← Back to Hub

Error loading visualization








  

    
Encoder Parameters

    
ModelSigLIP-So400M

    
Input224 x 224 x 3

    
Patch Size14 x 14

    
Num Patches256

    
Embed Dim1152

    
Layers27

    
WeightsFrozen

  

  

    
Encoding Progress

    

  

  

    
Visual Tokens (16 shown)

    

  





  

    
Legend

    
Raw Image (Pixels)

    
Patch Grid

    
SigLIP Encoder

    
Data Flow

    
Visual Tokens

  







  

    
Ready

    
Press Play to see how SigLIP converts a raw camera image into a sequence of visual tokens — the first step of the pi0 architecture.

  

  

    
    
    
    
Speed1x