{ "schemaVersion": 1, "repository": "RabbitDaisuke/Foot-Point-Tracker-onnx", "license": "apache-2.0", "modelId": "google/causal-bootstapir", "source": { "codeRepository": "google-deepmind/tapnet", "codeRevision": "730cda1c730877cfedbe01bf87fb1cadb78a565d", "weightRepository": "google/tapnet", "weightRevision": "5d3fb48e76c5422841e38501514121e251beabb7", "filename": "causal_bootstapir_checkpoint.pt", "bytes": 218887028, "sha256": "87c1e752cf5ce56e3e2f7da460aeb4d40fc826d04ef2939bade86a5c7495377f", "url": "https://huggingface.co/google/tapnet/resolve/5d3fb48e76c5422841e38501514121e251beabb7/causal_bootstapir_checkpoint.pt", "license": "apache-2.0", "sourceFiles": [ { "path": "tapnet/torch/__init__.py", "url": "https://raw.githubusercontent.com/google-deepmind/tapnet/730cda1c730877cfedbe01bf87fb1cadb78a565d/tapnet/torch/__init__.py", "sha256": "f7c8d8c66e2f93ac350ce82c03f02152ffc82524cd120ebc944c931b6ea879e1", "bytes": 656 }, { "path": "tapnet/torch/tapir_model.py", "url": "https://raw.githubusercontent.com/google-deepmind/tapnet/730cda1c730877cfedbe01bf87fb1cadb78a565d/tapnet/torch/tapir_model.py", "sha256": "87c05fb17d48d0f4353a00b67c9d40d3a264e13e43d88e3f7b5829119845e3ba", "bytes": 27917 }, { "path": "tapnet/torch/nets.py", "url": "https://raw.githubusercontent.com/google-deepmind/tapnet/730cda1c730877cfedbe01bf87fb1cadb78a565d/tapnet/torch/nets.py", "sha256": "4961a8e4882e8c9ac219c4714f6f2c811cc1ab665a439d4d2e79942002a9517c", "bytes": 12032 }, { "path": "tapnet/torch/utils.py", "url": "https://raw.githubusercontent.com/google-deepmind/tapnet/730cda1c730877cfedbe01bf87fb1cadb78a565d/tapnet/torch/utils.py", "sha256": "a4267b3a6bd8ecf2dfe5effacab9f7a43664dc5d08966c98f6a0cc68f5ccedd3", "bytes": 10477 }, { "path": "LICENSE", "url": "https://raw.githubusercontent.com/google-deepmind/tapnet/730cda1c730877cfedbe01bf87fb1cadb78a565d/LICENSE", "sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", "bytes": 11358 } ] }, "inputWH": [ 256, 256 ], "queryCount": 3, "stateShape": [ 737280 ], "stateType": "float32", "stateLayout": [ { "iteration": 0, "key": "block_0_causal_1", "offset": 0, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_0_causal_2", "offset": 3072, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 0, "key": "block_1_causal_1", "offset": 15360, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_1_causal_2", "offset": 18432, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 0, "key": "block_2_causal_1", "offset": 30720, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_2_causal_2", "offset": 33792, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 0, "key": "block_3_causal_1", "offset": 46080, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_3_causal_2", "offset": 49152, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 0, "key": "block_4_causal_1", "offset": 61440, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_4_causal_2", "offset": 64512, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 0, "key": "block_5_causal_1", "offset": 76800, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_5_causal_2", "offset": 79872, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 0, "key": "block_6_causal_1", "offset": 92160, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_6_causal_2", "offset": 95232, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 0, "key": "block_7_causal_1", "offset": 107520, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_7_causal_2", "offset": 110592, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 0, "key": "block_8_causal_1", "offset": 122880, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_8_causal_2", "offset": 125952, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 0, "key": "block_9_causal_1", "offset": 138240, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_9_causal_2", "offset": 141312, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 0, "key": "block_10_causal_1", "offset": 153600, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_10_causal_2", "offset": 156672, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 0, "key": "block_11_causal_1", "offset": 168960, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 0, "key": "block_11_causal_2", "offset": 172032, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_0_causal_1", "offset": 184320, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_0_causal_2", "offset": 187392, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_1_causal_1", "offset": 199680, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_1_causal_2", "offset": 202752, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_2_causal_1", "offset": 215040, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_2_causal_2", "offset": 218112, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_3_causal_1", "offset": 230400, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_3_causal_2", "offset": 233472, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_4_causal_1", "offset": 245760, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_4_causal_2", "offset": 248832, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_5_causal_1", "offset": 261120, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_5_causal_2", "offset": 264192, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_6_causal_1", "offset": 276480, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_6_causal_2", "offset": 279552, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_7_causal_1", "offset": 291840, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_7_causal_2", "offset": 294912, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_8_causal_1", "offset": 307200, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_8_causal_2", "offset": 310272, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_9_causal_1", "offset": 322560, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_9_causal_2", "offset": 325632, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_10_causal_1", "offset": 337920, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_10_causal_2", "offset": 340992, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 1, "key": "block_11_causal_1", "offset": 353280, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 1, "key": "block_11_causal_2", "offset": 356352, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_0_causal_1", "offset": 368640, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_0_causal_2", "offset": 371712, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_1_causal_1", "offset": 384000, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_1_causal_2", "offset": 387072, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_2_causal_1", "offset": 399360, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_2_causal_2", "offset": 402432, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_3_causal_1", "offset": 414720, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_3_causal_2", "offset": 417792, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_4_causal_1", "offset": 430080, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_4_causal_2", "offset": 433152, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_5_causal_1", "offset": 445440, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_5_causal_2", "offset": 448512, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_6_causal_1", "offset": 460800, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_6_causal_2", "offset": 463872, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_7_causal_1", "offset": 476160, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_7_causal_2", "offset": 479232, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_8_causal_1", "offset": 491520, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_8_causal_2", "offset": 494592, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_9_causal_1", "offset": 506880, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_9_causal_2", "offset": 509952, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_10_causal_1", "offset": 522240, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_10_causal_2", "offset": 525312, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 2, "key": "block_11_causal_1", "offset": 537600, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 2, "key": "block_11_causal_2", "offset": 540672, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_0_causal_1", "offset": 552960, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_0_causal_2", "offset": 556032, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_1_causal_1", "offset": 568320, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_1_causal_2", "offset": 571392, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_2_causal_1", "offset": 583680, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_2_causal_2", "offset": 586752, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_3_causal_1", "offset": 599040, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_3_causal_2", "offset": 602112, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_4_causal_1", "offset": 614400, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_4_causal_2", "offset": 617472, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_5_causal_1", "offset": 629760, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_5_causal_2", "offset": 632832, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_6_causal_1", "offset": 645120, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_6_causal_2", "offset": 648192, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_7_causal_1", "offset": 660480, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_7_causal_2", "offset": 663552, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_8_causal_1", "offset": 675840, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_8_causal_2", "offset": 678912, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_9_causal_1", "offset": 691200, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_9_causal_2", "offset": 694272, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_10_causal_1", "offset": 706560, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_10_causal_2", "offset": 709632, "length": 12288, "shape": [ 1, 3, 2, 2048 ] }, { "iteration": 3, "key": "block_11_causal_1", "offset": 721920, "length": 3072, "shape": [ 1, 3, 2, 512 ] }, { "iteration": 3, "key": "block_11_causal_2", "offset": 724992, "length": 12288, "shape": [ 1, 3, 2, 2048 ] } ], "preprocessing": { "input": "RGB uint8", "resize": "float bilinear align_corners=false", "tensor": "float32[1,1,256,256,3]", "range": [ -1, 1 ], "query": "first-frame t=0, model-pixel yx float32[1,3,2]" }, "models": [ { "filename": "causal-bootstapir-encoder-256-q3-fp32.onnx", "bytes": 115330894, "sha256": "2f74d4d6d3a052e0baeabf2ec79a8668ed06af8c8bf77150f8373b7e63956519", "opset": [ { "domain": "", "version": 17 } ], "inputs": [ { "name": "frame", "shape": [ 1, 1, 256, 256, 3 ], "type": "tensor(float)" } ], "outputs": [ { "name": "lowres", "shape": [ 1, 1, 32, 32, 256 ], "type": "tensor(float)" }, { "name": "hires", "shape": [ 1, 1, 64, 64, 128 ], "type": "tensor(float)" } ], "operators": { "Add": 23, "Cast": 9, "Clip": 2, "Concat": 14, "Constant": 122, "ConstantOfShape": 9, "Conv": 31, "Div": 2, "Gather": 6, "InstanceNormalization": 16, "LayerNormalization": 5, "Mul": 32, "Pad": 9, "ReduceSum": 2, "Relu": 16, "Reshape": 23, "Resize": 1, "Shape": 7, "Slice": 11, "Sqrt": 2, "Tanh": 5, "Transpose": 24, "Unsqueeze": 6 }, "webgpuStatic": { "runtimeVersion": "1.22.0", "unlistedComputeOps": [], "shapeRestrictions": [], "hostShapeOps": [ "Constant", "ConstantOfShape" ], "hardwareExecuted": false }, "id": "encoder", "exportSeconds": 2.887656843000002, "graphNormalization": [ { "node": "/Max", "from": "Max", "to": "Clip", "minimum": 9.999999960041972e-13 }, { "node": "/Max_1", "from": "Max", "to": "Clip", "minimum": 9.999999960041972e-13 } ], "revision": "4751f96d5413e48c05e63da133a5b4b1e6074c12", "url": "https://huggingface.co/RabbitDaisuke/Foot-Point-Tracker-onnx/resolve/4751f96d5413e48c05e63da133a5b4b1e6074c12/causal-bootstapir-encoder-256-q3-fp32.onnx" }, { "filename": "causal-bootstapir-query-256-q3-fp32.onnx", "bytes": 2291, "sha256": "3eabf9ac98fc426cf969c814b61a5f96f062aea656b149ec709d8e33d3b7ad7f", "opset": [ { "domain": "", "version": 17 } ], "inputs": [ { "name": "lowres", "shape": [ 1, 1, 32, 32, 256 ], "type": "tensor(float)" }, { "name": "hires", "shape": [ 1, 1, 64, 64, 128 ], "type": "tensor(float)" }, { "name": "query_yx", "shape": [ 1, 3, 2 ], "type": "tensor(float)" } ], "outputs": [ { "name": "query_lowres", "shape": [ 1, 3, 256 ], "type": "tensor(float)" }, { "name": "query_hires", "shape": [ 1, 3, 128 ], "type": "tensor(float)" } ], "operators": { "Constant": 10, "Gather": 2, "GridSample": 2, "Mul": 1, "Slice": 1, "Squeeze": 2, "Sub": 1, "Transpose": 4, "Unsqueeze": 1 }, "webgpuStatic": { "runtimeVersion": "1.22.0", "unlistedComputeOps": [], "shapeRestrictions": [], "hostShapeOps": [ "Constant" ], "hardwareExecuted": false }, "id": "query", "exportSeconds": 0.014053634999996234, "graphNormalization": [], "revision": "4751f96d5413e48c05e63da133a5b4b1e6074c12", "url": "https://huggingface.co/RabbitDaisuke/Foot-Point-Tracker-onnx/resolve/4751f96d5413e48c05e63da133a5b4b1e6074c12/causal-bootstapir-query-256-q3-fp32.onnx" }, { "filename": "causal-bootstapir-step-256-q3-fp32.onnx", "bytes": 105305753, "sha256": "953d8b48f703e70eb9d7c3b5bb53cf0634de0438bc0df0964fb6961fd5f4cd46", "opset": [ { "domain": "", "version": 17 } ], "inputs": [ { "name": "lowres", "shape": [ 1, 1, 32, 32, 256 ], "type": "tensor(float)" }, { "name": "hires", "shape": [ 1, 1, 64, 64, 128 ], "type": "tensor(float)" }, { "name": "query_lowres", "shape": [ 1, 3, 256 ], "type": "tensor(float)" }, { "name": "query_hires", "shape": [ 1, 3, 128 ], "type": "tensor(float)" }, { "name": "causal_state", "shape": [ 737280 ], "type": "tensor(float)" } ], "outputs": [ { "name": "tracks", "shape": [ 1, 3, 1, 2 ], "type": "tensor(float)" }, { "name": "occlusion", "shape": [ 1, 3, 1 ], "type": "tensor(float)" }, { "name": "expected_dist", "shape": [ 1, 3, 1 ], "type": "tensor(float)" }, { "name": "new_causal_state", "shape": [ 737280 ], "type": "tensor(float)" } ], "operators": { "Add": 565, "ArgMax": 1, "AveragePool": 4, "Cast": 363, "Clip": 1, "Concat": 438, "Constant": 4784, "ConstantOfShape": 100, "Conv": 99, "Div": 141, "Einsum": 13, "Expand": 26, "Gather": 579, "Gemm": 2, "Greater": 1, "GridSample": 12, "Identity": 78, "If": 2, "LayerNormalization": 100, "Less": 2, "MatMul": 104, "Mul": 628, "Pad": 97, "ReduceMean": 1, "ReduceSum": 3, "Relu": 3, "Reshape": 646, "ScatterND": 1, "Shape": 377, "Slice": 619, "Softmax": 1, "Sub": 14, "Tanh": 96, "Tile": 1, "Transpose": 342, "Unsqueeze": 586 }, "webgpuStatic": { "runtimeVersion": "1.22.0", "unlistedComputeOps": [], "shapeRestrictions": [], "hostShapeOps": [ "Constant", "ConstantOfShape", "Identity" ], "hardwareExecuted": false }, "id": "step", "exportSeconds": 8.792766373, "graphNormalization": [ { "node": "/Max", "from": "Max", "to": "Clip", "minimum": 9.999999960041972e-13 }, { "node": "/Size", "from": "Size", "to": "Constant", "inputShape": [ 4 ], "value": 4 } ], "revision": "4751f96d5413e48c05e63da133a5b4b1e6074c12", "url": "https://huggingface.co/RabbitDaisuke/Foot-Point-Tracker-onnx/resolve/4751f96d5413e48c05e63da133a5b4b1e6074c12/causal-bootstapir-step-256-q3-fp32.onnx" } ], "visibility": { "formula": "(1-sigmoid(occlusion))*(1-sigmoid(expected_dist))", "threshold": 0.5 }, "parityPassed": true, "webgpuStatic": { "operatorTable": "https://raw.githubusercontent.com/microsoft/onnxruntime/v1.22.0/js/web/docs/webgpu-operators.md", "operatorTableSha256": "e60f85a34d43d55cb6f4ea1b97571f6ef57b92c2aeb9b8d99fae50076caa06fe", "runtimeVersion": "1.22.0", "unlistedComputeOps": [], "hardwareExecuted": false }, "gitCommit": "274407d44a7fbdcb8411f9dc78286b1f49409c6d", "workflowRunId": "41384892970", "codecParityPassed": true, "requestedRepository": "RabbitDaisuke/Foot-Point-Tracker-onnx", "pathPrefix": "" }