-
Notifications
You must be signed in to change notification settings - Fork 382
/
app.py
1550 lines (1294 loc) · 75.7 KB
/
app.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
import os
import re
import time
import json
import copy
import types
from os import listdir
from os.path import isfile, join
import argparse
import gradio as gr
import global_vars
from chats import central
from transformers import AutoModelForCausalLM
from miscs.styles import MODEL_SELECTION_CSS
from miscs.js import GET_LOCAL_STORAGE, UPDATE_LEFT_BTNS_STATE, UPDATE_PLACEHOLDERS
from miscs.templates import templates
from utils import get_chat_manager, get_global_context
from pingpong.pingpong import PingPong
from pingpong.gradio import GradioAlpacaChatPPManager
from pingpong.gradio import GradioKoAlpacaChatPPManager
from pingpong.gradio import GradioStableLMChatPPManager
from pingpong.gradio import GradioFlanAlpacaChatPPManager
from pingpong.gradio import GradioOSStableLMChatPPManager
from pingpong.gradio import GradioVicunaChatPPManager
from pingpong.gradio import GradioStableVicunaChatPPManager
from pingpong.gradio import GradioStarChatPPManager
from pingpong.gradio import GradioMPTChatPPManager
from pingpong.gradio import GradioRedPajamaChatPPManager
from pingpong.gradio import GradioBaizeChatPPManager
# no cpu for
# - falcon families (too slow)
load_mode_list = ["cpu"]
ex_file = open("examples.txt", "r")
examples = ex_file.read().split("\n")
ex_btns = []
chl_file = open("channels.txt", "r")
channels = chl_file.read().split("\n")
channel_btns = []
default_ppm = GradioAlpacaChatPPManager()
default_ppm.ctx = "Context at top"
default_ppm.pingpongs = [
PingPong("user input #1...", "bot response #1..."),
PingPong("user input #2...", "bot response #2..."),
]
chosen_ppm = copy.deepcopy(default_ppm)
prompt_styles = {
"Alpaca": default_ppm,
"Baize": GradioBaizeChatPPManager(),
"Koalpaca": GradioKoAlpacaChatPPManager(),
"MPT": GradioMPTChatPPManager(),
"OpenAssistant StableLM": GradioOSStableLMChatPPManager(),
"RedPajama": GradioRedPajamaChatPPManager(),
"StableVicuna": GradioVicunaChatPPManager(),
"StableLM": GradioStableLMChatPPManager(),
"StarChat": GradioStarChatPPManager(),
"Vicuna": GradioVicunaChatPPManager(),
}
response_configs = [
f"configs/response_configs/{f}"
for f in listdir("configs/response_configs")
if isfile(join("configs/response_configs", f))
]
summarization_configs = [
f"configs/summarization_configs/{f}"
for f in listdir("configs/summarization_configs")
if isfile(join("configs/summarization_configs", f))
]
model_info = json.load(open("model_cards.json"))
table_data = []
for name, attributes in model_info.items():
thumbnail = attributes["thumb-tiny"]
parameters = float(attributes["parameters"])
olld_avg = float(attributes["ollb_average"])
olld_arc = float(attributes["ollb_arc"])
ollb_hellaswag = float(attributes["ollb_hellaswag"])
ollb_mmlu = float(attributes["ollb_mmlu"])
ollb_truthfulqa = float(attributes["ollb_truthfulqa"])
table_data.append(
[f"", name, parameters, olld_avg, olld_arc, ollb_hellaswag, ollb_mmlu, ollb_truthfulqa]
)
table_data.sort(key=lambda elem: elem[3], reverse=True)
###
def move_to_second_view_from_tb(tb, evt: gr.SelectData):
selected_model = tb.iloc[evt.index[0]]['Model']
info = model_info[selected_model]
guard_vram = 2 * 1024.
vram_req_full = int(info["vram(full)"]) + guard_vram
vram_req_8bit = int(info["vram(8bit)"]) + guard_vram
vram_req_4bit = int(info["vram(4bit)"]) + guard_vram
vram_req_gptq = info["vram(gptq)"]
if vram_req_gptq != "N/A":
vram_req_gptq = int(vram_req_gptq) + guard_vram
load_mode_list = []
if global_vars.cuda_availability:
print(f"total vram = {global_vars.available_vrams_mb}")
print(f"required vram(full={info['vram(full)']}, 8bit={info['vram(8bit)']}, 4bit={info['vram(4bit)']})")
if global_vars.available_vrams_mb >= vram_req_full:
load_mode_list.append("gpu(half)")
if global_vars.available_vrams_mb >= vram_req_8bit:
load_mode_list.append("gpu(load_in_8bit)")
if global_vars.available_vrams_mb >= vram_req_4bit:
load_mode_list.append("gpu(load_in_4bit)")
if vram_req_gptq != "N/A" and global_vars.available_vrams_mb >= vram_req_gptq:
load_mode_list.append("gpu(gptq)")
if global_vars.mps_availability:
load_mode_list.append("apple silicon")
# load_mode_list.append("apple silicon(gptq)")
# load_mode_list.append("cpu(gptq)")
load_mode_list.append("cpu")
load_mode_list.append("remote(TGI)")
print(info['hub(gptq_base)'])
vram_req_gptq_in_gb = vram_req_gptq
if vram_req_gptq != "N/A":
vram_req_gptq_in_gb = f"{round(vram_req_gptq_in_gb/1024., 1)}GiB"
return (
gr.update(visible=False),
gr.update(visible=True),
info["thumb"],
f"## {selected_model}",
f"**Parameters**\n: Approx. {info['parameters']}",
f"**Hugging Face Hub(base)**\n: {info['hub(base)']}",
f"**Hugging Face Hub(LoRA)**\n: {info['hub(ckpt)']}",
f"**Hugging Face Hub(GPTQ)**\n: {info['hub(gptq)']}",
f"**Hugging Face Hub(GPTQ_BASE)**\n: {info['hub(gptq_base)']}",
info['desc'],
f"""**Min VRAM requirements** :
| half precision | load_in_8bit | load_in_4bit |
| ------------------------------------- | ---------------------------------- | ---------------------------------- |
| {round(vram_req_full/1024., 1)}GiB | {round(vram_req_8bit/1024., 1)}GiB | {round(vram_req_4bit/1024., 1)}GiB |
| GPTQ |
| ------------------------------------- |
| {vram_req_gptq_in_gb} |
""",
info['default_gen_config'],
info['example1'],
info['example2'],
info['example3'],
info['example4'],
info['thumb-tiny'],
gr.update(choices=load_mode_list, value=load_mode_list[0]),
"",
)
def model_view_toggle(toggler):
if toggler == "Icon View(Recent)":
return (gr.update(visible=True), gr.update(visible=False), gr.update(visible=False), " ")
elif toggler == "Icon View(Full)":
return (gr.update(visible=False), gr.update(visible=True), gr.update(visible=False), " ")
else:
return (gr.update(visible=False), gr.update(visible=False), gr.update(visible=True), " ")
def get_placeholders(text):
"""Returns all substrings in between <placeholder> and </placeholder>."""
pattern = r"\[([^\]]*)\]"
matches = re.findall(pattern, text)
return matches
def fill_up_placeholders(txt):
placeholders = get_placeholders(txt)
highlighted_txt = txt
return (
gr.update(
visible=True,
value=highlighted_txt
),
gr.update(
visible=True if len(placeholders) >= 1 else False,
placeholder=placeholders[0] if len(placeholders) >= 1 else ""
),
gr.update(
visible=True if len(placeholders) >= 2 else False,
placeholder=placeholders[1] if len(placeholders) >= 2 else ""
),
gr.update(
visible=True if len(placeholders) >= 3 else False,
placeholder=placeholders[2] if len(placeholders) >= 3 else ""
),
"" if len(placeholders) >= 1 else txt
)
def get_final_template(
txt, placeholder_txt1, placeholder_txt2, placeholder_txt3
):
placeholders = get_placeholders(txt)
example_prompt = txt
if len(placeholders) >= 1:
if placeholder_txt1 != "":
example_prompt = example_prompt.replace(f"[{placeholders[0]}]", placeholder_txt1)
if len(placeholders) >= 2:
if placeholder_txt2 != "":
example_prompt = example_prompt.replace(f"[{placeholders[1]}]", placeholder_txt2)
if len(placeholders) >= 3:
if placeholder_txt3 != "":
example_prompt = example_prompt.replace(f"[{placeholders[2]}]", placeholder_txt3)
return (
example_prompt,
"",
"",
""
)
###
def move_to_model_select_view():
return (
"move to model select view",
gr.update(visible=False),
gr.update(visible=True),
)
def use_chosen_model():
try:
test = global_vars.model
except AttributeError:
try:
test2 = global_vars.remote_addr
if global_vars.remote_addr.strip() == "":
raise gr.Error("There is no previously chosen model")
except AttributeError:
raise gr.Error("There is no previously chosen model")
gen_config = global_vars.gen_config
gen_sum_config = global_vars.gen_config_summarization
if global_vars.model_type == "custom":
ppmanager_type = chosen_ppm
else:
ppmanager_type = get_chat_manager(global_vars.model_type)
return (
"Preparation done!",
gr.update(visible=False),
gr.update(visible=True),
gr.update(label=global_vars.model_name),
{
"ppmanager_type": ppmanager_type,
"model_type": global_vars.model_type,
},
get_global_context(global_vars.model_type),
gen_config.temperature,
gen_config.top_p,
gen_config.top_k,
gen_config.repetition_penalty,
gen_config.max_new_tokens,
gen_config.num_beams,
gen_config.use_cache,
gen_config.do_sample,
gen_config.eos_token_id,
gen_config.pad_token_id,
gen_sum_config.temperature,
gen_sum_config.top_p,
gen_sum_config.top_k,
gen_sum_config.repetition_penalty,
gen_sum_config.max_new_tokens,
gen_sum_config.num_beams,
gen_sum_config.use_cache,
gen_sum_config.do_sample,
gen_sum_config.eos_token_id,
gen_sum_config.pad_token_id,
)
def move_to_byom_view():
load_mode_list = []
if global_vars.cuda_availability:
load_mode_list.extend(["gpu(half)", "gpu(load_in_8bit)", "gpu(load_in_4bit)"])
if global_vars.mps_availability:
load_mode_list.append("apple silicon")
load_mode_list.append("cpu")
return (
"move to the byom view",
gr.update(visible=False),
gr.update(visible=True),
gr.update(choices=load_mode_list, value=load_mode_list[0])
)
def prompt_style_change(key):
ppm = prompt_styles[key]
ppm.ctx = "Context at top"
ppm.pingpongs = [
PingPong("user input #1...", "bot response #1..."),
PingPong("user input #2...", "bot response #2..."),
]
chosen_ppm = copy.deepcopy(ppm)
chosen_ppm.ctx = ""
chosen_ppm.pingpongs = []
return ppm.build_prompts()
def byom_load(
base, ckpt, model_cls, tokenizer_cls,
bos_token_id, eos_token_id, pad_token_id,
load_mode,
):
# mode_cpu, model_mps, mode_8bit, mode_4bit, mode_full_gpu
global_vars.initialize_globals_byom(
base, ckpt, model_cls, tokenizer_cls,
bos_token_id, eos_token_id, pad_token_id,
True if load_mode == "cpu" else False,
True if load_mode == "apple silicon" else False,
True if load_mode == "8bit" else False,
True if load_mode == "4bit" else False,
True if load_mode == "gpu(half)" else False,
)
return (
""
)
def channel_num(btn_title):
choice = 0
for idx, channel in enumerate(channels):
if channel == btn_title:
choice = idx
return choice
def set_chatbot(btn, ld, state):
choice = channel_num(btn)
res = [state["ppmanager_type"].from_json(json.dumps(ppm_str)) for ppm_str in ld]
empty = len(res[choice].pingpongs) == 0
return (res[choice].build_uis(), choice, gr.update(visible=empty), gr.update(interactive=not empty))
def set_example(btn):
return btn, gr.update(visible=False)
def set_popup_visibility(ld, example_block):
return example_block
def move_to_second_view(btn):
info = model_info[btn]
guard_vram = 2 * 1024.
vram_req_full = int(info["vram(full)"]) + guard_vram
vram_req_8bit = int(info["vram(8bit)"]) + guard_vram
vram_req_4bit = int(info["vram(4bit)"]) + guard_vram
vram_req_gptq = info["vram(gptq)"]
if vram_req_gptq != "N/A":
vram_req_gptq = int(vram_req_gptq) + guard_vram
load_mode_list = []
if global_vars.cuda_availability:
print(f"total vram = {global_vars.available_vrams_mb}")
print(f"required vram(full={info['vram(full)']}, 8bit={info['vram(8bit)']}, 4bit={info['vram(4bit)']})")
if global_vars.available_vrams_mb >= vram_req_full:
load_mode_list.append("gpu(half)")
if global_vars.available_vrams_mb >= vram_req_8bit:
load_mode_list.append("gpu(load_in_8bit)")
if global_vars.available_vrams_mb >= vram_req_4bit:
load_mode_list.append("gpu(load_in_4bit)")
if vram_req_gptq != "N/A" and global_vars.available_vrams_mb >= vram_req_gptq:
load_mode_list.append("gpu(gptq)")
if global_vars.mps_availability:
load_mode_list.append("apple silicon")
# load_mode_list.append("apple silicon(gptq)")
# load_mode_list.append("cpu(gptq)")
load_mode_list.append("cpu")
load_mode_list.append("remote(TGI)")
print(info['hub(gptq_base)'])
vram_req_gptq_in_gb = vram_req_gptq
if vram_req_gptq != "N/A":
vram_req_gptq_in_gb = f"{round(vram_req_gptq_in_gb/1024., 1)}GiB"
return (
gr.update(visible=False),
gr.update(visible=True),
info["thumb"],
f"## {btn}",
f"**Parameters**\n: Approx. {info['parameters']}",
f"**Hugging Face Hub(base)**\n: {info['hub(base)']}",
f"**Hugging Face Hub(LoRA)**\n: {info['hub(ckpt)']}",
f"**Hugging Face Hub(GPTQ)**\n: {info['hub(gptq)']}",
f"**Hugging Face Hub(GPTQ_BASE)**\n: {info['hub(gptq_base)']}",
info['desc'],
f"""**Min VRAM requirements** :
| half precision | load_in_8bit | load_in_4bit |
| ------------------------------------- | ---------------------------------- | ---------------------------------- |
| {round(vram_req_full/1024., 1)}GiB | {round(vram_req_8bit/1024., 1)}GiB | {round(vram_req_4bit/1024., 1)}GiB |
| GPTQ |
| ------------------------------------- |
| {vram_req_gptq_in_gb} |
""",
info['default_gen_config'],
info['example1'],
info['example2'],
info['example3'],
info['example4'],
info['thumb-tiny'],
gr.update(choices=load_mode_list, value=load_mode_list[0]),
"",
)
def move_to_first_view():
return (gr.update(visible=True), gr.update(visible=False))
def download_completed(
model_name,
model_base,
model_ckpt,
model_gptq,
model_gptq_base,
gen_config_path,
gen_config_sum_path,
load_mode,
thumbnail_tiny,
force_download,
remote_addr,
remote_port,
remote_token
):
global local_files_only
print(f"model_name: {model_name}")
print(f"model_base: {model_base}")
tmp_args = types.SimpleNamespace()
tmp_args.model_name = model_name[3:]
tmp_args.base_url = model_base.split(":")[-1].strip()
tmp_args.ft_ckpt_url = model_ckpt.split(":")[-1].strip()
tmp_args.gptq_url = model_gptq.split(":")[-1].strip()
tmp_args.gptq_base_url = model_gptq_base.split(":")[-1].strip().replace(' ', '')
tmp_args.gen_config_path = gen_config_path
tmp_args.gen_config_summarization_path = gen_config_sum_path
tmp_args.force_download_ckpt = force_download
tmp_args.thumbnail_tiny = thumbnail_tiny
tmp_args.mode_cpu = True if load_mode == "cpu" else False
tmp_args.mode_mps = True if load_mode == "apple silicon" else False
tmp_args.mode_8bit = True if load_mode == "gpu(load_in_8bit)" else False
tmp_args.mode_4bit = True if load_mode == "gpu(load_in_4bit)" else False
tmp_args.mode_gptq = True if load_mode == "gpu(gptq)" else False
tmp_args.mode_mps_gptq = True if load_mode == "apple silicon(gptq)" else False
tmp_args.mode_cpu_gptq = True if load_mode == "cpu(gptq)" else False
tmp_args.mode_full_gpu = True if load_mode == "gpu(half)" else False
tmp_args.mode_remote_tgi = True if load_mode == "remote(TGI)" else False
tmp_args.local_files_only = local_files_only
tmp_args.remote_addr = remote_addr
tmp_args.remote_port = remote_port
tmp_args.remote_token = remote_token
try:
global_vars.initialize_globals(tmp_args)
except RuntimeError as e:
raise gr.Error("GPU memory is not enough to load this model.")
return "Download completed!"
def move_to_third_view():
gen_config = global_vars.gen_config
gen_sum_config = global_vars.gen_config_summarization
if global_vars.model_type == "custom":
ppmanager_type = chosen_ppm
else:
ppmanager_type = get_chat_manager(global_vars.model_type)
return (
"Preparation done!",
gr.update(visible=False),
gr.update(visible=True),
gr.update(label=global_vars.model_name),
{
"ppmanager_type": ppmanager_type,
"model_type": global_vars.model_type,
},
get_global_context(global_vars.model_type),
gen_config.temperature,
gen_config.top_p,
gen_config.top_k,
gen_config.repetition_penalty,
gen_config.max_new_tokens,
gen_config.num_beams,
gen_config.use_cache,
gen_config.do_sample,
gen_config.eos_token_id,
gen_config.pad_token_id,
gen_sum_config.temperature,
gen_sum_config.top_p,
gen_sum_config.top_k,
gen_sum_config.repetition_penalty,
gen_sum_config.max_new_tokens,
gen_sum_config.num_beams,
gen_sum_config.use_cache,
gen_sum_config.do_sample,
gen_sum_config.eos_token_id,
gen_sum_config.pad_token_id,
)
def toggle_inspector(view_selector):
if view_selector == "with context inspector":
return gr.update(visible=True)
else:
return gr.update(visible=False)
def reset_chat(idx, ld, state):
res = [state["ppmanager_type"].from_json(json.dumps(ppm_str)) for ppm_str in ld]
res[idx].pingpongs = []
return (
"",
[],
str(res),
gr.update(visible=True),
gr.update(interactive=False),
)
def rollback_last(idx, ld, state):
res = [state["ppmanager_type"].from_json(json.dumps(ppm_str)) for ppm_str in ld]
last_user_message = res[idx].pingpongs[-1].ping
res[idx].pingpongs = res[idx].pingpongs[:-1]
return (
last_user_message,
res[idx].build_uis(),
str(res),
gr.update(interactive=False)
)
def gradio_main(args):
global local_files_only
local_files_only = args.local_files_only
with gr.Blocks(css=MODEL_SELECTION_CSS, theme='gradio/soft') as demo:
with gr.Column(visible=True, elem_id="landing-container") as landing_view:
gr.Markdown("# Chat with LLM", elem_classes=["center"])
with gr.Row(elem_id="landing-container-selection"):
with gr.Column():
gr.Markdown(
"This is the landing page of the project, [LLM As Chatbot](https://github.com/deep-diver/LLM-As-Chatbot). "
"This appliction is designed for personal use only. A single model will be selected at a time even if you "
"open up a new browser or a tab. As an initial choice, please select one of the following menu"
)
gr.Markdown(
"**Bring your own model**: You can chat with arbitrary models. If your own custom model is based on "
"🤗 Hugging Face's [transformers](https://huggingface.co/docs/transformers/index) library, you will "
"propbably be able to bring it into this application with this menu \n\n"
"**Select a model from model pool**: You can chat with one of the popular open source Large Language Model \n\n"
"**Use currently selected model**: If you have already selected, but if you came back to this landing page "
"accidently, you can directly go back to the chatting mode with this menu"
)
with gr.Row():
byom = gr.Button("custom model", elem_id="go-byom-select", elem_classes=["square", "landing-btn"])
select_model = gr.Button("model selection", elem_id="go-model-select", elem_classes=["square", "landing-btn"])
chosen_model = gr.Button("back to current model", elem_id="go-use-selected-model", elem_classes=["square", "landing-btn"])
with gr.Column(elem_id="landing-bottom"):
progress_view0 = gr.Textbox(label="Progress", elem_classes=["progress-view"])
gr.Markdown("""[project](https://github.com/deep-diver/LLM-As-Chatbot)
[developer](https://github.com/deep-diver)
""", elem_classes=["center"])
with gr.Column(visible=False, elem_id="model-selection-container") as model_choice_view:
gr.Markdown("# Choose a Model", elem_classes=["center"])
with gr.Row(elem_id="container"):
with gr.Column():
recent_normal_toggler = gr.Radio(
choices=["Icon View(Recent)", "Table View", "Icon View(Full)"], value="Icon View(Recent)",
label="Model list view modes", info="If you want to explore all models, choose Full Models"
)
with gr.Column(visible=False) as table_section:
gr.Markdown("## 🤗 Open LLM Leaderboard")
gr.Markdown(
"This view organizes the list of models based on [🤗 Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard). "
"Not all models are evaluated on the leader board, so those models' score is indicated with the value `-1`. Also, this application does not "
"come with all the open source LLMs on the leader board as well. That is because the actual functionalities are not fully tested, so if you "
"want to add more models in this application, please write an [issue](https://github.com/deep-diver/LLM-As-Chatbot/issues) for that."
)
gr.Markdown(
"If you are curious how the models are evaluated and what each score categories are, please find them on [🤗 Open LLM Leaderboard]"
"(https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard). For quick reference, please visit [ARC(AI2 Reasoning Challenge)]"
"(https://arxiv.org/abs/1803.05457), [HellaSwag](https://arxiv.org/abs/1905.07830), [MMLU(Measuring Massive Multitask Language Understanding)]"
"(https://arxiv.org/abs/2009.03300), and [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://arxiv.org/abs/2109.07958)."
)
model_table_view = gr.Dataframe(
value=table_data,
headers=["Icon", "Model", "Params(B)", "Avg.", "ARC", "HellaSwag", "MMLU", "TruthfulQA"],
datatype=["markdown", "str", "number", "number", "number", "number", "number", "number"],
col_count=(8, "fixed"),
row_count=1,
height=1000,
interactive=False,
wrap=True
)
with gr.Column() as recent_section:
gr.Markdown("## Recent Releases")
with gr.Row(elem_classes=["sub-container"]):
with gr.Column(min_width=20):
mistral_7b_rr = gr.Button("mistral-7b", elem_id="mistral-7b", elem_classes=["square"])
gr.Markdown("Mistral (7B)", elem_classes=["center"])
with gr.Column(min_width=20):
zephyr_7b_rr = gr.Button("zephyr-7b", elem_id="zephyr-7b", elem_classes=["square"])
gr.Markdown("Zephyr (7B)", elem_classes=["center"])
with gr.Column(min_width=20):
mistral_trismegistus_7b_rr = gr.Button("mistral-trismegistus-7b", elem_id="mistral-trismegistus-7b", elem_classes=["square"])
gr.Markdown("Mistral Trismegistus (7B)", elem_classes=["center"])
with gr.Column(min_width=20):
hermes_trismegistus_7b_rr = gr.Button("hermes-trismegistus-7b", elem_id="hermes-trismegistus-7b", elem_classes=["square"])
gr.Markdown("Hermes Trismegistus (7B)", elem_classes=["center"])
with gr.Column(min_width=20):
mistral_openhermes_2_5_7b_rr = gr.Button("mistral-openherems-2-5-7b", elem_id="mistral-openherems-2-5-7b", elem_classes=["square"])
gr.Markdown("Mistral OpenHermes 2.5 (7B)", elem_classes=["center"])
with gr.Column(visible=False) as full_section:
gr.Markdown("## ~ 10B Parameters")
with gr.Row(elem_classes=["sub-container"]):
with gr.Column(min_width=20):
t5_vicuna_3b = gr.Button("t5-vicuna-3b", elem_id="t5-vicuna-3b", elem_classes=["square"])
gr.Markdown("T5 Vicuna", elem_classes=["center"])
with gr.Column(min_width=20, visible=False):
flan3b = gr.Button("flan-3b", elem_id="flan-3b", elem_classes=["square"])
gr.Markdown("Flan-XL", elem_classes=["center"])
with gr.Column(min_width=20):
camel5b = gr.Button("camel-5b", elem_id="camel-5b", elem_classes=["square"])
gr.Markdown("Camel", elem_classes=["center"])
with gr.Column(min_width=20):
alpaca_lora7b = gr.Button("alpaca-lora-7b", elem_id="alpaca-lora-7b", elem_classes=["square"])
gr.Markdown("Alpaca-LoRA", elem_classes=["center"])
with gr.Column(min_width=20):
stablelm7b = gr.Button("stablelm-7b", elem_id="stablelm-7b", elem_classes=["square"])
gr.Markdown("StableLM", elem_classes=["center"])
with gr.Column(min_width=20, visible=False):
os_stablelm7b = gr.Button("os-stablelm-7b", elem_id="os-stablelm-7b", elem_classes=["square"])
gr.Markdown("OA+StableLM", elem_classes=["center"])
with gr.Column(min_width=20):
gpt4_alpaca_7b = gr.Button("gpt4-alpaca-7b", elem_id="gpt4-alpaca-7b", elem_classes=["square"])
gr.Markdown("GPT4-Alpaca-LoRA", elem_classes=["center"])
with gr.Column(min_width=20):
mpt_7b = gr.Button("mpt-7b", elem_id="mpt-7b", elem_classes=["square"])
gr.Markdown("MPT", elem_classes=["center"])
with gr.Column(min_width=20):
redpajama_7b = gr.Button("redpajama-7b", elem_id="redpajama-7b", elem_classes=["square"])
gr.Markdown("RedPajama", elem_classes=["center"])
with gr.Column(min_width=20, visible=False):
redpajama_instruct_7b = gr.Button("redpajama-instruct-7b", elem_id="redpajama-instruct-7b", elem_classes=["square"])
gr.Markdown("RedPajama Instruct", elem_classes=["center"])
with gr.Column(min_width=20):
vicuna_7b = gr.Button("vicuna-7b", elem_id="vicuna-7b", elem_classes=["square"])
gr.Markdown("Vicuna", elem_classes=["center"])
with gr.Column(min_width=20):
vicuna_7b_1_3 = gr.Button("vicuna-7b-1-3", elem_id="vicuna-7b-1-3", elem_classes=["square"])
gr.Markdown("Vicuna 1.3", elem_classes=["center"])
with gr.Column(min_width=20):
llama_deus_7b = gr.Button("llama-deus-7b", elem_id="llama-deus-7b",elem_classes=["square"])
gr.Markdown("LLaMA Deus", elem_classes=["center"])
with gr.Column(min_width=20):
evolinstruct_vicuna_7b = gr.Button("evolinstruct-vicuna-7b", elem_id="evolinstruct-vicuna-7b", elem_classes=["square"])
gr.Markdown("EvolInstruct Vicuna", elem_classes=["center"])
with gr.Column(min_width=20, visible=False):
alpacoom_7b = gr.Button("alpacoom-7b", elem_id="alpacoom-7b", elem_classes=["square"])
gr.Markdown("Alpacoom", elem_classes=["center"])
with gr.Column(min_width=20):
baize_7b = gr.Button("baize-7b", elem_id="baize-7b", elem_classes=["square"])
gr.Markdown("Baize", elem_classes=["center"])
with gr.Column(min_width=20):
guanaco_7b = gr.Button("guanaco-7b", elem_id="guanaco-7b", elem_classes=["square"])
gr.Markdown("Guanaco", elem_classes=["center"])
with gr.Column(min_width=20):
falcon_7b = gr.Button("falcon-7b", elem_id="falcon-7b", elem_classes=["square"])
gr.Markdown("Falcon", elem_classes=["center"])
with gr.Column(min_width=20):
wizard_falcon_7b = gr.Button("wizard-falcon-7b", elem_id="wizard-falcon-7b", elem_classes=["square"])
gr.Markdown("Wizard Falcon", elem_classes=["center"])
with gr.Column(min_width=20):
airoboros_7b = gr.Button("airoboros-7b", elem_id="airoboros-7b", elem_classes=["square"])
gr.Markdown("Airoboros", elem_classes=["center"])
with gr.Column(min_width=20):
samantha_7b = gr.Button("samantha-7b", elem_id="samantha-7b", elem_classes=["square"])
gr.Markdown("Samantha", elem_classes=["center"])
with gr.Column(min_width=20):
openllama_7b = gr.Button("openllama-7b", elem_id="openllama-7b", elem_classes=["square"])
gr.Markdown("OpenLLaMA", elem_classes=["center"])
with gr.Column(min_width=20):
orcamini_7b = gr.Button("orcamini-7b", elem_id="orcamini-7b", elem_classes=["square"])
gr.Markdown("Orca Mini", elem_classes=["center"])
with gr.Column(min_width=20):
xgen_7b = gr.Button("xgen-7b", elem_id="xgen-7b", elem_classes=["square"])
gr.Markdown("XGen", elem_classes=["center"])
with gr.Column(min_width=20):
llama2_7b = gr.Button("llama2-7b", elem_id="llama2-7b", elem_classes=["square"])
gr.Markdown("LLaMA 2", elem_classes=["center"])
with gr.Column(min_width=20):
nous_hermes_7b_v2 = gr.Button("nous-hermes-7b-llama2", elem_id="nous-hermes-7b-llama2", elem_classes=["square"])
gr.Markdown("Nous Hermes 2", elem_classes=["center"])
with gr.Column(min_width=20):
codellama_7b = gr.Button("codellama-7b", elem_id="codellama-7b", elem_classes=["square"])
gr.Markdown("Code LLaMA", elem_classes=["center"])
with gr.Column(min_width=20):
mistral_7b = gr.Button("mistral-7b", elem_id="mistral-7b", elem_classes=["square"])
gr.Markdown("Mistral", elem_classes=["center"])
with gr.Column(min_width=20):
zephyr_7b = gr.Button("zephyr-7b", elem_id="zephyr-7b", elem_classes=["square"])
gr.Markdown("Zephyr", elem_classes=["center"])
with gr.Column(min_width=20):
mistral_trismegistus_7b = gr.Button("mistral-trismegistus-7b", elem_id="mistral-trismegistus-7b", elem_classes=["square"])
gr.Markdown("Mistral Trismegistus (7B)", elem_classes=["center"])
with gr.Column(min_width=20):
hermes_trismegistus_7b = gr.Button("hermes-trismegistus-7b", elem_id="hermes-trismegistus-7b", elem_classes=["square"])
gr.Markdown("Hermes Trismegistus (7B)", elem_classes=["center"])
with gr.Column(min_width=20):
mistral_openhermes_2_5_7b = gr.Button("mistral-openherems-2-5-7b", elem_id="mistral-openherems-2-5-7b", elem_classes=["square"])
gr.Markdown("Mistral OpenHermes 2.5 (7B)", elem_classes=["center"])
gr.Markdown("## ~ 20B Parameters")
with gr.Row(elem_classes=["sub-container"]):
with gr.Column(min_width=20, visible=False):
flan11b = gr.Button("flan-11b", elem_id="flan-11b", elem_classes=["square"])
gr.Markdown("Flan-XXL", elem_classes=["center"])
with gr.Column(min_width=20):
koalpaca = gr.Button("koalpaca", elem_id="koalpaca", elem_classes=["square"])
gr.Markdown("koalpaca", elem_classes=["center"])
with gr.Column(min_width=20):
kullm = gr.Button("kullm", elem_id="kullm", elem_classes=["square"])
gr.Markdown("KULLM", elem_classes=["center"])
with gr.Column(min_width=20):
alpaca_lora13b = gr.Button("alpaca-lora-13b", elem_id="alpaca-lora-13b", elem_classes=["square"])
gr.Markdown("Alpaca-LoRA", elem_classes=["center"])
with gr.Column(min_width=20):
gpt4_alpaca_13b = gr.Button("gpt4-alpaca-13b", elem_id="gpt4-alpaca-13b", elem_classes=["square"])
gr.Markdown("GPT4-Alpaca-LoRA", elem_classes=["center"])
with gr.Column(min_width=20):
stable_vicuna_13b = gr.Button("stable-vicuna-13b", elem_id="stable-vicuna-13b", elem_classes=["square"])
gr.Markdown("Stable-Vicuna", elem_classes=["center"])
with gr.Column(min_width=20):
starchat_15b = gr.Button("starchat-15b", elem_id="starchat-15b", elem_classes=["square"])
gr.Markdown("StarChat", elem_classes=["center"])
with gr.Column(min_width=20):
starchat_beta_15b = gr.Button("starchat-beta-15b", elem_id="starchat-beta-15b", elem_classes=["square"])
gr.Markdown("StarChat β", elem_classes=["center"])
with gr.Column(min_width=20):
vicuna_13b = gr.Button("vicuna-13b", elem_id="vicuna-13b", elem_classes=["square"])
gr.Markdown("Vicuna", elem_classes=["center"])
with gr.Column(min_width=20):
vicuna_13b_1_3 = gr.Button("vicuna-13b-1-3", elem_id="vicuna-13b-1-3", elem_classes=["square"])
gr.Markdown("Vicuna 1.3", elem_classes=["center"])
with gr.Column(min_width=20):
evolinstruct_vicuna_13b = gr.Button("evolinstruct-vicuna-13b", elem_id="evolinstruct-vicuna-13b", elem_classes=["square"])
gr.Markdown("EvolInstruct Vicuna", elem_classes=["center"])
with gr.Column(min_width=20):
baize_13b = gr.Button("baize-13b", elem_id="baize-13b", elem_classes=["square"])
gr.Markdown("Baize", elem_classes=["center"])
with gr.Column(min_width=20):
guanaco_13b = gr.Button("guanaco-13b", elem_id="guanaco-13b", elem_classes=["square"])
gr.Markdown("Guanaco", elem_classes=["center"])
with gr.Column(min_width=20):
nous_hermes_13b = gr.Button("nous-hermes-13b", elem_id="nous-hermes-13b", elem_classes=["square"])
gr.Markdown("Nous Hermes", elem_classes=["center"])
with gr.Column(min_width=20):
airoboros_13b = gr.Button("airoboros-13b", elem_id="airoboros-13b", elem_classes=["square"])
gr.Markdown("Airoboros", elem_classes=["center"])
with gr.Column(min_width=20):
samantha_13b = gr.Button("samantha-13b", elem_id="samantha-13b", elem_classes=["square"])
gr.Markdown("Samantha", elem_classes=["center"])
with gr.Column(min_width=20):
chronos_13b = gr.Button("chronos-13b", elem_id="chronos-13b", elem_classes=["square"])
gr.Markdown("Chronos", elem_classes=["center"])
with gr.Column(min_width=20):
wizardlm_13b = gr.Button("wizardlm-13b", elem_id="wizardlm-13b", elem_classes=["square"])
gr.Markdown("WizardLM", elem_classes=["center"])
with gr.Column(min_width=20):
wizard_vicuna_13b = gr.Button("wizard-vicuna-13b", elem_id="wizard-vicuna-13b", elem_classes=["square"])
gr.Markdown("Wizard Vicuna (Uncensored)", elem_classes=["center"])
with gr.Column(min_width=20):
wizard_coder_15b = gr.Button("wizard-coder-15b", elem_id="wizard-coder-15b", elem_classes=["square"])
gr.Markdown("Wizard Coder", elem_classes=["center"])
with gr.Column(min_width=20):
openllama_13b = gr.Button("openllama-13b", elem_id="openllama-13b", elem_classes=["square"])
gr.Markdown("OpenLLaMA", elem_classes=["center"])
with gr.Column(min_width=20):
orcamini_13b = gr.Button("orcamini-13b", elem_id="orcamini-13b", elem_classes=["square"])
gr.Markdown("Orca Mini", elem_classes=["center"])
with gr.Column(min_width=20):
llama2_13b = gr.Button("llama2-13b", elem_id="llama2-13b", elem_classes=["square"])
gr.Markdown("LLaMA 2", elem_classes=["center"])
with gr.Column(min_width=20):
nous_hermes_13b_v2 = gr.Button("nous-hermes-13b-llama2", elem_id="nous-hermes-13b-llama2", elem_classes=["square"])
gr.Markdown("Nous Hermes 2", elem_classes=["center"])
with gr.Column(min_width=20):
nous_puffin_13b_v2 = gr.Button("nous-puffin-13b-llama2", elem_id="nous-puffin-13b-llama2", elem_classes=["square"])
gr.Markdown("Nous Puffin 2", elem_classes=["center"])
with gr.Column(min_width=20):
wizardlm_13b_1_2 = gr.Button("wizardlm-13b-1-2", elem_id="wizardlm-13b-1-2", elem_classes=["square"])
gr.Markdown("WizardLM 1.2", elem_classes=["center"])
with gr.Column(min_width=20):
codellama_13b = gr.Button("codellama-13b", elem_id="codellama-13b", elem_classes=["square"])
gr.Markdown("Code LLaMA", elem_classes=["center"])
gr.Markdown("## ~ 30B Parameters", visible=False)
with gr.Row(elem_classes=["sub-container"], visible=False):
with gr.Column(min_width=20):
camel20b = gr.Button("camel-20b", elem_id="camel-20b", elem_classes=["square"])
gr.Markdown("Camel", elem_classes=["center"])
gr.Markdown("## ~ 40B Parameters")
with gr.Row(elem_classes=["sub-container"]):
with gr.Column(min_width=20):
guanaco_33b = gr.Button("guanaco-33b", elem_id="guanaco-33b", elem_classes=["square"])
gr.Markdown("Guanaco", elem_classes=["center"])
with gr.Column(min_width=20):
falcon_40b = gr.Button("falcon-40b", elem_id="falcon-40b", elem_classes=["square"])
gr.Markdown("Falcon", elem_classes=["center"])
with gr.Column(min_width=20):
wizard_falcon_40b = gr.Button("wizard-falcon-40b", elem_id="wizard-falcon-40b", elem_classes=["square"])
gr.Markdown("Wizard Falcon", elem_classes=["center"])
with gr.Column(min_width=20):
samantha_33b = gr.Button("samantha-33b", elem_id="samantha-33b", elem_classes=["square"])
gr.Markdown("Samantha", elem_classes=["center"])
with gr.Column(min_width=20):
lazarus_30b = gr.Button("lazarus-30b", elem_id="lazarus-30b", elem_classes=["square"])
gr.Markdown("Lazarus", elem_classes=["center"])
with gr.Column(min_width=20):
chronos_33b = gr.Button("chronos-33b", elem_id="chronos-33b", elem_classes=["square"])
gr.Markdown("Chronos", elem_classes=["center"])
with gr.Column(min_width=20):
wizardlm_30b = gr.Button("wizardlm-30b", elem_id="wizardlm-30b", elem_classes=["square"])
gr.Markdown("WizardLM", elem_classes=["center"])
with gr.Column(min_width=20):
wizard_vicuna_30b = gr.Button("wizard-vicuna-30b", elem_id="wizard-vicuna-30b", elem_classes=["square"])
gr.Markdown("Wizard Vicuna (Uncensored)", elem_classes=["center"])
with gr.Column(min_width=20):
vicuna_33b_1_3 = gr.Button("vicuna-33b-1-3", elem_id="vicuna-33b-1-3", elem_classes=["square"])
gr.Markdown("Vicuna 1.3", elem_classes=["center"])
with gr.Column(min_width=20):
mpt_30b = gr.Button("mpt-30b", elem_id="mpt-30b", elem_classes=["square"])
gr.Markdown("MPT", elem_classes=["center"])
with gr.Column(min_width=20):
upstage_llama_30b = gr.Button("upstage-llama-30b", elem_id="upstage-llama-30b", elem_classes=["square"])
gr.Markdown("Upstage LLaMA", elem_classes=["center"])
with gr.Column(min_width=20):
codellama_34b = gr.Button("codellama-34b", elem_id="codellama-34b", elem_classes=["square"])
gr.Markdown("Code LLaMA", elem_classes=["center"])
gr.Markdown("## ~ 70B Parameters")
with gr.Row(elem_classes=["sub-container"]):
with gr.Column(min_width=20):
stable_beluga2_70b = gr.Button("stable-beluga2-70b", elem_id="stable-beluga2-70b", elem_classes=["square"])
gr.Markdown("Stable Beluga 2", elem_classes=["center"])
with gr.Column(min_width=20):
upstage_llama2_70b = gr.Button("upstage-llama2-70b", elem_id="upstage-llama2-70b", elem_classes=["square"])
gr.Markdown("Upstage2", elem_classes=["center"])
with gr.Column(min_width=20):
upstage_llama2_70b_2 = gr.Button("upstage-llama2-70b-2", elem_id="upstage-llama2-70b-2", elem_classes=["square"])
gr.Markdown("Upstage2 v2", elem_classes=["center"])
with gr.Column(min_width=20):
platypus2_70b = gr.Button("platypus2-70b", elem_id="platypus2-70b", elem_classes=["square"])
gr.Markdown("Platypus2", elem_classes=["center"])
with gr.Column(min_width=20):
wizardlm_70b = gr.Button("wizardlm-70b", elem_id="wizardlm-70b", elem_classes=["square"])
gr.Markdown("WizardLM", elem_classes=["center"])
with gr.Column(min_width=20):
orcamini_70b = gr.Button("orcamini-70b", elem_id="orcamini-70b", elem_classes=["square"])
gr.Markdown("Orca Mini", elem_classes=["center"])
with gr.Column(min_width=20):
samantha_70b = gr.Button("samantha-70b", elem_id="samantha-70b", elem_classes=["square"])
gr.Markdown("Samantha", elem_classes=["center"])
with gr.Column(min_width=20):
godzilla_70b = gr.Button("godzilla-70b", elem_id="godzilla-70b", elem_classes=["square"])
gr.Markdown("GadziLLa", elem_classes=["center"])
with gr.Column(min_width=20):
nous_hermes_70b = gr.Button("nous-hermes-70b", elem_id="nous-hermes-70b", elem_classes=["square"])
gr.Markdown("Nous Hermes 2", elem_classes=["center"])
progress_view = gr.Textbox(label="Progress", elem_classes=["progress-view"])
with gr.Column(visible=False) as byom_input_view: