-
Notifications
You must be signed in to change notification settings - Fork 24
/
Copy pathgradio_web_server_v2.py
1197 lines (1020 loc) · 48.1 KB
/
gradio_web_server_v2.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
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
"""
The gradio demo server for chatting with a single model.
"""
import argparse
from collections import defaultdict
import datetime
import json
import os
import random
import time
import uuid
import gradio as gr
import requests
from judgelm.conversation import SeparatorStyle
from judgelm.constants import (
LOGDIR,
WORKER_API_TIMEOUT,
ErrorCode,
MODERATION_MSG,
CONVERSATION_LIMIT_MSG,
SERVER_ERROR_MSG,
INACTIVE_MSG,
INPUT_CHAR_LEN_LIMIT,
CONVERSATION_TURN_LIMIT,
SESSION_EXPIRATION_TIME,
)
from judgelm.model.model_adapter import get_conversation_template
from judgelm.model.model_registry import model_info
from judgelm.serve.api_provider import (
anthropic_api_stream_iter,
openai_api_stream_iter,
palm_api_stream_iter,
init_palm_chat,
)
from judgelm.utils import (
build_logger,
violates_moderation,
get_window_url_params_js,
parse_gradio_auth_creds,
)
logger = build_logger("gradio_web_server", "gradio_web_server.log")
headers = {"User-Agent": "FastChat Client"}
no_change_btn = gr.Button.update()
enable_btn = gr.Button.update(interactive=True)
disable_btn = gr.Button.update(interactive=False)
controller_url = None
enable_moderation = False
learn_more_md = """
### License
The service is a research preview intended for non-commercial use only, subject to the model [License](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md) of LLaMA, [Terms of Use](https://openai.com/policies/terms-of-use) of the data generated by OpenAI, and [Privacy Practices](https://chrome.google.com/webstore/detail/sharegpt-share-your-chatg/daiacboceoaocpibfodeljbdfacokfjb) of ShareGPT. Please contact us if you find any potential violation.
"""
ip_expiration_dict = defaultdict(lambda: 0)
examples = [["What is the capital of France?",
"The capital of France is Paris.",
"The capital of France is London."],
["Write a sentence using the homophone \"won/one.\"\n",
"The cat meowed.",
"\"I won the game, but my one friend didn't.\""],
["Edit the following sentence \"I wants to visit Europe in my holidays.\"\nI wants to visit Europe in my holidays.",
"I want to visit Europe during my holidays.",
"I want to visit Europe during my holidays."],
["How is electricity generated?",
"Electricity is generated through various methods, such as burning fossil fuels, nuclear reactions, and harnessing renewable energy sources like wind and solar power.",
"Electricity is generated by simply plugging devices into an outlet."],
["What's your favorite type of music?",
"My favorite type of music is classical.",
"I enjoy listening to rock music."],
["Arrange the following words in alphabetical order: quickly, question, queue, quality.\n",
"quickly\nquestion\nqueue\nquality\n\nExplanation: The given words are already in alphabetical order.",
"Queuely, Question, Quality, Quickly."]]
examples_reference = [["Create a one-sentence description of the color yellow.",
"Yellow is a vibrant and warm color, often associated with sunshine, optimism, and cheerfulness.",
"Yellow is a bright and cheerful color that can make a room feel warm and inviting.",
"This is the description of the color yellow.Yellow is an emotion.",
],
["Who is the lead singer of the Gorillaz?",
"Damon Albarn is the lead singer of the Gorillaz",
"The lead singer of the Gorillaz is z世代.",
"Damon Albarn is the lead singer of the Gorillaz."
],
["I have a question the movie The Fugitive. What is the name of the character played by Harrison Ford?",
"The character played by Harrison Ford in The Fugitive is named Dr. Richard Kimble.",
"The character played by Harrison Ford in the movie The Fugitive is Dr. Richard Kimble.",
"The character played by Harrison Ford in The Fugitive is Doctor Richard Kimble."],
["Reverse the following number sequence\n59, 22, 15, 11, 5",
"The reversed sequence would be 5, 11, 15, 22, 59.",
"11, 15, 11, 5, 9",
"5, 11, 15, 22, 59"]]
# create num to words dict
num2words = {1:"one", 2:"two", 3:"three", 4:"four", 5:"five",
6:"six", 7:"seven", 8: "eight", 9: 'nine', 10: 'ten', \
11: 'eleven', 12: 'twelve', 13: 'thirteen', 14: 'fourteen', \
15: 'fifteen', 16: 'sixteen', 17: 'seventeen', 18: 'eighteen', 19: 'nineteen'}
class State:
def __init__(self, model_name):
self.conv = get_conversation_template(model_name)
self.conv_id = uuid.uuid4().hex
self.skip_next = False
self.model_name = model_name
if model_name == "palm-2":
# According to release note, "chat-bison@001" is PaLM 2 for chat.
# https://cloud.google.com/vertex-ai/docs/release-notes#May_10_2023
self.palm_chat = init_palm_chat("chat-bison@001")
def to_gradio_chatbot(self):
return self.conv.to_gradio_chatbot()
def dict(self):
base = self.conv.dict()
base.update(
{
"conv_id": self.conv_id,
"model_name": self.model_name,
}
)
return base
def set_global_vars(controller_url_, enable_moderation_):
global controller_url, enable_moderation
controller_url = controller_url_
enable_moderation = enable_moderation_
def get_conv_log_filename():
t = datetime.datetime.now()
name = os.path.join(LOGDIR, f"{t.year}-{t.month:02d}-{t.day:02d}-conv.json")
return name
def get_model_list(controller_url, add_chatgpt, add_claude, add_palm):
ret = requests.post(controller_url + "/refresh_all_workers")
assert ret.status_code == 200
ret = requests.post(controller_url + "/list_models")
models = ret.json()["models"]
# Add API providers
if add_chatgpt:
models += ["gpt-3.5-turbo", "gpt-4"]
if add_claude:
models += ["claude-2", "claude-instant-1"]
if add_palm:
models += ["palm-2"]
priority = {k: f"___{i:02d}" for i, k in enumerate(model_info)}
models.sort(key=lambda x: priority.get(x, x))
logger.info(f"Models: {models}")
return models
def load_demo_single_v2(models, url_params):
selected_model = models[0] if len(models) > 0 else ""
if "model" in url_params:
model = url_params["model"]
if model in models:
selected_model = model
dropdown_update = gr.Dropdown.update(
choices=models, value=selected_model, visible=True
)
state = None
# return (
# state,
# dropdown_update,
# gr.Chatbot.update(visible=True),
# gr.Textbox.update(visible=True),
# gr.Button.update(visible=True),
# gr.Row.update(visible=True),
# gr.Accordion.update(visible=True),
# gr.Textbox.update(visible=True),
# gr.Textbox.update(visible=True),
# gr.Textbox.update(visible=True),
# gr.Button.update(visible=True),
# )
def load_demo_v2(url_params, request: gr.Request):
global models
ip = request.client.host
logger.info(f"load_demo. ip: {ip}. params: {url_params}")
ip_expiration_dict[ip] = time.time() + SESSION_EXPIRATION_TIME
if args.model_list_mode == "reload":
models = get_model_list(
controller_url, args.add_chatgpt, args.add_claude, args.add_palm
)
return load_demo_single_v2(models, url_params)
def vote_last_response(state, vote_type, model_selector, request: gr.Request):
with open(get_conv_log_filename(), "a") as fout:
data = {
"tstamp": round(time.time(), 4),
"type": vote_type,
"model": model_selector,
"state": state.dict(),
"ip": request.client.host,
}
fout.write(json.dumps(data) + "\n")
def upvote_last_response(state, model_selector, request: gr.Request):
logger.info(f"upvote. ip: {request.client.host}")
vote_last_response(state, "upvote", model_selector, request)
return ("",) + (disable_btn,) * 3
def downvote_last_response(state, model_selector, request: gr.Request):
logger.info(f"downvote. ip: {request.client.host}")
vote_last_response(state, "downvote", model_selector, request)
return ("",) + (disable_btn,) * 3
def flag_last_response(state, model_selector, request: gr.Request):
logger.info(f"flag. ip: {request.client.host}")
vote_last_response(state, "flag", model_selector, request)
return ("",) + (disable_btn,) * 3
def regenerate(state, request: gr.Request):
logger.info(f"regenerate. ip: {request.client.host}")
state.conv.update_last_message(None)
return (state, state.to_gradio_chatbot(), "") + (disable_btn,) * 5
def regenerate_v2(state, request: gr.Request):
logger.info(f"regenerate. ip: {request.client.host}")
state.conv.update_last_message(None)
return (state, state.to_gradio_chatbot(), "") + (disable_btn,) * 6
def clear_history_v2(answer_num, request: gr.Request):
logger.info(f"clear_history. ip: {request.client.host}")
state = None
return_list = (state, [], "")
for i in range(int(answer_num)):
return_list += (f"Enter the assistant {i+1}'s answer text")
return return_list + (disable_btn,) * 5 + (enable_btn,)
def add_text_v2(state, model_selector, text, request: gr.Request):
ip = request.client.host
logger.info(f"add_text. ip: {ip}. len: {len(text)}")
if state is None:
state = State(model_selector)
if len(text) <= 0:
state.skip_next = True
return (state, state.to_gradio_chatbot(), "") + (no_change_btn,) * 6
if ip_expiration_dict[ip] < time.time():
logger.info(f"inactive. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), INACTIVE_MSG) + (no_change_btn,) * 6
if enable_moderation:
flagged = violates_moderation(text)
if flagged:
logger.info(f"violate moderation. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), MODERATION_MSG) + (
no_change_btn,
) * 6
conv = state.conv
if (len(conv.messages) - conv.offset) // 2 >= CONVERSATION_TURN_LIMIT:
logger.info(f"conversation turn limit. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), CONVERSATION_LIMIT_MSG) + (
no_change_btn,
) * 6
text = text[:INPUT_CHAR_LEN_LIMIT] # Hard cut-off
conv.append_message(conv.roles[0], text)
conv.append_message(conv.roles[1], None)
return (state, state.to_gradio_chatbot(), "") + (disable_btn,) * 6
def add_pretext_reference(state, model_selector, radio, question_text, answer_num, reference_box, answer1_text, answer2_text, answer3_text, answer4_text, answer5_text, answer6_text, answer7_text, answer8_text, answer9_text, answer10_text, request:gr.Request):
text = question_text
answers = [answer1_text, answer2_text, answer3_text, answer4_text, answer5_text, answer6_text, answer7_text, answer8_text, answer9_text, answer10_text]
answer1_text = answers[0]
answer2_text = answers[1]
ip = request.client.host
logger.info(f"add_text. ip: {ip}. len: {len(text)}")
state = State(model_selector)
if radio == "Fast Eval":
state.conv.stop_str = "\n"
# example questions
if "ep1" == question_text:
question_text = "I want to get a tattoo but I'm not sure what kind of design to get. What do you suggest?"
answer1_text = "As an AI language model, I cannot recommend any specific tattoo designs. However, I recommend researching various designs and styles to find one that resonates with you and has a personal meaning. It's also important to find a reputable tattoo artist that you trust and feel comfortable with. Best of luck in your tattoo journey"
answer2_text = "A friend of mine has a tattoo that he really likes, so I've been thinking about getting one too. What do you think?"
if len(text) <= 0:
state.skip_next = True
return (state, state.to_gradio_chatbot(), "", "", "") + (no_change_btn,) * 6
if ip_expiration_dict[ip] < time.time():
logger.info(f"inactive. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), INACTIVE_MSG, INACTIVE_MSG, INACTIVE_MSG) + (no_change_btn,) * 6
if enable_moderation:
flagged = violates_moderation(text)
if flagged:
logger.info(f"violate moderation. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), MODERATION_MSG, MODERATION_MSG, MODERATION_MSG) + (
no_change_btn,
) * 6
conv = state.conv
if (len(conv.messages) - conv.offset) // 2 >= CONVERSATION_TURN_LIMIT:
logger.info(f"conversation turn limit. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), CONVERSATION_LIMIT_MSG, CONVERSATION_LIMIT_MSG, CONVERSATION_LIMIT_MSG) + (
no_change_btn,
) * 6
# modify the input text
prompt_template = {"prompt_id": 1, "system_prompt": "You are a helpful and precise assistant for checking the quality of the answer.", "prompt_template": "[Question]\n{question}\n\n[The Start of Assistant 1's Answer]\n{answer_1}\n\n[The End of Assistant 1's Answer]\n\n[The Start of Assistant 2's Answer]\n{answer_2}\n\n[The End of Assistant 2's Answer]\n\n[System]\n{prompt}\n\n", "defaults": {"prompt": "We would like to request your feedback on the performance of two AI assistants in response to the user question displayed above.\nPlease rate the helpfulness, relevance, accuracy, level of details of their responses. Each assistant receives an overall score on a scale of 1 to 10, where a higher score indicates better overall performance.\nPlease first output a single line containing only two values indicating the scores for Assistant 1 and 2, respectively. The two scores are separated by a space. In the subsequent line, please provide a comprehensive explanation of your evaluation, avoiding any potential bias and ensuring that the order in which the responses were presented does not affect your judgment."}, "description": "Prompt for general questions", "category": "general"}
appendix = "### Response:"
text = prompt_template['system_prompt'] + '\n' + prompt_template['prompt_template'].format(question=question_text,
answer_1=answer1_text,
answer_2=answer2_text,
prompt=prompt_template['defaults'][
'prompt']) + appendix
plug_in_before_str = "[The Start of Assistant 1's Answer]"
plug_in_pos = text.find(plug_in_before_str)
new_text = text[:plug_in_pos] + "[Reference Answer]\n" + reference_box + "\n\n" + text[plug_in_pos:]
text = new_text
if answer_num > 2:
plug_in_before_str = "[System]"
plug_in_pos = text.find(plug_in_before_str)
new_answer = ""
for q_i in range(2, int(answer_num)):
new_answer += "[The Start of Assistant " + str(q_i+1) + "'s Answer]\n" + answers[q_i] + "\n\n" + "[The End of Assistant " + str(q_i+1) + "'s Answer]\n\n"
new_data_sample = text[:plug_in_pos] + new_answer + text[plug_in_pos:]
text = new_data_sample
text = text.replace(f"of two AI assistants", f"of {num2words[int(answer_num)]} AI assistants")
text = text.replace("containing only two values indicating ", f"containing only {num2words[int(answer_num)]} values indicating ")
text = text.replace("for Assistant 1 and 2", "for Assistant 1")
plug_in_after_str = "for Assistant 1"
plug_in_pos = text.find(plug_in_after_str) + len(plug_in_after_str)
new_answer = ""
for i in range(int(answer_num)-2):
new_answer += f", {i+2}"
new_answer += f" and {int(answer_num)}"
new_data_sample = text[:plug_in_pos] + new_answer + text[plug_in_pos:]
text = new_data_sample
text = text.replace("The two scores are", f"The {num2words[int(answer_num)]} scores are")
pass
text = text[:INPUT_CHAR_LEN_LIMIT] # Hard cut-off
conv.append_message(conv.roles[0], text)
conv.append_message(conv.roles[1], None)
return (state, state.to_gradio_chatbot()) + (disable_btn,) * 6
def add_pretext_single_answer(state, model_selector, radio, question_text, reference_box, answer_text, request:gr.Request):
text = question_text
ip = request.client.host
logger.info(f"add_text. ip: {ip}. len: {len(text)}")
state = State(model_selector)
if radio == "Fast Eval":
state.conv.stop_str = "\n"
# example questions
if "ep1" == question_text:
question_text = "I want to get a tattoo but I'm not sure what kind of design to get. What do you suggest?"
answer1_text = "As an AI language model, I cannot recommend any specific tattoo designs. However, I recommend researching various designs and styles to find one that resonates with you and has a personal meaning. It's also important to find a reputable tattoo artist that you trust and feel comfortable with. Best of luck in your tattoo journey"
answer2_text = "A friend of mine has a tattoo that he really likes, so I've been thinking about getting one too. What do you think?"
if len(text) <= 0:
state.skip_next = True
return (state, state.to_gradio_chatbot(), "", "", "") + (no_change_btn,) * 6
if ip_expiration_dict[ip] < time.time():
logger.info(f"inactive. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), INACTIVE_MSG, INACTIVE_MSG, INACTIVE_MSG) + (no_change_btn,) * 6
if enable_moderation:
flagged = violates_moderation(text)
if flagged:
logger.info(f"violate moderation. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), MODERATION_MSG, MODERATION_MSG, MODERATION_MSG) + (
no_change_btn,
) * 6
conv = state.conv
if (len(conv.messages) - conv.offset) // 2 >= CONVERSATION_TURN_LIMIT:
logger.info(f"conversation turn limit. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), CONVERSATION_LIMIT_MSG, CONVERSATION_LIMIT_MSG, CONVERSATION_LIMIT_MSG) + (
no_change_btn,
) * 6
# modify the input text
prompt_template = {"prompt_id": 1, "system_prompt": "You are a helpful and precise assistant for checking the quality of the answer.", "prompt_template": "[Question]\n{question}\n\n[The Start of Assistant 1's Answer]\n{answer_1}\n\n[The End of Assistant 1's Answer]\n\n[The Start of Assistant 2's Answer]\n{answer_2}\n\n[The End of Assistant 2's Answer]\n\n[System]\n{prompt}\n\n", "defaults": {"prompt": "We would like to request your feedback on the performance of two AI assistants in response to the user question displayed above.\nPlease rate the helpfulness, relevance, accuracy, level of details of their responses. Each assistant receives an overall score on a scale of 1 to 10, where a higher score indicates better overall performance.\nPlease first output a single line containing only two values indicating the scores for Assistant 1 and 2, respectively. The two scores are separated by a space. In the subsequent line, please provide a comprehensive explanation of your evaluation, avoiding any potential bias and ensuring that the order in which the responses were presented does not affect your judgment."}, "description": "Prompt for general questions", "category": "general"}
appendix = "### Response:10"
text = prompt_template['system_prompt'] + '\n' + prompt_template['prompt_template'].format(question=question_text,
answer_1=reference_box,
answer_2=answer_text,
prompt=prompt_template['defaults'][
'prompt']) + appendix
text = text[:INPUT_CHAR_LEN_LIMIT] # Hard cut-off
conv.append_message(conv.roles[0], text)
conv.append_message(conv.roles[1], None)
return (state, state.to_gradio_chatbot()) + (disable_btn,) * 6
def add_pretext(state, model_selector, radio, question_text, answer_num, answer1_text, answer2_text, answer3_text, answer4_text, answer5_text, answer6_text, answer7_text, answer8_text, answer9_text, answer10_text, request:gr.Request):
text = question_text
answers = [answer1_text, answer2_text, answer3_text, answer4_text, answer5_text, answer6_text, answer7_text, answer8_text, answer9_text, answer10_text]
answer1_text = answers[0]
answer2_text = answers[1]
ip = request.client.host
logger.info(f"add_text. ip: {ip}. len: {len(text)}")
state = State(model_selector)
if radio == "Fast Eval":
state.conv.stop_str = "\n"
# example questions
if "ep1" == question_text:
question_text = "I want to get a tattoo but I'm not sure what kind of design to get. What do you suggest?"
answer1_text = "As an AI language model, I cannot recommend any specific tattoo designs. However, I recommend researching various designs and styles to find one that resonates with you and has a personal meaning. It's also important to find a reputable tattoo artist that you trust and feel comfortable with. Best of luck in your tattoo journey"
answer2_text = "A friend of mine has a tattoo that he really likes, so I've been thinking about getting one too. What do you think?"
if len(text) <= 0:
state.skip_next = True
return (state, state.to_gradio_chatbot(), "", "", "") + (no_change_btn,) * 6
if ip_expiration_dict[ip] < time.time():
logger.info(f"inactive. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), INACTIVE_MSG, INACTIVE_MSG, INACTIVE_MSG) + (no_change_btn,) * 6
if enable_moderation:
flagged = violates_moderation(text)
if flagged:
logger.info(f"violate moderation. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), MODERATION_MSG, MODERATION_MSG, MODERATION_MSG) + (
no_change_btn,
) * 6
conv = state.conv
if (len(conv.messages) - conv.offset) // 2 >= CONVERSATION_TURN_LIMIT:
logger.info(f"conversation turn limit. ip: {request.client.host}. text: {text}")
state.skip_next = True
return (state, state.to_gradio_chatbot(), CONVERSATION_LIMIT_MSG, CONVERSATION_LIMIT_MSG, CONVERSATION_LIMIT_MSG) + (
no_change_btn,
) * 6
# modify the input text
prompt_template = {"prompt_id": 1, "system_prompt": "You are a helpful and precise assistant for checking the quality of the answer.", "prompt_template": "[Question]\n{question}\n\n[The Start of Assistant 1's Answer]\n{answer_1}\n\n[The End of Assistant 1's Answer]\n\n[The Start of Assistant 2's Answer]\n{answer_2}\n\n[The End of Assistant 2's Answer]\n\n[System]\n{prompt}\n\n", "defaults": {"prompt": "We would like to request your feedback on the performance of two AI assistants in response to the user question displayed above.\nPlease rate the helpfulness, relevance, accuracy, level of details of their responses. Each assistant receives an overall score on a scale of 1 to 10, where a higher score indicates better overall performance.\nPlease first output a single line containing only two values indicating the scores for Assistant 1 and 2, respectively. The two scores are separated by a space. In the subsequent line, please provide a comprehensive explanation of your evaluation, avoiding any potential bias and ensuring that the order in which the responses were presented does not affect your judgment."}, "description": "Prompt for general questions", "category": "general"}
appendix = "### Response:"
text = prompt_template['system_prompt'] + '\n' + prompt_template['prompt_template'].format(question=question_text,
answer_1=answer1_text,
answer_2=answer2_text,
prompt=prompt_template['defaults'][
'prompt']) + appendix
if answer_num > 2:
plug_in_before_str = "[System]"
plug_in_pos = text.find(plug_in_before_str)
new_answer = ""
for q_i in range(2, int(answer_num)):
new_answer += "[The Start of Assistant " + str(q_i+1) + "'s Answer]\n" + answers[q_i] + "\n\n" + "[The End of Assistant " + str(q_i+1) + "'s Answer]\n\n"
new_data_sample = text[:plug_in_pos] + new_answer + text[plug_in_pos:]
text = new_data_sample
text = text.replace(f"of two AI assistants", f"of {num2words[int(answer_num)]} AI assistants")
text = text.replace("containing only two values indicating ", f"containing only {num2words[int(answer_num)]} values indicating ")
text = text.replace("for Assistant 1 and 2", "for Assistant 1")
plug_in_after_str = "for Assistant 1"
plug_in_pos = text.find(plug_in_after_str) + len(plug_in_after_str)
new_answer = ""
for i in range(int(answer_num)-2):
new_answer += f", {i+2}"
new_answer += f" and {int(answer_num)}"
new_data_sample = text[:plug_in_pos] + new_answer + text[plug_in_pos:]
text = new_data_sample
text = text.replace("The two scores are", f"The {num2words[int(answer_num)]} scores are")
pass
text = text[:INPUT_CHAR_LEN_LIMIT] # Hard cut-off
conv.append_message(conv.roles[0], text)
conv.append_message(conv.roles[1], None)
return (state, state.to_gradio_chatbot()) + (disable_btn,) * 6
def post_process_code(code):
sep = "\n```"
if sep in code:
blocks = code.split(sep)
if len(blocks) % 2 == 1:
for i in range(1, len(blocks), 2):
blocks[i] = blocks[i].replace("\\_", "_")
code = sep.join(blocks)
return code
def model_worker_stream_iter(
conv,
model_name,
worker_addr,
prompt,
temperature,
repetition_penalty,
top_p,
max_new_tokens,
):
# Make requests
gen_params = {
"model": model_name,
"prompt": prompt,
"temperature": temperature,
"repetition_penalty": repetition_penalty,
"top_p": top_p,
"max_new_tokens": max_new_tokens,
"stop": conv.stop_str,
"stop_token_ids": conv.stop_token_ids,
"echo": False,
}
logger.info(f"==== request ====\n{gen_params}")
# Stream output
response = requests.post(
worker_addr + "/worker_generate_stream",
headers=headers,
json=gen_params,
stream=True,
timeout=WORKER_API_TIMEOUT,
)
for chunk in response.iter_lines(decode_unicode=False, delimiter=b"\0"):
if chunk:
data = json.loads(chunk.decode())
yield data
def bot_response_v2(state, temperature, top_p, max_new_tokens, request: gr.Request):
logger.info(f"bot_response. ip: {request.client.host}")
start_tstamp = time.time()
temperature = float(temperature)
top_p = float(top_p)
max_new_tokens = int(max_new_tokens)
if state.skip_next:
# This generate call is skipped due to invalid inputs
state.skip_next = False
yield (state, state.to_gradio_chatbot()) + (no_change_btn,) * 6
return
conv, model_name = state.conv, state.model_name
if model_name == "gpt-3.5-turbo" or model_name == "gpt-4":
prompt = conv.to_openai_api_messages()
stream_iter = openai_api_stream_iter(
model_name, prompt, temperature, top_p, max_new_tokens
)
elif model_name == "claude-2" or model_name == "claude-instant-1":
prompt = conv.get_prompt()
stream_iter = anthropic_api_stream_iter(
model_name, prompt, temperature, top_p, max_new_tokens
)
elif model_name == "palm-2":
stream_iter = palm_api_stream_iter(
state.palm_chat, conv.messages[-2][1], temperature, top_p, max_new_tokens
)
else:
# Query worker address
ret = requests.post(
controller_url + "/get_worker_address", json={"model": model_name}
)
worker_addr = ret.json()["address"]
logger.info(f"model_name: {model_name}, worker_addr: {worker_addr}")
# No available worker
if worker_addr == "":
conv.update_last_message(SERVER_ERROR_MSG)
yield (
state,
state.to_gradio_chatbot(),
disable_btn,
disable_btn,
disable_btn,
enable_btn,
enable_btn,
disable_btn,
)
return
# Construct prompt.
# We need to call it here, so it will not be affected by "▌".
prompt = conv.get_prompt()
# Set repetition_penalty
if "t5" in model_name:
repetition_penalty = 1.2
else:
repetition_penalty = 1.0
stream_iter = model_worker_stream_iter(
conv,
model_name,
worker_addr,
prompt,
temperature,
repetition_penalty,
top_p,
max_new_tokens,
)
conv.update_last_message("▌")
yield (state, state.to_gradio_chatbot()) + (disable_btn,) * 6
try:
for data in stream_iter:
if data["error_code"] == 0:
output = data["text"].strip()
if "vicuna" in model_name:
output = post_process_code(output)
conv.update_last_message(output + "▌")
yield (state, state.to_gradio_chatbot()) + (disable_btn,) * 6
else:
output = data["text"] + f"\n\n(error_code: {data['error_code']})"
conv.update_last_message(output)
yield (state, state.to_gradio_chatbot()) + (
disable_btn,
disable_btn,
disable_btn,
enable_btn,
enable_btn,
disable_btn,
)
return
time.sleep(0.015)
except requests.exceptions.RequestException as e:
conv.update_last_message(
f"{SERVER_ERROR_MSG}\n\n"
f"(error_code: {ErrorCode.GRADIO_REQUEST_ERROR}, {e})"
)
yield (state, state.to_gradio_chatbot()) + (
disable_btn,
disable_btn,
disable_btn,
enable_btn,
enable_btn,
disable_btn,
)
return
except Exception as e:
conv.update_last_message(
f"{SERVER_ERROR_MSG}\n\n"
f"(error_code: {ErrorCode.GRADIO_STREAM_UNKNOWN_ERROR}, {e})"
)
yield (state, state.to_gradio_chatbot()) + (
disable_btn,
disable_btn,
disable_btn,
enable_btn,
enable_btn,
disable_btn,
)
return
# Delete "▌"
conv.update_last_message(conv.messages[-1][-1][:-1])
yield (state, state.to_gradio_chatbot()) + (enable_btn,) * 6
finish_tstamp = time.time()
logger.info(f"{output}")
with open(get_conv_log_filename(), "a") as fout:
data = {
"tstamp": round(finish_tstamp, 4),
"type": "chat",
"model": model_name,
"gen_params": {
"temperature": temperature,
"top_p": top_p,
"max_new_tokens": max_new_tokens,
},
"start": round(start_tstamp, 4),
"finish": round(finish_tstamp, 4),
"state": state.dict(),
"ip": request.client.host,
}
fout.write(json.dumps(data) + "\n")
def add_answer(answer_num):
return_list = []
for i in range(int(answer_num)+1):
return_list.append(gr.update(visible=True))
for i in range(int(answer_num)+1, 10):
return_list.append(gr.update(visible=False))
return (gr.update(visible=True), ) * int(answer_num) + (gr.update(visible=True, value=""), ) + ("", ) * (9-int(answer_num)) + (answer_num+1, enable_btn)
def reduce_answer(answer_num):
if int(answer_num) - 1 == 2:
return (gr.update(visible=True), ) * int(answer_num-1) + (gr.update(visible=False, value=""), ) + ("", ) * (10-int(answer_num)) + (answer_num-1, disable_btn)
else:
return (gr.update(visible=True), ) * int(answer_num-1) + (gr.update(visible=False, value=""), ) + ("", ) * (10-int(answer_num)) + (answer_num-1, enable_btn)
block_css = """
h1 {
text-align: center;
}
#notice_markdown {
font-size: 104%
}
#notice_markdown th {
display: none;
}
#notice_markdown td {
padding-top: 6px;
padding-bottom: 6px;
}
#leaderboard_markdown {
font-size: 104%
}
#leaderboard_markdown td {
padding-top: 6px;
padding-bottom: 6px;
}
#leaderboard_dataframe td {
line-height: 0.1em;
}
"""
def get_model_description_md(models):
model_description_md = """
| | | |
| ---- | ---- | ---- |
"""
ct = 0
visited = set()
for i, name in enumerate(models):
if name in model_info:
minfo = model_info[name]
if minfo.simple_name in visited:
continue
visited.add(minfo.simple_name)
one_model_md = f"[{minfo.simple_name}]({minfo.link}): {minfo.description}"
else:
visited.add(name)
one_model_md = (
f"[{name}](): Add the description at fastchat/model/model_registry.py"
)
if ct % 3 == 0:
model_description_md += "|"
model_description_md += f" {one_model_md} |"
if ct % 3 == 2:
model_description_md += "\n"
ct += 1
return model_description_md
def build_single_model_ui_v2(model_selector, max_answer_num=10, reference=False, multi_answer=False, single_answer=False):
state = gr.State()
with gr.Row():
question_box = gr.Textbox(
placeholder="Enter the question text",
label="Question",
)
reference_box = None
if reference or single_answer:
with gr.Row():
reference_box = gr.Textbox(
placeholder="Enter the reference text",
label="Reference",
)
answer_list = []
with gr.Group():
if single_answer:
answer_box = gr.Textbox(
placeholder=f"Enter the assistant 1's answer text",
label=f"Assistant 1's Answer",
)
answer_list.append(answer_box)
else:
for i in range(2):
with gr.Row():
answer_box = gr.Textbox(
placeholder=f"Enter the assistant {i+1}'s answer text",
label=f"Assistant {i+1}'s Answer",
)
answer_list.append(answer_box)
for i in range(2, max_answer_num):
with gr.Row():
answer_box = gr.Textbox(
placeholder=f"Enter the assistant {i+1}'s answer text",
label=f"Assistant {i+1}'s Answer",
visible=False,
)
answer_list.append(answer_box)
with gr.Row().style():
answer_num = gr.Number(value=2, visible=False)
if multi_answer:
add_answer_btn = gr.Button(value="Add Answer")
reduce_answer_btn = gr.Button(value="Reduce Answer", interactive=False)
radio = gr.Radio(
label="Eval Type",
choices=["Detailed Eval", "Fast Eval"],
value="Detailed Eval",
)
with gr.Row().style(equal_height=True):
send_pretext_btn = gr.Button(value="Generate the judgement", variant="primary")
with gr.Row():
gr.Markdown('Click to add example as input.👇')
if reference:
with gr.Row():
gr.Examples(examples=examples_reference,
inputs=[question_box, reference_box, answer_list[0], answer_list[1]])
elif single_answer:
with gr.Row():
gr.Examples(examples=examples_reference,
inputs=[question_box, reference_box, answer_list[0]])
else:
with gr.Row():
gr.Examples(examples=examples,
inputs=[question_box, answer_list[0], answer_list[1]])
chatbot = gr.Chatbot(
elem_id="chatbot",
label="Scroll down and start chatting",
height=550,
)
with gr.Row():
with gr.Column(scale=20):
textbox = gr.Textbox(
show_label=False,
placeholder="Enter text and press ENTER",
)
with gr.Row():
send_btn = gr.Button(value="Send")
with gr.Row() as button_row:
upvote_btn = gr.Button(value="👍 Upvote", interactive=False)
downvote_btn = gr.Button(value="👎 Downvote", interactive=False)
flag_btn = gr.Button(value="⚠️ Flag", interactive=False)
regenerate_btn = gr.Button(value="🔄 Regenerate", interactive=False)
clear_btn = gr.Button(value="🗑️ Clear history", interactive=False)
with gr.Accordion("Parameters", open=False) as parameter_row:
temperature = gr.Slider(
minimum=0.0,
maximum=1.0,
value=0.2,
step=0.1,
interactive=True,
label="Temperature",
)
top_p = gr.Slider(
minimum=0.0,
maximum=1.0,
value=1.0,
step=0.1,
interactive=True,
label="Top P",
)
max_output_tokens = gr.Slider(
minimum=16,
maximum=2048,
value=2048,
step=64,
interactive=True,
label="Max output tokens",
)
gr.Markdown(learn_more_md)
# Register listeners
btn_list = [upvote_btn, downvote_btn, flag_btn, regenerate_btn, clear_btn, send_pretext_btn]
upvote_btn.click(
upvote_last_response,
[state, model_selector],
[textbox, upvote_btn, downvote_btn, flag_btn],
)
downvote_btn.click(
downvote_last_response,
[state, model_selector],
[textbox, upvote_btn, downvote_btn, flag_btn],
)
flag_btn.click(
flag_last_response,
[state, model_selector],
[textbox, upvote_btn, downvote_btn, flag_btn],
)
regenerate_btn.click(regenerate_v2, state, [state, chatbot, textbox] + btn_list).then(
bot_response_v2,
[state, temperature, top_p, max_output_tokens],
[state, chatbot] + btn_list,
)
clear_btn.click(clear_history_v2, answer_num, [state, chatbot, textbox, question_box] + answer_list + btn_list)
model_selector.change(clear_history_v2, answer_num, [state, chatbot, textbox, question_box] + answer_list + btn_list)
if multi_answer:
add_answer_btn.click(
add_answer, answer_num, answer_list + [answer_num, reduce_answer_btn]
)
reduce_answer_btn.click(
reduce_answer, answer_num, answer_list + [answer_num, reduce_answer_btn]
)
textbox.submit(
add_text_v2, [state, model_selector, textbox], [state, chatbot, textbox] + btn_list
).then(
bot_response_v2,
[state, temperature, top_p, max_output_tokens],
[state, chatbot] + btn_list,
)
send_btn.click(
add_text_v2, [state, model_selector, textbox], [state, chatbot, textbox] + btn_list
).then(
bot_response_v2,
[state, temperature, top_p, max_output_tokens],
[state, chatbot] + btn_list,
)
if reference:
send_pretext_btn.click(
add_pretext_reference,
[state, model_selector, radio, question_box, answer_num, reference_box] + answer_list,
[state, chatbot] + btn_list
).then(
bot_response_v2,
[state, temperature, top_p, max_output_tokens],
[state, chatbot] + btn_list,
)