class

GS::MASt3R::MASt3REncoderV2

Inherits Reference < Object

MASt3R Encoder v2: standard ViT without cross-attention Uses RoPE for position embeddings (computed, not stored)

Constructors

new(config : MASt3REncoderConfigV2, device : Tensor::Device = Tensor::Device::GPU)
Source

Instance methods

blocks
Source
config
Source
forward(x : Autograd::Variable) : Autograd::Variable

Encode single image x: [batch, channels, height, width] Returns: [batch, num_patches, embed_dim]

Source
forward_pair(x1 : Autograd::Variable, x2 : Autograd::Variable) : Tuple(Autograd::Variable, Autograd::Variable)

Encode image pair (for stereo) Simply encodes each image separately (cross-attention happens in decoder)

Source
norm
Source
parameters
Source
patch_embed
Source