Commit
·
192a1a0
1
Parent(s):
8a20dfb
Create models.py
Browse files
models.py
ADDED
|
@@ -0,0 +1,690 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
def return_task_name():
|
| 2 |
+
return ('text2text-generation', "text-generation")
|
| 3 |
+
|
| 4 |
+
|
| 5 |
+
def return_models(task_name):
|
| 6 |
+
if task_name == "text2text-generation":
|
| 7 |
+
models_parent = (
|
| 8 |
+
'google',
|
| 9 |
+
"mbzuai",
|
| 10 |
+
"bloom",
|
| 11 |
+
"lmsys",
|
| 12 |
+
"facebook"
|
| 13 |
+
)
|
| 14 |
+
else:
|
| 15 |
+
models_parent = (
|
| 16 |
+
'google',
|
| 17 |
+
"mbzuai",
|
| 18 |
+
"eleutherai",
|
| 19 |
+
"cerebras",
|
| 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",
|
| 48 |
+
"google/byt5-small",
|
| 49 |
+
"google/byt5-base",
|
| 50 |
+
"google/mt5-xxl",
|
| 51 |
+
"google/mt5-xl",
|
| 52 |
+
"google/mt5-large",
|
| 53 |
+
"google/mt5-small",
|
| 54 |
+
"google/long-t5-local-base",
|
| 55 |
+
"google/long-t5-local-large",
|
| 56 |
+
"google/long-t5-tglobal-base",
|
| 57 |
+
"google/long-t5-tglobal-large",
|
| 58 |
+
"google/pegasus-big_patent",
|
| 59 |
+
"google/pegasus-x-base",
|
| 60 |
+
"google/pegasus-x-large",
|
| 61 |
+
"google/pegasus-x-base-arxiv",
|
| 62 |
+
"google/roberta2roberta_L-24_wikisplit",
|
| 63 |
+
"google/roberta2roberta_L-24_discofuse",
|
| 64 |
+
"google/switch-base-8",
|
| 65 |
+
"google/switch-base-16",
|
| 66 |
+
"google/switch-base-32",
|
| 67 |
+
"google/switch-base-64",
|
| 68 |
+
"google/switch-base-128",
|
| 69 |
+
"google/switch-base-256",
|
| 70 |
+
"google/switch-large-128",
|
| 71 |
+
"google/switch-xxl-128",
|
| 72 |
+
"google/switch-c-2048",
|
| 73 |
+
"google/t5-11b-ssm",
|
| 74 |
+
"google/t5-11b-ssm-nq",
|
| 75 |
+
"google/t5-11b-ssm-nqo",
|
| 76 |
+
"google/t5-11b-ssm-tqa",
|
| 77 |
+
"google/t5-11b-ssm-tqao",
|
| 78 |
+
"google/t5-11b-ssm-wq",
|
| 79 |
+
"google/t5-11b-ssm-wqo",
|
| 80 |
+
"google/t5-3b-ssm",
|
| 81 |
+
"google/t5-3b-ssm-nq",
|
| 82 |
+
"google/t5-3b-ssm-nqo",
|
| 83 |
+
"google/t5-base-lm-adapt",
|
| 84 |
+
"google/t5-base-lm-adapt",
|
| 85 |
+
"google/t5-efficient-base",
|
| 86 |
+
"google/t5-efficient-base-dl2",
|
| 87 |
+
"google/t5-efficient-base-dl4",
|
| 88 |
+
"google/t5-efficient-base-dl6",
|
| 89 |
+
"google/t5-efficient-base-dl8",
|
| 90 |
+
"google/t5-efficient-base-dm256",
|
| 91 |
+
"google/t5-efficient-base-dm512",
|
| 92 |
+
"google/t5-efficient-base-dm1000",
|
| 93 |
+
"google/t5-efficient-base-dm2000",
|
| 94 |
+
"google/t5-efficient-base-el2",
|
| 95 |
+
"google/t5-efficient-base-el4",
|
| 96 |
+
"google/t5-efficient-base-el6",
|
| 97 |
+
"google/t5-efficient-base-el8",
|
| 98 |
+
"google/t5-efficient-base-el16",
|
| 99 |
+
"google/t5-efficient-base-nl40",
|
| 100 |
+
"google/t5-efficient-base-nl48",
|
| 101 |
+
"google/t5-efficient-base-nl8",
|
| 102 |
+
"google/t5-efficient-large",
|
| 103 |
+
"google/t5-efficient-large-dl12",
|
| 104 |
+
"google/t5-efficient-large-dl16",
|
| 105 |
+
"google/t5-efficient-large-dl2",
|
| 106 |
+
"google/t5-efficient-large-dl32",
|
| 107 |
+
"google/t5-efficient-large-dl4",
|
| 108 |
+
"google/t5-efficient-large-dl6",
|
| 109 |
+
"google/t5-efficient-large-dl8",
|
| 110 |
+
"google/t5-efficient-large-dm128",
|
| 111 |
+
"google/t5-efficient-large-dm2000",
|
| 112 |
+
"google/t5-efficient-large-dm256",
|
| 113 |
+
"google/t5-efficient-large-dm512",
|
| 114 |
+
"google/t5-efficient-large-dm768",
|
| 115 |
+
"google/t5-efficient-large-el12",
|
| 116 |
+
"google/t5-efficient-large-el2",
|
| 117 |
+
"google/t5-efficient-large-el4",
|
| 118 |
+
"google/t5-efficient-large-el6",
|
| 119 |
+
"google/t5-efficient-large-el8",
|
| 120 |
+
"google/t5-efficient-large-kv128",
|
| 121 |
+
"google/t5-efficient-large-kv16",
|
| 122 |
+
"google/t5-efficient-large-kv256",
|
| 123 |
+
"google/t5-efficient-large-kv32",
|
| 124 |
+
"google/t5-efficient-large-nh12",
|
| 125 |
+
"google/t5-efficient-large-nh2",
|
| 126 |
+
"google/t5-efficient-large-nh24",
|
| 127 |
+
"google/t5-efficient-large-nh32",
|
| 128 |
+
"google/t5-efficient-large-nh4",
|
| 129 |
+
"google/t5-efficient-large-nh8",
|
| 130 |
+
"google/t5-efficient-large-nh8-nl32",
|
| 131 |
+
"google/t5-efficient-large-nl10",
|
| 132 |
+
"google/t5-efficient-large-nl12",
|
| 133 |
+
"google/t5-efficient-large-nl16",
|
| 134 |
+
"google/t5-efficient-large-nl2",
|
| 135 |
+
"google/t5-efficient-large-nl20",
|
| 136 |
+
"google/t5-efficient-large-nl32",
|
| 137 |
+
"google/t5-efficient-large-nl36",
|
| 138 |
+
"google/t5-efficient-large-nl4",
|
| 139 |
+
"google/t5-efficient-large-nl8",
|
| 140 |
+
"google/t5-efficient-mini",
|
| 141 |
+
"google/t5-efficient-mini-nl12",
|
| 142 |
+
"google/t5-efficient-mini-nl24",
|
| 143 |
+
"google/t5-efficient-mini-nl6",
|
| 144 |
+
"google/t5-efficient-mini-nl8",
|
| 145 |
+
"google/t5-efficient-small",
|
| 146 |
+
"google/t5-efficient-small-dl12",
|
| 147 |
+
"google/t5-efficient-small-dl16",
|
| 148 |
+
"google/t5-efficient-small-dl2",
|
| 149 |
+
"google/t5-efficient-small-dl4",
|
| 150 |
+
"google/t5-efficient-small-dl8",
|
| 151 |
+
"google/t5-efficient-small-dm1000",
|
| 152 |
+
"google/t5-efficient-small-dm128",
|
| 153 |
+
"google/t5-efficient-small-dm2000",
|
| 154 |
+
"google/t5-efficient-small-dm256",
|
| 155 |
+
"google/t5-efficient-small-dm768",
|
| 156 |
+
"google/t5-efficient-small-el12",
|
| 157 |
+
"google/t5-efficient-small-el16",
|
| 158 |
+
"google/t5-efficient-small-el16-dl1",
|
| 159 |
+
"google/t5-efficient-small-el16-dl2",
|
| 160 |
+
"google/t5-efficient-small-el16-dl4",
|
| 161 |
+
"google/t5-efficient-small-el16-dl8",
|
| 162 |
+
"google/t5-efficient-small-el2",
|
| 163 |
+
"google/t5-efficient-small-el32",
|
| 164 |
+
"google/t5-efficient-small-el4",
|
| 165 |
+
"google/t5-efficient-small-el48",
|
| 166 |
+
"google/t5-efficient-small-el64",
|
| 167 |
+
"google/t5-efficient-small-el8",
|
| 168 |
+
"google/t5-efficient-small-el8-dl1",
|
| 169 |
+
"google/t5-efficient-small-el8-dl2",
|
| 170 |
+
"google/t5-efficient-small-el8-dl4",
|
| 171 |
+
"google/t5-efficient-small-ff1000",
|
| 172 |
+
"google/t5-efficient-small-ff12000",
|
| 173 |
+
"google/t5-efficient-small-ff3000",
|
| 174 |
+
"google/t5-efficient-small-ff6000",
|
| 175 |
+
"google/t5-efficient-small-ff9000",
|
| 176 |
+
"google/t5-efficient-small-kv128",
|
| 177 |
+
"google/t5-efficient-small-kv16",
|
| 178 |
+
"google/t5-efficient-small-kv256",
|
| 179 |
+
"google/t5-efficient-small-kv32",
|
| 180 |
+
"google/t5-efficient-small-nl16",
|
| 181 |
+
"google/t5-efficient-small-nl2",
|
| 182 |
+
"google/t5-efficient-small-nl20",
|
| 183 |
+
"google/t5-efficient-small-nl22",
|
| 184 |
+
"google/t5-efficient-small-nl24",
|
| 185 |
+
"google/t5-efficient-small-nl32",
|
| 186 |
+
"google/t5-efficient-small-nl36",
|
| 187 |
+
"google/t5-efficient-small-nl4",
|
| 188 |
+
"google/t5-efficient-small-nl40",
|
| 189 |
+
"google/t5-efficient-small-nl48",
|
| 190 |
+
"google/t5-efficient-small-nl8",
|
| 191 |
+
"google/t5-efficient-tiny",
|
| 192 |
+
"google/t5-efficient-tiny-dl2",
|
| 193 |
+
"google/t5-efficient-tiny-dl6",
|
| 194 |
+
"google/t5-efficient-tiny-dl8",
|
| 195 |
+
"google/t5-efficient-tiny-el12",
|
| 196 |
+
"google/t5-efficient-tiny-el2",
|
| 197 |
+
"google/t5-efficient-tiny-el6",
|
| 198 |
+
"google/t5-efficient-tiny-el8",
|
| 199 |
+
"google/t5-efficient-tiny-ff12000",
|
| 200 |
+
"google/t5-efficient-tiny-ff2000",
|
| 201 |
+
"google/t5-efficient-tiny-ff3000",
|
| 202 |
+
"google/t5-efficient-tiny-ff6000",
|
| 203 |
+
"google/t5-efficient-tiny-ff9000",
|
| 204 |
+
"google/t5-efficient-tiny-nh1",
|
| 205 |
+
"google/t5-efficient-tiny-nh16",
|
| 206 |
+
"google/t5-efficient-tiny-nh32",
|
| 207 |
+
"google/t5-efficient-tiny-nh8",
|
| 208 |
+
"google/t5-efficient-tiny-nl12",
|
| 209 |
+
"google/t5-efficient-tiny-nl16",
|
| 210 |
+
"google/t5-efficient-tiny-nl2",
|
| 211 |
+
"google/t5-efficient-tiny-nl24",
|
| 212 |
+
"google/t5-efficient-tiny-nl32",
|
| 213 |
+
"google/t5-efficient-tiny-nl6",
|
| 214 |
+
"google/t5-efficient-tiny-nl8",
|
| 215 |
+
"google/t5-efficient-xl",
|
| 216 |
+
"google/t5-efficient-xl-nl12",
|
| 217 |
+
"google/t5-efficient-xl-nl16",
|
| 218 |
+
"google/t5-efficient-xl-nl2",
|
| 219 |
+
"google/t5-efficient-xl-nl28",
|
| 220 |
+
"google/t5-efficient-xl-nl4",
|
| 221 |
+
"google/t5-efficient-xl-nl6",
|
| 222 |
+
"google/t5-efficient-xl-nl8",
|
| 223 |
+
"google/t5-efficient-xxl",
|
| 224 |
+
"google/t5-efficient-xxl-nl4",
|
| 225 |
+
"google/t5-large-lm-adapt",
|
| 226 |
+
"google/t5-large-ssm",
|
| 227 |
+
"google/t5-large-ssm-nq",
|
| 228 |
+
"google/t5-large-ssm-nqo",
|
| 229 |
+
"google/t5-small-lm-adapt",
|
| 230 |
+
"google/t5-small-ssm",
|
| 231 |
+
"google/t5-small-ssm-nq",
|
| 232 |
+
"google/t5-v1_1-base",
|
| 233 |
+
"google/t5-v1_1-large",
|
| 234 |
+
"google/t5-v1_1-small",
|
| 235 |
+
"google/t5-v1_1-xl",
|
| 236 |
+
"google/t5-v1_1-xxl",
|
| 237 |
+
"google/t5-xl-lm-adapt",
|
| 238 |
+
"google/t5-xl-ssm-nq",
|
| 239 |
+
"google/t5-xxl-lm-adapt",
|
| 240 |
+
"google/t5-xxl-ssm",
|
| 241 |
+
"google/t5-xxl-ssm-nq",
|
| 242 |
+
"google/t5-xxl-ssm-nqo",
|
| 243 |
+
"google/t5-xxl-ssm-tqa",
|
| 244 |
+
"google/t5-xxl-ssm-tqao",
|
| 245 |
+
"google/t5-xxl-ssm-wq",
|
| 246 |
+
"google/t5-xxl-ssm-wqo",
|
| 247 |
+
"google/t5_11b_trueteacher_and_anli",
|
| 248 |
+
"google/ul2",
|
| 249 |
+
"google/umt5-base",
|
| 250 |
+
"google/umt5-small",
|
| 251 |
+
"google/umt5-xl",
|
| 252 |
+
"google/umt5-xxl",
|
| 253 |
+
)
|
| 254 |
+
|
| 255 |
+
mbzuai_models_list = (
|
| 256 |
+
"MBZUAI/LaMini-Flan-T5-783M",
|
| 257 |
+
"MBZUAI/LaMini-Flan-T5-248M",
|
| 258 |
+
"MBZUAI/LaMini-Flan-T5-77M",
|
| 259 |
+
"MBZUAI/LaMini-T5-738M",
|
| 260 |
+
"MBZUAI/LaMini-T5-223M",
|
| 261 |
+
"MBZUAI/LaMini-T5-61M",
|
| 262 |
+
)
|
| 263 |
+
bloom_models_list = (
|
| 264 |
+
"bigscience/T0_3B",
|
| 265 |
+
"bigscience/T0_original_task_only",
|
| 266 |
+
"bigscience/T0_single_prompt",
|
| 267 |
+
"bigscience/T0p",
|
| 268 |
+
"bigscience/T0",
|
| 269 |
+
"bigscience/T0pp",
|
| 270 |
+
"bigscience/mt0-xxl-p3",
|
| 271 |
+
"bigscience/mt0-xxl",
|
| 272 |
+
"bigscience/mt0-large",
|
| 273 |
+
"bigscience/mt0-base",
|
| 274 |
+
"bigscience/mt0-small",
|
| 275 |
+
"bigscience/mt0-xxl-mt",
|
| 276 |
+
)
|
| 277 |
+
|
| 278 |
+
lmsys_models_list = (
|
| 279 |
+
"lmsys/fastchat-t5-3b-v1.0",
|
| 280 |
+
)
|
| 281 |
+
|
| 282 |
+
facebook_models_list = (
|
| 283 |
+
'facebook/mbart-large-50-many-to-many-mmt',
|
| 284 |
+
'facebook/musicgen-small',
|
| 285 |
+
'facebook/musicgen-medium',
|
| 286 |
+
"facebook/musicgen-large",
|
| 287 |
+
'facebook/m2m100_418M',
|
| 288 |
+
'facebook/mbart-large-50-one-to-many-mmt',
|
| 289 |
+
'facebook/mbart-large-50-many-to-one-mmt',
|
| 290 |
+
'facebook/mbart-large-50',
|
| 291 |
+
'facebook/mgenre-wiki',
|
| 292 |
+
'facebook/genre-linking-aidayago2',
|
| 293 |
+
'facebook/genre-linking-blink',
|
| 294 |
+
'facebook/genre-kilt',
|
| 295 |
+
'facebook/m2m100-12B-avg-10-ckpt',
|
| 296 |
+
'facebook/m2m100-12B-avg-5-ckpt',
|
| 297 |
+
'facebook/m2m100-12B-last-ckpt',
|
| 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
|
| 310 |
+
for i in model_dict:
|
| 311 |
+
models_count += len(model_dict[i])
|
| 312 |
+
return models_count
|
| 313 |
+
|
| 314 |
+
return model_dict[model_parent]
|
| 315 |
+
|
| 316 |
+
|
| 317 |
+
# Text Generation Models
|
| 318 |
+
def return_text_generation_models(model_parent, count=False):
|
| 319 |
+
google_models_list = (
|
| 320 |
+
"google/reformer-enwik8",
|
| 321 |
+
"google/reformer-crime-and-punishment",
|
| 322 |
+
)
|
| 323 |
+
mbzuai_models_list = (
|
| 324 |
+
"MBZUAI/LaMini-Cerebras-111M",
|
| 325 |
+
"MBZUAI/LaMini-Cerebras-256M",
|
| 326 |
+
"MBZUAI/LaMini-Cerebras-590M",
|
| 327 |
+
"MBZUAI/LaMini-Cerebras-1.3B",
|
| 328 |
+
"MBZUAI/LaMini-GPT-774M",
|
| 329 |
+
"MBZUAI/LaMini-GPT-124M",
|
| 330 |
+
"MBZUAI/LaMini-GPT-1.5B",
|
| 331 |
+
"MBZUAI/LaMini-Neo-125M",
|
| 332 |
+
"MBZUAI/LaMini-Neo-1.3B",
|
| 333 |
+
)
|
| 334 |
+
|
| 335 |
+
eleutherai_models_list=(
|
| 336 |
+
"EleutherAI/pythia-14m",
|
| 337 |
+
"EleutherAI/pythia-31m",
|
| 338 |
+
"EleutherAI/pythia-1b-deduped",
|
| 339 |
+
"EleutherAI/pythia-2.8b-v0",
|
| 340 |
+
"EleutherAI/pythia-1b-v0",
|
| 341 |
+
"EleutherAI/pythia-410m-v0",
|
| 342 |
+
"EleutherAI/pythia-70m-deduped-v0",
|
| 343 |
+
"EleutherAI/pythia-2.8b-deduped-v0",
|
| 344 |
+
"EleutherAI/pythia-1b-deduped-v0",
|
| 345 |
+
"EleutherAI/pythia-410m-deduped-v0",
|
| 346 |
+
"EleutherAI/pythia-160m-deduped-v0",
|
| 347 |
+
"EleutherAI/pythia-6.9b-deduped-v0",
|
| 348 |
+
"EleutherAI/pythia-70m-deduped",
|
| 349 |
+
"EleutherAI/pythia-70m",
|
| 350 |
+
"EleutherAI/pythia-2.8b-deduped",
|
| 351 |
+
"EleutherAI/pythia-1b",
|
| 352 |
+
"EleutherAI/pythia-410m-deduped",
|
| 353 |
+
"EleutherAI/pythia-160m-deduped",
|
| 354 |
+
"EleutherAI/pythia-160m-v0",
|
| 355 |
+
"EleutherAI/pythia-1.4b-deduped-v0",
|
| 356 |
+
"EleutherAI/pythia-1.4b",
|
| 357 |
+
"EleutherAI/pythia-410m",
|
| 358 |
+
"EleutherAI/pythia-intervention-410m-deduped",
|
| 359 |
+
"EleutherAI/gpt-neo-125m",
|
| 360 |
+
"EleutherAI/gpt-neo-2.7B",
|
| 361 |
+
"EleutherAI/gpt-neo-1.3B",
|
| 362 |
+
"EleutherAI/pythia-160m",
|
| 363 |
+
"EleutherAI/gpt-neox-20b",
|
| 364 |
+
"EleutherAI/gpt-j-6b",
|
| 365 |
+
"EleutherAI/pythia-2.8b",
|
| 366 |
+
"EleutherAI/pythia-12b-deduped",
|
| 367 |
+
"EleutherAI/pythia-6.9b-deduped",
|
| 368 |
+
"EleutherAI/pythia-1.4b-deduped",
|
| 369 |
+
"EleutherAI/pythia-12b",
|
| 370 |
+
"EleutherAI/pythia-6.9b",
|
| 371 |
+
"EleutherAI/polyglot-ko-12.8b",
|
| 372 |
+
"EleutherAI/polyglot-ko-5.8b",
|
| 373 |
+
"EleutherAI/polyglot-ko-3.8b",
|
| 374 |
+
"EleutherAI/polyglot-ko-1.3b",
|
| 375 |
+
"EleutherAI/pythia-intervention-6.9b-deduped",
|
| 376 |
+
"EleutherAI/pythia-intervention-1.4b-deduped",
|
| 377 |
+
"EleutherAI/pythia-intervention-70m-deduped",
|
| 378 |
+
"EleutherAI/pythia-intervention-long-1.4b-deduped",
|
| 379 |
+
"EleutherAI/pythia-70m-v0",
|
| 380 |
+
"EleutherAI/pythia-1.4b-v0",
|
| 381 |
+
"EleutherAI/pythia-6.9b-v0",
|
| 382 |
+
"EleutherAI/pythia-12b-deduped-v0",
|
| 383 |
+
"EleutherAI/pythia-12b-v0",
|
| 384 |
+
"EleutherAI/pythia-160m-seed3",
|
| 385 |
+
"EleutherAI/pythia-160m-seed2",
|
| 386 |
+
"EleutherAI/pythia-160m-seed1",
|
| 387 |
+
"EleutherAI/neox-ckpt-pythia-6.9b-deduped",
|
| 388 |
+
"EleutherAI/pythia-160m-hiddendropout",
|
| 389 |
+
"EleutherAI/pythia-160m-attndropout",
|
| 390 |
+
"EleutherAI/pythia-160m-alldropout",
|
| 391 |
+
"EleutherAI/pythia-6.9b-deduped-v0-seed42",
|
| 392 |
+
)
|
| 393 |
+
cerebras_models_list = (
|
| 394 |
+
"cerebras/btlm-3b-8k-base",
|
| 395 |
+
"cerebras/cerebras-GPT-13B",
|
| 396 |
+
"cerebras/cerebras-GPT-6.7B",
|
| 397 |
+
"cerebras/cerebras-GPT-2.7B",
|
| 398 |
+
"cerebras/cerebras-GPT-1.3B",
|
| 399 |
+
"cerebras/cerebras-GPT-590M",
|
| 400 |
+
"cerebras/cerebras-GPT-256M",
|
| 401 |
+
"cerebras/cerebras-GPT-111M",
|
| 402 |
+
)
|
| 403 |
+
open_assistant_models_list = (
|
| 404 |
+
"OpenAssistant/codellama-13b-oasst-sft-v10",
|
| 405 |
+
"OpenAssistant/llama2-70b-oasst-sft-v10",
|
| 406 |
+
"OpenAssistant/llama2-13b-megacode2-oasst",
|
| 407 |
+
"OpenAssistant/falcon-40b-megacode2-oasst",
|
| 408 |
+
"OpenAssistant/pythia-12b-sft-v8-rlhf-2k-steps",
|
| 409 |
+
"OpenAssistant/llama2-13b-orca-8k-3319",
|
| 410 |
+
"OpenAssistant/falcon-7b-sft-mix-2000",
|
| 411 |
+
"OpenAssistant/falcon-7b-sft-top1-696",
|
| 412 |
+
"OpenAssistant/falcon-40b-sft-mix-1226",
|
| 413 |
+
"OpenAssistant/falcon-40b-sft-top1-560",
|
| 414 |
+
"OpenAssistant/pythia-12b-sft-v8-2.5k-steps",
|
| 415 |
+
"OpenAssistant/pythia-12b-sft-v8-7k-steps",
|
| 416 |
+
"OpenAssistant/pythia-12b-pre-v8-12.5k-steps",
|
| 417 |
+
"OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5",
|
| 418 |
+
"OpenAssistant/stablelm-7b-sft-v7-epoch-3",
|
| 419 |
+
"OpenAssistant/oasst-sft-1-pythia-12b",
|
| 420 |
+
"OpenAssistant/galactica-6.7b-finetuned",
|
| 421 |
+
)
|
| 422 |
+
|
| 423 |
+
databricks_models_list = (
|
| 424 |
+
|
| 425 |
+
"databricks/dolly-v2-7b",
|
| 426 |
+
"databricks/dolly-v2-3b",
|
| 427 |
+
"databricks/dolly-v2-12b",
|
| 428 |
+
"databricks/dolly-v1-6b",
|
| 429 |
+
)
|
| 430 |
+
|
| 431 |
+
nomic_ai_models_list = (
|
| 432 |
+
"nomic-ai/gpt4all-falcon",
|
| 433 |
+
"nomic-ai/gpt4all-j",
|
| 434 |
+
"nomic-ai/gpt4all-mpt",
|
| 435 |
+
"nomic-ai/gpt4all-13b-snoozy",
|
| 436 |
+
"nomic-ai/gpt4all-j-lora",
|
| 437 |
+
)
|
| 438 |
+
|
| 439 |
+
blinkdl_models_list = (
|
| 440 |
+
"BlinkDL/rwkv-5-world",
|
| 441 |
+
"BlinkDL/rwkv-4-world",
|
| 442 |
+
"BlinkDL/rwkv-4-raven",
|
| 443 |
+
"BlinkDL/rwkv-4-pile-7b",
|
| 444 |
+
"BlinkDL/rwkv-4-pile-14b",
|
| 445 |
+
"BlinkDL/rwkv-4-novel",
|
| 446 |
+
"BlinkDL/rwkv-4-pileplus",
|
| 447 |
+
"BlinkDL/rwkv-4-pile-430m",
|
| 448 |
+
"BlinkDL/rwkv-4-pile-3b",
|
| 449 |
+
"BlinkDL/rwkv-4-pile-1b5",
|
| 450 |
+
"BlinkDL/rwkv-4-pile-169m",
|
| 451 |
+
"BlinkDL/rwkv-3-pile-1b5",
|
| 452 |
+
"BlinkDL/rwkv-3-pile-430m",
|
| 453 |
+
"BlinkDL/rwkv-2-pile-430m",
|
| 454 |
+
"BlinkDL/rwkv-3-pile-169m",
|
| 455 |
+
)
|
| 456 |
+
|
| 457 |
+
ai_squared_models_list = (
|
| 458 |
+
"aisquared/dlite-dais-2023",
|
| 459 |
+
"aisquared/chopt-1_3b",
|
| 460 |
+
"aisquared/chopt-350m",
|
| 461 |
+
"aisquared/chopt-125m",
|
| 462 |
+
"aisquared/chopt-2_7b",
|
| 463 |
+
"aisquared/dlite-v2-1_5b",
|
| 464 |
+
"aisquared/dlite-v2-774m",
|
| 465 |
+
"aisquared/dlite-v2-355m",
|
| 466 |
+
"aisquared/dlite-v2-124m",
|
| 467 |
+
"aisquared/dlite-v1-355m",
|
| 468 |
+
"aisquared/dlite-v1-774m",
|
| 469 |
+
"aisquared/dlite-v1-1_5b",
|
| 470 |
+
"aisquared/dlite-v1-124m",
|
| 471 |
+
"aisquared/chopt-research-350m",
|
| 472 |
+
"aisquared/chopt-research-125m",
|
| 473 |
+
"aisquared/chopt-research-2_7b",
|
| 474 |
+
"aisquared/chopt-research-1_3b",
|
| 475 |
+
)
|
| 476 |
+
|
| 477 |
+
bloom_models_list = (
|
| 478 |
+
"bigscience/bloom-3b-intermediate",
|
| 479 |
+
"bigscience/bloom",
|
| 480 |
+
"bigscience/bloomz-p3",
|
| 481 |
+
"bigscience/bloomz-mt",
|
| 482 |
+
"bigscience/bloomz-7b1-mt",
|
| 483 |
+
"bigscience/bloom-1b7-intermediate",
|
| 484 |
+
"bigscience/bloom-560m-intermediate",
|
| 485 |
+
"bigscience/bloomz-560m",
|
| 486 |
+
"bigscience/bloomz-1b1",
|
| 487 |
+
"bigscience/bloomz-1b7",
|
| 488 |
+
"bigscience/bloomz-3b",
|
| 489 |
+
"bigscience/bloomz-7b1",
|
| 490 |
+
'bigscience/bloomz',
|
| 491 |
+
"bigscience/bloom-1b7",
|
| 492 |
+
"bigscience/bloom-560m",
|
| 493 |
+
"bigscience/bloom-3b",
|
| 494 |
+
"bigscience/bigscience-small-testing",
|
| 495 |
+
"bigscience/distill-bloom-1b3",
|
| 496 |
+
"bigscience/bloom-1b1",
|
| 497 |
+
"bigscience/distill-bloom-1b3-10x",
|
| 498 |
+
"bigscience/test-bloomd",
|
| 499 |
+
"bigscience/test-bloomd-6b3",
|
| 500 |
+
"bigscience/bloom-7b1",
|
| 501 |
+
"bigscience/bloom-petals",
|
| 502 |
+
"bigscience/bloom-1b1-intermediate",
|
| 503 |
+
"bigscience/bloom-7b1-intermediate",
|
| 504 |
+
"bigscience/bloom-7b1-petals",
|
| 505 |
+
"bigscience/bloomz-petals",
|
| 506 |
+
"bigscience/bloomz-7b1-p3",
|
| 507 |
+
)
|
| 508 |
+
tiiuae_models_list = (
|
| 509 |
+
"tiiuae/falcon-180B",
|
| 510 |
+
"tiiuae/falcon-180B-chat",
|
| 511 |
+
"tiiuae/falcon-40b",
|
| 512 |
+
"tiiuae/falcon-7b",
|
| 513 |
+
"tiiuae/falcon-7b-instruct",
|
| 514 |
+
"tiiuae/falcon-40b-instruct",
|
| 515 |
+
"tiiuae/falcon-rw-7b",
|
| 516 |
+
"tiiuae/falcon-rw-1b",
|
| 517 |
+
)
|
| 518 |
+
|
| 519 |
+
openlm_models_list = (
|
| 520 |
+
"openlm-research/open_llama_3b_v2",
|
| 521 |
+
"openlm-research/open_llama_7b_v2",
|
| 522 |
+
"openlm-research/open_llama_13b",
|
| 523 |
+
"openlm-research/open_llama_7b",
|
| 524 |
+
"openlm-research/open_llama_3b",
|
| 525 |
+
)
|
| 526 |
+
stabilityai_models_list = (
|
| 527 |
+
"stabilityai/StableBeluga-7B",
|
| 528 |
+
"stabilityai/StableBeluga-13B",
|
| 529 |
+
"stabilityai/StableBeluga2",
|
| 530 |
+
"stabilityai/stablelm-base-alpha-3b-v2",
|
| 531 |
+
"stabilityai/stablelm-base-alpha-7b-v2",
|
| 532 |
+
"stabilityai/japanese-stablelm-instruct-alpha-7b",
|
| 533 |
+
"stabilityai/japanese-stablelm-base-alpha-7b",
|
| 534 |
+
"stabilityai/stablecode-completion-alpha-3b-4k",
|
| 535 |
+
"stabilityai/stablecode-instruct-alpha-3b",
|
| 536 |
+
"stabilityai/stablecode-completion-alpha-3b",
|
| 537 |
+
"stabilityai/StableBeluga1-Delta",
|
| 538 |
+
"stabilityai/stablelm-base-alpha-3b",
|
| 539 |
+
"stabilityai/stablelm-base-alpha-7b",
|
| 540 |
+
"stabilityai/stablelm-tuned-alpha-3b",
|
| 541 |
+
"stabilityai/stablelm-tuned-alpha-7b",
|
| 542 |
+
)
|
| 543 |
+
|
| 544 |
+
lmsys_models_list = (
|
| 545 |
+
"lmsys/vicuna-13b-v1.5-16k",
|
| 546 |
+
"lmsys/vicuna-13b-v1.5",
|
| 547 |
+
"lmsys/vicuna-7b-v1.5-16k",
|
| 548 |
+
"lmsys/longchat-7b-v1.5-32k",
|
| 549 |
+
"lmsys/vicuna-7b-v1.5",
|
| 550 |
+
"lmsys/vicuna-7b-v1.3",
|
| 551 |
+
"lmsys/vicuna-13b-v1.3",
|
| 552 |
+
"lmsys/vicuna-7b-v1.1",
|
| 553 |
+
"lmsys/vicuna-13b-v1.1",
|
| 554 |
+
"lmsys/vicuna-13b-delta-v0",
|
| 555 |
+
"lmsys/vicuna-7b-delta-v0",
|
| 556 |
+
"lmsys/vicuna-13b-delta-v1.1",
|
| 557 |
+
"lmsys/vicuna-7b-delta-v1.1",
|
| 558 |
+
"lmsys/vicuna-33b-v1.3",
|
| 559 |
+
"lmsys/longchat-13b-16k",
|
| 560 |
+
'lmsys/longchat-7b-16k',
|
| 561 |
+
)
|
| 562 |
+
|
| 563 |
+
togethercomputer_models_list = (
|
| 564 |
+
'togethercomputer/Llama-2-7B-32K-Instruct',
|
| 565 |
+
'togethercomputer/RedPajama-INCITE-7B-Instruct',
|
| 566 |
+
'togethercomputer/LLaMA-2-7B-32K',
|
| 567 |
+
'togethercomputer/RedPajama-INCITE-7B-Base',
|
| 568 |
+
'togethercomputer/RedPajama-INCITE-7B-Chat',
|
| 569 |
+
'togethercomputer/RedPajama-INCITE-Chat-3B-v1',
|
| 570 |
+
'togethercomputer/RedPajama-INCITE-Instruct-3B-v1',
|
| 571 |
+
'togethercomputer/RedPajama-INCITE-Base-3B-v1',
|
| 572 |
+
'togethercomputer/GPT-NeoXT-Chat-Base-20B',
|
| 573 |
+
'togethercomputer/Pythia-Chat-Base-7B',
|
| 574 |
+
'togethercomputer/GPT-JT-Moderation-6B',
|
| 575 |
+
'togethercomputer/GPT-JT-6B-v1',
|
| 576 |
+
'togethercomputer/GPT-JT-6B-v0'
|
| 577 |
+
)
|
| 578 |
+
mosaic_models_list = (
|
| 579 |
+
'mosaicml/mpt-7b-chat',
|
| 580 |
+
'mosaicml/mpt-30b-chat',
|
| 581 |
+
'mosaicml/mpt-7b-8k-chat',
|
| 582 |
+
'mosaicml/mpt-7b-instruct',
|
| 583 |
+
'mosaicml/mpt-7b-8k-instruct',
|
| 584 |
+
'mosaicml/mpt-7b-8k',
|
| 585 |
+
'mosaicml/mpt-30b',
|
| 586 |
+
'mosaicml/mpt-7b',
|
| 587 |
+
'mosaicml/mpt-7b-storywriter',
|
| 588 |
+
'mosaicml/mpt-30b-instruct',
|
| 589 |
+
'mosaicml/mpt-1b-redpajama-200b',
|
| 590 |
+
'mosaicml/mpt-1b-redpajama-200b-dolly'
|
| 591 |
+
)
|
| 592 |
+
|
| 593 |
+
h20ai_models_list = (
|
| 594 |
+
'h2oai/h2ogpt-16k-codellama-7b-python',
|
| 595 |
+
'h2oai/h2ogpt-16k-codellama-7b-instruct',
|
| 596 |
+
'h2oai/h2ogpt-16k-codellama-7b',
|
| 597 |
+
'h2oai/h2ogpt-16k-codellama-34b-python',
|
| 598 |
+
'h2oai/h2ogpt-16k-codellama-34b-instruct',
|
| 599 |
+
'h2oai/h2ogpt-16k-codellama-13b-python',
|
| 600 |
+
'h2oai/h2ogpt-16k-codellama-13b-instruct',
|
| 601 |
+
'h2oai/h2ogpt-16k-codellama-13b',
|
| 602 |
+
'h2oai/h2ogpt-16k-codellama-34b',
|
| 603 |
+
'h2oai/h2ogpt-4096-llama2-13b-chat',
|
| 604 |
+
'h2oai/h2ogpt-4096-llama2-70b-chat',
|
| 605 |
+
'h2oai/h2ogpt-4096-llama2-7b-chat',
|
| 606 |
+
'h2oai/h2ogpt-4096-llama2-13b',
|
| 607 |
+
'h2oai/h2ogpt-4096-llama2-7b',
|
| 608 |
+
'h2oai/h2ogpt-4096-llama2-70b',
|
| 609 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-falcon-40b-v2',
|
| 610 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-falcon-7b-v3',
|
| 611 |
+
'h2oai/h2ogpt-research-oasst1-llama-65b',
|
| 612 |
+
'h2oai/h2ogpt-gm-oasst1-en-xgen-7b-8k',
|
| 613 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-open-llama-3b',
|
| 614 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-open-llama-7b',
|
| 615 |
+
'h2oai/h2ogpt-oasst1-falcon-40b',
|
| 616 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-falcon-40b-v1',
|
| 617 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-open-llama-13b',
|
| 618 |
+
'h2oai/h2ogpt-oig-oasst1-falcon-40b',
|
| 619 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-falcon-7b',
|
| 620 |
+
'h2oai/h2ogpt-gm-oasst1-multilang-2048-falcon-7b',
|
| 621 |
+
'h2oai/h2ogpt-oasst1-512-12b',
|
| 622 |
+
'h2oai/h2ogpt-oig-oasst1-256-6_9b',
|
| 623 |
+
'h2oai/h2ogpt-oig-oasst1-512-6_9b',
|
| 624 |
+
'h2oai/h2ogpt-research-oig-oasst1-512-30b',
|
| 625 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-open-llama-7b-preview-700bt',
|
| 626 |
+
'h2oai/h2ogpt-gm-oasst1-en-1024-open-llama-7b-preview-400bt',
|
| 627 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-open-llama-7b-preview-300bt-v2',
|
| 628 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-open-llama-7b-preview-300bt',
|
| 629 |
+
'h2oai/h2ogpt-gm-oasst1-en-1024-12b',
|
| 630 |
+
'h2oai/h2ogpt-gm-oasst1-en-1024-20b',
|
| 631 |
+
'h2oai/h2ogpt-gm-oasst1-multilang-1024-20b',
|
| 632 |
+
'h2oai/h2ogpt-oasst1-512-20b'
|
| 633 |
+
)
|
| 634 |
+
|
| 635 |
+
facebook_models_list = (
|
| 636 |
+
'facebook/xglm-4.5B',
|
| 637 |
+
'facebook/galactica-125m',
|
| 638 |
+
'facebook/opt-iml-1.3b',
|
| 639 |
+
'facebook/opt-iml-max-1.3b',
|
| 640 |
+
'facebook/opt-iml-max-30b',
|
| 641 |
+
'facebook/opt-iml-30b',
|
| 642 |
+
'facebook/galactica-120b',
|
| 643 |
+
'facebook/galactica-30b',
|
| 644 |
+
'facebook/galactica-6.7b',
|
| 645 |
+
'facebook/galactica-1.3b',
|
| 646 |
+
'facebook/opt-66b',
|
| 647 |
+
'facebook/opt-30b',
|
| 648 |
+
'facebook/opt-13b',
|
| 649 |
+
'facebook/opt-6.7b',
|
| 650 |
+
'facebook/opt-2.7b',
|
| 651 |
+
'facebook/opt-1.3b',
|
| 652 |
+
'facebook/opt-350m',
|
| 653 |
+
'facebook/opt-125m',
|
| 654 |
+
'facebook/incoder-1B',
|
| 655 |
+
'facebook/incoder-6B',
|
| 656 |
+
'facebook/xglm-7.5B',
|
| 657 |
+
'facebook/xglm-564M',
|
| 658 |
+
'facebook/xglm-2.9B',
|
| 659 |
+
'facebook/xglm-1.7B'
|
| 660 |
+
)
|
| 661 |
+
|
| 662 |
+
model_dict = {
|
| 663 |
+
"google": google_models_list,
|
| 664 |
+
"mbzuai": mbzuai_models_list,
|
| 665 |
+
"eleutherai": eleutherai_models_list,
|
| 666 |
+
"cerebras": cerebras_models_list,
|
| 667 |
+
"open_assistant": open_assistant_models_list,
|
| 668 |
+
"databricks": databricks_models_list,
|
| 669 |
+
"nomic_ai": nomic_ai_models_list,
|
| 670 |
+
"blinkdl": blinkdl_models_list,
|
| 671 |
+
"aisquared": ai_squared_models_list,
|
| 672 |
+
"bloom": bloom_models_list,
|
| 673 |
+
"tiiuae": tiiuae_models_list,
|
| 674 |
+
"openlm": openlm_models_list,
|
| 675 |
+
"stabilityai": stabilityai_models_list,
|
| 676 |
+
"lmsys": lmsys_models_list,
|
| 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:
|
| 684 |
+
models_count = 0
|
| 685 |
+
for i in model_dict:
|
| 686 |
+
models_count += len(model_dict[i])
|
| 687 |
+
return models_count
|
| 688 |
+
|
| 689 |
+
return model_dict[model_parent]
|
| 690 |
+
|