Model Zoo

Note

• For all the pretraining and finetuning, we adopt spaese/uniform sampling.
• #Frame $=$ #input_frame $\times$ #crop $\times$ #clip
• #input_frame means how many frames are input for model per inference
• #crop means spatial crops (e.g., 3 for left/right/center)
• #clip means temporal clips (e.g., 4 means repeted sampling four clips with different start indices)

Pretraining

TBD

Distillation

TBD

Finetuning

K710

TBD

K400

Model	Setting	#Frame	Top-1	Model	Shell
$\text{InternVideo2}_{s1}$-1B	K-Mash PT + K710 FT	8x3x4	91.3	:hugs: HF link	TBD
$\text{InternVideo2}_{s1}$-1B	K-Mash PT + K710 FT	16x3x4	91.6	:hugs: HF link	TBD
$\text{InternVideo2}_{s1}$-6B	K-Mash PT + K710 FT	8x3x4	91.9	TBD	TBD
$\text{InternVideo2}_{s1}$-6B	K-Mash PT + K710 FT	16x3x4	92.1	TBD	TBD
$\text{InternVideo2}_{dist}$-S/14	K-Mash PT + K710 FT	8x3x4	85.4	:hugs: HF link	TBD
$\text{InternVideo2}_{dist}$-B/14	K-Mash PT + K710 FT	8x3x4	88.4	:hugs: HF link	TBD
$\text{InternVideo2}_{dist}$-L/14	K-Mash PT + K710 FT	8x3x4	90.4	:hugs: HF link	TBD
$\text{FluxViT}$-S/14	K-Mash PT + K710 FT	8x3x4	87.3	Link	run.sh
$\text{FluxViT}$-B/14	K-Mash PT + K710 FT	8x3x4	89.3	Link	run.sh

SthSth V2

TBD