร้าง
csv_data_state = gr.State()
json_data_state = gr.State()
dataset_card_state = gr.State()
hf_export_state = gr.State()
loading_status = gr.State()
with gr.TabItem("📈 รายงานคุณภาพ"):
quality_report = gr.JSON(
label="รายงานคุณภาพข้อมูล",
visible=True
)
quality_summary = gr.Markdown(
value="สร้างข้อมูลเสร็จแล้วจึงจะแสดงรายงานคุณภาพ"
)
with gr.TabItem("💾 ดาวน์โหลด"):
gr.Markdown("### 💾 ดาวน์โหลดชุดข้อมูลในรูปแบบต่างๆ")
download_status = gr.Markdown("สร้างข้อมูลเสร็จแล้วจึงจะสามารถดาวน์โหลดได้")
# Dynamic download buttons based on selected formats
download_buttons = {}
download_files = {}
with gr.Row():
csv_btn = gr.Button("📊 CSV", variant="secondary", visible=False)
json_btn = gr.Button("📋 JSON", variant="secondary", visible=False)
jsonl_btn = gr.Button("📄 JSONL", variant="secondary", visible=False)
txt_btn = gr.Button("📝 TXT", variant="secondary", visible=False)
with gr.Row():
parquet_btn = gr.Button("🗃️ Parquet", variant="secondary", visible=False)
tsv_btn = gr.Button("📋 TSV", variant="secondary", visible=False)
hf_btn = gr.Button("🤗 HF Dataset", variant="secondary", visible=False)
custom_btn = gr.Button("🎯 Custom", variant="secondary", visible=False)
# Download files
csv_download = gr.File(label="CSV File", visible=False)
json_download = gr.File(label="JSON File", visible=False)
jsonl_download = gr.File(label="JSONL File", visible=False)
txt_download = gr.File(label="TXT File", visible=False)
parquet_download = gr.File(label="Parquet File", visible=False)
tsv_download = gr.File(label="TSV File", visible=False)
hf_download = gr.File(label="HF Dataset Package", visible=False)
custom_download = gr.File(label="Custom Format", visible=False)
# All formats in one package
with gr.Row():
package_btn = gr.Button("📦 ดาวน์โหลดทั้งหมด (ZIP)", variant="primary")
package_download = gr.File(label="Complete Package", visible=False)
# ...existing code for states...
def update_export_format_visibility(selected_formats):
"""Update visibility of download buttons based on selected formats"""
return [
gr.update(visible=("csv" in selected_formats)),
gr.update(visible=("json" in selected_formats)),
gr.update(visible=("jsonl" in selected_formats)),
gr.update(visible=("txt" in selected_formats)),
gr.update(visible=("parquet" in selected_formats)),
gr.update(visible=("tsv" in selected_formats)),
gr.update(visible=("huggingface" in selected_formats)),
gr.update(visible=("custom" in selected_formats)),
gr.update(visible=("custom" in selected_formats))
]
def generate_multiple_formats(data, selected_formats, include_metadata, include_quality_score,
file_naming_pattern, custom_template_format, custom_file_extension,
task_type, compression):
"""Generate data in multiple formats"""
from datetime import datetime
import tempfile
import zipfile
import gzip
import pyarrow as pa
import pyarrow.parquet as pq
if not data:
return {}
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
model_name = data[0].get('model_used', 'unknown').replace('/', '_')
# Prepare data
export_data = []
for record in data:
export_record = {}
export_record['input'] = record.get('prompt', '')
export_record['output'] = record.get('generated_text', '')
if include_metadata:
export_record['metadata'] = {
'model_used': record.get('model_used', ''),
'task_type': record.get('task_type', ''),
'timestamp': record.get('generation_time', '')
}
if include_quality_score and 'quality_score' in record:
export_record['quality_score'] = record['quality_score']
export_data.append(export_record)
# Generate filename
filename_base = file_naming_pattern.format(
task=task_type,
timestamp=timestamp,
model=model_name,
count=len(export_data)
)
generated_files = {}
# Generate each format
if "csv" in selected_formats:
df = pd.DataFrame(export_data)
csv_content = df.to_csv(index=False)
generated_files['csv'] = (f"{filename_base}.csv", csv_content)
if "json" in selected_formats:
json_content = json.dumps(export_data, indent=2, ensure_ascii=False)
generated_files['json'] = (f"{filename_base}.json", json_content)
if "jsonl" in selected_formats:
jsonl_content = '\n'.join([json.dumps(record, ensure_ascii=False) for record in export_data])
generated_files['jsonl'] = (f"{filename_base}.jsonl", jsonl_content)
if "txt" in selected_formats:
txt_content = '\n'.join([f"Input: {record['input']}\nOutput: {record['output']}\n---" for record in export_data])
generated_files['txt'] = (f"{filename_base}.txt", txt_content)
if "tsv" in selected_formats:
df = pd.DataFrame(export_data)
tsv_content = df.to_csv(index=False, sep='\t')
generated_files['tsv'] = (f"{filename_base}.tsv", tsv_content)
if "parquet" in selected_formats:
df = pd.DataFrame(export_data)
temp_parquet = tempfile.mktemp(suffix='.parquet')
df.to_parquet(temp_parquet)
with open(temp_parquet, 'rb') as f:
parquet_content = f.read()
generated_files['parquet'] = (f"{filename_base}.parquet", parquet_content)
if "custom" in selected_formats:
custom_content = []
for record in export_data:
formatted = custom_template_format.format(**record)
custom_content.append(formatted)
custom_text = '\n'.join(custom_content)
generated_files['custom'] = (f"{filename_base}.{custom_file_extension}", custom_text)
# Apply compression if selected
if compression == "gzip":
for format_name, (filename, content) in generated_files.items():
if isinstance(content, str):
content = content.encode('utf-8')
compressed = gzip.compress(content)
generated_files[format_name] = (filename + '.gz', compressed)
return generated_files
def create_complete_package(generated_files, compression):
"""Create a complete package with all formats"""
import tempfile
import zipfile
if not generated_files:
return None
temp_zip = tempfile.mktemp(suffix='.zip')
with zipfile.ZipFile(temp_zip, 'w', zipfile.ZIP_DEFLATED) as zipf:
for format_name, (filename, content) in generated_files.items():
if isinstance(content, str):
content = content.encode('utf-8')
zipf.writestr(filename, content)
return temp_zip
def download_specific_format(format_name, generated_files):
"""Download specific format"""
if format_name in generated_files:
filename, content = generated_files[format_name]
if isinstance(content, str):
return gr.update(visible=True, value=io.StringIO(content))
else:
temp_file = tempfile.mktemp()
with open(temp_file, 'wb') as f:
f.write(content)
return gr.update(visible=True, value=temp_file)
return gr.update(visible=False)
# Event handlers
export_format.change(
fn=update_export_format_visibility,
inputs=[export_format],
outputs=[csv_btn, json_btn, jsonl_btn, txt_btn, parquet_btn, tsv_btn, hf_btn, custom_btn, custom_format_group]
)
# ...existing code for other event handlers...
# Download button handlers
csv_btn.click(
fn=lambda files: download_specific_format('csv', files),
inputs=[gr.State()], # Will be connected to generated files state
outputs=[csv_download]
)
json_btn.click(
fn=lambda files: download_specific_format('json', files),
inputs=[gr.State()],
outputs=[json_download]
)
jsonl_btn.click(
fn=lambda files: download_specific_format('jsonl', files),
inputs=[gr.State()],
outputs=[jsonl_download]
)
txt_btn.click(
fn=lambda files: download_specific_format('txt', files),
inputs=[gr.State()],
outputs=[txt_download]
)
parquet_btn.click(
fn=lambda files: download_specific_format('parquet', files),
inputs=[gr.State()],
outputs=[parquet_download]
)
tsv_btn.click(
fn=lambda files: download_specific_format('tsv', files),
inputs=[gr.State()],
outputs=[tsv_download]
)
hf_btn.click(
fn=lambda files: download_specific_format('huggingface', files),
inputs=[gr.State()],
outputs=[hf_download]
)
custom_btn.click(
fn=lambda files: download_specific_format('custom', files),
inputs=[gr.State()],
outputs=[custom_download]
)
package_btn.click(
fn=lambda files, comp: gr.update(visible=True, value=create_complete_package(files, comp)),
inputs=[gr.State(), compression], # Will be connected to generated files and compression
outputs=[package_download]
)
# Update generate button to use correct function
generate_btn.click(
fn=process_with_flexible_models,
inputs=[model_input_mode, single_model_name, suggested_models, multiple_model_names,
model_distribution_mode, task_dropdown, prompt_mode, custom_template,
multi_prompts, random_templates, random_variables, file_data_state,
row_preset, custom_rows, max_length, temperature, top_p, batch_size,
enable_cleaning, remove_duplicates, min_quality_score,
create_splits, export_format],
outputs=[dataset_preview, status_message, quality_report, quality_summary,
csv_data_state, json_data_state, dataset_card_state, hf_export_state,
loading_status]
)
return demo
def validate_models_before_generation(*args, **kwargs):
# TODO: implement validation logic
return None
def process_with_flexible_models(input_mode, single_model, suggested_model, multiple_models,
model_distribution_mode, task_type, prompt_mode, custom_template,
multi_prompts, random_templates, random_variables, file_data,
row_preset, custom_rows, max_length, temperature, top_p, batch_size,
enable_cleaning, remove_duplicates, min_quality_score,
create_splits, export_format):
"""Process generation with flexible model selection"""
# ฟังก์ชันเลือกโมเดลที่ใช้จริง
def get_selected_models(input_mode, single_model, suggested_model, multiple_models):
if input_mode == "manual":
return [single_model.strip()] if single_model and single_model.strip() else []
elif input_mode == "suggested":
return [suggested_model] if suggested_model else []
elif input_mode == "multiple":
# แยกชื่อโมเดลด้วย , และลบช่องว่าง
return [m.strip() for m in multiple_models.split(",") if m.strip()]
return []
# ฟังก์ชันนับจำนวนแถวข้อมูลที่ต้องการสร้าง
def get_final_row_count(row_preset, custom_rows):
try:
if custom_rows and str(custom_rows).strip():
return int(custom_rows)
return int(row_preset)
except Exception:
return 10
# Get selected models
selected_models = get_selected_models(input_mode, single_model, suggested_model, multiple_models)
if not selected_models:
yield (
gr.update(visible=False),
gr.update(visible=True, value="❌ กรุณาเลือกโมเดลอย่างน้อยหนึ่งตัว"),
{}, "ไม่มีโมเดล", None, None, None, None,
"❌ ไม่ได้เลือกโมเดล"
)
return
num_samples = get_final_row_count(row_preset, custom_rows)
try:
yield (
gr.update(visible=False),
gr.update(visible=True, value=f"🔄 กำลังสร้างข้อมูล {num_samples} แถว..."),
{}, "กำลังสร้าง...", None, None, None, None,
f"🔄 กำลังประมวลผล..."
)
# Simple generation for now
model_name = selected_models[0]
df, csv_data, json_data, error = generate_dataset_from_task(
model_name, task_type, custom_template, file_data,
num_samples, max_length, temperature, top_p
)
if error:
yield (
gr.update(visible=False),
gr.update(visible=True, value=f"❌ เกิดข้อผิดพลาด: {error}"),
{}, "เกิดข้อผิดพลาด", None, None, None, None,
f"❌ {error}"
)
return
# Basic quality processing
raw_data = df.to_dict('records')
quality_report = {
"total_records": len(raw_data),
"models_used": selected_models
}
final_df = pd.DataFrame(raw_data)
final_csv = final_df.to_csv(index=False)
final_json = json.dumps(raw_data, indent=2, ensure_ascii=False)
dataset_card = f"# Dataset generated with {model_name}\n\nRecords: {len(raw_data)}"
success_msg = f"✅ สร้างข้อมูลสำเร็จ! ได้ {len(raw_data)} แถว"
quality_summary = f"📊 จำนวนข้อมูล: {len(raw_data)} แถว"
yield (
gr.update(visible=True, value=final_df),
gr.update(visible=True, value=success_msg),
quality_report,
quality_summary,
final_csv,
final_json,
dataset_card,
None,
"✅ เสร็จสิ้น!"
)
except Exception as e:
yield (
gr.update(visible=False),
gr.update(visible=True, value=f"❌ ข้อผิดพลาด: {str(e)}"),
{}, "เกิดข้อผิดพลาด", None, None, None, None,
f"❌ {str(e)}"
)
if __name__ == "__main__":
demo = create_interface()
demo.launch(
server_name="0.0.0.0",
server_port=7860,
share=True
)
|