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

Track a series of values and provide access to smoothed values over a window or the global series average.

Multi-head Attention block with relative position embeddings.

Transformer blocks with support of window attention and residual propagation blocks

attend patch embeddings with high frequency componenets

Image to Patch Embedding.

This class computes an assignment between the targets and the predictions of the network For efficiency reasons, the targets don't include the no

Image to Patch Embedding.

Positional encoding using random spatial frequencies.

Positional encoding using random spatial frequencies.

This module converts the model's output into the format expected by the coco api

Resizes images to the longest side 'target_length', as well as provides methods for resizing coordinates and boxes. Provides methods for

This class computes the loss for DETR. The process happens in two steps: 1) we compute hungarian assignment between ground truth boxes an

attend patch embeddings with high frequency componenets

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

Randomly selects between transforms1 and transforms2, with probability p for transforms1 and (1 - p) for transforms2

This model should not be called directly, but is used in ONNX export. It combines the prompt encoder, mask decoder, and mask postprocessing o