open_clip performance and throughput

	B	C	D	E	F	G	H	I	J	K	L
1	Model	pretrain	Imagenet top1	Imagenet top5	image size	inference image T4 (ms) (bs1, 515, cuda 11.7, torch 1.13.1)	inference image 3090 (ms) (bs1, 510, cuda 11.6 torch 1.13.1)	average relative image inference (to ViT-L/14 openai)	text T4 (ms) (bs1, 510, cuda 11.6, torch 1.13.1)	text 3090 (ms) (bs1, 515, cuda 11.7, torch 1.13.1)	average relative text inference (to ViT-L/14 openai)
2	ViT-bigG-14	laion2B-39B-b160k	0.801	0.960	224x224	N/A	30.91	3.23	NA	14.18	2.92
3	ViT-H-14	laion2B-s32B-b79K	0.779	0.952	224x224	37.25	17.03	1.81	15.69	9.17	1.86
4	xlm-roberta-large-ViT-H-14	frozen_laion5b_s13b_b90k	0.770	0.951	224x224	37.94	13.95	1.66	21.31	11.41	2.42
5	convnext_large_d_320	laion2b_s29b_b131k_ft_soup	0.768	0.951	320x320	18.23	10.23	0.98	10.71	7.17	1.36
6	ViT-g-14	laion2B-s12B-b42K	0.767	0.948	224x224	56.21	21.88	2.53	15.20	9.19	1.83
7	convnext_large_d_320 -	laion2b_s29b_b131k_ft	0.766	0.949	320x320	17.93	10.34	0.98	11.00	6.95	1.36
8	ViT-L-14-336	openai	0.766	0.951	336x336	54.91	13.96	2.08	8.61	5.03	1.02
9	convnext_large_d	laion2b_s26b_b102k_augreg	0.759	0.946	256x256	18.98	9.45	0.96	10.92	6.40	1.30
10	coca_ViT-L-14	laion2b_s13b_b90k	0.757	0.943	224x224	18.68	9.63	0.96	8.40	5.16	1.02
11	ViT-L-14	openai	0.756	0.946	224x224	20.29	9.58	1.00	8.59	4.85	1.00
12	ViT-L-14	laion2B-s32B-b82K	0.753	0.943	224x224	18.14	9.17	0.93	7.93	4.92	0.97
13	ViT-L-14	laion400m_e32	0.728	0.929	224x224	18.54	8.54	0.90	8.10	4.79	0.97
14	ViT-L-14	laion400m_e31	0.727	0.930	224x224	19.81	9.11	0.96	8.13	4.67	0.95
15	convnext_base_w_320	laion_aesthetic_s13b_b82k	0.717	0.926	320x320	17.84	10.22	0.97	8.16	5.47	1.04
16	convnext_base_w	laion2b_s13b_b82k_augreg	0.715	0.928	256x256	18.64	9.55	0.96	8.21	4.83	0.98
17	convnext_base_w_320	laion_aesthetic_s13b_b82k_augreg	0.713	0.925	320x320	18.05	9.65	0.95	8.46	5.16	1.02
18	convnext_base_w	laion_aesthetic_s13b_b82k	0.710	0.922	256x256	18.00	10.18	0.97	8.29	4.80	0.98
19	convnext_base_w	laion2b_s13b_b82k	0.708	0.922	256x256	18.42	10.27	0.99	8.49	5.16	1.03
20	ViT-B-16	laion2b_s34b_b88k	0.702	0.918	224x224	8.86	4.91	0.47	8.14	4.91	0.98
21	ViT-B-16-plus-240	laion400m_e32	0.692	0.915	240x240	8.66	4.97	0.47	8.40	5.14	1.02
22	ViT-B-16-plus-240	laion400m_e31	0.691	0.914	240x240	8.71	4.80	0.47	8.05	4.98	0.98
23	ViT-B-16	openai	0.683	0.919	224x224	8.86	4.92	0.48	8.36	5.03	1.01
24	ViT-B-16	laion400m_e32	0.671	0.904	224x224	8.41	4.56	0.45	8.23	4.86	0.98
25	ViT-B-16	laion400m_e31	0.670	0.904	224x224	8.86	4.86	0.47	7.93	5.03	0.98
26	ViT-B-32	laion2b_s34b_b79k	0.665	0.899	224x224	8.21	4.48	0.44	8.09	4.82	0.97
27	convnext_base	laion400m_s13b_b51k	0.663	0.897	224x224	17.77	10.59	0.99	8.20	5.03	1.00
28	ViT-B-32	laion2b_e16	0.656	0.894	224x224	8.70	4.87	0.47	8.06	5.16	1.00
29	coca_ViT-B-32	laion2b_s13b_b90k	0.636	0.881	224x224	9.09	5.32	0.50	8.32	5.50	1.05
30	ViT-B-32	openai	0.634	0.888	224x224	9.22	5.50	0.51	8.61	5.34	1.05
31	ViT-B-32-quickgelu	openai	0.634	0.888	224x224	9.46	4.72	0.48	8.63	5.29	1.05
32	ViT-B-32-quickgelu	laion400m_e31	0.629	0.877	224x224	8.82	4.81	0.47	8.55	5.25	1.04
33	ViT-B-32-quickgelu	laion400m_e32	0.629	0.877	224x224	8.86	4.71	0.46	8.93	5.52	1.09
34	xlm-roberta-base-ViT-B-32	laion5b_s13b_b90k	0.624	0.881	224x224	8.25	4.63	0.45	11.21	6.29	1.30
35	roberta-ViT-B-32	laion2b_s12b_b32k	0.617	0.873	224x224	8.81	5.03	0.48	10.64	6.05	1.24
36	ViT-B-32	laion400m_e32	0.602	0.859	224x224	8.35	5.44	0.49	8.00	4.93	0.97
37	ViT-B-32	laion400m_e31	0.602	0.859	224x224	8.46	4.90	0.46	8.13	4.99	0.99
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