{ "schemaVersion": 1, "slug": "segment-anything", "title": "Turn a prompt into candidate masks", "className": "PromptableMaskDecoder", "question": "How do image features and a user prompt become a mask?", "description": "Reuse a cached image embedding, encode points or box corners, and follow two-way attention into three dynamic mask predictions.", "category": "Multimodal milestones", "tags": [ "SAM", "Segment Anything", "Segmentation", "Prompt encoder", "Two-way attention" ], "curator": "TensorViz", "attribution": "Original TensorViz teaching example. PyTorch provides the underlying operators.", "license": "No separate redistribution license has been declared for these project examples.", "limitations": "Untrained prompt/decoder subsystem, not a complete SAM model. Input is a cached synthetic embedding; the pretrained ViT image encoder, dense-mask encoder, tokenizer, image resizing and mask losses are omitted. The attention residual pattern follows the paper-level description with one head; this is not a parameter-compatible port. Only three candidate mask tokens are modeled, without the released implementation’s extra single-mask token. Quality scores are uncalibrated.", "openingStep": "prompt", "provenance": { "capturedAt": "2026-09-17T07:00:38.449307+00:00", "productRevision": "cf0ab1c9c58c1dc640235b742fa032cade00b43c", "sourceSha256": "bba1c4d0f25557c69579fade1e1ce55a8dd70d4af374ed69438d9a3652291d3e", "python": "3.13.13", "torch": "2.7.1", "execution": { "inputs": { "args": [ { "shape": [ 1, 8, 2, 2 ], "dtype": "float32", "fill": "randn" }, { "shape": [ 1, 2, 2 ], "dtype": "float32", "fill": "ones" }, { "shape": [ 1, 2 ], "dtype": "int64", "fill": "ones" } ] }, "device": "cpu", "mode": "eval", "seed": 0, "deterministic": true }, "outcome": "passed", "outputShape": [ 1, 195 ], "method": "TensorViz multifile projection and isolated execution worker", "scope": "SAM-inspired prompt-encoder and two-way mask-decoder subsystem with cached synthetic image features. Prompt permutation/location/type, hypernetwork dot-product and gradient checks. Output packs three 8 × 8 mask logits and three quality scores.", "numericalChecks": [ "Reordering sparse prompts preserves mask and quality outputs", "Moving a point or changing foreground/background type changes outputs without modifying cached image features", "Hypernetwork masks equal explicit per-pixel feature-weight dot products", "Mask gradients reach prompt coordinates and the image-to-token attention path" ] }, "layers": [ { "id": "input", "label": "input", "op": "Input", "kind": "input", "parent": null, "params": { "shape": "1,3,8" } }, { "id": "flatten@PromptableMaskDecoder#0", "label": "flatten@PromptableMaskDecoder#0", "op": "Flatten", "kind": "layer", "parent": null, "params": { "start_dim": 2 }, "source": { "line": 177, "endLine": 177 } }, { "id": "prompt_encoder", "label": "prompt_encoder", "op": "Module", "kind": "module", "parent": null, "params": {}, "source": { "line": 160, "endLine": 160 } }, { "id": "python@PromptableMaskDecoder#0", "label": "python@PromptableMaskDecoder#0", "op": "Python", "kind": "layer", "parent": null, "params": {}, "source": { "line": 178, "endLine": 178 } }, { "id": "output_tokens", "label": "output_tokens", "op": "Parameter", "kind": "layer", "parent": null, "params": {}, "source": { "line": 161, "endLine": 161 } }, { "id": "cat@PromptableMaskDecoder#0", "label": "cat@PromptableMaskDecoder#0", "op": "torch.cat", "kind": "layer", "parent": null, "params": { "dim": 1 }, "source": { "line": 174, "endLine": 176 } }, { "id": "no_mask", "label": "no_mask", "op": "Parameter", "kind": "layer", "parent": null, "params": {}, "source": { "line": 162, "endLine": 162 } }, { "id": "add@PromptableMaskDecoder#0", "label": "add@PromptableMaskDecoder#0", "op": "torch.add", "kind": "layer", "parent": null, "params": {}, "source": { "line": 177, "endLine": 177 } }, { "id": "decoder", "label": "decoder", "op": "Module", "kind": "module", "parent": null, "params": {}, "source": { "line": 167, "endLine": 167 } }, { "id": "upscale", "label": "upscale", "op": "Module", "kind": "module", "parent": null, "params": {}, "source": { "line": 168, "endLine": 168 } }, { "id": "masks", "label": "masks", "op": "Module", "kind": "module", "parent": null, "params": {}, "source": { "line": 169, "endLine": 169 } }, { "id": "flatten@PromptableMaskDecoder#1", "label": "flatten@PromptableMaskDecoder#1", "op": "Flatten", "kind": "layer", "parent": null, "params": { "start_dim": 1 }, "source": { "line": 183, "endLine": 183 } }, { "id": "quality", "label": "quality", "op": "Module", "kind": "module", "parent": null, "params": {}, "source": { "line": 170, "endLine": 170 } }, { "id": "cat@PromptableMaskDecoder#1", "label": "cat@PromptableMaskDecoder#1", "op": "torch.cat", "kind": "layer", "parent": null, "params": { "dim": -1 }, "source": { "line": 183, "endLine": 183 } }, { "id": "output", "label": "output", "op": "Output", "kind": "output", "parent": null, "params": {} }, { "id": "prompt_encoder.position", "label": "prompt_encoder.position", "op": "Module", "kind": "module", "parent": "prompt_encoder", "params": {}, "source": { "line": 26, "endLine": 26 } }, { "id": "prompt_encoder.type_embedding", "label": "prompt_encoder.type_embedding", "op": "Embedding", "kind": "layer", "parent": "prompt_encoder", "params": { "num_embeddings": 4, "embedding_dim": 8 }, "source": { "line": 27, "endLine": 27 } }, { "id": "add@prompt_encoder#0", "label": "add@prompt_encoder#0", "op": "torch.add", "kind": "layer", "parent": "prompt_encoder", "params": {}, "source": { "line": 31, "endLine": 31 } }, { "id": "cat@prompt_encoder.position#0", "label": "cat@prompt_encoder.position#0", "op": "torch.cat", "kind": "layer", "parent": "prompt_encoder.position", "params": { "dim": -1 }, "source": { "line": 20, "endLine": 20 } }, { "id": "decoder.first", "label": "decoder.first", "op": "Module", "kind": "module", "parent": "decoder", "params": {}, "source": { "line": 87, "endLine": 87 } }, { "id": "decoder.second", "label": "decoder.second", "op": "Module", "kind": "module", "parent": "decoder", "params": {}, "source": { "line": 88, "endLine": 88 } }, { "id": "add@decoder#0", "label": "add@decoder#0", "op": "torch.add", "kind": "layer", "parent": "decoder", "params": {}, "source": { "line": 96, "endLine": 99 } }, { "id": "add@decoder#1", "label": "add@decoder#1", "op": "torch.add", "kind": "layer", "parent": "decoder", "params": {}, "source": { "line": 96, "endLine": 99 } }, { "id": "decoder.final_attention", "label": "decoder.final_attention", "op": "Module", "kind": "module", "parent": "decoder", "params": {}, "source": { "line": 89, "endLine": 89 } }, { "id": "add@decoder#2", "label": "add@decoder#2", "op": "torch.add", "kind": "layer", "parent": "decoder", "params": {}, "source": { "line": 96, "endLine": 99 } }, { "id": "decoder.final_norm", "label": "decoder.final_norm", "op": "LayerNorm", "kind": "layer", "parent": "decoder", "params": { "normalized_shape": 8 }, "source": { "line": 90, "endLine": 90 } }, { "id": "decoder.first.self_attention", "label": "decoder.first.self_attention", "op": "Module", "kind": "module", "parent": "decoder.first", "params": { "arg_0": 8 }, "source": { "line": 54, "endLine": 54 } }, { "id": "add@decoder.first#0", "label": "add@decoder.first#0", "op": "torch.add", "kind": "layer", "parent": "decoder.first", "params": {}, "source": { "line": 67, "endLine": 67 } }, { "id": "decoder.first.self_norm", "label": "decoder.first.self_norm", "op": "LayerNorm", "kind": "layer", "parent": "decoder.first", "params": { "normalized_shape": 8 }, "source": { "line": 55, "endLine": 55 } }, { "id": "add@decoder.first#1", "label": "add@decoder.first#1", "op": "torch.add", "kind": "layer", "parent": "decoder.first", "params": {}, "source": { "line": 68, "endLine": 73 } }, { "id": "decoder.first.token_to_image", "label": "decoder.first.token_to_image", "op": "Module", "kind": "module", "parent": "decoder.first", "params": {}, "source": { "line": 56, "endLine": 56 } }, { "id": "add@decoder.first#2", "label": "add@decoder.first#2", "op": "torch.add", "kind": "layer", "parent": "decoder.first", "params": {}, "source": { "line": 68, "endLine": 73 } }, { "id": "decoder.first.token_norm", "label": "decoder.first.token_norm", "op": "LayerNorm", "kind": "layer", "parent": "decoder.first", "params": { "normalized_shape": 8 }, "source": { "line": 57, "endLine": 57 } }, { "id": "decoder.first.up", "label": "decoder.first.up", "op": "Linear", "kind": "layer", "parent": "decoder.first", "params": { "in_features": 8, "out_features": 32 }, "source": { "line": 58, "endLine": 58 } }, { "id": "decoder.first.relu", "label": "decoder.first.relu", "op": "ReLU", "kind": "layer", "parent": "decoder.first", "params": {}, "source": { "line": 59, "endLine": 59 } }, { "id": "decoder.first.down", "label": "decoder.first.down", "op": "Linear", "kind": "layer", "parent": "decoder.first", "params": { "in_features": 32, "out_features": 8 }, "source": { "line": 60, "endLine": 60 } }, { "id": "add@decoder.first#3", "label": "add@decoder.first#3", "op": "torch.add", "kind": "layer", "parent": "decoder.first", "params": {}, "source": { "line": 74, "endLine": 74 } }, { "id": "decoder.first.mlp_norm", "label": "decoder.first.mlp_norm", "op": "LayerNorm", "kind": "layer", "parent": "decoder.first", "params": { "normalized_shape": 8 }, "source": { "line": 61, "endLine": 61 } }, { "id": "add@decoder.first#4", "label": "add@decoder.first#4", "op": "torch.add", "kind": "layer", "parent": "decoder.first", "params": {}, "source": { "line": 75, "endLine": 80 } }, { "id": "decoder.first.image_to_token", "label": "decoder.first.image_to_token", "op": "Module", "kind": "module", "parent": "decoder.first", "params": {}, "source": { "line": 62, "endLine": 62 } }, { "id": "add@decoder.first#5", "label": "add@decoder.first#5", "op": "torch.add", "kind": "layer", "parent": "decoder.first", "params": {}, "source": { "line": 75, "endLine": 80 } }, { "id": "decoder.first.image_norm", "label": "decoder.first.image_norm", "op": "LayerNorm", "kind": "layer", "parent": "decoder.first", "params": { "normalized_shape": 8 }, "source": { "line": 63, "endLine": 63 } }, { "id": "decoder.first.self_attention.query", "label": "decoder.first.self_attention.query", "op": "Linear", "kind": "layer", "parent": "decoder.first.self_attention", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 38, "endLine": 38 } }, { "id": "decoder.first.self_attention.key", "label": "decoder.first.self_attention.key", "op": "Linear", "kind": "layer", "parent": "decoder.first.self_attention", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 39, "endLine": 39 } }, { "id": "div@decoder.first.self_attention#0", "label": "div@decoder.first.self_attention#0", "op": "torch.div", "kind": "layer", "parent": "decoder.first.self_attention", "params": {}, "source": { "line": 45, "endLine": 47 } }, { "id": "decoder.first.self_attention.value", "label": "decoder.first.self_attention.value", "op": "Linear", "kind": "layer", "parent": "decoder.first.self_attention", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 40, "endLine": 40 } }, { "id": "decoder.first.self_attention.softmax", "label": "decoder.first.self_attention.softmax", "op": "Softmax", "kind": "layer", "parent": "decoder.first.self_attention", "params": {}, "source": { "line": 41, "endLine": 41 } }, { "id": "decoder.first.self_attention.project", "label": "decoder.first.self_attention.project", "op": "Linear", "kind": "layer", "parent": "decoder.first.self_attention", "params": { "in_features": 4, "out_features": 8 }, "source": { "line": 42, "endLine": 42 } }, { "id": "decoder.first.token_to_image.query", "label": "decoder.first.token_to_image.query", "op": "Linear", "kind": "layer", "parent": "decoder.first.token_to_image", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 38, "endLine": 38 } }, { "id": "decoder.first.token_to_image.key", "label": "decoder.first.token_to_image.key", "op": "Linear", "kind": "layer", "parent": "decoder.first.token_to_image", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 39, "endLine": 39 } }, { "id": "div@decoder.first.token_to_image#0", "label": "div@decoder.first.token_to_image#0", "op": "torch.div", "kind": "layer", "parent": "decoder.first.token_to_image", "params": {}, "source": { "line": 45, "endLine": 47 } }, { "id": "decoder.first.token_to_image.value", "label": "decoder.first.token_to_image.value", "op": "Linear", "kind": "layer", "parent": "decoder.first.token_to_image", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 40, "endLine": 40 } }, { "id": "decoder.first.token_to_image.softmax", "label": "decoder.first.token_to_image.softmax", "op": "Softmax", "kind": "layer", "parent": "decoder.first.token_to_image", "params": {}, "source": { "line": 41, "endLine": 41 } }, { "id": "decoder.first.token_to_image.project", "label": "decoder.first.token_to_image.project", "op": "Linear", "kind": "layer", "parent": "decoder.first.token_to_image", "params": { "in_features": 4, "out_features": 8 }, "source": { "line": 42, "endLine": 42 } }, { "id": "decoder.first.image_to_token.query", "label": "decoder.first.image_to_token.query", "op": "Linear", "kind": "layer", "parent": "decoder.first.image_to_token", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 38, "endLine": 38 } }, { "id": "decoder.first.image_to_token.key", "label": "decoder.first.image_to_token.key", "op": "Linear", "kind": "layer", "parent": "decoder.first.image_to_token", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 39, "endLine": 39 } }, { "id": "div@decoder.first.image_to_token#0", "label": "div@decoder.first.image_to_token#0", "op": "torch.div", "kind": "layer", "parent": "decoder.first.image_to_token", "params": {}, "source": { "line": 45, "endLine": 47 } }, { "id": "decoder.first.image_to_token.value", "label": "decoder.first.image_to_token.value", "op": "Linear", "kind": "layer", "parent": "decoder.first.image_to_token", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 40, "endLine": 40 } }, { "id": "decoder.first.image_to_token.softmax", "label": "decoder.first.image_to_token.softmax", "op": "Softmax", "kind": "layer", "parent": "decoder.first.image_to_token", "params": {}, "source": { "line": 41, "endLine": 41 } }, { "id": "decoder.first.image_to_token.project", "label": "decoder.first.image_to_token.project", "op": "Linear", "kind": "layer", "parent": "decoder.first.image_to_token", "params": { "in_features": 4, "out_features": 8 }, "source": { "line": 42, "endLine": 42 } }, { "id": "decoder.second.self_attention", "label": "decoder.second.self_attention", "op": "Module", "kind": "module", "parent": "decoder.second", "params": { "arg_0": 8 }, "source": { "line": 54, "endLine": 54 } }, { "id": "add@decoder.second#0", "label": "add@decoder.second#0", "op": "torch.add", "kind": "layer", "parent": "decoder.second", "params": {}, "source": { "line": 67, "endLine": 67 } }, { "id": "decoder.second.self_norm", "label": "decoder.second.self_norm", "op": "LayerNorm", "kind": "layer", "parent": "decoder.second", "params": { "normalized_shape": 8 }, "source": { "line": 55, "endLine": 55 } }, { "id": "add@decoder.second#1", "label": "add@decoder.second#1", "op": "torch.add", "kind": "layer", "parent": "decoder.second", "params": {}, "source": { "line": 68, "endLine": 73 } }, { "id": "decoder.second.token_to_image", "label": "decoder.second.token_to_image", "op": "Module", "kind": "module", "parent": "decoder.second", "params": {}, "source": { "line": 56, "endLine": 56 } }, { "id": "add@decoder.second#2", "label": "add@decoder.second#2", "op": "torch.add", "kind": "layer", "parent": "decoder.second", "params": {}, "source": { "line": 68, "endLine": 73 } }, { "id": "decoder.second.token_norm", "label": "decoder.second.token_norm", "op": "LayerNorm", "kind": "layer", "parent": "decoder.second", "params": { "normalized_shape": 8 }, "source": { "line": 57, "endLine": 57 } }, { "id": "decoder.second.up", "label": "decoder.second.up", "op": "Linear", "kind": "layer", "parent": "decoder.second", "params": { "in_features": 8, "out_features": 32 }, "source": { "line": 58, "endLine": 58 } }, { "id": "decoder.second.relu", "label": "decoder.second.relu", "op": "ReLU", "kind": "layer", "parent": "decoder.second", "params": {}, "source": { "line": 59, "endLine": 59 } }, { "id": "decoder.second.down", "label": "decoder.second.down", "op": "Linear", "kind": "layer", "parent": "decoder.second", "params": { "in_features": 32, "out_features": 8 }, "source": { "line": 60, "endLine": 60 } }, { "id": "add@decoder.second#3", "label": "add@decoder.second#3", "op": "torch.add", "kind": "layer", "parent": "decoder.second", "params": {}, "source": { "line": 74, "endLine": 74 } }, { "id": "decoder.second.mlp_norm", "label": "decoder.second.mlp_norm", "op": "LayerNorm", "kind": "layer", "parent": "decoder.second", "params": { "normalized_shape": 8 }, "source": { "line": 61, "endLine": 61 } }, { "id": "add@decoder.second#4", "label": "add@decoder.second#4", "op": "torch.add", "kind": "layer", "parent": "decoder.second", "params": {}, "source": { "line": 75, "endLine": 80 } }, { "id": "decoder.second.image_to_token", "label": "decoder.second.image_to_token", "op": "Module", "kind": "module", "parent": "decoder.second", "params": {}, "source": { "line": 62, "endLine": 62 } }, { "id": "add@decoder.second#5", "label": "add@decoder.second#5", "op": "torch.add", "kind": "layer", "parent": "decoder.second", "params": {}, "source": { "line": 75, "endLine": 80 } }, { "id": "decoder.second.image_norm", "label": "decoder.second.image_norm", "op": "LayerNorm", "kind": "layer", "parent": "decoder.second", "params": { "normalized_shape": 8 }, "source": { "line": 63, "endLine": 63 } }, { "id": "decoder.second.self_attention.query", "label": "decoder.second.self_attention.query", "op": "Linear", "kind": "layer", "parent": "decoder.second.self_attention", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 38, "endLine": 38 } }, { "id": "decoder.second.self_attention.key", "label": "decoder.second.self_attention.key", "op": "Linear", "kind": "layer", "parent": "decoder.second.self_attention", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 39, "endLine": 39 } }, { "id": "div@decoder.second.self_attention#0", "label": "div@decoder.second.self_attention#0", "op": "torch.div", "kind": "layer", "parent": "decoder.second.self_attention", "params": {}, "source": { "line": 45, "endLine": 47 } }, { "id": "decoder.second.self_attention.value", "label": "decoder.second.self_attention.value", "op": "Linear", "kind": "layer", "parent": "decoder.second.self_attention", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 40, "endLine": 40 } }, { "id": "decoder.second.self_attention.softmax", "label": "decoder.second.self_attention.softmax", "op": "Softmax", "kind": "layer", "parent": "decoder.second.self_attention", "params": {}, "source": { "line": 41, "endLine": 41 } }, { "id": "decoder.second.self_attention.project", "label": "decoder.second.self_attention.project", "op": "Linear", "kind": "layer", "parent": "decoder.second.self_attention", "params": { "in_features": 4, "out_features": 8 }, "source": { "line": 42, "endLine": 42 } }, { "id": "decoder.second.token_to_image.query", "label": "decoder.second.token_to_image.query", "op": "Linear", "kind": "layer", "parent": "decoder.second.token_to_image", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 38, "endLine": 38 } }, { "id": "decoder.second.token_to_image.key", "label": "decoder.second.token_to_image.key", "op": "Linear", "kind": "layer", "parent": "decoder.second.token_to_image", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 39, "endLine": 39 } }, { "id": "div@decoder.second.token_to_image#0", "label": "div@decoder.second.token_to_image#0", "op": "torch.div", "kind": "layer", "parent": "decoder.second.token_to_image", "params": {}, "source": { "line": 45, "endLine": 47 } }, { "id": "decoder.second.token_to_image.value", "label": "decoder.second.token_to_image.value", "op": "Linear", "kind": "layer", "parent": "decoder.second.token_to_image", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 40, "endLine": 40 } }, { "id": "decoder.second.token_to_image.softmax", "label": "decoder.second.token_to_image.softmax", "op": "Softmax", "kind": "layer", "parent": "decoder.second.token_to_image", "params": {}, "source": { "line": 41, "endLine": 41 } }, { "id": "decoder.second.token_to_image.project", "label": "decoder.second.token_to_image.project", "op": "Linear", "kind": "layer", "parent": "decoder.second.token_to_image", "params": { "in_features": 4, "out_features": 8 }, "source": { "line": 42, "endLine": 42 } }, { "id": "decoder.second.image_to_token.query", "label": "decoder.second.image_to_token.query", "op": "Linear", "kind": "layer", "parent": "decoder.second.image_to_token", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 38, "endLine": 38 } }, { "id": "decoder.second.image_to_token.key", "label": "decoder.second.image_to_token.key", "op": "Linear", "kind": "layer", "parent": "decoder.second.image_to_token", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 39, "endLine": 39 } }, { "id": "div@decoder.second.image_to_token#0", "label": "div@decoder.second.image_to_token#0", "op": "torch.div", "kind": "layer", "parent": "decoder.second.image_to_token", "params": {}, "source": { "line": 45, "endLine": 47 } }, { "id": "decoder.second.image_to_token.value", "label": "decoder.second.image_to_token.value", "op": "Linear", "kind": "layer", "parent": "decoder.second.image_to_token", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 40, "endLine": 40 } }, { "id": "decoder.second.image_to_token.softmax", "label": "decoder.second.image_to_token.softmax", "op": "Softmax", "kind": "layer", "parent": "decoder.second.image_to_token", "params": {}, "source": { "line": 41, "endLine": 41 } }, { "id": "decoder.second.image_to_token.project", "label": "decoder.second.image_to_token.project", "op": "Linear", "kind": "layer", "parent": "decoder.second.image_to_token", "params": { "in_features": 4, "out_features": 8 }, "source": { "line": 42, "endLine": 42 } }, { "id": "decoder.final_attention.query", "label": "decoder.final_attention.query", "op": "Linear", "kind": "layer", "parent": "decoder.final_attention", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 38, "endLine": 38 } }, { "id": "decoder.final_attention.key", "label": "decoder.final_attention.key", "op": "Linear", "kind": "layer", "parent": "decoder.final_attention", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 39, "endLine": 39 } }, { "id": "div@decoder.final_attention#0", "label": "div@decoder.final_attention#0", "op": "torch.div", "kind": "layer", "parent": "decoder.final_attention", "params": {}, "source": { "line": 45, "endLine": 47 } }, { "id": "decoder.final_attention.value", "label": "decoder.final_attention.value", "op": "Linear", "kind": "layer", "parent": "decoder.final_attention", "params": { "in_features": 8, "out_features": 4 }, "source": { "line": 40, "endLine": 40 } }, { "id": "decoder.final_attention.softmax", "label": "decoder.final_attention.softmax", "op": "Softmax", "kind": "layer", "parent": "decoder.final_attention", "params": {}, "source": { "line": 41, "endLine": 41 } }, { "id": "decoder.final_attention.project", "label": "decoder.final_attention.project", "op": "Linear", "kind": "layer", "parent": "decoder.final_attention", "params": { "in_features": 4, "out_features": 8 }, "source": { "line": 42, "endLine": 42 } }, { "id": "upscale.first", "label": "upscale.first", "op": "ConvTranspose2d", "kind": "layer", "parent": "upscale", "params": { "in_channels": 8, "out_channels": 4, "kernel_size": 2, "stride": 2 }, "source": { "line": 106, "endLine": 106 } }, { "id": "permute@upscale#0", "label": "permute@upscale#0", "op": "torch.permute", "kind": "layer", "parent": "upscale", "params": { "dims": [ 0, 2, 3, 1 ] }, "source": { "line": 112, "endLine": 112 } }, { "id": "upscale.norm", "label": "upscale.norm", "op": "LayerNorm", "kind": "layer", "parent": "upscale", "params": { "normalized_shape": 4 }, "source": { "line": 107, "endLine": 107 } }, { "id": "permute@upscale#1", "label": "permute@upscale#1", "op": "torch.permute", "kind": "layer", "parent": "upscale", "params": { "dims": [ 0, 3, 1, 2 ] }, "source": { "line": 112, "endLine": 112 } }, { "id": "gelu@upscale#0", "label": "gelu@upscale#0", "op": "GELU", "kind": "layer", "parent": "upscale", "params": {}, "source": { "line": 112, "endLine": 112 } }, { "id": "upscale.second", "label": "upscale.second", "op": "ConvTranspose2d", "kind": "layer", "parent": "upscale", "params": { "in_channels": 4, "out_channels": 2, "kernel_size": 2, "stride": 2 }, "source": { "line": 108, "endLine": 108 } }, { "id": "gelu@upscale#1", "label": "gelu@upscale#1", "op": "GELU", "kind": "layer", "parent": "upscale", "params": {}, "source": { "line": 113, "endLine": 113 } }, { "id": "flatten@masks#0", "label": "flatten@masks#0", "op": "Flatten", "kind": "layer", "parent": "masks", "params": { "start_dim": 2 }, "source": { "line": 143, "endLine": 143 } }, { "id": "masks.mask0", "label": "masks.mask0", "op": "Module", "kind": "module", "parent": "masks", "params": {}, "source": { "line": 130, "endLine": 130 } }, { "id": "masks.mask1", "label": "masks.mask1", "op": "Module", "kind": "module", "parent": "masks", "params": {}, "source": { "line": 131, "endLine": 131 } }, { "id": "masks.mask2", "label": "masks.mask2", "op": "Module", "kind": "module", "parent": "masks", "params": {}, "source": { "line": 132, "endLine": 132 } }, { "id": "stack@masks#0", "label": "stack@masks#0", "op": "torch.stack", "kind": "layer", "parent": "masks", "params": { "dim": 1 }, "source": { "line": 135, "endLine": 142 } }, { "id": "python@masks#0", "label": "python@masks#0", "op": "Python", "kind": "layer", "parent": "masks", "params": {}, "source": { "line": 143, "endLine": 143 } }, { "id": "masks.mask0.first", "label": "masks.mask0.first", "op": "Linear", "kind": "layer", "parent": "masks.mask0", "params": { "in_features": 8, "out_features": 8 }, "source": { "line": 119, "endLine": 119 } }, { "id": "relu@masks.mask0#0", "label": "relu@masks.mask0#0", "op": "ReLU", "kind": "layer", "parent": "masks.mask0", "params": {}, "source": { "line": 124, "endLine": 124 } }, { "id": "masks.mask0.second", "label": "masks.mask0.second", "op": "Linear", "kind": "layer", "parent": "masks.mask0", "params": { "in_features": 8, "out_features": 8 }, "source": { "line": 120, "endLine": 120 } }, { "id": "relu@masks.mask0#1", "label": "relu@masks.mask0#1", "op": "ReLU", "kind": "layer", "parent": "masks.mask0", "params": {}, "source": { "line": 124, "endLine": 124 } }, { "id": "masks.mask0.project", "label": "masks.mask0.project", "op": "Linear", "kind": "layer", "parent": "masks.mask0", "params": { "in_features": 8, "out_features": 2 }, "source": { "line": 121, "endLine": 121 } }, { "id": "masks.mask1.first", "label": "masks.mask1.first", "op": "Linear", "kind": "layer", "parent": "masks.mask1", "params": { "in_features": 8, "out_features": 8 }, "source": { "line": 119, "endLine": 119 } }, { "id": "relu@masks.mask1#0", "label": "relu@masks.mask1#0", "op": "ReLU", "kind": "layer", "parent": "masks.mask1", "params": {}, "source": { "line": 124, "endLine": 124 } }, { "id": "masks.mask1.second", "label": "masks.mask1.second", "op": "Linear", "kind": "layer", "parent": "masks.mask1", "params": { "in_features": 8, "out_features": 8 }, "source": { "line": 120, "endLine": 120 } }, { "id": "relu@masks.mask1#1", "label": "relu@masks.mask1#1", "op": "ReLU", "kind": "layer", "parent": "masks.mask1", "params": {}, "source": { "line": 124, "endLine": 124 } }, { "id": "masks.mask1.project", "label": "masks.mask1.project", "op": "Linear", "kind": "layer", "parent": "masks.mask1", "params": { "in_features": 8, "out_features": 2 }, "source": { "line": 121, "endLine": 121 } }, { "id": "masks.mask2.first", "label": "masks.mask2.first", "op": "Linear", "kind": "layer", "parent": "masks.mask2", "params": { "in_features": 8, "out_features": 8 }, "source": { "line": 119, "endLine": 119 } }, { "id": "relu@masks.mask2#0", "label": "relu@masks.mask2#0", "op": "ReLU", "kind": "layer", "parent": "masks.mask2", "params": {}, "source": { "line": 124, "endLine": 124 } }, { "id": "masks.mask2.second", "label": "masks.mask2.second", "op": "Linear", "kind": "layer", "parent": "masks.mask2", "params": { "in_features": 8, "out_features": 8 }, "source": { "line": 120, "endLine": 120 } }, { "id": "relu@masks.mask2#1", "label": "relu@masks.mask2#1", "op": "ReLU", "kind": "layer", "parent": "masks.mask2", "params": {}, "source": { "line": 124, "endLine": 124 } }, { "id": "masks.mask2.project", "label": "masks.mask2.project", "op": "Linear", "kind": "layer", "parent": "masks.mask2", "params": { "in_features": 8, "out_features": 2 }, "source": { "line": 121, "endLine": 121 } }, { "id": "quality.first", "label": "quality.first", "op": "Linear", "kind": "layer", "parent": "quality", "params": { "in_features": 8, "out_features": 8 }, "source": { "line": 149, "endLine": 149 } }, { "id": "relu@quality#0", "label": "relu@quality#0", "op": "ReLU", "kind": "layer", "parent": "quality", "params": {}, "source": { "line": 154, "endLine": 154 } }, { "id": "quality.second", "label": "quality.second", "op": "Linear", "kind": "layer", "parent": "quality", "params": { "in_features": 8, "out_features": 8 }, "source": { "line": 150, "endLine": 150 } }, { "id": "relu@quality#1", "label": "relu@quality#1", "op": "ReLU", "kind": "layer", "parent": "quality", "params": {}, "source": { "line": 154, "endLine": 154 } }, { "id": "quality.scores", "label": "quality.scores", "op": "Linear", "kind": "layer", "parent": "quality", "params": { "in_features": 8, "out_features": 3 }, "source": { "line": 151, "endLine": 151 } } ], "connections": [ { "id": "edge-1", "source": "cat@prompt_encoder.position#0", "target": "add@prompt_encoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-2", "source": "prompt_encoder.position", "target": "add@prompt_encoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-3", "source": "prompt_encoder.type_embedding", "target": "add@prompt_encoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-4", "source": "decoder.first.self_attention.query", "target": "div@decoder.first.self_attention#0", "sourcePort": null, "targetPort": null }, { "id": "edge-5", "source": "decoder.first.self_attention.key", "target": "div@decoder.first.self_attention#0", "sourcePort": null, "targetPort": null }, { "id": "edge-6", "source": "div@decoder.first.self_attention#0", "target": "decoder.first.self_attention.softmax", "sourcePort": null, "targetPort": null }, { "id": "edge-7", "source": "decoder.first.self_attention.softmax", "target": "decoder.first.self_attention.project", "sourcePort": null, "targetPort": null }, { "id": "edge-8", "source": "decoder.first.self_attention.value", "target": "decoder.first.self_attention.project", "sourcePort": null, "targetPort": null }, { "id": "edge-9", "source": "decoder.first.token_to_image.query", "target": "div@decoder.first.token_to_image#0", "sourcePort": null, "targetPort": null }, { "id": "edge-10", "source": "decoder.first.token_to_image.key", "target": "div@decoder.first.token_to_image#0", "sourcePort": null, "targetPort": null }, { "id": "edge-11", "source": "div@decoder.first.token_to_image#0", "target": "decoder.first.token_to_image.softmax", "sourcePort": null, "targetPort": null }, { "id": "edge-12", "source": "decoder.first.token_to_image.softmax", "target": "decoder.first.token_to_image.project", "sourcePort": null, "targetPort": null }, { "id": "edge-13", "source": "decoder.first.token_to_image.value", "target": "decoder.first.token_to_image.project", "sourcePort": null, "targetPort": null }, { "id": "edge-14", "source": "decoder.first.image_to_token.query", "target": "div@decoder.first.image_to_token#0", "sourcePort": null, "targetPort": null }, { "id": "edge-15", "source": "decoder.first.image_to_token.key", "target": "div@decoder.first.image_to_token#0", "sourcePort": null, "targetPort": null }, { "id": "edge-16", "source": "div@decoder.first.image_to_token#0", "target": "decoder.first.image_to_token.softmax", "sourcePort": null, "targetPort": null }, { "id": "edge-17", "source": "decoder.first.image_to_token.softmax", "target": "decoder.first.image_to_token.project", "sourcePort": null, "targetPort": null }, { "id": "edge-18", "source": "decoder.first.image_to_token.value", "target": "decoder.first.image_to_token.project", "sourcePort": null, "targetPort": null }, { "id": "edge-19", "source": "decoder.first.self_attention.project", "target": "add@decoder.first#0", "sourcePort": null, "targetPort": null }, { "id": "edge-20", "source": "decoder.first.self_attention", "target": "add@decoder.first#0", "sourcePort": null, "targetPort": null }, { "id": "edge-21", "source": "add@decoder.first#0", "target": "decoder.first.self_norm", "sourcePort": null, "targetPort": null }, { "id": "edge-22", "source": "decoder.first.self_norm", "target": "add@decoder.first#1", "sourcePort": null, "targetPort": null }, { "id": "edge-23", "source": "add@decoder.first#1", "target": "decoder.first.token_to_image.query", "sourcePort": null, "targetPort": null }, { "id": "edge-24", "source": "add@decoder.first#1", "target": "decoder.first.token_to_image.key", "sourcePort": null, "targetPort": null }, { "id": "edge-25", "source": "add@decoder.first#1", "target": "decoder.first.token_to_image.value", "sourcePort": null, "targetPort": null }, { "id": "edge-26", "source": "add@decoder.first#1", "target": "decoder.first.token_to_image", "sourcePort": null, "targetPort": null }, { "id": "edge-27", "source": "decoder.first.self_norm", "target": "add@decoder.first#2", "sourcePort": null, "targetPort": null }, { "id": "edge-28", "source": "decoder.first.token_to_image.project", "target": "add@decoder.first#2", "sourcePort": null, "targetPort": null }, { "id": "edge-29", "source": "decoder.first.token_to_image", "target": "add@decoder.first#2", "sourcePort": null, "targetPort": null }, { "id": "edge-30", "source": "add@decoder.first#2", "target": "decoder.first.token_norm", "sourcePort": null, "targetPort": null }, { "id": "edge-31", "source": "decoder.first.token_norm", "target": "decoder.first.up", "sourcePort": null, "targetPort": null }, { "id": "edge-32", "source": "decoder.first.up", "target": "decoder.first.relu", "sourcePort": null, "targetPort": null }, { "id": "edge-33", "source": "decoder.first.relu", "target": "decoder.first.down", "sourcePort": null, "targetPort": null }, { "id": "edge-34", "source": "decoder.first.token_norm", "target": "add@decoder.first#3", "sourcePort": null, "targetPort": null }, { "id": "edge-35", "source": "decoder.first.down", "target": "add@decoder.first#3", "sourcePort": null, "targetPort": null }, { "id": "edge-36", "source": "add@decoder.first#3", "target": "decoder.first.mlp_norm", "sourcePort": null, "targetPort": null }, { "id": "edge-37", "source": "decoder.first.mlp_norm", "target": "add@decoder.first#4", "sourcePort": null, "targetPort": null }, { "id": "edge-38", "source": "add@decoder.first#4", "target": "decoder.first.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-39", "source": "add@decoder.first#4", "target": "decoder.first.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-40", "source": "add@decoder.first#4", "target": "decoder.first.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-41", "source": "add@decoder.first#4", "target": "decoder.first.image_to_token", "sourcePort": null, "targetPort": null }, { "id": "edge-42", "source": "decoder.first.mlp_norm", "target": "decoder.first.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-43", "source": "decoder.first.mlp_norm", "target": "decoder.first.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-44", "source": "decoder.first.mlp_norm", "target": "decoder.first.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-45", "source": "decoder.first.mlp_norm", "target": "decoder.first.image_to_token", "sourcePort": null, "targetPort": null }, { "id": "edge-46", "source": "decoder.first.image_to_token.project", "target": "add@decoder.first#5", "sourcePort": null, "targetPort": null }, { "id": "edge-47", "source": "decoder.first.image_to_token", "target": "add@decoder.first#5", "sourcePort": null, "targetPort": null }, { "id": "edge-48", "source": "add@decoder.first#5", "target": "decoder.first.image_norm", "sourcePort": null, "targetPort": null }, { "id": "edge-49", "source": "decoder.second.self_attention.query", "target": "div@decoder.second.self_attention#0", "sourcePort": null, "targetPort": null }, { "id": "edge-50", "source": "decoder.second.self_attention.key", "target": "div@decoder.second.self_attention#0", "sourcePort": null, "targetPort": null }, { "id": "edge-51", "source": "div@decoder.second.self_attention#0", "target": "decoder.second.self_attention.softmax", "sourcePort": null, "targetPort": null }, { "id": "edge-52", "source": "decoder.second.self_attention.softmax", "target": "decoder.second.self_attention.project", "sourcePort": null, "targetPort": null }, { "id": "edge-53", "source": "decoder.second.self_attention.value", "target": "decoder.second.self_attention.project", "sourcePort": null, "targetPort": null }, { "id": "edge-54", "source": "decoder.second.token_to_image.query", "target": "div@decoder.second.token_to_image#0", "sourcePort": null, "targetPort": null }, { "id": "edge-55", "source": "decoder.second.token_to_image.key", "target": "div@decoder.second.token_to_image#0", "sourcePort": null, "targetPort": null }, { "id": "edge-56", "source": "div@decoder.second.token_to_image#0", "target": "decoder.second.token_to_image.softmax", "sourcePort": null, "targetPort": null }, { "id": "edge-57", "source": "decoder.second.token_to_image.softmax", "target": "decoder.second.token_to_image.project", "sourcePort": null, "targetPort": null }, { "id": "edge-58", "source": "decoder.second.token_to_image.value", "target": "decoder.second.token_to_image.project", "sourcePort": null, "targetPort": null }, { "id": "edge-59", "source": "decoder.second.image_to_token.query", "target": "div@decoder.second.image_to_token#0", "sourcePort": null, "targetPort": null }, { "id": "edge-60", "source": "decoder.second.image_to_token.key", "target": "div@decoder.second.image_to_token#0", "sourcePort": null, "targetPort": null }, { "id": "edge-61", "source": "div@decoder.second.image_to_token#0", "target": "decoder.second.image_to_token.softmax", "sourcePort": null, "targetPort": null }, { "id": "edge-62", "source": "decoder.second.image_to_token.softmax", "target": "decoder.second.image_to_token.project", "sourcePort": null, "targetPort": null }, { "id": "edge-63", "source": "decoder.second.image_to_token.value", "target": "decoder.second.image_to_token.project", "sourcePort": null, "targetPort": null }, { "id": "edge-64", "source": "decoder.second.self_attention.project", "target": "add@decoder.second#0", "sourcePort": null, "targetPort": null }, { "id": "edge-65", "source": "decoder.second.self_attention", "target": "add@decoder.second#0", "sourcePort": null, "targetPort": null }, { "id": "edge-66", "source": "add@decoder.second#0", "target": "decoder.second.self_norm", "sourcePort": null, "targetPort": null }, { "id": "edge-67", "source": "decoder.second.self_norm", "target": "add@decoder.second#1", "sourcePort": null, "targetPort": null }, { "id": "edge-68", "source": "add@decoder.second#1", "target": "decoder.second.token_to_image.query", "sourcePort": null, "targetPort": null }, { "id": "edge-69", "source": "add@decoder.second#1", "target": "decoder.second.token_to_image.key", "sourcePort": null, "targetPort": null }, { "id": "edge-70", "source": "add@decoder.second#1", "target": "decoder.second.token_to_image.value", "sourcePort": null, "targetPort": null }, { "id": "edge-71", "source": "add@decoder.second#1", "target": "decoder.second.token_to_image", "sourcePort": null, "targetPort": null }, { "id": "edge-72", "source": "decoder.second.self_norm", "target": "add@decoder.second#2", "sourcePort": null, "targetPort": null }, { "id": "edge-73", "source": "decoder.second.token_to_image.project", "target": "add@decoder.second#2", "sourcePort": null, "targetPort": null }, { "id": "edge-74", "source": "decoder.second.token_to_image", "target": "add@decoder.second#2", "sourcePort": null, "targetPort": null }, { "id": "edge-75", "source": "add@decoder.second#2", "target": "decoder.second.token_norm", "sourcePort": null, "targetPort": null }, { "id": "edge-76", "source": "decoder.second.token_norm", "target": "decoder.second.up", "sourcePort": null, "targetPort": null }, { "id": "edge-77", "source": "decoder.second.up", "target": "decoder.second.relu", "sourcePort": null, "targetPort": null }, { "id": "edge-78", "source": "decoder.second.relu", "target": "decoder.second.down", "sourcePort": null, "targetPort": null }, { "id": "edge-79", "source": "decoder.second.token_norm", "target": "add@decoder.second#3", "sourcePort": null, "targetPort": null }, { "id": "edge-80", "source": "decoder.second.down", "target": "add@decoder.second#3", "sourcePort": null, "targetPort": null }, { "id": "edge-81", "source": "add@decoder.second#3", "target": "decoder.second.mlp_norm", "sourcePort": null, "targetPort": null }, { "id": "edge-82", "source": "decoder.second.mlp_norm", "target": "add@decoder.second#4", "sourcePort": null, "targetPort": null }, { "id": "edge-83", "source": "add@decoder.second#4", "target": "decoder.second.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-84", "source": "add@decoder.second#4", "target": "decoder.second.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-85", "source": "add@decoder.second#4", "target": "decoder.second.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-86", "source": "add@decoder.second#4", "target": "decoder.second.image_to_token", "sourcePort": null, "targetPort": null }, { "id": "edge-87", "source": "decoder.second.mlp_norm", "target": "decoder.second.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-88", "source": "decoder.second.mlp_norm", "target": "decoder.second.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-89", "source": "decoder.second.mlp_norm", "target": "decoder.second.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-90", "source": "decoder.second.mlp_norm", "target": "decoder.second.image_to_token", "sourcePort": null, "targetPort": null }, { "id": "edge-91", "source": "decoder.second.image_to_token.project", "target": "add@decoder.second#5", "sourcePort": null, "targetPort": null }, { "id": "edge-92", "source": "decoder.second.image_to_token", "target": "add@decoder.second#5", "sourcePort": null, "targetPort": null }, { "id": "edge-93", "source": "add@decoder.second#5", "target": "decoder.second.image_norm", "sourcePort": null, "targetPort": null }, { "id": "edge-94", "source": "decoder.final_attention.query", "target": "div@decoder.final_attention#0", "sourcePort": null, "targetPort": null }, { "id": "edge-95", "source": "decoder.final_attention.key", "target": "div@decoder.final_attention#0", "sourcePort": null, "targetPort": null }, { "id": "edge-96", "source": "div@decoder.final_attention#0", "target": "decoder.final_attention.softmax", "sourcePort": null, "targetPort": null }, { "id": "edge-97", "source": "decoder.final_attention.softmax", "target": "decoder.final_attention.project", "sourcePort": null, "targetPort": null }, { "id": "edge-98", "source": "decoder.final_attention.value", "target": "decoder.final_attention.project", "sourcePort": null, "targetPort": null }, { "id": "edge-99", "source": "decoder.first.mlp_norm", "target": "decoder.second.self_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-100", "source": "decoder.first.mlp_norm", "target": "decoder.second.self_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-101", "source": "decoder.first.mlp_norm", "target": "decoder.second.self_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-102", "source": "decoder.first.mlp_norm", "target": "add@decoder.second#0", "sourcePort": null, "targetPort": null }, { "id": "edge-103", "source": "decoder.first.mlp_norm", "target": "add@decoder.second#1", "sourcePort": null, "targetPort": null }, { "id": "edge-104", "source": "decoder.first.mlp_norm", "target": "decoder.second.token_to_image.query", "sourcePort": null, "targetPort": null }, { "id": "edge-105", "source": "decoder.first.mlp_norm", "target": "decoder.second.token_to_image.key", "sourcePort": null, "targetPort": null }, { "id": "edge-106", "source": "decoder.first.mlp_norm", "target": "decoder.second.token_to_image.value", "sourcePort": null, "targetPort": null }, { "id": "edge-107", "source": "decoder.first.mlp_norm", "target": "add@decoder.second#4", "sourcePort": null, "targetPort": null }, { "id": "edge-108", "source": "decoder.first.mlp_norm", "target": "decoder.second.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-109", "source": "decoder.first.mlp_norm", "target": "decoder.second.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-110", "source": "decoder.first.mlp_norm", "target": "decoder.second.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-111", "source": "decoder.first.mlp_norm", "target": "add@decoder.second#5", "sourcePort": null, "targetPort": null }, { "id": "edge-112", "source": "decoder.first.image_norm", "target": "decoder.second.self_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-113", "source": "decoder.first.image_norm", "target": "decoder.second.self_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-114", "source": "decoder.first.image_norm", "target": "decoder.second.self_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-115", "source": "decoder.first.image_norm", "target": "add@decoder.second#0", "sourcePort": null, "targetPort": null }, { "id": "edge-116", "source": "decoder.first.image_norm", "target": "add@decoder.second#1", "sourcePort": null, "targetPort": null }, { "id": "edge-117", "source": "decoder.first.image_norm", "target": "decoder.second.token_to_image.query", "sourcePort": null, "targetPort": null }, { "id": "edge-118", "source": "decoder.first.image_norm", "target": "decoder.second.token_to_image.key", "sourcePort": null, "targetPort": null }, { "id": "edge-119", "source": "decoder.first.image_norm", "target": "decoder.second.token_to_image.value", "sourcePort": null, "targetPort": null }, { "id": "edge-120", "source": "decoder.first.image_norm", "target": "add@decoder.second#4", "sourcePort": null, "targetPort": null }, { "id": "edge-121", "source": "decoder.first.image_norm", "target": "decoder.second.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-122", "source": "decoder.first.image_norm", "target": "decoder.second.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-123", "source": "decoder.first.image_norm", "target": "decoder.second.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-124", "source": "decoder.first.image_norm", "target": "add@decoder.second#5", "sourcePort": null, "targetPort": null }, { "id": "edge-125", "source": "decoder.first", "target": "decoder.second", "sourcePort": null, "targetPort": null }, { "id": "edge-126", "source": "decoder.second.mlp_norm", "target": "add@decoder#1", "sourcePort": null, "targetPort": null }, { "id": "edge-127", "source": "decoder.second.image_norm", "target": "add@decoder#1", "sourcePort": null, "targetPort": null }, { "id": "edge-128", "source": "decoder.second", "target": "add@decoder#1", "sourcePort": null, "targetPort": null }, { "id": "edge-129", "source": "decoder.second.mlp_norm", "target": "add@decoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-130", "source": "decoder.second.image_norm", "target": "add@decoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-131", "source": "decoder.second", "target": "add@decoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-132", "source": "add@decoder#1", "target": "decoder.final_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-133", "source": "add@decoder#1", "target": "decoder.final_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-134", "source": "add@decoder#1", "target": "decoder.final_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-135", "source": "add@decoder#1", "target": "decoder.final_attention", "sourcePort": null, "targetPort": null }, { "id": "edge-136", "source": "add@decoder#0", "target": "decoder.final_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-137", "source": "add@decoder#0", "target": "decoder.final_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-138", "source": "add@decoder#0", "target": "decoder.final_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-139", "source": "add@decoder#0", "target": "decoder.final_attention", "sourcePort": null, "targetPort": null }, { "id": "edge-140", "source": "decoder.second.mlp_norm", "target": "decoder.final_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-141", "source": "decoder.second.mlp_norm", "target": "decoder.final_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-142", "source": "decoder.second.mlp_norm", "target": "decoder.final_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-143", "source": "decoder.second.image_norm", "target": "decoder.final_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-144", "source": "decoder.second.image_norm", "target": "decoder.final_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-145", "source": "decoder.second.image_norm", "target": "decoder.final_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-146", "source": "decoder.second", "target": "decoder.final_attention", "sourcePort": null, "targetPort": null }, { "id": "edge-147", "source": "decoder.second.mlp_norm", "target": "add@decoder#2", "sourcePort": null, "targetPort": null }, { "id": "edge-148", "source": "decoder.second.image_norm", "target": "add@decoder#2", "sourcePort": null, "targetPort": null }, { "id": "edge-149", "source": "decoder.second", "target": "add@decoder#2", "sourcePort": null, "targetPort": null }, { "id": "edge-150", "source": "decoder.final_attention.project", "target": "add@decoder#2", "sourcePort": null, "targetPort": null }, { "id": "edge-151", "source": "decoder.final_attention", "target": "add@decoder#2", "sourcePort": null, "targetPort": null }, { "id": "edge-152", "source": "add@decoder#2", "target": "decoder.final_norm", "sourcePort": null, "targetPort": null }, { "id": "edge-153", "source": "upscale.first", "target": "permute@upscale#0", "sourcePort": null, "targetPort": null }, { "id": "edge-154", "source": "permute@upscale#0", "target": "upscale.norm", "sourcePort": null, "targetPort": null }, { "id": "edge-155", "source": "upscale.norm", "target": "permute@upscale#1", "sourcePort": null, "targetPort": null }, { "id": "edge-156", "source": "permute@upscale#1", "target": "gelu@upscale#0", "sourcePort": null, "targetPort": null }, { "id": "edge-157", "source": "gelu@upscale#0", "target": "upscale.second", "sourcePort": null, "targetPort": null }, { "id": "edge-158", "source": "upscale.second", "target": "gelu@upscale#1", "sourcePort": null, "targetPort": null }, { "id": "edge-159", "source": "masks.mask0.first", "target": "relu@masks.mask0#0", "sourcePort": null, "targetPort": null }, { "id": "edge-160", "source": "relu@masks.mask0#0", "target": "masks.mask0.second", "sourcePort": null, "targetPort": null }, { "id": "edge-161", "source": "masks.mask0.second", "target": "relu@masks.mask0#1", "sourcePort": null, "targetPort": null }, { "id": "edge-162", "source": "relu@masks.mask0#1", "target": "masks.mask0.project", "sourcePort": null, "targetPort": null }, { "id": "edge-163", "source": "masks.mask1.first", "target": "relu@masks.mask1#0", "sourcePort": null, "targetPort": null }, { "id": "edge-164", "source": "relu@masks.mask1#0", "target": "masks.mask1.second", "sourcePort": null, "targetPort": null }, { "id": "edge-165", "source": "masks.mask1.second", "target": "relu@masks.mask1#1", "sourcePort": null, "targetPort": null }, { "id": "edge-166", "source": "relu@masks.mask1#1", "target": "masks.mask1.project", "sourcePort": null, "targetPort": null }, { "id": "edge-167", "source": "masks.mask2.first", "target": "relu@masks.mask2#0", "sourcePort": null, "targetPort": null }, { "id": "edge-168", "source": "relu@masks.mask2#0", "target": "masks.mask2.second", "sourcePort": null, "targetPort": null }, { "id": "edge-169", "source": "masks.mask2.second", "target": "relu@masks.mask2#1", "sourcePort": null, "targetPort": null }, { "id": "edge-170", "source": "relu@masks.mask2#1", "target": "masks.mask2.project", "sourcePort": null, "targetPort": null }, { "id": "edge-171", "source": "masks.mask0.project", "target": "stack@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-172", "source": "masks.mask0", "target": "stack@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-173", "source": "masks.mask1.project", "target": "stack@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-174", "source": "masks.mask1", "target": "stack@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-175", "source": "masks.mask2.project", "target": "stack@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-176", "source": "masks.mask2", "target": "stack@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-177", "source": "stack@masks#0", "target": "python@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-178", "source": "flatten@masks#0", "target": "python@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-179", "source": "quality.first", "target": "relu@quality#0", "sourcePort": null, "targetPort": null }, { "id": "edge-180", "source": "relu@quality#0", "target": "quality.second", "sourcePort": null, "targetPort": null }, { "id": "edge-181", "source": "quality.second", "target": "relu@quality#1", "sourcePort": null, "targetPort": null }, { "id": "edge-182", "source": "relu@quality#1", "target": "quality.scores", "sourcePort": null, "targetPort": null }, { "id": "edge-183", "source": "input", "target": "cat@prompt_encoder.position#0", "sourcePort": null, "targetPort": null }, { "id": "edge-184", "source": "input", "target": "prompt_encoder.type_embedding", "sourcePort": null, "targetPort": null }, { "id": "edge-185", "source": "input", "target": "prompt_encoder", "sourcePort": null, "targetPort": null }, { "id": "edge-186", "source": "output_tokens", "target": "cat@PromptableMaskDecoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-187", "source": "input", "target": "cat@PromptableMaskDecoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-188", "source": "add@prompt_encoder#0", "target": "cat@PromptableMaskDecoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-189", "source": "prompt_encoder", "target": "cat@PromptableMaskDecoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-190", "source": "input", "target": "flatten@PromptableMaskDecoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-191", "source": "flatten@PromptableMaskDecoder#0", "target": "add@PromptableMaskDecoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-192", "source": "no_mask", "target": "add@PromptableMaskDecoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-193", "source": "add@prompt_encoder#0", "target": "python@PromptableMaskDecoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-194", "source": "prompt_encoder", "target": "python@PromptableMaskDecoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-195", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.first.self_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-196", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.first.self_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-197", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.first.self_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-198", "source": "cat@PromptableMaskDecoder#0", "target": "add@decoder.first#0", "sourcePort": null, "targetPort": null }, { "id": "edge-199", "source": "cat@PromptableMaskDecoder#0", "target": "add@decoder.first#1", "sourcePort": null, "targetPort": null }, { "id": "edge-200", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.first.token_to_image.query", "sourcePort": null, "targetPort": null }, { "id": "edge-201", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.first.token_to_image.key", "sourcePort": null, "targetPort": null }, { "id": "edge-202", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.first.token_to_image.value", "sourcePort": null, "targetPort": null }, { "id": "edge-203", "source": "cat@PromptableMaskDecoder#0", "target": "add@decoder.first#4", "sourcePort": null, "targetPort": null }, { "id": "edge-204", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.first.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-205", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.first.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-206", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.first.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-207", "source": "cat@PromptableMaskDecoder#0", "target": "add@decoder.first#5", "sourcePort": null, "targetPort": null }, { "id": "edge-208", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.second.self_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-209", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.second.self_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-210", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.second.self_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-211", "source": "cat@PromptableMaskDecoder#0", "target": "add@decoder.second#0", "sourcePort": null, "targetPort": null }, { "id": "edge-212", "source": "cat@PromptableMaskDecoder#0", "target": "add@decoder.second#1", "sourcePort": null, "targetPort": null }, { "id": "edge-213", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.second.token_to_image.query", "sourcePort": null, "targetPort": null }, { "id": "edge-214", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.second.token_to_image.key", "sourcePort": null, "targetPort": null }, { "id": "edge-215", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.second.token_to_image.value", "sourcePort": null, "targetPort": null }, { "id": "edge-216", "source": "cat@PromptableMaskDecoder#0", "target": "add@decoder.second#4", "sourcePort": null, "targetPort": null }, { "id": "edge-217", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.second.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-218", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.second.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-219", "source": "cat@PromptableMaskDecoder#0", "target": "decoder.second.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-220", "source": "cat@PromptableMaskDecoder#0", "target": "add@decoder.second#5", "sourcePort": null, "targetPort": null }, { "id": "edge-221", "source": "cat@PromptableMaskDecoder#0", "target": "add@decoder#1", "sourcePort": null, "targetPort": null }, { "id": "edge-222", "source": "cat@PromptableMaskDecoder#0", "target": "add@decoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-223", "source": "cat@PromptableMaskDecoder#0", "target": "decoder", "sourcePort": null, "targetPort": null }, { "id": "edge-224", "source": "add@PromptableMaskDecoder#0", "target": "decoder.first.self_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-225", "source": "add@PromptableMaskDecoder#0", "target": "decoder.first.self_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-226", "source": "add@PromptableMaskDecoder#0", "target": "decoder.first.self_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-227", "source": "add@PromptableMaskDecoder#0", "target": "add@decoder.first#0", "sourcePort": null, "targetPort": null }, { "id": "edge-228", "source": "add@PromptableMaskDecoder#0", "target": "add@decoder.first#1", "sourcePort": null, "targetPort": null }, { "id": "edge-229", "source": "add@PromptableMaskDecoder#0", "target": "decoder.first.token_to_image.query", "sourcePort": null, "targetPort": null }, { "id": "edge-230", "source": "add@PromptableMaskDecoder#0", "target": "decoder.first.token_to_image.key", "sourcePort": null, "targetPort": null }, { "id": "edge-231", "source": "add@PromptableMaskDecoder#0", "target": "decoder.first.token_to_image.value", "sourcePort": null, "targetPort": null }, { "id": "edge-232", "source": "add@PromptableMaskDecoder#0", "target": "add@decoder.first#4", "sourcePort": null, "targetPort": null }, { "id": "edge-233", "source": "add@PromptableMaskDecoder#0", "target": "decoder.first.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-234", "source": "add@PromptableMaskDecoder#0", "target": "decoder.first.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-235", "source": "add@PromptableMaskDecoder#0", "target": "decoder.first.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-236", "source": "add@PromptableMaskDecoder#0", "target": "add@decoder.first#5", "sourcePort": null, "targetPort": null }, { "id": "edge-237", "source": "add@PromptableMaskDecoder#0", "target": "decoder.second.self_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-238", "source": "add@PromptableMaskDecoder#0", "target": "decoder.second.self_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-239", "source": "add@PromptableMaskDecoder#0", "target": "decoder.second.self_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-240", "source": "add@PromptableMaskDecoder#0", "target": "add@decoder.second#0", "sourcePort": null, "targetPort": null }, { "id": "edge-241", "source": "add@PromptableMaskDecoder#0", "target": "add@decoder.second#1", "sourcePort": null, "targetPort": null }, { "id": "edge-242", "source": "add@PromptableMaskDecoder#0", "target": "decoder.second.token_to_image.query", "sourcePort": null, "targetPort": null }, { "id": "edge-243", "source": "add@PromptableMaskDecoder#0", "target": "decoder.second.token_to_image.key", "sourcePort": null, "targetPort": null }, { "id": "edge-244", "source": "add@PromptableMaskDecoder#0", "target": "decoder.second.token_to_image.value", "sourcePort": null, "targetPort": null }, { "id": "edge-245", "source": "add@PromptableMaskDecoder#0", "target": "add@decoder.second#4", "sourcePort": null, "targetPort": null }, { "id": "edge-246", "source": "add@PromptableMaskDecoder#0", "target": "decoder.second.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-247", "source": "add@PromptableMaskDecoder#0", "target": "decoder.second.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-248", "source": "add@PromptableMaskDecoder#0", "target": "decoder.second.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-249", "source": "add@PromptableMaskDecoder#0", "target": "add@decoder.second#5", "sourcePort": null, "targetPort": null }, { "id": "edge-250", "source": "add@PromptableMaskDecoder#0", "target": "add@decoder#1", "sourcePort": null, "targetPort": null }, { "id": "edge-251", "source": "add@PromptableMaskDecoder#0", "target": "add@decoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-252", "source": "add@PromptableMaskDecoder#0", "target": "decoder", "sourcePort": null, "targetPort": null }, { "id": "edge-253", "source": "python@PromptableMaskDecoder#0", "target": "decoder.first.self_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-254", "source": "python@PromptableMaskDecoder#0", "target": "decoder.first.self_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-255", "source": "python@PromptableMaskDecoder#0", "target": "decoder.first.self_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-256", "source": "python@PromptableMaskDecoder#0", "target": "add@decoder.first#0", "sourcePort": null, "targetPort": null }, { "id": "edge-257", "source": "python@PromptableMaskDecoder#0", "target": "add@decoder.first#1", "sourcePort": null, "targetPort": null }, { "id": "edge-258", "source": "python@PromptableMaskDecoder#0", "target": "decoder.first.token_to_image.query", "sourcePort": null, "targetPort": null }, { "id": "edge-259", "source": "python@PromptableMaskDecoder#0", "target": "decoder.first.token_to_image.key", "sourcePort": null, "targetPort": null }, { "id": "edge-260", "source": "python@PromptableMaskDecoder#0", "target": "decoder.first.token_to_image.value", "sourcePort": null, "targetPort": null }, { "id": "edge-261", "source": "python@PromptableMaskDecoder#0", "target": "add@decoder.first#4", "sourcePort": null, "targetPort": null }, { "id": "edge-262", "source": "python@PromptableMaskDecoder#0", "target": "decoder.first.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-263", "source": "python@PromptableMaskDecoder#0", "target": "decoder.first.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-264", "source": "python@PromptableMaskDecoder#0", "target": "decoder.first.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-265", "source": "python@PromptableMaskDecoder#0", "target": "add@decoder.first#5", "sourcePort": null, "targetPort": null }, { "id": "edge-266", "source": "python@PromptableMaskDecoder#0", "target": "decoder.second.self_attention.query", "sourcePort": null, "targetPort": null }, { "id": "edge-267", "source": "python@PromptableMaskDecoder#0", "target": "decoder.second.self_attention.key", "sourcePort": null, "targetPort": null }, { "id": "edge-268", "source": "python@PromptableMaskDecoder#0", "target": "decoder.second.self_attention.value", "sourcePort": null, "targetPort": null }, { "id": "edge-269", "source": "python@PromptableMaskDecoder#0", "target": "add@decoder.second#0", "sourcePort": null, "targetPort": null }, { "id": "edge-270", "source": "python@PromptableMaskDecoder#0", "target": "add@decoder.second#1", "sourcePort": null, "targetPort": null }, { "id": "edge-271", "source": "python@PromptableMaskDecoder#0", "target": "decoder.second.token_to_image.query", "sourcePort": null, "targetPort": null }, { "id": "edge-272", "source": "python@PromptableMaskDecoder#0", "target": "decoder.second.token_to_image.key", "sourcePort": null, "targetPort": null }, { "id": "edge-273", "source": "python@PromptableMaskDecoder#0", "target": "decoder.second.token_to_image.value", "sourcePort": null, "targetPort": null }, { "id": "edge-274", "source": "python@PromptableMaskDecoder#0", "target": "add@decoder.second#4", "sourcePort": null, "targetPort": null }, { "id": "edge-275", "source": "python@PromptableMaskDecoder#0", "target": "decoder.second.image_to_token.query", "sourcePort": null, "targetPort": null }, { "id": "edge-276", "source": "python@PromptableMaskDecoder#0", "target": "decoder.second.image_to_token.key", "sourcePort": null, "targetPort": null }, { "id": "edge-277", "source": "python@PromptableMaskDecoder#0", "target": "decoder.second.image_to_token.value", "sourcePort": null, "targetPort": null }, { "id": "edge-278", "source": "python@PromptableMaskDecoder#0", "target": "add@decoder.second#5", "sourcePort": null, "targetPort": null }, { "id": "edge-279", "source": "python@PromptableMaskDecoder#0", "target": "add@decoder#1", "sourcePort": null, "targetPort": null }, { "id": "edge-280", "source": "python@PromptableMaskDecoder#0", "target": "add@decoder#0", "sourcePort": null, "targetPort": null }, { "id": "edge-281", "source": "python@PromptableMaskDecoder#0", "target": "decoder", "sourcePort": null, "targetPort": null }, { "id": "edge-282", "source": "decoder.final_norm", "target": "upscale.first", "sourcePort": null, "targetPort": null }, { "id": "edge-283", "source": "decoder.second.mlp_norm", "target": "upscale.first", "sourcePort": null, "targetPort": null }, { "id": "edge-284", "source": "decoder.second.image_norm", "target": "upscale.first", "sourcePort": null, "targetPort": null }, { "id": "edge-285", "source": "decoder", "target": "upscale", "sourcePort": null, "targetPort": null }, { "id": "edge-286", "source": "decoder.final_norm", "target": "masks.mask0.first", "sourcePort": null, "targetPort": null }, { "id": "edge-287", "source": "decoder.final_norm", "target": "masks.mask1.first", "sourcePort": null, "targetPort": null }, { "id": "edge-288", "source": "decoder.final_norm", "target": "masks.mask2.first", "sourcePort": null, "targetPort": null }, { "id": "edge-289", "source": "decoder.final_norm", "target": "flatten@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-290", "source": "decoder.second.mlp_norm", "target": "masks.mask0.first", "sourcePort": null, "targetPort": null }, { "id": "edge-291", "source": "decoder.second.mlp_norm", "target": "masks.mask1.first", "sourcePort": null, "targetPort": null }, { "id": "edge-292", "source": "decoder.second.mlp_norm", "target": "masks.mask2.first", "sourcePort": null, "targetPort": null }, { "id": "edge-293", "source": "decoder.second.mlp_norm", "target": "flatten@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-294", "source": "decoder.second.image_norm", "target": "masks.mask0.first", "sourcePort": null, "targetPort": null }, { "id": "edge-295", "source": "decoder.second.image_norm", "target": "masks.mask1.first", "sourcePort": null, "targetPort": null }, { "id": "edge-296", "source": "decoder.second.image_norm", "target": "masks.mask2.first", "sourcePort": null, "targetPort": null }, { "id": "edge-297", "source": "decoder.second.image_norm", "target": "flatten@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-298", "source": "decoder", "target": "masks", "sourcePort": null, "targetPort": null }, { "id": "edge-299", "source": "gelu@upscale#1", "target": "masks.mask0.first", "sourcePort": null, "targetPort": null }, { "id": "edge-300", "source": "gelu@upscale#1", "target": "masks.mask1.first", "sourcePort": null, "targetPort": null }, { "id": "edge-301", "source": "gelu@upscale#1", "target": "masks.mask2.first", "sourcePort": null, "targetPort": null }, { "id": "edge-302", "source": "gelu@upscale#1", "target": "flatten@masks#0", "sourcePort": null, "targetPort": null }, { "id": "edge-303", "source": "upscale", "target": "masks", "sourcePort": null, "targetPort": null }, { "id": "edge-304", "source": "decoder.final_norm", "target": "quality.first", "sourcePort": null, "targetPort": null }, { "id": "edge-305", "source": "decoder.second.mlp_norm", "target": "quality.first", "sourcePort": null, "targetPort": null }, { "id": "edge-306", "source": "decoder.second.image_norm", "target": "quality.first", "sourcePort": null, "targetPort": null }, { "id": "edge-307", "source": "decoder", "target": "quality", "sourcePort": null, "targetPort": null }, { "id": "edge-308", "source": "python@masks#0", "target": "flatten@PromptableMaskDecoder#1", "sourcePort": null, "targetPort": null }, { "id": "edge-309", "source": "masks", "target": "flatten@PromptableMaskDecoder#1", "sourcePort": null, "targetPort": null }, { "id": "edge-310", "source": "flatten@PromptableMaskDecoder#1", "target": "cat@PromptableMaskDecoder#1", "sourcePort": null, "targetPort": null }, { "id": "edge-311", "source": "quality.scores", "target": "cat@PromptableMaskDecoder#1", "sourcePort": null, "targetPort": null }, { "id": "edge-312", "source": "quality", "target": "cat@PromptableMaskDecoder#1", "sourcePort": null, "targetPort": null }, { "id": "edge-313", "source": "cat@PromptableMaskDecoder#1", "target": "output", "sourcePort": null, "targetPort": null } ], "steps": [ { "id": "prompt", "target": "prompt_encoder.type_embedding", "title": "Combine prompt geometry with its meaning", "note": "Random Fourier features encode normalized (x,y) coordinates. A learned type embedding distinguishes foreground/background points and the two box corners. The same image features are reused when a prompt changes." }, { "id": "tokens", "target": "decoder.first.token_to_image", "title": "Let prompt and output tokens read image features", "note": "Three mask tokens, one quality token and the sparse prompts pass through self-attention, token-to-image attention and an MLP. Prompt and image positions are added to attention queries and keys." }, { "id": "image", "target": "decoder.first.image_to_token", "title": "Update image features with prompt information", "note": "Attention also runs in the reverse direction: image locations query the updated tokens. A second two-way block repeats this exchange, followed by a final token-to-image read. Both representation streams can now depend on the prompt." }, { "id": "masks", "target": "masks.mask0.project", "title": "Use mask tokens as dynamic pixel classifiers", "note": "Each mask token passes through its own MLP to produce two feature weights. A dot product with every upscaled image location creates one candidate mask logit. A separate quality head scores the three candidates; none of these random outputs is a trained segmentation result." } ], "milestone": { "name": "Segment Anything · promptable decoder", "year": 2023, "kind": "Component", "fidelity": "Width eight, two two-way blocks with one attention head, three mask tokens and one quality token. Random Fourier coordinate features plus point/box type embeddings. Two transposed convolutions upscale a 2 × 2 image embedding to 8 × 8 mask features.", "papers": [ { "title": "Segment Anything", "url": "https://arxiv.org/abs/2304.02643" } ] }, "experiment": { "kind": "matrices", "title": "Reuse the image; change the prompt.", "description": "Recorded prompt-to-mask computation from cached synthetic image features. Explore point location, prompt type and a box while the two-way decoder updates image and token features.", "controlLabel": "Segmentation prompt", "cases": [ { "id": "points", "label": "Foreground + background points", "target": "prompt_encoder.type_embedding", "note": "The same cached 2 × 2 image embedding is reused for each prompt. Coordinates are normalized (x,y) in [0,1]. Type IDs are 0 = background, 1 = foreground, 2 = top-left box corner, 3 = bottom-right. Three mask logits express the multi-mask interface; these untrained outputs do not segment a real object and quality scores are not calibrated IoUs.", "matrices": [ { "label": "Candidate 1 · mask logits", "values": [ [ 0.08098679913165055, -0.005838083167316226, 0.07063712119657557, -0.001284980490123819, -0.002424238874847102, 0.045513075403544344, 0.013761522381180082, 0.061024551322869186 ], [ 0.12263730365794653, 0.025385730542043423, 0.08357136836491912, 0.048107847491274086, 0.07269710629562388, 0.01616772678650197, 0.08023548179568549, 0.003486905182956408 ], [ 0.016226885362688806, 0.053655445490188886, 0.014849274625457091, 0.003523421208497987, -0.012129199247292053, 0.004863235614241405, 0.10393273494912302, -0.006139820352728196 ], [ 0.05469153696274688, 0.02535431972038923, 0.07575751190725871, 0.04806029741108182, 0.059135887138483306, 0.06118248975152672, 0.1300542064962185, 0.02120816042164226 ], [ 0.05908956120080295, -0.0005823370223078333, 0.0705233751785966, -0.005872781173809721, 0.032046115086304124, -0.0017034766316091686, 0.024371584924116358, 0.06231566533368841 ], [ 0.10098998937299028, 0.033573374898908104, 0.11827048562645487, 0.0282156693015281, 0.09264192628940288, 0.04094273687468716, 0.08509010768635909, 0.0003713153423974925 ], [ 0.022378646707228035, 0.014843810415984423, 0.006470452620553114, 0.008071856650141292, 0.01645807187860608, 0.03088117444393799, 0.010361673624542286, -0.000546743078004723 ], [ 0.021479614013632735, 0.09486705974432492, 0.0568970344226771, 0.059859682859569094, 0.03509149991112538, 0.06093580159317687, 0.07947882783221216, 0.04989987856687473 ] ] }, { "label": "Candidate 2 · mask logits", "values": [ [ -0.10136377340241084, -0.016230393487613095, -0.09132051960733184, -0.016585398459828996, 0.0004579814342304774, -0.0727743330618845, -0.02307370500007136, -0.08952769670064349 ], [ -0.18301312531515235, -0.039764320605063144, -0.11878802413218716, -0.0635520737644157, -0.10861015638133091, -0.020198591463255134, -0.11567833923381561, -0.0019320548676481186 ], [ -0.029134260081880307, -0.07801888543959741, -0.01701542871834716, -0.025607946719521366, 0.018515842771617174, -0.02848586279366714, -0.13139494965045806, -0.015014360228033844 ], [ -0.07398683668777471, -0.02796274190234071, -0.11720737539271134, -0.06851040176014345, -0.09769287049647406, -0.08789784831841334, -0.19184588483750245, -0.034016280691784875 ], [ -0.07423393606214145, -0.020398189303758005, -0.08769383816137317, -0.016548303479177674, -0.03836475331643968, -0.020823065166345963, -0.037409608053616976, -0.0905621508433296 ], [ -0.15034553290998312, -0.048862263565368526, -0.17744686255299236, -0.043515867139630945, -0.14184583549797472, -0.059837741222607724, -0.12104840610817028, 0.0027251144797966953 ], [ -0.03600663590076566, -0.0292004526624181, -0.007867596695478967, -0.028681191786969495, -0.028431383857466446, -0.049523843073389376, -0.010150840734791472, -0.022667794778545046 ], [ -0.02524788501660664, -0.11949802743112692, -0.08716371540609857, -0.08183896951875215, -0.045856389008649634, -0.0753785251123925, -0.12491957469491075, -0.07218465008236381 ] ] }, { "label": "Candidate 3 · mask logits", "values": [ [ 0.007400319250212546, 0.1274781367503821, 0.02228374359862451, 0.09883168840609666, 0.013789620257577032, 0.09014267482379607, 0.033071763449396455, 0.07708818406809399 ], [ 0.1717509124262336, 0.04578166755556891, 0.08480756029587276, 0.022560372686194954, 0.10248180748600598, 0.0012757222305557342, 0.09029783252556653, -0.012909182205040016 ], [ 0.049476400156321265, 0.06398437679760136, -0.007182076472487078, 0.11561046120861615, -0.01924531438468727, 0.12226466823734282, 0.016631716561290297, 0.12289092966781696 ], [ 0.0350974152692115, -0.018192035477106405, 0.12868635189538793, 0.04984631122600079, 0.13417937174588462, 0.06716334576995642, 0.16997962313202003, 0.04257533225929281 ], [ 0.006905563910633368, 0.1148493897473243, 0.0033233772043988106, 0.12944015853958338, -0.006559121824925758, 0.12468935919696963, 0.039785892557443324, 0.0740435045209202 ], [ 0.13945984111401036, 0.04027656199609883, 0.17080412836006983, 0.047180040975426066, 0.1492959158011781, 0.050478034640895925, 0.08690161283259046, -0.017314578321167208 ], [ 0.04553956101827288, 0.05912469180016636, -0.0006644114795915898, 0.10177891122712752, 0.04410112740697815, 0.061954626753514916, -0.014378855962462158, 0.1269539403942008 ], [ -0.006935893421326083, 0.012810732592496002, 0.09195041535040878, 0.04309478085164013, 0.013733417448822147, 0.0007318165001537961, 0.14563912005609775, 0.05747521485744992 ] ] } ], "vectors": [ { "label": "Prompt coordinates · x₀, y₀, x₁, y₁", "values": [ 0.25, 0.25, 0.75, 0.75 ] }, { "label": "Prompt type IDs", "values": [ 1, 0 ] }, { "label": "Predicted quality scores · untrained", "values": [ -0.09329061722099478, -0.09946665440873959, 0.12908435977633292 ] } ], "metrics": [ { "label": "Image encoder calls per changed prompt", "value": 0 }, { "label": "Candidate masks", "value": 3 } ] }, { "id": "move", "label": "Move the foreground point", "target": "prompt_encoder.type_embedding", "note": "The same cached 2 × 2 image embedding is reused for each prompt. Coordinates are normalized (x,y) in [0,1]. Type IDs are 0 = background, 1 = foreground, 2 = top-left box corner, 3 = bottom-right. Three mask logits express the multi-mask interface; these untrained outputs do not segment a real object and quality scores are not calibrated IoUs.", "matrices": [ { "label": "Candidate 1 · mask logits", "values": [ [ 0.06275956807362652, -0.04110915987107775, 0.05118716104475162, -0.02804427084215687, -0.007094659055198218, 0.006842106021733423, 0.004029267444918, 0.03458063939833618 ], [ 0.053340149785418116, 0.009192888390991938, 0.04536229729858209, 0.03581137485176314, 0.02911311029544304, 0.015906750648802998, 0.04492647908958831, 0.0044829440543716305 ], [ 0.0005783528439077116, 0.03135752473499308, 0.01072438453438998, -0.028865307511446083, -0.00453056608923407, -0.03129297112111905, 0.08278539329382108, -0.039721566829266794 ], [ 0.039152789819226284, 0.019185109804575183, 0.024945899449488725, 0.028132947021031116, 0.011959900114848937, 0.032559507942382965, 0.06146477250396872, 0.006229448037790837 ], [ 0.0438396860931971, -0.03227882360981866, 0.07319532842819024, -0.04024327129079577, 0.02743770934177828, -0.037337993415876146, 0.010620583676537561, 0.028196594311990564 ], [ 0.04367193566384373, 0.01825030598522454, 0.05767856623388725, 0.00706943949388513, 0.035015014707809815, 0.02070661231281276, 0.046534129436754036, 0.006716148587596592 ], [ 0.00875769241695974, -0.005170932781378592, 0.0028194472220165285, -0.02063836560848873, 0.0027477463815665038, 0.029595686139544067, 0.006599296461666282, -0.034862427388747536 ], [ 0.021483481682018673, 0.07426780933622673, 0.0202154183381342, 0.0414135418485978, 0.03969198420223469, 0.01759810109618791, 0.021320392922641093, 0.028274150003229352 ] ] }, { "label": "Candidate 2 · mask logits", "values": [ [ -0.09863305612953964, -0.008760078766553657, -0.0906230529766489, -0.01130769206868811, 0.006177887019798434, -0.06285362023006245, -0.02721721752541797, -0.09419535214346195 ], [ -0.17649714569526412, -0.03971575774183906, -0.11356835613033514, -0.06657471807508902, -0.10333451534960757, -0.028482310328341445, -0.11791671646699316, 0.002615887267391878 ], [ -0.027653776876258734, -0.08440477788613696, -0.011714723220817503, -0.019986496672710838, 0.017888054101809, -0.020994767039786668, -0.13536178075464467, -0.007618529613666105 ], [ -0.08603762675740824, -0.0181115882535415, -0.10785113903833725, -0.07120095242692757, -0.09363508949204884, -0.08723052829753858, -0.18715881856580943, -0.03285860821821759 ], [ -0.06985954148270117, -0.014395307010852084, -0.11772173803675462, -0.008368892525308397, -0.03741328080691993, -0.013233011727971708, -0.037646170283302936, -0.08432194369880024 ], [ -0.14236084633610685, -0.05034754978828503, -0.18291295811126793, -0.03545431963426384, -0.13687922020803311, -0.06015840969641389, -0.11842203214377763, -0.0016244741953217027 ], [ -0.03838092192672614, -0.02563074019566744, -0.004376259501054832, -0.023981047684422524, -0.029580260575417128, -0.08247295170497485, -0.000944737248319212, -0.016908597480091176 ], [ -0.030260291621270627, -0.11988657162200535, -0.07787955792349618, -0.0865268685096622, -0.09027395163865755, -0.01701752374495276, -0.11119976255520041, -0.07654887312637564 ] ] }, { "label": "Candidate 3 · mask logits", "values": [ [ -0.0020540089286198385, 0.1371667820703113, 0.01716531817209776, 0.10344407825435202, 0.009436789808526035, 0.09622730162668768, 0.03853117251460447, 0.0726375648070019 ], [ 0.16980543294933279, 0.04647692246432969, 0.07678052031914817, 0.017883803480176402, 0.10564279341080839, 0.005927736986812969, 0.08611267557425328, -0.01803219828385537 ], [ 0.049492492124700284, 0.06399566365541372, -0.00986594148066136, 0.12192634912210601, -0.019785643870893578, 0.1309354828421551, 0.0070209444113642225, 0.13097096846054573 ], [ 0.044083570693908475, -0.0229165315004913, 0.1262647957802113, 0.049039280369973154, 0.1381541826290275, 0.06569041319934157, 0.16563894220053582, 0.04250152543772239 ], [ 0.00034427220030480303, 0.12161875655297717, 0.0025802803601167094, 0.13389501630358297, -0.01146470032018153, 0.1343519360525216, 0.03844508931483402, 0.07314227568739648 ], [ 0.13505597533437683, 0.03951068423184216, 0.16891840607639164, 0.04483545211061751, 0.15037814166179098, 0.05044621104452534, 0.08231817955709923, -0.01675266460206592 ], [ 0.04528622118241788, 0.06266265301987212, -0.00019370388932966854, 0.10511629925728325, 0.046676198029888635, 0.06560297615207956, -0.017667231196113752, 0.1338728618325349 ], [ -0.007188400085860083, 0.0034325229061752693, 0.08469768522544155, 0.03833778900081292, 0.0503396783533091, -0.020272688662244635, 0.14313075591608243, 0.05852418540116836 ] ] } ], "vectors": [ { "label": "Prompt coordinates · x₀, y₀, x₁, y₁", "values": [ 0.6, 0.25, 0.75, 0.75 ] }, { "label": "Prompt type IDs", "values": [ 1, 0 ] }, { "label": "Predicted quality scores · untrained", "values": [ -0.08266494259276634, -0.10268568305588599, 0.13795196121355457 ] } ], "metrics": [ { "label": "Image encoder calls per changed prompt", "value": 0 }, { "label": "Candidate masks", "value": 3 } ] }, { "id": "type", "label": "Mark both points as background", "target": "prompt_encoder.type_embedding", "note": "The same cached 2 × 2 image embedding is reused for each prompt. Coordinates are normalized (x,y) in [0,1]. Type IDs are 0 = background, 1 = foreground, 2 = top-left box corner, 3 = bottom-right. Three mask logits express the multi-mask interface; these untrained outputs do not segment a real object and quality scores are not calibrated IoUs.", "matrices": [ { "label": "Candidate 1 · mask logits", "values": [ [ 0.08353360713684845, 0.002009776370770226, 0.07416776088861363, 0.004341380263261048, -0.0006408804594445342, 0.05380497764378969, 0.017331202129137448, 0.06276119387855689 ], [ 0.13527868538649154, 0.028333161318600905, 0.09100613746334744, 0.0499588013444328, 0.08140616718402939, 0.014043384139066861, 0.08695663869763076, 0.004649399752513548 ], [ 0.019696810435250626, 0.0619819688045628, 0.01662402523187531, 0.010413392277600315, -0.014233791885862131, 0.012452580262354681, 0.10474135811196897, 0.0017379980270065952 ], [ 0.06255310494231785, 0.01859728661013372, 0.086856556241987, 0.05054547025688176, 0.06763362855117874, 0.06691321003473673, 0.13971937972591858, 0.024965453637481032 ], [ 0.06030125991582091, 0.006827712228792327, 0.06740593189320432, 0.0026406600282712148, 0.031671652843288184, 0.006240614056200144, 0.027426533824473777, 0.06637124877135861 ], [ 0.1094574874470131, 0.03695692664278655, 0.1286107255601403, 0.03201495136485392, 0.10265553196233969, 0.04463080201357291, 0.09086537047035086, 0.000215479158832993 ], [ 0.02676925285053413, 0.018599454821777335, 0.007403862993984203, 0.014576487497748087, 0.021234666535105782, 0.04907956219883077, 0.010332081536093598, 0.007170832259927013 ], [ 0.02088977979017917, 0.09727934223507988, 0.0629672373565176, 0.06331069198061365, 0.052869295577320875, 0.03464868379443908, 0.0904401323125698, 0.05368548960032947 ] ] }, { "label": "Candidate 2 · mask logits", "values": [ [ -0.10875142164462051, -0.015591079049131045, -0.0981086622201758, -0.01579264458771996, -0.0007344933422587283, -0.07864137399030821, -0.025596777239199134, -0.08907033201873735 ], [ -0.1923674911438589, -0.0412821765375512, -0.12638678506976586, -0.0668713572702367, -0.11565906824154781, -0.018041075743027057, -0.12171162580451741, -0.00493626287790927 ], [ -0.030436110270799448, -0.08674678069432673, -0.020772019500617057, -0.02532455501802042, 0.020407110257083045, -0.028607728175127824, -0.1370933266539468, -0.014640232372831638 ], [ -0.08514884514421707, -0.022041684352267953, -0.12560664913855665, -0.07038824950287342, -0.10112520538525142, -0.09342698316841799, -0.19746082221090164, -0.03647119491491396 ], [ -0.07871252001733561, -0.02040793272467842, -0.08732582284065904, -0.016613974519523626, -0.04026757136182598, -0.020749383655475157, -0.03946956629789078, -0.09333444722555409 ], [ -0.15537361154656348, -0.05178361014708139, -0.18358169958817178, -0.04620064838673389, -0.1478436447467554, -0.06281941773967387, -0.1262221734967201, 0.0014259026678700183 ], [ -0.03937141510858201, -0.029988403497639142, -0.009586345214415431, -0.029134403254071725, -0.03211890189361824, -0.06998145168380117, -0.011886254599567065, -0.022267464963127877 ], [ -0.026153918522764184, -0.12697560077110084, -0.0905686645666248, -0.08612691120989595, -0.07162531435173451, -0.04369896437552284, -0.13183972424314835, -0.0752352182795855 ] ] }, { "label": "Candidate 3 · mask logits", "values": [ [ 0.019030928683115567, 0.11617792655653998, 0.030726285369428768, 0.09143340858789778, 0.01384652649508423, 0.08890161934631564, 0.031004101753907436, 0.07996247598932367 ], [ 0.17575106382602676, 0.04565896172039217, 0.09125544933175071, 0.02770871955160486, 0.10485826974598574, 0.001042437520006251, 0.09565604625432039, -0.008900910035301475 ], [ 0.047236671448256135, 0.06810921821181004, -0.003977556001038703, 0.10732642450373367, -0.01997752182184067, 0.11339549681432126, 0.03038931940914188, 0.11079118130254553 ], [ 0.04735561109194883, -0.015116447548374325, 0.13153501374023233, 0.05239736976393533, 0.1320133912569233, 0.07155408244657582, 0.1706268825012274, 0.04108685606023785 ], [ 0.01558444414129786, 0.10429344311829163, 0.011528920836737701, 0.11811932700050153, -0.00139467220826959, 0.11402218842341272, 0.03981309123690709, 0.07689827985511793 ], [ 0.13974392639427308, 0.04115074834592658, 0.17329676260774343, 0.04761435193317942, 0.15001567500330226, 0.052221870928225916, 0.09138973015986844, -0.015170556044857987 ], [ 0.04641997127875655, 0.055735547584728984, 0.0012171275889413932, 0.0939149826527175, 0.04473897469571271, 0.0654566025927554, -0.011603972753931028, 0.11697260991305986 ], [ -0.004536541843376932, 0.02509480998513487, 0.09098054607130661, 0.047454825479535154, 0.036977004651970485, -0.004679751622789233, 0.14633337047418243, 0.05988237997912065 ] ] } ], "vectors": [ { "label": "Prompt coordinates · x₀, y₀, x₁, y₁", "values": [ 0.25, 0.25, 0.75, 0.75 ] }, { "label": "Prompt type IDs", "values": [ 0, 0 ] }, { "label": "Predicted quality scores · untrained", "values": [ -0.08693018375392456, -0.10139353606615217, 0.13439242640996635 ] } ], "metrics": [ { "label": "Image encoder calls per changed prompt", "value": 0 }, { "label": "Candidate masks", "value": 3 } ] }, { "id": "box", "label": "Two box corners", "target": "prompt_encoder.type_embedding", "note": "The same cached 2 × 2 image embedding is reused for each prompt. Coordinates are normalized (x,y) in [0,1]. Type IDs are 0 = background, 1 = foreground, 2 = top-left box corner, 3 = bottom-right. Three mask logits express the multi-mask interface; these untrained outputs do not segment a real object and quality scores are not calibrated IoUs.", "matrices": [ { "label": "Candidate 1 · mask logits", "values": [ [ 0.08015658486588924, 0.018317512122278, 0.0717711924666817, 0.018108138615807165, 0.005498051865254593, 0.06943287194040541, 0.011417285848724973, 0.0771663128087772 ], [ 0.15720939908415577, 0.036156713172170174, 0.09276621266397285, 0.057698682903803554, 0.09998547394819705, 0.008995372917456052, 0.09901662370771812, 0.0027393396493222 ], [ 0.027096650176441115, 0.07912597372374619, 0.013896747050818337, 0.025568298555558935, -0.007067702337425635, 0.02553393817315301, 0.10290113596944059, 0.02106096846632771 ], [ 0.08270433652361153, 0.004907585986363545, 0.10312235589809553, 0.058781565211504606, 0.09874998133425844, 0.06983371818783479, 0.1655699852589081, 0.023081584599757673 ], [ 0.053347845151708144, 0.022068062133014554, 0.10096142769969471, 0.017745759554232323, 0.022300336844405057, 0.023293099414273707, 0.029922102674532162, 0.08213831666347861 ], [ 0.12354075465487299, 0.04524514189770564, 0.1621130255354853, 0.03146196541719403, 0.11596628097831913, 0.055056672024664556, 0.09795508988743111, -0.002807203806616097 ], [ 0.04159122611590881, 0.0242849339441813, 0.004898455072847453, 0.028041868291503277, 0.025103118205391014, 0.09241320840068626, 0.0011754628287579125, 0.027840890926384365 ], [ 0.029165271184444678, 0.09349325361625152, 0.06887027326231678, 0.07338859206441362, 0.10383762820786119, -0.009701271766846655, 0.09585605759905204, 0.06291437135189794 ] ] }, { "label": "Candidate 2 · mask logits", "values": [ [ -0.11519440393930745, -0.010778362126919426, -0.10084472076401545, -0.014631528668295432, -0.00567977175902099, -0.08970840488554575, -0.013143543798022489, -0.1016632886428069 ], [ -0.2063128108867817, -0.046600919111636184, -0.1251562700585869, -0.08090699656203047, -0.1318762226311755, -0.013539074118067084, -0.13156190306539425, -0.0053775639978907745 ], [ -0.03342591600887176, -0.10612884934644867, -0.021002302837833765, -0.022776081776455605, 0.007916549706477495, -0.02136904138812765, -0.14668072700909765, -0.0159481610538776 ], [ -0.11233743903549953, -0.009660482790251777, -0.13323005717680694, -0.07885990450805415, -0.12528247198883546, -0.09320164695547874, -0.21896236108856623, -0.029274752587339248 ], [ -0.07649945448393908, -0.01803871215563883, -0.14426389968382486, -0.010624910419471967, -0.03326552523927514, -0.018674022194679896, -0.03798908864854001, -0.11002955080491432 ], [ -0.16224396680012618, -0.060452282849076604, -0.21382068950648903, -0.04024035661379759, -0.1500779709443392, -0.07330025844198033, -0.13198107909453008, 0.0016592595418260258 ], [ -0.05465708283820998, -0.027507792117398847, -0.006898556755827583, -0.028537082848450922, -0.031000104959781758, -0.12476706535335273, -0.0030578059861111387, -0.02604151325635472 ], [ -0.04226265225020236, -0.13380377814434008, -0.08960469683496111, -0.1011078276092877, -0.13941611061511097, 0.011303403489068074, -0.12280923691188518, -0.08438290509990243 ] ] }, { "label": "Candidate 3 · mask logits", "values": [ [ 0.03445904483225733, 0.10378056484615936, 0.04503640454020917, 0.07806352623205773, 0.016068813571267628, 0.09200171986806667, 0.025047037816930514, 0.09013790090167254 ], [ 0.18859486880600168, 0.04861347068509627, 0.0902179746528129, 0.037221557531310116, 0.11586958433680736, 0.00009359994207775559, 0.10880253794120485, -0.0077109541989906485 ], [ 0.04567252265573436, 0.08080558131766889, -0.0003865334829749061, 0.09716587883965903, -0.016860775937649376, 0.1055267204421172, 0.05164144658820484, 0.09738961232500663 ], [ 0.07576720287813642, -0.013977582802724802, 0.13667705590940898, 0.059916516255572504, 0.1450626736121269, 0.07417752335973533, 0.18827529569676005, 0.033958873558020554 ], [ 0.023968313152828064, 0.09385431663531264, 0.048520293560443906, 0.09941239709864384, 0.0020769563686109252, 0.1013225595086288, 0.04378615030988589, 0.08474336951359378 ], [ 0.1474889595412929, 0.04764552664313209, 0.18786022024134313, 0.044211824068705226, 0.15213287070108417, 0.05958936225792009, 0.09634888841906673, -0.015858672781792097 ], [ 0.04943132451804163, 0.05604548236440118, 0.00297627868215969, 0.08461892862169751, 0.04210646314503388, 0.08933761922024198, -0.00793732991779224, 0.09814625179517114 ], [ 0.010386618990803708, 0.043628535989884226, 0.08741176769650992, 0.05844810364287735, 0.10516330576882356, -0.020447607118830487, 0.13341941751159597, 0.06426157403550768 ] ] } ], "vectors": [ { "label": "Prompt coordinates · x₀, y₀, x₁, y₁", "values": [ 0.2, 0.2, 0.8, 0.8 ] }, { "label": "Prompt type IDs", "values": [ 2, 3 ] }, { "label": "Predicted quality scores · untrained", "values": [ -0.0651719163264563, -0.10798516329095176, 0.15255067520470086 ] } ], "metrics": [ { "label": "Image encoder calls per changed prompt", "value": 0 }, { "label": "Candidate masks", "value": 3 } ] } ] }, "revision": "06fa171ea414ac32", "files": { "graph.tensorviz.json": { "url": "/models/segment-anything/06fa171ea414ac32/graph.tensorviz.json", "sha256": "60274b9bc4e3a736dab315e7002d25191d05dd133d88013a5895957b10929b99", "bytes": 357615 }, "model.py": { "url": "/models/segment-anything/06fa171ea414ac32/model.py", "sha256": "bba1c4d0f25557c69579fade1e1ce55a8dd70d4af374ed69438d9a3652291d3e", "bytes": 10829 }, "run.py": { "url": "/models/segment-anything/06fa171ea414ac32/run.py", "sha256": "75bb7dad878fc9a168a747e5f68fe0cfbdc0151fdb58e2aa54ad00853911c154", "bytes": 768 }, "requirements.txt": { "url": "/models/segment-anything/06fa171ea414ac32/requirements.txt", "sha256": "edb86016b42cb3cc6cdc0c490aab929d3bb098ced4515394f0b55766033e0daf", "bytes": 91 }, "preview.svg": { "url": "/models/segment-anything/06fa171ea414ac32/preview.svg", "sha256": "df6b127ac8257b4ba4035c8359439d20d444e62b44e047e6d6580b7fc11886c2", "bytes": 2281 } } }