An attention layer that allows for downscaling the size of the embedding after projection to queries, keys, and values.

A structure for storing masks and their related data in batched format. Implements basic filtering and concatenation.

A list of video frames to be load asynchronously without blocking session start.

Dice coeff for individual examples

Dice coeff for individual examples

Image to Patch Embedding.

Positional encoding using random spatial frequencies.

r"""ConvNeXt Block. There are two equivalent implementations: (1) DwConv -> LayerNorm (channels_first) -> 1x1 Conv -> GELU -> 1x1 Conv; all in (N,

A modified variant of Feature Pyramid Network (FPN) neck (we remove output conv and also do bicubic interpolation similar to ViT pos embe

Reference: https://arxiv.org/abs/2306.00989

Progressively downsample a mask by total_stride, each time by stride. Note that LayerNorm is applied per *token*, like in ViT. With each

This is a more standard version of the position embedding, very similar to the one used by the Attention is all you need paper, generalized t

Attention with rotary position encoding.

The predictor class to handle user interactions and manage inference states.