Spaces:
Running
Running
models.tsx
Browse files- src/types/{types.ts → _types.ts} +0 -0
- src/utils/models.ts +127 -9
src/types/{types.ts → _types.ts}
RENAMED
File without changes
|
src/utils/models.ts
CHANGED
@@ -1,11 +1,10 @@
|
|
1 |
-
// Contains information about pre-set models
|
2 |
type ModelInfo = {
|
3 |
-
name: string
|
4 |
-
params: number
|
5 |
-
hidden_size: number
|
6 |
-
intermediate_size: number
|
7 |
-
num_hidden_layers: number
|
8 |
-
}
|
9 |
|
10 |
export const MODELS: ModelInfo[] = [
|
11 |
{
|
@@ -29,6 +28,125 @@ export const MODELS: ModelInfo[] = [
|
|
29 |
intermediate_size: 28672,
|
30 |
num_hidden_layers: 80,
|
31 |
},
|
32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
-
export default MODELS
|
|
|
|
|
1 |
type ModelInfo = {
|
2 |
+
name: string;
|
3 |
+
params: number;
|
4 |
+
hidden_size: number;
|
5 |
+
intermediate_size: number;
|
6 |
+
num_hidden_layers: number;
|
7 |
+
};
|
8 |
|
9 |
export const MODELS: ModelInfo[] = [
|
10 |
{
|
|
|
28 |
intermediate_size: 28672,
|
29 |
num_hidden_layers: 80,
|
30 |
},
|
31 |
+
{
|
32 |
+
name: 'Mistral (13B)',
|
33 |
+
params: 13,
|
34 |
+
hidden_size: 4096,
|
35 |
+
intermediate_size: 14336,
|
36 |
+
num_hidden_layers: 60,
|
37 |
+
},
|
38 |
+
{
|
39 |
+
name: 'Mistral (13B)',
|
40 |
+
params: 13,
|
41 |
+
hidden_size: 5120,
|
42 |
+
intermediate_size: 13824,
|
43 |
+
num_hidden_layers: 40,
|
44 |
+
},
|
45 |
+
{
|
46 |
+
name: 'Mistral (7B)',
|
47 |
+
params: 7,
|
48 |
+
hidden_size: 4096,
|
49 |
+
intermediate_size: 14336,
|
50 |
+
num_hidden_layers: 32,
|
51 |
+
},
|
52 |
+
{
|
53 |
+
name: 'Qwen (7B)',
|
54 |
+
params: 7,
|
55 |
+
hidden_size: 4096,
|
56 |
+
intermediate_size: 22016,
|
57 |
+
num_hidden_layers: 32,
|
58 |
+
},
|
59 |
+
{
|
60 |
+
name: 'Qwen (1.5 7B)',
|
61 |
+
params: 7,
|
62 |
+
hidden_size: 4096,
|
63 |
+
intermediate_size: 11008,
|
64 |
+
num_hidden_layers: 32,
|
65 |
+
},
|
66 |
+
{
|
67 |
+
name: 'Qwen (7B)',
|
68 |
+
params: 7,
|
69 |
+
hidden_size: 3584,
|
70 |
+
intermediate_size: 18944,
|
71 |
+
num_hidden_layers: 28,
|
72 |
+
},
|
73 |
+
{
|
74 |
+
name: 'Qwen (7B)',
|
75 |
+
params: 7,
|
76 |
+
hidden_size: 4096,
|
77 |
+
intermediate_size: 11008,
|
78 |
+
num_hidden_layers: 32,
|
79 |
+
},
|
80 |
+
{
|
81 |
+
name: 'Llava (7B)',
|
82 |
+
params: 7,
|
83 |
+
hidden_size: 4096,
|
84 |
+
intermediate_size: 11008,
|
85 |
+
num_hidden_layers: 32,
|
86 |
+
},
|
87 |
+
{
|
88 |
+
name: 'Llava (1.5 13B)',
|
89 |
+
params: 13,
|
90 |
+
hidden_size: 5120,
|
91 |
+
intermediate_size: 13824,
|
92 |
+
num_hidden_layers: 40,
|
93 |
+
},
|
94 |
+
{
|
95 |
+
name: 'Llava (1.6 34B)',
|
96 |
+
params: 34,
|
97 |
+
hidden_size: 7168,
|
98 |
+
intermediate_size: 20480,
|
99 |
+
num_hidden_layers: 60,
|
100 |
+
},
|
101 |
+
{
|
102 |
+
name: 'Gemma (2B)',
|
103 |
+
params: 2,
|
104 |
+
hidden_size: 2048,
|
105 |
+
intermediate_size: 16384,
|
106 |
+
num_hidden_layers: 18,
|
107 |
+
},
|
108 |
+
{
|
109 |
+
name: 'Gemma (2.9B)',
|
110 |
+
params: 2.9,
|
111 |
+
hidden_size: 3584,
|
112 |
+
intermediate_size: 14336,
|
113 |
+
num_hidden_layers: 42,
|
114 |
+
},
|
115 |
+
{
|
116 |
+
name: 'Gemma (27B)',
|
117 |
+
params: 27,
|
118 |
+
hidden_size: 4608,
|
119 |
+
intermediate_size: 36864,
|
120 |
+
num_hidden_layers: 46,
|
121 |
+
},
|
122 |
+
{
|
123 |
+
name: 'Mixtral (46B)',
|
124 |
+
params: 46, // Example parameter value
|
125 |
+
hidden_size: 4096, // Example hidden size
|
126 |
+
intermediate_size: 14336, // Example intermediate size
|
127 |
+
num_hidden_layers: 32,
|
128 |
+
},
|
129 |
+
{
|
130 |
+
name: 'LLaMA 3 (8B)',
|
131 |
+
params: 8,
|
132 |
+
hidden_size: 4096,
|
133 |
+
intermediate_size: 14336,
|
134 |
+
num_hidden_layers: 32,
|
135 |
+
},
|
136 |
+
{
|
137 |
+
name: 'LLaMA 3 (8B)',
|
138 |
+
params: 8,
|
139 |
+
hidden_size: 4096,
|
140 |
+
intermediate_size: 14336,
|
141 |
+
num_hidden_layers: 32,
|
142 |
+
},
|
143 |
+
{
|
144 |
+
name: 'LLaMA 3 (70B)',
|
145 |
+
params: 70,
|
146 |
+
hidden_size: 8192,
|
147 |
+
intermediate_size: 28672,
|
148 |
+
num_hidden_layers: 80,
|
149 |
+
},
|
150 |
+
];
|
151 |
|
152 |
+
export default MODELS;
|