{ "producer": "pytorch", "producer_version": "2.13.0", "ir_version": 9, "opsets": [ [ "", 19 ] ], "inputs": [ { "name": "input", "dtype": "FLOAT", "shape": [ "batch", 3, 576, 576 ] } ], "outputs": [ { "name": "dets", "dtype": "FLOAT", "shape": [ "batch", "Concatdets_dim_1", "Concatdets_dim_2" ] }, { "name": "labels", "dtype": "FLOAT", "shape": [ "batch", "Addlabels_dim_1", 91 ] } ], "nodes": 3367, "external_initializers": 0, "ops": { "Constant": 1128, "Shape": 236, "Gather": 204, "Cast": 123, "Conv": 9, "Slice": 71, "Concat": 170, "Reshape": 149, "Transpose": 125, "Unsqueeze": 349, "ConstantOfShape": 8, "Mul": 159, "Equal": 4, "Where": 6, "Expand": 8, "Add": 188, "Div": 77, "Tile": 3, "LayerNormalization": 51, "MatMul": 151, "Sqrt": 48, "Softmax": 20, "Erf": 12, "Sigmoid": 8, "Split": 5, "Sub": 10, "Range": 2, "Greater": 2, "Less": 1, "And": 1, "Not": 3, "ReduceSum": 5, "Relu": 9, "Exp": 3, "ReduceMax": 1, "TopK": 1, "GatherElements": 1, "Sin": 4, "Cos": 4, "Gemm": 4, "GridSample": 4 }, "input_consumers": { "input": [ { "name": "/backbone/backbone.0/encoder/encoder/embeddings/Shape", "op": "Shape", "inputs": [ "input" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Shape_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/embeddings/Shape_1", "op": "Shape", "inputs": [ "input" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Shape_1_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/embeddings/Shape_2", "op": "Shape", "inputs": [ "input" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Shape_2_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/embeddings/Cast", "op": "Cast", "inputs": [ "input" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "attrs": { "to": 1 } } ] }, "output_paths": { "dets": [ { "name": "/Concat", "op": "Concat", "inputs": [ "/Add_output_0", "/Mul_1_output_0" ], "outputs": [ "dets" ], "attrs": { "axis": -1 } }, { "name": "/Add", "op": "Add", "inputs": [ "/Mul_output_0", "/Slice_2_output_0" ], "outputs": [ "/Add_output_0" ], "attrs": {} }, { "name": "/Mul", "op": "Mul", "inputs": [ "/Slice_output_0", "/Slice_1_output_0" ], "outputs": [ "/Mul_output_0" ], "attrs": {} }, { "name": "/Slice", "op": "Slice", "inputs": [ "/bbox_embed/layers.2/Add_output_0", "/Constant_1_output_0", "/Constant_2_output_0", "/Constant_output_0", "/Constant_3_output_0" ], "outputs": [ "/Slice_output_0" ], "attrs": {} }, { "name": "/bbox_embed/layers.2/Add", "op": "Add", "inputs": [ "bbox_embed.layers.2.bias", "/bbox_embed/layers.2/MatMul_output_0" ], "outputs": [ "/bbox_embed/layers.2/Add_output_0" ], "attrs": {} }, { "name": "/bbox_embed/layers.2/MatMul", "op": "MatMul", "inputs": [ "/bbox_embed/Relu_1_output_0", "onnx::MatMul_4938" ], "outputs": [ "/bbox_embed/layers.2/MatMul_output_0" ], "attrs": {} }, { "name": "/Constant_1", "op": "Constant", "inputs": [], "outputs": [ "/Constant_1_output_0" ], "attrs": { "value": [ 0 ] } }, { "name": "/Constant_2", "op": "Constant", "inputs": [], "outputs": [ "/Constant_2_output_0" ], "attrs": { "value": [ 2 ] } }, { "name": "/Constant", "op": "Constant", "inputs": [], "outputs": [ "/Constant_output_0" ], "attrs": { "value": [ 2 ] } }, { "name": "/Constant_3", "op": "Constant", "inputs": [], "outputs": [ "/Constant_3_output_0" ], "attrs": { "value": [ 1 ] } }, { "name": "/Slice_1", "op": "Slice", "inputs": [ "/transformer/Concat_15_output_0", "/Constant_5_output_0", "/Constant_6_output_0", "/Constant_4_output_0", "/Constant_7_output_0" ], "outputs": [ "/Slice_1_output_0" ], "attrs": {} }, { "name": "/transformer/Concat_15", "op": "Concat", "inputs": [ "/transformer/Concat_14_output_0", "/transformer/Slice_6_output_0" ], "outputs": [ "/transformer/Concat_15_output_0" ], "attrs": { "axis": -2 } }, { "name": "/transformer/Concat_14", "op": "Concat", "inputs": [ "/transformer/Add_6_output_0", "/transformer/Mul_9_output_0" ], "outputs": [ "/transformer/Concat_14_output_0" ], "attrs": { "axis": -1 } }, { "name": "/transformer/Slice_6", "op": "Slice", "inputs": [ "/transformer/Expand_4_output_0", "/transformer/Unsqueeze_10_output_0", "/transformer/Constant_77_output_0", "/transformer/Constant_75_output_0", "/transformer/Constant_78_output_0" ], "outputs": [ "/transformer/Slice_6_output_0" ], "attrs": {} }, { "name": "/Constant_5", "op": "Constant", "inputs": [], "outputs": [ "/Constant_5_output_0" ], "attrs": { "value": [ 2 ] } }, { "name": "/Constant_6", "op": "Constant", "inputs": [], "outputs": [ "/Constant_6_output_0" ], "attrs": { "value": [ 9223372036854775807 ] } }, { "name": "/Constant_4", "op": "Constant", "inputs": [], "outputs": [ "/Constant_4_output_0" ], "attrs": { "value": [ 2 ] } }, { "name": "/Constant_7", "op": "Constant", "inputs": [], "outputs": [ "/Constant_7_output_0" ], "attrs": { "value": [ 1 ] } }, { "name": "/Slice_2", "op": "Slice", "inputs": [ "/transformer/Concat_15_output_0", "/Constant_9_output_0", "/Constant_10_output_0", "/Constant_8_output_0", "/Constant_11_output_0" ], "outputs": [ "/Slice_2_output_0" ], "attrs": {} }, { "name": "/Constant_9", "op": "Constant", "inputs": [], "outputs": [ "/Constant_9_output_0" ], "attrs": { "value": [ 0 ] } }, { "name": "/Constant_10", "op": "Constant", "inputs": [], "outputs": [ "/Constant_10_output_0" ], "attrs": { "value": [ 2 ] } }, { "name": "/Constant_8", "op": "Constant", "inputs": [], "outputs": [ "/Constant_8_output_0" ], "attrs": { "value": [ 2 ] } }, { "name": "/Constant_11", "op": "Constant", "inputs": [], "outputs": [ "/Constant_11_output_0" ], "attrs": { "value": [ 1 ] } }, { "name": "/Mul_1", "op": "Mul", "inputs": [ "/Exp_output_0", "/Slice_1_output_0" ], "outputs": [ "/Mul_1_output_0" ], "attrs": {} }, { "name": "/Exp", "op": "Exp", "inputs": [ "/Slice_3_output_0" ], "outputs": [ "/Exp_output_0" ], "attrs": {} }, { "name": "/Slice_3", "op": "Slice", "inputs": [ "/bbox_embed/layers.2/Add_output_0", "/Constant_13_output_0", "/Constant_14_output_0", "/Constant_12_output_0", "/Constant_15_output_0" ], "outputs": [ "/Slice_3_output_0" ], "attrs": {} }, { "name": "/Constant_13", "op": "Constant", "inputs": [], "outputs": [ "/Constant_13_output_0" ], "attrs": { "value": [ 2 ] } }, { "name": "/Constant_14", "op": "Constant", "inputs": [], "outputs": [ "/Constant_14_output_0" ], "attrs": { "value": [ 9223372036854775807 ] } }, { "name": "/Constant_12", "op": "Constant", "inputs": [], "outputs": [ "/Constant_12_output_0" ], "attrs": { "value": [ 2 ] } }, { "name": "/Constant_15", "op": "Constant", "inputs": [], "outputs": [ "/Constant_15_output_0" ], "attrs": { "value": [ 1 ] } } ], "labels": [ { "name": "/class_embed/Add", "op": "Add", "inputs": [ "class_embed.bias", "/class_embed/MatMul_output_0" ], "outputs": [ "labels" ], "attrs": {} }, { "name": "/class_embed/MatMul", "op": "MatMul", "inputs": [ "/transformer/decoder/norm/LayerNormalization_output_0", "onnx::MatMul_4955" ], "outputs": [ "/class_embed/MatMul_output_0" ], "attrs": {} }, { "name": "/transformer/decoder/norm/LayerNormalization", "op": "LayerNormalization", "inputs": [ "/transformer/decoder/layers.3/norm3/LayerNormalization_output_0", "transformer.decoder.norm.weight", "transformer.decoder.norm.bias" ], "outputs": [ "/transformer/decoder/norm/LayerNormalization_output_0" ], "attrs": { "axis": -1, "epsilon": 9.999999747378752e-06 } }, { "name": "/transformer/decoder/layers.3/norm3/LayerNormalization", "op": "LayerNormalization", "inputs": [ "/transformer/decoder/layers.3/Add_4_output_0", "transformer.decoder.layers.3.norm3.weight", "transformer.decoder.layers.3.norm3.bias" ], "outputs": [ "/transformer/decoder/layers.3/norm3/LayerNormalization_output_0" ], "attrs": { "axis": -1, "epsilon": 9.999999747378752e-06 } }, { "name": "/transformer/decoder/layers.3/Add_4", "op": "Add", "inputs": [ "/transformer/decoder/layers.3/norm2/LayerNormalization_output_0", "/transformer/decoder/layers.3/linear2/Add_output_0" ], "outputs": [ "/transformer/decoder/layers.3/Add_4_output_0" ], "attrs": {} }, { "name": "/transformer/decoder/layers.3/norm2/LayerNormalization", "op": "LayerNormalization", "inputs": [ "/transformer/decoder/layers.3/Add_3_output_0", "transformer.decoder.layers.3.norm2.weight", "transformer.decoder.layers.3.norm2.bias" ], "outputs": [ "/transformer/decoder/layers.3/norm2/LayerNormalization_output_0" ], "attrs": { "axis": -1, "epsilon": 9.999999747378752e-06 } }, { "name": "/transformer/decoder/layers.3/linear2/Add", "op": "Add", "inputs": [ "transformer.decoder.layers.3.linear2.bias", "/transformer/decoder/layers.3/linear2/MatMul_output_0" ], "outputs": [ "/transformer/decoder/layers.3/linear2/Add_output_0" ], "attrs": {} } ] }, "after_input_cast": [ { "name": "/backbone/backbone.0/encoder/encoder/embeddings/patch_embeddings/projection/Conv", "op": "Conv", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0", "backbone.0.encoder.encoder.embeddings.patch_embeddings.projection.weight", "backbone.0.encoder.encoder.embeddings.patch_embeddings.projection.bias" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/patch_embeddings/projection/Conv_output_0" ], "attrs": { "dilations": [ 1, 1 ], "group": 1, "kernel_shape": [ 16, 16 ], "pads": [ 0, 0, 0, 0 ], "strides": [ 16, 16 ] } }, { "name": "/backbone/backbone.0/encoder/encoder/Shape", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/Shape_1", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_1_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/Shape_2", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_2_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/Shape_6", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_6_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/Shape_7", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_7_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/Shape_8", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_8_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/Shape_12", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_12_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/Shape_13", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_13_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/Shape_14", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_14_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/Shape_18", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_18_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/Shape_19", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_19_output_0" ], "attrs": {} }, { "name": "/backbone/backbone.0/encoder/encoder/Shape_20", "op": "Shape", "inputs": [ "/backbone/backbone.0/encoder/encoder/embeddings/Cast_output_0" ], "outputs": [ "/backbone/backbone.0/encoder/encoder/Shape_20_output_0" ], "attrs": {} } ], "topk": [ { "name": "/transformer/TopK", "op": "TopK", "inputs": [ "/transformer/ReduceMax_output_0", "/transformer/Constant_57_output_0" ], "outputs": [ "/transformer/TopK_output_0", "/transformer/TopK_output_1" ], "attrs": { "axis": 1, "largest": 1, "sorted": 1 } } ], "head_initializers": [ { "name": "class_embed.bias", "dims": [ 91 ] } ], "checker": "PASS", "inferred_outputs": [ { "name": "dets", "dtype": "FLOAT", "shape": [ "batch", 300, 4 ] }, { "name": "labels", "dtype": "FLOAT", "shape": [ "batch", 300, 91 ] } ] }