Sharathhebbar24
commited on
Commit
•
93eaf4a
1
Parent(s):
b52c479
Update models.py
Browse files
models.py
CHANGED
@@ -6,42 +6,51 @@ def return_models(task_name):
|
|
6 |
if task_name == "text2text-generation":
|
7 |
models_parent = (
|
8 |
'google',
|
9 |
-
"
|
|
|
|
|
10 |
"bloom",
|
|
|
|
|
11 |
"lmsys",
|
12 |
-
"
|
|
|
|
|
|
|
13 |
)
|
14 |
else:
|
15 |
models_parent = (
|
16 |
'google',
|
17 |
-
"
|
18 |
-
"
|
19 |
-
"
|
20 |
-
"open_assistant",
|
21 |
"databricks",
|
22 |
-
"nomic_ai",
|
23 |
-
"blinkdl",
|
24 |
"aisquared",
|
25 |
"bloom",
|
|
|
26 |
"tiiuae",
|
27 |
"openlm",
|
28 |
"stabilityai",
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
"lmsys",
|
30 |
"together_computer",
|
31 |
"mosaic_ml",
|
32 |
"h20ai",
|
33 |
-
"facebook"
|
34 |
-
|
35 |
)
|
36 |
return models_parent
|
37 |
|
38 |
def return_text2text_generation_models(model_parent, count=False):
|
39 |
google_models_list = (
|
|
|
|
|
|
|
40 |
'google/flan-t5-xl',
|
41 |
'google/flan-t5-xxl',
|
42 |
-
'google/flan-t5-large',
|
43 |
-
"google/flan-t5-small",
|
44 |
-
"google/flan-t5-base",
|
45 |
"google/byt5-xxl",
|
46 |
"google/byt5-xl",
|
47 |
"google/byt5-large",
|
@@ -298,12 +307,129 @@ def return_text2text_generation_models(model_parent, count=False):
|
|
298 |
'facebook/m2m100_1.2B'
|
299 |
)
|
300 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
301 |
model_dict = {
|
302 |
"google": google_models_list,
|
303 |
"mbzuai": mbzuai_models_list,
|
304 |
"bloom": bloom_models_list,
|
305 |
"lmsys": lmsys_models_list,
|
306 |
-
"facebook": facebook_models_list
|
|
|
|
|
|
|
|
|
|
|
|
|
307 |
}
|
308 |
if count is True:
|
309 |
models_count = 0
|
@@ -658,6 +784,72 @@ def return_text_generation_models(model_parent, count=False):
|
|
658 |
'facebook/xglm-2.9B',
|
659 |
'facebook/xglm-1.7B'
|
660 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
661 |
|
662 |
model_dict = {
|
663 |
"google": google_models_list,
|
@@ -677,7 +869,10 @@ def return_text_generation_models(model_parent, count=False):
|
|
677 |
"together_computer": togethercomputer_models_list,
|
678 |
"mosaic_ml": mosaic_models_list,
|
679 |
"h20ai": h20ai_models_list,
|
680 |
-
"facebook": facebook_models_list
|
|
|
|
|
|
|
681 |
}
|
682 |
|
683 |
if count is True:
|
|
|
6 |
if task_name == "text2text-generation":
|
7 |
models_parent = (
|
8 |
'google',
|
9 |
+
"facebook",
|
10 |
+
"microsoft",
|
11 |
+
"langboat",
|
12 |
"bloom",
|
13 |
+
"allenai",
|
14 |
+
"mbzuai",
|
15 |
"lmsys",
|
16 |
+
"starmpcc",
|
17 |
+
"haining",
|
18 |
+
"kaludi"
|
19 |
+
|
20 |
)
|
21 |
else:
|
22 |
models_parent = (
|
23 |
'google',
|
24 |
+
"facebook",
|
25 |
+
"microsoft",
|
26 |
+
"langboat",
|
|
|
27 |
"databricks",
|
|
|
|
|
28 |
"aisquared",
|
29 |
"bloom",
|
30 |
+
"allenai",
|
31 |
"tiiuae",
|
32 |
"openlm",
|
33 |
"stabilityai",
|
34 |
+
"eleutherai",
|
35 |
+
"mbzuai",
|
36 |
+
"cerebras",
|
37 |
+
"open_assistant",
|
38 |
+
"nomic_ai",
|
39 |
+
"blinkdl",
|
40 |
"lmsys",
|
41 |
"together_computer",
|
42 |
"mosaic_ml",
|
43 |
"h20ai",
|
|
|
|
|
44 |
)
|
45 |
return models_parent
|
46 |
|
47 |
def return_text2text_generation_models(model_parent, count=False):
|
48 |
google_models_list = (
|
49 |
+
"google/flan-t5-small",
|
50 |
+
"google/flan-t5-base",
|
51 |
+
'google/flan-t5-large',
|
52 |
'google/flan-t5-xl',
|
53 |
'google/flan-t5-xxl',
|
|
|
|
|
|
|
54 |
"google/byt5-xxl",
|
55 |
"google/byt5-xl",
|
56 |
"google/byt5-large",
|
|
|
307 |
'facebook/m2m100_1.2B'
|
308 |
)
|
309 |
|
310 |
+
microsoft_models_list = (
|
311 |
+
"microsoft/prophetnet-large-uncased",
|
312 |
+
"microsoft/codereviewer",
|
313 |
+
"microsoft/xprophetnet-large-wiki100-cased",
|
314 |
+
"microsoft/xprophetnet-large-wiki100-cased-xglue-qg",
|
315 |
+
"microsoft/xprophetnet-large-wiki100-cased-xglue-ntg",
|
316 |
+
"microsoft/prophetnet-large-uncased-squad-qg",
|
317 |
+
"microsoft/prophetnet-large-uncased-cnndm",
|
318 |
+
)
|
319 |
+
starmpcc_models_list = (
|
320 |
+
"starmpcc/Asclepius-13B",
|
321 |
+
"starmpcc/Asclepius-7B"
|
322 |
+
)
|
323 |
+
|
324 |
+
langboat_models_list = (
|
325 |
+
"Langboat/mengzi-t5-base",
|
326 |
+
"Langboat/mengzi-t5-base-mt"
|
327 |
+
)
|
328 |
+
|
329 |
+
haining_models_list = (
|
330 |
+
"haining/scientific_abstract_simplification",
|
331 |
+
"haining/poem_interpretation_allpoetry169k_baseline",
|
332 |
+
"haining/poem_interpretation_allpoetry169k_full",
|
333 |
+
"haining/lyrics_interpretation_nonnegative",
|
334 |
+
"haining/poem_interpretation_allpoetry169k",
|
335 |
+
"haining/sas_baseline",
|
336 |
+
)
|
337 |
+
|
338 |
+
kaludi_models_list = (
|
339 |
+
"Kaludi/chatgpt-gpt4-prompts-bart-large-cnn-samsum",
|
340 |
+
"Kaludi/Customer-Support-Assistant",
|
341 |
+
"Kaludi/Customer-Support-Assistant-V2",
|
342 |
+
"Kaludi/historical-events-reimagined"
|
343 |
+
)
|
344 |
+
|
345 |
+
allenai_models_list = (
|
346 |
+
'allenai/macaw-large',
|
347 |
+
'allenai/entailer-11b',
|
348 |
+
'allenai/entailer-large',
|
349 |
+
'allenai/System4_classify_FigLang2022',
|
350 |
+
'allenai/System4_explain_FigLang2022',
|
351 |
+
'allenai/System3_DREAM_FLUTE_all_dimensions_FigLang2022',
|
352 |
+
'allenai/System3_DREAM_FLUTE_social_norm_FigLang2022',
|
353 |
+
'allenai/System3_DREAM_FLUTE_consequence_FigLang2022',
|
354 |
+
'allenai/System3_DREAM_FLUTE_motivation_FigLang2022',
|
355 |
+
'allenai/System3_DREAM_FLUTE_emotion_FigLang2022',
|
356 |
+
'allenai/System2_FigLang2022',
|
357 |
+
'allenai/System1_FigLang2022',
|
358 |
+
'allenai/DREAM',
|
359 |
+
'allenai/led-base-16384-cochrane',
|
360 |
+
'allenai/bart-large-multi_lexsum-long-multitask',
|
361 |
+
'allenai/bart-large-multi_lexsum-source-multitask',
|
362 |
+
'allenai/pegasus-multi_lexsum-short-tiny',
|
363 |
+
'allenai/pegasus-multi_lexsum-long-tiny',
|
364 |
+
'allenai/pegasus-multi_lexsum-long-short',
|
365 |
+
'allenai/bart-large-multi_lexsum-short-tiny',
|
366 |
+
'allenai/bart-large-multi_lexsum-long-tiny',
|
367 |
+
'allenai/bart-large-multi_lexsum-long-short',
|
368 |
+
'allenai/primera-multi_lexsum-source-tiny',
|
369 |
+
'allenai/primera-multi_lexsum-source-short',
|
370 |
+
'allenai/primera-multi_lexsum-source-long',
|
371 |
+
'allenai/led-base-16384-multi_lexsum-source-tiny',
|
372 |
+
'allenai/led-base-16384-multi_lexsum-source-short',
|
373 |
+
'allenai/led-base-16384-multi_lexsum-source-long',
|
374 |
+
'allenai/led-base-16384-ms2',
|
375 |
+
'allenai/mtk-instruct-11b-def-pos',
|
376 |
+
'allenai/tk-instruct-small-def-pos',
|
377 |
+
'allenai/tk-instruct-base-def-pos',
|
378 |
+
'allenai/tk-instruct-large-def-pos',
|
379 |
+
'allenai/mtk-instruct-3b-def-pos',
|
380 |
+
'allenai/tk-instruct-3b-def-pos-neg-expl',
|
381 |
+
'allenai/tk-instruct-3b-def-pos-neg',
|
382 |
+
'allenai/tk-instruct-3b-pos',
|
383 |
+
'allenai/tk-instruct-3b-def-pos',
|
384 |
+
'allenai/tk-instruct-3b-def',
|
385 |
+
'allenai/tk-instruct-11b-def-pos-neg-expl',
|
386 |
+
'allenai/tk-instruct-11b-def-pos',
|
387 |
+
'allenai/tk-instruct-11b-def',
|
388 |
+
'allenai/PRIMERA-arxiv',
|
389 |
+
'allenai/PRIMERA-wcep',
|
390 |
+
'allenai/PRIMERA-multixscience',
|
391 |
+
'allenai/PRIMERA-multinews',
|
392 |
+
'allenai/PRIMERA',
|
393 |
+
'allenai/unifiedqa-v2-t5-small-1363200',
|
394 |
+
'allenai/unifiedqa-v2-t5-small-1251000',
|
395 |
+
'allenai/unifiedqa-v2-t5-large-1363200',
|
396 |
+
'allenai/unifiedqa-v2-t5-large-1251000',
|
397 |
+
'allenai/unifiedqa-v2-t5-base-1363200',
|
398 |
+
'allenai/unifiedqa-v2-t5-base-1251000',
|
399 |
+
'allenai/unifiedqa-v2-t5-3b-1363200',
|
400 |
+
'allenai/unifiedqa-v2-t5-3b-1251000',
|
401 |
+
'allenai/unifiedqa-v2-t5-11b-1363200',
|
402 |
+
'allenai/unifiedqa-v2-t5-11b-1251000',
|
403 |
+
'allenai/unifiedqa-t5-small',
|
404 |
+
'allenai/unifiedqa-t5-large',
|
405 |
+
'allenai/unifiedqa-t5-base',
|
406 |
+
'allenai/unifiedqa-t5-3b',
|
407 |
+
'allenai/unifiedqa-t5-11b',
|
408 |
+
'allenai/tailor',
|
409 |
+
'allenai/t5-small-squad2-question-generation',
|
410 |
+
'allenai/t5-small-squad2-next-word-generator-squad',
|
411 |
+
'allenai/t5-small-squad11',
|
412 |
+
'allenai/t5-small-next-word-generator-qoogle',
|
413 |
+
'allenai/macaw-answer-11b',
|
414 |
+
'allenai/macaw-3b',
|
415 |
+
'allenai/macaw-11b',
|
416 |
+
'allenai/led-large-16384',
|
417 |
+
'allenai/led-large-16384-arxiv',
|
418 |
+
'allenai/led-base-16384'
|
419 |
+
)
|
420 |
+
|
421 |
model_dict = {
|
422 |
"google": google_models_list,
|
423 |
"mbzuai": mbzuai_models_list,
|
424 |
"bloom": bloom_models_list,
|
425 |
"lmsys": lmsys_models_list,
|
426 |
+
"facebook": facebook_models_list,
|
427 |
+
"microsoft": microsoft_models_list,
|
428 |
+
"starmpcc": starmpcc_models_list,
|
429 |
+
"langboat": langboat_models_list,
|
430 |
+
"haining": haining_models_list,
|
431 |
+
"kaludi": kaludi_models_list,
|
432 |
+
"allenai": allenai_models_list
|
433 |
}
|
434 |
if count is True:
|
435 |
models_count = 0
|
|
|
784 |
'facebook/xglm-2.9B',
|
785 |
'facebook/xglm-1.7B'
|
786 |
)
|
787 |
+
|
788 |
+
microsoft_models_list = (
|
789 |
+
"microsoft/dolly-v2-7b-olive-optimized",
|
790 |
+
"microsoft/lts-gpt2-sm",
|
791 |
+
"microsoft/BioGPT-Large",
|
792 |
+
"microsoft/BioGPT-Large-PubMedQA",
|
793 |
+
"microsoft/biogpt",
|
794 |
+
"microsoft/git-large-vatex",
|
795 |
+
"microsoft/git-base-msrvtt-qa",
|
796 |
+
"microsoft/git-base-vatex",
|
797 |
+
"microsoft/Promptist",
|
798 |
+
"microsoft/CodeGPT-small-py",
|
799 |
+
"microsoft/CodeGPT-small-py-adaptedGPT2",
|
800 |
+
"microsoft/CodeGPT-small-java",
|
801 |
+
"microsoft/CodeGPT-small-java-adaptedGPT2",
|
802 |
+
)
|
803 |
+
|
804 |
+
langboat_models_list = (
|
805 |
+
"Langboat/bloom-1b4-zh",
|
806 |
+
"Langboat/ReGPT-125M-200G",
|
807 |
+
"Langboat/bloom-6b4-zh",
|
808 |
+
"Langboat/bloom-2b5-zh",
|
809 |
+
"Langboat/bloom-800m-zh",
|
810 |
+
"Langboat/bloom-389m-zh",
|
811 |
+
"Langboat/mengzi-gpt-neo-base",
|
812 |
+
)
|
813 |
+
|
814 |
+
allenai_models_list = (
|
815 |
+
'allenai/bhaskara',
|
816 |
+
'allenai/open-instruct-pythia-6.9b-tulu',
|
817 |
+
'allenai/open-instruct-opt-6.7b-tulu',
|
818 |
+
'allenai/open-instruct-sni-13b',
|
819 |
+
'allenai/open-instruct-flan-v2-13b',
|
820 |
+
'allenai/open-instruct-dolly-13b',
|
821 |
+
'allenai/open-instruct-self-instruct-13b',
|
822 |
+
'allenai/open-instruct-self-instruct-7b',
|
823 |
+
'allenai/open-instruct-oasst1-13b',
|
824 |
+
'allenai/open-instruct-oasst1-7b',
|
825 |
+
'allenai/open-instruct-sni-7b',
|
826 |
+
'allenai/open-instruct-baize-13b',
|
827 |
+
'allenai/open-instruct-baize-7b',
|
828 |
+
'allenai/open-instruct-cot-13b',
|
829 |
+
'allenai/open-instruct-cot-7b',
|
830 |
+
'allenai/open-instruct-gpt4-alpaca-13b',
|
831 |
+
'allenai/open-instruct-gpt4-alpaca-7b',
|
832 |
+
'allenai/open-instruct-human-mix-7b',
|
833 |
+
'allenai/open-instruct-human-mix-13b',
|
834 |
+
'allenai/open-instruct-sharegpt-7b',
|
835 |
+
'allenai/open-instruct-sharegpt-13b',
|
836 |
+
'allenai/open-instruct-sharegpt-30b',
|
837 |
+
'allenai/open-instruct-human-mix-30b',
|
838 |
+
'allenai/tulu-7b',
|
839 |
+
'allenai/tulu-13b',
|
840 |
+
'allenai/tulu-30b',
|
841 |
+
'allenai/open-instruct-code-alpaca-13b',
|
842 |
+
'allenai/open-instruct-code-alpaca-7b',
|
843 |
+
'allenai/open-instruct-unnatural-instructions-13b',
|
844 |
+
'allenai/open-instruct-unnatural-instructions-7b',
|
845 |
+
'allenai/open-instruct-stanford-alpaca-13b',
|
846 |
+
'allenai/open-instruct-stanford-alpaca-7b',
|
847 |
+
'allenai/open-instruct-flan-v2-7b',
|
848 |
+
'allenai/open-instruct-dolly-7b',
|
849 |
+
'allenai/tulu-65b',
|
850 |
+
'allenai/open-instruct-sharegpt-65b',
|
851 |
+
'allenai/open-instruct-human-mix-65b'
|
852 |
+
)
|
853 |
|
854 |
model_dict = {
|
855 |
"google": google_models_list,
|
|
|
869 |
"together_computer": togethercomputer_models_list,
|
870 |
"mosaic_ml": mosaic_models_list,
|
871 |
"h20ai": h20ai_models_list,
|
872 |
+
"facebook": facebook_models_list,
|
873 |
+
"microsoft": microsoft_models_list,
|
874 |
+
"langboat": langboat_models_list,
|
875 |
+
"allenai": allenai_models_list
|
876 |
}
|
877 |
|
878 |
if count is True:
|