SamDecoderConfig

interface SamDecoderConfig

import type { SamDecoderConfig } from '@driftengine/capture';

Properties

Name	Type	Description
`dim`readonly	`number`
`heads`readonly	`number`
`mlp`readonly	`number`
`depth`readonly	`number`
`downsample`readonly	`number`	The width attention between tokens and image narrows by.
`masks`readonly	`number`	The mask tokens: one for an unambiguous prompt and three for an ambiguous one.
`iouHidden`readonly	`number`
`iouDepth`readonly	`number`
`maskChannels`readonly	`number`	The channels a mask input is downscaled through.