diff --git a/ANE/.DS_Store b/ANE/.DS_Store new file mode 100644 index 0000000000000000000000000000000000000000..1d34ffbe2f38b024e661ae2658bda93620cf9a44 Binary files /dev/null and b/ANE/.DS_Store differ diff --git a/ANE/styletts2_ane_alignment.mlmodelc/analytics/coremldata.bin b/ANE/styletts2_ane_alignment.mlmodelc/analytics/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..d9fd13a5d0f529dc1e2b4aff23c0d0229bf0b403 --- /dev/null +++ b/ANE/styletts2_ane_alignment.mlmodelc/analytics/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79b2d334b136d0db3149ae324b9423c3c3985d12512d8bb977d6e157bf37c7d4 +size 243 diff --git a/ANE/styletts2_ane_alignment.mlmodelc/coremldata.bin b/ANE/styletts2_ane_alignment.mlmodelc/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..466b14ed5f7b748bb6e9c9b6c1c9627a4dc89202 --- /dev/null +++ b/ANE/styletts2_ane_alignment.mlmodelc/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2578251e92caabd70dd2d68c453e0ed86d05c2e7db9b5511e25a9287c9fdeee9 +size 519 diff --git a/ANE/styletts2_ane_alignment.mlmodelc/metadata.json b/ANE/styletts2_ane_alignment.mlmodelc/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..b40819623bcaa3412111638257692c0542cd7669 --- /dev/null +++ b/ANE/styletts2_ane_alignment.mlmodelc/metadata.json @@ -0,0 +1,104 @@ +[ + { + "metadataOutputVersion" : "3.0", + "shortDescription" : "StyleTTS2-ANE Stage 3: Alignment (cumsum+broadcast, T_a≤2000) — fp16, ANE.", + "outputSchema" : [ + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float16", + "formattedType" : "MultiArray (Float16 1 × 640 × 2000)", + "shortDescription" : "", + "shape" : "[1, 640, 2000]", + "name" : "en", + "type" : "MultiArray" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float16", + "formattedType" : "MultiArray (Float16 1 × 512 × 2000)", + "shortDescription" : "", + "shape" : "[1, 512, 2000]", + "name" : "asr", + "type" : "MultiArray" + } + ], + "storagePrecision" : "Float16", + "modelParameters" : [ + + ], + "specificationVersion" : 8, + "mlProgramOperationTypeHistogram" : { + "Ios16.cumsum" : 1, + "Ios17.logicalAnd" : 1, + "Ios17.cast" : 4, + "Ios17.less" : 1, + "Ios17.matmul" : 2, + "Ios17.greaterEqual" : 1, + "Ios17.sub" : 1, + "Ios17.expandDims" : 2 + }, + "computePrecision" : "Mixed (Float16, Int32)", + "isUpdatable" : "0", + "stateSchema" : [ + + ], + "availability" : { + "macOS" : "14.0", + "tvOS" : "17.0", + "visionOS" : "1.0", + "watchOS" : "10.0", + "iOS" : "17.0", + "macCatalyst" : "17.0" + }, + "modelType" : { + "name" : "MLModelType_mlProgram" + }, + "inputSchema" : [ + { + "dataType" : "Float32", + "hasShapeFlexibility" : "1", + "isOptional" : "0", + "shapeFlexibility" : "1 × 2...512", + "shapeRange" : "[[1, 1], [2, 512]]", + "formattedType" : "MultiArray (Float32 1 × 64)", + "type" : "MultiArray", + "shape" : "[1, 64]", + "name" : "pred_dur", + "shortDescription" : "" + }, + { + "dataType" : "Float32", + "hasShapeFlexibility" : "1", + "isOptional" : "0", + "shapeFlexibility" : "1 × 2...512 × 640", + "shapeRange" : "[[1, 1], [2, 512], [640, 640]]", + "formattedType" : "MultiArray (Float32 1 × 64 × 640)", + "type" : "MultiArray", + "shape" : "[1, 64, 640]", + "name" : "d", + "shortDescription" : "" + }, + { + "dataType" : "Float32", + "hasShapeFlexibility" : "1", + "isOptional" : "0", + "shapeFlexibility" : "1 × 512 × 2...512", + "shapeRange" : "[[1, 1], [512, 512], [2, 512]]", + "formattedType" : "MultiArray (Float32 1 × 512 × 64)", + "type" : "MultiArray", + "shape" : "[1, 512, 64]", + "name" : "t_en", + "shortDescription" : "" + } + ], + "userDefinedMetadata" : { + "com.github.apple.coremltools.source_dialect" : "TorchScript", + "com.github.apple.coremltools.source" : "torch==2.11.0", + "com.github.apple.coremltools.version" : "8.3.0" + }, + "generatedClassName" : "styletts2_ane_alignment", + "method" : "predict" + } +] \ No newline at end of file diff --git a/ANE/styletts2_ane_alignment.mlmodelc/model.mil b/ANE/styletts2_ane_alignment.mlmodelc/model.mil new file mode 100644 index 0000000000000000000000000000000000000000..0d87fcdb7ed718e7c65ed1054ffba06057ef30ff --- /dev/null +++ b/ANE/styletts2_ane_alignment.mlmodelc/model.mil @@ -0,0 +1,33 @@ +program(1.0) +[buildInfo = dict, tensor>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}})] +{ + func main(tensor d, tensor pred_dur, tensor t_en) [FlexibleShapeInformation = tuple, dict, tensor>>, tuple, dict, list, ?>>>>((("DefaultShapes", {{"d", [1, 64, 640]}, {"pred_dur", [1, 64]}, {"t_en", [1, 512, 64]}}), ("RangeDims", {{"d", [[1, 1], [2, 512], [640, 640]]}, {"pred_dur", [[1, 1], [2, 512]]}, {"t_en", [[1, 1], [512, 512], [2, 512]]}})))] { + tensor var_9 = const()[name = tensor("op_9"), val = tensor(-1)]; + tensor cum_exclusive_0 = const()[name = tensor("cum_exclusive_0"), val = tensor(false)]; + tensor cum_reverse_0 = const()[name = tensor("cum_reverse_0"), val = tensor(false)]; + tensor pred_dur_to_fp16_dtype_0 = const()[name = tensor("pred_dur_to_fp16_dtype_0"), val = tensor("fp16")]; + tensor pred_dur_to_fp16 = cast(dtype = pred_dur_to_fp16_dtype_0, x = pred_dur)[name = tensor("cast_3")]; + tensor cum_cast_fp16 = cumsum(axis = var_9, exclusive = cum_exclusive_0, reverse = cum_reverse_0, x = pred_dur_to_fp16)[name = tensor("cum_cast_fp16")]; + tensor starts_cast_fp16 = sub(x = cum_cast_fp16, y = pred_dur_to_fp16)[name = tensor("starts_cast_fp16")]; + tensor var_25_axes_0 = const()[name = tensor("op_25_axes_0"), val = tensor([-1])]; + tensor var_25_cast_fp16 = expand_dims(axes = var_25_axes_0, x = starts_cast_fp16)[name = tensor("op_25_cast_fp16")]; + tensor frames_promoted_to_fp16 = const()[name = tensor("frames_promoted_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(64)))]; + tensor var_26_cast_fp16 = greater_equal(x = frames_promoted_to_fp16, y = var_25_cast_fp16)[name = tensor("op_26_cast_fp16")]; + tensor var_28_axes_0 = const()[name = tensor("op_28_axes_0"), val = tensor([-1])]; + tensor var_28_cast_fp16 = expand_dims(axes = var_28_axes_0, x = cum_cast_fp16)[name = tensor("op_28_cast_fp16")]; + tensor var_29_cast_fp16 = less(x = frames_promoted_to_fp16, y = var_28_cast_fp16)[name = tensor("op_29_cast_fp16")]; + tensor in_seg = logical_and(x = var_26_cast_fp16, y = var_29_cast_fp16)[name = tensor("in_seg")]; + tensor d_to_fp16_dtype_0 = const()[name = tensor("d_to_fp16_dtype_0"), val = tensor("fp16")]; + tensor var_44_transpose_x_1 = const()[name = tensor("op_44_transpose_x_1"), val = tensor(true)]; + tensor var_44_transpose_y_1 = const()[name = tensor("op_44_transpose_y_1"), val = tensor(false)]; + tensor cast_3_to_fp16_dtype_0 = const()[name = tensor("cast_3_to_fp16_dtype_0"), val = tensor("fp16")]; + tensor d_to_fp16 = cast(dtype = d_to_fp16_dtype_0, x = d)[name = tensor("cast_1")]; + tensor in_seg_to_fp16 = cast(dtype = cast_3_to_fp16_dtype_0, x = in_seg)[name = tensor("cast_2")]; + tensor en = matmul(transpose_x = var_44_transpose_x_1, transpose_y = var_44_transpose_y_1, x = d_to_fp16, y = in_seg_to_fp16)[name = tensor("op_44_cast_fp16")]; + tensor var_50_transpose_x_0 = const()[name = tensor("op_50_transpose_x_0"), val = tensor(false)]; + tensor var_50_transpose_y_0 = const()[name = tensor("op_50_transpose_y_0"), val = tensor(false)]; + tensor t_en_to_fp16_dtype_0 = const()[name = tensor("t_en_to_fp16_dtype_0"), val = tensor("fp16")]; + tensor t_en_to_fp16 = cast(dtype = t_en_to_fp16_dtype_0, x = t_en)[name = tensor("cast_0")]; + tensor asr = matmul(transpose_x = var_50_transpose_x_0, transpose_y = var_50_transpose_y_0, x = t_en_to_fp16, y = in_seg_to_fp16)[name = tensor("op_50_cast_fp16")]; + } -> (en, asr); +} \ No newline at end of file diff --git a/ANE/styletts2_ane_alignment.mlmodelc/weights/weight.bin b/ANE/styletts2_ane_alignment.mlmodelc/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..6104dbdaf764de798e11c12c90ac8fa596b41bd5 --- /dev/null +++ b/ANE/styletts2_ane_alignment.mlmodelc/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e7d69128b59d615fc3d3cf85637a687235fc086b1eb136359adb11a61615f6b +size 4128 diff --git a/ANE/styletts2_ane_alignment.mlpackage/Data/com.apple.CoreML/model.mlmodel b/ANE/styletts2_ane_alignment.mlpackage/Data/com.apple.CoreML/model.mlmodel new file mode 100644 index 0000000000000000000000000000000000000000..f34d06211c7579ab2a8d83edbd7fefecb5042bfc --- /dev/null +++ b/ANE/styletts2_ane_alignment.mlpackage/Data/com.apple.CoreML/model.mlmodel @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0a866791eb7d8e7fe4a54551d9c98d7e8a10bd442799862fa52dd81ac9c550e6 +size 4162 diff --git a/ANE/styletts2_ane_alignment.mlpackage/Data/com.apple.CoreML/weights/weight.bin b/ANE/styletts2_ane_alignment.mlpackage/Data/com.apple.CoreML/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..6104dbdaf764de798e11c12c90ac8fa596b41bd5 --- /dev/null +++ b/ANE/styletts2_ane_alignment.mlpackage/Data/com.apple.CoreML/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e7d69128b59d615fc3d3cf85637a687235fc086b1eb136359adb11a61615f6b +size 4128 diff --git a/ANE/styletts2_ane_alignment.mlpackage/Manifest.json b/ANE/styletts2_ane_alignment.mlpackage/Manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..214caaab8633cc1cac6c4160b77eb2b995500280 --- /dev/null +++ b/ANE/styletts2_ane_alignment.mlpackage/Manifest.json @@ -0,0 +1,18 @@ +{ + "fileFormatVersion": "1.0.0", + "itemInfoEntries": { + "B5C37E4A-EDAD-42AA-A450-1BF7CC840C90": { + "author": "com.apple.CoreML", + "description": "CoreML Model Weights", + "name": "weights", + "path": "com.apple.CoreML/weights" + }, + "DBA3FF95-5E96-49D4-A75D-BD281A186EE8": { + "author": "com.apple.CoreML", + "description": "CoreML Model Specification", + "name": "model.mlmodel", + "path": "com.apple.CoreML/model.mlmodel" + } + }, + "rootModelIdentifier": "DBA3FF95-5E96-49D4-A75D-BD281A186EE8" +} diff --git a/ANE/styletts2_ane_diffusion_step.mlmodelc/analytics/coremldata.bin b/ANE/styletts2_ane_diffusion_step.mlmodelc/analytics/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..eeb98f1bf470d80fd2036dc3a8f2b64b9f513f16 --- /dev/null +++ b/ANE/styletts2_ane_diffusion_step.mlmodelc/analytics/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:78d70f424ec17bfb8c861db787faf81ca410d2011942e5f50a7e2a251108d2f3 +size 243 diff --git a/ANE/styletts2_ane_diffusion_step.mlmodelc/coremldata.bin b/ANE/styletts2_ane_diffusion_step.mlmodelc/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..730af6e0d867037ffc7c8984b2bdf6baf84fd558 --- /dev/null +++ b/ANE/styletts2_ane_diffusion_step.mlmodelc/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:979bc3282748bfae788bd403e1bd2c91ff6e9f348e90545b14cb80c0aab0cf57 +size 472 diff --git a/ANE/styletts2_ane_diffusion_step.mlmodelc/metadata.json b/ANE/styletts2_ane_diffusion_step.mlmodelc/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..bc3ac116d274ca17216a89f28565d35a14f35e5c --- /dev/null +++ b/ANE/styletts2_ane_diffusion_step.mlmodelc/metadata.json @@ -0,0 +1,115 @@ +[ + { + "metadataOutputVersion" : "3.0", + "shortDescription" : "StyleTTS2-ANE Stage 4: ADPM2 single denoise step (static shapes) — fp16, ANE.", + "outputSchema" : [ + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float16", + "formattedType" : "MultiArray (Float16 1 × 1 × 256)", + "shortDescription" : "", + "shape" : "[1, 1, 256]", + "name" : "denoised", + "type" : "MultiArray" + } + ], + "storagePrecision" : "Mixed (Float16, Palettized (8 bits))", + "modelParameters" : [ + + ], + "specificationVersion" : 8, + "mlProgramOperationTypeHistogram" : { + "Ios17.cos" : 1, + "Ios17.reshape" : 20, + "Ios17.inverse" : 2, + "Ios16.softmax" : 3, + "Ios17.matmul" : 6, + "Ios17.sin" : 1, + "Ios17.transpose" : 27, + "Split" : 9, + "Ios16.reduceMean" : 1, + "Ios17.expandDims" : 2, + "Ios17.add" : 23, + "Tile" : 2, + "Ios16.reduceSum" : 1, + "Ios17.log" : 1, + "Ios17.layerNorm" : 3, + "Ios17.sqrt" : 1, + "Ios16.gelu" : 7, + "Ios16.constexprLutToDense" : 26, + "Ios17.cast" : 4, + "Ios17.realDiv" : 1, + "Ios17.linear" : 25, + "Stack" : 1, + "Ios17.concat" : 2, + "Ios17.conv" : 1, + "Ios17.mul" : 19 + }, + "computePrecision" : "Mixed (Float16, Float32, Int32)", + "isUpdatable" : "0", + "stateSchema" : [ + + ], + "availability" : { + "macOS" : "14.0", + "tvOS" : "17.0", + "visionOS" : "1.0", + "watchOS" : "10.0", + "iOS" : "17.0", + "macCatalyst" : "17.0" + }, + "modelType" : { + "name" : "MLModelType_mlProgram" + }, + "inputSchema" : [ + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float32", + "formattedType" : "MultiArray (Float32 1 × 1 × 256)", + "shortDescription" : "", + "shape" : "[1, 1, 256]", + "name" : "x_noisy", + "type" : "MultiArray" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float32", + "formattedType" : "MultiArray (Float32 1)", + "shortDescription" : "", + "shape" : "[1]", + "name" : "sigma", + "type" : "MultiArray" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float32", + "formattedType" : "MultiArray (Float32 1 × 512 × 768)", + "shortDescription" : "", + "shape" : "[1, 512, 768]", + "name" : "embedding", + "type" : "MultiArray" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float32", + "formattedType" : "MultiArray (Float32 1 × 256)", + "shortDescription" : "", + "shape" : "[1, 256]", + "name" : "features", + "type" : "MultiArray" + } + ], + "userDefinedMetadata" : { + "com.github.apple.coremltools.source_dialect" : "TorchScript", + "com.github.apple.coremltools.version" : "8.3.0", + "com.github.apple.coremltools.source" : "torch==2.11.0" + }, + "generatedClassName" : "styletts2_ane_diffusion_step", + "method" : "predict" + } +] \ No newline at end of file diff --git a/ANE/styletts2_ane_diffusion_step.mlmodelc/model.mil b/ANE/styletts2_ane_diffusion_step.mlmodelc/model.mil new file mode 100644 index 0000000000000000000000000000000000000000..2cd5afd8368c4b2938373e090d7b2745ef9fc1fb --- /dev/null +++ b/ANE/styletts2_ane_diffusion_step.mlmodelc/model.mil @@ -0,0 +1,351 @@ +program(1.0) +[buildInfo = dict, tensor>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}})] +{ + func main(tensor embedding, tensor features, tensor sigma, tensor x_noisy) { + tensor var_25 = const()[name = tensor("op_25"), val = tensor([-1, 1, 1])]; + tensor sigma_to_fp16_dtype_0 = const()[name = tensor("sigma_to_fp16_dtype_0"), val = tensor("fp16")]; + tensor sigma_to_fp16 = cast(dtype = sigma_to_fp16_dtype_0, x = sigma)[name = tensor("cast_3")]; + tensor s_cast_fp16 = reshape(shape = var_25, x = sigma_to_fp16)[name = tensor("s_cast_fp16")]; + tensor var_27_cast_fp16 = mul(x = s_cast_fp16, y = s_cast_fp16)[name = tensor("op_27_cast_fp16")]; + tensor var_29_to_fp16 = const()[name = tensor("op_29_to_fp16"), val = tensor(0x1.47cp-5)]; + tensor var_30_cast_fp16 = add(x = var_27_cast_fp16, y = var_29_to_fp16)[name = tensor("op_30_cast_fp16")]; + tensor var_31_epsilon_0 = const()[name = tensor("op_31_epsilon_0"), val = tensor(0x1.a36e2ep-14)]; + tensor var_31_cast_fp16 = inverse(epsilon = var_31_epsilon_0, x = var_30_cast_fp16)[name = tensor("op_31_cast_fp16")]; + tensor var_32_to_fp16 = const()[name = tensor("op_32_to_fp16"), val = tensor(0x1.47cp-5)]; + tensor c_skip_cast_fp16 = mul(x = var_31_cast_fp16, y = var_32_to_fp16)[name = tensor("c_skip_cast_fp16")]; + tensor var_34_to_fp16 = const()[name = tensor("op_34_to_fp16"), val = tensor(0x1.998p-3)]; + tensor var_35_cast_fp16 = mul(x = s_cast_fp16, y = var_34_to_fp16)[name = tensor("op_35_cast_fp16")]; + tensor var_40_cast_fp16 = sqrt(x = var_30_cast_fp16)[name = tensor("op_40_cast_fp16")]; + tensor c_out_cast_fp16 = real_div(x = var_35_cast_fp16, y = var_40_cast_fp16)[name = tensor("c_out_cast_fp16")]; + tensor var_47_epsilon_0 = const()[name = tensor("op_47_epsilon_0"), val = tensor(0x1.a36e2ep-14)]; + tensor var_47_cast_fp16 = inverse(epsilon = var_47_epsilon_0, x = var_40_cast_fp16)[name = tensor("op_47_cast_fp16")]; + tensor var_50_epsilon_0 = const()[name = tensor("op_50_epsilon_0"), val = tensor(0x1p-149)]; + tensor var_50_cast_fp16 = log(epsilon = var_50_epsilon_0, x = sigma_to_fp16)[name = tensor("op_50_cast_fp16")]; + tensor var_51_to_fp16 = const()[name = tensor("op_51_to_fp16"), val = tensor(0x1p-2)]; + tensor x_1_cast_fp16 = mul(x = var_50_cast_fp16, y = var_51_to_fp16)[name = tensor("x_1_cast_fp16")]; + tensor x_noisy_to_fp16_dtype_0 = const()[name = tensor("x_noisy_to_fp16_dtype_0"), val = tensor("fp16")]; + tensor x_noisy_to_fp16 = cast(dtype = x_noisy_to_fp16_dtype_0, x = x_noisy)[name = tensor("cast_2")]; + tensor x_11_cast_fp16 = mul(x = var_47_cast_fp16, y = x_noisy_to_fp16)[name = tensor("x_11_cast_fp16")]; + tensor var_55 = const()[name = tensor("op_55"), val = tensor(-1)]; + tensor var_67 = const()[name = tensor("op_67"), val = tensor([1, 1])]; + tensor x_5_cast_fp16 = reshape(shape = var_67, x = x_1_cast_fp16)[name = tensor("x_5_cast_fp16")]; + tensor var_75_to_fp16 = const()[name = tensor("op_75_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(64)))]; + tensor var_76_cast_fp16 = mul(x = x_5_cast_fp16, y = var_75_to_fp16)[name = tensor("op_76_cast_fp16")]; + tensor var_77_promoted_to_fp16 = const()[name = tensor("op_77_promoted_to_fp16"), val = tensor(0x1p+1)]; + tensor var_78_cast_fp16 = mul(x = var_76_cast_fp16, y = var_77_promoted_to_fp16)[name = tensor("op_78_cast_fp16")]; + tensor var_79_to_fp16 = const()[name = tensor("op_79_to_fp16"), val = tensor(0x1.92p+1)]; + tensor freqs_cast_fp16 = mul(x = var_78_cast_fp16, y = var_79_to_fp16)[name = tensor("freqs_cast_fp16")]; + tensor var_81_cast_fp16 = sin(x = freqs_cast_fp16)[name = tensor("op_81_cast_fp16")]; + tensor var_82_cast_fp16 = cos(x = freqs_cast_fp16)[name = tensor("op_82_cast_fp16")]; + tensor input_1_interleave_0 = const()[name = tensor("input_1_interleave_0"), val = tensor(false)]; + tensor input_1_cast_fp16 = concat(axis = var_55, interleave = input_1_interleave_0, values = (x_5_cast_fp16, var_81_cast_fp16, var_82_cast_fp16))[name = tensor("input_1_cast_fp16")]; + tensor transformer_to_time_0_1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(384))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(263616))), name = tensor("transformer_to_time_0_1_weight_to_fp16_palettized"), shape = tensor([1024, 257])]; + tensor transformer_to_time_0_1_bias_to_fp16 = const()[name = tensor("transformer_to_time_0_1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(264192)))]; + tensor linear_0_cast_fp16 = linear(bias = transformer_to_time_0_1_bias_to_fp16, weight = transformer_to_time_0_1_weight_to_fp16_palettized, x = input_1_cast_fp16)[name = tensor("linear_0_cast_fp16")]; + tensor var_88_mode_0 = const()[name = tensor("op_88_mode_0"), val = tensor("EXACT")]; + tensor var_88_cast_fp16 = gelu(mode = var_88_mode_0, x = linear_0_cast_fp16)[name = tensor("op_88_cast_fp16")]; + tensor features_to_fp16_dtype_0 = const()[name = tensor("features_to_fp16_dtype_0"), val = tensor("fp16")]; + tensor transformer_to_features_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(266304))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(528512))), name = tensor("transformer_to_features_0_weight_to_fp16_palettized"), shape = tensor([1024, 256])]; + tensor transformer_to_features_0_bias_to_fp16 = const()[name = tensor("transformer_to_features_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(529088)))]; + tensor features_to_fp16 = cast(dtype = features_to_fp16_dtype_0, x = features)[name = tensor("cast_1")]; + tensor linear_1_cast_fp16 = linear(bias = transformer_to_features_0_bias_to_fp16, weight = transformer_to_features_0_weight_to_fp16_palettized, x = features_to_fp16)[name = tensor("linear_1_cast_fp16")]; + tensor var_94_mode_0 = const()[name = tensor("op_94_mode_0"), val = tensor("EXACT")]; + tensor var_94_cast_fp16 = gelu(mode = var_94_mode_0, x = linear_1_cast_fp16)[name = tensor("op_94_cast_fp16")]; + tensor x_7_axis_0 = const()[name = tensor("x_7_axis_0"), val = tensor(0)]; + tensor x_7_cast_fp16 = stack(axis = x_7_axis_0, values = (var_88_cast_fp16, var_94_cast_fp16))[name = tensor("x_7_cast_fp16")]; + tensor var_101 = const()[name = tensor("op_101"), val = tensor([1, 2, 0])]; + tensor input_7_axes_0 = const()[name = tensor("input_7_axes_0"), val = tensor([2])]; + tensor input_7_keep_dims_0 = const()[name = tensor("input_7_keep_dims_0"), val = tensor(false)]; + tensor x_9_cast_fp16 = transpose(perm = var_101, x = x_7_cast_fp16)[name = tensor("transpose_44")]; + tensor input_7_cast_fp16 = reduce_sum(axes = input_7_axes_0, keep_dims = input_7_keep_dims_0, x = x_9_cast_fp16)[name = tensor("input_7_cast_fp16")]; + tensor transformer_to_mapping_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(531200))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1579840))), name = tensor("transformer_to_mapping_0_weight_to_fp16_palettized"), shape = tensor([1024, 1024])]; + tensor transformer_to_mapping_0_bias_to_fp16 = const()[name = tensor("transformer_to_mapping_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1580416)))]; + tensor linear_2_cast_fp16 = linear(bias = transformer_to_mapping_0_bias_to_fp16, weight = transformer_to_mapping_0_weight_to_fp16_palettized, x = input_7_cast_fp16)[name = tensor("linear_2_cast_fp16")]; + tensor input_11_mode_0 = const()[name = tensor("input_11_mode_0"), val = tensor("EXACT")]; + tensor input_11_cast_fp16 = gelu(mode = input_11_mode_0, x = linear_2_cast_fp16)[name = tensor("input_11_cast_fp16")]; + tensor transformer_to_mapping_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1582528))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2631168))), name = tensor("transformer_to_mapping_2_weight_to_fp16_palettized"), shape = tensor([1024, 1024])]; + tensor transformer_to_mapping_2_bias_to_fp16 = const()[name = tensor("transformer_to_mapping_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2631744)))]; + tensor linear_3_cast_fp16 = linear(bias = transformer_to_mapping_2_bias_to_fp16, weight = transformer_to_mapping_2_weight_to_fp16_palettized, x = input_11_cast_fp16)[name = tensor("linear_3_cast_fp16")]; + tensor mapping_1_mode_0 = const()[name = tensor("mapping_1_mode_0"), val = tensor("EXACT")]; + tensor mapping_1_cast_fp16 = gelu(mode = mapping_1_mode_0, x = linear_3_cast_fp16)[name = tensor("mapping_1_cast_fp16")]; + tensor var_127_reps_0 = const()[name = tensor("op_127_reps_0"), val = tensor([1, 512, 1])]; + tensor var_127_cast_fp16 = tile(reps = var_127_reps_0, x = x_11_cast_fp16)[name = tensor("op_127_cast_fp16")]; + tensor var_129 = const()[name = tensor("op_129"), val = tensor(-1)]; + tensor x_13_interleave_0 = const()[name = tensor("x_13_interleave_0"), val = tensor(false)]; + tensor embedding_to_fp16_dtype_0 = const()[name = tensor("embedding_to_fp16_dtype_0"), val = tensor("fp16")]; + tensor embedding_to_fp16 = cast(dtype = embedding_to_fp16_dtype_0, x = embedding)[name = tensor("cast_0")]; + tensor x_13_cast_fp16 = concat(axis = var_129, interleave = x_13_interleave_0, values = (var_127_cast_fp16, embedding_to_fp16))[name = tensor("x_13_cast_fp16")]; + tensor var_132_axes_0 = const()[name = tensor("op_132_axes_0"), val = tensor([1])]; + tensor var_132_cast_fp16 = expand_dims(axes = var_132_axes_0, x = mapping_1_cast_fp16)[name = tensor("op_132_cast_fp16")]; + tensor mapping_reps_0 = const()[name = tensor("mapping_reps_0"), val = tensor([1, 512, 1])]; + tensor mapping_cast_fp16 = tile(reps = mapping_reps_0, x = var_132_cast_fp16)[name = tensor("mapping_cast_fp16")]; + tensor x_15_cast_fp16 = add(x = x_13_cast_fp16, y = mapping_cast_fp16)[name = tensor("x_15_cast_fp16")]; + tensor var_153 = const()[name = tensor("op_153"), val = tensor(-1)]; + tensor transformer_blocks_0_attention_norm_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2633856))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3158208))), name = tensor("transformer_blocks_0_attention_norm_fc_weight_to_fp16_palettized"), shape = tensor([2048, 256])]; + tensor transformer_blocks_0_attention_norm_fc_bias_to_fp16 = const()[name = tensor("transformer_blocks_0_attention_norm_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3158784)))]; + tensor linear_4_cast_fp16 = linear(bias = transformer_blocks_0_attention_norm_fc_bias_to_fp16, weight = transformer_blocks_0_attention_norm_fc_weight_to_fp16_palettized, x = features_to_fp16)[name = tensor("linear_4_cast_fp16")]; + tensor var_172 = const()[name = tensor("op_172"), val = tensor([1, 2048, 1])]; + tensor h_3_cast_fp16 = reshape(shape = var_172, x = linear_4_cast_fp16)[name = tensor("h_3_cast_fp16")]; + tensor var_174_split_sizes_0 = const()[name = tensor("op_174_split_sizes_0"), val = tensor([1024, 1024])]; + tensor var_174_axis_0 = const()[name = tensor("op_174_axis_0"), val = tensor(1)]; + tensor var_174_cast_fp16_0, tensor var_174_cast_fp16_1 = split(axis = var_174_axis_0, split_sizes = var_174_split_sizes_0, x = h_3_cast_fp16)[name = tensor("op_174_cast_fp16")]; + tensor gamma_3_perm_0 = const()[name = tensor("gamma_3_perm_0"), val = tensor([0, -1, 1])]; + tensor beta_3_perm_0 = const()[name = tensor("beta_3_perm_0"), val = tensor([0, -1, 1])]; + tensor x_19_axes_0 = const()[name = tensor("x_19_axes_0"), val = tensor([-1])]; + tensor var_146_to_fp16 = const()[name = tensor("op_146_to_fp16"), val = tensor(0x1.5p-17)]; + tensor x_19_cast_fp16 = layer_norm(axes = x_19_axes_0, epsilon = var_146_to_fp16, x = x_15_cast_fp16)[name = tensor("x_19_cast_fp16")]; + tensor var_180_promoted_to_fp16 = const()[name = tensor("op_180_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor gamma_3_cast_fp16 = transpose(perm = gamma_3_perm_0, x = var_174_cast_fp16_0)[name = tensor("transpose_43")]; + tensor var_181_cast_fp16 = add(x = gamma_3_cast_fp16, y = var_180_promoted_to_fp16)[name = tensor("op_181_cast_fp16")]; + tensor var_182_cast_fp16 = mul(x = var_181_cast_fp16, y = x_19_cast_fp16)[name = tensor("op_182_cast_fp16")]; + tensor beta_3_cast_fp16 = transpose(perm = beta_3_perm_0, x = var_174_cast_fp16_1)[name = tensor("transpose_42")]; + tensor x_21_cast_fp16 = add(x = var_182_cast_fp16, y = beta_3_cast_fp16)[name = tensor("x_21_cast_fp16")]; + tensor transformer_blocks_0_attention_norm_context_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3162944))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3687296))), name = tensor("transformer_blocks_0_attention_norm_context_fc_weight_to_fp16_palettized"), shape = tensor([2048, 256])]; + tensor transformer_blocks_0_attention_norm_context_fc_bias_to_fp16 = const()[name = tensor("transformer_blocks_0_attention_norm_context_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3687872)))]; + tensor linear_5_cast_fp16 = linear(bias = transformer_blocks_0_attention_norm_context_fc_bias_to_fp16, weight = transformer_blocks_0_attention_norm_context_fc_weight_to_fp16_palettized, x = features_to_fp16)[name = tensor("linear_5_cast_fp16")]; + tensor var_194 = const()[name = tensor("op_194"), val = tensor([1, 2048, 1])]; + tensor h_7_cast_fp16 = reshape(shape = var_194, x = linear_5_cast_fp16)[name = tensor("h_7_cast_fp16")]; + tensor var_196_split_sizes_0 = const()[name = tensor("op_196_split_sizes_0"), val = tensor([1024, 1024])]; + tensor var_196_axis_0 = const()[name = tensor("op_196_axis_0"), val = tensor(1)]; + tensor var_196_cast_fp16_0, tensor var_196_cast_fp16_1 = split(axis = var_196_axis_0, split_sizes = var_196_split_sizes_0, x = h_7_cast_fp16)[name = tensor("op_196_cast_fp16")]; + tensor gamma_7_perm_0 = const()[name = tensor("gamma_7_perm_0"), val = tensor([0, -1, 1])]; + tensor beta_7_perm_0 = const()[name = tensor("beta_7_perm_0"), val = tensor([0, -1, 1])]; + tensor var_202_promoted_to_fp16 = const()[name = tensor("op_202_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor gamma_7_cast_fp16 = transpose(perm = gamma_7_perm_0, x = var_196_cast_fp16_0)[name = tensor("transpose_41")]; + tensor var_203_cast_fp16 = add(x = gamma_7_cast_fp16, y = var_202_promoted_to_fp16)[name = tensor("op_203_cast_fp16")]; + tensor var_204_cast_fp16 = mul(x = var_203_cast_fp16, y = x_19_cast_fp16)[name = tensor("op_204_cast_fp16")]; + tensor beta_7_cast_fp16 = transpose(perm = beta_7_perm_0, x = var_196_cast_fp16_1)[name = tensor("transpose_40")]; + tensor x_27_cast_fp16 = add(x = var_204_cast_fp16, y = beta_7_cast_fp16)[name = tensor("x_27_cast_fp16")]; + tensor transformer_blocks_0_attention_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3692032))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4216384))), name = tensor("transformer_blocks_0_attention_to_q_weight_to_fp16_palettized"), shape = tensor([512, 1024])]; + tensor linear_6_bias_0_to_fp16 = const()[name = tensor("linear_6_bias_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4216960)))]; + tensor linear_6_cast_fp16 = linear(bias = linear_6_bias_0_to_fp16, weight = transformer_blocks_0_attention_to_q_weight_to_fp16_palettized, x = x_21_cast_fp16)[name = tensor("linear_6_cast_fp16")]; + tensor transformer_blocks_0_attention_to_kv_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4218048))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5266688))), name = tensor("transformer_blocks_0_attention_to_kv_weight_to_fp16_palettized"), shape = tensor([1024, 1024])]; + tensor linear_7_bias_0_to_fp16 = const()[name = tensor("linear_7_bias_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5267264)))]; + tensor linear_7_cast_fp16 = linear(bias = linear_7_bias_0_to_fp16, weight = transformer_blocks_0_attention_to_kv_weight_to_fp16_palettized, x = x_27_cast_fp16)[name = tensor("linear_7_cast_fp16")]; + tensor var_212_split_sizes_0 = const()[name = tensor("op_212_split_sizes_0"), val = tensor([512, 512])]; + tensor var_212_axis_0 = const()[name = tensor("op_212_axis_0"), val = tensor(-1)]; + tensor var_212_cast_fp16_0, tensor var_212_cast_fp16_1 = split(axis = var_212_axis_0, split_sizes = var_212_split_sizes_0, x = linear_7_cast_fp16)[name = tensor("op_212_cast_fp16")]; + tensor var_221 = const()[name = tensor("op_221"), val = tensor([1, 512, 8, 64])]; + tensor x_31_cast_fp16 = reshape(shape = var_221, x = linear_6_cast_fp16)[name = tensor("x_31_cast_fp16")]; + tensor var_231 = const()[name = tensor("op_231"), val = tensor([1, 512, 8, 64])]; + tensor x_35_cast_fp16 = reshape(shape = var_231, x = var_212_cast_fp16_0)[name = tensor("x_35_cast_fp16")]; + tensor var_241 = const()[name = tensor("op_241"), val = tensor([1, 512, 8, 64])]; + tensor x_39_cast_fp16 = reshape(shape = var_241, x = var_212_cast_fp16_1)[name = tensor("x_39_cast_fp16")]; + tensor var_243 = const()[name = tensor("op_243"), val = tensor([0, 2, -3, -1])]; + tensor sim_1_transpose_x_0 = const()[name = tensor("sim_1_transpose_x_0"), val = tensor(false)]; + tensor sim_1_transpose_y_0 = const()[name = tensor("sim_1_transpose_y_0"), val = tensor(false)]; + tensor transpose_12_perm_0 = const()[name = tensor("transpose_12_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_13_perm_0 = const()[name = tensor("transpose_13_perm_0"), val = tensor([0, 2, -1, -3])]; + tensor transpose_13 = transpose(perm = transpose_13_perm_0, x = x_35_cast_fp16)[name = tensor("transpose_38")]; + tensor transpose_12 = transpose(perm = transpose_12_perm_0, x = x_31_cast_fp16)[name = tensor("transpose_39")]; + tensor sim_1_cast_fp16 = matmul(transpose_x = sim_1_transpose_x_0, transpose_y = sim_1_transpose_y_0, x = transpose_12, y = transpose_13)[name = tensor("sim_1_cast_fp16")]; + tensor var_247_to_fp16 = const()[name = tensor("op_247_to_fp16"), val = tensor(0x1p-3)]; + tensor sim_3_cast_fp16 = mul(x = sim_1_cast_fp16, y = var_247_to_fp16)[name = tensor("sim_3_cast_fp16")]; + tensor attn_1_cast_fp16 = softmax(axis = var_153, x = sim_3_cast_fp16)[name = tensor("attn_1_cast_fp16")]; + tensor x_41_transpose_x_0 = const()[name = tensor("x_41_transpose_x_0"), val = tensor(false)]; + tensor x_41_transpose_y_0 = const()[name = tensor("x_41_transpose_y_0"), val = tensor(false)]; + tensor v_1_cast_fp16 = transpose(perm = var_243, x = x_39_cast_fp16)[name = tensor("transpose_37")]; + tensor x_41_cast_fp16 = matmul(transpose_x = x_41_transpose_x_0, transpose_y = x_41_transpose_y_0, x = attn_1_cast_fp16, y = v_1_cast_fp16)[name = tensor("x_41_cast_fp16")]; + tensor var_269 = const()[name = tensor("op_269"), val = tensor([0, 2, 1, 3])]; + tensor var_271 = const()[name = tensor("op_271"), val = tensor([1, 512, 512])]; + tensor x_43_cast_fp16 = transpose(perm = var_269, x = x_41_cast_fp16)[name = tensor("transpose_36")]; + tensor input_23_cast_fp16 = reshape(shape = var_271, x = x_43_cast_fp16)[name = tensor("input_23_cast_fp16")]; + tensor transformer_blocks_0_attention_attention_to_out_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5269376))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5793728))), name = tensor("transformer_blocks_0_attention_attention_to_out_weight_to_fp16_palettized"), shape = tensor([1024, 512])]; + tensor transformer_blocks_0_attention_attention_to_out_bias_to_fp16 = const()[name = tensor("transformer_blocks_0_attention_attention_to_out_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5794304)))]; + tensor linear_8_cast_fp16 = linear(bias = transformer_blocks_0_attention_attention_to_out_bias_to_fp16, weight = transformer_blocks_0_attention_attention_to_out_weight_to_fp16_palettized, x = input_23_cast_fp16)[name = tensor("linear_8_cast_fp16")]; + tensor input_25_cast_fp16 = add(x = linear_8_cast_fp16, y = x_15_cast_fp16)[name = tensor("input_25_cast_fp16")]; + tensor transformer_blocks_0_feed_forward_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5796416))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7893632))), name = tensor("transformer_blocks_0_feed_forward_0_weight_to_fp16_palettized"), shape = tensor([2048, 1024])]; + tensor transformer_blocks_0_feed_forward_0_bias_to_fp16 = const()[name = tensor("transformer_blocks_0_feed_forward_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7894208)))]; + tensor linear_9_cast_fp16 = linear(bias = transformer_blocks_0_feed_forward_0_bias_to_fp16, weight = transformer_blocks_0_feed_forward_0_weight_to_fp16_palettized, x = input_25_cast_fp16)[name = tensor("linear_9_cast_fp16")]; + tensor input_29_mode_0 = const()[name = tensor("input_29_mode_0"), val = tensor("EXACT")]; + tensor input_29_cast_fp16 = gelu(mode = input_29_mode_0, x = linear_9_cast_fp16)[name = tensor("input_29_cast_fp16")]; + tensor transformer_blocks_0_feed_forward_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7898368))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9995584))), name = tensor("transformer_blocks_0_feed_forward_2_weight_to_fp16_palettized"), shape = tensor([1024, 2048])]; + tensor transformer_blocks_0_feed_forward_2_bias_to_fp16 = const()[name = tensor("transformer_blocks_0_feed_forward_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9996160)))]; + tensor linear_10_cast_fp16 = linear(bias = transformer_blocks_0_feed_forward_2_bias_to_fp16, weight = transformer_blocks_0_feed_forward_2_weight_to_fp16_palettized, x = input_29_cast_fp16)[name = tensor("linear_10_cast_fp16")]; + tensor x_45_cast_fp16 = add(x = linear_10_cast_fp16, y = input_25_cast_fp16)[name = tensor("x_45_cast_fp16")]; + tensor x_47_cast_fp16 = add(x = x_45_cast_fp16, y = mapping_cast_fp16)[name = tensor("x_47_cast_fp16")]; + tensor var_298 = const()[name = tensor("op_298"), val = tensor(-1)]; + tensor transformer_blocks_1_attention_norm_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9998272))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(10522624))), name = tensor("transformer_blocks_1_attention_norm_fc_weight_to_fp16_palettized"), shape = tensor([2048, 256])]; + tensor transformer_blocks_1_attention_norm_fc_bias_to_fp16 = const()[name = tensor("transformer_blocks_1_attention_norm_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(10523200)))]; + tensor linear_11_cast_fp16 = linear(bias = transformer_blocks_1_attention_norm_fc_bias_to_fp16, weight = transformer_blocks_1_attention_norm_fc_weight_to_fp16_palettized, x = features_to_fp16)[name = tensor("linear_11_cast_fp16")]; + tensor var_317 = const()[name = tensor("op_317"), val = tensor([1, 2048, 1])]; + tensor h_11_cast_fp16 = reshape(shape = var_317, x = linear_11_cast_fp16)[name = tensor("h_11_cast_fp16")]; + tensor var_319_split_sizes_0 = const()[name = tensor("op_319_split_sizes_0"), val = tensor([1024, 1024])]; + tensor var_319_axis_0 = const()[name = tensor("op_319_axis_0"), val = tensor(1)]; + tensor var_319_cast_fp16_0, tensor var_319_cast_fp16_1 = split(axis = var_319_axis_0, split_sizes = var_319_split_sizes_0, x = h_11_cast_fp16)[name = tensor("op_319_cast_fp16")]; + tensor gamma_11_perm_0 = const()[name = tensor("gamma_11_perm_0"), val = tensor([0, -1, 1])]; + tensor beta_11_perm_0 = const()[name = tensor("beta_11_perm_0"), val = tensor([0, -1, 1])]; + tensor x_51_axes_0 = const()[name = tensor("x_51_axes_0"), val = tensor([-1])]; + tensor var_291_to_fp16 = const()[name = tensor("op_291_to_fp16"), val = tensor(0x1.5p-17)]; + tensor x_51_cast_fp16 = layer_norm(axes = x_51_axes_0, epsilon = var_291_to_fp16, x = x_47_cast_fp16)[name = tensor("x_51_cast_fp16")]; + tensor var_325_promoted_to_fp16 = const()[name = tensor("op_325_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor gamma_11_cast_fp16 = transpose(perm = gamma_11_perm_0, x = var_319_cast_fp16_0)[name = tensor("transpose_35")]; + tensor var_326_cast_fp16 = add(x = gamma_11_cast_fp16, y = var_325_promoted_to_fp16)[name = tensor("op_326_cast_fp16")]; + tensor var_327_cast_fp16 = mul(x = var_326_cast_fp16, y = x_51_cast_fp16)[name = tensor("op_327_cast_fp16")]; + tensor beta_11_cast_fp16 = transpose(perm = beta_11_perm_0, x = var_319_cast_fp16_1)[name = tensor("transpose_34")]; + tensor x_53_cast_fp16 = add(x = var_327_cast_fp16, y = beta_11_cast_fp16)[name = tensor("x_53_cast_fp16")]; + tensor transformer_blocks_1_attention_norm_context_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(10527360))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11051712))), name = tensor("transformer_blocks_1_attention_norm_context_fc_weight_to_fp16_palettized"), shape = tensor([2048, 256])]; + tensor transformer_blocks_1_attention_norm_context_fc_bias_to_fp16 = const()[name = tensor("transformer_blocks_1_attention_norm_context_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11052288)))]; + tensor linear_12_cast_fp16 = linear(bias = transformer_blocks_1_attention_norm_context_fc_bias_to_fp16, weight = transformer_blocks_1_attention_norm_context_fc_weight_to_fp16_palettized, x = features_to_fp16)[name = tensor("linear_12_cast_fp16")]; + tensor var_339 = const()[name = tensor("op_339"), val = tensor([1, 2048, 1])]; + tensor h_15_cast_fp16 = reshape(shape = var_339, x = linear_12_cast_fp16)[name = tensor("h_15_cast_fp16")]; + tensor var_341_split_sizes_0 = const()[name = tensor("op_341_split_sizes_0"), val = tensor([1024, 1024])]; + tensor var_341_axis_0 = const()[name = tensor("op_341_axis_0"), val = tensor(1)]; + tensor var_341_cast_fp16_0, tensor var_341_cast_fp16_1 = split(axis = var_341_axis_0, split_sizes = var_341_split_sizes_0, x = h_15_cast_fp16)[name = tensor("op_341_cast_fp16")]; + tensor gamma_15_perm_0 = const()[name = tensor("gamma_15_perm_0"), val = tensor([0, -1, 1])]; + tensor beta_15_perm_0 = const()[name = tensor("beta_15_perm_0"), val = tensor([0, -1, 1])]; + tensor var_347_promoted_to_fp16 = const()[name = tensor("op_347_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor gamma_15_cast_fp16 = transpose(perm = gamma_15_perm_0, x = var_341_cast_fp16_0)[name = tensor("transpose_33")]; + tensor var_348_cast_fp16 = add(x = gamma_15_cast_fp16, y = var_347_promoted_to_fp16)[name = tensor("op_348_cast_fp16")]; + tensor var_349_cast_fp16 = mul(x = var_348_cast_fp16, y = x_51_cast_fp16)[name = tensor("op_349_cast_fp16")]; + tensor beta_15_cast_fp16 = transpose(perm = beta_15_perm_0, x = var_341_cast_fp16_1)[name = tensor("transpose_32")]; + tensor x_59_cast_fp16 = add(x = var_349_cast_fp16, y = beta_15_cast_fp16)[name = tensor("x_59_cast_fp16")]; + tensor transformer_blocks_1_attention_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11056448))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11580800))), name = tensor("transformer_blocks_1_attention_to_q_weight_to_fp16_palettized"), shape = tensor([512, 1024])]; + tensor linear_13_cast_fp16 = linear(bias = linear_6_bias_0_to_fp16, weight = transformer_blocks_1_attention_to_q_weight_to_fp16_palettized, x = x_53_cast_fp16)[name = tensor("linear_13_cast_fp16")]; + tensor transformer_blocks_1_attention_to_kv_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11581376))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(12630016))), name = tensor("transformer_blocks_1_attention_to_kv_weight_to_fp16_palettized"), shape = tensor([1024, 1024])]; + tensor linear_14_cast_fp16 = linear(bias = linear_7_bias_0_to_fp16, weight = transformer_blocks_1_attention_to_kv_weight_to_fp16_palettized, x = x_59_cast_fp16)[name = tensor("linear_14_cast_fp16")]; + tensor var_357_split_sizes_0 = const()[name = tensor("op_357_split_sizes_0"), val = tensor([512, 512])]; + tensor var_357_axis_0 = const()[name = tensor("op_357_axis_0"), val = tensor(-1)]; + tensor var_357_cast_fp16_0, tensor var_357_cast_fp16_1 = split(axis = var_357_axis_0, split_sizes = var_357_split_sizes_0, x = linear_14_cast_fp16)[name = tensor("op_357_cast_fp16")]; + tensor var_366 = const()[name = tensor("op_366"), val = tensor([1, 512, 8, 64])]; + tensor x_63_cast_fp16 = reshape(shape = var_366, x = linear_13_cast_fp16)[name = tensor("x_63_cast_fp16")]; + tensor var_376 = const()[name = tensor("op_376"), val = tensor([1, 512, 8, 64])]; + tensor x_67_cast_fp16 = reshape(shape = var_376, x = var_357_cast_fp16_0)[name = tensor("x_67_cast_fp16")]; + tensor var_386 = const()[name = tensor("op_386"), val = tensor([1, 512, 8, 64])]; + tensor x_71_cast_fp16 = reshape(shape = var_386, x = var_357_cast_fp16_1)[name = tensor("x_71_cast_fp16")]; + tensor var_388 = const()[name = tensor("op_388"), val = tensor([0, 2, -3, -1])]; + tensor sim_5_transpose_x_0 = const()[name = tensor("sim_5_transpose_x_0"), val = tensor(false)]; + tensor sim_5_transpose_y_0 = const()[name = tensor("sim_5_transpose_y_0"), val = tensor(false)]; + tensor transpose_14_perm_0 = const()[name = tensor("transpose_14_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_15_perm_0 = const()[name = tensor("transpose_15_perm_0"), val = tensor([0, 2, -1, -3])]; + tensor transpose_15 = transpose(perm = transpose_15_perm_0, x = x_67_cast_fp16)[name = tensor("transpose_30")]; + tensor transpose_14 = transpose(perm = transpose_14_perm_0, x = x_63_cast_fp16)[name = tensor("transpose_31")]; + tensor sim_5_cast_fp16 = matmul(transpose_x = sim_5_transpose_x_0, transpose_y = sim_5_transpose_y_0, x = transpose_14, y = transpose_15)[name = tensor("sim_5_cast_fp16")]; + tensor var_392_to_fp16 = const()[name = tensor("op_392_to_fp16"), val = tensor(0x1p-3)]; + tensor sim_7_cast_fp16 = mul(x = sim_5_cast_fp16, y = var_392_to_fp16)[name = tensor("sim_7_cast_fp16")]; + tensor attn_3_cast_fp16 = softmax(axis = var_298, x = sim_7_cast_fp16)[name = tensor("attn_3_cast_fp16")]; + tensor x_73_transpose_x_0 = const()[name = tensor("x_73_transpose_x_0"), val = tensor(false)]; + tensor x_73_transpose_y_0 = const()[name = tensor("x_73_transpose_y_0"), val = tensor(false)]; + tensor v_3_cast_fp16 = transpose(perm = var_388, x = x_71_cast_fp16)[name = tensor("transpose_29")]; + tensor x_73_cast_fp16 = matmul(transpose_x = x_73_transpose_x_0, transpose_y = x_73_transpose_y_0, x = attn_3_cast_fp16, y = v_3_cast_fp16)[name = tensor("x_73_cast_fp16")]; + tensor var_414 = const()[name = tensor("op_414"), val = tensor([0, 2, 1, 3])]; + tensor var_416 = const()[name = tensor("op_416"), val = tensor([1, 512, 512])]; + tensor x_75_cast_fp16 = transpose(perm = var_414, x = x_73_cast_fp16)[name = tensor("transpose_28")]; + tensor input_39_cast_fp16 = reshape(shape = var_416, x = x_75_cast_fp16)[name = tensor("input_39_cast_fp16")]; + tensor transformer_blocks_1_attention_attention_to_out_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(12630592))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(13154944))), name = tensor("transformer_blocks_1_attention_attention_to_out_weight_to_fp16_palettized"), shape = tensor([1024, 512])]; + tensor transformer_blocks_1_attention_attention_to_out_bias_to_fp16 = const()[name = tensor("transformer_blocks_1_attention_attention_to_out_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(13155520)))]; + tensor linear_15_cast_fp16 = linear(bias = transformer_blocks_1_attention_attention_to_out_bias_to_fp16, weight = transformer_blocks_1_attention_attention_to_out_weight_to_fp16_palettized, x = input_39_cast_fp16)[name = tensor("linear_15_cast_fp16")]; + tensor input_41_cast_fp16 = add(x = linear_15_cast_fp16, y = x_47_cast_fp16)[name = tensor("input_41_cast_fp16")]; + tensor transformer_blocks_1_feed_forward_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(13157632))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15254848))), name = tensor("transformer_blocks_1_feed_forward_0_weight_to_fp16_palettized"), shape = tensor([2048, 1024])]; + tensor transformer_blocks_1_feed_forward_0_bias_to_fp16 = const()[name = tensor("transformer_blocks_1_feed_forward_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15255424)))]; + tensor linear_16_cast_fp16 = linear(bias = transformer_blocks_1_feed_forward_0_bias_to_fp16, weight = transformer_blocks_1_feed_forward_0_weight_to_fp16_palettized, x = input_41_cast_fp16)[name = tensor("linear_16_cast_fp16")]; + tensor input_45_mode_0 = const()[name = tensor("input_45_mode_0"), val = tensor("EXACT")]; + tensor input_45_cast_fp16 = gelu(mode = input_45_mode_0, x = linear_16_cast_fp16)[name = tensor("input_45_cast_fp16")]; + tensor transformer_blocks_1_feed_forward_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15259584))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17356800))), name = tensor("transformer_blocks_1_feed_forward_2_weight_to_fp16_palettized"), shape = tensor([1024, 2048])]; + tensor transformer_blocks_1_feed_forward_2_bias_to_fp16 = const()[name = tensor("transformer_blocks_1_feed_forward_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17357376)))]; + tensor linear_17_cast_fp16 = linear(bias = transformer_blocks_1_feed_forward_2_bias_to_fp16, weight = transformer_blocks_1_feed_forward_2_weight_to_fp16_palettized, x = input_45_cast_fp16)[name = tensor("linear_17_cast_fp16")]; + tensor x_77_cast_fp16 = add(x = linear_17_cast_fp16, y = input_41_cast_fp16)[name = tensor("x_77_cast_fp16")]; + tensor x_79_cast_fp16 = add(x = x_77_cast_fp16, y = mapping_cast_fp16)[name = tensor("x_79_cast_fp16")]; + tensor var_443 = const()[name = tensor("op_443"), val = tensor(-1)]; + tensor transformer_blocks_2_attention_norm_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17359488))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17883840))), name = tensor("transformer_blocks_2_attention_norm_fc_weight_to_fp16_palettized"), shape = tensor([2048, 256])]; + tensor transformer_blocks_2_attention_norm_fc_bias_to_fp16 = const()[name = tensor("transformer_blocks_2_attention_norm_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17884416)))]; + tensor linear_18_cast_fp16 = linear(bias = transformer_blocks_2_attention_norm_fc_bias_to_fp16, weight = transformer_blocks_2_attention_norm_fc_weight_to_fp16_palettized, x = features_to_fp16)[name = tensor("linear_18_cast_fp16")]; + tensor var_462 = const()[name = tensor("op_462"), val = tensor([1, 2048, 1])]; + tensor h_19_cast_fp16 = reshape(shape = var_462, x = linear_18_cast_fp16)[name = tensor("h_19_cast_fp16")]; + tensor var_464_split_sizes_0 = const()[name = tensor("op_464_split_sizes_0"), val = tensor([1024, 1024])]; + tensor var_464_axis_0 = const()[name = tensor("op_464_axis_0"), val = tensor(1)]; + tensor var_464_cast_fp16_0, tensor var_464_cast_fp16_1 = split(axis = var_464_axis_0, split_sizes = var_464_split_sizes_0, x = h_19_cast_fp16)[name = tensor("op_464_cast_fp16")]; + tensor gamma_19_perm_0 = const()[name = tensor("gamma_19_perm_0"), val = tensor([0, -1, 1])]; + tensor beta_19_perm_0 = const()[name = tensor("beta_19_perm_0"), val = tensor([0, -1, 1])]; + tensor x_83_axes_0 = const()[name = tensor("x_83_axes_0"), val = tensor([-1])]; + tensor var_436_to_fp16 = const()[name = tensor("op_436_to_fp16"), val = tensor(0x1.5p-17)]; + tensor x_83_cast_fp16 = layer_norm(axes = x_83_axes_0, epsilon = var_436_to_fp16, x = x_79_cast_fp16)[name = tensor("x_83_cast_fp16")]; + tensor var_470_promoted_to_fp16 = const()[name = tensor("op_470_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor gamma_19_cast_fp16 = transpose(perm = gamma_19_perm_0, x = var_464_cast_fp16_0)[name = tensor("transpose_27")]; + tensor var_471_cast_fp16 = add(x = gamma_19_cast_fp16, y = var_470_promoted_to_fp16)[name = tensor("op_471_cast_fp16")]; + tensor var_472_cast_fp16 = mul(x = var_471_cast_fp16, y = x_83_cast_fp16)[name = tensor("op_472_cast_fp16")]; + tensor beta_19_cast_fp16 = transpose(perm = beta_19_perm_0, x = var_464_cast_fp16_1)[name = tensor("transpose_26")]; + tensor x_85_cast_fp16 = add(x = var_472_cast_fp16, y = beta_19_cast_fp16)[name = tensor("x_85_cast_fp16")]; + tensor transformer_blocks_2_attention_norm_context_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17888576))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(18412928))), name = tensor("transformer_blocks_2_attention_norm_context_fc_weight_to_fp16_palettized"), shape = tensor([2048, 256])]; + tensor transformer_blocks_2_attention_norm_context_fc_bias_to_fp16 = const()[name = tensor("transformer_blocks_2_attention_norm_context_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(18413504)))]; + tensor linear_19_cast_fp16 = linear(bias = transformer_blocks_2_attention_norm_context_fc_bias_to_fp16, weight = transformer_blocks_2_attention_norm_context_fc_weight_to_fp16_palettized, x = features_to_fp16)[name = tensor("linear_19_cast_fp16")]; + tensor var_484 = const()[name = tensor("op_484"), val = tensor([1, 2048, 1])]; + tensor h_cast_fp16 = reshape(shape = var_484, x = linear_19_cast_fp16)[name = tensor("h_cast_fp16")]; + tensor var_486_split_sizes_0 = const()[name = tensor("op_486_split_sizes_0"), val = tensor([1024, 1024])]; + tensor var_486_axis_0 = const()[name = tensor("op_486_axis_0"), val = tensor(1)]; + tensor var_486_cast_fp16_0, tensor var_486_cast_fp16_1 = split(axis = var_486_axis_0, split_sizes = var_486_split_sizes_0, x = h_cast_fp16)[name = tensor("op_486_cast_fp16")]; + tensor gamma_perm_0 = const()[name = tensor("gamma_perm_0"), val = tensor([0, -1, 1])]; + tensor beta_perm_0 = const()[name = tensor("beta_perm_0"), val = tensor([0, -1, 1])]; + tensor var_492_promoted_to_fp16 = const()[name = tensor("op_492_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor gamma_cast_fp16 = transpose(perm = gamma_perm_0, x = var_486_cast_fp16_0)[name = tensor("transpose_25")]; + tensor var_493_cast_fp16 = add(x = gamma_cast_fp16, y = var_492_promoted_to_fp16)[name = tensor("op_493_cast_fp16")]; + tensor var_494_cast_fp16 = mul(x = var_493_cast_fp16, y = x_83_cast_fp16)[name = tensor("op_494_cast_fp16")]; + tensor beta_cast_fp16 = transpose(perm = beta_perm_0, x = var_486_cast_fp16_1)[name = tensor("transpose_24")]; + tensor x_91_cast_fp16 = add(x = var_494_cast_fp16, y = beta_cast_fp16)[name = tensor("x_91_cast_fp16")]; + tensor transformer_blocks_2_attention_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(18417664))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(18942016))), name = tensor("transformer_blocks_2_attention_to_q_weight_to_fp16_palettized"), shape = tensor([512, 1024])]; + tensor linear_20_cast_fp16 = linear(bias = linear_6_bias_0_to_fp16, weight = transformer_blocks_2_attention_to_q_weight_to_fp16_palettized, x = x_85_cast_fp16)[name = tensor("linear_20_cast_fp16")]; + tensor transformer_blocks_2_attention_to_kv_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(18942592))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19991232))), name = tensor("transformer_blocks_2_attention_to_kv_weight_to_fp16_palettized"), shape = tensor([1024, 1024])]; + tensor linear_21_cast_fp16 = linear(bias = linear_7_bias_0_to_fp16, weight = transformer_blocks_2_attention_to_kv_weight_to_fp16_palettized, x = x_91_cast_fp16)[name = tensor("linear_21_cast_fp16")]; + tensor var_502_split_sizes_0 = const()[name = tensor("op_502_split_sizes_0"), val = tensor([512, 512])]; + tensor var_502_axis_0 = const()[name = tensor("op_502_axis_0"), val = tensor(-1)]; + tensor var_502_cast_fp16_0, tensor var_502_cast_fp16_1 = split(axis = var_502_axis_0, split_sizes = var_502_split_sizes_0, x = linear_21_cast_fp16)[name = tensor("op_502_cast_fp16")]; + tensor var_511 = const()[name = tensor("op_511"), val = tensor([1, 512, 8, 64])]; + tensor x_95_cast_fp16 = reshape(shape = var_511, x = linear_20_cast_fp16)[name = tensor("x_95_cast_fp16")]; + tensor var_521 = const()[name = tensor("op_521"), val = tensor([1, 512, 8, 64])]; + tensor x_99_cast_fp16 = reshape(shape = var_521, x = var_502_cast_fp16_0)[name = tensor("x_99_cast_fp16")]; + tensor var_531 = const()[name = tensor("op_531"), val = tensor([1, 512, 8, 64])]; + tensor x_103_cast_fp16 = reshape(shape = var_531, x = var_502_cast_fp16_1)[name = tensor("x_103_cast_fp16")]; + tensor var_533 = const()[name = tensor("op_533"), val = tensor([0, 2, -3, -1])]; + tensor sim_9_transpose_x_0 = const()[name = tensor("sim_9_transpose_x_0"), val = tensor(false)]; + tensor sim_9_transpose_y_0 = const()[name = tensor("sim_9_transpose_y_0"), val = tensor(false)]; + tensor transpose_16_perm_0 = const()[name = tensor("transpose_16_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_17_perm_0 = const()[name = tensor("transpose_17_perm_0"), val = tensor([0, 2, -1, -3])]; + tensor transpose_17 = transpose(perm = transpose_17_perm_0, x = x_99_cast_fp16)[name = tensor("transpose_22")]; + tensor transpose_16 = transpose(perm = transpose_16_perm_0, x = x_95_cast_fp16)[name = tensor("transpose_23")]; + tensor sim_9_cast_fp16 = matmul(transpose_x = sim_9_transpose_x_0, transpose_y = sim_9_transpose_y_0, x = transpose_16, y = transpose_17)[name = tensor("sim_9_cast_fp16")]; + tensor var_537_to_fp16 = const()[name = tensor("op_537_to_fp16"), val = tensor(0x1p-3)]; + tensor sim_cast_fp16 = mul(x = sim_9_cast_fp16, y = var_537_to_fp16)[name = tensor("sim_cast_fp16")]; + tensor attn_cast_fp16 = softmax(axis = var_443, x = sim_cast_fp16)[name = tensor("attn_cast_fp16")]; + tensor x_105_transpose_x_0 = const()[name = tensor("x_105_transpose_x_0"), val = tensor(false)]; + tensor x_105_transpose_y_0 = const()[name = tensor("x_105_transpose_y_0"), val = tensor(false)]; + tensor v_cast_fp16 = transpose(perm = var_533, x = x_103_cast_fp16)[name = tensor("transpose_21")]; + tensor x_105_cast_fp16 = matmul(transpose_x = x_105_transpose_x_0, transpose_y = x_105_transpose_y_0, x = attn_cast_fp16, y = v_cast_fp16)[name = tensor("x_105_cast_fp16")]; + tensor var_559 = const()[name = tensor("op_559"), val = tensor([0, 2, 1, 3])]; + tensor var_561 = const()[name = tensor("op_561"), val = tensor([1, 512, 512])]; + tensor x_107_cast_fp16 = transpose(perm = var_559, x = x_105_cast_fp16)[name = tensor("transpose_20")]; + tensor input_55_cast_fp16 = reshape(shape = var_561, x = x_107_cast_fp16)[name = tensor("input_55_cast_fp16")]; + tensor transformer_blocks_2_attention_attention_to_out_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19991808))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(20516160))), name = tensor("transformer_blocks_2_attention_attention_to_out_weight_to_fp16_palettized"), shape = tensor([1024, 512])]; + tensor transformer_blocks_2_attention_attention_to_out_bias_to_fp16 = const()[name = tensor("transformer_blocks_2_attention_attention_to_out_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(20516736)))]; + tensor linear_22_cast_fp16 = linear(bias = transformer_blocks_2_attention_attention_to_out_bias_to_fp16, weight = transformer_blocks_2_attention_attention_to_out_weight_to_fp16_palettized, x = input_55_cast_fp16)[name = tensor("linear_22_cast_fp16")]; + tensor input_57_cast_fp16 = add(x = linear_22_cast_fp16, y = x_79_cast_fp16)[name = tensor("input_57_cast_fp16")]; + tensor transformer_blocks_2_feed_forward_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(20518848))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(22616064))), name = tensor("transformer_blocks_2_feed_forward_0_weight_to_fp16_palettized"), shape = tensor([2048, 1024])]; + tensor transformer_blocks_2_feed_forward_0_bias_to_fp16 = const()[name = tensor("transformer_blocks_2_feed_forward_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(22616640)))]; + tensor linear_23_cast_fp16 = linear(bias = transformer_blocks_2_feed_forward_0_bias_to_fp16, weight = transformer_blocks_2_feed_forward_0_weight_to_fp16_palettized, x = input_57_cast_fp16)[name = tensor("linear_23_cast_fp16")]; + tensor input_61_mode_0 = const()[name = tensor("input_61_mode_0"), val = tensor("EXACT")]; + tensor input_61_cast_fp16 = gelu(mode = input_61_mode_0, x = linear_23_cast_fp16)[name = tensor("input_61_cast_fp16")]; + tensor transformer_blocks_2_feed_forward_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(22620800))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24718016))), name = tensor("transformer_blocks_2_feed_forward_2_weight_to_fp16_palettized"), shape = tensor([1024, 2048])]; + tensor transformer_blocks_2_feed_forward_2_bias_to_fp16 = const()[name = tensor("transformer_blocks_2_feed_forward_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24718592)))]; + tensor linear_24_cast_fp16 = linear(bias = transformer_blocks_2_feed_forward_2_bias_to_fp16, weight = transformer_blocks_2_feed_forward_2_weight_to_fp16_palettized, x = input_61_cast_fp16)[name = tensor("linear_24_cast_fp16")]; + tensor x_109_cast_fp16 = add(x = linear_24_cast_fp16, y = input_57_cast_fp16)[name = tensor("x_109_cast_fp16")]; + tensor var_581_axes_0 = const()[name = tensor("op_581_axes_0"), val = tensor([1])]; + tensor var_581_keep_dims_0 = const()[name = tensor("op_581_keep_dims_0"), val = tensor(false)]; + tensor var_581_cast_fp16 = reduce_mean(axes = var_581_axes_0, keep_dims = var_581_keep_dims_0, x = x_109_cast_fp16)[name = tensor("op_581_cast_fp16")]; + tensor x_111_axes_0 = const()[name = tensor("x_111_axes_0"), val = tensor([1])]; + tensor x_111_cast_fp16 = expand_dims(axes = x_111_axes_0, x = var_581_cast_fp16)[name = tensor("x_111_cast_fp16")]; + tensor var_590 = const()[name = tensor("op_590"), val = tensor([0, 2, 1])]; + tensor x_pad_type_0 = const()[name = tensor("x_pad_type_0"), val = tensor("valid")]; + tensor x_strides_0 = const()[name = tensor("x_strides_0"), val = tensor([1])]; + tensor x_pad_0 = const()[name = tensor("x_pad_0"), val = tensor([0, 0])]; + tensor x_dilations_0 = const()[name = tensor("x_dilations_0"), val = tensor([1])]; + tensor x_groups_0 = const()[name = tensor("x_groups_0"), val = tensor(1)]; + tensor transformer_to_out_1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24720704))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24982912))), name = tensor("transformer_to_out_1_weight_to_fp16_palettized"), shape = tensor([256, 1024, 1])]; + tensor transformer_to_out_1_bias_to_fp16 = const()[name = tensor("transformer_to_out_1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24983488)))]; + tensor input_cast_fp16 = transpose(perm = var_590, x = x_111_cast_fp16)[name = tensor("transpose_19")]; + tensor x_cast_fp16 = conv(bias = transformer_to_out_1_bias_to_fp16, dilations = x_dilations_0, groups = x_groups_0, pad = x_pad_0, pad_type = x_pad_type_0, strides = x_strides_0, weight = transformer_to_out_1_weight_to_fp16_palettized, x = input_cast_fp16)[name = tensor("x_cast_fp16")]; + tensor x_pred_perm_0 = const()[name = tensor("x_pred_perm_0"), val = tensor([0, -1, -2])]; + tensor var_602_cast_fp16 = mul(x = c_skip_cast_fp16, y = x_noisy_to_fp16)[name = tensor("op_602_cast_fp16")]; + tensor x_pred_cast_fp16 = transpose(perm = x_pred_perm_0, x = x_cast_fp16)[name = tensor("transpose_18")]; + tensor var_603_cast_fp16 = mul(x = c_out_cast_fp16, y = x_pred_cast_fp16)[name = tensor("op_603_cast_fp16")]; + tensor denoised = add(x = var_602_cast_fp16, y = var_603_cast_fp16)[name = tensor("op_605_cast_fp16")]; + } -> (denoised); +} \ No newline at end of file diff --git a/ANE/styletts2_ane_diffusion_step.mlmodelc/weights/weight.bin b/ANE/styletts2_ane_diffusion_step.mlmodelc/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..088c545fffa8cb6c5a53ada162b8e21be75d0603 --- /dev/null +++ b/ANE/styletts2_ane_diffusion_step.mlmodelc/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c7ec58d48cbfd5094ca3c0a9de778a789878e9d2cae35cc4ad277cff669d538d +size 24984064 diff --git a/ANE/styletts2_ane_diffusion_step.mlpackage/Data/com.apple.CoreML/model.mlmodel b/ANE/styletts2_ane_diffusion_step.mlpackage/Data/com.apple.CoreML/model.mlmodel new file mode 100644 index 0000000000000000000000000000000000000000..3908df93d062bb9f950886d0136c29a2f5b55611 --- /dev/null +++ b/ANE/styletts2_ane_diffusion_step.mlpackage/Data/com.apple.CoreML/model.mlmodel @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f7012e3db65ab67a3ad9a85a421d83b9ceb2363b370c7fc3f27d23271af6ca9e +size 58407 diff --git a/ANE/styletts2_ane_diffusion_step.mlpackage/Data/com.apple.CoreML/weights/weight.bin b/ANE/styletts2_ane_diffusion_step.mlpackage/Data/com.apple.CoreML/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..088c545fffa8cb6c5a53ada162b8e21be75d0603 --- /dev/null +++ b/ANE/styletts2_ane_diffusion_step.mlpackage/Data/com.apple.CoreML/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c7ec58d48cbfd5094ca3c0a9de778a789878e9d2cae35cc4ad277cff669d538d +size 24984064 diff --git a/ANE/styletts2_ane_diffusion_step.mlpackage/Manifest.json b/ANE/styletts2_ane_diffusion_step.mlpackage/Manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..aba12ebce3a54ab37f9b50e06c9aa36b033cdb6d --- /dev/null +++ b/ANE/styletts2_ane_diffusion_step.mlpackage/Manifest.json @@ -0,0 +1,18 @@ +{ + "fileFormatVersion": "1.0.0", + "itemInfoEntries": { + "0D2868A2-D1E7-4D21-BDBB-2078B73600CD": { + "author": "com.apple.CoreML", + "description": "CoreML Model Specification", + "name": "model.mlmodel", + "path": "com.apple.CoreML/model.mlmodel" + }, + "99F44601-8B02-4B68-8756-183CA2AB7B21": { + "author": "com.apple.CoreML", + "description": "CoreML Model Weights", + "name": "weights", + "path": "com.apple.CoreML/weights" + } + }, + "rootModelIdentifier": "0D2868A2-D1E7-4D21-BDBB-2078B73600CD" +} diff --git a/ANE/styletts2_ane_noise.mlmodelc/analytics/coremldata.bin b/ANE/styletts2_ane_noise.mlmodelc/analytics/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..af64aafee611cd3ed08749ff2461ad3aefe9b3e1 --- /dev/null +++ b/ANE/styletts2_ane_noise.mlmodelc/analytics/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4f4c8e0ce15ecb4ac16b403361160287c7638ac0dfdd8327835be795979b6395 +size 243 diff --git a/ANE/styletts2_ane_noise.mlmodelc/coremldata.bin b/ANE/styletts2_ane_noise.mlmodelc/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..f477cc4fbb2b3896dc8589a63c58cdedd1599768 --- /dev/null +++ b/ANE/styletts2_ane_noise.mlmodelc/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:848949d5bf9dd6db5ef5cd49faa59a9ecfe47efcf442562eb568471c9787ff16 +size 407 diff --git a/ANE/styletts2_ane_noise.mlmodelc/metadata.json b/ANE/styletts2_ane_noise.mlmodelc/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..a122cb5d75488e34337e6f70c1625145ef11e085 --- /dev/null +++ b/ANE/styletts2_ane_noise.mlmodelc/metadata.json @@ -0,0 +1,87 @@ +[ + { + "metadataOutputVersion" : "3.0", + "shortDescription" : "StyleTTS2-ANE Stage 6: SineGen (static T_a=2000) — fp32, ALL.", + "outputSchema" : [ + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float32", + "formattedType" : "MultiArray (Float32 1 × 1200000 × 9)", + "shortDescription" : "", + "shape" : "[1, 1200000, 9]", + "name" : "sine_waves", + "type" : "MultiArray" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float32", + "formattedType" : "MultiArray (Float32 1 × 1200000 × 1)", + "shortDescription" : "", + "shape" : "[1, 1200000, 1]", + "name" : "uv", + "type" : "MultiArray" + } + ], + "storagePrecision" : "Palettized (8 bits)", + "modelParameters" : [ + + ], + "specificationVersion" : 8, + "mlProgramOperationTypeHistogram" : { + "Ios17.mul" : 14, + "Ios17.floor" : 1, + "Ios17.transpose" : 7, + "Ios17.sub" : 2, + "Ios16.constexprLutToDense" : 2, + "Ios17.sin" : 2, + "Ios16.cumsum" : 1, + "Ios17.greater" : 1, + "Ios17.concat" : 1, + "Ios17.sliceByIndex" : 3, + "UpsampleNearestNeighbor" : 1, + "Ios17.add" : 3, + "Tile" : 2, + "Ios17.expandDims" : 2, + "Ios17.squeeze" : 1, + "Ios17.reshape" : 4, + "Ios17.cast" : 1 + }, + "computePrecision" : "Mixed (Float32, Int32)", + "isUpdatable" : "0", + "stateSchema" : [ + + ], + "availability" : { + "macOS" : "14.0", + "tvOS" : "17.0", + "visionOS" : "1.0", + "watchOS" : "10.0", + "iOS" : "17.0", + "macCatalyst" : "17.0" + }, + "modelType" : { + "name" : "MLModelType_mlProgram" + }, + "inputSchema" : [ + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float32", + "formattedType" : "MultiArray (Float32 1 × 4000)", + "shortDescription" : "", + "shape" : "[1, 4000]", + "name" : "F0_curve", + "type" : "MultiArray" + } + ], + "userDefinedMetadata" : { + "com.github.apple.coremltools.version" : "8.3.0", + "com.github.apple.coremltools.source_dialect" : "TorchScript", + "com.github.apple.coremltools.source" : "torch==2.11.0" + }, + "generatedClassName" : "styletts2_ane_noise", + "method" : "predict" + } +] \ No newline at end of file diff --git a/ANE/styletts2_ane_noise.mlmodelc/model.mil b/ANE/styletts2_ane_noise.mlmodelc/model.mil new file mode 100644 index 0000000000000000000000000000000000000000..8bed6c184db39cc077c18d819cf677587c99bd29 --- /dev/null +++ b/ANE/styletts2_ane_noise.mlmodelc/model.mil @@ -0,0 +1,99 @@ +program(1.0) +[buildInfo = dict, tensor>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}})] +{ + func main(tensor F0_curve) { + tensor input_axes_0 = const()[name = tensor("input_axes_0"), val = tensor([1])]; + tensor input = expand_dims(axes = input_axes_0, x = F0_curve)[name = tensor("input")]; + tensor expand_dims_0_axes_0 = const()[name = tensor("expand_dims_0_axes_0"), val = tensor([3])]; + tensor expand_dims_0 = expand_dims(axes = expand_dims_0_axes_0, x = input)[name = tensor("expand_dims_0")]; + tensor upsample_nearest_neighbor_0_scale_factor_height_0 = const()[name = tensor("upsample_nearest_neighbor_0_scale_factor_height_0"), val = tensor(300)]; + tensor upsample_nearest_neighbor_0_scale_factor_width_0 = const()[name = tensor("upsample_nearest_neighbor_0_scale_factor_width_0"), val = tensor(1)]; + tensor upsample_nearest_neighbor_0 = upsample_nearest_neighbor(scale_factor_height = upsample_nearest_neighbor_0_scale_factor_height_0, scale_factor_width = upsample_nearest_neighbor_0_scale_factor_width_0, x = expand_dims_0)[name = tensor("upsample_nearest_neighbor_0")]; + tensor var_12_axes_0 = const()[name = tensor("op_12_axes_0"), val = tensor([3])]; + tensor var_12 = squeeze(axes = var_12_axes_0, x = upsample_nearest_neighbor_0)[name = tensor("op_12")]; + tensor f0_perm_0 = const()[name = tensor("f0_perm_0"), val = tensor([0, 2, 1])]; + tensor var_23 = const()[name = tensor("op_23"), val = tensor(1)]; + tensor const_1 = const()[name = tensor("const_1"), val = tensor([[[0x1p+0, 0x1p+1, 0x1.8p+1, 0x1p+2, 0x1.4p+2, 0x1.8p+2, 0x1.cp+2, 0x1p+3, 0x1.2p+3]]])]; + tensor f0 = transpose(perm = f0_perm_0, x = var_12)[name = tensor("transpose_6")]; + tensor f0_values = mul(x = f0, y = const_1)[name = tensor("f0_values")]; + tensor _inversed_rad_div_y_0 = const()[name = tensor("_inversed_rad_div_y_0"), val = tensor(0x1.5d867cp-15)]; + tensor _inversed_rad_div = mul(x = f0_values, y = _inversed_rad_div_y_0)[name = tensor("_inversed_rad_div")]; + tensor var_35 = floor(x = _inversed_rad_div)[name = tensor("op_35")]; + tensor rad_mod = sub(x = _inversed_rad_div, y = var_35)[name = tensor("rad_mod")]; + tensor var_38_begin_0 = const()[name = tensor("op_38_begin_0"), val = tensor([0, 0, 0])]; + tensor var_38_end_0 = const()[name = tensor("op_38_end_0"), val = tensor([1, 1200000, 9])]; + tensor var_38_end_mask_0 = const()[name = tensor("op_38_end_mask_0"), val = tensor([true, true, true])]; + tensor var_38_stride_0 = const()[name = tensor("op_38_stride_0"), val = tensor([1, 300, 1])]; + tensor var_38 = slice_by_index(begin = var_38_begin_0, end = var_38_end_0, end_mask = var_38_end_mask_0, stride = var_38_stride_0, x = rad_mod)[name = tensor("op_38")]; + tensor var_40_exclusive_0 = const()[name = tensor("op_40_exclusive_0"), val = tensor(false)]; + tensor var_40_reverse_0 = const()[name = tensor("op_40_reverse_0"), val = tensor(false)]; + tensor var_40 = cumsum(axis = var_23, exclusive = var_40_exclusive_0, reverse = var_40_reverse_0, x = var_38)[name = tensor("op_40")]; + tensor var_41_promoted = const()[name = tensor("op_41_promoted"), val = tensor(0x1p+1)]; + tensor var_42 = mul(x = var_40, y = var_41_promoted)[name = tensor("op_42")]; + tensor var_43 = const()[name = tensor("op_43"), val = tensor(0x1.921fb6p+1)]; + tensor phase = mul(x = var_42, y = var_43)[name = tensor("phase")]; + tensor var_45_promoted = const()[name = tensor("op_45_promoted"), val = tensor(0x1.2cp+8)]; + tensor phase_scaled = mul(x = phase, y = var_45_promoted)[name = tensor("phase_scaled")]; + tensor transpose_0_perm_0 = const()[name = tensor("transpose_0_perm_0"), val = tensor([1, 0, 2])]; + tensor tile_0_reps_0 = const()[name = tensor("tile_0_reps_0"), val = tensor([300, 1, 1])]; + tensor transpose_0 = transpose(perm = transpose_0_perm_0, x = phase_scaled)[name = tensor("transpose_5")]; + tensor tile_0 = tile(reps = tile_0_reps_0, x = transpose_0)[name = tensor("tile_0")]; + tensor concat_0 = const()[name = tensor("concat_0"), val = tensor([300, 4000, 1, 9])]; + tensor reshape_0 = reshape(shape = concat_0, x = tile_0)[name = tensor("reshape_0")]; + tensor transpose_1_perm_0 = const()[name = tensor("transpose_1_perm_0"), val = tensor([1, 0, 2, 3])]; + tensor concat_1 = const()[name = tensor("concat_1"), val = tensor([-1, 1, 9])]; + tensor transpose_1 = transpose(perm = transpose_1_perm_0, x = reshape_0)[name = tensor("transpose_4")]; + tensor reshape_1 = reshape(shape = concat_1, x = transpose_1)[name = tensor("reshape_1")]; + tensor left_perm_0 = const()[name = tensor("left_perm_0"), val = tensor([1, 0, 2])]; + tensor var_49_begin_0 = const()[name = tensor("op_49_begin_0"), val = tensor([0, -1, 0])]; + tensor var_49_end_0 = const()[name = tensor("op_49_end_0"), val = tensor([1, 4000, 9])]; + tensor var_49_end_mask_0 = const()[name = tensor("op_49_end_mask_0"), val = tensor([true, true, true])]; + tensor var_49 = slice_by_index(begin = var_49_begin_0, end = var_49_end_0, end_mask = var_49_end_mask_0, x = phase_scaled)[name = tensor("op_49")]; + tensor var_52_begin_0 = const()[name = tensor("op_52_begin_0"), val = tensor([0, 1, 0])]; + tensor var_52_end_0 = const()[name = tensor("op_52_end_0"), val = tensor([1, 4000, 9])]; + tensor var_52_end_mask_0 = const()[name = tensor("op_52_end_mask_0"), val = tensor([true, true, true])]; + tensor var_52 = slice_by_index(begin = var_52_begin_0, end = var_52_end_0, end_mask = var_52_end_mask_0, x = phase_scaled)[name = tensor("op_52")]; + tensor shifted_interleave_0 = const()[name = tensor("shifted_interleave_0"), val = tensor(false)]; + tensor shifted = concat(axis = var_23, interleave = shifted_interleave_0, values = (var_52, var_49))[name = tensor("shifted")]; + tensor transpose_2_perm_0 = const()[name = tensor("transpose_2_perm_0"), val = tensor([1, 0, 2])]; + tensor tile_1_reps_0 = const()[name = tensor("tile_1_reps_0"), val = tensor([300, 1, 1])]; + tensor transpose_2 = transpose(perm = transpose_2_perm_0, x = shifted)[name = tensor("transpose_3")]; + tensor tile_1 = tile(reps = tile_1_reps_0, x = transpose_2)[name = tensor("tile_1")]; + tensor concat_2 = const()[name = tensor("concat_2"), val = tensor([300, 4000, 1, 9])]; + tensor reshape_2 = reshape(shape = concat_2, x = tile_1)[name = tensor("reshape_2")]; + tensor transpose_3_perm_0 = const()[name = tensor("transpose_3_perm_0"), val = tensor([1, 0, 2, 3])]; + tensor concat_3 = const()[name = tensor("concat_3"), val = tensor([-1, 1, 9])]; + tensor transpose_3 = transpose(perm = transpose_3_perm_0, x = reshape_2)[name = tensor("transpose_2")]; + tensor reshape_3 = reshape(shape = concat_3, x = transpose_3)[name = tensor("reshape_3")]; + tensor right_perm_0 = const()[name = tensor("right_perm_0"), val = tensor([1, 0, 2])]; + tensor const_2_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(64))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1200128))), name = tensor("const_2_palettized"), shape = tensor([1, 1200000, 1])]; + tensor op_58_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1201216))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2401280))), name = tensor("op_58_palettized"), shape = tensor([1, 1200000, 1])]; + tensor left = transpose(perm = left_perm_0, x = reshape_1)[name = tensor("transpose_1")]; + tensor var_59 = mul(x = left, y = op_58_palettized)[name = tensor("op_59")]; + tensor right = transpose(perm = right_perm_0, x = reshape_3)[name = tensor("transpose_0")]; + tensor var_60 = mul(x = right, y = const_2_palettized)[name = tensor("op_60")]; + tensor phase_up = add(x = var_59, y = var_60)[name = tensor("phase_up")]; + tensor sines = sin(x = phase_up)[name = tensor("sines")]; + tensor var_63 = const()[name = tensor("op_63"), val = tensor(0x1.99999ap-4)]; + tensor sine_waves_1 = mul(x = sines, y = var_63)[name = tensor("sine_waves")]; + tensor var_16_promoted = const()[name = tensor("op_16_promoted"), val = tensor(0x1.4p+3)]; + tensor var_65 = greater(x = f0, y = var_16_promoted)[name = tensor("op_65")]; + tensor cast_2_dtype_0 = const()[name = tensor("cast_2_dtype_0"), val = tensor("fp32")]; + tensor var_67 = const()[name = tensor("op_67"), val = tensor(0x1.89374cp-9)]; + tensor uv = cast(dtype = cast_2_dtype_0, x = var_65)[name = tensor("cast_0")]; + tensor var_68 = mul(x = uv, y = var_67)[name = tensor("op_68")]; + tensor var_23_promoted = const()[name = tensor("op_23_promoted"), val = tensor(0x1p+0)]; + tensor var_69 = sub(x = var_23_promoted, y = uv)[name = tensor("op_69")]; + tensor var_70 = const()[name = tensor("op_70"), val = tensor(0x1.99999ap-4)]; + tensor var_71 = mul(x = var_69, y = var_70)[name = tensor("op_71")]; + tensor _inversed_73_y_0 = const()[name = tensor("_inversed_73_y_0"), val = tensor(0x1.555556p-2)]; + tensor _inversed_73 = mul(x = var_71, y = _inversed_73_y_0)[name = tensor("_inversed_73")]; + tensor noise_amp = add(x = var_68, y = _inversed_73)[name = tensor("noise_amp")]; + tensor var_75 = const()[name = tensor("op_75"), val = tensor(0x1.9p+6)]; + tensor var_76 = mul(x = f0_values, y = var_75)[name = tensor("op_76")]; + tensor var_77 = sin(x = var_76)[name = tensor("op_77")]; + tensor noise = mul(x = noise_amp, y = var_77)[name = tensor("noise")]; + tensor var_79 = mul(x = sine_waves_1, y = uv)[name = tensor("op_79")]; + tensor sine_waves = add(x = var_79, y = noise)[name = tensor("op_80")]; + } -> (sine_waves, uv); +} \ No newline at end of file diff --git a/ANE/styletts2_ane_noise.mlmodelc/weights/weight.bin b/ANE/styletts2_ane_noise.mlmodelc/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..9f9cc6d4dc3ada4c4836e0a6cf8ed278c571cc24 --- /dev/null +++ b/ANE/styletts2_ane_noise.mlmodelc/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b988942b88ad44925a6cf81bf3d1108b1da6c2f2c84edc19091e75cd29e53d6 +size 2402368 diff --git a/ANE/styletts2_ane_noise.mlpackage/Data/com.apple.CoreML/model.mlmodel b/ANE/styletts2_ane_noise.mlpackage/Data/com.apple.CoreML/model.mlmodel new file mode 100644 index 0000000000000000000000000000000000000000..ac752e8ac0ea75a1b8a3a945a2615654f2857f47 --- /dev/null +++ b/ANE/styletts2_ane_noise.mlpackage/Data/com.apple.CoreML/model.mlmodel @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:54d28925e686100531855b225d53c4c50f94bac6eac7a2d0193340d4dcecb0bc +size 12114 diff --git a/ANE/styletts2_ane_noise.mlpackage/Data/com.apple.CoreML/weights/weight.bin b/ANE/styletts2_ane_noise.mlpackage/Data/com.apple.CoreML/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..9f9cc6d4dc3ada4c4836e0a6cf8ed278c571cc24 --- /dev/null +++ b/ANE/styletts2_ane_noise.mlpackage/Data/com.apple.CoreML/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b988942b88ad44925a6cf81bf3d1108b1da6c2f2c84edc19091e75cd29e53d6 +size 2402368 diff --git a/ANE/styletts2_ane_noise.mlpackage/Manifest.json b/ANE/styletts2_ane_noise.mlpackage/Manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..2b78629f3301ebac8955a2dc86a4fdbf1d516d65 --- /dev/null +++ b/ANE/styletts2_ane_noise.mlpackage/Manifest.json @@ -0,0 +1,18 @@ +{ + "fileFormatVersion": "1.0.0", + "itemInfoEntries": { + "47473672-28F9-487C-9ED6-63C6D82ABB57": { + "author": "com.apple.CoreML", + "description": "CoreML Model Weights", + "name": "weights", + "path": "com.apple.CoreML/weights" + }, + "88C15F91-FFC7-4B76-B3DF-CFFD20CF2957": { + "author": "com.apple.CoreML", + "description": "CoreML Model Specification", + "name": "model.mlmodel", + "path": "com.apple.CoreML/model.mlmodel" + } + }, + "rootModelIdentifier": "88C15F91-FFC7-4B76-B3DF-CFFD20CF2957" +} diff --git a/ANE/styletts2_ane_plbert.mlmodelc/analytics/coremldata.bin b/ANE/styletts2_ane_plbert.mlmodelc/analytics/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..c326b6a7a2f109863299ebabba930098f2a20dc3 --- /dev/null +++ b/ANE/styletts2_ane_plbert.mlmodelc/analytics/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a828a4caef0ea1661731eb81369d07f3c0bcbfa9bbb9d2aaebe0c75fd1136221 +size 243 diff --git a/ANE/styletts2_ane_plbert.mlmodelc/coremldata.bin b/ANE/styletts2_ane_plbert.mlmodelc/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..9b1340ea61c46131f6af7888b26667d0aea74a01 --- /dev/null +++ b/ANE/styletts2_ane_plbert.mlmodelc/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e968b81438916aee673e4daa1410ca1b6490946302de377e2a3c96389f8c9bed +size 380 diff --git a/ANE/styletts2_ane_plbert.mlmodelc/metadata.json b/ANE/styletts2_ane_plbert.mlmodelc/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..3d217ed51da3597a8d07eb6d638e2a83cdfc2693 --- /dev/null +++ b/ANE/styletts2_ane_plbert.mlmodelc/metadata.json @@ -0,0 +1,83 @@ +[ + { + "metadataOutputVersion" : "3.0", + "shortDescription" : "StyleTTS2-ANE Stage 1: PLBERT (Albert) — fp16, ANE.", + "outputSchema" : [ + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float16", + "formattedType" : "MultiArray (Float16)", + "shortDescription" : "", + "shape" : "[]", + "name" : "bert_dur", + "type" : "MultiArray" + } + ], + "storagePrecision" : "Mixed (Float16, Palettized (8 bits))", + "modelParameters" : [ + + ], + "specificationVersion" : 8, + "mlProgramOperationTypeHistogram" : { + "Fill" : 1, + "Ios17.reshape" : 48, + "Ios16.softmax" : 12, + "Ios17.matmul" : 24, + "Ios17.transpose" : 48, + "Select" : 1, + "Ios17.expandDims" : 2, + "Ios17.add" : 38, + "Tile" : 2, + "Ios17.sliceByIndex" : 2, + "Shape" : 16, + "Ios17.gather" : 29, + "Ios17.layerNorm" : 25, + "Ios17.cast" : 44, + "Ios17.sub" : 1, + "Ios16.constexprLutToDense" : 9, + "Ios16.gelu" : 12, + "Ios17.realDiv" : 2, + "Ios17.linear" : 73, + "Ios17.concat" : 26, + "Ios17.mul" : 12 + }, + "computePrecision" : "Mixed (Float16, Int16, Int32, UInt16)", + "isUpdatable" : "0", + "stateSchema" : [ + + ], + "availability" : { + "macOS" : "14.0", + "tvOS" : "17.0", + "visionOS" : "1.0", + "watchOS" : "10.0", + "iOS" : "17.0", + "macCatalyst" : "17.0" + }, + "modelType" : { + "name" : "MLModelType_mlProgram" + }, + "inputSchema" : [ + { + "dataType" : "Int32", + "hasShapeFlexibility" : "1", + "isOptional" : "0", + "shapeFlexibility" : "1 × 2...512", + "shapeRange" : "[[1, 1], [2, 512]]", + "formattedType" : "MultiArray (Int32 1 × 64)", + "type" : "MultiArray", + "shape" : "[1, 64]", + "name" : "tokens", + "shortDescription" : "" + } + ], + "userDefinedMetadata" : { + "com.github.apple.coremltools.version" : "8.3.0", + "com.github.apple.coremltools.source_dialect" : "TorchScript", + "com.github.apple.coremltools.source" : "torch==2.11.0" + }, + "generatedClassName" : "styletts2_ane_plbert", + "method" : "predict" + } +] \ No newline at end of file diff --git a/ANE/styletts2_ane_plbert.mlmodelc/model.mil b/ANE/styletts2_ane_plbert.mlmodelc/model.mil new file mode 100644 index 0000000000000000000000000000000000000000..924374557fe26dc5448330bdebb915459f57020c --- /dev/null +++ b/ANE/styletts2_ane_plbert.mlmodelc/model.mil @@ -0,0 +1,832 @@ +program(1.0) +[buildInfo = dict, tensor>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}})] +{ + func main(tensor tokens) [FlexibleShapeInformation = tuple, dict, tensor>>, tuple, dict, list, ?>>>>((("DefaultShapes", {{"tokens", [1, 64]}}), ("RangeDims", {{"tokens", [[1, 1], [2, 512]]}})))] { + tensor bert_embeddings_token_type_ids = const()[name = tensor("bert_embeddings_token_type_ids"), val = tensor([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]])]; + tensor bert_embeddings_position_ids = const()[name = tensor("bert_embeddings_position_ids"), val = tensor([[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 256, 257, 258, 259, 260, 261, 262, 263, 264, 265, 266, 267, 268, 269, 270, 271, 272, 273, 274, 275, 276, 277, 278, 279, 280, 281, 282, 283, 284, 285, 286, 287, 288, 289, 290, 291, 292, 293, 294, 295, 296, 297, 298, 299, 300, 301, 302, 303, 304, 305, 306, 307, 308, 309, 310, 311, 312, 313, 314, 315, 316, 317, 318, 319, 320, 321, 322, 323, 324, 325, 326, 327, 328, 329, 330, 331, 332, 333, 334, 335, 336, 337, 338, 339, 340, 341, 342, 343, 344, 345, 346, 347, 348, 349, 350, 351, 352, 353, 354, 355, 356, 357, 358, 359, 360, 361, 362, 363, 364, 365, 366, 367, 368, 369, 370, 371, 372, 373, 374, 375, 376, 377, 378, 379, 380, 381, 382, 383, 384, 385, 386, 387, 388, 389, 390, 391, 392, 393, 394, 395, 396, 397, 398, 399, 400, 401, 402, 403, 404, 405, 406, 407, 408, 409, 410, 411, 412, 413, 414, 415, 416, 417, 418, 419, 420, 421, 422, 423, 424, 425, 426, 427, 428, 429, 430, 431, 432, 433, 434, 435, 436, 437, 438, 439, 440, 441, 442, 443, 444, 445, 446, 447, 448, 449, 450, 451, 452, 453, 454, 455, 456, 457, 458, 459, 460, 461, 462, 463, 464, 465, 466, 467, 468, 469, 470, 471, 472, 473, 474, 475, 476, 477, 478, 479, 480, 481, 482, 483, 484, 485, 486, 487, 488, 489, 490, 491, 492, 493, 494, 495, 496, 497, 498, 499, 500, 501, 502, 503, 504, 505, 506, 507, 508, 509, 510, 511]])]; + tensor var_3 = const()[name = tensor("op_3"), val = tensor(768)]; + tensor var_4 = const()[name = tensor("op_4"), val = tensor(64)]; + tensor var_5 = const()[name = tensor("op_5"), val = tensor(12)]; + tensor var_13 = const()[name = tensor("op_13"), val = tensor(-1)]; + tensor var_22 = const()[name = tensor("op_22"), val = tensor(1)]; + tensor var_28_shape = shape(x = tokens)[name = tensor("op_28_shape")]; + tensor gather_0 = const()[name = tensor("gather_0"), val = tensor(1)]; + tensor gather_1_axis_0 = const()[name = tensor("gather_1_axis_0"), val = tensor(0)]; + tensor gather_1_batch_dims_0 = const()[name = tensor("gather_1_batch_dims_0"), val = tensor(0)]; + tensor gather_1_validate_indices_0 = const()[name = tensor("gather_1_validate_indices_0"), val = tensor(false)]; + tensor var_28_shape_to_int16_dtype_0 = const()[name = tensor("op_28_shape_to_int16_dtype_0"), val = tensor("int16")]; + tensor select_1_to_uint16 = const()[name = tensor("select_1_to_uint16"), val = tensor(1)]; + tensor var_28_shape_to_int16 = cast(dtype = var_28_shape_to_int16_dtype_0, x = var_28_shape)[name = tensor("cast_43")]; + tensor gather_1_cast_uint16 = gather(axis = gather_1_axis_0, batch_dims = gather_1_batch_dims_0, indices = select_1_to_uint16, validate_indices = gather_1_validate_indices_0, x = var_28_shape_to_int16)[name = tensor("gather_1_cast_uint16")]; + tensor gather_1_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_1_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor concat_0_axis_0 = const()[name = tensor("concat_0_axis_0"), val = tensor(0)]; + tensor concat_0_interleave_0 = const()[name = tensor("concat_0_interleave_0"), val = tensor(false)]; + tensor gather_1_cast_uint16_to_int32 = cast(dtype = gather_1_cast_uint16_to_int32_dtype_0, x = gather_1_cast_uint16)[name = tensor("cast_42")]; + tensor concat_0 = concat(axis = concat_0_axis_0, interleave = concat_0_interleave_0, values = (gather_0, gather_1_cast_uint16_to_int32))[name = tensor("concat_0")]; + tensor fill_0_value_0_to_fp16 = const()[name = tensor("fill_0_value_0_to_fp16"), val = tensor(0x1p+0)]; + tensor fill_0_cast_fp16 = fill(shape = concat_0, value = fill_0_value_0_to_fp16)[name = tensor("fill_0_cast_fp16")]; + tensor buffered_token_type_ids_begin_0 = const()[name = tensor("buffered_token_type_ids_begin_0"), val = tensor([0, 0])]; + tensor buffered_token_type_ids_end_mask_0 = const()[name = tensor("buffered_token_type_ids_end_mask_0"), val = tensor([true, false])]; + tensor buffered_token_type_ids = slice_by_index(begin = buffered_token_type_ids_begin_0, end = concat_0, end_mask = buffered_token_type_ids_end_mask_0, x = bert_embeddings_token_type_ids)[name = tensor("buffered_token_type_ids")]; + tensor shape_2 = shape(x = buffered_token_type_ids)[name = tensor("shape_2")]; + tensor real_div_0 = real_div(x = concat_0, y = shape_2)[name = tensor("real_div_0")]; + tensor input_1 = tile(reps = real_div_0, x = buffered_token_type_ids)[name = tensor("input_1")]; + tensor input_3_begin_0 = const()[name = tensor("input_3_begin_0"), val = tensor([0, 0])]; + tensor input_3_end_mask_0 = const()[name = tensor("input_3_end_mask_0"), val = tensor([true, false])]; + tensor input_3 = slice_by_index(begin = input_3_begin_0, end = concat_0, end_mask = input_3_end_mask_0, x = bert_embeddings_position_ids)[name = tensor("input_3")]; + tensor inputs_embeds_axis_0 = const()[name = tensor("inputs_embeds_axis_0"), val = tensor(0)]; + tensor inputs_embeds_batch_dims_0 = const()[name = tensor("inputs_embeds_batch_dims_0"), val = tensor(0)]; + tensor inputs_embeds_validate_indices_0 = const()[name = tensor("inputs_embeds_validate_indices_0"), val = tensor(false)]; + tensor bert_embeddings_word_embeddings_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(64))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(22912))), name = tensor("bert_embeddings_word_embeddings_weight_to_fp16_palettized"), shape = tensor([178, 128])]; + tensor tokens_to_uint16_dtype_0 = const()[name = tensor("tokens_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor tokens_to_uint16 = cast(dtype = tokens_to_uint16_dtype_0, x = tokens)[name = tensor("cast_41")]; + tensor inputs_embeds_cast_fp16_cast_uint16 = gather(axis = inputs_embeds_axis_0, batch_dims = inputs_embeds_batch_dims_0, indices = tokens_to_uint16, validate_indices = inputs_embeds_validate_indices_0, x = bert_embeddings_word_embeddings_weight_to_fp16_palettized)[name = tensor("inputs_embeds_cast_fp16_cast_uint16")]; + tensor token_type_embeddings_1_axis_0 = const()[name = tensor("token_type_embeddings_1_axis_0"), val = tensor(0)]; + tensor token_type_embeddings_1_batch_dims_0 = const()[name = tensor("token_type_embeddings_1_batch_dims_0"), val = tensor(0)]; + tensor token_type_embeddings_1_validate_indices_0 = const()[name = tensor("token_type_embeddings_1_validate_indices_0"), val = tensor(false)]; + tensor bert_embeddings_token_type_embeddings_weight_to_fp16 = const()[name = tensor("bert_embeddings_token_type_embeddings_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23488)))]; + tensor input_1_to_uint16_dtype_0 = const()[name = tensor("input_1_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor input_1_to_uint16 = cast(dtype = input_1_to_uint16_dtype_0, x = input_1)[name = tensor("cast_40")]; + tensor token_type_embeddings_1_cast_fp16_cast_uint16 = gather(axis = token_type_embeddings_1_axis_0, batch_dims = token_type_embeddings_1_batch_dims_0, indices = input_1_to_uint16, validate_indices = token_type_embeddings_1_validate_indices_0, x = bert_embeddings_token_type_embeddings_weight_to_fp16)[name = tensor("token_type_embeddings_1_cast_fp16_cast_uint16")]; + tensor embeddings_1_cast_fp16 = add(x = inputs_embeds_cast_fp16_cast_uint16, y = token_type_embeddings_1_cast_fp16_cast_uint16)[name = tensor("embeddings_1_cast_fp16")]; + tensor position_embeddings_1_axis_0 = const()[name = tensor("position_embeddings_1_axis_0"), val = tensor(0)]; + tensor position_embeddings_1_batch_dims_0 = const()[name = tensor("position_embeddings_1_batch_dims_0"), val = tensor(0)]; + tensor position_embeddings_1_validate_indices_0 = const()[name = tensor("position_embeddings_1_validate_indices_0"), val = tensor(false)]; + tensor bert_embeddings_position_embeddings_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24064))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(89664))), name = tensor("bert_embeddings_position_embeddings_weight_to_fp16_palettized"), shape = tensor([512, 128])]; + tensor input_3_to_uint16_dtype_0 = const()[name = tensor("input_3_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor input_3_to_uint16 = cast(dtype = input_3_to_uint16_dtype_0, x = input_3)[name = tensor("cast_39")]; + tensor position_embeddings_1_cast_fp16_cast_uint16 = gather(axis = position_embeddings_1_axis_0, batch_dims = position_embeddings_1_batch_dims_0, indices = input_3_to_uint16, validate_indices = position_embeddings_1_validate_indices_0, x = bert_embeddings_position_embeddings_weight_to_fp16_palettized)[name = tensor("position_embeddings_1_cast_fp16_cast_uint16")]; + tensor input_5_cast_fp16 = add(x = embeddings_1_cast_fp16, y = position_embeddings_1_cast_fp16_cast_uint16)[name = tensor("input_5_cast_fp16")]; + tensor input_7_axes_0 = const()[name = tensor("input_7_axes_0"), val = tensor([-1])]; + tensor bert_embeddings_LayerNorm_weight_to_fp16 = const()[name = tensor("bert_embeddings_LayerNorm_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(90240)))]; + tensor bert_embeddings_LayerNorm_bias_to_fp16 = const()[name = tensor("bert_embeddings_LayerNorm_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(90560)))]; + tensor var_15_to_fp16 = const()[name = tensor("op_15_to_fp16"), val = tensor(0x1p-24)]; + tensor input_7_cast_fp16 = layer_norm(axes = input_7_axes_0, beta = bert_embeddings_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_embeddings_LayerNorm_weight_to_fp16, x = input_5_cast_fp16)[name = tensor("input_7_cast_fp16")]; + tensor var_61_shape_cast_fp16 = shape(x = fill_0_cast_fp16)[name = tensor("op_61_shape_cast_fp16")]; + tensor gather_3 = const()[name = tensor("gather_3"), val = tensor(1)]; + tensor gather_4_axis_0 = const()[name = tensor("gather_4_axis_0"), val = tensor(0)]; + tensor gather_4_batch_dims_0 = const()[name = tensor("gather_4_batch_dims_0"), val = tensor(0)]; + tensor gather_4_validate_indices_0 = const()[name = tensor("gather_4_validate_indices_0"), val = tensor(false)]; + tensor var_61_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_61_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_4_to_uint16 = const()[name = tensor("select_4_to_uint16"), val = tensor(1)]; + tensor var_61_shape_cast_fp16_to_uint16 = cast(dtype = var_61_shape_cast_fp16_to_uint16_dtype_0, x = var_61_shape_cast_fp16)[name = tensor("cast_38")]; + tensor gather_4_cast_uint16 = gather(axis = gather_4_axis_0, batch_dims = gather_4_batch_dims_0, indices = select_4_to_uint16, validate_indices = gather_4_validate_indices_0, x = var_61_shape_cast_fp16_to_uint16)[name = tensor("gather_4_cast_uint16")]; + tensor gather_4_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_4_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor var_64_axes_0 = const()[name = tensor("op_64_axes_0"), val = tensor([1])]; + tensor var_64_cast_fp16 = expand_dims(axes = var_64_axes_0, x = fill_0_cast_fp16)[name = tensor("op_64_cast_fp16")]; + tensor var_65_axes_0 = const()[name = tensor("op_65_axes_0"), val = tensor([2])]; + tensor var_65_cast_fp16 = expand_dims(axes = var_65_axes_0, x = var_64_cast_fp16)[name = tensor("op_65_cast_fp16")]; + tensor concat_4_axis_0 = const()[name = tensor("concat_4_axis_0"), val = tensor(0)]; + tensor concat_4_interleave_0 = const()[name = tensor("concat_4_interleave_0"), val = tensor(false)]; + tensor gather_4_cast_uint16_to_int32 = cast(dtype = gather_4_cast_uint16_to_int32_dtype_0, x = gather_4_cast_uint16)[name = tensor("cast_37")]; + tensor concat_4 = concat(axis = concat_4_axis_0, interleave = concat_4_interleave_0, values = (gather_3, var_22, gather_1_cast_uint16_to_int32, gather_4_cast_uint16_to_int32))[name = tensor("concat_4")]; + tensor shape_6_cast_fp16 = shape(x = var_65_cast_fp16)[name = tensor("shape_6_cast_fp16")]; + tensor real_div_1 = real_div(x = concat_4, y = shape_6_cast_fp16)[name = tensor("real_div_1")]; + tensor var_68_cast_fp16 = tile(reps = real_div_1, x = var_65_cast_fp16)[name = tensor("op_68_cast_fp16")]; + tensor const_0_to_fp16 = const()[name = tensor("const_0_to_fp16"), val = tensor(0x1p+0)]; + tensor inverted_mask_cast_fp16 = sub(x = const_0_to_fp16, y = var_68_cast_fp16)[name = tensor("inverted_mask_cast_fp16")]; + tensor cast_6_dtype_0 = const()[name = tensor("cast_6_dtype_0"), val = tensor("bool")]; + tensor var_7_to_fp16 = const()[name = tensor("op_7_to_fp16"), val = tensor(-inf)]; + tensor inverted_mask_cast_fp16_to_bool = cast(dtype = cast_6_dtype_0, x = inverted_mask_cast_fp16)[name = tensor("cast_36")]; + tensor attention_mask_cast_fp16 = select(a = var_7_to_fp16, b = inverted_mask_cast_fp16, cond = inverted_mask_cast_fp16_to_bool)[name = tensor("attention_mask_cast_fp16")]; + tensor bert_encoder_embedding_hidden_mapping_in_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(90880))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(189248))), name = tensor("bert_encoder_embedding_hidden_mapping_in_weight_to_fp16_palettized"), shape = tensor([768, 128])]; + tensor bert_encoder_embedding_hidden_mapping_in_bias_to_fp16 = const()[name = tensor("bert_encoder_embedding_hidden_mapping_in_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(189824)))]; + tensor linear_0_cast_fp16 = linear(bias = bert_encoder_embedding_hidden_mapping_in_bias_to_fp16, weight = bert_encoder_embedding_hidden_mapping_in_weight_to_fp16_palettized, x = input_7_cast_fp16)[name = tensor("linear_0_cast_fp16")]; + tensor var_92_shape_cast_fp16 = shape(x = linear_0_cast_fp16)[name = tensor("op_92_shape_cast_fp16")]; + tensor gather_5_axis_0 = const()[name = tensor("gather_5_axis_0"), val = tensor(0)]; + tensor gather_5_batch_dims_0 = const()[name = tensor("gather_5_batch_dims_0"), val = tensor(0)]; + tensor gather_5_validate_indices_0 = const()[name = tensor("gather_5_validate_indices_0"), val = tensor(false)]; + tensor var_92_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_92_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_5_to_uint16 = const()[name = tensor("select_5_to_uint16"), val = tensor(0)]; + tensor var_92_shape_cast_fp16_to_uint16 = cast(dtype = var_92_shape_cast_fp16_to_uint16_dtype_0, x = var_92_shape_cast_fp16)[name = tensor("cast_35")]; + tensor gather_5_cast_uint16 = gather(axis = gather_5_axis_0, batch_dims = gather_5_batch_dims_0, indices = select_5_to_uint16, validate_indices = gather_5_validate_indices_0, x = var_92_shape_cast_fp16_to_uint16)[name = tensor("gather_5_cast_uint16")]; + tensor gather_5_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_5_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_6_axis_0 = const()[name = tensor("gather_6_axis_0"), val = tensor(0)]; + tensor gather_6_batch_dims_0 = const()[name = tensor("gather_6_batch_dims_0"), val = tensor(0)]; + tensor gather_6_validate_indices_0 = const()[name = tensor("gather_6_validate_indices_0"), val = tensor(false)]; + tensor select_6_to_uint16 = const()[name = tensor("select_6_to_uint16"), val = tensor(1)]; + tensor gather_6_cast_uint16 = gather(axis = gather_6_axis_0, batch_dims = gather_6_batch_dims_0, indices = select_6_to_uint16, validate_indices = gather_6_validate_indices_0, x = var_92_shape_cast_fp16_to_uint16)[name = tensor("gather_6_cast_uint16")]; + tensor gather_6_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_6_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(191424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(781312))), name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized"), shape = tensor([768, 768])]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16 = const()[name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(781888)))]; + tensor linear_1_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = linear_0_cast_fp16)[name = tensor("linear_1_cast_fp16")]; + tensor concat_5_axis_0 = const()[name = tensor("concat_5_axis_0"), val = tensor(0)]; + tensor concat_5_interleave_0 = const()[name = tensor("concat_5_interleave_0"), val = tensor(false)]; + tensor gather_5_cast_uint16_to_int32 = cast(dtype = gather_5_cast_uint16_to_int32_dtype_0, x = gather_5_cast_uint16)[name = tensor("cast_34")]; + tensor concat_5 = concat(axis = concat_5_axis_0, interleave = concat_5_interleave_0, values = (gather_5_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_5")]; + tensor var_98_cast_fp16 = reshape(shape = concat_5, x = linear_1_cast_fp16)[name = tensor("op_98_cast_fp16")]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(783488))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1373376))), name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized"), shape = tensor([768, 768])]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16 = const()[name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1373952)))]; + tensor linear_2_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = linear_0_cast_fp16)[name = tensor("linear_2_cast_fp16")]; + tensor var_104_cast_fp16 = reshape(shape = concat_5, x = linear_2_cast_fp16)[name = tensor("op_104_cast_fp16")]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1375552))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1965440))), name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized"), shape = tensor([768, 768])]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16 = const()[name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1966016)))]; + tensor linear_3_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = linear_0_cast_fp16)[name = tensor("linear_3_cast_fp16")]; + tensor var_110_cast_fp16 = reshape(shape = concat_5, x = linear_3_cast_fp16)[name = tensor("op_110_cast_fp16")]; + tensor value_layer_1_perm_0 = const()[name = tensor("value_layer_1_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_0_y_0_to_fp16 = const()[name = tensor("mul_0_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_0_cast_fp16 = mul(x = var_98_cast_fp16, y = mul_0_y_0_to_fp16)[name = tensor("mul_0_cast_fp16")]; + tensor matmul_0_transpose_y_0 = const()[name = tensor("matmul_0_transpose_y_0"), val = tensor(true)]; + tensor matmul_0_transpose_x_0 = const()[name = tensor("matmul_0_transpose_x_0"), val = tensor(false)]; + tensor transpose_48_perm_0 = const()[name = tensor("transpose_48_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_49_perm_0 = const()[name = tensor("transpose_49_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_49 = transpose(perm = transpose_49_perm_0, x = var_104_cast_fp16)[name = tensor("transpose_118")]; + tensor transpose_48 = transpose(perm = transpose_48_perm_0, x = mul_0_cast_fp16)[name = tensor("transpose_119")]; + tensor matmul_0_cast_fp16 = matmul(transpose_x = matmul_0_transpose_x_0, transpose_y = matmul_0_transpose_y_0, x = transpose_48, y = transpose_49)[name = tensor("matmul_0_cast_fp16")]; + tensor add_7_cast_fp16 = add(x = matmul_0_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_7_cast_fp16")]; + tensor softmax_0_axis_0 = const()[name = tensor("softmax_0_axis_0"), val = tensor(-1)]; + tensor softmax_0_cast_fp16 = softmax(axis = softmax_0_axis_0, x = add_7_cast_fp16)[name = tensor("softmax_0_cast_fp16")]; + tensor attention_output_1_transpose_x_0 = const()[name = tensor("attention_output_1_transpose_x_0"), val = tensor(false)]; + tensor attention_output_1_transpose_y_0 = const()[name = tensor("attention_output_1_transpose_y_0"), val = tensor(false)]; + tensor value_layer_1_cast_fp16 = transpose(perm = value_layer_1_perm_0, x = var_110_cast_fp16)[name = tensor("transpose_117")]; + tensor attention_output_1_cast_fp16 = matmul(transpose_x = attention_output_1_transpose_x_0, transpose_y = attention_output_1_transpose_y_0, x = softmax_0_cast_fp16, y = value_layer_1_cast_fp16)[name = tensor("attention_output_1_cast_fp16")]; + tensor attention_output_3_perm_0 = const()[name = tensor("attention_output_3_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_8_axis_0 = const()[name = tensor("concat_8_axis_0"), val = tensor(0)]; + tensor concat_8_interleave_0 = const()[name = tensor("concat_8_interleave_0"), val = tensor(false)]; + tensor gather_6_cast_uint16_to_int32 = cast(dtype = gather_6_cast_uint16_to_int32_dtype_0, x = gather_6_cast_uint16)[name = tensor("cast_33")]; + tensor concat_8 = concat(axis = concat_8_axis_0, interleave = concat_8_interleave_0, values = (gather_5_cast_uint16_to_int32, gather_6_cast_uint16_to_int32, var_3))[name = tensor("concat_8")]; + tensor attention_output_3_cast_fp16 = transpose(perm = attention_output_3_perm_0, x = attention_output_1_cast_fp16)[name = tensor("transpose_116")]; + tensor input_11_cast_fp16 = reshape(shape = concat_8, x = attention_output_3_cast_fp16)[name = tensor("input_11_cast_fp16")]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1967616))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2557504))), name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized"), shape = tensor([768, 768])]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16 = const()[name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2558080)))]; + tensor linear_4_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_11_cast_fp16)[name = tensor("linear_4_cast_fp16")]; + tensor input_15_cast_fp16 = add(x = linear_0_cast_fp16, y = linear_4_cast_fp16)[name = tensor("input_15_cast_fp16")]; + tensor input_17_axes_0 = const()[name = tensor("input_17_axes_0"), val = tensor([-1])]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16 = const()[name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2559680)))]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16 = const()[name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2561280)))]; + tensor input_17_cast_fp16 = layer_norm(axes = input_17_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_15_cast_fp16)[name = tensor("input_17_cast_fp16")]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2562880))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4135808))), name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized"), shape = tensor([2048, 768])]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16 = const()[name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4136384)))]; + tensor linear_5_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_17_cast_fp16)[name = tensor("linear_5_cast_fp16")]; + tensor input_21_mode_0 = const()[name = tensor("input_21_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_21_cast_fp16 = gelu(mode = input_21_mode_0, x = linear_5_cast_fp16)[name = tensor("input_21_cast_fp16")]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4140544))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5713472))), name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized"), shape = tensor([768, 2048])]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16 = const()[name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5714048)))]; + tensor linear_6_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_21_cast_fp16)[name = tensor("linear_6_cast_fp16")]; + tensor input_23_cast_fp16 = add(x = linear_6_cast_fp16, y = input_17_cast_fp16)[name = tensor("input_23_cast_fp16")]; + tensor hidden_states_3_axes_0 = const()[name = tensor("hidden_states_3_axes_0"), val = tensor([-1])]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16 = const()[name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5715648)))]; + tensor bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16 = const()[name = tensor("bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5717248)))]; + tensor hidden_states_3_cast_fp16 = layer_norm(axes = hidden_states_3_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_23_cast_fp16)[name = tensor("hidden_states_3_cast_fp16")]; + tensor var_159_shape_cast_fp16 = shape(x = hidden_states_3_cast_fp16)[name = tensor("op_159_shape_cast_fp16")]; + tensor gather_7_axis_0 = const()[name = tensor("gather_7_axis_0"), val = tensor(0)]; + tensor gather_7_batch_dims_0 = const()[name = tensor("gather_7_batch_dims_0"), val = tensor(0)]; + tensor gather_7_validate_indices_0 = const()[name = tensor("gather_7_validate_indices_0"), val = tensor(false)]; + tensor var_159_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_159_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_7_to_uint16 = const()[name = tensor("select_7_to_uint16"), val = tensor(0)]; + tensor var_159_shape_cast_fp16_to_uint16 = cast(dtype = var_159_shape_cast_fp16_to_uint16_dtype_0, x = var_159_shape_cast_fp16)[name = tensor("cast_32")]; + tensor gather_7_cast_uint16 = gather(axis = gather_7_axis_0, batch_dims = gather_7_batch_dims_0, indices = select_7_to_uint16, validate_indices = gather_7_validate_indices_0, x = var_159_shape_cast_fp16_to_uint16)[name = tensor("gather_7_cast_uint16")]; + tensor gather_7_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_7_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_8_axis_0 = const()[name = tensor("gather_8_axis_0"), val = tensor(0)]; + tensor gather_8_batch_dims_0 = const()[name = tensor("gather_8_batch_dims_0"), val = tensor(0)]; + tensor gather_8_validate_indices_0 = const()[name = tensor("gather_8_validate_indices_0"), val = tensor(false)]; + tensor select_8_to_uint16 = const()[name = tensor("select_8_to_uint16"), val = tensor(1)]; + tensor gather_8_cast_uint16 = gather(axis = gather_8_axis_0, batch_dims = gather_8_batch_dims_0, indices = select_8_to_uint16, validate_indices = gather_8_validate_indices_0, x = var_159_shape_cast_fp16_to_uint16)[name = tensor("gather_8_cast_uint16")]; + tensor gather_8_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_8_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor linear_7_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = hidden_states_3_cast_fp16)[name = tensor("linear_7_cast_fp16")]; + tensor concat_9_axis_0 = const()[name = tensor("concat_9_axis_0"), val = tensor(0)]; + tensor concat_9_interleave_0 = const()[name = tensor("concat_9_interleave_0"), val = tensor(false)]; + tensor gather_7_cast_uint16_to_int32 = cast(dtype = gather_7_cast_uint16_to_int32_dtype_0, x = gather_7_cast_uint16)[name = tensor("cast_31")]; + tensor concat_9 = concat(axis = concat_9_axis_0, interleave = concat_9_interleave_0, values = (gather_7_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_9")]; + tensor var_165_cast_fp16 = reshape(shape = concat_9, x = linear_7_cast_fp16)[name = tensor("op_165_cast_fp16")]; + tensor linear_8_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = hidden_states_3_cast_fp16)[name = tensor("linear_8_cast_fp16")]; + tensor var_171_cast_fp16 = reshape(shape = concat_9, x = linear_8_cast_fp16)[name = tensor("op_171_cast_fp16")]; + tensor linear_9_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = hidden_states_3_cast_fp16)[name = tensor("linear_9_cast_fp16")]; + tensor var_177_cast_fp16 = reshape(shape = concat_9, x = linear_9_cast_fp16)[name = tensor("op_177_cast_fp16")]; + tensor value_layer_3_perm_0 = const()[name = tensor("value_layer_3_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_1_y_0_to_fp16 = const()[name = tensor("mul_1_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_1_cast_fp16 = mul(x = var_165_cast_fp16, y = mul_1_y_0_to_fp16)[name = tensor("mul_1_cast_fp16")]; + tensor matmul_1_transpose_y_0 = const()[name = tensor("matmul_1_transpose_y_0"), val = tensor(true)]; + tensor matmul_1_transpose_x_0 = const()[name = tensor("matmul_1_transpose_x_0"), val = tensor(false)]; + tensor transpose_50_perm_0 = const()[name = tensor("transpose_50_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_51_perm_0 = const()[name = tensor("transpose_51_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_51 = transpose(perm = transpose_51_perm_0, x = var_171_cast_fp16)[name = tensor("transpose_114")]; + tensor transpose_50 = transpose(perm = transpose_50_perm_0, x = mul_1_cast_fp16)[name = tensor("transpose_115")]; + tensor matmul_1_cast_fp16 = matmul(transpose_x = matmul_1_transpose_x_0, transpose_y = matmul_1_transpose_y_0, x = transpose_50, y = transpose_51)[name = tensor("matmul_1_cast_fp16")]; + tensor add_10_cast_fp16 = add(x = matmul_1_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_10_cast_fp16")]; + tensor softmax_1_axis_0 = const()[name = tensor("softmax_1_axis_0"), val = tensor(-1)]; + tensor softmax_1_cast_fp16 = softmax(axis = softmax_1_axis_0, x = add_10_cast_fp16)[name = tensor("softmax_1_cast_fp16")]; + tensor attention_output_5_transpose_x_0 = const()[name = tensor("attention_output_5_transpose_x_0"), val = tensor(false)]; + tensor attention_output_5_transpose_y_0 = const()[name = tensor("attention_output_5_transpose_y_0"), val = tensor(false)]; + tensor value_layer_3_cast_fp16 = transpose(perm = value_layer_3_perm_0, x = var_177_cast_fp16)[name = tensor("transpose_113")]; + tensor attention_output_5_cast_fp16 = matmul(transpose_x = attention_output_5_transpose_x_0, transpose_y = attention_output_5_transpose_y_0, x = softmax_1_cast_fp16, y = value_layer_3_cast_fp16)[name = tensor("attention_output_5_cast_fp16")]; + tensor attention_output_7_perm_0 = const()[name = tensor("attention_output_7_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_12_axis_0 = const()[name = tensor("concat_12_axis_0"), val = tensor(0)]; + tensor concat_12_interleave_0 = const()[name = tensor("concat_12_interleave_0"), val = tensor(false)]; + tensor gather_8_cast_uint16_to_int32 = cast(dtype = gather_8_cast_uint16_to_int32_dtype_0, x = gather_8_cast_uint16)[name = tensor("cast_30")]; + tensor concat_12 = concat(axis = concat_12_axis_0, interleave = concat_12_interleave_0, values = (gather_7_cast_uint16_to_int32, gather_8_cast_uint16_to_int32, var_3))[name = tensor("concat_12")]; + tensor attention_output_7_cast_fp16 = transpose(perm = attention_output_7_perm_0, x = attention_output_5_cast_fp16)[name = tensor("transpose_112")]; + tensor input_25_cast_fp16 = reshape(shape = concat_12, x = attention_output_7_cast_fp16)[name = tensor("input_25_cast_fp16")]; + tensor linear_10_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_25_cast_fp16)[name = tensor("linear_10_cast_fp16")]; + tensor input_29_cast_fp16 = add(x = hidden_states_3_cast_fp16, y = linear_10_cast_fp16)[name = tensor("input_29_cast_fp16")]; + tensor input_31_axes_0 = const()[name = tensor("input_31_axes_0"), val = tensor([-1])]; + tensor input_31_cast_fp16 = layer_norm(axes = input_31_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_29_cast_fp16)[name = tensor("input_31_cast_fp16")]; + tensor linear_11_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_31_cast_fp16)[name = tensor("linear_11_cast_fp16")]; + tensor input_35_mode_0 = const()[name = tensor("input_35_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_35_cast_fp16 = gelu(mode = input_35_mode_0, x = linear_11_cast_fp16)[name = tensor("input_35_cast_fp16")]; + tensor linear_12_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_35_cast_fp16)[name = tensor("linear_12_cast_fp16")]; + tensor input_37_cast_fp16 = add(x = linear_12_cast_fp16, y = input_31_cast_fp16)[name = tensor("input_37_cast_fp16")]; + tensor hidden_states_5_axes_0 = const()[name = tensor("hidden_states_5_axes_0"), val = tensor([-1])]; + tensor hidden_states_5_cast_fp16 = layer_norm(axes = hidden_states_5_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_37_cast_fp16)[name = tensor("hidden_states_5_cast_fp16")]; + tensor var_226_shape_cast_fp16 = shape(x = hidden_states_5_cast_fp16)[name = tensor("op_226_shape_cast_fp16")]; + tensor gather_9_axis_0 = const()[name = tensor("gather_9_axis_0"), val = tensor(0)]; + tensor gather_9_batch_dims_0 = const()[name = tensor("gather_9_batch_dims_0"), val = tensor(0)]; + tensor gather_9_validate_indices_0 = const()[name = tensor("gather_9_validate_indices_0"), val = tensor(false)]; + tensor var_226_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_226_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_9_to_uint16 = const()[name = tensor("select_9_to_uint16"), val = tensor(0)]; + tensor var_226_shape_cast_fp16_to_uint16 = cast(dtype = var_226_shape_cast_fp16_to_uint16_dtype_0, x = var_226_shape_cast_fp16)[name = tensor("cast_29")]; + tensor gather_9_cast_uint16 = gather(axis = gather_9_axis_0, batch_dims = gather_9_batch_dims_0, indices = select_9_to_uint16, validate_indices = gather_9_validate_indices_0, x = var_226_shape_cast_fp16_to_uint16)[name = tensor("gather_9_cast_uint16")]; + tensor gather_9_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_9_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_10_axis_0 = const()[name = tensor("gather_10_axis_0"), val = tensor(0)]; + tensor gather_10_batch_dims_0 = const()[name = tensor("gather_10_batch_dims_0"), val = tensor(0)]; + tensor gather_10_validate_indices_0 = const()[name = tensor("gather_10_validate_indices_0"), val = tensor(false)]; + tensor select_10_to_uint16 = const()[name = tensor("select_10_to_uint16"), val = tensor(1)]; + tensor gather_10_cast_uint16 = gather(axis = gather_10_axis_0, batch_dims = gather_10_batch_dims_0, indices = select_10_to_uint16, validate_indices = gather_10_validate_indices_0, x = var_226_shape_cast_fp16_to_uint16)[name = tensor("gather_10_cast_uint16")]; + tensor gather_10_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_10_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor linear_13_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = hidden_states_5_cast_fp16)[name = tensor("linear_13_cast_fp16")]; + tensor concat_13_axis_0 = const()[name = tensor("concat_13_axis_0"), val = tensor(0)]; + tensor concat_13_interleave_0 = const()[name = tensor("concat_13_interleave_0"), val = tensor(false)]; + tensor gather_9_cast_uint16_to_int32 = cast(dtype = gather_9_cast_uint16_to_int32_dtype_0, x = gather_9_cast_uint16)[name = tensor("cast_28")]; + tensor concat_13 = concat(axis = concat_13_axis_0, interleave = concat_13_interleave_0, values = (gather_9_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_13")]; + tensor var_232_cast_fp16 = reshape(shape = concat_13, x = linear_13_cast_fp16)[name = tensor("op_232_cast_fp16")]; + tensor linear_14_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = hidden_states_5_cast_fp16)[name = tensor("linear_14_cast_fp16")]; + tensor var_238_cast_fp16 = reshape(shape = concat_13, x = linear_14_cast_fp16)[name = tensor("op_238_cast_fp16")]; + tensor linear_15_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = hidden_states_5_cast_fp16)[name = tensor("linear_15_cast_fp16")]; + tensor var_244_cast_fp16 = reshape(shape = concat_13, x = linear_15_cast_fp16)[name = tensor("op_244_cast_fp16")]; + tensor value_layer_5_perm_0 = const()[name = tensor("value_layer_5_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_2_y_0_to_fp16 = const()[name = tensor("mul_2_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_2_cast_fp16 = mul(x = var_232_cast_fp16, y = mul_2_y_0_to_fp16)[name = tensor("mul_2_cast_fp16")]; + tensor matmul_2_transpose_y_0 = const()[name = tensor("matmul_2_transpose_y_0"), val = tensor(true)]; + tensor matmul_2_transpose_x_0 = const()[name = tensor("matmul_2_transpose_x_0"), val = tensor(false)]; + tensor transpose_52_perm_0 = const()[name = tensor("transpose_52_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_53_perm_0 = const()[name = tensor("transpose_53_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_53 = transpose(perm = transpose_53_perm_0, x = var_238_cast_fp16)[name = tensor("transpose_110")]; + tensor transpose_52 = transpose(perm = transpose_52_perm_0, x = mul_2_cast_fp16)[name = tensor("transpose_111")]; + tensor matmul_2_cast_fp16 = matmul(transpose_x = matmul_2_transpose_x_0, transpose_y = matmul_2_transpose_y_0, x = transpose_52, y = transpose_53)[name = tensor("matmul_2_cast_fp16")]; + tensor add_13_cast_fp16 = add(x = matmul_2_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_13_cast_fp16")]; + tensor softmax_2_axis_0 = const()[name = tensor("softmax_2_axis_0"), val = tensor(-1)]; + tensor softmax_2_cast_fp16 = softmax(axis = softmax_2_axis_0, x = add_13_cast_fp16)[name = tensor("softmax_2_cast_fp16")]; + tensor attention_output_9_transpose_x_0 = const()[name = tensor("attention_output_9_transpose_x_0"), val = tensor(false)]; + tensor attention_output_9_transpose_y_0 = const()[name = tensor("attention_output_9_transpose_y_0"), val = tensor(false)]; + tensor value_layer_5_cast_fp16 = transpose(perm = value_layer_5_perm_0, x = var_244_cast_fp16)[name = tensor("transpose_109")]; + tensor attention_output_9_cast_fp16 = matmul(transpose_x = attention_output_9_transpose_x_0, transpose_y = attention_output_9_transpose_y_0, x = softmax_2_cast_fp16, y = value_layer_5_cast_fp16)[name = tensor("attention_output_9_cast_fp16")]; + tensor attention_output_11_perm_0 = const()[name = tensor("attention_output_11_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_16_axis_0 = const()[name = tensor("concat_16_axis_0"), val = tensor(0)]; + tensor concat_16_interleave_0 = const()[name = tensor("concat_16_interleave_0"), val = tensor(false)]; + tensor gather_10_cast_uint16_to_int32 = cast(dtype = gather_10_cast_uint16_to_int32_dtype_0, x = gather_10_cast_uint16)[name = tensor("cast_27")]; + tensor concat_16 = concat(axis = concat_16_axis_0, interleave = concat_16_interleave_0, values = (gather_9_cast_uint16_to_int32, gather_10_cast_uint16_to_int32, var_3))[name = tensor("concat_16")]; + tensor attention_output_11_cast_fp16 = transpose(perm = attention_output_11_perm_0, x = attention_output_9_cast_fp16)[name = tensor("transpose_108")]; + tensor input_39_cast_fp16 = reshape(shape = concat_16, x = attention_output_11_cast_fp16)[name = tensor("input_39_cast_fp16")]; + tensor linear_16_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_39_cast_fp16)[name = tensor("linear_16_cast_fp16")]; + tensor input_43_cast_fp16 = add(x = hidden_states_5_cast_fp16, y = linear_16_cast_fp16)[name = tensor("input_43_cast_fp16")]; + tensor input_45_axes_0 = const()[name = tensor("input_45_axes_0"), val = tensor([-1])]; + tensor input_45_cast_fp16 = layer_norm(axes = input_45_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_43_cast_fp16)[name = tensor("input_45_cast_fp16")]; + tensor linear_17_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_45_cast_fp16)[name = tensor("linear_17_cast_fp16")]; + tensor input_49_mode_0 = const()[name = tensor("input_49_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_49_cast_fp16 = gelu(mode = input_49_mode_0, x = linear_17_cast_fp16)[name = tensor("input_49_cast_fp16")]; + tensor linear_18_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_49_cast_fp16)[name = tensor("linear_18_cast_fp16")]; + tensor input_51_cast_fp16 = add(x = linear_18_cast_fp16, y = input_45_cast_fp16)[name = tensor("input_51_cast_fp16")]; + tensor hidden_states_7_axes_0 = const()[name = tensor("hidden_states_7_axes_0"), val = tensor([-1])]; + tensor hidden_states_7_cast_fp16 = layer_norm(axes = hidden_states_7_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_51_cast_fp16)[name = tensor("hidden_states_7_cast_fp16")]; + tensor var_293_shape_cast_fp16 = shape(x = hidden_states_7_cast_fp16)[name = tensor("op_293_shape_cast_fp16")]; + tensor gather_11_axis_0 = const()[name = tensor("gather_11_axis_0"), val = tensor(0)]; + tensor gather_11_batch_dims_0 = const()[name = tensor("gather_11_batch_dims_0"), val = tensor(0)]; + tensor gather_11_validate_indices_0 = const()[name = tensor("gather_11_validate_indices_0"), val = tensor(false)]; + tensor var_293_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_293_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_11_to_uint16 = const()[name = tensor("select_11_to_uint16"), val = tensor(0)]; + tensor var_293_shape_cast_fp16_to_uint16 = cast(dtype = var_293_shape_cast_fp16_to_uint16_dtype_0, x = var_293_shape_cast_fp16)[name = tensor("cast_26")]; + tensor gather_11_cast_uint16 = gather(axis = gather_11_axis_0, batch_dims = gather_11_batch_dims_0, indices = select_11_to_uint16, validate_indices = gather_11_validate_indices_0, x = var_293_shape_cast_fp16_to_uint16)[name = tensor("gather_11_cast_uint16")]; + tensor gather_11_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_11_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_12_axis_0 = const()[name = tensor("gather_12_axis_0"), val = tensor(0)]; + tensor gather_12_batch_dims_0 = const()[name = tensor("gather_12_batch_dims_0"), val = tensor(0)]; + tensor gather_12_validate_indices_0 = const()[name = tensor("gather_12_validate_indices_0"), val = tensor(false)]; + tensor select_12_to_uint16 = const()[name = tensor("select_12_to_uint16"), val = tensor(1)]; + tensor gather_12_cast_uint16 = gather(axis = gather_12_axis_0, batch_dims = gather_12_batch_dims_0, indices = select_12_to_uint16, validate_indices = gather_12_validate_indices_0, x = var_293_shape_cast_fp16_to_uint16)[name = tensor("gather_12_cast_uint16")]; + tensor gather_12_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_12_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor linear_19_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = hidden_states_7_cast_fp16)[name = tensor("linear_19_cast_fp16")]; + tensor concat_17_axis_0 = const()[name = tensor("concat_17_axis_0"), val = tensor(0)]; + tensor concat_17_interleave_0 = const()[name = tensor("concat_17_interleave_0"), val = tensor(false)]; + tensor gather_11_cast_uint16_to_int32 = cast(dtype = gather_11_cast_uint16_to_int32_dtype_0, x = gather_11_cast_uint16)[name = tensor("cast_25")]; + tensor concat_17 = concat(axis = concat_17_axis_0, interleave = concat_17_interleave_0, values = (gather_11_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_17")]; + tensor var_299_cast_fp16 = reshape(shape = concat_17, x = linear_19_cast_fp16)[name = tensor("op_299_cast_fp16")]; + tensor linear_20_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = hidden_states_7_cast_fp16)[name = tensor("linear_20_cast_fp16")]; + tensor var_305_cast_fp16 = reshape(shape = concat_17, x = linear_20_cast_fp16)[name = tensor("op_305_cast_fp16")]; + tensor linear_21_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = hidden_states_7_cast_fp16)[name = tensor("linear_21_cast_fp16")]; + tensor var_311_cast_fp16 = reshape(shape = concat_17, x = linear_21_cast_fp16)[name = tensor("op_311_cast_fp16")]; + tensor value_layer_7_perm_0 = const()[name = tensor("value_layer_7_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_3_y_0_to_fp16 = const()[name = tensor("mul_3_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_3_cast_fp16 = mul(x = var_299_cast_fp16, y = mul_3_y_0_to_fp16)[name = tensor("mul_3_cast_fp16")]; + tensor matmul_3_transpose_y_0 = const()[name = tensor("matmul_3_transpose_y_0"), val = tensor(true)]; + tensor matmul_3_transpose_x_0 = const()[name = tensor("matmul_3_transpose_x_0"), val = tensor(false)]; + tensor transpose_54_perm_0 = const()[name = tensor("transpose_54_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_55_perm_0 = const()[name = tensor("transpose_55_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_55 = transpose(perm = transpose_55_perm_0, x = var_305_cast_fp16)[name = tensor("transpose_106")]; + tensor transpose_54 = transpose(perm = transpose_54_perm_0, x = mul_3_cast_fp16)[name = tensor("transpose_107")]; + tensor matmul_3_cast_fp16 = matmul(transpose_x = matmul_3_transpose_x_0, transpose_y = matmul_3_transpose_y_0, x = transpose_54, y = transpose_55)[name = tensor("matmul_3_cast_fp16")]; + tensor add_16_cast_fp16 = add(x = matmul_3_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_16_cast_fp16")]; + tensor softmax_3_axis_0 = const()[name = tensor("softmax_3_axis_0"), val = tensor(-1)]; + tensor softmax_3_cast_fp16 = softmax(axis = softmax_3_axis_0, x = add_16_cast_fp16)[name = tensor("softmax_3_cast_fp16")]; + tensor attention_output_13_transpose_x_0 = const()[name = tensor("attention_output_13_transpose_x_0"), val = tensor(false)]; + tensor attention_output_13_transpose_y_0 = const()[name = tensor("attention_output_13_transpose_y_0"), val = tensor(false)]; + tensor value_layer_7_cast_fp16 = transpose(perm = value_layer_7_perm_0, x = var_311_cast_fp16)[name = tensor("transpose_105")]; + tensor attention_output_13_cast_fp16 = matmul(transpose_x = attention_output_13_transpose_x_0, transpose_y = attention_output_13_transpose_y_0, x = softmax_3_cast_fp16, y = value_layer_7_cast_fp16)[name = tensor("attention_output_13_cast_fp16")]; + tensor attention_output_15_perm_0 = const()[name = tensor("attention_output_15_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_20_axis_0 = const()[name = tensor("concat_20_axis_0"), val = tensor(0)]; + tensor concat_20_interleave_0 = const()[name = tensor("concat_20_interleave_0"), val = tensor(false)]; + tensor gather_12_cast_uint16_to_int32 = cast(dtype = gather_12_cast_uint16_to_int32_dtype_0, x = gather_12_cast_uint16)[name = tensor("cast_24")]; + tensor concat_20 = concat(axis = concat_20_axis_0, interleave = concat_20_interleave_0, values = (gather_11_cast_uint16_to_int32, gather_12_cast_uint16_to_int32, var_3))[name = tensor("concat_20")]; + tensor attention_output_15_cast_fp16 = transpose(perm = attention_output_15_perm_0, x = attention_output_13_cast_fp16)[name = tensor("transpose_104")]; + tensor input_53_cast_fp16 = reshape(shape = concat_20, x = attention_output_15_cast_fp16)[name = tensor("input_53_cast_fp16")]; + tensor linear_22_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_53_cast_fp16)[name = tensor("linear_22_cast_fp16")]; + tensor input_57_cast_fp16 = add(x = hidden_states_7_cast_fp16, y = linear_22_cast_fp16)[name = tensor("input_57_cast_fp16")]; + tensor input_59_axes_0 = const()[name = tensor("input_59_axes_0"), val = tensor([-1])]; + tensor input_59_cast_fp16 = layer_norm(axes = input_59_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_57_cast_fp16)[name = tensor("input_59_cast_fp16")]; + tensor linear_23_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_59_cast_fp16)[name = tensor("linear_23_cast_fp16")]; + tensor input_63_mode_0 = const()[name = tensor("input_63_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_63_cast_fp16 = gelu(mode = input_63_mode_0, x = linear_23_cast_fp16)[name = tensor("input_63_cast_fp16")]; + tensor linear_24_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_63_cast_fp16)[name = tensor("linear_24_cast_fp16")]; + tensor input_65_cast_fp16 = add(x = linear_24_cast_fp16, y = input_59_cast_fp16)[name = tensor("input_65_cast_fp16")]; + tensor hidden_states_9_axes_0 = const()[name = tensor("hidden_states_9_axes_0"), val = tensor([-1])]; + tensor hidden_states_9_cast_fp16 = layer_norm(axes = hidden_states_9_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_65_cast_fp16)[name = tensor("hidden_states_9_cast_fp16")]; + tensor var_360_shape_cast_fp16 = shape(x = hidden_states_9_cast_fp16)[name = tensor("op_360_shape_cast_fp16")]; + tensor gather_13_axis_0 = const()[name = tensor("gather_13_axis_0"), val = tensor(0)]; + tensor gather_13_batch_dims_0 = const()[name = tensor("gather_13_batch_dims_0"), val = tensor(0)]; + tensor gather_13_validate_indices_0 = const()[name = tensor("gather_13_validate_indices_0"), val = tensor(false)]; + tensor var_360_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_360_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_13_to_uint16 = const()[name = tensor("select_13_to_uint16"), val = tensor(0)]; + tensor var_360_shape_cast_fp16_to_uint16 = cast(dtype = var_360_shape_cast_fp16_to_uint16_dtype_0, x = var_360_shape_cast_fp16)[name = tensor("cast_23")]; + tensor gather_13_cast_uint16 = gather(axis = gather_13_axis_0, batch_dims = gather_13_batch_dims_0, indices = select_13_to_uint16, validate_indices = gather_13_validate_indices_0, x = var_360_shape_cast_fp16_to_uint16)[name = tensor("gather_13_cast_uint16")]; + tensor gather_13_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_13_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_14_axis_0 = const()[name = tensor("gather_14_axis_0"), val = tensor(0)]; + tensor gather_14_batch_dims_0 = const()[name = tensor("gather_14_batch_dims_0"), val = tensor(0)]; + tensor gather_14_validate_indices_0 = const()[name = tensor("gather_14_validate_indices_0"), val = tensor(false)]; + tensor select_14_to_uint16 = const()[name = tensor("select_14_to_uint16"), val = tensor(1)]; + tensor gather_14_cast_uint16 = gather(axis = gather_14_axis_0, batch_dims = gather_14_batch_dims_0, indices = select_14_to_uint16, validate_indices = gather_14_validate_indices_0, x = var_360_shape_cast_fp16_to_uint16)[name = tensor("gather_14_cast_uint16")]; + tensor gather_14_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_14_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor linear_25_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = hidden_states_9_cast_fp16)[name = tensor("linear_25_cast_fp16")]; + tensor concat_21_axis_0 = const()[name = tensor("concat_21_axis_0"), val = tensor(0)]; + tensor concat_21_interleave_0 = const()[name = tensor("concat_21_interleave_0"), val = tensor(false)]; + tensor gather_13_cast_uint16_to_int32 = cast(dtype = gather_13_cast_uint16_to_int32_dtype_0, x = gather_13_cast_uint16)[name = tensor("cast_22")]; + tensor concat_21 = concat(axis = concat_21_axis_0, interleave = concat_21_interleave_0, values = (gather_13_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_21")]; + tensor var_366_cast_fp16 = reshape(shape = concat_21, x = linear_25_cast_fp16)[name = tensor("op_366_cast_fp16")]; + tensor linear_26_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = hidden_states_9_cast_fp16)[name = tensor("linear_26_cast_fp16")]; + tensor var_372_cast_fp16 = reshape(shape = concat_21, x = linear_26_cast_fp16)[name = tensor("op_372_cast_fp16")]; + tensor linear_27_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = hidden_states_9_cast_fp16)[name = tensor("linear_27_cast_fp16")]; + tensor var_378_cast_fp16 = reshape(shape = concat_21, x = linear_27_cast_fp16)[name = tensor("op_378_cast_fp16")]; + tensor value_layer_9_perm_0 = const()[name = tensor("value_layer_9_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_4_y_0_to_fp16 = const()[name = tensor("mul_4_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_4_cast_fp16 = mul(x = var_366_cast_fp16, y = mul_4_y_0_to_fp16)[name = tensor("mul_4_cast_fp16")]; + tensor matmul_4_transpose_y_0 = const()[name = tensor("matmul_4_transpose_y_0"), val = tensor(true)]; + tensor matmul_4_transpose_x_0 = const()[name = tensor("matmul_4_transpose_x_0"), val = tensor(false)]; + tensor transpose_56_perm_0 = const()[name = tensor("transpose_56_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_57_perm_0 = const()[name = tensor("transpose_57_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_57 = transpose(perm = transpose_57_perm_0, x = var_372_cast_fp16)[name = tensor("transpose_102")]; + tensor transpose_56 = transpose(perm = transpose_56_perm_0, x = mul_4_cast_fp16)[name = tensor("transpose_103")]; + tensor matmul_4_cast_fp16 = matmul(transpose_x = matmul_4_transpose_x_0, transpose_y = matmul_4_transpose_y_0, x = transpose_56, y = transpose_57)[name = tensor("matmul_4_cast_fp16")]; + tensor add_19_cast_fp16 = add(x = matmul_4_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_19_cast_fp16")]; + tensor softmax_4_axis_0 = const()[name = tensor("softmax_4_axis_0"), val = tensor(-1)]; + tensor softmax_4_cast_fp16 = softmax(axis = softmax_4_axis_0, x = add_19_cast_fp16)[name = tensor("softmax_4_cast_fp16")]; + tensor attention_output_17_transpose_x_0 = const()[name = tensor("attention_output_17_transpose_x_0"), val = tensor(false)]; + tensor attention_output_17_transpose_y_0 = const()[name = tensor("attention_output_17_transpose_y_0"), val = tensor(false)]; + tensor value_layer_9_cast_fp16 = transpose(perm = value_layer_9_perm_0, x = var_378_cast_fp16)[name = tensor("transpose_101")]; + tensor attention_output_17_cast_fp16 = matmul(transpose_x = attention_output_17_transpose_x_0, transpose_y = attention_output_17_transpose_y_0, x = softmax_4_cast_fp16, y = value_layer_9_cast_fp16)[name = tensor("attention_output_17_cast_fp16")]; + tensor attention_output_19_perm_0 = const()[name = tensor("attention_output_19_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_24_axis_0 = const()[name = tensor("concat_24_axis_0"), val = tensor(0)]; + tensor concat_24_interleave_0 = const()[name = tensor("concat_24_interleave_0"), val = tensor(false)]; + tensor gather_14_cast_uint16_to_int32 = cast(dtype = gather_14_cast_uint16_to_int32_dtype_0, x = gather_14_cast_uint16)[name = tensor("cast_21")]; + tensor concat_24 = concat(axis = concat_24_axis_0, interleave = concat_24_interleave_0, values = (gather_13_cast_uint16_to_int32, gather_14_cast_uint16_to_int32, var_3))[name = tensor("concat_24")]; + tensor attention_output_19_cast_fp16 = transpose(perm = attention_output_19_perm_0, x = attention_output_17_cast_fp16)[name = tensor("transpose_100")]; + tensor input_67_cast_fp16 = reshape(shape = concat_24, x = attention_output_19_cast_fp16)[name = tensor("input_67_cast_fp16")]; + tensor linear_28_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_67_cast_fp16)[name = tensor("linear_28_cast_fp16")]; + tensor input_71_cast_fp16 = add(x = hidden_states_9_cast_fp16, y = linear_28_cast_fp16)[name = tensor("input_71_cast_fp16")]; + tensor input_73_axes_0 = const()[name = tensor("input_73_axes_0"), val = tensor([-1])]; + tensor input_73_cast_fp16 = layer_norm(axes = input_73_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_71_cast_fp16)[name = tensor("input_73_cast_fp16")]; + tensor linear_29_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_73_cast_fp16)[name = tensor("linear_29_cast_fp16")]; + tensor input_77_mode_0 = const()[name = tensor("input_77_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_77_cast_fp16 = gelu(mode = input_77_mode_0, x = linear_29_cast_fp16)[name = tensor("input_77_cast_fp16")]; + tensor linear_30_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_77_cast_fp16)[name = tensor("linear_30_cast_fp16")]; + tensor input_79_cast_fp16 = add(x = linear_30_cast_fp16, y = input_73_cast_fp16)[name = tensor("input_79_cast_fp16")]; + tensor hidden_states_11_axes_0 = const()[name = tensor("hidden_states_11_axes_0"), val = tensor([-1])]; + tensor hidden_states_11_cast_fp16 = layer_norm(axes = hidden_states_11_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_79_cast_fp16)[name = tensor("hidden_states_11_cast_fp16")]; + tensor var_427_shape_cast_fp16 = shape(x = hidden_states_11_cast_fp16)[name = tensor("op_427_shape_cast_fp16")]; + tensor gather_15_axis_0 = const()[name = tensor("gather_15_axis_0"), val = tensor(0)]; + tensor gather_15_batch_dims_0 = const()[name = tensor("gather_15_batch_dims_0"), val = tensor(0)]; + tensor gather_15_validate_indices_0 = const()[name = tensor("gather_15_validate_indices_0"), val = tensor(false)]; + tensor var_427_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_427_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_15_to_uint16 = const()[name = tensor("select_15_to_uint16"), val = tensor(0)]; + tensor var_427_shape_cast_fp16_to_uint16 = cast(dtype = var_427_shape_cast_fp16_to_uint16_dtype_0, x = var_427_shape_cast_fp16)[name = tensor("cast_20")]; + tensor gather_15_cast_uint16 = gather(axis = gather_15_axis_0, batch_dims = gather_15_batch_dims_0, indices = select_15_to_uint16, validate_indices = gather_15_validate_indices_0, x = var_427_shape_cast_fp16_to_uint16)[name = tensor("gather_15_cast_uint16")]; + tensor gather_15_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_15_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_16_axis_0 = const()[name = tensor("gather_16_axis_0"), val = tensor(0)]; + tensor gather_16_batch_dims_0 = const()[name = tensor("gather_16_batch_dims_0"), val = tensor(0)]; + tensor gather_16_validate_indices_0 = const()[name = tensor("gather_16_validate_indices_0"), val = tensor(false)]; + tensor select_16_to_uint16 = const()[name = tensor("select_16_to_uint16"), val = tensor(1)]; + tensor gather_16_cast_uint16 = gather(axis = gather_16_axis_0, batch_dims = gather_16_batch_dims_0, indices = select_16_to_uint16, validate_indices = gather_16_validate_indices_0, x = var_427_shape_cast_fp16_to_uint16)[name = tensor("gather_16_cast_uint16")]; + tensor gather_16_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_16_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor linear_31_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = hidden_states_11_cast_fp16)[name = tensor("linear_31_cast_fp16")]; + tensor concat_25_axis_0 = const()[name = tensor("concat_25_axis_0"), val = tensor(0)]; + tensor concat_25_interleave_0 = const()[name = tensor("concat_25_interleave_0"), val = tensor(false)]; + tensor gather_15_cast_uint16_to_int32 = cast(dtype = gather_15_cast_uint16_to_int32_dtype_0, x = gather_15_cast_uint16)[name = tensor("cast_19")]; + tensor concat_25 = concat(axis = concat_25_axis_0, interleave = concat_25_interleave_0, values = (gather_15_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_25")]; + tensor var_433_cast_fp16 = reshape(shape = concat_25, x = linear_31_cast_fp16)[name = tensor("op_433_cast_fp16")]; + tensor linear_32_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = hidden_states_11_cast_fp16)[name = tensor("linear_32_cast_fp16")]; + tensor var_439_cast_fp16 = reshape(shape = concat_25, x = linear_32_cast_fp16)[name = tensor("op_439_cast_fp16")]; + tensor linear_33_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = hidden_states_11_cast_fp16)[name = tensor("linear_33_cast_fp16")]; + tensor var_445_cast_fp16 = reshape(shape = concat_25, x = linear_33_cast_fp16)[name = tensor("op_445_cast_fp16")]; + tensor value_layer_11_perm_0 = const()[name = tensor("value_layer_11_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_5_y_0_to_fp16 = const()[name = tensor("mul_5_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_5_cast_fp16 = mul(x = var_433_cast_fp16, y = mul_5_y_0_to_fp16)[name = tensor("mul_5_cast_fp16")]; + tensor matmul_5_transpose_y_0 = const()[name = tensor("matmul_5_transpose_y_0"), val = tensor(true)]; + tensor matmul_5_transpose_x_0 = const()[name = tensor("matmul_5_transpose_x_0"), val = tensor(false)]; + tensor transpose_58_perm_0 = const()[name = tensor("transpose_58_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_59_perm_0 = const()[name = tensor("transpose_59_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_59 = transpose(perm = transpose_59_perm_0, x = var_439_cast_fp16)[name = tensor("transpose_98")]; + tensor transpose_58 = transpose(perm = transpose_58_perm_0, x = mul_5_cast_fp16)[name = tensor("transpose_99")]; + tensor matmul_5_cast_fp16 = matmul(transpose_x = matmul_5_transpose_x_0, transpose_y = matmul_5_transpose_y_0, x = transpose_58, y = transpose_59)[name = tensor("matmul_5_cast_fp16")]; + tensor add_22_cast_fp16 = add(x = matmul_5_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_22_cast_fp16")]; + tensor softmax_5_axis_0 = const()[name = tensor("softmax_5_axis_0"), val = tensor(-1)]; + tensor softmax_5_cast_fp16 = softmax(axis = softmax_5_axis_0, x = add_22_cast_fp16)[name = tensor("softmax_5_cast_fp16")]; + tensor attention_output_21_transpose_x_0 = const()[name = tensor("attention_output_21_transpose_x_0"), val = tensor(false)]; + tensor attention_output_21_transpose_y_0 = const()[name = tensor("attention_output_21_transpose_y_0"), val = tensor(false)]; + tensor value_layer_11_cast_fp16 = transpose(perm = value_layer_11_perm_0, x = var_445_cast_fp16)[name = tensor("transpose_97")]; + tensor attention_output_21_cast_fp16 = matmul(transpose_x = attention_output_21_transpose_x_0, transpose_y = attention_output_21_transpose_y_0, x = softmax_5_cast_fp16, y = value_layer_11_cast_fp16)[name = tensor("attention_output_21_cast_fp16")]; + tensor attention_output_23_perm_0 = const()[name = tensor("attention_output_23_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_28_axis_0 = const()[name = tensor("concat_28_axis_0"), val = tensor(0)]; + tensor concat_28_interleave_0 = const()[name = tensor("concat_28_interleave_0"), val = tensor(false)]; + tensor gather_16_cast_uint16_to_int32 = cast(dtype = gather_16_cast_uint16_to_int32_dtype_0, x = gather_16_cast_uint16)[name = tensor("cast_18")]; + tensor concat_28 = concat(axis = concat_28_axis_0, interleave = concat_28_interleave_0, values = (gather_15_cast_uint16_to_int32, gather_16_cast_uint16_to_int32, var_3))[name = tensor("concat_28")]; + tensor attention_output_23_cast_fp16 = transpose(perm = attention_output_23_perm_0, x = attention_output_21_cast_fp16)[name = tensor("transpose_96")]; + tensor input_81_cast_fp16 = reshape(shape = concat_28, x = attention_output_23_cast_fp16)[name = tensor("input_81_cast_fp16")]; + tensor linear_34_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_81_cast_fp16)[name = tensor("linear_34_cast_fp16")]; + tensor input_85_cast_fp16 = add(x = hidden_states_11_cast_fp16, y = linear_34_cast_fp16)[name = tensor("input_85_cast_fp16")]; + tensor input_87_axes_0 = const()[name = tensor("input_87_axes_0"), val = tensor([-1])]; + tensor input_87_cast_fp16 = layer_norm(axes = input_87_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_85_cast_fp16)[name = tensor("input_87_cast_fp16")]; + tensor linear_35_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_87_cast_fp16)[name = tensor("linear_35_cast_fp16")]; + tensor input_91_mode_0 = const()[name = tensor("input_91_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_91_cast_fp16 = gelu(mode = input_91_mode_0, x = linear_35_cast_fp16)[name = tensor("input_91_cast_fp16")]; + tensor linear_36_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_91_cast_fp16)[name = tensor("linear_36_cast_fp16")]; + tensor input_93_cast_fp16 = add(x = linear_36_cast_fp16, y = input_87_cast_fp16)[name = tensor("input_93_cast_fp16")]; + tensor hidden_states_13_axes_0 = const()[name = tensor("hidden_states_13_axes_0"), val = tensor([-1])]; + tensor hidden_states_13_cast_fp16 = layer_norm(axes = hidden_states_13_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_93_cast_fp16)[name = tensor("hidden_states_13_cast_fp16")]; + tensor var_494_shape_cast_fp16 = shape(x = hidden_states_13_cast_fp16)[name = tensor("op_494_shape_cast_fp16")]; + tensor gather_17_axis_0 = const()[name = tensor("gather_17_axis_0"), val = tensor(0)]; + tensor gather_17_batch_dims_0 = const()[name = tensor("gather_17_batch_dims_0"), val = tensor(0)]; + tensor gather_17_validate_indices_0 = const()[name = tensor("gather_17_validate_indices_0"), val = tensor(false)]; + tensor var_494_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_494_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_17_to_uint16 = const()[name = tensor("select_17_to_uint16"), val = tensor(0)]; + tensor var_494_shape_cast_fp16_to_uint16 = cast(dtype = var_494_shape_cast_fp16_to_uint16_dtype_0, x = var_494_shape_cast_fp16)[name = tensor("cast_17")]; + tensor gather_17_cast_uint16 = gather(axis = gather_17_axis_0, batch_dims = gather_17_batch_dims_0, indices = select_17_to_uint16, validate_indices = gather_17_validate_indices_0, x = var_494_shape_cast_fp16_to_uint16)[name = tensor("gather_17_cast_uint16")]; + tensor gather_17_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_17_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_18_axis_0 = const()[name = tensor("gather_18_axis_0"), val = tensor(0)]; + tensor gather_18_batch_dims_0 = const()[name = tensor("gather_18_batch_dims_0"), val = tensor(0)]; + tensor gather_18_validate_indices_0 = const()[name = tensor("gather_18_validate_indices_0"), val = tensor(false)]; + tensor select_18_to_uint16 = const()[name = tensor("select_18_to_uint16"), val = tensor(1)]; + tensor gather_18_cast_uint16 = gather(axis = gather_18_axis_0, batch_dims = gather_18_batch_dims_0, indices = select_18_to_uint16, validate_indices = gather_18_validate_indices_0, x = var_494_shape_cast_fp16_to_uint16)[name = tensor("gather_18_cast_uint16")]; + tensor gather_18_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_18_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor linear_37_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = hidden_states_13_cast_fp16)[name = tensor("linear_37_cast_fp16")]; + tensor concat_29_axis_0 = const()[name = tensor("concat_29_axis_0"), val = tensor(0)]; + tensor concat_29_interleave_0 = const()[name = tensor("concat_29_interleave_0"), val = tensor(false)]; + tensor gather_17_cast_uint16_to_int32 = cast(dtype = gather_17_cast_uint16_to_int32_dtype_0, x = gather_17_cast_uint16)[name = tensor("cast_16")]; + tensor concat_29 = concat(axis = concat_29_axis_0, interleave = concat_29_interleave_0, values = (gather_17_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_29")]; + tensor var_500_cast_fp16 = reshape(shape = concat_29, x = linear_37_cast_fp16)[name = tensor("op_500_cast_fp16")]; + tensor linear_38_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = hidden_states_13_cast_fp16)[name = tensor("linear_38_cast_fp16")]; + tensor var_506_cast_fp16 = reshape(shape = concat_29, x = linear_38_cast_fp16)[name = tensor("op_506_cast_fp16")]; + tensor linear_39_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = hidden_states_13_cast_fp16)[name = tensor("linear_39_cast_fp16")]; + tensor var_512_cast_fp16 = reshape(shape = concat_29, x = linear_39_cast_fp16)[name = tensor("op_512_cast_fp16")]; + tensor value_layer_13_perm_0 = const()[name = tensor("value_layer_13_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_6_y_0_to_fp16 = const()[name = tensor("mul_6_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_6_cast_fp16 = mul(x = var_500_cast_fp16, y = mul_6_y_0_to_fp16)[name = tensor("mul_6_cast_fp16")]; + tensor matmul_6_transpose_y_0 = const()[name = tensor("matmul_6_transpose_y_0"), val = tensor(true)]; + tensor matmul_6_transpose_x_0 = const()[name = tensor("matmul_6_transpose_x_0"), val = tensor(false)]; + tensor transpose_60_perm_0 = const()[name = tensor("transpose_60_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_61_perm_0 = const()[name = tensor("transpose_61_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_61 = transpose(perm = transpose_61_perm_0, x = var_506_cast_fp16)[name = tensor("transpose_94")]; + tensor transpose_60 = transpose(perm = transpose_60_perm_0, x = mul_6_cast_fp16)[name = tensor("transpose_95")]; + tensor matmul_6_cast_fp16 = matmul(transpose_x = matmul_6_transpose_x_0, transpose_y = matmul_6_transpose_y_0, x = transpose_60, y = transpose_61)[name = tensor("matmul_6_cast_fp16")]; + tensor add_25_cast_fp16 = add(x = matmul_6_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_25_cast_fp16")]; + tensor softmax_6_axis_0 = const()[name = tensor("softmax_6_axis_0"), val = tensor(-1)]; + tensor softmax_6_cast_fp16 = softmax(axis = softmax_6_axis_0, x = add_25_cast_fp16)[name = tensor("softmax_6_cast_fp16")]; + tensor attention_output_25_transpose_x_0 = const()[name = tensor("attention_output_25_transpose_x_0"), val = tensor(false)]; + tensor attention_output_25_transpose_y_0 = const()[name = tensor("attention_output_25_transpose_y_0"), val = tensor(false)]; + tensor value_layer_13_cast_fp16 = transpose(perm = value_layer_13_perm_0, x = var_512_cast_fp16)[name = tensor("transpose_93")]; + tensor attention_output_25_cast_fp16 = matmul(transpose_x = attention_output_25_transpose_x_0, transpose_y = attention_output_25_transpose_y_0, x = softmax_6_cast_fp16, y = value_layer_13_cast_fp16)[name = tensor("attention_output_25_cast_fp16")]; + tensor attention_output_27_perm_0 = const()[name = tensor("attention_output_27_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_32_axis_0 = const()[name = tensor("concat_32_axis_0"), val = tensor(0)]; + tensor concat_32_interleave_0 = const()[name = tensor("concat_32_interleave_0"), val = tensor(false)]; + tensor gather_18_cast_uint16_to_int32 = cast(dtype = gather_18_cast_uint16_to_int32_dtype_0, x = gather_18_cast_uint16)[name = tensor("cast_15")]; + tensor concat_32 = concat(axis = concat_32_axis_0, interleave = concat_32_interleave_0, values = (gather_17_cast_uint16_to_int32, gather_18_cast_uint16_to_int32, var_3))[name = tensor("concat_32")]; + tensor attention_output_27_cast_fp16 = transpose(perm = attention_output_27_perm_0, x = attention_output_25_cast_fp16)[name = tensor("transpose_92")]; + tensor input_95_cast_fp16 = reshape(shape = concat_32, x = attention_output_27_cast_fp16)[name = tensor("input_95_cast_fp16")]; + tensor linear_40_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_95_cast_fp16)[name = tensor("linear_40_cast_fp16")]; + tensor input_99_cast_fp16 = add(x = hidden_states_13_cast_fp16, y = linear_40_cast_fp16)[name = tensor("input_99_cast_fp16")]; + tensor input_101_axes_0 = const()[name = tensor("input_101_axes_0"), val = tensor([-1])]; + tensor input_101_cast_fp16 = layer_norm(axes = input_101_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_99_cast_fp16)[name = tensor("input_101_cast_fp16")]; + tensor linear_41_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_101_cast_fp16)[name = tensor("linear_41_cast_fp16")]; + tensor input_105_mode_0 = const()[name = tensor("input_105_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_105_cast_fp16 = gelu(mode = input_105_mode_0, x = linear_41_cast_fp16)[name = tensor("input_105_cast_fp16")]; + tensor linear_42_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_105_cast_fp16)[name = tensor("linear_42_cast_fp16")]; + tensor input_107_cast_fp16 = add(x = linear_42_cast_fp16, y = input_101_cast_fp16)[name = tensor("input_107_cast_fp16")]; + tensor hidden_states_15_axes_0 = const()[name = tensor("hidden_states_15_axes_0"), val = tensor([-1])]; + tensor hidden_states_15_cast_fp16 = layer_norm(axes = hidden_states_15_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_107_cast_fp16)[name = tensor("hidden_states_15_cast_fp16")]; + tensor var_561_shape_cast_fp16 = shape(x = hidden_states_15_cast_fp16)[name = tensor("op_561_shape_cast_fp16")]; + tensor gather_19_axis_0 = const()[name = tensor("gather_19_axis_0"), val = tensor(0)]; + tensor gather_19_batch_dims_0 = const()[name = tensor("gather_19_batch_dims_0"), val = tensor(0)]; + tensor gather_19_validate_indices_0 = const()[name = tensor("gather_19_validate_indices_0"), val = tensor(false)]; + tensor var_561_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_561_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_19_to_uint16 = const()[name = tensor("select_19_to_uint16"), val = tensor(0)]; + tensor var_561_shape_cast_fp16_to_uint16 = cast(dtype = var_561_shape_cast_fp16_to_uint16_dtype_0, x = var_561_shape_cast_fp16)[name = tensor("cast_14")]; + tensor gather_19_cast_uint16 = gather(axis = gather_19_axis_0, batch_dims = gather_19_batch_dims_0, indices = select_19_to_uint16, validate_indices = gather_19_validate_indices_0, x = var_561_shape_cast_fp16_to_uint16)[name = tensor("gather_19_cast_uint16")]; + tensor gather_19_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_19_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_20_axis_0 = const()[name = tensor("gather_20_axis_0"), val = tensor(0)]; + tensor gather_20_batch_dims_0 = const()[name = tensor("gather_20_batch_dims_0"), val = tensor(0)]; + tensor gather_20_validate_indices_0 = const()[name = tensor("gather_20_validate_indices_0"), val = tensor(false)]; + tensor select_20_to_uint16 = const()[name = tensor("select_20_to_uint16"), val = tensor(1)]; + tensor gather_20_cast_uint16 = gather(axis = gather_20_axis_0, batch_dims = gather_20_batch_dims_0, indices = select_20_to_uint16, validate_indices = gather_20_validate_indices_0, x = var_561_shape_cast_fp16_to_uint16)[name = tensor("gather_20_cast_uint16")]; + tensor gather_20_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_20_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor linear_43_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = hidden_states_15_cast_fp16)[name = tensor("linear_43_cast_fp16")]; + tensor concat_33_axis_0 = const()[name = tensor("concat_33_axis_0"), val = tensor(0)]; + tensor concat_33_interleave_0 = const()[name = tensor("concat_33_interleave_0"), val = tensor(false)]; + tensor gather_19_cast_uint16_to_int32 = cast(dtype = gather_19_cast_uint16_to_int32_dtype_0, x = gather_19_cast_uint16)[name = tensor("cast_13")]; + tensor concat_33 = concat(axis = concat_33_axis_0, interleave = concat_33_interleave_0, values = (gather_19_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_33")]; + tensor var_567_cast_fp16 = reshape(shape = concat_33, x = linear_43_cast_fp16)[name = tensor("op_567_cast_fp16")]; + tensor linear_44_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = hidden_states_15_cast_fp16)[name = tensor("linear_44_cast_fp16")]; + tensor var_573_cast_fp16 = reshape(shape = concat_33, x = linear_44_cast_fp16)[name = tensor("op_573_cast_fp16")]; + tensor linear_45_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = hidden_states_15_cast_fp16)[name = tensor("linear_45_cast_fp16")]; + tensor var_579_cast_fp16 = reshape(shape = concat_33, x = linear_45_cast_fp16)[name = tensor("op_579_cast_fp16")]; + tensor value_layer_15_perm_0 = const()[name = tensor("value_layer_15_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_7_y_0_to_fp16 = const()[name = tensor("mul_7_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_7_cast_fp16 = mul(x = var_567_cast_fp16, y = mul_7_y_0_to_fp16)[name = tensor("mul_7_cast_fp16")]; + tensor matmul_7_transpose_y_0 = const()[name = tensor("matmul_7_transpose_y_0"), val = tensor(true)]; + tensor matmul_7_transpose_x_0 = const()[name = tensor("matmul_7_transpose_x_0"), val = tensor(false)]; + tensor transpose_62_perm_0 = const()[name = tensor("transpose_62_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_63_perm_0 = const()[name = tensor("transpose_63_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_63 = transpose(perm = transpose_63_perm_0, x = var_573_cast_fp16)[name = tensor("transpose_90")]; + tensor transpose_62 = transpose(perm = transpose_62_perm_0, x = mul_7_cast_fp16)[name = tensor("transpose_91")]; + tensor matmul_7_cast_fp16 = matmul(transpose_x = matmul_7_transpose_x_0, transpose_y = matmul_7_transpose_y_0, x = transpose_62, y = transpose_63)[name = tensor("matmul_7_cast_fp16")]; + tensor add_28_cast_fp16 = add(x = matmul_7_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_28_cast_fp16")]; + tensor softmax_7_axis_0 = const()[name = tensor("softmax_7_axis_0"), val = tensor(-1)]; + tensor softmax_7_cast_fp16 = softmax(axis = softmax_7_axis_0, x = add_28_cast_fp16)[name = tensor("softmax_7_cast_fp16")]; + tensor attention_output_29_transpose_x_0 = const()[name = tensor("attention_output_29_transpose_x_0"), val = tensor(false)]; + tensor attention_output_29_transpose_y_0 = const()[name = tensor("attention_output_29_transpose_y_0"), val = tensor(false)]; + tensor value_layer_15_cast_fp16 = transpose(perm = value_layer_15_perm_0, x = var_579_cast_fp16)[name = tensor("transpose_89")]; + tensor attention_output_29_cast_fp16 = matmul(transpose_x = attention_output_29_transpose_x_0, transpose_y = attention_output_29_transpose_y_0, x = softmax_7_cast_fp16, y = value_layer_15_cast_fp16)[name = tensor("attention_output_29_cast_fp16")]; + tensor attention_output_31_perm_0 = const()[name = tensor("attention_output_31_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_36_axis_0 = const()[name = tensor("concat_36_axis_0"), val = tensor(0)]; + tensor concat_36_interleave_0 = const()[name = tensor("concat_36_interleave_0"), val = tensor(false)]; + tensor gather_20_cast_uint16_to_int32 = cast(dtype = gather_20_cast_uint16_to_int32_dtype_0, x = gather_20_cast_uint16)[name = tensor("cast_12")]; + tensor concat_36 = concat(axis = concat_36_axis_0, interleave = concat_36_interleave_0, values = (gather_19_cast_uint16_to_int32, gather_20_cast_uint16_to_int32, var_3))[name = tensor("concat_36")]; + tensor attention_output_31_cast_fp16 = transpose(perm = attention_output_31_perm_0, x = attention_output_29_cast_fp16)[name = tensor("transpose_88")]; + tensor input_109_cast_fp16 = reshape(shape = concat_36, x = attention_output_31_cast_fp16)[name = tensor("input_109_cast_fp16")]; + tensor linear_46_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_109_cast_fp16)[name = tensor("linear_46_cast_fp16")]; + tensor input_113_cast_fp16 = add(x = hidden_states_15_cast_fp16, y = linear_46_cast_fp16)[name = tensor("input_113_cast_fp16")]; + tensor input_115_axes_0 = const()[name = tensor("input_115_axes_0"), val = tensor([-1])]; + tensor input_115_cast_fp16 = layer_norm(axes = input_115_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_113_cast_fp16)[name = tensor("input_115_cast_fp16")]; + tensor linear_47_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_115_cast_fp16)[name = tensor("linear_47_cast_fp16")]; + tensor input_119_mode_0 = const()[name = tensor("input_119_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_119_cast_fp16 = gelu(mode = input_119_mode_0, x = linear_47_cast_fp16)[name = tensor("input_119_cast_fp16")]; + tensor linear_48_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_119_cast_fp16)[name = tensor("linear_48_cast_fp16")]; + tensor input_121_cast_fp16 = add(x = linear_48_cast_fp16, y = input_115_cast_fp16)[name = tensor("input_121_cast_fp16")]; + tensor hidden_states_17_axes_0 = const()[name = tensor("hidden_states_17_axes_0"), val = tensor([-1])]; + tensor hidden_states_17_cast_fp16 = layer_norm(axes = hidden_states_17_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_121_cast_fp16)[name = tensor("hidden_states_17_cast_fp16")]; + tensor var_628_shape_cast_fp16 = shape(x = hidden_states_17_cast_fp16)[name = tensor("op_628_shape_cast_fp16")]; + tensor gather_21_axis_0 = const()[name = tensor("gather_21_axis_0"), val = tensor(0)]; + tensor gather_21_batch_dims_0 = const()[name = tensor("gather_21_batch_dims_0"), val = tensor(0)]; + tensor gather_21_validate_indices_0 = const()[name = tensor("gather_21_validate_indices_0"), val = tensor(false)]; + tensor var_628_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_628_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_21_to_uint16 = const()[name = tensor("select_21_to_uint16"), val = tensor(0)]; + tensor var_628_shape_cast_fp16_to_uint16 = cast(dtype = var_628_shape_cast_fp16_to_uint16_dtype_0, x = var_628_shape_cast_fp16)[name = tensor("cast_11")]; + tensor gather_21_cast_uint16 = gather(axis = gather_21_axis_0, batch_dims = gather_21_batch_dims_0, indices = select_21_to_uint16, validate_indices = gather_21_validate_indices_0, x = var_628_shape_cast_fp16_to_uint16)[name = tensor("gather_21_cast_uint16")]; + tensor gather_21_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_21_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_22_axis_0 = const()[name = tensor("gather_22_axis_0"), val = tensor(0)]; + tensor gather_22_batch_dims_0 = const()[name = tensor("gather_22_batch_dims_0"), val = tensor(0)]; + tensor gather_22_validate_indices_0 = const()[name = tensor("gather_22_validate_indices_0"), val = tensor(false)]; + tensor select_22_to_uint16 = const()[name = tensor("select_22_to_uint16"), val = tensor(1)]; + tensor gather_22_cast_uint16 = gather(axis = gather_22_axis_0, batch_dims = gather_22_batch_dims_0, indices = select_22_to_uint16, validate_indices = gather_22_validate_indices_0, x = var_628_shape_cast_fp16_to_uint16)[name = tensor("gather_22_cast_uint16")]; + tensor gather_22_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_22_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor linear_49_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = hidden_states_17_cast_fp16)[name = tensor("linear_49_cast_fp16")]; + tensor concat_37_axis_0 = const()[name = tensor("concat_37_axis_0"), val = tensor(0)]; + tensor concat_37_interleave_0 = const()[name = tensor("concat_37_interleave_0"), val = tensor(false)]; + tensor gather_21_cast_uint16_to_int32 = cast(dtype = gather_21_cast_uint16_to_int32_dtype_0, x = gather_21_cast_uint16)[name = tensor("cast_10")]; + tensor concat_37 = concat(axis = concat_37_axis_0, interleave = concat_37_interleave_0, values = (gather_21_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_37")]; + tensor var_634_cast_fp16 = reshape(shape = concat_37, x = linear_49_cast_fp16)[name = tensor("op_634_cast_fp16")]; + tensor linear_50_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = hidden_states_17_cast_fp16)[name = tensor("linear_50_cast_fp16")]; + tensor var_640_cast_fp16 = reshape(shape = concat_37, x = linear_50_cast_fp16)[name = tensor("op_640_cast_fp16")]; + tensor linear_51_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = hidden_states_17_cast_fp16)[name = tensor("linear_51_cast_fp16")]; + tensor var_646_cast_fp16 = reshape(shape = concat_37, x = linear_51_cast_fp16)[name = tensor("op_646_cast_fp16")]; + tensor value_layer_17_perm_0 = const()[name = tensor("value_layer_17_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_8_y_0_to_fp16 = const()[name = tensor("mul_8_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_8_cast_fp16 = mul(x = var_634_cast_fp16, y = mul_8_y_0_to_fp16)[name = tensor("mul_8_cast_fp16")]; + tensor matmul_8_transpose_y_0 = const()[name = tensor("matmul_8_transpose_y_0"), val = tensor(true)]; + tensor matmul_8_transpose_x_0 = const()[name = tensor("matmul_8_transpose_x_0"), val = tensor(false)]; + tensor transpose_64_perm_0 = const()[name = tensor("transpose_64_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_65_perm_0 = const()[name = tensor("transpose_65_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_65 = transpose(perm = transpose_65_perm_0, x = var_640_cast_fp16)[name = tensor("transpose_86")]; + tensor transpose_64 = transpose(perm = transpose_64_perm_0, x = mul_8_cast_fp16)[name = tensor("transpose_87")]; + tensor matmul_8_cast_fp16 = matmul(transpose_x = matmul_8_transpose_x_0, transpose_y = matmul_8_transpose_y_0, x = transpose_64, y = transpose_65)[name = tensor("matmul_8_cast_fp16")]; + tensor add_31_cast_fp16 = add(x = matmul_8_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_31_cast_fp16")]; + tensor softmax_8_axis_0 = const()[name = tensor("softmax_8_axis_0"), val = tensor(-1)]; + tensor softmax_8_cast_fp16 = softmax(axis = softmax_8_axis_0, x = add_31_cast_fp16)[name = tensor("softmax_8_cast_fp16")]; + tensor attention_output_33_transpose_x_0 = const()[name = tensor("attention_output_33_transpose_x_0"), val = tensor(false)]; + tensor attention_output_33_transpose_y_0 = const()[name = tensor("attention_output_33_transpose_y_0"), val = tensor(false)]; + tensor value_layer_17_cast_fp16 = transpose(perm = value_layer_17_perm_0, x = var_646_cast_fp16)[name = tensor("transpose_85")]; + tensor attention_output_33_cast_fp16 = matmul(transpose_x = attention_output_33_transpose_x_0, transpose_y = attention_output_33_transpose_y_0, x = softmax_8_cast_fp16, y = value_layer_17_cast_fp16)[name = tensor("attention_output_33_cast_fp16")]; + tensor attention_output_35_perm_0 = const()[name = tensor("attention_output_35_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_40_axis_0 = const()[name = tensor("concat_40_axis_0"), val = tensor(0)]; + tensor concat_40_interleave_0 = const()[name = tensor("concat_40_interleave_0"), val = tensor(false)]; + tensor gather_22_cast_uint16_to_int32 = cast(dtype = gather_22_cast_uint16_to_int32_dtype_0, x = gather_22_cast_uint16)[name = tensor("cast_9")]; + tensor concat_40 = concat(axis = concat_40_axis_0, interleave = concat_40_interleave_0, values = (gather_21_cast_uint16_to_int32, gather_22_cast_uint16_to_int32, var_3))[name = tensor("concat_40")]; + tensor attention_output_35_cast_fp16 = transpose(perm = attention_output_35_perm_0, x = attention_output_33_cast_fp16)[name = tensor("transpose_84")]; + tensor input_123_cast_fp16 = reshape(shape = concat_40, x = attention_output_35_cast_fp16)[name = tensor("input_123_cast_fp16")]; + tensor linear_52_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_123_cast_fp16)[name = tensor("linear_52_cast_fp16")]; + tensor input_127_cast_fp16 = add(x = hidden_states_17_cast_fp16, y = linear_52_cast_fp16)[name = tensor("input_127_cast_fp16")]; + tensor input_129_axes_0 = const()[name = tensor("input_129_axes_0"), val = tensor([-1])]; + tensor input_129_cast_fp16 = layer_norm(axes = input_129_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_127_cast_fp16)[name = tensor("input_129_cast_fp16")]; + tensor linear_53_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_129_cast_fp16)[name = tensor("linear_53_cast_fp16")]; + tensor input_133_mode_0 = const()[name = tensor("input_133_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_133_cast_fp16 = gelu(mode = input_133_mode_0, x = linear_53_cast_fp16)[name = tensor("input_133_cast_fp16")]; + tensor linear_54_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_133_cast_fp16)[name = tensor("linear_54_cast_fp16")]; + tensor input_135_cast_fp16 = add(x = linear_54_cast_fp16, y = input_129_cast_fp16)[name = tensor("input_135_cast_fp16")]; + tensor hidden_states_19_axes_0 = const()[name = tensor("hidden_states_19_axes_0"), val = tensor([-1])]; + tensor hidden_states_19_cast_fp16 = layer_norm(axes = hidden_states_19_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_135_cast_fp16)[name = tensor("hidden_states_19_cast_fp16")]; + tensor var_695_shape_cast_fp16 = shape(x = hidden_states_19_cast_fp16)[name = tensor("op_695_shape_cast_fp16")]; + tensor gather_23_axis_0 = const()[name = tensor("gather_23_axis_0"), val = tensor(0)]; + tensor gather_23_batch_dims_0 = const()[name = tensor("gather_23_batch_dims_0"), val = tensor(0)]; + tensor gather_23_validate_indices_0 = const()[name = tensor("gather_23_validate_indices_0"), val = tensor(false)]; + tensor var_695_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_695_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_23_to_uint16 = const()[name = tensor("select_23_to_uint16"), val = tensor(0)]; + tensor var_695_shape_cast_fp16_to_uint16 = cast(dtype = var_695_shape_cast_fp16_to_uint16_dtype_0, x = var_695_shape_cast_fp16)[name = tensor("cast_8")]; + tensor gather_23_cast_uint16 = gather(axis = gather_23_axis_0, batch_dims = gather_23_batch_dims_0, indices = select_23_to_uint16, validate_indices = gather_23_validate_indices_0, x = var_695_shape_cast_fp16_to_uint16)[name = tensor("gather_23_cast_uint16")]; + tensor gather_23_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_23_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_24_axis_0 = const()[name = tensor("gather_24_axis_0"), val = tensor(0)]; + tensor gather_24_batch_dims_0 = const()[name = tensor("gather_24_batch_dims_0"), val = tensor(0)]; + tensor gather_24_validate_indices_0 = const()[name = tensor("gather_24_validate_indices_0"), val = tensor(false)]; + tensor select_24_to_uint16 = const()[name = tensor("select_24_to_uint16"), val = tensor(1)]; + tensor gather_24_cast_uint16 = gather(axis = gather_24_axis_0, batch_dims = gather_24_batch_dims_0, indices = select_24_to_uint16, validate_indices = gather_24_validate_indices_0, x = var_695_shape_cast_fp16_to_uint16)[name = tensor("gather_24_cast_uint16")]; + tensor gather_24_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_24_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor linear_55_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = hidden_states_19_cast_fp16)[name = tensor("linear_55_cast_fp16")]; + tensor concat_41_axis_0 = const()[name = tensor("concat_41_axis_0"), val = tensor(0)]; + tensor concat_41_interleave_0 = const()[name = tensor("concat_41_interleave_0"), val = tensor(false)]; + tensor gather_23_cast_uint16_to_int32 = cast(dtype = gather_23_cast_uint16_to_int32_dtype_0, x = gather_23_cast_uint16)[name = tensor("cast_7")]; + tensor concat_41 = concat(axis = concat_41_axis_0, interleave = concat_41_interleave_0, values = (gather_23_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_41")]; + tensor var_701_cast_fp16 = reshape(shape = concat_41, x = linear_55_cast_fp16)[name = tensor("op_701_cast_fp16")]; + tensor linear_56_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = hidden_states_19_cast_fp16)[name = tensor("linear_56_cast_fp16")]; + tensor var_707_cast_fp16 = reshape(shape = concat_41, x = linear_56_cast_fp16)[name = tensor("op_707_cast_fp16")]; + tensor linear_57_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = hidden_states_19_cast_fp16)[name = tensor("linear_57_cast_fp16")]; + tensor var_713_cast_fp16 = reshape(shape = concat_41, x = linear_57_cast_fp16)[name = tensor("op_713_cast_fp16")]; + tensor value_layer_19_perm_0 = const()[name = tensor("value_layer_19_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_9_y_0_to_fp16 = const()[name = tensor("mul_9_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_9_cast_fp16 = mul(x = var_701_cast_fp16, y = mul_9_y_0_to_fp16)[name = tensor("mul_9_cast_fp16")]; + tensor matmul_9_transpose_y_0 = const()[name = tensor("matmul_9_transpose_y_0"), val = tensor(true)]; + tensor matmul_9_transpose_x_0 = const()[name = tensor("matmul_9_transpose_x_0"), val = tensor(false)]; + tensor transpose_66_perm_0 = const()[name = tensor("transpose_66_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_67_perm_0 = const()[name = tensor("transpose_67_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_67 = transpose(perm = transpose_67_perm_0, x = var_707_cast_fp16)[name = tensor("transpose_82")]; + tensor transpose_66 = transpose(perm = transpose_66_perm_0, x = mul_9_cast_fp16)[name = tensor("transpose_83")]; + tensor matmul_9_cast_fp16 = matmul(transpose_x = matmul_9_transpose_x_0, transpose_y = matmul_9_transpose_y_0, x = transpose_66, y = transpose_67)[name = tensor("matmul_9_cast_fp16")]; + tensor add_34_cast_fp16 = add(x = matmul_9_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_34_cast_fp16")]; + tensor softmax_9_axis_0 = const()[name = tensor("softmax_9_axis_0"), val = tensor(-1)]; + tensor softmax_9_cast_fp16 = softmax(axis = softmax_9_axis_0, x = add_34_cast_fp16)[name = tensor("softmax_9_cast_fp16")]; + tensor attention_output_37_transpose_x_0 = const()[name = tensor("attention_output_37_transpose_x_0"), val = tensor(false)]; + tensor attention_output_37_transpose_y_0 = const()[name = tensor("attention_output_37_transpose_y_0"), val = tensor(false)]; + tensor value_layer_19_cast_fp16 = transpose(perm = value_layer_19_perm_0, x = var_713_cast_fp16)[name = tensor("transpose_81")]; + tensor attention_output_37_cast_fp16 = matmul(transpose_x = attention_output_37_transpose_x_0, transpose_y = attention_output_37_transpose_y_0, x = softmax_9_cast_fp16, y = value_layer_19_cast_fp16)[name = tensor("attention_output_37_cast_fp16")]; + tensor attention_output_39_perm_0 = const()[name = tensor("attention_output_39_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_44_axis_0 = const()[name = tensor("concat_44_axis_0"), val = tensor(0)]; + tensor concat_44_interleave_0 = const()[name = tensor("concat_44_interleave_0"), val = tensor(false)]; + tensor gather_24_cast_uint16_to_int32 = cast(dtype = gather_24_cast_uint16_to_int32_dtype_0, x = gather_24_cast_uint16)[name = tensor("cast_6")]; + tensor concat_44 = concat(axis = concat_44_axis_0, interleave = concat_44_interleave_0, values = (gather_23_cast_uint16_to_int32, gather_24_cast_uint16_to_int32, var_3))[name = tensor("concat_44")]; + tensor attention_output_39_cast_fp16 = transpose(perm = attention_output_39_perm_0, x = attention_output_37_cast_fp16)[name = tensor("transpose_80")]; + tensor input_137_cast_fp16 = reshape(shape = concat_44, x = attention_output_39_cast_fp16)[name = tensor("input_137_cast_fp16")]; + tensor linear_58_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_137_cast_fp16)[name = tensor("linear_58_cast_fp16")]; + tensor input_141_cast_fp16 = add(x = hidden_states_19_cast_fp16, y = linear_58_cast_fp16)[name = tensor("input_141_cast_fp16")]; + tensor input_143_axes_0 = const()[name = tensor("input_143_axes_0"), val = tensor([-1])]; + tensor input_143_cast_fp16 = layer_norm(axes = input_143_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_141_cast_fp16)[name = tensor("input_143_cast_fp16")]; + tensor linear_59_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_143_cast_fp16)[name = tensor("linear_59_cast_fp16")]; + tensor input_147_mode_0 = const()[name = tensor("input_147_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_147_cast_fp16 = gelu(mode = input_147_mode_0, x = linear_59_cast_fp16)[name = tensor("input_147_cast_fp16")]; + tensor linear_60_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_147_cast_fp16)[name = tensor("linear_60_cast_fp16")]; + tensor input_149_cast_fp16 = add(x = linear_60_cast_fp16, y = input_143_cast_fp16)[name = tensor("input_149_cast_fp16")]; + tensor hidden_states_21_axes_0 = const()[name = tensor("hidden_states_21_axes_0"), val = tensor([-1])]; + tensor hidden_states_21_cast_fp16 = layer_norm(axes = hidden_states_21_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_149_cast_fp16)[name = tensor("hidden_states_21_cast_fp16")]; + tensor var_762_shape_cast_fp16 = shape(x = hidden_states_21_cast_fp16)[name = tensor("op_762_shape_cast_fp16")]; + tensor gather_25_axis_0 = const()[name = tensor("gather_25_axis_0"), val = tensor(0)]; + tensor gather_25_batch_dims_0 = const()[name = tensor("gather_25_batch_dims_0"), val = tensor(0)]; + tensor gather_25_validate_indices_0 = const()[name = tensor("gather_25_validate_indices_0"), val = tensor(false)]; + tensor var_762_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_762_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_25_to_uint16 = const()[name = tensor("select_25_to_uint16"), val = tensor(0)]; + tensor var_762_shape_cast_fp16_to_uint16 = cast(dtype = var_762_shape_cast_fp16_to_uint16_dtype_0, x = var_762_shape_cast_fp16)[name = tensor("cast_5")]; + tensor gather_25_cast_uint16 = gather(axis = gather_25_axis_0, batch_dims = gather_25_batch_dims_0, indices = select_25_to_uint16, validate_indices = gather_25_validate_indices_0, x = var_762_shape_cast_fp16_to_uint16)[name = tensor("gather_25_cast_uint16")]; + tensor gather_25_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_25_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_26_axis_0 = const()[name = tensor("gather_26_axis_0"), val = tensor(0)]; + tensor gather_26_batch_dims_0 = const()[name = tensor("gather_26_batch_dims_0"), val = tensor(0)]; + tensor gather_26_validate_indices_0 = const()[name = tensor("gather_26_validate_indices_0"), val = tensor(false)]; + tensor select_26_to_uint16 = const()[name = tensor("select_26_to_uint16"), val = tensor(1)]; + tensor gather_26_cast_uint16 = gather(axis = gather_26_axis_0, batch_dims = gather_26_batch_dims_0, indices = select_26_to_uint16, validate_indices = gather_26_validate_indices_0, x = var_762_shape_cast_fp16_to_uint16)[name = tensor("gather_26_cast_uint16")]; + tensor gather_26_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_26_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor linear_61_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = hidden_states_21_cast_fp16)[name = tensor("linear_61_cast_fp16")]; + tensor concat_45_axis_0 = const()[name = tensor("concat_45_axis_0"), val = tensor(0)]; + tensor concat_45_interleave_0 = const()[name = tensor("concat_45_interleave_0"), val = tensor(false)]; + tensor gather_25_cast_uint16_to_int32 = cast(dtype = gather_25_cast_uint16_to_int32_dtype_0, x = gather_25_cast_uint16)[name = tensor("cast_4")]; + tensor concat_45 = concat(axis = concat_45_axis_0, interleave = concat_45_interleave_0, values = (gather_25_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_45")]; + tensor var_768_cast_fp16 = reshape(shape = concat_45, x = linear_61_cast_fp16)[name = tensor("op_768_cast_fp16")]; + tensor linear_62_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = hidden_states_21_cast_fp16)[name = tensor("linear_62_cast_fp16")]; + tensor var_774_cast_fp16 = reshape(shape = concat_45, x = linear_62_cast_fp16)[name = tensor("op_774_cast_fp16")]; + tensor linear_63_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = hidden_states_21_cast_fp16)[name = tensor("linear_63_cast_fp16")]; + tensor var_780_cast_fp16 = reshape(shape = concat_45, x = linear_63_cast_fp16)[name = tensor("op_780_cast_fp16")]; + tensor value_layer_21_perm_0 = const()[name = tensor("value_layer_21_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_10_y_0_to_fp16 = const()[name = tensor("mul_10_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_10_cast_fp16 = mul(x = var_768_cast_fp16, y = mul_10_y_0_to_fp16)[name = tensor("mul_10_cast_fp16")]; + tensor matmul_10_transpose_y_0 = const()[name = tensor("matmul_10_transpose_y_0"), val = tensor(true)]; + tensor matmul_10_transpose_x_0 = const()[name = tensor("matmul_10_transpose_x_0"), val = tensor(false)]; + tensor transpose_68_perm_0 = const()[name = tensor("transpose_68_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_69_perm_0 = const()[name = tensor("transpose_69_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_69 = transpose(perm = transpose_69_perm_0, x = var_774_cast_fp16)[name = tensor("transpose_78")]; + tensor transpose_68 = transpose(perm = transpose_68_perm_0, x = mul_10_cast_fp16)[name = tensor("transpose_79")]; + tensor matmul_10_cast_fp16 = matmul(transpose_x = matmul_10_transpose_x_0, transpose_y = matmul_10_transpose_y_0, x = transpose_68, y = transpose_69)[name = tensor("matmul_10_cast_fp16")]; + tensor add_37_cast_fp16 = add(x = matmul_10_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_37_cast_fp16")]; + tensor softmax_10_axis_0 = const()[name = tensor("softmax_10_axis_0"), val = tensor(-1)]; + tensor softmax_10_cast_fp16 = softmax(axis = softmax_10_axis_0, x = add_37_cast_fp16)[name = tensor("softmax_10_cast_fp16")]; + tensor attention_output_41_transpose_x_0 = const()[name = tensor("attention_output_41_transpose_x_0"), val = tensor(false)]; + tensor attention_output_41_transpose_y_0 = const()[name = tensor("attention_output_41_transpose_y_0"), val = tensor(false)]; + tensor value_layer_21_cast_fp16 = transpose(perm = value_layer_21_perm_0, x = var_780_cast_fp16)[name = tensor("transpose_77")]; + tensor attention_output_41_cast_fp16 = matmul(transpose_x = attention_output_41_transpose_x_0, transpose_y = attention_output_41_transpose_y_0, x = softmax_10_cast_fp16, y = value_layer_21_cast_fp16)[name = tensor("attention_output_41_cast_fp16")]; + tensor attention_output_43_perm_0 = const()[name = tensor("attention_output_43_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_48_axis_0 = const()[name = tensor("concat_48_axis_0"), val = tensor(0)]; + tensor concat_48_interleave_0 = const()[name = tensor("concat_48_interleave_0"), val = tensor(false)]; + tensor gather_26_cast_uint16_to_int32 = cast(dtype = gather_26_cast_uint16_to_int32_dtype_0, x = gather_26_cast_uint16)[name = tensor("cast_3")]; + tensor concat_48 = concat(axis = concat_48_axis_0, interleave = concat_48_interleave_0, values = (gather_25_cast_uint16_to_int32, gather_26_cast_uint16_to_int32, var_3))[name = tensor("concat_48")]; + tensor attention_output_43_cast_fp16 = transpose(perm = attention_output_43_perm_0, x = attention_output_41_cast_fp16)[name = tensor("transpose_76")]; + tensor input_151_cast_fp16 = reshape(shape = concat_48, x = attention_output_43_cast_fp16)[name = tensor("input_151_cast_fp16")]; + tensor linear_64_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_151_cast_fp16)[name = tensor("linear_64_cast_fp16")]; + tensor input_155_cast_fp16 = add(x = hidden_states_21_cast_fp16, y = linear_64_cast_fp16)[name = tensor("input_155_cast_fp16")]; + tensor input_157_axes_0 = const()[name = tensor("input_157_axes_0"), val = tensor([-1])]; + tensor input_157_cast_fp16 = layer_norm(axes = input_157_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_155_cast_fp16)[name = tensor("input_157_cast_fp16")]; + tensor linear_65_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_157_cast_fp16)[name = tensor("linear_65_cast_fp16")]; + tensor input_161_mode_0 = const()[name = tensor("input_161_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_161_cast_fp16 = gelu(mode = input_161_mode_0, x = linear_65_cast_fp16)[name = tensor("input_161_cast_fp16")]; + tensor linear_66_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_161_cast_fp16)[name = tensor("linear_66_cast_fp16")]; + tensor input_163_cast_fp16 = add(x = linear_66_cast_fp16, y = input_157_cast_fp16)[name = tensor("input_163_cast_fp16")]; + tensor hidden_states_axes_0 = const()[name = tensor("hidden_states_axes_0"), val = tensor([-1])]; + tensor hidden_states_cast_fp16 = layer_norm(axes = hidden_states_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_163_cast_fp16)[name = tensor("hidden_states_cast_fp16")]; + tensor var_829_shape_cast_fp16 = shape(x = hidden_states_cast_fp16)[name = tensor("op_829_shape_cast_fp16")]; + tensor gather_27_axis_0 = const()[name = tensor("gather_27_axis_0"), val = tensor(0)]; + tensor gather_27_batch_dims_0 = const()[name = tensor("gather_27_batch_dims_0"), val = tensor(0)]; + tensor gather_27_validate_indices_0 = const()[name = tensor("gather_27_validate_indices_0"), val = tensor(false)]; + tensor var_829_shape_cast_fp16_to_uint16_dtype_0 = const()[name = tensor("op_829_shape_cast_fp16_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor select_27_to_uint16 = const()[name = tensor("select_27_to_uint16"), val = tensor(0)]; + tensor var_829_shape_cast_fp16_to_uint16 = cast(dtype = var_829_shape_cast_fp16_to_uint16_dtype_0, x = var_829_shape_cast_fp16)[name = tensor("cast_2")]; + tensor gather_27_cast_uint16 = gather(axis = gather_27_axis_0, batch_dims = gather_27_batch_dims_0, indices = select_27_to_uint16, validate_indices = gather_27_validate_indices_0, x = var_829_shape_cast_fp16_to_uint16)[name = tensor("gather_27_cast_uint16")]; + tensor gather_27_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_27_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor gather_28_axis_0 = const()[name = tensor("gather_28_axis_0"), val = tensor(0)]; + tensor gather_28_batch_dims_0 = const()[name = tensor("gather_28_batch_dims_0"), val = tensor(0)]; + tensor gather_28_validate_indices_0 = const()[name = tensor("gather_28_validate_indices_0"), val = tensor(false)]; + tensor select_28_to_uint16 = const()[name = tensor("select_28_to_uint16"), val = tensor(1)]; + tensor gather_28_cast_uint16 = gather(axis = gather_28_axis_0, batch_dims = gather_28_batch_dims_0, indices = select_28_to_uint16, validate_indices = gather_28_validate_indices_0, x = var_829_shape_cast_fp16_to_uint16)[name = tensor("gather_28_cast_uint16")]; + tensor gather_28_cast_uint16_to_int32_dtype_0 = const()[name = tensor("gather_28_cast_uint16_to_int32_dtype_0"), val = tensor("int32")]; + tensor linear_67_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_query_weight_to_fp16_palettized, x = hidden_states_cast_fp16)[name = tensor("linear_67_cast_fp16")]; + tensor concat_49_axis_0 = const()[name = tensor("concat_49_axis_0"), val = tensor(0)]; + tensor concat_49_interleave_0 = const()[name = tensor("concat_49_interleave_0"), val = tensor(false)]; + tensor gather_27_cast_uint16_to_int32 = cast(dtype = gather_27_cast_uint16_to_int32_dtype_0, x = gather_27_cast_uint16)[name = tensor("cast_1")]; + tensor concat_49 = concat(axis = concat_49_axis_0, interleave = concat_49_interleave_0, values = (gather_27_cast_uint16_to_int32, var_13, var_5, var_4))[name = tensor("concat_49")]; + tensor var_835_cast_fp16 = reshape(shape = concat_49, x = linear_67_cast_fp16)[name = tensor("op_835_cast_fp16")]; + tensor linear_68_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_key_weight_to_fp16_palettized, x = hidden_states_cast_fp16)[name = tensor("linear_68_cast_fp16")]; + tensor var_841_cast_fp16 = reshape(shape = concat_49, x = linear_68_cast_fp16)[name = tensor("op_841_cast_fp16")]; + tensor linear_69_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_value_weight_to_fp16_palettized, x = hidden_states_cast_fp16)[name = tensor("linear_69_cast_fp16")]; + tensor var_847_cast_fp16 = reshape(shape = concat_49, x = linear_69_cast_fp16)[name = tensor("op_847_cast_fp16")]; + tensor value_layer_perm_0 = const()[name = tensor("value_layer_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_11_y_0_to_fp16 = const()[name = tensor("mul_11_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_11_cast_fp16 = mul(x = var_835_cast_fp16, y = mul_11_y_0_to_fp16)[name = tensor("mul_11_cast_fp16")]; + tensor matmul_11_transpose_y_0 = const()[name = tensor("matmul_11_transpose_y_0"), val = tensor(true)]; + tensor matmul_11_transpose_x_0 = const()[name = tensor("matmul_11_transpose_x_0"), val = tensor(false)]; + tensor transpose_70_perm_0 = const()[name = tensor("transpose_70_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_71_perm_0 = const()[name = tensor("transpose_71_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_71 = transpose(perm = transpose_71_perm_0, x = var_841_cast_fp16)[name = tensor("transpose_74")]; + tensor transpose_70 = transpose(perm = transpose_70_perm_0, x = mul_11_cast_fp16)[name = tensor("transpose_75")]; + tensor matmul_11_cast_fp16 = matmul(transpose_x = matmul_11_transpose_x_0, transpose_y = matmul_11_transpose_y_0, x = transpose_70, y = transpose_71)[name = tensor("matmul_11_cast_fp16")]; + tensor add_40_cast_fp16 = add(x = matmul_11_cast_fp16, y = attention_mask_cast_fp16)[name = tensor("add_40_cast_fp16")]; + tensor softmax_11_axis_0 = const()[name = tensor("softmax_11_axis_0"), val = tensor(-1)]; + tensor softmax_11_cast_fp16 = softmax(axis = softmax_11_axis_0, x = add_40_cast_fp16)[name = tensor("softmax_11_cast_fp16")]; + tensor attention_output_45_transpose_x_0 = const()[name = tensor("attention_output_45_transpose_x_0"), val = tensor(false)]; + tensor attention_output_45_transpose_y_0 = const()[name = tensor("attention_output_45_transpose_y_0"), val = tensor(false)]; + tensor value_layer_cast_fp16 = transpose(perm = value_layer_perm_0, x = var_847_cast_fp16)[name = tensor("transpose_73")]; + tensor attention_output_45_cast_fp16 = matmul(transpose_x = attention_output_45_transpose_x_0, transpose_y = attention_output_45_transpose_y_0, x = softmax_11_cast_fp16, y = value_layer_cast_fp16)[name = tensor("attention_output_45_cast_fp16")]; + tensor attention_output_perm_0 = const()[name = tensor("attention_output_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor concat_52_axis_0 = const()[name = tensor("concat_52_axis_0"), val = tensor(0)]; + tensor concat_52_interleave_0 = const()[name = tensor("concat_52_interleave_0"), val = tensor(false)]; + tensor gather_28_cast_uint16_to_int32 = cast(dtype = gather_28_cast_uint16_to_int32_dtype_0, x = gather_28_cast_uint16)[name = tensor("cast_0")]; + tensor concat_52 = concat(axis = concat_52_axis_0, interleave = concat_52_interleave_0, values = (gather_27_cast_uint16_to_int32, gather_28_cast_uint16_to_int32, var_3))[name = tensor("concat_52")]; + tensor attention_output_cast_fp16 = transpose(perm = attention_output_perm_0, x = attention_output_45_cast_fp16)[name = tensor("transpose_72")]; + tensor input_165_cast_fp16 = reshape(shape = concat_52, x = attention_output_cast_fp16)[name = tensor("input_165_cast_fp16")]; + tensor linear_70_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_dense_weight_to_fp16_palettized, x = input_165_cast_fp16)[name = tensor("linear_70_cast_fp16")]; + tensor input_169_cast_fp16 = add(x = hidden_states_cast_fp16, y = linear_70_cast_fp16)[name = tensor("input_169_cast_fp16")]; + tensor input_171_axes_0 = const()[name = tensor("input_171_axes_0"), val = tensor([-1])]; + tensor input_171_cast_fp16 = layer_norm(axes = input_171_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_attention_LayerNorm_weight_to_fp16, x = input_169_cast_fp16)[name = tensor("input_171_cast_fp16")]; + tensor linear_71_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_weight_to_fp16_palettized, x = input_171_cast_fp16)[name = tensor("linear_71_cast_fp16")]; + tensor input_175_mode_0 = const()[name = tensor("input_175_mode_0"), val = tensor("TANH_APPROXIMATION")]; + tensor input_175_cast_fp16 = gelu(mode = input_175_mode_0, x = linear_71_cast_fp16)[name = tensor("input_175_cast_fp16")]; + tensor linear_72_cast_fp16 = linear(bias = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_bias_to_fp16, weight = bert_encoder_albert_layer_groups_0_albert_layers_0_ffn_output_weight_to_fp16_palettized, x = input_175_cast_fp16)[name = tensor("linear_72_cast_fp16")]; + tensor input_177_cast_fp16 = add(x = linear_72_cast_fp16, y = input_171_cast_fp16)[name = tensor("input_177_cast_fp16")]; + tensor sequence_output_axes_0 = const()[name = tensor("sequence_output_axes_0"), val = tensor([-1])]; + tensor bert_dur = layer_norm(axes = sequence_output_axes_0, beta = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_bias_to_fp16, epsilon = var_15_to_fp16, gamma = bert_encoder_albert_layer_groups_0_albert_layers_0_full_layer_layer_norm_weight_to_fp16, x = input_177_cast_fp16)[name = tensor("sequence_output_cast_fp16")]; + } -> (bert_dur); +} \ No newline at end of file diff --git a/ANE/styletts2_ane_plbert.mlmodelc/weights/weight.bin b/ANE/styletts2_ane_plbert.mlmodelc/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..28d6d0c4416d1c2bed8db487fbeb0d68a8bf84c1 --- /dev/null +++ b/ANE/styletts2_ane_plbert.mlmodelc/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59c7da8562b26c0842a3941543ebf48d94629105022320c9d8cd590761b6fcab +size 5718848 diff --git a/ANE/styletts2_ane_plbert.mlpackage/Data/com.apple.CoreML/model.mlmodel b/ANE/styletts2_ane_plbert.mlpackage/Data/com.apple.CoreML/model.mlmodel new file mode 100644 index 0000000000000000000000000000000000000000..00b738d0733ff54b3434e52265444787029f128a --- /dev/null +++ b/ANE/styletts2_ane_plbert.mlpackage/Data/com.apple.CoreML/model.mlmodel @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8a91ff072efb2aa10a26f76e280108f110a877429d74ec76769aa560c3ebbc7e +size 148365 diff --git a/ANE/styletts2_ane_plbert.mlpackage/Data/com.apple.CoreML/weights/weight.bin b/ANE/styletts2_ane_plbert.mlpackage/Data/com.apple.CoreML/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..28d6d0c4416d1c2bed8db487fbeb0d68a8bf84c1 --- /dev/null +++ b/ANE/styletts2_ane_plbert.mlpackage/Data/com.apple.CoreML/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59c7da8562b26c0842a3941543ebf48d94629105022320c9d8cd590761b6fcab +size 5718848 diff --git a/ANE/styletts2_ane_plbert.mlpackage/Manifest.json b/ANE/styletts2_ane_plbert.mlpackage/Manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..4a68041feee0aaa9edc5f82563c6dee868f9c793 --- /dev/null +++ b/ANE/styletts2_ane_plbert.mlpackage/Manifest.json @@ -0,0 +1,18 @@ +{ + "fileFormatVersion": "1.0.0", + "itemInfoEntries": { + "3B613FCB-B965-4C2D-94A7-4D5C80EDE23F": { + "author": "com.apple.CoreML", + "description": "CoreML Model Weights", + "name": "weights", + "path": "com.apple.CoreML/weights" + }, + "440AF525-7EC6-4E07-8019-7BF58173E928": { + "author": "com.apple.CoreML", + "description": "CoreML Model Specification", + "name": "model.mlmodel", + "path": "com.apple.CoreML/model.mlmodel" + } + }, + "rootModelIdentifier": "440AF525-7EC6-4E07-8019-7BF58173E928" +} diff --git a/ANE/styletts2_ane_postbert.mlmodelc/analytics/coremldata.bin b/ANE/styletts2_ane_postbert.mlmodelc/analytics/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..9bf02468a5ef4e3b4e0a28815003a271db17fb1f --- /dev/null +++ b/ANE/styletts2_ane_postbert.mlmodelc/analytics/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:140544065220d26a02c8c8c6592a298c95dd14b80c0b22bc4005189cc86698d2 +size 243 diff --git a/ANE/styletts2_ane_postbert.mlmodelc/coremldata.bin b/ANE/styletts2_ane_postbert.mlmodelc/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..4b653afd725704f638b48615c89328da70c551cf --- /dev/null +++ b/ANE/styletts2_ane_postbert.mlmodelc/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dde3c94cb2d81022ada31a29ca27224bdc7d3c75672565969188fe7be761c172 +size 542 diff --git a/ANE/styletts2_ane_postbert.mlmodelc/metadata.json b/ANE/styletts2_ane_postbert.mlmodelc/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..4c940d4255ce185010ec0e123e91c52f82f9a2ce --- /dev/null +++ b/ANE/styletts2_ane_postbert.mlmodelc/metadata.json @@ -0,0 +1,133 @@ +[ + { + "metadataOutputVersion" : "3.0", + "shortDescription" : "StyleTTS2-ANE Stage 2: TextEncoder + DurationEncoder + duration head — fp16, ANE.", + "outputSchema" : [ + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float16", + "formattedType" : "MultiArray (Float16)", + "shortDescription" : "", + "shape" : "[]", + "name" : "t_en", + "type" : "MultiArray" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float16", + "formattedType" : "MultiArray (Float16)", + "shortDescription" : "", + "shape" : "[]", + "name" : "d", + "type" : "MultiArray" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float16", + "formattedType" : "MultiArray (Float16)", + "shortDescription" : "", + "shape" : "[]", + "name" : "pred_dur_log", + "type" : "MultiArray" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float16", + "formattedType" : "MultiArray (Float16)", + "shortDescription" : "", + "shape" : "[]", + "name" : "fixed_embedding", + "type" : "MultiArray" + } + ], + "storagePrecision" : "Float16", + "modelParameters" : [ + + ], + "specificationVersion" : 8, + "mlProgramOperationTypeHistogram" : { + "Split" : 3, + "Ios17.mul" : 3, + "Ios17.linear" : 5, + "Ios17.transpose" : 27, + "Ios17.sub" : 2, + "Ios16.fillLike" : 1, + "Ios17.conv" : 3, + "Ios16.cumsum" : 1, + "Ios17.leakyRelu" : 3, + "Ios17.gather" : 2, + "Ios17.add" : 7, + "Ios17.sliceByIndex" : 3, + "Ios17.lstm" : 5, + "Ios17.concat" : 4, + "Ios17.expandDims" : 2, + "Ios17.squeeze" : 1, + "Ios17.cast" : 3, + "Ios17.layerNorm" : 6, + "Ios17.reshape" : 3 + }, + "computePrecision" : "Mixed (Float16, Float32, Int16, Int32, UInt16)", + "isUpdatable" : "0", + "stateSchema" : [ + + ], + "availability" : { + "macOS" : "14.0", + "tvOS" : "17.0", + "visionOS" : "1.0", + "watchOS" : "10.0", + "iOS" : "17.0", + "macCatalyst" : "17.0" + }, + "modelType" : { + "name" : "MLModelType_mlProgram" + }, + "inputSchema" : [ + { + "dataType" : "Float16", + "hasShapeFlexibility" : "1", + "isOptional" : "0", + "shapeFlexibility" : "1 × 2...512 × 768", + "shapeRange" : "[[1, 1], [2, 512], [768, 768]]", + "formattedType" : "MultiArray (Float16 1 × 64 × 768)", + "type" : "MultiArray", + "shape" : "[1, 64, 768]", + "name" : "bert_dur", + "shortDescription" : "" + }, + { + "dataType" : "Int32", + "hasShapeFlexibility" : "1", + "isOptional" : "0", + "shapeFlexibility" : "1 × 2...512", + "shapeRange" : "[[1, 1], [2, 512]]", + "formattedType" : "MultiArray (Int32 1 × 64)", + "type" : "MultiArray", + "shape" : "[1, 64]", + "name" : "tokens", + "shortDescription" : "" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float16", + "formattedType" : "MultiArray (Float16 1 × 256)", + "shortDescription" : "", + "shape" : "[1, 256]", + "name" : "style", + "type" : "MultiArray" + } + ], + "userDefinedMetadata" : { + "com.github.apple.coremltools.version" : "8.3.0", + "com.github.apple.coremltools.source" : "torch==2.11.0", + "com.github.apple.coremltools.source_dialect" : "TorchScript" + }, + "generatedClassName" : "styletts2_ane_postbert", + "method" : "predict" + } +] \ No newline at end of file diff --git a/ANE/styletts2_ane_postbert.mlmodelc/model.mil b/ANE/styletts2_ane_postbert.mlmodelc/model.mil new file mode 100644 index 0000000000000000000000000000000000000000..c7aa34e9d832db32bfbf7355a89ae1fcf7bab892 --- /dev/null +++ b/ANE/styletts2_ane_postbert.mlmodelc/model.mil @@ -0,0 +1,265 @@ +program(1.0) +[buildInfo = dict, tensor>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}, {"coremltools-component-torch", "2.11.0"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "8.3.0"}})] +{ + func main(tensor bert_dur, tensor style, tensor tokens) [FlexibleShapeInformation = tuple, dict, tensor>>, tuple, dict, list, ?>>>>((("DefaultShapes", {{"bert_dur", [1, 64, 768]}, {"tokens", [1, 64]}}), ("RangeDims", {{"bert_dur", [[1, 1], [2, 512], [768, 768]]}, {"tokens", [[1, 1], [2, 512]]}})))] { + tensor style_begin_0 = const()[name = tensor("style_begin_0"), val = tensor([0, 0])]; + tensor style_end_0 = const()[name = tensor("style_end_0"), val = tensor([1, 128])]; + tensor style_end_mask_0 = const()[name = tensor("style_end_mask_0"), val = tensor([true, false])]; + tensor style_cast_fp16 = slice_by_index(begin = style_begin_0, end = style_end_0, end_mask = style_end_mask_0, x = style)[name = tensor("style_cast_fp16")]; + tensor var_30 = const()[name = tensor("op_30"), val = tensor(0x1.99999ap-3)]; + tensor x_1_axis_0 = const()[name = tensor("x_1_axis_0"), val = tensor(0)]; + tensor x_1_batch_dims_0 = const()[name = tensor("x_1_batch_dims_0"), val = tensor(0)]; + tensor x_1_validate_indices_0 = const()[name = tensor("x_1_validate_indices_0"), val = tensor(false)]; + tensor text_encoder_embedding_weight_to_fp16 = const()[name = tensor("text_encoder_embedding_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(64)))]; + tensor tokens_to_int16_dtype_0 = const()[name = tensor("tokens_to_int16_dtype_0"), val = tensor("int16")]; + tensor tokens_to_int16 = cast(dtype = tokens_to_int16_dtype_0, x = tokens)[name = tensor("cast_14")]; + tensor x_1_cast_fp16_cast_uint16 = gather(axis = x_1_axis_0, batch_dims = x_1_batch_dims_0, indices = tokens_to_int16, validate_indices = x_1_validate_indices_0, x = text_encoder_embedding_weight_to_fp16)[name = tensor("x_1_cast_fp16_cast_uint16")]; + tensor input_1_perm_0 = const()[name = tensor("input_1_perm_0"), val = tensor([0, 2, 1])]; + tensor x_3_pad_type_0 = const()[name = tensor("x_3_pad_type_0"), val = tensor("custom")]; + tensor x_3_pad_0 = const()[name = tensor("x_3_pad_0"), val = tensor([2, 2])]; + tensor x_3_strides_0 = const()[name = tensor("x_3_strides_0"), val = tensor([1])]; + tensor x_3_dilations_0 = const()[name = tensor("x_3_dilations_0"), val = tensor([1])]; + tensor x_3_groups_0 = const()[name = tensor("x_3_groups_0"), val = tensor(1)]; + tensor weight_3_to_fp16 = const()[name = tensor("weight_3_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(182400)))]; + tensor text_encoder_cnn_0_0_bias_to_fp16 = const()[name = tensor("text_encoder_cnn_0_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2803904)))]; + tensor input_1_cast_fp16 = transpose(perm = input_1_perm_0, x = x_1_cast_fp16_cast_uint16)[name = tensor("transpose_36")]; + tensor x_3_cast_fp16 = conv(bias = text_encoder_cnn_0_0_bias_to_fp16, dilations = x_3_dilations_0, groups = x_3_groups_0, pad = x_3_pad_0, pad_type = x_3_pad_type_0, strides = x_3_strides_0, weight = weight_3_to_fp16, x = input_1_cast_fp16)[name = tensor("x_3_cast_fp16")]; + tensor input_3_perm_0 = const()[name = tensor("input_3_perm_0"), val = tensor([0, -1, 1])]; + tensor x_5_axes_0 = const()[name = tensor("x_5_axes_0"), val = tensor([-1])]; + tensor text_encoder_cnn_0_1_gamma_to_fp16 = const()[name = tensor("text_encoder_cnn_0_1_gamma_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2804992)))]; + tensor text_encoder_cnn_0_1_beta_to_fp16 = const()[name = tensor("text_encoder_cnn_0_1_beta_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2806080)))]; + tensor var_28_to_fp16 = const()[name = tensor("op_28_to_fp16"), val = tensor(0x1.5p-17)]; + tensor input_3_cast_fp16 = transpose(perm = input_3_perm_0, x = x_3_cast_fp16)[name = tensor("transpose_35")]; + tensor x_5_cast_fp16 = layer_norm(axes = x_5_axes_0, beta = text_encoder_cnn_0_1_beta_to_fp16, epsilon = var_28_to_fp16, gamma = text_encoder_cnn_0_1_gamma_to_fp16, x = input_3_cast_fp16)[name = tensor("x_5_cast_fp16")]; + tensor input_5_perm_0 = const()[name = tensor("input_5_perm_0"), val = tensor([0, -1, 1])]; + tensor input_5_cast_fp16 = transpose(perm = input_5_perm_0, x = x_5_cast_fp16)[name = tensor("transpose_34")]; + tensor input_7_cast_fp16 = leaky_relu(alpha = var_30, x = input_5_cast_fp16)[name = tensor("input_7_cast_fp16")]; + tensor x_7_pad_type_0 = const()[name = tensor("x_7_pad_type_0"), val = tensor("custom")]; + tensor x_7_pad_0 = const()[name = tensor("x_7_pad_0"), val = tensor([2, 2])]; + tensor x_7_strides_0 = const()[name = tensor("x_7_strides_0"), val = tensor([1])]; + tensor x_7_dilations_0 = const()[name = tensor("x_7_dilations_0"), val = tensor([1])]; + tensor x_7_groups_0 = const()[name = tensor("x_7_groups_0"), val = tensor(1)]; + tensor weight_7_to_fp16 = const()[name = tensor("weight_7_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2807168)))]; + tensor text_encoder_cnn_1_0_bias_to_fp16 = const()[name = tensor("text_encoder_cnn_1_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5428672)))]; + tensor x_7_cast_fp16 = conv(bias = text_encoder_cnn_1_0_bias_to_fp16, dilations = x_7_dilations_0, groups = x_7_groups_0, pad = x_7_pad_0, pad_type = x_7_pad_type_0, strides = x_7_strides_0, weight = weight_7_to_fp16, x = input_7_cast_fp16)[name = tensor("x_7_cast_fp16")]; + tensor input_11_perm_0 = const()[name = tensor("input_11_perm_0"), val = tensor([0, -1, 1])]; + tensor x_9_axes_0 = const()[name = tensor("x_9_axes_0"), val = tensor([-1])]; + tensor text_encoder_cnn_1_1_gamma_to_fp16 = const()[name = tensor("text_encoder_cnn_1_1_gamma_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5429760)))]; + tensor text_encoder_cnn_1_1_beta_to_fp16 = const()[name = tensor("text_encoder_cnn_1_1_beta_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5430848)))]; + tensor input_11_cast_fp16 = transpose(perm = input_11_perm_0, x = x_7_cast_fp16)[name = tensor("transpose_33")]; + tensor x_9_cast_fp16 = layer_norm(axes = x_9_axes_0, beta = text_encoder_cnn_1_1_beta_to_fp16, epsilon = var_28_to_fp16, gamma = text_encoder_cnn_1_1_gamma_to_fp16, x = input_11_cast_fp16)[name = tensor("x_9_cast_fp16")]; + tensor input_13_perm_0 = const()[name = tensor("input_13_perm_0"), val = tensor([0, -1, 1])]; + tensor input_13_cast_fp16 = transpose(perm = input_13_perm_0, x = x_9_cast_fp16)[name = tensor("transpose_32")]; + tensor input_15_cast_fp16 = leaky_relu(alpha = var_30, x = input_13_cast_fp16)[name = tensor("input_15_cast_fp16")]; + tensor x_11_pad_type_0 = const()[name = tensor("x_11_pad_type_0"), val = tensor("custom")]; + tensor x_11_pad_0 = const()[name = tensor("x_11_pad_0"), val = tensor([2, 2])]; + tensor x_11_strides_0 = const()[name = tensor("x_11_strides_0"), val = tensor([1])]; + tensor x_11_dilations_0 = const()[name = tensor("x_11_dilations_0"), val = tensor([1])]; + tensor x_11_groups_0 = const()[name = tensor("x_11_groups_0"), val = tensor(1)]; + tensor weight_11_to_fp16 = const()[name = tensor("weight_11_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5431936)))]; + tensor text_encoder_cnn_2_0_bias_to_fp16 = const()[name = tensor("text_encoder_cnn_2_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8053440)))]; + tensor x_11_cast_fp16 = conv(bias = text_encoder_cnn_2_0_bias_to_fp16, dilations = x_11_dilations_0, groups = x_11_groups_0, pad = x_11_pad_0, pad_type = x_11_pad_type_0, strides = x_11_strides_0, weight = weight_11_to_fp16, x = input_15_cast_fp16)[name = tensor("x_11_cast_fp16")]; + tensor input_19_perm_0 = const()[name = tensor("input_19_perm_0"), val = tensor([0, -1, 1])]; + tensor x_13_axes_0 = const()[name = tensor("x_13_axes_0"), val = tensor([-1])]; + tensor text_encoder_cnn_2_1_gamma_to_fp16 = const()[name = tensor("text_encoder_cnn_2_1_gamma_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8054528)))]; + tensor text_encoder_cnn_2_1_beta_to_fp16 = const()[name = tensor("text_encoder_cnn_2_1_beta_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8055616)))]; + tensor input_19_cast_fp16 = transpose(perm = input_19_perm_0, x = x_11_cast_fp16)[name = tensor("transpose_31")]; + tensor x_13_cast_fp16 = layer_norm(axes = x_13_axes_0, beta = text_encoder_cnn_2_1_beta_to_fp16, epsilon = var_28_to_fp16, gamma = text_encoder_cnn_2_1_gamma_to_fp16, x = input_19_cast_fp16)[name = tensor("x_13_cast_fp16")]; + tensor input_23_cast_fp16 = leaky_relu(alpha = var_30, x = x_13_cast_fp16)[name = tensor("input_23_cast_fp16")]; + tensor input_25_batch_first_transpose_perm_0 = const()[name = tensor("input_25_batch_first_transpose_perm_0"), val = tensor([1, 0, 2])]; + tensor x_17_batch_first_direction_0 = const()[name = tensor("x_17_batch_first_direction_0"), val = tensor("bidirectional")]; + tensor x_17_batch_first_output_sequence_0 = const()[name = tensor("x_17_batch_first_output_sequence_0"), val = tensor(true)]; + tensor x_17_batch_first_recurrent_activation_0 = const()[name = tensor("x_17_batch_first_recurrent_activation_0"), val = tensor("sigmoid")]; + tensor x_17_batch_first_cell_activation_0 = const()[name = tensor("x_17_batch_first_cell_activation_0"), val = tensor("tanh")]; + tensor x_17_batch_first_activation_0 = const()[name = tensor("x_17_batch_first_activation_0"), val = tensor("tanh")]; + tensor x_17_batch_first_lstm_h0_reshaped_to_fp16 = const()[name = tensor("x_17_batch_first_lstm_h0_reshaped_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8056704)))]; + tensor concat_4_to_fp16 = const()[name = tensor("concat_4_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8057792)))]; + tensor concat_5_to_fp16 = const()[name = tensor("concat_5_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9106432)))]; + tensor add_1_to_fp16 = const()[name = tensor("add_1_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9630784)))]; + tensor concat_6_to_fp16 = const()[name = tensor("concat_6_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9632896)))]; + tensor concat_7_to_fp16 = const()[name = tensor("concat_7_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(10681536)))]; + tensor add_2_to_fp16 = const()[name = tensor("add_2_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11205888)))]; + tensor input_25_batch_first_transpose_cast_fp16 = transpose(perm = input_25_batch_first_transpose_perm_0, x = input_23_cast_fp16)[name = tensor("transpose_30")]; + tensor x_17_batch_first_cast_fp16_0, tensor x_17_batch_first_cast_fp16_1, tensor x_17_batch_first_cast_fp16_2 = lstm(activation = x_17_batch_first_activation_0, bias = add_1_to_fp16, bias_back = add_2_to_fp16, cell_activation = x_17_batch_first_cell_activation_0, direction = x_17_batch_first_direction_0, initial_c = x_17_batch_first_lstm_h0_reshaped_to_fp16, initial_h = x_17_batch_first_lstm_h0_reshaped_to_fp16, output_sequence = x_17_batch_first_output_sequence_0, recurrent_activation = x_17_batch_first_recurrent_activation_0, weight_hh = concat_5_to_fp16, weight_hh_back = concat_7_to_fp16, weight_ih = concat_4_to_fp16, weight_ih_back = concat_6_to_fp16, x = input_25_batch_first_transpose_cast_fp16)[name = tensor("x_17_batch_first_cast_fp16")]; + tensor transpose_0_perm_0 = const()[name = tensor("transpose_0_perm_0"), val = tensor([1, 2, 0])]; + tensor bert_encoder_weight_to_fp16 = const()[name = tensor("bert_encoder_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11208000)))]; + tensor bert_encoder_bias_to_fp16 = const()[name = tensor("bert_encoder_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11994496)))]; + tensor linear_0_cast_fp16 = linear(bias = bert_encoder_bias_to_fp16, weight = bert_encoder_weight_to_fp16, x = bert_dur)[name = tensor("linear_0_cast_fp16")]; + tensor x_19_perm_0 = const()[name = tensor("x_19_perm_0"), val = tensor([0, -1, -2])]; + tensor var_147 = const()[name = tensor("op_147"), val = tensor(1)]; + tensor var_163_begin_0 = const()[name = tensor("op_163_begin_0"), val = tensor([0, 0, 0])]; + tensor var_163_end_0 = const()[name = tensor("op_163_end_0"), val = tensor([1, 1, 0])]; + tensor var_163_end_mask_0 = const()[name = tensor("op_163_end_mask_0"), val = tensor([true, false, true])]; + tensor x_19_cast_fp16 = transpose(perm = x_19_perm_0, x = linear_0_cast_fp16)[name = tensor("transpose_28")]; + tensor var_163_cast_fp16 = slice_by_index(begin = var_163_begin_0, end = var_163_end_0, end_mask = var_163_end_mask_0, x = x_19_cast_fp16)[name = tensor("op_163_cast_fp16")]; + tensor sub_0_cast_fp16 = sub(x = var_163_cast_fp16, y = var_163_cast_fp16)[name = tensor("sub_0_cast_fp16")]; + tensor var_166_axes_0 = const()[name = tensor("op_166_axes_0"), val = tensor([-1])]; + tensor var_166_cast_fp16 = expand_dims(axes = var_166_axes_0, x = style_cast_fp16)[name = tensor("op_166_cast_fp16")]; + tensor s_chan_cast_fp16 = add(x = var_166_cast_fp16, y = sub_0_cast_fp16)[name = tensor("s_chan_cast_fp16")]; + tensor x_21_interleave_0 = const()[name = tensor("x_21_interleave_0"), val = tensor(false)]; + tensor x_21_cast_fp16 = concat(axis = var_147, interleave = x_21_interleave_0, values = (x_19_cast_fp16, s_chan_cast_fp16))[name = tensor("x_21_cast_fp16")]; + tensor transpose_1_perm_0 = const()[name = tensor("transpose_1_perm_0"), val = tensor([-1, 0, -2])]; + tensor input_29_batch_first_direction_0 = const()[name = tensor("input_29_batch_first_direction_0"), val = tensor("bidirectional")]; + tensor input_29_batch_first_output_sequence_0 = const()[name = tensor("input_29_batch_first_output_sequence_0"), val = tensor(true)]; + tensor input_29_batch_first_recurrent_activation_0 = const()[name = tensor("input_29_batch_first_recurrent_activation_0"), val = tensor("sigmoid")]; + tensor input_29_batch_first_cell_activation_0 = const()[name = tensor("input_29_batch_first_cell_activation_0"), val = tensor("tanh")]; + tensor input_29_batch_first_activation_0 = const()[name = tensor("input_29_batch_first_activation_0"), val = tensor("tanh")]; + tensor concat_14_to_fp16 = const()[name = tensor("concat_14_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11995584)))]; + tensor concat_15_to_fp16 = const()[name = tensor("concat_15_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(13306368)))]; + tensor add_4_to_fp16 = const()[name = tensor("add_4_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(13830720)))]; + tensor concat_16_to_fp16 = const()[name = tensor("concat_16_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(13832832)))]; + tensor concat_17_to_fp16 = const()[name = tensor("concat_17_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15143616)))]; + tensor add_5_to_fp16 = const()[name = tensor("add_5_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15667968)))]; + tensor transpose_1_cast_fp16 = transpose(perm = transpose_1_perm_0, x = x_21_cast_fp16)[name = tensor("transpose_27")]; + tensor input_29_batch_first_cast_fp16_0, tensor input_29_batch_first_cast_fp16_1, tensor input_29_batch_first_cast_fp16_2 = lstm(activation = input_29_batch_first_activation_0, bias = add_4_to_fp16, bias_back = add_5_to_fp16, cell_activation = input_29_batch_first_cell_activation_0, direction = input_29_batch_first_direction_0, initial_c = x_17_batch_first_lstm_h0_reshaped_to_fp16, initial_h = x_17_batch_first_lstm_h0_reshaped_to_fp16, output_sequence = input_29_batch_first_output_sequence_0, recurrent_activation = input_29_batch_first_recurrent_activation_0, weight_hh = concat_15_to_fp16, weight_hh_back = concat_17_to_fp16, weight_ih = concat_14_to_fp16, weight_ih_back = concat_16_to_fp16, x = transpose_1_cast_fp16)[name = tensor("input_29_batch_first_cast_fp16")]; + tensor transpose_7_perm_0 = const()[name = tensor("transpose_7_perm_0"), val = tensor([1, 0, 2])]; + tensor duration_encoder_norms_0_fc_weight_to_fp16 = const()[name = tensor("duration_encoder_norms_0_fc_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15670080)))]; + tensor duration_encoder_norms_0_fc_bias_to_fp16 = const()[name = tensor("duration_encoder_norms_0_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15932288)))]; + tensor linear_1_cast_fp16 = linear(bias = duration_encoder_norms_0_fc_bias_to_fp16, weight = duration_encoder_norms_0_fc_weight_to_fp16, x = style_cast_fp16)[name = tensor("linear_1_cast_fp16")]; + tensor var_200 = const()[name = tensor("op_200"), val = tensor([1, 1024, 1])]; + tensor h_3_cast_fp16 = reshape(shape = var_200, x = linear_1_cast_fp16)[name = tensor("h_3_cast_fp16")]; + tensor var_202_split_sizes_0 = const()[name = tensor("op_202_split_sizes_0"), val = tensor([512, 512])]; + tensor var_202_axis_0 = const()[name = tensor("op_202_axis_0"), val = tensor(1)]; + tensor var_202_cast_fp16_0, tensor var_202_cast_fp16_1 = split(axis = var_202_axis_0, split_sizes = var_202_split_sizes_0, x = h_3_cast_fp16)[name = tensor("op_202_cast_fp16")]; + tensor gamma_3_perm_0 = const()[name = tensor("gamma_3_perm_0"), val = tensor([0, -1, 1])]; + tensor beta_3_perm_0 = const()[name = tensor("beta_3_perm_0"), val = tensor([0, -1, 1])]; + tensor x_31_axes_0 = const()[name = tensor("x_31_axes_0"), val = tensor([-1])]; + tensor var_136_to_fp16 = const()[name = tensor("op_136_to_fp16"), val = tensor(0x1.5p-17)]; + tensor transpose_7_cast_fp16 = transpose(perm = transpose_7_perm_0, x = input_29_batch_first_cast_fp16_0)[name = tensor("transpose_26")]; + tensor x_31_cast_fp16 = layer_norm(axes = x_31_axes_0, epsilon = var_136_to_fp16, x = transpose_7_cast_fp16)[name = tensor("x_31_cast_fp16")]; + tensor var_208_promoted_to_fp16 = const()[name = tensor("op_208_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor gamma_3_cast_fp16 = transpose(perm = gamma_3_perm_0, x = var_202_cast_fp16_0)[name = tensor("transpose_25")]; + tensor var_209_cast_fp16 = add(x = gamma_3_cast_fp16, y = var_208_promoted_to_fp16)[name = tensor("op_209_cast_fp16")]; + tensor var_210_cast_fp16 = mul(x = var_209_cast_fp16, y = x_31_cast_fp16)[name = tensor("op_210_cast_fp16")]; + tensor beta_3_cast_fp16 = transpose(perm = beta_3_perm_0, x = var_202_cast_fp16_1)[name = tensor("transpose_24")]; + tensor x_33_cast_fp16 = add(x = var_210_cast_fp16, y = beta_3_cast_fp16)[name = tensor("x_33_cast_fp16")]; + tensor x_35_perm_0 = const()[name = tensor("x_35_perm_0"), val = tensor([0, -1, -2])]; + tensor x_37_interleave_0 = const()[name = tensor("x_37_interleave_0"), val = tensor(false)]; + tensor x_35_cast_fp16 = transpose(perm = x_35_perm_0, x = x_33_cast_fp16)[name = tensor("transpose_23")]; + tensor x_37_cast_fp16 = concat(axis = var_147, interleave = x_37_interleave_0, values = (x_35_cast_fp16, s_chan_cast_fp16))[name = tensor("x_37_cast_fp16")]; + tensor transpose_3_perm_0 = const()[name = tensor("transpose_3_perm_0"), val = tensor([-1, 0, -2])]; + tensor input_35_batch_first_direction_0 = const()[name = tensor("input_35_batch_first_direction_0"), val = tensor("bidirectional")]; + tensor input_35_batch_first_output_sequence_0 = const()[name = tensor("input_35_batch_first_output_sequence_0"), val = tensor(true)]; + tensor input_35_batch_first_recurrent_activation_0 = const()[name = tensor("input_35_batch_first_recurrent_activation_0"), val = tensor("sigmoid")]; + tensor input_35_batch_first_cell_activation_0 = const()[name = tensor("input_35_batch_first_cell_activation_0"), val = tensor("tanh")]; + tensor input_35_batch_first_activation_0 = const()[name = tensor("input_35_batch_first_activation_0"), val = tensor("tanh")]; + tensor concat_24_to_fp16 = const()[name = tensor("concat_24_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15934400)))]; + tensor concat_25_to_fp16 = const()[name = tensor("concat_25_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17245184)))]; + tensor add_7_to_fp16 = const()[name = tensor("add_7_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17769536)))]; + tensor concat_26_to_fp16 = const()[name = tensor("concat_26_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17771648)))]; + tensor concat_27_to_fp16 = const()[name = tensor("concat_27_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19082432)))]; + tensor add_8_to_fp16 = const()[name = tensor("add_8_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19606784)))]; + tensor transpose_3_cast_fp16 = transpose(perm = transpose_3_perm_0, x = x_37_cast_fp16)[name = tensor("transpose_22")]; + tensor input_35_batch_first_cast_fp16_0, tensor input_35_batch_first_cast_fp16_1, tensor input_35_batch_first_cast_fp16_2 = lstm(activation = input_35_batch_first_activation_0, bias = add_7_to_fp16, bias_back = add_8_to_fp16, cell_activation = input_35_batch_first_cell_activation_0, direction = input_35_batch_first_direction_0, initial_c = x_17_batch_first_lstm_h0_reshaped_to_fp16, initial_h = x_17_batch_first_lstm_h0_reshaped_to_fp16, output_sequence = input_35_batch_first_output_sequence_0, recurrent_activation = input_35_batch_first_recurrent_activation_0, weight_hh = concat_25_to_fp16, weight_hh_back = concat_27_to_fp16, weight_ih = concat_24_to_fp16, weight_ih_back = concat_26_to_fp16, x = transpose_3_cast_fp16)[name = tensor("input_35_batch_first_cast_fp16")]; + tensor transpose_8_perm_0 = const()[name = tensor("transpose_8_perm_0"), val = tensor([1, 0, 2])]; + tensor duration_encoder_norms_1_fc_weight_to_fp16 = const()[name = tensor("duration_encoder_norms_1_fc_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19608896)))]; + tensor duration_encoder_norms_1_fc_bias_to_fp16 = const()[name = tensor("duration_encoder_norms_1_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19871104)))]; + tensor linear_2_cast_fp16 = linear(bias = duration_encoder_norms_1_fc_bias_to_fp16, weight = duration_encoder_norms_1_fc_weight_to_fp16, x = style_cast_fp16)[name = tensor("linear_2_cast_fp16")]; + tensor var_247 = const()[name = tensor("op_247"), val = tensor([1, 1024, 1])]; + tensor h_7_cast_fp16 = reshape(shape = var_247, x = linear_2_cast_fp16)[name = tensor("h_7_cast_fp16")]; + tensor var_249_split_sizes_0 = const()[name = tensor("op_249_split_sizes_0"), val = tensor([512, 512])]; + tensor var_249_axis_0 = const()[name = tensor("op_249_axis_0"), val = tensor(1)]; + tensor var_249_cast_fp16_0, tensor var_249_cast_fp16_1 = split(axis = var_249_axis_0, split_sizes = var_249_split_sizes_0, x = h_7_cast_fp16)[name = tensor("op_249_cast_fp16")]; + tensor gamma_7_perm_0 = const()[name = tensor("gamma_7_perm_0"), val = tensor([0, -1, 1])]; + tensor beta_7_perm_0 = const()[name = tensor("beta_7_perm_0"), val = tensor([0, -1, 1])]; + tensor x_47_axes_0 = const()[name = tensor("x_47_axes_0"), val = tensor([-1])]; + tensor transpose_8_cast_fp16 = transpose(perm = transpose_8_perm_0, x = input_35_batch_first_cast_fp16_0)[name = tensor("transpose_21")]; + tensor x_47_cast_fp16 = layer_norm(axes = x_47_axes_0, epsilon = var_136_to_fp16, x = transpose_8_cast_fp16)[name = tensor("x_47_cast_fp16")]; + tensor var_255_promoted_to_fp16 = const()[name = tensor("op_255_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor gamma_7_cast_fp16 = transpose(perm = gamma_7_perm_0, x = var_249_cast_fp16_0)[name = tensor("transpose_20")]; + tensor var_256_cast_fp16 = add(x = gamma_7_cast_fp16, y = var_255_promoted_to_fp16)[name = tensor("op_256_cast_fp16")]; + tensor var_257_cast_fp16 = mul(x = var_256_cast_fp16, y = x_47_cast_fp16)[name = tensor("op_257_cast_fp16")]; + tensor beta_7_cast_fp16 = transpose(perm = beta_7_perm_0, x = var_249_cast_fp16_1)[name = tensor("transpose_19")]; + tensor x_49_cast_fp16 = add(x = var_257_cast_fp16, y = beta_7_cast_fp16)[name = tensor("x_49_cast_fp16")]; + tensor x_51_perm_0 = const()[name = tensor("x_51_perm_0"), val = tensor([0, -1, -2])]; + tensor x_53_interleave_0 = const()[name = tensor("x_53_interleave_0"), val = tensor(false)]; + tensor x_51_cast_fp16 = transpose(perm = x_51_perm_0, x = x_49_cast_fp16)[name = tensor("transpose_18")]; + tensor x_53_cast_fp16 = concat(axis = var_147, interleave = x_53_interleave_0, values = (x_51_cast_fp16, s_chan_cast_fp16))[name = tensor("x_53_cast_fp16")]; + tensor transpose_5_perm_0 = const()[name = tensor("transpose_5_perm_0"), val = tensor([-1, 0, -2])]; + tensor input_41_batch_first_direction_0 = const()[name = tensor("input_41_batch_first_direction_0"), val = tensor("bidirectional")]; + tensor input_41_batch_first_output_sequence_0 = const()[name = tensor("input_41_batch_first_output_sequence_0"), val = tensor(true)]; + tensor input_41_batch_first_recurrent_activation_0 = const()[name = tensor("input_41_batch_first_recurrent_activation_0"), val = tensor("sigmoid")]; + tensor input_41_batch_first_cell_activation_0 = const()[name = tensor("input_41_batch_first_cell_activation_0"), val = tensor("tanh")]; + tensor input_41_batch_first_activation_0 = const()[name = tensor("input_41_batch_first_activation_0"), val = tensor("tanh")]; + tensor concat_34_to_fp16 = const()[name = tensor("concat_34_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19873216)))]; + tensor concat_35_to_fp16 = const()[name = tensor("concat_35_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(21184000)))]; + tensor add_10_to_fp16 = const()[name = tensor("add_10_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(21708352)))]; + tensor concat_36_to_fp16 = const()[name = tensor("concat_36_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(21710464)))]; + tensor concat_37_to_fp16 = const()[name = tensor("concat_37_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23021248)))]; + tensor add_11_to_fp16 = const()[name = tensor("add_11_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23545600)))]; + tensor transpose_5_cast_fp16 = transpose(perm = transpose_5_perm_0, x = x_53_cast_fp16)[name = tensor("transpose_17")]; + tensor input_41_batch_first_cast_fp16_0, tensor input_41_batch_first_cast_fp16_1, tensor input_41_batch_first_cast_fp16_2 = lstm(activation = input_41_batch_first_activation_0, bias = add_10_to_fp16, bias_back = add_11_to_fp16, cell_activation = input_41_batch_first_cell_activation_0, direction = input_41_batch_first_direction_0, initial_c = x_17_batch_first_lstm_h0_reshaped_to_fp16, initial_h = x_17_batch_first_lstm_h0_reshaped_to_fp16, output_sequence = input_41_batch_first_output_sequence_0, recurrent_activation = input_41_batch_first_recurrent_activation_0, weight_hh = concat_35_to_fp16, weight_hh_back = concat_37_to_fp16, weight_ih = concat_34_to_fp16, weight_ih_back = concat_36_to_fp16, x = transpose_5_cast_fp16)[name = tensor("input_41_batch_first_cast_fp16")]; + tensor transpose_9_perm_0 = const()[name = tensor("transpose_9_perm_0"), val = tensor([1, 0, 2])]; + tensor duration_encoder_norms_2_fc_weight_to_fp16 = const()[name = tensor("duration_encoder_norms_2_fc_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23547712)))]; + tensor duration_encoder_norms_2_fc_bias_to_fp16 = const()[name = tensor("duration_encoder_norms_2_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23809920)))]; + tensor linear_3_cast_fp16 = linear(bias = duration_encoder_norms_2_fc_bias_to_fp16, weight = duration_encoder_norms_2_fc_weight_to_fp16, x = style_cast_fp16)[name = tensor("linear_3_cast_fp16")]; + tensor var_294 = const()[name = tensor("op_294"), val = tensor([1, 1024, 1])]; + tensor h_cast_fp16 = reshape(shape = var_294, x = linear_3_cast_fp16)[name = tensor("h_cast_fp16")]; + tensor var_296_split_sizes_0 = const()[name = tensor("op_296_split_sizes_0"), val = tensor([512, 512])]; + tensor var_296_axis_0 = const()[name = tensor("op_296_axis_0"), val = tensor(1)]; + tensor var_296_cast_fp16_0, tensor var_296_cast_fp16_1 = split(axis = var_296_axis_0, split_sizes = var_296_split_sizes_0, x = h_cast_fp16)[name = tensor("op_296_cast_fp16")]; + tensor gamma_11_perm_0 = const()[name = tensor("gamma_11_perm_0"), val = tensor([0, -1, 1])]; + tensor beta_11_perm_0 = const()[name = tensor("beta_11_perm_0"), val = tensor([0, -1, 1])]; + tensor x_63_axes_0 = const()[name = tensor("x_63_axes_0"), val = tensor([-1])]; + tensor transpose_9_cast_fp16 = transpose(perm = transpose_9_perm_0, x = input_41_batch_first_cast_fp16_0)[name = tensor("transpose_16")]; + tensor x_63_cast_fp16 = layer_norm(axes = x_63_axes_0, epsilon = var_136_to_fp16, x = transpose_9_cast_fp16)[name = tensor("x_63_cast_fp16")]; + tensor var_302_promoted_to_fp16 = const()[name = tensor("op_302_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor gamma_11_cast_fp16 = transpose(perm = gamma_11_perm_0, x = var_296_cast_fp16_0)[name = tensor("transpose_15")]; + tensor var_303_cast_fp16 = add(x = gamma_11_cast_fp16, y = var_302_promoted_to_fp16)[name = tensor("op_303_cast_fp16")]; + tensor var_304_cast_fp16 = mul(x = var_303_cast_fp16, y = x_63_cast_fp16)[name = tensor("op_304_cast_fp16")]; + tensor beta_11_cast_fp16 = transpose(perm = beta_11_perm_0, x = var_296_cast_fp16_1)[name = tensor("transpose_14")]; + tensor x_65_cast_fp16 = add(x = var_304_cast_fp16, y = beta_11_cast_fp16)[name = tensor("x_65_cast_fp16")]; + tensor x_67_perm_0 = const()[name = tensor("x_67_perm_0"), val = tensor([0, -1, -2])]; + tensor x_interleave_0 = const()[name = tensor("x_interleave_0"), val = tensor(false)]; + tensor x_67_cast_fp16 = transpose(perm = x_67_perm_0, x = x_65_cast_fp16)[name = tensor("transpose_13")]; + tensor x_cast_fp16 = concat(axis = var_147, interleave = x_interleave_0, values = (x_67_cast_fp16, s_chan_cast_fp16))[name = tensor("x_cast_fp16")]; + tensor input_45_perm_0 = const()[name = tensor("input_45_perm_0"), val = tensor([0, -1, -2])]; + tensor input_45_batch_first_transpose_perm_0 = const()[name = tensor("input_45_batch_first_transpose_perm_0"), val = tensor([1, 0, 2])]; + tensor input_47_batch_first_direction_0 = const()[name = tensor("input_47_batch_first_direction_0"), val = tensor("bidirectional")]; + tensor input_47_batch_first_output_sequence_0 = const()[name = tensor("input_47_batch_first_output_sequence_0"), val = tensor(true)]; + tensor input_47_batch_first_recurrent_activation_0 = const()[name = tensor("input_47_batch_first_recurrent_activation_0"), val = tensor("sigmoid")]; + tensor input_47_batch_first_cell_activation_0 = const()[name = tensor("input_47_batch_first_cell_activation_0"), val = tensor("tanh")]; + tensor input_47_batch_first_activation_0 = const()[name = tensor("input_47_batch_first_activation_0"), val = tensor("tanh")]; + tensor concat_44_to_fp16 = const()[name = tensor("concat_44_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23812032)))]; + tensor concat_45_to_fp16 = const()[name = tensor("concat_45_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(25122816)))]; + tensor add_13_to_fp16 = const()[name = tensor("add_13_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(25647168)))]; + tensor concat_46_to_fp16 = const()[name = tensor("concat_46_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(25649280)))]; + tensor concat_47_to_fp16 = const()[name = tensor("concat_47_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(26960064)))]; + tensor add_14_to_fp16 = const()[name = tensor("add_14_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(27484416)))]; + tensor d = transpose(perm = input_45_perm_0, x = x_cast_fp16)[name = tensor("transpose_12")]; + tensor input_45_batch_first_transpose_cast_fp16 = transpose(perm = input_45_batch_first_transpose_perm_0, x = d)[name = tensor("transpose_11")]; + tensor input_47_batch_first_cast_fp16_0, tensor input_47_batch_first_cast_fp16_1, tensor input_47_batch_first_cast_fp16_2 = lstm(activation = input_47_batch_first_activation_0, bias = add_13_to_fp16, bias_back = add_14_to_fp16, cell_activation = input_47_batch_first_cell_activation_0, direction = input_47_batch_first_direction_0, initial_c = x_17_batch_first_lstm_h0_reshaped_to_fp16, initial_h = x_17_batch_first_lstm_h0_reshaped_to_fp16, output_sequence = input_47_batch_first_output_sequence_0, recurrent_activation = input_47_batch_first_recurrent_activation_0, weight_hh = concat_45_to_fp16, weight_hh_back = concat_47_to_fp16, weight_ih = concat_44_to_fp16, weight_ih_back = concat_46_to_fp16, x = input_45_batch_first_transpose_cast_fp16)[name = tensor("input_47_batch_first_cast_fp16")]; + tensor input_47_perm_0 = const()[name = tensor("input_47_perm_0"), val = tensor([1, 0, 2])]; + tensor duration_proj_linear_layer_weight_to_fp16 = const()[name = tensor("duration_proj_linear_layer_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(27486528)))]; + tensor duration_proj_linear_layer_bias_to_fp16 = const()[name = tensor("duration_proj_linear_layer_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(27537792)))]; + tensor input_47_cast_fp16 = transpose(perm = input_47_perm_0, x = input_47_batch_first_cast_fp16_0)[name = tensor("transpose_10")]; + tensor pred_dur_log = linear(bias = duration_proj_linear_layer_bias_to_fp16, weight = duration_proj_linear_layer_weight_to_fp16, x = input_47_cast_fp16)[name = tensor("linear_4_cast_fp16")]; + tensor var_345 = const()[name = tensor("op_345"), val = tensor(-1)]; + tensor var_355_begin_0 = const()[name = tensor("op_355_begin_0"), val = tensor([0, 0, 0])]; + tensor var_355_end_0 = const()[name = tensor("op_355_end_0"), val = tensor([1, 0, 1])]; + tensor var_355_end_mask_0 = const()[name = tensor("op_355_end_mask_0"), val = tensor([true, true, false])]; + tensor var_355_squeeze_mask_0 = const()[name = tensor("op_355_squeeze_mask_0"), val = tensor([false, false, true])]; + tensor var_355_cast_fp16 = slice_by_index(begin = var_355_begin_0, end = var_355_end_0, end_mask = var_355_end_mask_0, squeeze_mask = var_355_squeeze_mask_0, x = bert_dur)[name = tensor("op_355_cast_fp16")]; + tensor fill_like_0_value_0_to_fp16 = const()[name = tensor("fill_like_0_value_0_to_fp16"), val = tensor(0x1p+0)]; + tensor fill_like_0_cast_fp16 = fill_like(ref_tensor = var_355_cast_fp16, value = fill_like_0_value_0_to_fp16)[name = tensor("fill_like_0_cast_fp16")]; + tensor var_357_exclusive_0 = const()[name = tensor("op_357_exclusive_0"), val = tensor(false)]; + tensor var_357_reverse_0 = const()[name = tensor("op_357_reverse_0"), val = tensor(false)]; + tensor var_357_cast_fp16 = cumsum(axis = var_345, exclusive = var_357_exclusive_0, reverse = var_357_reverse_0, x = fill_like_0_cast_fp16)[name = tensor("op_357_cast_fp16")]; + tensor cast_6_dtype_0 = const()[name = tensor("cast_6_dtype_0"), val = tensor("int32")]; + tensor var_359 = const()[name = tensor("op_359"), val = tensor(1)]; + tensor var_357_cast_fp16_to_int32 = cast(dtype = cast_6_dtype_0, x = var_357_cast_fp16)[name = tensor("cast_13")]; + tensor indices = sub(x = var_357_cast_fp16_to_int32, y = var_359)[name = tensor("indices")]; + tensor input_axes_0 = const()[name = tensor("input_axes_0"), val = tensor([0])]; + tensor indices_to_uint16_dtype_0 = const()[name = tensor("indices_to_uint16_dtype_0"), val = tensor("uint16")]; + tensor indices_to_uint16 = cast(dtype = indices_to_uint16_dtype_0, x = indices)[name = tensor("cast_12")]; + tensor input_cast_uint16 = squeeze(axes = input_axes_0, x = indices_to_uint16)[name = tensor("input_cast_uint16")]; + tensor fe_axis_0 = const()[name = tensor("fe_axis_0"), val = tensor(0)]; + tensor fe_batch_dims_0 = const()[name = tensor("fe_batch_dims_0"), val = tensor(0)]; + tensor fe_validate_indices_0 = const()[name = tensor("fe_validate_indices_0"), val = tensor(false)]; + tensor fixed_embedding_embedding_weight_to_fp16 = const()[name = tensor("fixed_embedding_embedding_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(27537984)))]; + tensor fe_cast_fp16_cast_uint16 = gather(axis = fe_axis_0, batch_dims = fe_batch_dims_0, indices = input_cast_uint16, validate_indices = fe_validate_indices_0, x = fixed_embedding_embedding_weight_to_fp16)[name = tensor("fe_cast_fp16_cast_uint16")]; + tensor var_364_axes_0 = const()[name = tensor("op_364_axes_0"), val = tensor([0])]; + tensor fixed_embedding = expand_dims(axes = var_364_axes_0, x = fe_cast_fp16_cast_uint16)[name = tensor("op_364_cast_fp16")]; + tensor t_en = transpose(perm = transpose_0_perm_0, x = x_17_batch_first_cast_fp16_0)[name = tensor("transpose_29")]; + } -> (t_en, d, pred_dur_log, fixed_embedding); +} \ No newline at end of file diff --git a/ANE/styletts2_ane_postbert.mlmodelc/weights/weight.bin b/ANE/styletts2_ane_postbert.mlmodelc/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..8b3c44f64887db89ecfe75de6e5e9701dbec5787 --- /dev/null +++ b/ANE/styletts2_ane_postbert.mlmodelc/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bffc0f40bfeee94d93537de68e3057f8e18e847c2197bef2045118d25e950a0d +size 28324480 diff --git a/ANE/styletts2_ane_postbert.mlpackage/Data/com.apple.CoreML/model.mlmodel b/ANE/styletts2_ane_postbert.mlpackage/Data/com.apple.CoreML/model.mlmodel new file mode 100644 index 0000000000000000000000000000000000000000..cb6eb23ef06b0f13ff217129d890818d155e8a11 --- /dev/null +++ b/ANE/styletts2_ane_postbert.mlpackage/Data/com.apple.CoreML/model.mlmodel @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:982078ab6840241c1b6717dab503af3f123373606fd6b31967f33845468aa0fd +size 43623 diff --git a/ANE/styletts2_ane_postbert.mlpackage/Data/com.apple.CoreML/weights/weight.bin b/ANE/styletts2_ane_postbert.mlpackage/Data/com.apple.CoreML/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..8b3c44f64887db89ecfe75de6e5e9701dbec5787 --- /dev/null +++ b/ANE/styletts2_ane_postbert.mlpackage/Data/com.apple.CoreML/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bffc0f40bfeee94d93537de68e3057f8e18e847c2197bef2045118d25e950a0d +size 28324480 diff --git a/ANE/styletts2_ane_postbert.mlpackage/Manifest.json b/ANE/styletts2_ane_postbert.mlpackage/Manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..cc72c11557c12399403a105981e8b30e86c54b33 --- /dev/null +++ b/ANE/styletts2_ane_postbert.mlpackage/Manifest.json @@ -0,0 +1,18 @@ +{ + "fileFormatVersion": "1.0.0", + "itemInfoEntries": { + "4577D915-6FB9-4738-8BE7-3DAA463EE8BD": { + "author": "com.apple.CoreML", + "description": "CoreML Model Specification", + "name": "model.mlmodel", + "path": "com.apple.CoreML/model.mlmodel" + }, + "8CBB3422-373A-42D0-8E4F-AF8A2A241439": { + "author": "com.apple.CoreML", + "description": "CoreML Model Weights", + "name": "weights", + "path": "com.apple.CoreML/weights" + } + }, + "rootModelIdentifier": "4577D915-6FB9-4738-8BE7-3DAA463EE8BD" +} diff --git a/ANE/styletts2_ane_prosody.mlmodelc/analytics/coremldata.bin b/ANE/styletts2_ane_prosody.mlmodelc/analytics/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..a5100b4e70906f1f0005b0bca084d76bdf9a95c5 --- /dev/null +++ b/ANE/styletts2_ane_prosody.mlmodelc/analytics/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a569de5bd6b2a4cb48a7ec8c1544f80e7c6484933bb9160dc7c88b462c23c9cc +size 243 diff --git a/ANE/styletts2_ane_prosody.mlmodelc/coremldata.bin b/ANE/styletts2_ane_prosody.mlmodelc/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..3859a5eacafc29e4641e383e261740dcac1c2f2b --- /dev/null +++ b/ANE/styletts2_ane_prosody.mlmodelc/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:97801e9d7ade59ab0c78110290f93222ad0430e0c479125ef09755433bff3ac7 +size 417 diff --git a/ANE/styletts2_ane_prosody.mlmodelc/metadata.json b/ANE/styletts2_ane_prosody.mlmodelc/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..4fbd70f7be2ed3a342d1bd919e1a8e56b9256623 --- /dev/null +++ b/ANE/styletts2_ane_prosody.mlmodelc/metadata.json @@ -0,0 +1,97 @@ +[ + { + "metadataOutputVersion" : "3.0", + "shortDescription" : "StyleTTS2-ANE Stage 5: Prosody F0Ntrain (static T_a=2000) — fp16, ANE.", + "outputSchema" : [ + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float16", + "formattedType" : "MultiArray (Float16 1 × 4000)", + "shortDescription" : "", + "shape" : "[1, 4000]", + "name" : "F0", + "type" : "MultiArray" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float16", + "formattedType" : "MultiArray (Float16 1 × 4000)", + "shortDescription" : "", + "shape" : "[1, 4000]", + "name" : "N", + "type" : "MultiArray" + } + ], + "storagePrecision" : "Mixed (Float16, Palettized (8 bits))", + "modelParameters" : [ + + ], + "specificationVersion" : 8, + "mlProgramOperationTypeHistogram" : { + "Ios17.squeeze" : 4, + "Ios17.mul" : 18, + "Ios17.linear" : 12, + "Ios17.transpose" : 2, + "Ios16.constexprLutToDense" : 30, + "Ios17.conv" : 16, + "Ios17.leakyRelu" : 12, + "Ios17.add" : 30, + "Ios17.lstm" : 1, + "Ios17.convTranspose" : 2, + "Ios17.sliceByIndex" : 2, + "UpsampleNearestNeighbor" : 2, + "Ios17.expandDims" : 2, + "Ios17.instanceNorm" : 11, + "Ios17.reshape" : 12, + "Ios17.cast" : 2, + "Split" : 12 + }, + "computePrecision" : "Mixed (Float16, Float32, Int32)", + "isUpdatable" : "0", + "stateSchema" : [ + + ], + "availability" : { + "macOS" : "14.0", + "tvOS" : "17.0", + "visionOS" : "1.0", + "watchOS" : "10.0", + "iOS" : "17.0", + "macCatalyst" : "17.0" + }, + "modelType" : { + "name" : "MLModelType_mlProgram" + }, + "inputSchema" : [ + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float32", + "formattedType" : "MultiArray (Float32 1 × 640 × 2000)", + "shortDescription" : "", + "shape" : "[1, 640, 2000]", + "name" : "en", + "type" : "MultiArray" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float32", + "formattedType" : "MultiArray (Float32 1 × 128)", + "shortDescription" : "", + "shape" : "[1, 128]", + "name" : "s", + "type" : "MultiArray" + } + ], + "userDefinedMetadata" : { + "com.github.apple.coremltools.version" : "8.3.0", + "com.github.apple.coremltools.source" : "torch==2.11.0", + "com.github.apple.coremltools.source_dialect" : "TorchScript" + }, + "generatedClassName" : "styletts2_ane_prosody", + "method" : "predict" + } +] \ No newline at end of file diff --git a/ANE/styletts2_ane_prosody.mlmodelc/model.mil b/ANE/styletts2_ane_prosody.mlmodelc/model.mil new file mode 100644 index 0000000000000000000000000000000000000000..f1170668f0dcb0da1b3b9185c2d49d86afa9706b --- /dev/null +++ b/ANE/styletts2_ane_prosody.mlmodelc/model.mil @@ -0,0 +1,396 @@ +program(1.0) +[buildInfo = dict, tensor>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}})] +{ + func main(tensor en, tensor s) { + tensor transpose_0_perm_0 = const()[name = tensor("transpose_0_perm_0"), val = tensor([-1, 0, -2])]; + tensor en_to_fp16_dtype_0 = const()[name = tensor("en_to_fp16_dtype_0"), val = tensor("fp16")]; + tensor x_batch_first_direction_0 = const()[name = tensor("x_batch_first_direction_0"), val = tensor("bidirectional")]; + tensor x_batch_first_output_sequence_0 = const()[name = tensor("x_batch_first_output_sequence_0"), val = tensor(true)]; + tensor x_batch_first_recurrent_activation_0 = const()[name = tensor("x_batch_first_recurrent_activation_0"), val = tensor("sigmoid")]; + tensor x_batch_first_cell_activation_0 = const()[name = tensor("x_batch_first_cell_activation_0"), val = tensor("tanh")]; + tensor x_batch_first_activation_0 = const()[name = tensor("x_batch_first_activation_0"), val = tensor("tanh")]; + tensor x_batch_first_lstm_h0_reshaped_to_fp16 = const()[name = tensor("x_batch_first_lstm_h0_reshaped_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(64)))]; + tensor concat_4_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(656576))), name = tensor("concat_4_to_fp16_palettized"), shape = tensor([1024, 640])]; + tensor concat_5_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(657152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(919360))), name = tensor("concat_5_to_fp16_palettized"), shape = tensor([1024, 256])]; + tensor add_0_to_fp16 = const()[name = tensor("add_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(919936)))]; + tensor concat_6_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(922048))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1577472))), name = tensor("concat_6_to_fp16_palettized"), shape = tensor([1024, 640])]; + tensor concat_7_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1578048))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1840256))), name = tensor("concat_7_to_fp16_palettized"), shape = tensor([1024, 256])]; + tensor add_1_to_fp16 = const()[name = tensor("add_1_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1840832)))]; + tensor en_to_fp16 = cast(dtype = en_to_fp16_dtype_0, x = en)[name = tensor("cast_1")]; + tensor transpose_0_cast_fp16 = transpose(perm = transpose_0_perm_0, x = en_to_fp16)[name = tensor("transpose_2")]; + tensor x_batch_first_cast_fp16_0, tensor x_batch_first_cast_fp16_1, tensor x_batch_first_cast_fp16_2 = lstm(activation = x_batch_first_activation_0, bias = add_0_to_fp16, bias_back = add_1_to_fp16, cell_activation = x_batch_first_cell_activation_0, direction = x_batch_first_direction_0, initial_c = x_batch_first_lstm_h0_reshaped_to_fp16, initial_h = x_batch_first_lstm_h0_reshaped_to_fp16, output_sequence = x_batch_first_output_sequence_0, recurrent_activation = x_batch_first_recurrent_activation_0, weight_hh = concat_5_to_fp16_palettized, weight_hh_back = concat_7_to_fp16_palettized, weight_ih = concat_4_to_fp16_palettized, weight_ih_back = concat_6_to_fp16_palettized, x = transpose_0_cast_fp16)[name = tensor("x_batch_first_cast_fp16")]; + tensor var_53 = const()[name = tensor("op_53"), val = tensor(0x1.99999ap-3)]; + tensor s_to_fp16_dtype_0 = const()[name = tensor("s_to_fp16_dtype_0"), val = tensor("fp16")]; + tensor F0_0_norm1_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1842944))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1974080))), name = tensor("F0_0_norm1_fc_weight_to_fp16_palettized"), shape = tensor([1024, 128])]; + tensor F0_0_norm1_fc_bias_to_fp16 = const()[name = tensor("F0_0_norm1_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1974656)))]; + tensor s_to_fp16 = cast(dtype = s_to_fp16_dtype_0, x = s)[name = tensor("cast_0")]; + tensor linear_0_cast_fp16 = linear(bias = F0_0_norm1_fc_bias_to_fp16, weight = F0_0_norm1_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_0_cast_fp16")]; + tensor var_78 = const()[name = tensor("op_78"), val = tensor([1, 1024, 1])]; + tensor h_3_cast_fp16 = reshape(shape = var_78, x = linear_0_cast_fp16)[name = tensor("h_3_cast_fp16")]; + tensor var_80_split_sizes_0 = const()[name = tensor("op_80_split_sizes_0"), val = tensor([512, 512])]; + tensor var_80_axis_0 = const()[name = tensor("op_80_axis_0"), val = tensor(1)]; + tensor var_80_cast_fp16_0, tensor var_80_cast_fp16_1 = split(axis = var_80_axis_0, split_sizes = var_80_split_sizes_0, x = h_3_cast_fp16)[name = tensor("op_80_cast_fp16")]; + tensor var_82_promoted_to_fp16 = const()[name = tensor("op_82_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_83_cast_fp16 = add(x = var_80_cast_fp16_0, y = var_82_promoted_to_fp16)[name = tensor("op_83_cast_fp16")]; + tensor var_56_to_fp16 = const()[name = tensor("op_56_to_fp16"), val = tensor(0x1.5p-17)]; + tensor transpose_0_perm_0_1 = const()[name = tensor("transpose_0_perm_0_1"), val = tensor([1, 2, 0])]; + tensor transpose_0 = transpose(perm = transpose_0_perm_0_1, x = x_batch_first_cast_fp16_0)[name = tensor("transpose_1")]; + tensor var_84_cast_fp16 = instance_norm(epsilon = var_56_to_fp16, x = transpose_0)[name = tensor("op_84_cast_fp16")]; + tensor var_85_cast_fp16 = mul(x = var_83_cast_fp16, y = var_84_cast_fp16)[name = tensor("op_85_cast_fp16")]; + tensor input_5_cast_fp16 = add(x = var_85_cast_fp16, y = var_80_cast_fp16_1)[name = tensor("input_5_cast_fp16")]; + tensor input_7_cast_fp16 = leaky_relu(alpha = var_53, x = input_5_cast_fp16)[name = tensor("input_7_cast_fp16")]; + tensor input_11_pad_type_0 = const()[name = tensor("input_11_pad_type_0"), val = tensor("custom")]; + tensor input_11_pad_0 = const()[name = tensor("input_11_pad_0"), val = tensor([1, 1])]; + tensor input_11_strides_0 = const()[name = tensor("input_11_strides_0"), val = tensor([1])]; + tensor input_11_dilations_0 = const()[name = tensor("input_11_dilations_0"), val = tensor([1])]; + tensor input_11_groups_0 = const()[name = tensor("input_11_groups_0"), val = tensor(1)]; + tensor weight_1_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1976768))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2763264))), name = tensor("weight_1_to_fp16_palettized"), shape = tensor([512, 512, 3])]; + tensor F0_0_conv1_bias_to_fp16 = const()[name = tensor("F0_0_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2763840)))]; + tensor input_11_cast_fp16 = conv(bias = F0_0_conv1_bias_to_fp16, dilations = input_11_dilations_0, groups = input_11_groups_0, pad = input_11_pad_0, pad_type = input_11_pad_type_0, strides = input_11_strides_0, weight = weight_1_to_fp16_palettized, x = input_7_cast_fp16)[name = tensor("input_11_cast_fp16")]; + tensor F0_0_norm2_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2764928))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2896064))), name = tensor("F0_0_norm2_fc_weight_to_fp16_palettized"), shape = tensor([1024, 128])]; + tensor F0_0_norm2_fc_bias_to_fp16 = const()[name = tensor("F0_0_norm2_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2896640)))]; + tensor linear_1_cast_fp16 = linear(bias = F0_0_norm2_fc_bias_to_fp16, weight = F0_0_norm2_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_1_cast_fp16")]; + tensor var_102 = const()[name = tensor("op_102"), val = tensor([1, 1024, 1])]; + tensor h_7_cast_fp16 = reshape(shape = var_102, x = linear_1_cast_fp16)[name = tensor("h_7_cast_fp16")]; + tensor var_104_split_sizes_0 = const()[name = tensor("op_104_split_sizes_0"), val = tensor([512, 512])]; + tensor var_104_axis_0 = const()[name = tensor("op_104_axis_0"), val = tensor(1)]; + tensor var_104_cast_fp16_0, tensor var_104_cast_fp16_1 = split(axis = var_104_axis_0, split_sizes = var_104_split_sizes_0, x = h_7_cast_fp16)[name = tensor("op_104_cast_fp16")]; + tensor var_106_promoted_to_fp16 = const()[name = tensor("op_106_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_107_cast_fp16 = add(x = var_104_cast_fp16_0, y = var_106_promoted_to_fp16)[name = tensor("op_107_cast_fp16")]; + tensor var_108_cast_fp16 = instance_norm(epsilon = var_56_to_fp16, x = input_11_cast_fp16)[name = tensor("op_108_cast_fp16")]; + tensor var_109_cast_fp16 = mul(x = var_107_cast_fp16, y = var_108_cast_fp16)[name = tensor("op_109_cast_fp16")]; + tensor input_13_cast_fp16 = add(x = var_109_cast_fp16, y = var_104_cast_fp16_1)[name = tensor("input_13_cast_fp16")]; + tensor input_15_cast_fp16 = leaky_relu(alpha = var_53, x = input_13_cast_fp16)[name = tensor("input_15_cast_fp16")]; + tensor out_1_pad_type_0 = const()[name = tensor("out_1_pad_type_0"), val = tensor("custom")]; + tensor out_1_pad_0 = const()[name = tensor("out_1_pad_0"), val = tensor([1, 1])]; + tensor out_1_strides_0 = const()[name = tensor("out_1_strides_0"), val = tensor([1])]; + tensor out_1_dilations_0 = const()[name = tensor("out_1_dilations_0"), val = tensor([1])]; + tensor out_1_groups_0 = const()[name = tensor("out_1_groups_0"), val = tensor(1)]; + tensor weight_3_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2898752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3685248))), name = tensor("weight_3_to_fp16_palettized"), shape = tensor([512, 512, 3])]; + tensor F0_0_conv2_bias_to_fp16 = const()[name = tensor("F0_0_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3685824)))]; + tensor out_1_cast_fp16 = conv(bias = F0_0_conv2_bias_to_fp16, dilations = out_1_dilations_0, groups = out_1_groups_0, pad = out_1_pad_0, pad_type = out_1_pad_type_0, strides = out_1_strides_0, weight = weight_3_to_fp16_palettized, x = input_15_cast_fp16)[name = tensor("out_1_cast_fp16")]; + tensor var_120_cast_fp16 = add(x = out_1_cast_fp16, y = transpose_0)[name = tensor("op_120_cast_fp16")]; + tensor _inversed_input_19_y_0_to_fp16 = const()[name = tensor("_inversed_input_19_y_0_to_fp16"), val = tensor(0x1.6ap-1)]; + tensor _inversed_input_19_cast_fp16 = mul(x = var_120_cast_fp16, y = _inversed_input_19_y_0_to_fp16)[name = tensor("_inversed_input_19_cast_fp16")]; + tensor var_126 = const()[name = tensor("op_126"), val = tensor(0x1.99999ap-3)]; + tensor F0_1_norm1_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3686912))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3818048))), name = tensor("F0_1_norm1_fc_weight_to_fp16_palettized"), shape = tensor([1024, 128])]; + tensor F0_1_norm1_fc_bias_to_fp16 = const()[name = tensor("F0_1_norm1_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3818624)))]; + tensor linear_2_cast_fp16 = linear(bias = F0_1_norm1_fc_bias_to_fp16, weight = F0_1_norm1_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_2_cast_fp16")]; + tensor var_161 = const()[name = tensor("op_161"), val = tensor([1, 1024, 1])]; + tensor h_11_cast_fp16 = reshape(shape = var_161, x = linear_2_cast_fp16)[name = tensor("h_11_cast_fp16")]; + tensor var_163_split_sizes_0 = const()[name = tensor("op_163_split_sizes_0"), val = tensor([512, 512])]; + tensor var_163_axis_0 = const()[name = tensor("op_163_axis_0"), val = tensor(1)]; + tensor var_163_cast_fp16_0, tensor var_163_cast_fp16_1 = split(axis = var_163_axis_0, split_sizes = var_163_split_sizes_0, x = h_11_cast_fp16)[name = tensor("op_163_cast_fp16")]; + tensor var_165_promoted_to_fp16 = const()[name = tensor("op_165_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_166_cast_fp16 = add(x = var_163_cast_fp16_0, y = var_165_promoted_to_fp16)[name = tensor("op_166_cast_fp16")]; + tensor var_130_to_fp16 = const()[name = tensor("op_130_to_fp16"), val = tensor(0x1.5p-17)]; + tensor var_167_cast_fp16 = instance_norm(epsilon = var_130_to_fp16, x = _inversed_input_19_cast_fp16)[name = tensor("op_167_cast_fp16")]; + tensor var_168_cast_fp16 = mul(x = var_166_cast_fp16, y = var_167_cast_fp16)[name = tensor("op_168_cast_fp16")]; + tensor input_21_cast_fp16 = add(x = var_168_cast_fp16, y = var_163_cast_fp16_1)[name = tensor("input_21_cast_fp16")]; + tensor input_23_cast_fp16 = leaky_relu(alpha = var_126, x = input_21_cast_fp16)[name = tensor("input_23_cast_fp16")]; + tensor conv_transpose_0_pad_type_0 = const()[name = tensor("conv_transpose_0_pad_type_0"), val = tensor("custom")]; + tensor conv_transpose_0_pad_0 = const()[name = tensor("conv_transpose_0_pad_0"), val = tensor([0, 0])]; + tensor conv_transpose_0_strides_0 = const()[name = tensor("conv_transpose_0_strides_0"), val = tensor([2])]; + tensor conv_transpose_0_groups_0 = const()[name = tensor("conv_transpose_0_groups_0"), val = tensor(512)]; + tensor conv_transpose_0_dilations_0 = const()[name = tensor("conv_transpose_0_dilations_0"), val = tensor([1])]; + tensor conv_transpose_0_has_output_shape_output_shape_0 = const()[name = tensor("conv_transpose_0_has_output_shape_output_shape_0"), val = tensor([1, 512, 4001])]; + tensor var_171_to_fp16 = const()[name = tensor("op_171_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3820736)))]; + tensor F0_1_pool_bias_to_fp16 = const()[name = tensor("F0_1_pool_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3823872)))]; + tensor conv_transpose_0_has_output_shape_cast_fp16 = conv_transpose(bias = F0_1_pool_bias_to_fp16, dilations = conv_transpose_0_dilations_0, groups = conv_transpose_0_groups_0, output_shape = conv_transpose_0_has_output_shape_output_shape_0, pad = conv_transpose_0_pad_0, pad_type = conv_transpose_0_pad_type_0, strides = conv_transpose_0_strides_0, weight = var_171_to_fp16, x = input_23_cast_fp16)[name = tensor("conv_transpose_0_has_output_shape_cast_fp16")]; + tensor input_25_begin_0 = const()[name = tensor("input_25_begin_0"), val = tensor([0, 0, 1])]; + tensor input_25_end_0 = const()[name = tensor("input_25_end_0"), val = tensor([0, 0, 0])]; + tensor input_25_begin_mask_0 = const()[name = tensor("input_25_begin_mask_0"), val = tensor([true, true, false])]; + tensor input_25_end_mask_0 = const()[name = tensor("input_25_end_mask_0"), val = tensor([true, true, true])]; + tensor input_25_cast_fp16 = slice_by_index(begin = input_25_begin_0, begin_mask = input_25_begin_mask_0, end = input_25_end_0, end_mask = input_25_end_mask_0, x = conv_transpose_0_has_output_shape_cast_fp16)[name = tensor("input_25_cast_fp16")]; + tensor input_29_pad_type_0 = const()[name = tensor("input_29_pad_type_0"), val = tensor("custom")]; + tensor input_29_pad_0 = const()[name = tensor("input_29_pad_0"), val = tensor([1, 1])]; + tensor input_29_strides_0 = const()[name = tensor("input_29_strides_0"), val = tensor([1])]; + tensor input_29_dilations_0 = const()[name = tensor("input_29_dilations_0"), val = tensor([1])]; + tensor input_29_groups_0 = const()[name = tensor("input_29_groups_0"), val = tensor(1)]; + tensor weight_5_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3824960))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4218240))), name = tensor("weight_5_to_fp16_palettized"), shape = tensor([256, 512, 3])]; + tensor F0_1_conv1_bias_to_fp16 = const()[name = tensor("F0_1_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4218816)))]; + tensor input_29_cast_fp16 = conv(bias = F0_1_conv1_bias_to_fp16, dilations = input_29_dilations_0, groups = input_29_groups_0, pad = input_29_pad_0, pad_type = input_29_pad_type_0, strides = input_29_strides_0, weight = weight_5_to_fp16_palettized, x = input_25_cast_fp16)[name = tensor("input_29_cast_fp16")]; + tensor F0_1_norm2_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4219392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4284992))), name = tensor("F0_1_norm2_fc_weight_to_fp16_palettized"), shape = tensor([512, 128])]; + tensor F0_1_norm2_fc_bias_to_fp16 = const()[name = tensor("F0_1_norm2_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4285568)))]; + tensor linear_3_cast_fp16 = linear(bias = F0_1_norm2_fc_bias_to_fp16, weight = F0_1_norm2_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_3_cast_fp16")]; + tensor var_192 = const()[name = tensor("op_192"), val = tensor([1, 512, 1])]; + tensor h_15_cast_fp16 = reshape(shape = var_192, x = linear_3_cast_fp16)[name = tensor("h_15_cast_fp16")]; + tensor var_194_split_sizes_0 = const()[name = tensor("op_194_split_sizes_0"), val = tensor([256, 256])]; + tensor var_194_axis_0 = const()[name = tensor("op_194_axis_0"), val = tensor(1)]; + tensor var_194_cast_fp16_0, tensor var_194_cast_fp16_1 = split(axis = var_194_axis_0, split_sizes = var_194_split_sizes_0, x = h_15_cast_fp16)[name = tensor("op_194_cast_fp16")]; + tensor var_196_promoted_to_fp16 = const()[name = tensor("op_196_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_197_cast_fp16 = add(x = var_194_cast_fp16_0, y = var_196_promoted_to_fp16)[name = tensor("op_197_cast_fp16")]; + tensor var_198_cast_fp16 = instance_norm(epsilon = var_130_to_fp16, x = input_29_cast_fp16)[name = tensor("op_198_cast_fp16")]; + tensor var_199_cast_fp16 = mul(x = var_197_cast_fp16, y = var_198_cast_fp16)[name = tensor("op_199_cast_fp16")]; + tensor input_31_cast_fp16 = add(x = var_199_cast_fp16, y = var_194_cast_fp16_1)[name = tensor("input_31_cast_fp16")]; + tensor input_33_cast_fp16 = leaky_relu(alpha = var_126, x = input_31_cast_fp16)[name = tensor("input_33_cast_fp16")]; + tensor out_3_pad_type_0 = const()[name = tensor("out_3_pad_type_0"), val = tensor("custom")]; + tensor out_3_pad_0 = const()[name = tensor("out_3_pad_0"), val = tensor([1, 1])]; + tensor out_3_strides_0 = const()[name = tensor("out_3_strides_0"), val = tensor([1])]; + tensor out_3_dilations_0 = const()[name = tensor("out_3_dilations_0"), val = tensor([1])]; + tensor out_3_groups_0 = const()[name = tensor("out_3_groups_0"), val = tensor(1)]; + tensor weight_7_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4286656))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4483328))), name = tensor("weight_7_to_fp16_palettized"), shape = tensor([256, 256, 3])]; + tensor F0_1_conv2_bias_to_fp16 = const()[name = tensor("F0_1_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4483904)))]; + tensor out_3_cast_fp16 = conv(bias = F0_1_conv2_bias_to_fp16, dilations = out_3_dilations_0, groups = out_3_groups_0, pad = out_3_pad_0, pad_type = out_3_pad_type_0, strides = out_3_strides_0, weight = weight_7_to_fp16_palettized, x = input_33_cast_fp16)[name = tensor("out_3_cast_fp16")]; + tensor expand_dims_0_axes_0 = const()[name = tensor("expand_dims_0_axes_0"), val = tensor([3])]; + tensor expand_dims_0_cast_fp16 = expand_dims(axes = expand_dims_0_axes_0, x = _inversed_input_19_cast_fp16)[name = tensor("expand_dims_0_cast_fp16")]; + tensor upsample_nearest_neighbor_0_scale_factor_height_0 = const()[name = tensor("upsample_nearest_neighbor_0_scale_factor_height_0"), val = tensor(2)]; + tensor upsample_nearest_neighbor_0_scale_factor_width_0 = const()[name = tensor("upsample_nearest_neighbor_0_scale_factor_width_0"), val = tensor(1)]; + tensor upsample_nearest_neighbor_0_cast_fp16 = upsample_nearest_neighbor(scale_factor_height = upsample_nearest_neighbor_0_scale_factor_height_0, scale_factor_width = upsample_nearest_neighbor_0_scale_factor_width_0, x = expand_dims_0_cast_fp16)[name = tensor("upsample_nearest_neighbor_0_cast_fp16")]; + tensor input_37_axes_0 = const()[name = tensor("input_37_axes_0"), val = tensor([3])]; + tensor input_37_cast_fp16 = squeeze(axes = input_37_axes_0, x = upsample_nearest_neighbor_0_cast_fp16)[name = tensor("input_37_cast_fp16")]; + tensor var_217_pad_type_0 = const()[name = tensor("op_217_pad_type_0"), val = tensor("valid")]; + tensor var_217_strides_0 = const()[name = tensor("op_217_strides_0"), val = tensor([1])]; + tensor var_217_pad_0 = const()[name = tensor("op_217_pad_0"), val = tensor([0, 0])]; + tensor var_217_dilations_0 = const()[name = tensor("op_217_dilations_0"), val = tensor([1])]; + tensor var_217_groups_0 = const()[name = tensor("op_217_groups_0"), val = tensor(1)]; + tensor weight_9_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4484480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4615616))), name = tensor("weight_9_to_fp16_palettized"), shape = tensor([256, 512, 1])]; + tensor var_217_cast_fp16 = conv(dilations = var_217_dilations_0, groups = var_217_groups_0, pad = var_217_pad_0, pad_type = var_217_pad_type_0, strides = var_217_strides_0, weight = weight_9_to_fp16_palettized, x = input_37_cast_fp16)[name = tensor("op_217_cast_fp16")]; + tensor var_218_cast_fp16 = add(x = out_3_cast_fp16, y = var_217_cast_fp16)[name = tensor("op_218_cast_fp16")]; + tensor _inversed_input_39_y_0_to_fp16 = const()[name = tensor("_inversed_input_39_y_0_to_fp16"), val = tensor(0x1.6ap-1)]; + tensor _inversed_input_39_cast_fp16 = mul(x = var_218_cast_fp16, y = _inversed_input_39_y_0_to_fp16)[name = tensor("_inversed_input_39_cast_fp16")]; + tensor var_222 = const()[name = tensor("op_222"), val = tensor(0x1.99999ap-3)]; + tensor F0_2_norm1_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4616192))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4681792))), name = tensor("F0_2_norm1_fc_weight_to_fp16_palettized"), shape = tensor([512, 128])]; + tensor F0_2_norm1_fc_bias_to_fp16 = const()[name = tensor("F0_2_norm1_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4682368)))]; + tensor linear_4_cast_fp16 = linear(bias = F0_2_norm1_fc_bias_to_fp16, weight = F0_2_norm1_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_4_cast_fp16")]; + tensor var_247 = const()[name = tensor("op_247"), val = tensor([1, 512, 1])]; + tensor h_19_cast_fp16 = reshape(shape = var_247, x = linear_4_cast_fp16)[name = tensor("h_19_cast_fp16")]; + tensor var_249_split_sizes_0 = const()[name = tensor("op_249_split_sizes_0"), val = tensor([256, 256])]; + tensor var_249_axis_0 = const()[name = tensor("op_249_axis_0"), val = tensor(1)]; + tensor var_249_cast_fp16_0, tensor var_249_cast_fp16_1 = split(axis = var_249_axis_0, split_sizes = var_249_split_sizes_0, x = h_19_cast_fp16)[name = tensor("op_249_cast_fp16")]; + tensor var_251_promoted_to_fp16 = const()[name = tensor("op_251_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_252_cast_fp16 = add(x = var_249_cast_fp16_0, y = var_251_promoted_to_fp16)[name = tensor("op_252_cast_fp16")]; + tensor var_225_to_fp16 = const()[name = tensor("op_225_to_fp16"), val = tensor(0x1.5p-17)]; + tensor var_253_cast_fp16 = instance_norm(epsilon = var_225_to_fp16, x = _inversed_input_39_cast_fp16)[name = tensor("op_253_cast_fp16")]; + tensor var_254_cast_fp16 = mul(x = var_252_cast_fp16, y = var_253_cast_fp16)[name = tensor("op_254_cast_fp16")]; + tensor input_41_cast_fp16 = add(x = var_254_cast_fp16, y = var_249_cast_fp16_1)[name = tensor("input_41_cast_fp16")]; + tensor input_43_cast_fp16 = leaky_relu(alpha = var_222, x = input_41_cast_fp16)[name = tensor("input_43_cast_fp16")]; + tensor input_47_pad_type_0 = const()[name = tensor("input_47_pad_type_0"), val = tensor("custom")]; + tensor input_47_pad_0 = const()[name = tensor("input_47_pad_0"), val = tensor([1, 1])]; + tensor input_47_strides_0 = const()[name = tensor("input_47_strides_0"), val = tensor([1])]; + tensor input_47_dilations_0 = const()[name = tensor("input_47_dilations_0"), val = tensor([1])]; + tensor input_47_groups_0 = const()[name = tensor("input_47_groups_0"), val = tensor(1)]; + tensor weight_11_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4683456))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4880128))), name = tensor("weight_11_to_fp16_palettized"), shape = tensor([256, 256, 3])]; + tensor F0_2_conv1_bias_to_fp16 = const()[name = tensor("F0_2_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4880704)))]; + tensor input_47_cast_fp16 = conv(bias = F0_2_conv1_bias_to_fp16, dilations = input_47_dilations_0, groups = input_47_groups_0, pad = input_47_pad_0, pad_type = input_47_pad_type_0, strides = input_47_strides_0, weight = weight_11_to_fp16_palettized, x = input_43_cast_fp16)[name = tensor("input_47_cast_fp16")]; + tensor F0_2_norm2_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4881280))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4946880))), name = tensor("F0_2_norm2_fc_weight_to_fp16_palettized"), shape = tensor([512, 128])]; + tensor F0_2_norm2_fc_bias_to_fp16 = const()[name = tensor("F0_2_norm2_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4947456)))]; + tensor linear_5_cast_fp16 = linear(bias = F0_2_norm2_fc_bias_to_fp16, weight = F0_2_norm2_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_5_cast_fp16")]; + tensor var_271 = const()[name = tensor("op_271"), val = tensor([1, 512, 1])]; + tensor h_23_cast_fp16 = reshape(shape = var_271, x = linear_5_cast_fp16)[name = tensor("h_23_cast_fp16")]; + tensor var_273_split_sizes_0 = const()[name = tensor("op_273_split_sizes_0"), val = tensor([256, 256])]; + tensor var_273_axis_0 = const()[name = tensor("op_273_axis_0"), val = tensor(1)]; + tensor var_273_cast_fp16_0, tensor var_273_cast_fp16_1 = split(axis = var_273_axis_0, split_sizes = var_273_split_sizes_0, x = h_23_cast_fp16)[name = tensor("op_273_cast_fp16")]; + tensor var_275_promoted_to_fp16 = const()[name = tensor("op_275_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_276_cast_fp16 = add(x = var_273_cast_fp16_0, y = var_275_promoted_to_fp16)[name = tensor("op_276_cast_fp16")]; + tensor var_277_cast_fp16 = instance_norm(epsilon = var_225_to_fp16, x = input_47_cast_fp16)[name = tensor("op_277_cast_fp16")]; + tensor var_278_cast_fp16 = mul(x = var_276_cast_fp16, y = var_277_cast_fp16)[name = tensor("op_278_cast_fp16")]; + tensor input_49_cast_fp16 = add(x = var_278_cast_fp16, y = var_273_cast_fp16_1)[name = tensor("input_49_cast_fp16")]; + tensor input_51_cast_fp16 = leaky_relu(alpha = var_222, x = input_49_cast_fp16)[name = tensor("input_51_cast_fp16")]; + tensor out_5_pad_type_0 = const()[name = tensor("out_5_pad_type_0"), val = tensor("custom")]; + tensor out_5_pad_0 = const()[name = tensor("out_5_pad_0"), val = tensor([1, 1])]; + tensor out_5_strides_0 = const()[name = tensor("out_5_strides_0"), val = tensor([1])]; + tensor out_5_dilations_0 = const()[name = tensor("out_5_dilations_0"), val = tensor([1])]; + tensor out_5_groups_0 = const()[name = tensor("out_5_groups_0"), val = tensor(1)]; + tensor weight_13_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4948544))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5145216))), name = tensor("weight_13_to_fp16_palettized"), shape = tensor([256, 256, 3])]; + tensor F0_2_conv2_bias_to_fp16 = const()[name = tensor("F0_2_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5145792)))]; + tensor out_5_cast_fp16 = conv(bias = F0_2_conv2_bias_to_fp16, dilations = out_5_dilations_0, groups = out_5_groups_0, pad = out_5_pad_0, pad_type = out_5_pad_type_0, strides = out_5_strides_0, weight = weight_13_to_fp16_palettized, x = input_51_cast_fp16)[name = tensor("out_5_cast_fp16")]; + tensor var_289_cast_fp16 = add(x = out_5_cast_fp16, y = _inversed_input_39_cast_fp16)[name = tensor("op_289_cast_fp16")]; + tensor _inversed_input_55_y_0_to_fp16 = const()[name = tensor("_inversed_input_55_y_0_to_fp16"), val = tensor(0x1.6ap-1)]; + tensor _inversed_input_55_cast_fp16 = mul(x = var_289_cast_fp16, y = _inversed_input_55_y_0_to_fp16)[name = tensor("_inversed_input_55_cast_fp16")]; + tensor var_302_pad_type_0 = const()[name = tensor("op_302_pad_type_0"), val = tensor("valid")]; + tensor var_302_strides_0 = const()[name = tensor("op_302_strides_0"), val = tensor([1])]; + tensor var_302_pad_0 = const()[name = tensor("op_302_pad_0"), val = tensor([0, 0])]; + tensor var_302_dilations_0 = const()[name = tensor("op_302_dilations_0"), val = tensor([1])]; + tensor var_302_groups_0 = const()[name = tensor("op_302_groups_0"), val = tensor(1)]; + tensor F0_proj_weight_to_fp16 = const()[name = tensor("F0_proj_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5146368)))]; + tensor F0_proj_bias_to_fp16 = const()[name = tensor("F0_proj_bias_to_fp16"), val = tensor([0x1.adp-4])]; + tensor var_302_cast_fp16 = conv(bias = F0_proj_bias_to_fp16, dilations = var_302_dilations_0, groups = var_302_groups_0, pad = var_302_pad_0, pad_type = var_302_pad_type_0, strides = var_302_strides_0, weight = F0_proj_weight_to_fp16, x = _inversed_input_55_cast_fp16)[name = tensor("op_302_cast_fp16")]; + tensor var_304_axes_0 = const()[name = tensor("op_304_axes_0"), val = tensor([1])]; + tensor F0 = squeeze(axes = var_304_axes_0, x = var_302_cast_fp16)[name = tensor("op_304_cast_fp16")]; + tensor var_309 = const()[name = tensor("op_309"), val = tensor(0x1.99999ap-3)]; + tensor N_0_norm1_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5146944))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5278080))), name = tensor("N_0_norm1_fc_weight_to_fp16_palettized"), shape = tensor([1024, 128])]; + tensor N_0_norm1_fc_bias_to_fp16 = const()[name = tensor("N_0_norm1_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5278656)))]; + tensor linear_6_cast_fp16 = linear(bias = N_0_norm1_fc_bias_to_fp16, weight = N_0_norm1_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_6_cast_fp16")]; + tensor var_334 = const()[name = tensor("op_334"), val = tensor([1, 1024, 1])]; + tensor h_27_cast_fp16 = reshape(shape = var_334, x = linear_6_cast_fp16)[name = tensor("h_27_cast_fp16")]; + tensor var_336_split_sizes_0 = const()[name = tensor("op_336_split_sizes_0"), val = tensor([512, 512])]; + tensor var_336_axis_0 = const()[name = tensor("op_336_axis_0"), val = tensor(1)]; + tensor var_336_cast_fp16_0, tensor var_336_cast_fp16_1 = split(axis = var_336_axis_0, split_sizes = var_336_split_sizes_0, x = h_27_cast_fp16)[name = tensor("op_336_cast_fp16")]; + tensor var_338_promoted_to_fp16 = const()[name = tensor("op_338_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_339_cast_fp16 = add(x = var_336_cast_fp16_0, y = var_338_promoted_to_fp16)[name = tensor("op_339_cast_fp16")]; + tensor var_341_cast_fp16 = mul(x = var_339_cast_fp16, y = var_84_cast_fp16)[name = tensor("op_341_cast_fp16")]; + tensor input_59_cast_fp16 = add(x = var_341_cast_fp16, y = var_336_cast_fp16_1)[name = tensor("input_59_cast_fp16")]; + tensor input_61_cast_fp16 = leaky_relu(alpha = var_309, x = input_59_cast_fp16)[name = tensor("input_61_cast_fp16")]; + tensor input_65_pad_type_0 = const()[name = tensor("input_65_pad_type_0"), val = tensor("custom")]; + tensor input_65_pad_0 = const()[name = tensor("input_65_pad_0"), val = tensor([1, 1])]; + tensor input_65_strides_0 = const()[name = tensor("input_65_strides_0"), val = tensor([1])]; + tensor input_65_dilations_0 = const()[name = tensor("input_65_dilations_0"), val = tensor([1])]; + tensor input_65_groups_0 = const()[name = tensor("input_65_groups_0"), val = tensor(1)]; + tensor weight_17_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5280768))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6067264))), name = tensor("weight_17_to_fp16_palettized"), shape = tensor([512, 512, 3])]; + tensor N_0_conv1_bias_to_fp16 = const()[name = tensor("N_0_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6067840)))]; + tensor input_65_cast_fp16 = conv(bias = N_0_conv1_bias_to_fp16, dilations = input_65_dilations_0, groups = input_65_groups_0, pad = input_65_pad_0, pad_type = input_65_pad_type_0, strides = input_65_strides_0, weight = weight_17_to_fp16_palettized, x = input_61_cast_fp16)[name = tensor("input_65_cast_fp16")]; + tensor N_0_norm2_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6068928))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6200064))), name = tensor("N_0_norm2_fc_weight_to_fp16_palettized"), shape = tensor([1024, 128])]; + tensor N_0_norm2_fc_bias_to_fp16 = const()[name = tensor("N_0_norm2_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6200640)))]; + tensor linear_7_cast_fp16 = linear(bias = N_0_norm2_fc_bias_to_fp16, weight = N_0_norm2_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_7_cast_fp16")]; + tensor var_358 = const()[name = tensor("op_358"), val = tensor([1, 1024, 1])]; + tensor h_31_cast_fp16 = reshape(shape = var_358, x = linear_7_cast_fp16)[name = tensor("h_31_cast_fp16")]; + tensor var_360_split_sizes_0 = const()[name = tensor("op_360_split_sizes_0"), val = tensor([512, 512])]; + tensor var_360_axis_0 = const()[name = tensor("op_360_axis_0"), val = tensor(1)]; + tensor var_360_cast_fp16_0, tensor var_360_cast_fp16_1 = split(axis = var_360_axis_0, split_sizes = var_360_split_sizes_0, x = h_31_cast_fp16)[name = tensor("op_360_cast_fp16")]; + tensor var_362_promoted_to_fp16 = const()[name = tensor("op_362_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_363_cast_fp16 = add(x = var_360_cast_fp16_0, y = var_362_promoted_to_fp16)[name = tensor("op_363_cast_fp16")]; + tensor var_312_to_fp16 = const()[name = tensor("op_312_to_fp16"), val = tensor(0x1.5p-17)]; + tensor var_364_cast_fp16 = instance_norm(epsilon = var_312_to_fp16, x = input_65_cast_fp16)[name = tensor("op_364_cast_fp16")]; + tensor var_365_cast_fp16 = mul(x = var_363_cast_fp16, y = var_364_cast_fp16)[name = tensor("op_365_cast_fp16")]; + tensor input_67_cast_fp16 = add(x = var_365_cast_fp16, y = var_360_cast_fp16_1)[name = tensor("input_67_cast_fp16")]; + tensor input_69_cast_fp16 = leaky_relu(alpha = var_309, x = input_67_cast_fp16)[name = tensor("input_69_cast_fp16")]; + tensor out_7_pad_type_0 = const()[name = tensor("out_7_pad_type_0"), val = tensor("custom")]; + tensor out_7_pad_0 = const()[name = tensor("out_7_pad_0"), val = tensor([1, 1])]; + tensor out_7_strides_0 = const()[name = tensor("out_7_strides_0"), val = tensor([1])]; + tensor out_7_dilations_0 = const()[name = tensor("out_7_dilations_0"), val = tensor([1])]; + tensor out_7_groups_0 = const()[name = tensor("out_7_groups_0"), val = tensor(1)]; + tensor weight_19_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6202752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6989248))), name = tensor("weight_19_to_fp16_palettized"), shape = tensor([512, 512, 3])]; + tensor N_0_conv2_bias_to_fp16 = const()[name = tensor("N_0_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6989824)))]; + tensor out_7_cast_fp16 = conv(bias = N_0_conv2_bias_to_fp16, dilations = out_7_dilations_0, groups = out_7_groups_0, pad = out_7_pad_0, pad_type = out_7_pad_type_0, strides = out_7_strides_0, weight = weight_19_to_fp16_palettized, x = input_69_cast_fp16)[name = tensor("out_7_cast_fp16")]; + tensor var_376_cast_fp16 = add(x = out_7_cast_fp16, y = transpose_0)[name = tensor("op_376_cast_fp16")]; + tensor _inversed_input_73_y_0_to_fp16 = const()[name = tensor("_inversed_input_73_y_0_to_fp16"), val = tensor(0x1.6ap-1)]; + tensor _inversed_input_73_cast_fp16 = mul(x = var_376_cast_fp16, y = _inversed_input_73_y_0_to_fp16)[name = tensor("_inversed_input_73_cast_fp16")]; + tensor var_382 = const()[name = tensor("op_382"), val = tensor(0x1.99999ap-3)]; + tensor N_1_norm1_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6990912))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7122048))), name = tensor("N_1_norm1_fc_weight_to_fp16_palettized"), shape = tensor([1024, 128])]; + tensor N_1_norm1_fc_bias_to_fp16 = const()[name = tensor("N_1_norm1_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7122624)))]; + tensor linear_8_cast_fp16 = linear(bias = N_1_norm1_fc_bias_to_fp16, weight = N_1_norm1_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_8_cast_fp16")]; + tensor var_417 = const()[name = tensor("op_417"), val = tensor([1, 1024, 1])]; + tensor h_35_cast_fp16 = reshape(shape = var_417, x = linear_8_cast_fp16)[name = tensor("h_35_cast_fp16")]; + tensor var_419_split_sizes_0 = const()[name = tensor("op_419_split_sizes_0"), val = tensor([512, 512])]; + tensor var_419_axis_0 = const()[name = tensor("op_419_axis_0"), val = tensor(1)]; + tensor var_419_cast_fp16_0, tensor var_419_cast_fp16_1 = split(axis = var_419_axis_0, split_sizes = var_419_split_sizes_0, x = h_35_cast_fp16)[name = tensor("op_419_cast_fp16")]; + tensor var_421_promoted_to_fp16 = const()[name = tensor("op_421_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_422_cast_fp16 = add(x = var_419_cast_fp16_0, y = var_421_promoted_to_fp16)[name = tensor("op_422_cast_fp16")]; + tensor var_386_to_fp16 = const()[name = tensor("op_386_to_fp16"), val = tensor(0x1.5p-17)]; + tensor var_423_cast_fp16 = instance_norm(epsilon = var_386_to_fp16, x = _inversed_input_73_cast_fp16)[name = tensor("op_423_cast_fp16")]; + tensor var_424_cast_fp16 = mul(x = var_422_cast_fp16, y = var_423_cast_fp16)[name = tensor("op_424_cast_fp16")]; + tensor input_75_cast_fp16 = add(x = var_424_cast_fp16, y = var_419_cast_fp16_1)[name = tensor("input_75_cast_fp16")]; + tensor input_77_cast_fp16 = leaky_relu(alpha = var_382, x = input_75_cast_fp16)[name = tensor("input_77_cast_fp16")]; + tensor conv_transpose_1_pad_type_0 = const()[name = tensor("conv_transpose_1_pad_type_0"), val = tensor("custom")]; + tensor conv_transpose_1_pad_0 = const()[name = tensor("conv_transpose_1_pad_0"), val = tensor([0, 0])]; + tensor conv_transpose_1_strides_0 = const()[name = tensor("conv_transpose_1_strides_0"), val = tensor([2])]; + tensor conv_transpose_1_groups_0 = const()[name = tensor("conv_transpose_1_groups_0"), val = tensor(512)]; + tensor conv_transpose_1_dilations_0 = const()[name = tensor("conv_transpose_1_dilations_0"), val = tensor([1])]; + tensor conv_transpose_1_has_output_shape_output_shape_0 = const()[name = tensor("conv_transpose_1_has_output_shape_output_shape_0"), val = tensor([1, 512, 4001])]; + tensor var_427_to_fp16 = const()[name = tensor("op_427_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7124736)))]; + tensor N_1_pool_bias_to_fp16 = const()[name = tensor("N_1_pool_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7127872)))]; + tensor conv_transpose_1_has_output_shape_cast_fp16 = conv_transpose(bias = N_1_pool_bias_to_fp16, dilations = conv_transpose_1_dilations_0, groups = conv_transpose_1_groups_0, output_shape = conv_transpose_1_has_output_shape_output_shape_0, pad = conv_transpose_1_pad_0, pad_type = conv_transpose_1_pad_type_0, strides = conv_transpose_1_strides_0, weight = var_427_to_fp16, x = input_77_cast_fp16)[name = tensor("conv_transpose_1_has_output_shape_cast_fp16")]; + tensor input_79_begin_0 = const()[name = tensor("input_79_begin_0"), val = tensor([0, 0, 1])]; + tensor input_79_end_0 = const()[name = tensor("input_79_end_0"), val = tensor([0, 0, 0])]; + tensor input_79_begin_mask_0 = const()[name = tensor("input_79_begin_mask_0"), val = tensor([true, true, false])]; + tensor input_79_end_mask_0 = const()[name = tensor("input_79_end_mask_0"), val = tensor([true, true, true])]; + tensor input_79_cast_fp16 = slice_by_index(begin = input_79_begin_0, begin_mask = input_79_begin_mask_0, end = input_79_end_0, end_mask = input_79_end_mask_0, x = conv_transpose_1_has_output_shape_cast_fp16)[name = tensor("input_79_cast_fp16")]; + tensor input_83_pad_type_0 = const()[name = tensor("input_83_pad_type_0"), val = tensor("custom")]; + tensor input_83_pad_0 = const()[name = tensor("input_83_pad_0"), val = tensor([1, 1])]; + tensor input_83_strides_0 = const()[name = tensor("input_83_strides_0"), val = tensor([1])]; + tensor input_83_dilations_0 = const()[name = tensor("input_83_dilations_0"), val = tensor([1])]; + tensor input_83_groups_0 = const()[name = tensor("input_83_groups_0"), val = tensor(1)]; + tensor weight_21_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7128960))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7522240))), name = tensor("weight_21_to_fp16_palettized"), shape = tensor([256, 512, 3])]; + tensor N_1_conv1_bias_to_fp16 = const()[name = tensor("N_1_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7522816)))]; + tensor input_83_cast_fp16 = conv(bias = N_1_conv1_bias_to_fp16, dilations = input_83_dilations_0, groups = input_83_groups_0, pad = input_83_pad_0, pad_type = input_83_pad_type_0, strides = input_83_strides_0, weight = weight_21_to_fp16_palettized, x = input_79_cast_fp16)[name = tensor("input_83_cast_fp16")]; + tensor N_1_norm2_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7523392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7588992))), name = tensor("N_1_norm2_fc_weight_to_fp16_palettized"), shape = tensor([512, 128])]; + tensor N_1_norm2_fc_bias_to_fp16 = const()[name = tensor("N_1_norm2_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7589568)))]; + tensor linear_9_cast_fp16 = linear(bias = N_1_norm2_fc_bias_to_fp16, weight = N_1_norm2_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_9_cast_fp16")]; + tensor var_448 = const()[name = tensor("op_448"), val = tensor([1, 512, 1])]; + tensor h_39_cast_fp16 = reshape(shape = var_448, x = linear_9_cast_fp16)[name = tensor("h_39_cast_fp16")]; + tensor var_450_split_sizes_0 = const()[name = tensor("op_450_split_sizes_0"), val = tensor([256, 256])]; + tensor var_450_axis_0 = const()[name = tensor("op_450_axis_0"), val = tensor(1)]; + tensor var_450_cast_fp16_0, tensor var_450_cast_fp16_1 = split(axis = var_450_axis_0, split_sizes = var_450_split_sizes_0, x = h_39_cast_fp16)[name = tensor("op_450_cast_fp16")]; + tensor var_452_promoted_to_fp16 = const()[name = tensor("op_452_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_453_cast_fp16 = add(x = var_450_cast_fp16_0, y = var_452_promoted_to_fp16)[name = tensor("op_453_cast_fp16")]; + tensor var_454_cast_fp16 = instance_norm(epsilon = var_386_to_fp16, x = input_83_cast_fp16)[name = tensor("op_454_cast_fp16")]; + tensor var_455_cast_fp16 = mul(x = var_453_cast_fp16, y = var_454_cast_fp16)[name = tensor("op_455_cast_fp16")]; + tensor input_85_cast_fp16 = add(x = var_455_cast_fp16, y = var_450_cast_fp16_1)[name = tensor("input_85_cast_fp16")]; + tensor input_87_cast_fp16 = leaky_relu(alpha = var_382, x = input_85_cast_fp16)[name = tensor("input_87_cast_fp16")]; + tensor out_9_pad_type_0 = const()[name = tensor("out_9_pad_type_0"), val = tensor("custom")]; + tensor out_9_pad_0 = const()[name = tensor("out_9_pad_0"), val = tensor([1, 1])]; + tensor out_9_strides_0 = const()[name = tensor("out_9_strides_0"), val = tensor([1])]; + tensor out_9_dilations_0 = const()[name = tensor("out_9_dilations_0"), val = tensor([1])]; + tensor out_9_groups_0 = const()[name = tensor("out_9_groups_0"), val = tensor(1)]; + tensor weight_23_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7590656))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7787328))), name = tensor("weight_23_to_fp16_palettized"), shape = tensor([256, 256, 3])]; + tensor N_1_conv2_bias_to_fp16 = const()[name = tensor("N_1_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7787904)))]; + tensor out_9_cast_fp16 = conv(bias = N_1_conv2_bias_to_fp16, dilations = out_9_dilations_0, groups = out_9_groups_0, pad = out_9_pad_0, pad_type = out_9_pad_type_0, strides = out_9_strides_0, weight = weight_23_to_fp16_palettized, x = input_87_cast_fp16)[name = tensor("out_9_cast_fp16")]; + tensor expand_dims_1_axes_0 = const()[name = tensor("expand_dims_1_axes_0"), val = tensor([3])]; + tensor expand_dims_1_cast_fp16 = expand_dims(axes = expand_dims_1_axes_0, x = _inversed_input_73_cast_fp16)[name = tensor("expand_dims_1_cast_fp16")]; + tensor upsample_nearest_neighbor_1_scale_factor_height_0 = const()[name = tensor("upsample_nearest_neighbor_1_scale_factor_height_0"), val = tensor(2)]; + tensor upsample_nearest_neighbor_1_scale_factor_width_0 = const()[name = tensor("upsample_nearest_neighbor_1_scale_factor_width_0"), val = tensor(1)]; + tensor upsample_nearest_neighbor_1_cast_fp16 = upsample_nearest_neighbor(scale_factor_height = upsample_nearest_neighbor_1_scale_factor_height_0, scale_factor_width = upsample_nearest_neighbor_1_scale_factor_width_0, x = expand_dims_1_cast_fp16)[name = tensor("upsample_nearest_neighbor_1_cast_fp16")]; + tensor input_91_axes_0 = const()[name = tensor("input_91_axes_0"), val = tensor([3])]; + tensor input_91_cast_fp16 = squeeze(axes = input_91_axes_0, x = upsample_nearest_neighbor_1_cast_fp16)[name = tensor("input_91_cast_fp16")]; + tensor var_473_pad_type_0 = const()[name = tensor("op_473_pad_type_0"), val = tensor("valid")]; + tensor var_473_strides_0 = const()[name = tensor("op_473_strides_0"), val = tensor([1])]; + tensor var_473_pad_0 = const()[name = tensor("op_473_pad_0"), val = tensor([0, 0])]; + tensor var_473_dilations_0 = const()[name = tensor("op_473_dilations_0"), val = tensor([1])]; + tensor var_473_groups_0 = const()[name = tensor("op_473_groups_0"), val = tensor(1)]; + tensor weight_25_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7788480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7919616))), name = tensor("weight_25_to_fp16_palettized"), shape = tensor([256, 512, 1])]; + tensor var_473_cast_fp16 = conv(dilations = var_473_dilations_0, groups = var_473_groups_0, pad = var_473_pad_0, pad_type = var_473_pad_type_0, strides = var_473_strides_0, weight = weight_25_to_fp16_palettized, x = input_91_cast_fp16)[name = tensor("op_473_cast_fp16")]; + tensor var_474_cast_fp16 = add(x = out_9_cast_fp16, y = var_473_cast_fp16)[name = tensor("op_474_cast_fp16")]; + tensor _inversed_input_93_y_0_to_fp16 = const()[name = tensor("_inversed_input_93_y_0_to_fp16"), val = tensor(0x1.6ap-1)]; + tensor _inversed_input_93_cast_fp16 = mul(x = var_474_cast_fp16, y = _inversed_input_93_y_0_to_fp16)[name = tensor("_inversed_input_93_cast_fp16")]; + tensor var_478 = const()[name = tensor("op_478"), val = tensor(0x1.99999ap-3)]; + tensor N_2_norm1_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7920192))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7985792))), name = tensor("N_2_norm1_fc_weight_to_fp16_palettized"), shape = tensor([512, 128])]; + tensor N_2_norm1_fc_bias_to_fp16 = const()[name = tensor("N_2_norm1_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7986368)))]; + tensor linear_10_cast_fp16 = linear(bias = N_2_norm1_fc_bias_to_fp16, weight = N_2_norm1_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_10_cast_fp16")]; + tensor var_503 = const()[name = tensor("op_503"), val = tensor([1, 512, 1])]; + tensor h_43_cast_fp16 = reshape(shape = var_503, x = linear_10_cast_fp16)[name = tensor("h_43_cast_fp16")]; + tensor var_505_split_sizes_0 = const()[name = tensor("op_505_split_sizes_0"), val = tensor([256, 256])]; + tensor var_505_axis_0 = const()[name = tensor("op_505_axis_0"), val = tensor(1)]; + tensor var_505_cast_fp16_0, tensor var_505_cast_fp16_1 = split(axis = var_505_axis_0, split_sizes = var_505_split_sizes_0, x = h_43_cast_fp16)[name = tensor("op_505_cast_fp16")]; + tensor var_507_promoted_to_fp16 = const()[name = tensor("op_507_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_508_cast_fp16 = add(x = var_505_cast_fp16_0, y = var_507_promoted_to_fp16)[name = tensor("op_508_cast_fp16")]; + tensor var_481_to_fp16 = const()[name = tensor("op_481_to_fp16"), val = tensor(0x1.5p-17)]; + tensor var_509_cast_fp16 = instance_norm(epsilon = var_481_to_fp16, x = _inversed_input_93_cast_fp16)[name = tensor("op_509_cast_fp16")]; + tensor var_510_cast_fp16 = mul(x = var_508_cast_fp16, y = var_509_cast_fp16)[name = tensor("op_510_cast_fp16")]; + tensor input_95_cast_fp16 = add(x = var_510_cast_fp16, y = var_505_cast_fp16_1)[name = tensor("input_95_cast_fp16")]; + tensor input_97_cast_fp16 = leaky_relu(alpha = var_478, x = input_95_cast_fp16)[name = tensor("input_97_cast_fp16")]; + tensor input_101_pad_type_0 = const()[name = tensor("input_101_pad_type_0"), val = tensor("custom")]; + tensor input_101_pad_0 = const()[name = tensor("input_101_pad_0"), val = tensor([1, 1])]; + tensor input_101_strides_0 = const()[name = tensor("input_101_strides_0"), val = tensor([1])]; + tensor input_101_dilations_0 = const()[name = tensor("input_101_dilations_0"), val = tensor([1])]; + tensor input_101_groups_0 = const()[name = tensor("input_101_groups_0"), val = tensor(1)]; + tensor weight_27_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7987456))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8184128))), name = tensor("weight_27_to_fp16_palettized"), shape = tensor([256, 256, 3])]; + tensor N_2_conv1_bias_to_fp16 = const()[name = tensor("N_2_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8184704)))]; + tensor input_101_cast_fp16 = conv(bias = N_2_conv1_bias_to_fp16, dilations = input_101_dilations_0, groups = input_101_groups_0, pad = input_101_pad_0, pad_type = input_101_pad_type_0, strides = input_101_strides_0, weight = weight_27_to_fp16_palettized, x = input_97_cast_fp16)[name = tensor("input_101_cast_fp16")]; + tensor N_2_norm2_fc_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8185280))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8250880))), name = tensor("N_2_norm2_fc_weight_to_fp16_palettized"), shape = tensor([512, 128])]; + tensor N_2_norm2_fc_bias_to_fp16 = const()[name = tensor("N_2_norm2_fc_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8251456)))]; + tensor linear_11_cast_fp16 = linear(bias = N_2_norm2_fc_bias_to_fp16, weight = N_2_norm2_fc_weight_to_fp16_palettized, x = s_to_fp16)[name = tensor("linear_11_cast_fp16")]; + tensor var_527 = const()[name = tensor("op_527"), val = tensor([1, 512, 1])]; + tensor h_cast_fp16 = reshape(shape = var_527, x = linear_11_cast_fp16)[name = tensor("h_cast_fp16")]; + tensor var_529_split_sizes_0 = const()[name = tensor("op_529_split_sizes_0"), val = tensor([256, 256])]; + tensor var_529_axis_0 = const()[name = tensor("op_529_axis_0"), val = tensor(1)]; + tensor var_529_cast_fp16_0, tensor var_529_cast_fp16_1 = split(axis = var_529_axis_0, split_sizes = var_529_split_sizes_0, x = h_cast_fp16)[name = tensor("op_529_cast_fp16")]; + tensor var_531_promoted_to_fp16 = const()[name = tensor("op_531_promoted_to_fp16"), val = tensor(0x1p+0)]; + tensor var_532_cast_fp16 = add(x = var_529_cast_fp16_0, y = var_531_promoted_to_fp16)[name = tensor("op_532_cast_fp16")]; + tensor var_533_cast_fp16 = instance_norm(epsilon = var_481_to_fp16, x = input_101_cast_fp16)[name = tensor("op_533_cast_fp16")]; + tensor var_534_cast_fp16 = mul(x = var_532_cast_fp16, y = var_533_cast_fp16)[name = tensor("op_534_cast_fp16")]; + tensor input_103_cast_fp16 = add(x = var_534_cast_fp16, y = var_529_cast_fp16_1)[name = tensor("input_103_cast_fp16")]; + tensor input_105_cast_fp16 = leaky_relu(alpha = var_478, x = input_103_cast_fp16)[name = tensor("input_105_cast_fp16")]; + tensor out_pad_type_0 = const()[name = tensor("out_pad_type_0"), val = tensor("custom")]; + tensor out_pad_0 = const()[name = tensor("out_pad_0"), val = tensor([1, 1])]; + tensor out_strides_0 = const()[name = tensor("out_strides_0"), val = tensor([1])]; + tensor out_dilations_0 = const()[name = tensor("out_dilations_0"), val = tensor([1])]; + tensor out_groups_0 = const()[name = tensor("out_groups_0"), val = tensor(1)]; + tensor weight_29_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8252544))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8449216))), name = tensor("weight_29_to_fp16_palettized"), shape = tensor([256, 256, 3])]; + tensor N_2_conv2_bias_to_fp16 = const()[name = tensor("N_2_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8449792)))]; + tensor out_cast_fp16 = conv(bias = N_2_conv2_bias_to_fp16, dilations = out_dilations_0, groups = out_groups_0, pad = out_pad_0, pad_type = out_pad_type_0, strides = out_strides_0, weight = weight_29_to_fp16_palettized, x = input_105_cast_fp16)[name = tensor("out_cast_fp16")]; + tensor var_545_cast_fp16 = add(x = out_cast_fp16, y = _inversed_input_93_cast_fp16)[name = tensor("op_545_cast_fp16")]; + tensor _inversed_input_y_0_to_fp16 = const()[name = tensor("_inversed_input_y_0_to_fp16"), val = tensor(0x1.6ap-1)]; + tensor _inversed_input_cast_fp16 = mul(x = var_545_cast_fp16, y = _inversed_input_y_0_to_fp16)[name = tensor("_inversed_input_cast_fp16")]; + tensor var_558_pad_type_0 = const()[name = tensor("op_558_pad_type_0"), val = tensor("valid")]; + tensor var_558_strides_0 = const()[name = tensor("op_558_strides_0"), val = tensor([1])]; + tensor var_558_pad_0 = const()[name = tensor("op_558_pad_0"), val = tensor([0, 0])]; + tensor var_558_dilations_0 = const()[name = tensor("op_558_dilations_0"), val = tensor([1])]; + tensor var_558_groups_0 = const()[name = tensor("op_558_groups_0"), val = tensor(1)]; + tensor N_proj_weight_to_fp16 = const()[name = tensor("N_proj_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8450368)))]; + tensor N_proj_bias_to_fp16 = const()[name = tensor("N_proj_bias_to_fp16"), val = tensor([0x1.6ep-4])]; + tensor var_558_cast_fp16 = conv(bias = N_proj_bias_to_fp16, dilations = var_558_dilations_0, groups = var_558_groups_0, pad = var_558_pad_0, pad_type = var_558_pad_type_0, strides = var_558_strides_0, weight = N_proj_weight_to_fp16, x = _inversed_input_cast_fp16)[name = tensor("op_558_cast_fp16")]; + tensor var_560_axes_0 = const()[name = tensor("op_560_axes_0"), val = tensor([1])]; + tensor N = squeeze(axes = var_560_axes_0, x = var_558_cast_fp16)[name = tensor("op_560_cast_fp16")]; + } -> (F0, N); +} \ No newline at end of file diff --git a/ANE/styletts2_ane_prosody.mlmodelc/weights/weight.bin b/ANE/styletts2_ane_prosody.mlmodelc/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..b2b14b8f90bb987917197519514a2fead5f4dc84 --- /dev/null +++ b/ANE/styletts2_ane_prosody.mlmodelc/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4242f6c193f8d1c995e9da98e63cda3f9559daaa9154be4d7a325c976d347ad8 +size 8450944 diff --git a/ANE/styletts2_ane_prosody.mlpackage/Data/com.apple.CoreML/model.mlmodel b/ANE/styletts2_ane_prosody.mlpackage/Data/com.apple.CoreML/model.mlmodel new file mode 100644 index 0000000000000000000000000000000000000000..8bd5c2994d6c79a732c7bfc602c04e16719f821b --- /dev/null +++ b/ANE/styletts2_ane_prosody.mlpackage/Data/com.apple.CoreML/model.mlmodel @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c8bb395a1d0b34fceb6da4e42c4cd201fe6c4121f788475940abaeb7ccd9e6f +size 65948 diff --git a/ANE/styletts2_ane_prosody.mlpackage/Data/com.apple.CoreML/weights/weight.bin b/ANE/styletts2_ane_prosody.mlpackage/Data/com.apple.CoreML/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..b2b14b8f90bb987917197519514a2fead5f4dc84 --- /dev/null +++ b/ANE/styletts2_ane_prosody.mlpackage/Data/com.apple.CoreML/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4242f6c193f8d1c995e9da98e63cda3f9559daaa9154be4d7a325c976d347ad8 +size 8450944 diff --git a/ANE/styletts2_ane_prosody.mlpackage/Manifest.json b/ANE/styletts2_ane_prosody.mlpackage/Manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..ba161dae4e5e8d85e1f4334b29eb8f419f6c25b8 --- /dev/null +++ b/ANE/styletts2_ane_prosody.mlpackage/Manifest.json @@ -0,0 +1,18 @@ +{ + "fileFormatVersion": "1.0.0", + "itemInfoEntries": { + "5239C997-4433-48DE-B7F7-8F278231936E": { + "author": "com.apple.CoreML", + "description": "CoreML Model Weights", + "name": "weights", + "path": "com.apple.CoreML/weights" + }, + "EAD2ABD5-31E9-4A23-A7AB-80FE05510307": { + "author": "com.apple.CoreML", + "description": "CoreML Model Specification", + "name": "model.mlmodel", + "path": "com.apple.CoreML/model.mlmodel" + } + }, + "rootModelIdentifier": "EAD2ABD5-31E9-4A23-A7AB-80FE05510307" +} diff --git a/ANE/styletts2_ane_vocoder.mlmodelc/analytics/coremldata.bin b/ANE/styletts2_ane_vocoder.mlmodelc/analytics/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..cbb424da59a3f1a68bd8d1469d80776e2ce244a9 --- /dev/null +++ b/ANE/styletts2_ane_vocoder.mlmodelc/analytics/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e00e88bd1cdc41d123c27fe2f02f15f06711c8f6a271a0e6f458a482f066a233 +size 243 diff --git a/ANE/styletts2_ane_vocoder.mlmodelc/coremldata.bin b/ANE/styletts2_ane_vocoder.mlmodelc/coremldata.bin new file mode 100644 index 0000000000000000000000000000000000000000..5abeb989ee5b56c02e4dddc6d0c50d2ae5ae478e --- /dev/null +++ b/ANE/styletts2_ane_vocoder.mlmodelc/coremldata.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b5bd3b5278ffc740bb257d2e837104d36fdc269adac04b7e9ac374676f923f51 +size 579 diff --git a/ANE/styletts2_ane_vocoder.mlmodelc/metadata.json b/ANE/styletts2_ane_vocoder.mlmodelc/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..c218ee81eea804529c48402b44296c86d5bc41cc --- /dev/null +++ b/ANE/styletts2_ane_vocoder.mlmodelc/metadata.json @@ -0,0 +1,127 @@ +[ + { + "metadataOutputVersion" : "3.0", + "shortDescription" : "StyleTTS2-ANE Stage 7: HiFi-GAN body (cos-Snake, RangeDim T_a∈[2,2000] (default=120)) — fp16, ANE.", + "outputSchema" : [ + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float32", + "formattedType" : "MultiArray (Float32)", + "shortDescription" : "", + "shape" : "[]", + "name" : "audio", + "type" : "MultiArray" + } + ], + "storagePrecision" : "Mixed (Float32, Palettized (8 bits))", + "modelParameters" : [ + + ], + "specificationVersion" : 8, + "mlProgramOperationTypeHistogram" : { + "Ios17.mul" : 418, + "Ios17.linear" : 107, + "Ios17.transpose" : 1, + "Ios17.cos" : 96, + "Ios16.constexprLutToDense" : 224, + "Ios17.conv" : 119, + "Ios17.sin" : 5, + "Ios17.leakyRelu" : 10, + "Ios17.concat" : 5, + "Ios17.add" : 474, + "Ios17.convTranspose" : 5, + "Ios17.sliceByIndex" : 3, + "UpsampleNearestNeighbor" : 1, + "Ios17.expandDims" : 3, + "Ios17.tanh" : 2, + "Ios17.instanceNorm" : 98, + "Ios17.reshape" : 106, + "Split" : 106, + "Ios17.squeeze" : 2 + }, + "computePrecision" : "Mixed (Float32, Int32)", + "isUpdatable" : "0", + "stateSchema" : [ + + ], + "availability" : { + "macOS" : "14.0", + "tvOS" : "17.0", + "visionOS" : "1.0", + "watchOS" : "10.0", + "iOS" : "17.0", + "macCatalyst" : "17.0" + }, + "modelType" : { + "name" : "MLModelType_mlProgram" + }, + "inputSchema" : [ + { + "dataType" : "Float32", + "hasShapeFlexibility" : "1", + "isOptional" : "0", + "shapeFlexibility" : "1 × 512 × 2...2000", + "shapeRange" : "[[1, 1], [512, 512], [2, 2000]]", + "formattedType" : "MultiArray (Float32 1 × 512 × 120)", + "type" : "MultiArray", + "shape" : "[1, 512, 120]", + "name" : "asr", + "shortDescription" : "" + }, + { + "dataType" : "Float32", + "hasShapeFlexibility" : "1", + "isOptional" : "0", + "shapeFlexibility" : "1 × 4...4000", + "shapeRange" : "[[1, 1], [4, 4000]]", + "formattedType" : "MultiArray (Float32 1 × 240)", + "type" : "MultiArray", + "shape" : "[1, 240]", + "name" : "F0_curve", + "shortDescription" : "" + }, + { + "dataType" : "Float32", + "hasShapeFlexibility" : "1", + "isOptional" : "0", + "shapeFlexibility" : "1 × 4...4000", + "shapeRange" : "[[1, 1], [4, 4000]]", + "formattedType" : "MultiArray (Float32 1 × 240)", + "type" : "MultiArray", + "shape" : "[1, 240]", + "name" : "N", + "shortDescription" : "" + }, + { + "hasShapeFlexibility" : "0", + "isOptional" : "0", + "dataType" : "Float32", + "formattedType" : "MultiArray (Float32 1 × 128)", + "shortDescription" : "", + "shape" : "[1, 128]", + "name" : "s", + "type" : "MultiArray" + }, + { + "dataType" : "Float32", + "hasShapeFlexibility" : "1", + "isOptional" : "0", + "shapeFlexibility" : "1 × 1200...1200000 × 9", + "shapeRange" : "[[1, 1], [1200, 1200000], [9, 9]]", + "formattedType" : "MultiArray (Float32 1 × 72000 × 9)", + "type" : "MultiArray", + "shape" : "[1, 72000, 9]", + "name" : "sine_waves", + "shortDescription" : "" + } + ], + "userDefinedMetadata" : { + "com.github.apple.coremltools.source_dialect" : "TorchScript", + "com.github.apple.coremltools.source" : "torch==2.11.0", + "com.github.apple.coremltools.version" : "8.3.0" + }, + "generatedClassName" : "styletts2_ane_vocoder", + "method" : "predict" + } +] \ No newline at end of file diff --git a/ANE/styletts2_ane_vocoder.mlmodelc/model.mil b/ANE/styletts2_ane_vocoder.mlmodelc/model.mil new file mode 100644 index 0000000000000000000000000000000000000000..ef082cdc2bececb7ecf66459c6c82d4d1f1fca50 --- /dev/null +++ b/ANE/styletts2_ane_vocoder.mlmodelc/model.mil @@ -0,0 +1,3527 @@ +program(1.0) +[buildInfo = dict, tensor>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}})] +{ + func main(tensor F0_curve, tensor N, tensor asr, tensor s, tensor sine_waves) [FlexibleShapeInformation = tuple, dict, tensor>>, tuple, dict, list, ?>>>>((("DefaultShapes", {{"F0_curve", [1, 240]}, {"N", [1, 240]}, {"asr", [1, 512, 120]}, {"sine_waves", [1, 72000, 9]}}), ("RangeDims", {{"F0_curve", [[1, 1], [4, 4000]]}, {"N", [[1, 1], [4, 4000]]}, {"asr", [[1, 1], [512, 512], [2, 2000]]}, {"sine_waves", [[1, 1], [1200, 1200000], [9, 9]]}})))] { + tensor generator_alphas_4 = const()[name = tensor("generator_alphas_4"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(64)))]; + tensor generator_alphas_3 = const()[name = tensor("generator_alphas_3"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(256)))]; + tensor generator_alphas_2 = const()[name = tensor("generator_alphas_2"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(576)))]; + tensor generator_alphas_1 = const()[name = tensor("generator_alphas_1"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1152)))]; + tensor generator_alphas_0 = const()[name = tensor("generator_alphas_0"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2240)))]; + tensor F0_conv_bias = const()[name = tensor("F0_conv_bias"), val = tensor([0x1.8443d2p-2])]; + tensor N_conv_bias = const()[name = tensor("N_conv_bias"), val = tensor([0x1.44d67ep-1])]; + tensor encode_norm1_fc_bias = const()[name = tensor("encode_norm1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4352)))]; + tensor encode_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8576))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(140224))), name = tensor("encode_norm1_fc_weight_palettized"), shape = tensor([1028, 128])]; + tensor encode_conv1_bias = const()[name = tensor("encode_conv1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(141312)))]; + tensor encode_norm2_fc_bias = const()[name = tensor("encode_norm2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(145472)))]; + tensor encode_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(153728))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(415936))), name = tensor("encode_norm2_fc_weight_palettized"), shape = tensor([2048, 128])]; + tensor encode_conv2_bias = const()[name = tensor("encode_conv2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(417024)))]; + tensor asr_res_0_bias = const()[name = tensor("asr_res_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(421184)))]; + tensor decode_0_norm1_fc_bias = const()[name = tensor("decode_0_norm1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(421504)))]; + tensor decode_0_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(430336))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(709440))), name = tensor("decode_0_norm1_fc_weight_palettized"), shape = tensor([2180, 128])]; + tensor decode_0_conv1_bias = const()[name = tensor("decode_0_conv1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(710528)))]; + tensor decode_0_norm2_fc_bias = const()[name = tensor("decode_0_norm2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(714688)))]; + tensor decode_0_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(722944))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(985152))), name = tensor("decode_0_norm2_fc_weight_palettized"), shape = tensor([2048, 128])]; + tensor decode_0_conv2_bias = const()[name = tensor("decode_0_conv2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(986240)))]; + tensor decode_1_norm1_fc_bias = const()[name = tensor("decode_1_norm1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(990400)))]; + tensor decode_1_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(999232))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1278336))), name = tensor("decode_1_norm1_fc_weight_palettized"), shape = tensor([2180, 128])]; + tensor decode_1_conv1_bias = const()[name = tensor("decode_1_conv1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1279424)))]; + tensor decode_1_norm2_fc_bias = const()[name = tensor("decode_1_norm2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1283584)))]; + tensor decode_1_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1291840))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1554048))), name = tensor("decode_1_norm2_fc_weight_palettized"), shape = tensor([2048, 128])]; + tensor decode_1_conv2_bias = const()[name = tensor("decode_1_conv2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1555136)))]; + tensor decode_2_norm1_fc_bias = const()[name = tensor("decode_2_norm1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1559296)))]; + tensor decode_2_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1568128))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1847232))), name = tensor("decode_2_norm1_fc_weight_palettized"), shape = tensor([2180, 128])]; + tensor decode_2_conv1_bias = const()[name = tensor("decode_2_conv1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1848320)))]; + tensor decode_2_norm2_fc_bias = const()[name = tensor("decode_2_norm2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1852480)))]; + tensor decode_2_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1860736))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2122944))), name = tensor("decode_2_norm2_fc_weight_palettized"), shape = tensor([2048, 128])]; + tensor decode_2_conv2_bias = const()[name = tensor("decode_2_conv2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2124032)))]; + tensor decode_3_norm1_fc_bias = const()[name = tensor("decode_3_norm1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2128192)))]; + tensor decode_3_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2137024))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2416128))), name = tensor("decode_3_norm1_fc_weight_palettized"), shape = tensor([2180, 128])]; + tensor decode_3_pool_bias = const()[name = tensor("decode_3_pool_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2417216)))]; + tensor decode_3_conv1_bias = const()[name = tensor("decode_3_conv1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2421696)))]; + tensor decode_3_norm2_fc_bias = const()[name = tensor("decode_3_norm2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2423808)))]; + tensor decode_3_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2427968))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2559104))), name = tensor("decode_3_norm2_fc_weight_palettized"), shape = tensor([1024, 128])]; + tensor decode_3_conv2_bias = const()[name = tensor("decode_3_conv2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2560192)))]; + tensor generator_m_source_l_linear_bias = const()[name = tensor("generator_m_source_l_linear_bias"), val = tensor([0x1.23e7f2p-6])]; + tensor generator_m_source_l_linear_weight = const()[name = tensor("generator_m_source_l_linear_weight"), val = tensor([[-0x1.2aaee4p-12, 0x1.3872d6p-4, -0x1.e6ccbep-7, 0x1.e1debcp-8, -0x1.6ca714p-10, 0x1.8ba75p-10, 0x1.b33eecp-9, -0x1.cd2accp-8, 0x1.07f694p-7]])]; + tensor generator_noise_convs_0_bias = const()[name = tensor("generator_noise_convs_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2562304)))]; + tensor generator_noise_convs_0_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2563392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2578816))), name = tensor("generator_noise_convs_0_weight_palettized"), shape = tensor([256, 1, 60])]; + tensor generator_noise_res_0_adain1_0_fc_bias = const()[name = tensor("generator_noise_res_0_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2579904)))]; + tensor generator_noise_res_0_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2582016))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2647616))), name = tensor("generator_noise_res_0_adain1_0_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_noise_res_0_convs1_0_bias = const()[name = tensor("generator_noise_res_0_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2648704)))]; + tensor generator_noise_res_0_adain2_0_fc_bias = const()[name = tensor("generator_noise_res_0_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2649792)))]; + tensor generator_noise_res_0_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2651904))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2717504))), name = tensor("generator_noise_res_0_adain2_0_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_noise_res_0_convs2_0_bias = const()[name = tensor("generator_noise_res_0_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2718592)))]; + tensor generator_noise_res_0_adain1_1_fc_bias = const()[name = tensor("generator_noise_res_0_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2719680)))]; + tensor generator_noise_res_0_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2721792))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2787392))), name = tensor("generator_noise_res_0_adain1_1_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_noise_res_0_convs1_1_bias = const()[name = tensor("generator_noise_res_0_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2788480)))]; + tensor generator_noise_res_0_adain2_1_fc_bias = const()[name = tensor("generator_noise_res_0_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2789568)))]; + tensor generator_noise_res_0_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2791680))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2857280))), name = tensor("generator_noise_res_0_adain2_1_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_noise_res_0_convs2_1_bias = const()[name = tensor("generator_noise_res_0_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2858368)))]; + tensor generator_noise_res_0_adain1_2_fc_bias = const()[name = tensor("generator_noise_res_0_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2859456)))]; + tensor generator_noise_res_0_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2861568))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2927168))), name = tensor("generator_noise_res_0_adain1_2_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_noise_res_0_convs1_2_bias = const()[name = tensor("generator_noise_res_0_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2928256)))]; + tensor generator_noise_res_0_adain2_2_fc_bias = const()[name = tensor("generator_noise_res_0_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2929344)))]; + tensor generator_noise_res_0_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2931456))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2997056))), name = tensor("generator_noise_res_0_adain2_2_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_noise_res_0_convs2_2_bias = const()[name = tensor("generator_noise_res_0_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2998144)))]; + tensor generator_ups_0_bias = const()[name = tensor("generator_ups_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(2999232)))]; + tensor generator_resblocks_0_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_0_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3000320)))]; + tensor generator_resblocks_0_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3002432))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3068032))), name = tensor("generator_resblocks_0_adain1_0_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_0_convs1_0_bias = const()[name = tensor("generator_resblocks_0_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3069120)))]; + tensor generator_resblocks_0_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_0_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3070208)))]; + tensor generator_resblocks_0_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3072320))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3137920))), name = tensor("generator_resblocks_0_adain2_0_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_0_convs2_0_bias = const()[name = tensor("generator_resblocks_0_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3139008)))]; + tensor generator_resblocks_0_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_0_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3140096)))]; + tensor generator_resblocks_0_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3142208))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3207808))), name = tensor("generator_resblocks_0_adain1_1_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_0_convs1_1_bias = const()[name = tensor("generator_resblocks_0_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3208896)))]; + tensor generator_resblocks_0_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_0_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3209984)))]; + tensor generator_resblocks_0_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3212096))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3277696))), name = tensor("generator_resblocks_0_adain2_1_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_0_convs2_1_bias = const()[name = tensor("generator_resblocks_0_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3278784)))]; + tensor generator_resblocks_0_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_0_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3279872)))]; + tensor generator_resblocks_0_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3281984))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3347584))), name = tensor("generator_resblocks_0_adain1_2_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_0_convs1_2_bias = const()[name = tensor("generator_resblocks_0_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3348672)))]; + tensor generator_resblocks_0_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_0_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3349760)))]; + tensor generator_resblocks_0_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3351872))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3417472))), name = tensor("generator_resblocks_0_adain2_2_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_0_convs2_2_bias = const()[name = tensor("generator_resblocks_0_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3418560)))]; + tensor generator_resblocks_1_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_1_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3419648)))]; + tensor generator_resblocks_1_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3421760))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3487360))), name = tensor("generator_resblocks_1_adain1_0_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_1_convs1_0_bias = const()[name = tensor("generator_resblocks_1_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3488448)))]; + tensor generator_resblocks_1_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_1_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3489536)))]; + tensor generator_resblocks_1_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3491648))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3557248))), name = tensor("generator_resblocks_1_adain2_0_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_1_convs2_0_bias = const()[name = tensor("generator_resblocks_1_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3558336)))]; + tensor generator_resblocks_1_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_1_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3559424)))]; + tensor generator_resblocks_1_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3561536))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3627136))), name = tensor("generator_resblocks_1_adain1_1_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_1_convs1_1_bias = const()[name = tensor("generator_resblocks_1_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3628224)))]; + tensor generator_resblocks_1_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_1_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3629312)))]; + tensor generator_resblocks_1_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3631424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3697024))), name = tensor("generator_resblocks_1_adain2_1_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_1_convs2_1_bias = const()[name = tensor("generator_resblocks_1_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3698112)))]; + tensor generator_resblocks_1_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_1_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3699200)))]; + tensor generator_resblocks_1_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3701312))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3766912))), name = tensor("generator_resblocks_1_adain1_2_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_1_convs1_2_bias = const()[name = tensor("generator_resblocks_1_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3768000)))]; + tensor generator_resblocks_1_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_1_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3769088)))]; + tensor generator_resblocks_1_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3771200))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3836800))), name = tensor("generator_resblocks_1_adain2_2_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_1_convs2_2_bias = const()[name = tensor("generator_resblocks_1_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3837888)))]; + tensor generator_resblocks_2_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_2_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3838976)))]; + tensor generator_resblocks_2_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3841088))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3906688))), name = tensor("generator_resblocks_2_adain1_0_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_2_convs1_0_bias = const()[name = tensor("generator_resblocks_2_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3907776)))]; + tensor generator_resblocks_2_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_2_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3908864)))]; + tensor generator_resblocks_2_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3910976))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3976576))), name = tensor("generator_resblocks_2_adain2_0_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_2_convs2_0_bias = const()[name = tensor("generator_resblocks_2_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3977664)))]; + tensor generator_resblocks_2_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_2_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3978752)))]; + tensor generator_resblocks_2_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(3980864))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4046464))), name = tensor("generator_resblocks_2_adain1_1_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_2_convs1_1_bias = const()[name = tensor("generator_resblocks_2_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4047552)))]; + tensor generator_resblocks_2_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_2_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4048640)))]; + tensor generator_resblocks_2_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4050752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4116352))), name = tensor("generator_resblocks_2_adain2_1_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_2_convs2_1_bias = const()[name = tensor("generator_resblocks_2_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4117440)))]; + tensor generator_resblocks_2_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_2_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4118528)))]; + tensor generator_resblocks_2_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4120640))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4186240))), name = tensor("generator_resblocks_2_adain1_2_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_2_convs1_2_bias = const()[name = tensor("generator_resblocks_2_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4187328)))]; + tensor generator_resblocks_2_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_2_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4188416)))]; + tensor generator_resblocks_2_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4190528))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4256128))), name = tensor("generator_resblocks_2_adain2_2_fc_weight_palettized"), shape = tensor([512, 128])]; + tensor generator_resblocks_2_convs2_2_bias = const()[name = tensor("generator_resblocks_2_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4257216)))]; + tensor generator_noise_convs_1_bias = const()[name = tensor("generator_noise_convs_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4258304)))]; + tensor generator_noise_convs_1_weight = const()[name = tensor("generator_noise_convs_1_weight"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4258880)))]; + tensor generator_noise_res_1_adain1_0_fc_bias = const()[name = tensor("generator_noise_res_1_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4265088)))]; + tensor generator_noise_res_1_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4266176))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4299008))), name = tensor("generator_noise_res_1_adain1_0_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_noise_res_1_convs1_0_bias = const()[name = tensor("generator_noise_res_1_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4300096)))]; + tensor generator_noise_res_1_adain2_0_fc_bias = const()[name = tensor("generator_noise_res_1_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4300672)))]; + tensor generator_noise_res_1_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4301760))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4334592))), name = tensor("generator_noise_res_1_adain2_0_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_noise_res_1_convs2_0_bias = const()[name = tensor("generator_noise_res_1_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4335680)))]; + tensor generator_noise_res_1_adain1_1_fc_bias = const()[name = tensor("generator_noise_res_1_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4336256)))]; + tensor generator_noise_res_1_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4337344))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4370176))), name = tensor("generator_noise_res_1_adain1_1_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_noise_res_1_convs1_1_bias = const()[name = tensor("generator_noise_res_1_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4371264)))]; + tensor generator_noise_res_1_adain2_1_fc_bias = const()[name = tensor("generator_noise_res_1_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4371840)))]; + tensor generator_noise_res_1_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4372928))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4405760))), name = tensor("generator_noise_res_1_adain2_1_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_noise_res_1_convs2_1_bias = const()[name = tensor("generator_noise_res_1_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4406848)))]; + tensor generator_noise_res_1_adain1_2_fc_bias = const()[name = tensor("generator_noise_res_1_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4407424)))]; + tensor generator_noise_res_1_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4408512))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4441344))), name = tensor("generator_noise_res_1_adain1_2_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_noise_res_1_convs1_2_bias = const()[name = tensor("generator_noise_res_1_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4442432)))]; + tensor generator_noise_res_1_adain2_2_fc_bias = const()[name = tensor("generator_noise_res_1_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4443008)))]; + tensor generator_noise_res_1_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4444096))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4476928))), name = tensor("generator_noise_res_1_adain2_2_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_noise_res_1_convs2_2_bias = const()[name = tensor("generator_noise_res_1_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4478016)))]; + tensor generator_ups_1_bias = const()[name = tensor("generator_ups_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4478592)))]; + tensor generator_resblocks_3_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_3_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4479168)))]; + tensor generator_resblocks_3_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4480256))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4513088))), name = tensor("generator_resblocks_3_adain1_0_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_3_convs1_0_bias = const()[name = tensor("generator_resblocks_3_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4514176)))]; + tensor generator_resblocks_3_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_3_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4514752)))]; + tensor generator_resblocks_3_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4515840))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4548672))), name = tensor("generator_resblocks_3_adain2_0_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_3_convs2_0_bias = const()[name = tensor("generator_resblocks_3_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4549760)))]; + tensor generator_resblocks_3_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_3_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4550336)))]; + tensor generator_resblocks_3_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4551424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4584256))), name = tensor("generator_resblocks_3_adain1_1_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_3_convs1_1_bias = const()[name = tensor("generator_resblocks_3_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4585344)))]; + tensor generator_resblocks_3_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_3_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4585920)))]; + tensor generator_resblocks_3_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4587008))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4619840))), name = tensor("generator_resblocks_3_adain2_1_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_3_convs2_1_bias = const()[name = tensor("generator_resblocks_3_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4620928)))]; + tensor generator_resblocks_3_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_3_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4621504)))]; + tensor generator_resblocks_3_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4622592))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4655424))), name = tensor("generator_resblocks_3_adain1_2_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_3_convs1_2_bias = const()[name = tensor("generator_resblocks_3_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4656512)))]; + tensor generator_resblocks_3_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_3_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4657088)))]; + tensor generator_resblocks_3_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4658176))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4691008))), name = tensor("generator_resblocks_3_adain2_2_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_3_convs2_2_bias = const()[name = tensor("generator_resblocks_3_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4692096)))]; + tensor generator_resblocks_4_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_4_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4692672)))]; + tensor generator_resblocks_4_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4693760))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4726592))), name = tensor("generator_resblocks_4_adain1_0_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_4_convs1_0_bias = const()[name = tensor("generator_resblocks_4_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4727680)))]; + tensor generator_resblocks_4_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_4_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4728256)))]; + tensor generator_resblocks_4_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4729344))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4762176))), name = tensor("generator_resblocks_4_adain2_0_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_4_convs2_0_bias = const()[name = tensor("generator_resblocks_4_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4763264)))]; + tensor generator_resblocks_4_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_4_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4763840)))]; + tensor generator_resblocks_4_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4764928))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4797760))), name = tensor("generator_resblocks_4_adain1_1_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_4_convs1_1_bias = const()[name = tensor("generator_resblocks_4_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4798848)))]; + tensor generator_resblocks_4_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_4_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4799424)))]; + tensor generator_resblocks_4_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4800512))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4833344))), name = tensor("generator_resblocks_4_adain2_1_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_4_convs2_1_bias = const()[name = tensor("generator_resblocks_4_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4834432)))]; + tensor generator_resblocks_4_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_4_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4835008)))]; + tensor generator_resblocks_4_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4836096))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4868928))), name = tensor("generator_resblocks_4_adain1_2_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_4_convs1_2_bias = const()[name = tensor("generator_resblocks_4_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4870016)))]; + tensor generator_resblocks_4_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_4_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4870592)))]; + tensor generator_resblocks_4_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4871680))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4904512))), name = tensor("generator_resblocks_4_adain2_2_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_4_convs2_2_bias = const()[name = tensor("generator_resblocks_4_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4905600)))]; + tensor generator_resblocks_5_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_5_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4906176)))]; + tensor generator_resblocks_5_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4907264))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4940096))), name = tensor("generator_resblocks_5_adain1_0_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_5_convs1_0_bias = const()[name = tensor("generator_resblocks_5_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4941184)))]; + tensor generator_resblocks_5_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_5_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4941760)))]; + tensor generator_resblocks_5_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4942848))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4975680))), name = tensor("generator_resblocks_5_adain2_0_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_5_convs2_0_bias = const()[name = tensor("generator_resblocks_5_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4976768)))]; + tensor generator_resblocks_5_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_5_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4977344)))]; + tensor generator_resblocks_5_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4978432))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5011264))), name = tensor("generator_resblocks_5_adain1_1_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_5_convs1_1_bias = const()[name = tensor("generator_resblocks_5_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5012352)))]; + tensor generator_resblocks_5_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_5_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5012928)))]; + tensor generator_resblocks_5_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5014016))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5046848))), name = tensor("generator_resblocks_5_adain2_1_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_5_convs2_1_bias = const()[name = tensor("generator_resblocks_5_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5047936)))]; + tensor generator_resblocks_5_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_5_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5048512)))]; + tensor generator_resblocks_5_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5049600))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5082432))), name = tensor("generator_resblocks_5_adain1_2_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_5_convs1_2_bias = const()[name = tensor("generator_resblocks_5_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5083520)))]; + tensor generator_resblocks_5_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_5_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5084096)))]; + tensor generator_resblocks_5_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5085184))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5118016))), name = tensor("generator_resblocks_5_adain2_2_fc_weight_palettized"), shape = tensor([256, 128])]; + tensor generator_resblocks_5_convs2_2_bias = const()[name = tensor("generator_resblocks_5_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5119104)))]; + tensor generator_noise_convs_2_bias = const()[name = tensor("generator_noise_convs_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5119680)))]; + tensor generator_noise_convs_2_weight = const()[name = tensor("generator_noise_convs_2_weight"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5120000)))]; + tensor generator_noise_res_2_adain1_0_fc_bias = const()[name = tensor("generator_noise_res_2_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5121088)))]; + tensor generator_noise_res_2_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5121664))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5138112))), name = tensor("generator_noise_res_2_adain1_0_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_noise_res_2_convs1_0_bias = const()[name = tensor("generator_noise_res_2_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5139200)))]; + tensor generator_noise_res_2_adain2_0_fc_bias = const()[name = tensor("generator_noise_res_2_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5139520)))]; + tensor generator_noise_res_2_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5140096))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5156544))), name = tensor("generator_noise_res_2_adain2_0_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_noise_res_2_convs2_0_bias = const()[name = tensor("generator_noise_res_2_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5157632)))]; + tensor generator_noise_res_2_adain1_1_fc_bias = const()[name = tensor("generator_noise_res_2_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5157952)))]; + tensor generator_noise_res_2_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5158528))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5174976))), name = tensor("generator_noise_res_2_adain1_1_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_noise_res_2_convs1_1_bias = const()[name = tensor("generator_noise_res_2_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5176064)))]; + tensor generator_noise_res_2_adain2_1_fc_bias = const()[name = tensor("generator_noise_res_2_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5176384)))]; + tensor generator_noise_res_2_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5176960))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5193408))), name = tensor("generator_noise_res_2_adain2_1_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_noise_res_2_convs2_1_bias = const()[name = tensor("generator_noise_res_2_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5194496)))]; + tensor generator_noise_res_2_adain1_2_fc_bias = const()[name = tensor("generator_noise_res_2_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5194816)))]; + tensor generator_noise_res_2_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5195392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5211840))), name = tensor("generator_noise_res_2_adain1_2_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_noise_res_2_convs1_2_bias = const()[name = tensor("generator_noise_res_2_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5212928)))]; + tensor generator_noise_res_2_adain2_2_fc_bias = const()[name = tensor("generator_noise_res_2_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5213248)))]; + tensor generator_noise_res_2_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5213824))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5230272))), name = tensor("generator_noise_res_2_adain2_2_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_noise_res_2_convs2_2_bias = const()[name = tensor("generator_noise_res_2_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5231360)))]; + tensor generator_ups_2_bias = const()[name = tensor("generator_ups_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5231680)))]; + tensor generator_resblocks_6_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_6_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5232000)))]; + tensor generator_resblocks_6_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5232576))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5249024))), name = tensor("generator_resblocks_6_adain1_0_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_6_convs1_0_bias = const()[name = tensor("generator_resblocks_6_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5250112)))]; + tensor generator_resblocks_6_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_6_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5250432)))]; + tensor generator_resblocks_6_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5251008))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5267456))), name = tensor("generator_resblocks_6_adain2_0_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_6_convs2_0_bias = const()[name = tensor("generator_resblocks_6_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5268544)))]; + tensor generator_resblocks_6_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_6_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5268864)))]; + tensor generator_resblocks_6_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5269440))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5285888))), name = tensor("generator_resblocks_6_adain1_1_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_6_convs1_1_bias = const()[name = tensor("generator_resblocks_6_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5286976)))]; + tensor generator_resblocks_6_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_6_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5287296)))]; + tensor generator_resblocks_6_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5287872))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5304320))), name = tensor("generator_resblocks_6_adain2_1_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_6_convs2_1_bias = const()[name = tensor("generator_resblocks_6_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5305408)))]; + tensor generator_resblocks_6_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_6_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5305728)))]; + tensor generator_resblocks_6_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5306304))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5322752))), name = tensor("generator_resblocks_6_adain1_2_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_6_convs1_2_bias = const()[name = tensor("generator_resblocks_6_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5323840)))]; + tensor generator_resblocks_6_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_6_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5324160)))]; + tensor generator_resblocks_6_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5324736))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5341184))), name = tensor("generator_resblocks_6_adain2_2_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_6_convs2_2_bias = const()[name = tensor("generator_resblocks_6_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5342272)))]; + tensor generator_resblocks_7_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_7_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5342592)))]; + tensor generator_resblocks_7_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5343168))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5359616))), name = tensor("generator_resblocks_7_adain1_0_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_7_convs1_0_bias = const()[name = tensor("generator_resblocks_7_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5360704)))]; + tensor generator_resblocks_7_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_7_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5361024)))]; + tensor generator_resblocks_7_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5361600))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5378048))), name = tensor("generator_resblocks_7_adain2_0_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_7_convs2_0_bias = const()[name = tensor("generator_resblocks_7_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5379136)))]; + tensor generator_resblocks_7_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_7_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5379456)))]; + tensor generator_resblocks_7_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5380032))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5396480))), name = tensor("generator_resblocks_7_adain1_1_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_7_convs1_1_bias = const()[name = tensor("generator_resblocks_7_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5397568)))]; + tensor generator_resblocks_7_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_7_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5397888)))]; + tensor generator_resblocks_7_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5398464))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5414912))), name = tensor("generator_resblocks_7_adain2_1_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_7_convs2_1_bias = const()[name = tensor("generator_resblocks_7_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5416000)))]; + tensor generator_resblocks_7_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_7_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5416320)))]; + tensor generator_resblocks_7_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5416896))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5433344))), name = tensor("generator_resblocks_7_adain1_2_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_7_convs1_2_bias = const()[name = tensor("generator_resblocks_7_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5434432)))]; + tensor generator_resblocks_7_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_7_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5434752)))]; + tensor generator_resblocks_7_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5435328))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5451776))), name = tensor("generator_resblocks_7_adain2_2_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_7_convs2_2_bias = const()[name = tensor("generator_resblocks_7_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5452864)))]; + tensor generator_resblocks_8_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_8_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5453184)))]; + tensor generator_resblocks_8_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5453760))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5470208))), name = tensor("generator_resblocks_8_adain1_0_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_8_convs1_0_bias = const()[name = tensor("generator_resblocks_8_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5471296)))]; + tensor generator_resblocks_8_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_8_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5471616)))]; + tensor generator_resblocks_8_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5472192))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5488640))), name = tensor("generator_resblocks_8_adain2_0_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_8_convs2_0_bias = const()[name = tensor("generator_resblocks_8_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5489728)))]; + tensor generator_resblocks_8_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_8_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5490048)))]; + tensor generator_resblocks_8_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5490624))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5507072))), name = tensor("generator_resblocks_8_adain1_1_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_8_convs1_1_bias = const()[name = tensor("generator_resblocks_8_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5508160)))]; + tensor generator_resblocks_8_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_8_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5508480)))]; + tensor generator_resblocks_8_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5509056))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5525504))), name = tensor("generator_resblocks_8_adain2_1_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_8_convs2_1_bias = const()[name = tensor("generator_resblocks_8_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5526592)))]; + tensor generator_resblocks_8_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_8_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5526912)))]; + tensor generator_resblocks_8_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5527488))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5543936))), name = tensor("generator_resblocks_8_adain1_2_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_8_convs1_2_bias = const()[name = tensor("generator_resblocks_8_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5545024)))]; + tensor generator_resblocks_8_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_8_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5545344)))]; + tensor generator_resblocks_8_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5545920))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5562368))), name = tensor("generator_resblocks_8_adain2_2_fc_weight_palettized"), shape = tensor([128, 128])]; + tensor generator_resblocks_8_convs2_2_bias = const()[name = tensor("generator_resblocks_8_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5563456)))]; + tensor generator_noise_convs_3_bias = const()[name = tensor("generator_noise_convs_3_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5563776)))]; + tensor generator_noise_convs_3_weight = const()[name = tensor("generator_noise_convs_3_weight"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5563968)))]; + tensor generator_noise_res_3_adain1_0_fc_bias = const()[name = tensor("generator_noise_res_3_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5564160)))]; + tensor generator_noise_res_3_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5564480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5572736))), name = tensor("generator_noise_res_3_adain1_0_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_noise_res_3_convs1_0_bias = const()[name = tensor("generator_noise_res_3_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5573824)))]; + tensor generator_noise_res_3_adain2_0_fc_bias = const()[name = tensor("generator_noise_res_3_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5574016)))]; + tensor generator_noise_res_3_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5574336))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5582592))), name = tensor("generator_noise_res_3_adain2_0_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_noise_res_3_convs2_0_bias = const()[name = tensor("generator_noise_res_3_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5583680)))]; + tensor generator_noise_res_3_adain1_1_fc_bias = const()[name = tensor("generator_noise_res_3_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5583872)))]; + tensor generator_noise_res_3_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5584192))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5592448))), name = tensor("generator_noise_res_3_adain1_1_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_noise_res_3_convs1_1_bias = const()[name = tensor("generator_noise_res_3_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5593536)))]; + tensor generator_noise_res_3_adain2_1_fc_bias = const()[name = tensor("generator_noise_res_3_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5593728)))]; + tensor generator_noise_res_3_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5594048))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5602304))), name = tensor("generator_noise_res_3_adain2_1_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_noise_res_3_convs2_1_bias = const()[name = tensor("generator_noise_res_3_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5603392)))]; + tensor generator_noise_res_3_adain1_2_fc_bias = const()[name = tensor("generator_noise_res_3_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5603584)))]; + tensor generator_noise_res_3_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5603904))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5612160))), name = tensor("generator_noise_res_3_adain1_2_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_noise_res_3_convs1_2_bias = const()[name = tensor("generator_noise_res_3_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5613248)))]; + tensor generator_noise_res_3_adain2_2_fc_bias = const()[name = tensor("generator_noise_res_3_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5613440)))]; + tensor generator_noise_res_3_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5613760))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5622016))), name = tensor("generator_noise_res_3_adain2_2_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_noise_res_3_convs2_2_bias = const()[name = tensor("generator_noise_res_3_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5623104)))]; + tensor generator_ups_3_bias = const()[name = tensor("generator_ups_3_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5623296)))]; + tensor generator_resblocks_9_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_9_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5623488)))]; + tensor generator_resblocks_9_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5623808))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5632064))), name = tensor("generator_resblocks_9_adain1_0_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_9_convs1_0_bias = const()[name = tensor("generator_resblocks_9_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5633152)))]; + tensor generator_resblocks_9_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_9_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5633344)))]; + tensor generator_resblocks_9_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5633664))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5641920))), name = tensor("generator_resblocks_9_adain2_0_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_9_convs2_0_bias = const()[name = tensor("generator_resblocks_9_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5643008)))]; + tensor generator_resblocks_9_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_9_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5643200)))]; + tensor generator_resblocks_9_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5643520))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5651776))), name = tensor("generator_resblocks_9_adain1_1_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_9_convs1_1_bias = const()[name = tensor("generator_resblocks_9_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5652864)))]; + tensor generator_resblocks_9_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_9_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5653056)))]; + tensor generator_resblocks_9_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5653376))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5661632))), name = tensor("generator_resblocks_9_adain2_1_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_9_convs2_1_bias = const()[name = tensor("generator_resblocks_9_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5662720)))]; + tensor generator_resblocks_9_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_9_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5662912)))]; + tensor generator_resblocks_9_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5663232))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5671488))), name = tensor("generator_resblocks_9_adain1_2_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_9_convs1_2_bias = const()[name = tensor("generator_resblocks_9_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5672576)))]; + tensor generator_resblocks_9_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_9_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5672768)))]; + tensor generator_resblocks_9_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5673088))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5681344))), name = tensor("generator_resblocks_9_adain2_2_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_9_convs2_2_bias = const()[name = tensor("generator_resblocks_9_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5682432)))]; + tensor generator_resblocks_10_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_10_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5682624)))]; + tensor generator_resblocks_10_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5682944))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5691200))), name = tensor("generator_resblocks_10_adain1_0_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_10_convs1_0_bias = const()[name = tensor("generator_resblocks_10_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5692288)))]; + tensor generator_resblocks_10_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_10_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5692480)))]; + tensor generator_resblocks_10_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5692800))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5701056))), name = tensor("generator_resblocks_10_adain2_0_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_10_convs2_0_bias = const()[name = tensor("generator_resblocks_10_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5702144)))]; + tensor generator_resblocks_10_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_10_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5702336)))]; + tensor generator_resblocks_10_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5702656))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5710912))), name = tensor("generator_resblocks_10_adain1_1_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_10_convs1_1_bias = const()[name = tensor("generator_resblocks_10_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5712000)))]; + tensor generator_resblocks_10_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_10_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5712192)))]; + tensor generator_resblocks_10_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5712512))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5720768))), name = tensor("generator_resblocks_10_adain2_1_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_10_convs2_1_bias = const()[name = tensor("generator_resblocks_10_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5721856)))]; + tensor generator_resblocks_10_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_10_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5722048)))]; + tensor generator_resblocks_10_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5722368))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5730624))), name = tensor("generator_resblocks_10_adain1_2_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_10_convs1_2_bias = const()[name = tensor("generator_resblocks_10_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5731712)))]; + tensor generator_resblocks_10_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_10_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5731904)))]; + tensor generator_resblocks_10_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5732224))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5740480))), name = tensor("generator_resblocks_10_adain2_2_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_10_convs2_2_bias = const()[name = tensor("generator_resblocks_10_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5741568)))]; + tensor generator_resblocks_11_adain1_0_fc_bias = const()[name = tensor("generator_resblocks_11_adain1_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5741760)))]; + tensor generator_resblocks_11_adain1_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5742080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5750336))), name = tensor("generator_resblocks_11_adain1_0_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_11_convs1_0_bias = const()[name = tensor("generator_resblocks_11_convs1_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5751424)))]; + tensor generator_resblocks_11_adain2_0_fc_bias = const()[name = tensor("generator_resblocks_11_adain2_0_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5751616)))]; + tensor generator_resblocks_11_adain2_0_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5751936))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5760192))), name = tensor("generator_resblocks_11_adain2_0_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_11_convs2_0_bias = const()[name = tensor("generator_resblocks_11_convs2_0_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5761280)))]; + tensor generator_resblocks_11_adain1_1_fc_bias = const()[name = tensor("generator_resblocks_11_adain1_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5761472)))]; + tensor generator_resblocks_11_adain1_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5761792))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5770048))), name = tensor("generator_resblocks_11_adain1_1_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_11_convs1_1_bias = const()[name = tensor("generator_resblocks_11_convs1_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5771136)))]; + tensor generator_resblocks_11_adain2_1_fc_bias = const()[name = tensor("generator_resblocks_11_adain2_1_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5771328)))]; + tensor generator_resblocks_11_adain2_1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5771648))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5779904))), name = tensor("generator_resblocks_11_adain2_1_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_11_convs2_1_bias = const()[name = tensor("generator_resblocks_11_convs2_1_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5780992)))]; + tensor generator_resblocks_11_adain1_2_fc_bias = const()[name = tensor("generator_resblocks_11_adain1_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5781184)))]; + tensor generator_resblocks_11_adain1_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5781504))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5789760))), name = tensor("generator_resblocks_11_adain1_2_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_11_convs1_2_bias = const()[name = tensor("generator_resblocks_11_convs1_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5790848)))]; + tensor generator_resblocks_11_adain2_2_fc_bias = const()[name = tensor("generator_resblocks_11_adain2_2_fc_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5791040)))]; + tensor generator_resblocks_11_adain2_2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5791360))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5799616))), name = tensor("generator_resblocks_11_adain2_2_fc_weight_palettized"), shape = tensor([64, 128])]; + tensor generator_resblocks_11_convs2_2_bias = const()[name = tensor("generator_resblocks_11_convs2_2_bias"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5800704)))]; + tensor generator_conv_post_bias = const()[name = tensor("generator_conv_post_bias"), val = tensor([0x1.25d976p-16])]; + tensor input_1_axes_0 = const()[name = tensor("input_1_axes_0"), val = tensor([1])]; + tensor input_1 = expand_dims(axes = input_1_axes_0, x = F0_curve)[name = tensor("input_1")]; + tensor weight_1 = const()[name = tensor("weight_1"), val = tensor([[[0x1.5663dap-5, -0x1.864e36p-5, -0x1.5048p-4]]])]; + tensor F0_pad_type_0 = const()[name = tensor("F0_pad_type_0"), val = tensor("custom")]; + tensor F0_pad_0 = const()[name = tensor("F0_pad_0"), val = tensor([1, 1])]; + tensor F0_strides_0 = const()[name = tensor("F0_strides_0"), val = tensor([2])]; + tensor F0_dilations_0 = const()[name = tensor("F0_dilations_0"), val = tensor([1])]; + tensor F0_groups_0 = const()[name = tensor("F0_groups_0"), val = tensor(1)]; + tensor F0 = conv(bias = F0_conv_bias, dilations = F0_dilations_0, groups = F0_groups_0, pad = F0_pad_0, pad_type = F0_pad_type_0, strides = F0_strides_0, weight = weight_1, x = input_1)[name = tensor("F0")]; + tensor input_3_axes_0 = const()[name = tensor("input_3_axes_0"), val = tensor([1])]; + tensor input_3 = expand_dims(axes = input_3_axes_0, x = N)[name = tensor("input_3")]; + tensor weight_3 = const()[name = tensor("weight_3"), val = tensor([[[0x1.1d454p-5, -0x1.710ff4p-1, -0x1.89a072p-1]]])]; + tensor Nc_pad_type_0 = const()[name = tensor("Nc_pad_type_0"), val = tensor("custom")]; + tensor Nc_pad_0 = const()[name = tensor("Nc_pad_0"), val = tensor([1, 1])]; + tensor Nc_strides_0 = const()[name = tensor("Nc_strides_0"), val = tensor([2])]; + tensor Nc_dilations_0 = const()[name = tensor("Nc_dilations_0"), val = tensor([1])]; + tensor Nc_groups_0 = const()[name = tensor("Nc_groups_0"), val = tensor(1)]; + tensor Nc = conv(bias = N_conv_bias, dilations = Nc_dilations_0, groups = Nc_groups_0, pad = Nc_pad_0, pad_type = Nc_pad_type_0, strides = Nc_strides_0, weight = weight_3, x = input_3)[name = tensor("Nc")]; + tensor var_187 = const()[name = tensor("op_187"), val = tensor(1)]; + tensor input_5_interleave_0 = const()[name = tensor("input_5_interleave_0"), val = tensor(false)]; + tensor input_5 = concat(axis = var_187, interleave = input_5_interleave_0, values = (asr, F0, Nc))[name = tensor("input_5")]; + tensor var_191 = const()[name = tensor("op_191"), val = tensor(0x1.99999ap-3)]; + tensor var_194 = const()[name = tensor("op_194"), val = tensor(0x1.4f8b58p-17)]; + tensor h_1 = linear(bias = encode_norm1_fc_bias, weight = encode_norm1_fc_weight_palettized, x = s)[name = tensor("linear_0")]; + tensor var_220 = const()[name = tensor("op_220"), val = tensor([1, 1028, 1])]; + tensor h_3 = reshape(shape = var_220, x = h_1)[name = tensor("h_3")]; + tensor var_222_split_sizes_0 = const()[name = tensor("op_222_split_sizes_0"), val = tensor([514, 514])]; + tensor var_222_axis_0 = const()[name = tensor("op_222_axis_0"), val = tensor(1)]; + tensor var_222_0, tensor var_222_1 = split(axis = var_222_axis_0, split_sizes = var_222_split_sizes_0, x = h_3)[name = tensor("op_222")]; + tensor var_224_promoted = const()[name = tensor("op_224_promoted"), val = tensor(0x1p+0)]; + tensor var_225 = add(x = var_222_0, y = var_224_promoted)[name = tensor("op_225")]; + tensor var_226 = instance_norm(epsilon = var_194, x = input_5)[name = tensor("op_226")]; + tensor var_227 = mul(x = var_225, y = var_226)[name = tensor("op_227")]; + tensor input_7 = add(x = var_227, y = var_222_1)[name = tensor("input_7")]; + tensor input_9 = leaky_relu(alpha = var_191, x = input_7)[name = tensor("input_9")]; + tensor weight_5_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5800896))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7379968))), name = tensor("weight_5_palettized"), shape = tensor([1024, 514, 3])]; + tensor input_13_pad_type_0 = const()[name = tensor("input_13_pad_type_0"), val = tensor("custom")]; + tensor input_13_pad_0 = const()[name = tensor("input_13_pad_0"), val = tensor([1, 1])]; + tensor input_13_strides_0 = const()[name = tensor("input_13_strides_0"), val = tensor([1])]; + tensor input_13_dilations_0 = const()[name = tensor("input_13_dilations_0"), val = tensor([1])]; + tensor input_13_groups_0 = const()[name = tensor("input_13_groups_0"), val = tensor(1)]; + tensor input_13 = conv(bias = encode_conv1_bias, dilations = input_13_dilations_0, groups = input_13_groups_0, pad = input_13_pad_0, pad_type = input_13_pad_type_0, strides = input_13_strides_0, weight = weight_5_palettized, x = input_9)[name = tensor("input_13")]; + tensor h_5 = linear(bias = encode_norm2_fc_bias, weight = encode_norm2_fc_weight_palettized, x = s)[name = tensor("linear_1")]; + tensor var_244 = const()[name = tensor("op_244"), val = tensor([1, 2048, 1])]; + tensor h_7 = reshape(shape = var_244, x = h_5)[name = tensor("h_7")]; + tensor var_246_split_sizes_0 = const()[name = tensor("op_246_split_sizes_0"), val = tensor([1024, 1024])]; + tensor var_246_axis_0 = const()[name = tensor("op_246_axis_0"), val = tensor(1)]; + tensor var_246_0, tensor var_246_1 = split(axis = var_246_axis_0, split_sizes = var_246_split_sizes_0, x = h_7)[name = tensor("op_246")]; + tensor var_248_promoted = const()[name = tensor("op_248_promoted"), val = tensor(0x1p+0)]; + tensor var_249 = add(x = var_246_0, y = var_248_promoted)[name = tensor("op_249")]; + tensor var_250 = instance_norm(epsilon = var_194, x = input_13)[name = tensor("op_250")]; + tensor var_251 = mul(x = var_249, y = var_250)[name = tensor("op_251")]; + tensor input_15 = add(x = var_251, y = var_246_1)[name = tensor("input_15")]; + tensor input_17 = leaky_relu(alpha = var_191, x = input_15)[name = tensor("input_17")]; + tensor weight_7_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7381056))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(10526848))), name = tensor("weight_7_palettized"), shape = tensor([1024, 1024, 3])]; + tensor out_1_pad_type_0 = const()[name = tensor("out_1_pad_type_0"), val = tensor("custom")]; + tensor out_1_pad_0 = const()[name = tensor("out_1_pad_0"), val = tensor([1, 1])]; + tensor out_1_strides_0 = const()[name = tensor("out_1_strides_0"), val = tensor([1])]; + tensor out_1_dilations_0 = const()[name = tensor("out_1_dilations_0"), val = tensor([1])]; + tensor out_1_groups_0 = const()[name = tensor("out_1_groups_0"), val = tensor(1)]; + tensor out_1 = conv(bias = encode_conv2_bias, dilations = out_1_dilations_0, groups = out_1_groups_0, pad = out_1_pad_0, pad_type = out_1_pad_type_0, strides = out_1_strides_0, weight = weight_7_palettized, x = input_17)[name = tensor("out_1")]; + tensor weight_9_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(10527936))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11054336))), name = tensor("weight_9_palettized"), shape = tensor([1024, 514, 1])]; + tensor var_267_pad_type_0 = const()[name = tensor("op_267_pad_type_0"), val = tensor("valid")]; + tensor var_267_strides_0 = const()[name = tensor("op_267_strides_0"), val = tensor([1])]; + tensor var_267_pad_0 = const()[name = tensor("op_267_pad_0"), val = tensor([0, 0])]; + tensor var_267_dilations_0 = const()[name = tensor("op_267_dilations_0"), val = tensor([1])]; + tensor var_267_groups_0 = const()[name = tensor("op_267_groups_0"), val = tensor(1)]; + tensor var_267 = conv(dilations = var_267_dilations_0, groups = var_267_groups_0, pad = var_267_pad_0, pad_type = var_267_pad_type_0, strides = var_267_strides_0, weight = weight_9_palettized, x = input_5)[name = tensor("op_267")]; + tensor var_268 = add(x = out_1, y = var_267)[name = tensor("op_268")]; + tensor _inversed_x_1_y_0 = const()[name = tensor("_inversed_x_1_y_0"), val = tensor(0x1.6a09e6p-1)]; + tensor _inversed_x_1 = mul(x = var_268, y = _inversed_x_1_y_0)[name = tensor("_inversed_x_1")]; + tensor weight_11_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11055424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11088256))), name = tensor("weight_11_palettized"), shape = tensor([64, 512, 1])]; + tensor asr_res_1_pad_type_0 = const()[name = tensor("asr_res_1_pad_type_0"), val = tensor("valid")]; + tensor asr_res_1_strides_0 = const()[name = tensor("asr_res_1_strides_0"), val = tensor([1])]; + tensor asr_res_1_pad_0 = const()[name = tensor("asr_res_1_pad_0"), val = tensor([0, 0])]; + tensor asr_res_1_dilations_0 = const()[name = tensor("asr_res_1_dilations_0"), val = tensor([1])]; + tensor asr_res_1_groups_0 = const()[name = tensor("asr_res_1_groups_0"), val = tensor(1)]; + tensor asr_res_1 = conv(bias = asr_res_0_bias, dilations = asr_res_1_dilations_0, groups = asr_res_1_groups_0, pad = asr_res_1_pad_0, pad_type = asr_res_1_pad_type_0, strides = asr_res_1_strides_0, weight = weight_11_palettized, x = asr)[name = tensor("asr_res_1")]; + tensor var_288 = const()[name = tensor("op_288"), val = tensor(1)]; + tensor input_21_interleave_0 = const()[name = tensor("input_21_interleave_0"), val = tensor(false)]; + tensor input_21 = concat(axis = var_288, interleave = input_21_interleave_0, values = (_inversed_x_1, asr_res_1, F0, Nc))[name = tensor("input_21")]; + tensor var_292 = const()[name = tensor("op_292"), val = tensor(0x1.99999ap-3)]; + tensor var_295 = const()[name = tensor("op_295"), val = tensor(0x1.4f8b58p-17)]; + tensor h_9 = linear(bias = decode_0_norm1_fc_bias, weight = decode_0_norm1_fc_weight_palettized, x = s)[name = tensor("linear_2")]; + tensor var_321 = const()[name = tensor("op_321"), val = tensor([1, 2180, 1])]; + tensor h_11 = reshape(shape = var_321, x = h_9)[name = tensor("h_11")]; + tensor var_323_split_sizes_0 = const()[name = tensor("op_323_split_sizes_0"), val = tensor([1090, 1090])]; + tensor var_323_axis_0 = const()[name = tensor("op_323_axis_0"), val = tensor(1)]; + tensor var_323_0, tensor var_323_1 = split(axis = var_323_axis_0, split_sizes = var_323_split_sizes_0, x = h_11)[name = tensor("op_323")]; + tensor var_325_promoted = const()[name = tensor("op_325_promoted"), val = tensor(0x1p+0)]; + tensor var_326 = add(x = var_323_0, y = var_325_promoted)[name = tensor("op_326")]; + tensor var_327 = instance_norm(epsilon = var_295, x = input_21)[name = tensor("op_327")]; + tensor var_328 = mul(x = var_326, y = var_327)[name = tensor("op_328")]; + tensor input_23 = add(x = var_328, y = var_323_1)[name = tensor("input_23")]; + tensor input_25 = leaky_relu(alpha = var_292, x = input_23)[name = tensor("input_25")]; + tensor weight_13_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11089344))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(14437888))), name = tensor("weight_13_palettized"), shape = tensor([1024, 1090, 3])]; + tensor input_29_pad_type_0 = const()[name = tensor("input_29_pad_type_0"), val = tensor("custom")]; + tensor input_29_pad_0 = const()[name = tensor("input_29_pad_0"), val = tensor([1, 1])]; + tensor input_29_strides_0 = const()[name = tensor("input_29_strides_0"), val = tensor([1])]; + tensor input_29_dilations_0 = const()[name = tensor("input_29_dilations_0"), val = tensor([1])]; + tensor input_29_groups_0 = const()[name = tensor("input_29_groups_0"), val = tensor(1)]; + tensor input_29 = conv(bias = decode_0_conv1_bias, dilations = input_29_dilations_0, groups = input_29_groups_0, pad = input_29_pad_0, pad_type = input_29_pad_type_0, strides = input_29_strides_0, weight = weight_13_palettized, x = input_25)[name = tensor("input_29")]; + tensor h_13 = linear(bias = decode_0_norm2_fc_bias, weight = decode_0_norm2_fc_weight_palettized, x = s)[name = tensor("linear_3")]; + tensor var_345 = const()[name = tensor("op_345"), val = tensor([1, 2048, 1])]; + tensor h_15 = reshape(shape = var_345, x = h_13)[name = tensor("h_15")]; + tensor var_347_split_sizes_0 = const()[name = tensor("op_347_split_sizes_0"), val = tensor([1024, 1024])]; + tensor var_347_axis_0 = const()[name = tensor("op_347_axis_0"), val = tensor(1)]; + tensor var_347_0, tensor var_347_1 = split(axis = var_347_axis_0, split_sizes = var_347_split_sizes_0, x = h_15)[name = tensor("op_347")]; + tensor var_349_promoted = const()[name = tensor("op_349_promoted"), val = tensor(0x1p+0)]; + tensor var_350 = add(x = var_347_0, y = var_349_promoted)[name = tensor("op_350")]; + tensor var_351 = instance_norm(epsilon = var_295, x = input_29)[name = tensor("op_351")]; + tensor var_352 = mul(x = var_350, y = var_351)[name = tensor("op_352")]; + tensor input_31 = add(x = var_352, y = var_347_1)[name = tensor("input_31")]; + tensor input_33 = leaky_relu(alpha = var_292, x = input_31)[name = tensor("input_33")]; + tensor weight_15_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(14438976))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17584768))), name = tensor("weight_15_palettized"), shape = tensor([1024, 1024, 3])]; + tensor out_3_pad_type_0 = const()[name = tensor("out_3_pad_type_0"), val = tensor("custom")]; + tensor out_3_pad_0 = const()[name = tensor("out_3_pad_0"), val = tensor([1, 1])]; + tensor out_3_strides_0 = const()[name = tensor("out_3_strides_0"), val = tensor([1])]; + tensor out_3_dilations_0 = const()[name = tensor("out_3_dilations_0"), val = tensor([1])]; + tensor out_3_groups_0 = const()[name = tensor("out_3_groups_0"), val = tensor(1)]; + tensor out_3 = conv(bias = decode_0_conv2_bias, dilations = out_3_dilations_0, groups = out_3_groups_0, pad = out_3_pad_0, pad_type = out_3_pad_type_0, strides = out_3_strides_0, weight = weight_15_palettized, x = input_33)[name = tensor("out_3")]; + tensor weight_17_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17585856))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(18702080))), name = tensor("weight_17_palettized"), shape = tensor([1024, 1090, 1])]; + tensor var_368_pad_type_0 = const()[name = tensor("op_368_pad_type_0"), val = tensor("valid")]; + tensor var_368_strides_0 = const()[name = tensor("op_368_strides_0"), val = tensor([1])]; + tensor var_368_pad_0 = const()[name = tensor("op_368_pad_0"), val = tensor([0, 0])]; + tensor var_368_dilations_0 = const()[name = tensor("op_368_dilations_0"), val = tensor([1])]; + tensor var_368_groups_0 = const()[name = tensor("op_368_groups_0"), val = tensor(1)]; + tensor var_368 = conv(dilations = var_368_dilations_0, groups = var_368_groups_0, pad = var_368_pad_0, pad_type = var_368_pad_type_0, strides = var_368_strides_0, weight = weight_17_palettized, x = input_21)[name = tensor("op_368")]; + tensor var_369 = add(x = out_3, y = var_368)[name = tensor("op_369")]; + tensor _inversed_x_3_y_0 = const()[name = tensor("_inversed_x_3_y_0"), val = tensor(0x1.6a09e6p-1)]; + tensor _inversed_x_3 = mul(x = var_369, y = _inversed_x_3_y_0)[name = tensor("_inversed_x_3")]; + tensor var_373 = const()[name = tensor("op_373"), val = tensor(1)]; + tensor input_37_interleave_0 = const()[name = tensor("input_37_interleave_0"), val = tensor(false)]; + tensor input_37 = concat(axis = var_373, interleave = input_37_interleave_0, values = (_inversed_x_3, asr_res_1, F0, Nc))[name = tensor("input_37")]; + tensor var_377 = const()[name = tensor("op_377"), val = tensor(0x1.99999ap-3)]; + tensor var_380 = const()[name = tensor("op_380"), val = tensor(0x1.4f8b58p-17)]; + tensor h_17 = linear(bias = decode_1_norm1_fc_bias, weight = decode_1_norm1_fc_weight_palettized, x = s)[name = tensor("linear_4")]; + tensor var_406 = const()[name = tensor("op_406"), val = tensor([1, 2180, 1])]; + tensor h_19 = reshape(shape = var_406, x = h_17)[name = tensor("h_19")]; + tensor var_408_split_sizes_0 = const()[name = tensor("op_408_split_sizes_0"), val = tensor([1090, 1090])]; + tensor var_408_axis_0 = const()[name = tensor("op_408_axis_0"), val = tensor(1)]; + tensor var_408_0, tensor var_408_1 = split(axis = var_408_axis_0, split_sizes = var_408_split_sizes_0, x = h_19)[name = tensor("op_408")]; + tensor var_410_promoted = const()[name = tensor("op_410_promoted"), val = tensor(0x1p+0)]; + tensor var_411 = add(x = var_408_0, y = var_410_promoted)[name = tensor("op_411")]; + tensor var_412 = instance_norm(epsilon = var_380, x = input_37)[name = tensor("op_412")]; + tensor var_413 = mul(x = var_411, y = var_412)[name = tensor("op_413")]; + tensor input_39 = add(x = var_413, y = var_408_1)[name = tensor("input_39")]; + tensor input_41 = leaky_relu(alpha = var_377, x = input_39)[name = tensor("input_41")]; + tensor weight_19_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(18703168))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(22051712))), name = tensor("weight_19_palettized"), shape = tensor([1024, 1090, 3])]; + tensor input_45_pad_type_0 = const()[name = tensor("input_45_pad_type_0"), val = tensor("custom")]; + tensor input_45_pad_0 = const()[name = tensor("input_45_pad_0"), val = tensor([1, 1])]; + tensor input_45_strides_0 = const()[name = tensor("input_45_strides_0"), val = tensor([1])]; + tensor input_45_dilations_0 = const()[name = tensor("input_45_dilations_0"), val = tensor([1])]; + tensor input_45_groups_0 = const()[name = tensor("input_45_groups_0"), val = tensor(1)]; + tensor input_45 = conv(bias = decode_1_conv1_bias, dilations = input_45_dilations_0, groups = input_45_groups_0, pad = input_45_pad_0, pad_type = input_45_pad_type_0, strides = input_45_strides_0, weight = weight_19_palettized, x = input_41)[name = tensor("input_45")]; + tensor h_21 = linear(bias = decode_1_norm2_fc_bias, weight = decode_1_norm2_fc_weight_palettized, x = s)[name = tensor("linear_5")]; + tensor var_430 = const()[name = tensor("op_430"), val = tensor([1, 2048, 1])]; + tensor h_23 = reshape(shape = var_430, x = h_21)[name = tensor("h_23")]; + tensor var_432_split_sizes_0 = const()[name = tensor("op_432_split_sizes_0"), val = tensor([1024, 1024])]; + tensor var_432_axis_0 = const()[name = tensor("op_432_axis_0"), val = tensor(1)]; + tensor var_432_0, tensor var_432_1 = split(axis = var_432_axis_0, split_sizes = var_432_split_sizes_0, x = h_23)[name = tensor("op_432")]; + tensor var_434_promoted = const()[name = tensor("op_434_promoted"), val = tensor(0x1p+0)]; + tensor var_435 = add(x = var_432_0, y = var_434_promoted)[name = tensor("op_435")]; + tensor var_436 = instance_norm(epsilon = var_380, x = input_45)[name = tensor("op_436")]; + tensor var_437 = mul(x = var_435, y = var_436)[name = tensor("op_437")]; + tensor input_47 = add(x = var_437, y = var_432_1)[name = tensor("input_47")]; + tensor input_49 = leaky_relu(alpha = var_377, x = input_47)[name = tensor("input_49")]; + tensor weight_21_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(22052800))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(25198592))), name = tensor("weight_21_palettized"), shape = tensor([1024, 1024, 3])]; + tensor out_5_pad_type_0 = const()[name = tensor("out_5_pad_type_0"), val = tensor("custom")]; + tensor out_5_pad_0 = const()[name = tensor("out_5_pad_0"), val = tensor([1, 1])]; + tensor out_5_strides_0 = const()[name = tensor("out_5_strides_0"), val = tensor([1])]; + tensor out_5_dilations_0 = const()[name = tensor("out_5_dilations_0"), val = tensor([1])]; + tensor out_5_groups_0 = const()[name = tensor("out_5_groups_0"), val = tensor(1)]; + tensor out_5 = conv(bias = decode_1_conv2_bias, dilations = out_5_dilations_0, groups = out_5_groups_0, pad = out_5_pad_0, pad_type = out_5_pad_type_0, strides = out_5_strides_0, weight = weight_21_palettized, x = input_49)[name = tensor("out_5")]; + tensor weight_23_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(25199680))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(26315904))), name = tensor("weight_23_palettized"), shape = tensor([1024, 1090, 1])]; + tensor var_453_pad_type_0 = const()[name = tensor("op_453_pad_type_0"), val = tensor("valid")]; + tensor var_453_strides_0 = const()[name = tensor("op_453_strides_0"), val = tensor([1])]; + tensor var_453_pad_0 = const()[name = tensor("op_453_pad_0"), val = tensor([0, 0])]; + tensor var_453_dilations_0 = const()[name = tensor("op_453_dilations_0"), val = tensor([1])]; + tensor var_453_groups_0 = const()[name = tensor("op_453_groups_0"), val = tensor(1)]; + tensor var_453 = conv(dilations = var_453_dilations_0, groups = var_453_groups_0, pad = var_453_pad_0, pad_type = var_453_pad_type_0, strides = var_453_strides_0, weight = weight_23_palettized, x = input_37)[name = tensor("op_453")]; + tensor var_454 = add(x = out_5, y = var_453)[name = tensor("op_454")]; + tensor _inversed_x_5_y_0 = const()[name = tensor("_inversed_x_5_y_0"), val = tensor(0x1.6a09e6p-1)]; + tensor _inversed_x_5 = mul(x = var_454, y = _inversed_x_5_y_0)[name = tensor("_inversed_x_5")]; + tensor var_458 = const()[name = tensor("op_458"), val = tensor(1)]; + tensor input_53_interleave_0 = const()[name = tensor("input_53_interleave_0"), val = tensor(false)]; + tensor input_53 = concat(axis = var_458, interleave = input_53_interleave_0, values = (_inversed_x_5, asr_res_1, F0, Nc))[name = tensor("input_53")]; + tensor var_462 = const()[name = tensor("op_462"), val = tensor(0x1.99999ap-3)]; + tensor var_465 = const()[name = tensor("op_465"), val = tensor(0x1.4f8b58p-17)]; + tensor h_25 = linear(bias = decode_2_norm1_fc_bias, weight = decode_2_norm1_fc_weight_palettized, x = s)[name = tensor("linear_6")]; + tensor var_491 = const()[name = tensor("op_491"), val = tensor([1, 2180, 1])]; + tensor h_27 = reshape(shape = var_491, x = h_25)[name = tensor("h_27")]; + tensor var_493_split_sizes_0 = const()[name = tensor("op_493_split_sizes_0"), val = tensor([1090, 1090])]; + tensor var_493_axis_0 = const()[name = tensor("op_493_axis_0"), val = tensor(1)]; + tensor var_493_0, tensor var_493_1 = split(axis = var_493_axis_0, split_sizes = var_493_split_sizes_0, x = h_27)[name = tensor("op_493")]; + tensor var_495_promoted = const()[name = tensor("op_495_promoted"), val = tensor(0x1p+0)]; + tensor var_496 = add(x = var_493_0, y = var_495_promoted)[name = tensor("op_496")]; + tensor var_497 = instance_norm(epsilon = var_465, x = input_53)[name = tensor("op_497")]; + tensor var_498 = mul(x = var_496, y = var_497)[name = tensor("op_498")]; + tensor input_55 = add(x = var_498, y = var_493_1)[name = tensor("input_55")]; + tensor input_57 = leaky_relu(alpha = var_462, x = input_55)[name = tensor("input_57")]; + tensor weight_25_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(26316992))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(29665536))), name = tensor("weight_25_palettized"), shape = tensor([1024, 1090, 3])]; + tensor input_61_pad_type_0 = const()[name = tensor("input_61_pad_type_0"), val = tensor("custom")]; + tensor input_61_pad_0 = const()[name = tensor("input_61_pad_0"), val = tensor([1, 1])]; + tensor input_61_strides_0 = const()[name = tensor("input_61_strides_0"), val = tensor([1])]; + tensor input_61_dilations_0 = const()[name = tensor("input_61_dilations_0"), val = tensor([1])]; + tensor input_61_groups_0 = const()[name = tensor("input_61_groups_0"), val = tensor(1)]; + tensor input_61 = conv(bias = decode_2_conv1_bias, dilations = input_61_dilations_0, groups = input_61_groups_0, pad = input_61_pad_0, pad_type = input_61_pad_type_0, strides = input_61_strides_0, weight = weight_25_palettized, x = input_57)[name = tensor("input_61")]; + tensor h_29 = linear(bias = decode_2_norm2_fc_bias, weight = decode_2_norm2_fc_weight_palettized, x = s)[name = tensor("linear_7")]; + tensor var_515 = const()[name = tensor("op_515"), val = tensor([1, 2048, 1])]; + tensor h_31 = reshape(shape = var_515, x = h_29)[name = tensor("h_31")]; + tensor var_517_split_sizes_0 = const()[name = tensor("op_517_split_sizes_0"), val = tensor([1024, 1024])]; + tensor var_517_axis_0 = const()[name = tensor("op_517_axis_0"), val = tensor(1)]; + tensor var_517_0, tensor var_517_1 = split(axis = var_517_axis_0, split_sizes = var_517_split_sizes_0, x = h_31)[name = tensor("op_517")]; + tensor var_519_promoted = const()[name = tensor("op_519_promoted"), val = tensor(0x1p+0)]; + tensor var_520 = add(x = var_517_0, y = var_519_promoted)[name = tensor("op_520")]; + tensor var_521 = instance_norm(epsilon = var_465, x = input_61)[name = tensor("op_521")]; + tensor var_522 = mul(x = var_520, y = var_521)[name = tensor("op_522")]; + tensor input_63 = add(x = var_522, y = var_517_1)[name = tensor("input_63")]; + tensor input_65 = leaky_relu(alpha = var_462, x = input_63)[name = tensor("input_65")]; + tensor weight_27_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(29666624))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(32812416))), name = tensor("weight_27_palettized"), shape = tensor([1024, 1024, 3])]; + tensor out_7_pad_type_0 = const()[name = tensor("out_7_pad_type_0"), val = tensor("custom")]; + tensor out_7_pad_0 = const()[name = tensor("out_7_pad_0"), val = tensor([1, 1])]; + tensor out_7_strides_0 = const()[name = tensor("out_7_strides_0"), val = tensor([1])]; + tensor out_7_dilations_0 = const()[name = tensor("out_7_dilations_0"), val = tensor([1])]; + tensor out_7_groups_0 = const()[name = tensor("out_7_groups_0"), val = tensor(1)]; + tensor out_7 = conv(bias = decode_2_conv2_bias, dilations = out_7_dilations_0, groups = out_7_groups_0, pad = out_7_pad_0, pad_type = out_7_pad_type_0, strides = out_7_strides_0, weight = weight_27_palettized, x = input_65)[name = tensor("out_7")]; + tensor weight_29_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(32813504))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(33929728))), name = tensor("weight_29_palettized"), shape = tensor([1024, 1090, 1])]; + tensor var_538_pad_type_0 = const()[name = tensor("op_538_pad_type_0"), val = tensor("valid")]; + tensor var_538_strides_0 = const()[name = tensor("op_538_strides_0"), val = tensor([1])]; + tensor var_538_pad_0 = const()[name = tensor("op_538_pad_0"), val = tensor([0, 0])]; + tensor var_538_dilations_0 = const()[name = tensor("op_538_dilations_0"), val = tensor([1])]; + tensor var_538_groups_0 = const()[name = tensor("op_538_groups_0"), val = tensor(1)]; + tensor var_538 = conv(dilations = var_538_dilations_0, groups = var_538_groups_0, pad = var_538_pad_0, pad_type = var_538_pad_type_0, strides = var_538_strides_0, weight = weight_29_palettized, x = input_53)[name = tensor("op_538")]; + tensor var_539 = add(x = out_7, y = var_538)[name = tensor("op_539")]; + tensor _inversed_x_7_y_0 = const()[name = tensor("_inversed_x_7_y_0"), val = tensor(0x1.6a09e6p-1)]; + tensor _inversed_x_7 = mul(x = var_539, y = _inversed_x_7_y_0)[name = tensor("_inversed_x_7")]; + tensor var_543 = const()[name = tensor("op_543"), val = tensor(1)]; + tensor input_69_interleave_0 = const()[name = tensor("input_69_interleave_0"), val = tensor(false)]; + tensor input_69 = concat(axis = var_543, interleave = input_69_interleave_0, values = (_inversed_x_7, asr_res_1, F0, Nc))[name = tensor("input_69")]; + tensor var_549 = const()[name = tensor("op_549"), val = tensor(0x1.99999ap-3)]; + tensor var_553 = const()[name = tensor("op_553"), val = tensor(0x1.4f8b58p-17)]; + tensor h_33 = linear(bias = decode_3_norm1_fc_bias, weight = decode_3_norm1_fc_weight_palettized, x = s)[name = tensor("linear_8")]; + tensor var_584 = const()[name = tensor("op_584"), val = tensor([1, 2180, 1])]; + tensor h_35 = reshape(shape = var_584, x = h_33)[name = tensor("h_35")]; + tensor var_586_split_sizes_0 = const()[name = tensor("op_586_split_sizes_0"), val = tensor([1090, 1090])]; + tensor var_586_axis_0 = const()[name = tensor("op_586_axis_0"), val = tensor(1)]; + tensor var_586_0, tensor var_586_1 = split(axis = var_586_axis_0, split_sizes = var_586_split_sizes_0, x = h_35)[name = tensor("op_586")]; + tensor var_588_promoted = const()[name = tensor("op_588_promoted"), val = tensor(0x1p+0)]; + tensor var_589 = add(x = var_586_0, y = var_588_promoted)[name = tensor("op_589")]; + tensor var_590 = instance_norm(epsilon = var_553, x = input_69)[name = tensor("op_590")]; + tensor var_591 = mul(x = var_589, y = var_590)[name = tensor("op_591")]; + tensor input_71 = add(x = var_591, y = var_586_1)[name = tensor("input_71")]; + tensor input_73 = leaky_relu(alpha = var_549, x = input_71)[name = tensor("input_73")]; + tensor op_594_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(33930816))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(33934208))), name = tensor("op_594_palettized"), shape = tensor([1090, 1, 3])]; + tensor conv_transpose_0_pad_type_0 = const()[name = tensor("conv_transpose_0_pad_type_0"), val = tensor("custom")]; + tensor conv_transpose_0_pad_0 = const()[name = tensor("conv_transpose_0_pad_0"), val = tensor([0, 0])]; + tensor conv_transpose_0_strides_0 = const()[name = tensor("conv_transpose_0_strides_0"), val = tensor([2])]; + tensor conv_transpose_0_groups_0 = const()[name = tensor("conv_transpose_0_groups_0"), val = tensor(1090)]; + tensor conv_transpose_0_dilations_0 = const()[name = tensor("conv_transpose_0_dilations_0"), val = tensor([1])]; + tensor conv_transpose_0 = conv_transpose(bias = decode_3_pool_bias, dilations = conv_transpose_0_dilations_0, groups = conv_transpose_0_groups_0, pad = conv_transpose_0_pad_0, pad_type = conv_transpose_0_pad_type_0, strides = conv_transpose_0_strides_0, weight = op_594_palettized, x = input_73)[name = tensor("conv_transpose_0")]; + tensor input_75_begin_0 = const()[name = tensor("input_75_begin_0"), val = tensor([0, 0, 1])]; + tensor input_75_end_0 = const()[name = tensor("input_75_end_0"), val = tensor([0, 0, 0])]; + tensor input_75_begin_mask_0 = const()[name = tensor("input_75_begin_mask_0"), val = tensor([true, true, false])]; + tensor input_75_end_mask_0 = const()[name = tensor("input_75_end_mask_0"), val = tensor([true, true, true])]; + tensor input_75 = slice_by_index(begin = input_75_begin_0, begin_mask = input_75_begin_mask_0, end = input_75_end_0, end_mask = input_75_end_mask_0, x = conv_transpose_0)[name = tensor("input_75")]; + tensor weight_31_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(33935296))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(35609600))), name = tensor("weight_31_palettized"), shape = tensor([512, 1090, 3])]; + tensor input_79_pad_type_0 = const()[name = tensor("input_79_pad_type_0"), val = tensor("custom")]; + tensor input_79_pad_0 = const()[name = tensor("input_79_pad_0"), val = tensor([1, 1])]; + tensor input_79_strides_0 = const()[name = tensor("input_79_strides_0"), val = tensor([1])]; + tensor input_79_dilations_0 = const()[name = tensor("input_79_dilations_0"), val = tensor([1])]; + tensor input_79_groups_0 = const()[name = tensor("input_79_groups_0"), val = tensor(1)]; + tensor input_79 = conv(bias = decode_3_conv1_bias, dilations = input_79_dilations_0, groups = input_79_groups_0, pad = input_79_pad_0, pad_type = input_79_pad_type_0, strides = input_79_strides_0, weight = weight_31_palettized, x = input_75)[name = tensor("input_79")]; + tensor h_37 = linear(bias = decode_3_norm2_fc_bias, weight = decode_3_norm2_fc_weight_palettized, x = s)[name = tensor("linear_9")]; + tensor var_615 = const()[name = tensor("op_615"), val = tensor([1, 1024, 1])]; + tensor h_39 = reshape(shape = var_615, x = h_37)[name = tensor("h_39")]; + tensor var_617_split_sizes_0 = const()[name = tensor("op_617_split_sizes_0"), val = tensor([512, 512])]; + tensor var_617_axis_0 = const()[name = tensor("op_617_axis_0"), val = tensor(1)]; + tensor var_617_0, tensor var_617_1 = split(axis = var_617_axis_0, split_sizes = var_617_split_sizes_0, x = h_39)[name = tensor("op_617")]; + tensor var_619_promoted = const()[name = tensor("op_619_promoted"), val = tensor(0x1p+0)]; + tensor var_620 = add(x = var_617_0, y = var_619_promoted)[name = tensor("op_620")]; + tensor var_621 = instance_norm(epsilon = var_553, x = input_79)[name = tensor("op_621")]; + tensor var_622 = mul(x = var_620, y = var_621)[name = tensor("op_622")]; + tensor input_81 = add(x = var_622, y = var_617_1)[name = tensor("input_81")]; + tensor input_83 = leaky_relu(alpha = var_549, x = input_81)[name = tensor("input_83")]; + tensor weight_33_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(35610688))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(36397184))), name = tensor("weight_33_palettized"), shape = tensor([512, 512, 3])]; + tensor out_pad_type_0 = const()[name = tensor("out_pad_type_0"), val = tensor("custom")]; + tensor out_pad_0 = const()[name = tensor("out_pad_0"), val = tensor([1, 1])]; + tensor out_strides_0 = const()[name = tensor("out_strides_0"), val = tensor([1])]; + tensor out_dilations_0 = const()[name = tensor("out_dilations_0"), val = tensor([1])]; + tensor out_groups_0 = const()[name = tensor("out_groups_0"), val = tensor(1)]; + tensor out = conv(bias = decode_3_conv2_bias, dilations = out_dilations_0, groups = out_groups_0, pad = out_pad_0, pad_type = out_pad_type_0, strides = out_strides_0, weight = weight_33_palettized, x = input_83)[name = tensor("out")]; + tensor expand_dims_0_axes_0 = const()[name = tensor("expand_dims_0_axes_0"), val = tensor([3])]; + tensor expand_dims_0 = expand_dims(axes = expand_dims_0_axes_0, x = input_69)[name = tensor("expand_dims_0")]; + tensor upsample_nearest_neighbor_0_scale_factor_height_0 = const()[name = tensor("upsample_nearest_neighbor_0_scale_factor_height_0"), val = tensor(2)]; + tensor upsample_nearest_neighbor_0_scale_factor_width_0 = const()[name = tensor("upsample_nearest_neighbor_0_scale_factor_width_0"), val = tensor(1)]; + tensor upsample_nearest_neighbor_0 = upsample_nearest_neighbor(scale_factor_height = upsample_nearest_neighbor_0_scale_factor_height_0, scale_factor_width = upsample_nearest_neighbor_0_scale_factor_width_0, x = expand_dims_0)[name = tensor("upsample_nearest_neighbor_0")]; + tensor input_87_axes_0 = const()[name = tensor("input_87_axes_0"), val = tensor([3])]; + tensor input_87 = squeeze(axes = input_87_axes_0, x = upsample_nearest_neighbor_0)[name = tensor("input_87")]; + tensor weight_35_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(36398272))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(36956416))), name = tensor("weight_35_palettized"), shape = tensor([512, 1090, 1])]; + tensor var_640_pad_type_0 = const()[name = tensor("op_640_pad_type_0"), val = tensor("valid")]; + tensor var_640_strides_0 = const()[name = tensor("op_640_strides_0"), val = tensor([1])]; + tensor var_640_pad_0 = const()[name = tensor("op_640_pad_0"), val = tensor([0, 0])]; + tensor var_640_dilations_0 = const()[name = tensor("op_640_dilations_0"), val = tensor([1])]; + tensor var_640_groups_0 = const()[name = tensor("op_640_groups_0"), val = tensor(1)]; + tensor var_640 = conv(dilations = var_640_dilations_0, groups = var_640_groups_0, pad = var_640_pad_0, pad_type = var_640_pad_type_0, strides = var_640_strides_0, weight = weight_35_palettized, x = input_87)[name = tensor("op_640")]; + tensor var_641 = add(x = out, y = var_640)[name = tensor("op_641")]; + tensor _inversed_z_1_y_0 = const()[name = tensor("_inversed_z_1_y_0"), val = tensor(0x1.6a09e6p-1)]; + tensor _inversed_z_1 = mul(x = var_641, y = _inversed_z_1_y_0)[name = tensor("_inversed_z_1")]; + tensor input_89 = linear(bias = generator_m_source_l_linear_bias, weight = generator_m_source_l_linear_weight, x = sine_waves)[name = tensor("linear_10")]; + tensor sine_merge = tanh(x = input_89)[name = tensor("sine_merge")]; + tensor input_91_perm_0 = const()[name = tensor("input_91_perm_0"), val = tensor([0, 2, 1])]; + tensor var_651 = mul(x = _inversed_z_1, y = generator_alphas_0)[name = tensor("op_651")]; + tensor sv_1 = sin(x = var_651)[name = tensor("sv_1")]; + tensor var_653 = mul(x = sv_1, y = sv_1)[name = tensor("op_653")]; + tensor _inversed_654_y_0 = const()[name = tensor("_inversed_654_y_0"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(36957504)))]; + tensor _inversed_654 = mul(x = var_653, y = _inversed_654_y_0)[name = tensor("_inversed_654")]; + tensor input_117 = add(x = _inversed_z_1, y = _inversed_654)[name = tensor("input_117")]; + tensor input_93_pad_type_0 = const()[name = tensor("input_93_pad_type_0"), val = tensor("custom")]; + tensor input_93_pad_0 = const()[name = tensor("input_93_pad_0"), val = tensor([15, 15])]; + tensor input_93_strides_0 = const()[name = tensor("input_93_strides_0"), val = tensor([30])]; + tensor input_93_dilations_0 = const()[name = tensor("input_93_dilations_0"), val = tensor([1])]; + tensor input_93_groups_0 = const()[name = tensor("input_93_groups_0"), val = tensor(1)]; + tensor input_91 = transpose(perm = input_91_perm_0, x = sine_merge)[name = tensor("transpose_0")]; + tensor input_93 = conv(bias = generator_noise_convs_0_bias, dilations = input_93_dilations_0, groups = input_93_groups_0, pad = input_93_pad_0, pad_type = input_93_pad_type_0, strides = input_93_strides_0, weight = generator_noise_convs_0_weight_palettized, x = input_91)[name = tensor("input_93")]; + tensor var_677 = const()[name = tensor("op_677"), val = tensor(0x1.4f8b58p-17)]; + tensor h_41 = linear(bias = generator_noise_res_0_adain1_0_fc_bias, weight = generator_noise_res_0_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_11")]; + tensor var_759 = const()[name = tensor("op_759"), val = tensor([1, 512, 1])]; + tensor h_43 = reshape(shape = var_759, x = h_41)[name = tensor("h_43")]; + tensor var_761_split_sizes_0 = const()[name = tensor("op_761_split_sizes_0"), val = tensor([256, 256])]; + tensor var_761_axis_0 = const()[name = tensor("op_761_axis_0"), val = tensor(1)]; + tensor var_761_0, tensor var_761_1 = split(axis = var_761_axis_0, split_sizes = var_761_split_sizes_0, x = h_43)[name = tensor("op_761")]; + tensor var_763_promoted = const()[name = tensor("op_763_promoted"), val = tensor(0x1p+0)]; + tensor var_764 = add(x = var_761_0, y = var_763_promoted)[name = tensor("op_764")]; + tensor var_765 = instance_norm(epsilon = var_677, x = input_93)[name = tensor("op_765")]; + tensor var_766 = mul(x = var_764, y = var_765)[name = tensor("op_766")]; + tensor xt_1 = add(x = var_766, y = var_761_1)[name = tensor("xt_1")]; + tensor var_769 = const()[name = tensor("op_769"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(36959616)))]; + tensor var_770 = mul(x = xt_1, y = var_769)[name = tensor("op_770")]; + tensor cv_1 = cos(x = var_770)[name = tensor("cv_1")]; + tensor var_772 = const()[name = tensor("op_772"), val = tensor(-0x1p-1)]; + tensor var_773 = mul(x = cv_1, y = var_772)[name = tensor("op_773")]; + tensor var_774 = const()[name = tensor("op_774"), val = tensor(0x1p-1)]; + tensor var_775 = add(x = var_773, y = var_774)[name = tensor("op_775")]; + tensor var_776 = const()[name = tensor("op_776"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(36960704)))]; + tensor var_779 = mul(x = var_775, y = var_776)[name = tensor("op_779")]; + tensor input_95 = add(x = xt_1, y = var_779)[name = tensor("input_95")]; + tensor weight_39_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(36961792))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37420608))), name = tensor("weight_39_palettized"), shape = tensor([256, 256, 7])]; + tensor input_97_pad_type_0 = const()[name = tensor("input_97_pad_type_0"), val = tensor("custom")]; + tensor input_97_pad_0 = const()[name = tensor("input_97_pad_0"), val = tensor([3, 3])]; + tensor input_97_strides_0 = const()[name = tensor("input_97_strides_0"), val = tensor([1])]; + tensor input_97_dilations_0 = const()[name = tensor("input_97_dilations_0"), val = tensor([1])]; + tensor input_97_groups_0 = const()[name = tensor("input_97_groups_0"), val = tensor(1)]; + tensor input_97 = conv(bias = generator_noise_res_0_convs1_0_bias, dilations = input_97_dilations_0, groups = input_97_groups_0, pad = input_97_pad_0, pad_type = input_97_pad_type_0, strides = input_97_strides_0, weight = weight_39_palettized, x = input_95)[name = tensor("input_97")]; + tensor h_45 = linear(bias = generator_noise_res_0_adain2_0_fc_bias, weight = generator_noise_res_0_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_12")]; + tensor var_794 = const()[name = tensor("op_794"), val = tensor([1, 512, 1])]; + tensor h_47 = reshape(shape = var_794, x = h_45)[name = tensor("h_47")]; + tensor var_796_split_sizes_0 = const()[name = tensor("op_796_split_sizes_0"), val = tensor([256, 256])]; + tensor var_796_axis_0 = const()[name = tensor("op_796_axis_0"), val = tensor(1)]; + tensor var_796_0, tensor var_796_1 = split(axis = var_796_axis_0, split_sizes = var_796_split_sizes_0, x = h_47)[name = tensor("op_796")]; + tensor var_798_promoted = const()[name = tensor("op_798_promoted"), val = tensor(0x1p+0)]; + tensor var_799 = add(x = var_796_0, y = var_798_promoted)[name = tensor("op_799")]; + tensor var_800 = instance_norm(epsilon = var_677, x = input_97)[name = tensor("op_800")]; + tensor var_801 = mul(x = var_799, y = var_800)[name = tensor("op_801")]; + tensor xt_3 = add(x = var_801, y = var_796_1)[name = tensor("xt_3")]; + tensor var_804 = const()[name = tensor("op_804"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37421696)))]; + tensor var_805 = mul(x = xt_3, y = var_804)[name = tensor("op_805")]; + tensor cv_3 = cos(x = var_805)[name = tensor("cv_3")]; + tensor var_807 = const()[name = tensor("op_807"), val = tensor(-0x1p-1)]; + tensor var_808 = mul(x = cv_3, y = var_807)[name = tensor("op_808")]; + tensor var_809 = const()[name = tensor("op_809"), val = tensor(0x1p-1)]; + tensor var_810 = add(x = var_808, y = var_809)[name = tensor("op_810")]; + tensor var_811 = const()[name = tensor("op_811"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37422784)))]; + tensor var_814 = mul(x = var_810, y = var_811)[name = tensor("op_814")]; + tensor input_99 = add(x = xt_3, y = var_814)[name = tensor("input_99")]; + tensor weight_41_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37423872))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37882688))), name = tensor("weight_41_palettized"), shape = tensor([256, 256, 7])]; + tensor xt_5_pad_type_0 = const()[name = tensor("xt_5_pad_type_0"), val = tensor("custom")]; + tensor xt_5_pad_0 = const()[name = tensor("xt_5_pad_0"), val = tensor([3, 3])]; + tensor xt_5_strides_0 = const()[name = tensor("xt_5_strides_0"), val = tensor([1])]; + tensor xt_5_dilations_0 = const()[name = tensor("xt_5_dilations_0"), val = tensor([1])]; + tensor xt_5_groups_0 = const()[name = tensor("xt_5_groups_0"), val = tensor(1)]; + tensor xt_5 = conv(bias = generator_noise_res_0_convs2_0_bias, dilations = xt_5_dilations_0, groups = xt_5_groups_0, pad = xt_5_pad_0, pad_type = xt_5_pad_type_0, strides = xt_5_strides_0, weight = weight_41_palettized, x = input_99)[name = tensor("xt_5")]; + tensor input_101 = add(x = xt_5, y = input_93)[name = tensor("input_101")]; + tensor h_49 = linear(bias = generator_noise_res_0_adain1_1_fc_bias, weight = generator_noise_res_0_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_13")]; + tensor var_830 = const()[name = tensor("op_830"), val = tensor([1, 512, 1])]; + tensor h_51 = reshape(shape = var_830, x = h_49)[name = tensor("h_51")]; + tensor var_832_split_sizes_0 = const()[name = tensor("op_832_split_sizes_0"), val = tensor([256, 256])]; + tensor var_832_axis_0 = const()[name = tensor("op_832_axis_0"), val = tensor(1)]; + tensor var_832_0, tensor var_832_1 = split(axis = var_832_axis_0, split_sizes = var_832_split_sizes_0, x = h_51)[name = tensor("op_832")]; + tensor var_834_promoted = const()[name = tensor("op_834_promoted"), val = tensor(0x1p+0)]; + tensor var_835 = add(x = var_832_0, y = var_834_promoted)[name = tensor("op_835")]; + tensor var_836 = instance_norm(epsilon = var_677, x = input_101)[name = tensor("op_836")]; + tensor var_837 = mul(x = var_835, y = var_836)[name = tensor("op_837")]; + tensor xt_7 = add(x = var_837, y = var_832_1)[name = tensor("xt_7")]; + tensor var_840 = const()[name = tensor("op_840"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37883776)))]; + tensor var_841 = mul(x = xt_7, y = var_840)[name = tensor("op_841")]; + tensor cv_5 = cos(x = var_841)[name = tensor("cv_5")]; + tensor var_843 = const()[name = tensor("op_843"), val = tensor(-0x1p-1)]; + tensor var_844 = mul(x = cv_5, y = var_843)[name = tensor("op_844")]; + tensor var_845 = const()[name = tensor("op_845"), val = tensor(0x1p-1)]; + tensor var_846 = add(x = var_844, y = var_845)[name = tensor("op_846")]; + tensor var_847 = const()[name = tensor("op_847"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37884864)))]; + tensor var_850 = mul(x = var_846, y = var_847)[name = tensor("op_850")]; + tensor input_103 = add(x = xt_7, y = var_850)[name = tensor("input_103")]; + tensor weight_43_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37885952))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38344768))), name = tensor("weight_43_palettized"), shape = tensor([256, 256, 7])]; + tensor input_105_pad_type_0 = const()[name = tensor("input_105_pad_type_0"), val = tensor("custom")]; + tensor input_105_pad_0 = const()[name = tensor("input_105_pad_0"), val = tensor([9, 9])]; + tensor input_105_dilations_0 = const()[name = tensor("input_105_dilations_0"), val = tensor([3])]; + tensor input_105_strides_0 = const()[name = tensor("input_105_strides_0"), val = tensor([1])]; + tensor input_105_groups_0 = const()[name = tensor("input_105_groups_0"), val = tensor(1)]; + tensor input_105 = conv(bias = generator_noise_res_0_convs1_1_bias, dilations = input_105_dilations_0, groups = input_105_groups_0, pad = input_105_pad_0, pad_type = input_105_pad_type_0, strides = input_105_strides_0, weight = weight_43_palettized, x = input_103)[name = tensor("input_105")]; + tensor h_53 = linear(bias = generator_noise_res_0_adain2_1_fc_bias, weight = generator_noise_res_0_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_14")]; + tensor var_865 = const()[name = tensor("op_865"), val = tensor([1, 512, 1])]; + tensor h_55 = reshape(shape = var_865, x = h_53)[name = tensor("h_55")]; + tensor var_867_split_sizes_0 = const()[name = tensor("op_867_split_sizes_0"), val = tensor([256, 256])]; + tensor var_867_axis_0 = const()[name = tensor("op_867_axis_0"), val = tensor(1)]; + tensor var_867_0, tensor var_867_1 = split(axis = var_867_axis_0, split_sizes = var_867_split_sizes_0, x = h_55)[name = tensor("op_867")]; + tensor var_869_promoted = const()[name = tensor("op_869_promoted"), val = tensor(0x1p+0)]; + tensor var_870 = add(x = var_867_0, y = var_869_promoted)[name = tensor("op_870")]; + tensor var_871 = instance_norm(epsilon = var_677, x = input_105)[name = tensor("op_871")]; + tensor var_872 = mul(x = var_870, y = var_871)[name = tensor("op_872")]; + tensor xt_9 = add(x = var_872, y = var_867_1)[name = tensor("xt_9")]; + tensor var_875 = const()[name = tensor("op_875"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38345856)))]; + tensor var_876 = mul(x = xt_9, y = var_875)[name = tensor("op_876")]; + tensor cv_7 = cos(x = var_876)[name = tensor("cv_7")]; + tensor var_878 = const()[name = tensor("op_878"), val = tensor(-0x1p-1)]; + tensor var_879 = mul(x = cv_7, y = var_878)[name = tensor("op_879")]; + tensor var_880 = const()[name = tensor("op_880"), val = tensor(0x1p-1)]; + tensor var_881 = add(x = var_879, y = var_880)[name = tensor("op_881")]; + tensor var_882 = const()[name = tensor("op_882"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38346944)))]; + tensor var_885 = mul(x = var_881, y = var_882)[name = tensor("op_885")]; + tensor input_107 = add(x = xt_9, y = var_885)[name = tensor("input_107")]; + tensor weight_45_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38348032))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38806848))), name = tensor("weight_45_palettized"), shape = tensor([256, 256, 7])]; + tensor xt_11_pad_type_0 = const()[name = tensor("xt_11_pad_type_0"), val = tensor("custom")]; + tensor xt_11_pad_0 = const()[name = tensor("xt_11_pad_0"), val = tensor([3, 3])]; + tensor xt_11_strides_0 = const()[name = tensor("xt_11_strides_0"), val = tensor([1])]; + tensor xt_11_dilations_0 = const()[name = tensor("xt_11_dilations_0"), val = tensor([1])]; + tensor xt_11_groups_0 = const()[name = tensor("xt_11_groups_0"), val = tensor(1)]; + tensor xt_11 = conv(bias = generator_noise_res_0_convs2_1_bias, dilations = xt_11_dilations_0, groups = xt_11_groups_0, pad = xt_11_pad_0, pad_type = xt_11_pad_type_0, strides = xt_11_strides_0, weight = weight_45_palettized, x = input_107)[name = tensor("xt_11")]; + tensor input_109 = add(x = xt_11, y = input_101)[name = tensor("input_109")]; + tensor h_57 = linear(bias = generator_noise_res_0_adain1_2_fc_bias, weight = generator_noise_res_0_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_15")]; + tensor var_901 = const()[name = tensor("op_901"), val = tensor([1, 512, 1])]; + tensor h_59 = reshape(shape = var_901, x = h_57)[name = tensor("h_59")]; + tensor var_903_split_sizes_0 = const()[name = tensor("op_903_split_sizes_0"), val = tensor([256, 256])]; + tensor var_903_axis_0 = const()[name = tensor("op_903_axis_0"), val = tensor(1)]; + tensor var_903_0, tensor var_903_1 = split(axis = var_903_axis_0, split_sizes = var_903_split_sizes_0, x = h_59)[name = tensor("op_903")]; + tensor var_905_promoted = const()[name = tensor("op_905_promoted"), val = tensor(0x1p+0)]; + tensor var_906 = add(x = var_903_0, y = var_905_promoted)[name = tensor("op_906")]; + tensor var_907 = instance_norm(epsilon = var_677, x = input_109)[name = tensor("op_907")]; + tensor var_908 = mul(x = var_906, y = var_907)[name = tensor("op_908")]; + tensor xt_13 = add(x = var_908, y = var_903_1)[name = tensor("xt_13")]; + tensor var_911 = const()[name = tensor("op_911"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38807936)))]; + tensor var_912 = mul(x = xt_13, y = var_911)[name = tensor("op_912")]; + tensor cv_9 = cos(x = var_912)[name = tensor("cv_9")]; + tensor var_914 = const()[name = tensor("op_914"), val = tensor(-0x1p-1)]; + tensor var_915 = mul(x = cv_9, y = var_914)[name = tensor("op_915")]; + tensor var_916 = const()[name = tensor("op_916"), val = tensor(0x1p-1)]; + tensor var_917 = add(x = var_915, y = var_916)[name = tensor("op_917")]; + tensor var_918 = const()[name = tensor("op_918"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38809024)))]; + tensor var_921 = mul(x = var_917, y = var_918)[name = tensor("op_921")]; + tensor input_111 = add(x = xt_13, y = var_921)[name = tensor("input_111")]; + tensor weight_47_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38810112))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(39268928))), name = tensor("weight_47_palettized"), shape = tensor([256, 256, 7])]; + tensor input_113_pad_type_0 = const()[name = tensor("input_113_pad_type_0"), val = tensor("custom")]; + tensor input_113_pad_0 = const()[name = tensor("input_113_pad_0"), val = tensor([15, 15])]; + tensor input_113_dilations_0 = const()[name = tensor("input_113_dilations_0"), val = tensor([5])]; + tensor input_113_strides_0 = const()[name = tensor("input_113_strides_0"), val = tensor([1])]; + tensor input_113_groups_0 = const()[name = tensor("input_113_groups_0"), val = tensor(1)]; + tensor input_113 = conv(bias = generator_noise_res_0_convs1_2_bias, dilations = input_113_dilations_0, groups = input_113_groups_0, pad = input_113_pad_0, pad_type = input_113_pad_type_0, strides = input_113_strides_0, weight = weight_47_palettized, x = input_111)[name = tensor("input_113")]; + tensor h_61 = linear(bias = generator_noise_res_0_adain2_2_fc_bias, weight = generator_noise_res_0_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_16")]; + tensor var_936 = const()[name = tensor("op_936"), val = tensor([1, 512, 1])]; + tensor h_63 = reshape(shape = var_936, x = h_61)[name = tensor("h_63")]; + tensor var_938_split_sizes_0 = const()[name = tensor("op_938_split_sizes_0"), val = tensor([256, 256])]; + tensor var_938_axis_0 = const()[name = tensor("op_938_axis_0"), val = tensor(1)]; + tensor var_938_0, tensor var_938_1 = split(axis = var_938_axis_0, split_sizes = var_938_split_sizes_0, x = h_63)[name = tensor("op_938")]; + tensor var_940_promoted = const()[name = tensor("op_940_promoted"), val = tensor(0x1p+0)]; + tensor var_941 = add(x = var_938_0, y = var_940_promoted)[name = tensor("op_941")]; + tensor var_942 = instance_norm(epsilon = var_677, x = input_113)[name = tensor("op_942")]; + tensor var_943 = mul(x = var_941, y = var_942)[name = tensor("op_943")]; + tensor xt_15 = add(x = var_943, y = var_938_1)[name = tensor("xt_15")]; + tensor var_946 = const()[name = tensor("op_946"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(39270016)))]; + tensor var_947 = mul(x = xt_15, y = var_946)[name = tensor("op_947")]; + tensor cv_11 = cos(x = var_947)[name = tensor("cv_11")]; + tensor var_949 = const()[name = tensor("op_949"), val = tensor(-0x1p-1)]; + tensor var_950 = mul(x = cv_11, y = var_949)[name = tensor("op_950")]; + tensor var_951 = const()[name = tensor("op_951"), val = tensor(0x1p-1)]; + tensor var_952 = add(x = var_950, y = var_951)[name = tensor("op_952")]; + tensor var_953 = const()[name = tensor("op_953"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(39271104)))]; + tensor var_956 = mul(x = var_952, y = var_953)[name = tensor("op_956")]; + tensor input_115 = add(x = xt_15, y = var_956)[name = tensor("input_115")]; + tensor weight_49_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(39272192))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(39731008))), name = tensor("weight_49_palettized"), shape = tensor([256, 256, 7])]; + tensor xt_17_pad_type_0 = const()[name = tensor("xt_17_pad_type_0"), val = tensor("custom")]; + tensor xt_17_pad_0 = const()[name = tensor("xt_17_pad_0"), val = tensor([3, 3])]; + tensor xt_17_strides_0 = const()[name = tensor("xt_17_strides_0"), val = tensor([1])]; + tensor xt_17_dilations_0 = const()[name = tensor("xt_17_dilations_0"), val = tensor([1])]; + tensor xt_17_groups_0 = const()[name = tensor("xt_17_groups_0"), val = tensor(1)]; + tensor xt_17 = conv(bias = generator_noise_res_0_convs2_2_bias, dilations = xt_17_dilations_0, groups = xt_17_groups_0, pad = xt_17_pad_0, pad_type = xt_17_pad_type_0, strides = xt_17_strides_0, weight = weight_49_palettized, x = input_115)[name = tensor("xt_17")]; + tensor x_source_1 = add(x = xt_17, y = input_109)[name = tensor("x_source_1")]; + tensor op_967_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(39732096))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42353600))), name = tensor("op_967_palettized"), shape = tensor([512, 256, 20])]; + tensor x_9_pad_type_0 = const()[name = tensor("x_9_pad_type_0"), val = tensor("custom")]; + tensor x_9_pad_0 = const()[name = tensor("x_9_pad_0"), val = tensor([5, 5])]; + tensor x_9_strides_0 = const()[name = tensor("x_9_strides_0"), val = tensor([10])]; + tensor x_9_dilations_0 = const()[name = tensor("x_9_dilations_0"), val = tensor([1])]; + tensor x_9_groups_0 = const()[name = tensor("x_9_groups_0"), val = tensor(1)]; + tensor x_9 = conv_transpose(bias = generator_ups_0_bias, dilations = x_9_dilations_0, groups = x_9_groups_0, pad = x_9_pad_0, pad_type = x_9_pad_type_0, strides = x_9_strides_0, weight = op_967_palettized, x = input_117)[name = tensor("x_9")]; + tensor input_119 = add(x = x_9, y = x_source_1)[name = tensor("input_119")]; + tensor var_987 = const()[name = tensor("op_987"), val = tensor(0x1.4f8b58p-17)]; + tensor h_65 = linear(bias = generator_resblocks_0_adain1_0_fc_bias, weight = generator_resblocks_0_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_17")]; + tensor var_1069 = const()[name = tensor("op_1069"), val = tensor([1, 512, 1])]; + tensor h_67 = reshape(shape = var_1069, x = h_65)[name = tensor("h_67")]; + tensor var_1071_split_sizes_0 = const()[name = tensor("op_1071_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1071_axis_0 = const()[name = tensor("op_1071_axis_0"), val = tensor(1)]; + tensor var_1071_0, tensor var_1071_1 = split(axis = var_1071_axis_0, split_sizes = var_1071_split_sizes_0, x = h_67)[name = tensor("op_1071")]; + tensor var_1073_promoted = const()[name = tensor("op_1073_promoted"), val = tensor(0x1p+0)]; + tensor var_1074 = add(x = var_1071_0, y = var_1073_promoted)[name = tensor("op_1074")]; + tensor var_1075 = instance_norm(epsilon = var_987, x = input_119)[name = tensor("op_1075")]; + tensor var_1076 = mul(x = var_1074, y = var_1075)[name = tensor("op_1076")]; + tensor xt_19 = add(x = var_1076, y = var_1071_1)[name = tensor("xt_19")]; + tensor var_1079 = const()[name = tensor("op_1079"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42354688)))]; + tensor var_1080 = mul(x = xt_19, y = var_1079)[name = tensor("op_1080")]; + tensor cv_13 = cos(x = var_1080)[name = tensor("cv_13")]; + tensor var_1082 = const()[name = tensor("op_1082"), val = tensor(-0x1p-1)]; + tensor var_1083 = mul(x = cv_13, y = var_1082)[name = tensor("op_1083")]; + tensor var_1084 = const()[name = tensor("op_1084"), val = tensor(0x1p-1)]; + tensor var_1085 = add(x = var_1083, y = var_1084)[name = tensor("op_1085")]; + tensor var_1086 = const()[name = tensor("op_1086"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42355776)))]; + tensor var_1089 = mul(x = var_1085, y = var_1086)[name = tensor("op_1089")]; + tensor input_121 = add(x = xt_19, y = var_1089)[name = tensor("input_121")]; + tensor weight_51_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42356864))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42553536))), name = tensor("weight_51_palettized"), shape = tensor([256, 256, 3])]; + tensor input_123_pad_type_0 = const()[name = tensor("input_123_pad_type_0"), val = tensor("custom")]; + tensor input_123_pad_0 = const()[name = tensor("input_123_pad_0"), val = tensor([1, 1])]; + tensor input_123_strides_0 = const()[name = tensor("input_123_strides_0"), val = tensor([1])]; + tensor input_123_dilations_0 = const()[name = tensor("input_123_dilations_0"), val = tensor([1])]; + tensor input_123_groups_0 = const()[name = tensor("input_123_groups_0"), val = tensor(1)]; + tensor input_123 = conv(bias = generator_resblocks_0_convs1_0_bias, dilations = input_123_dilations_0, groups = input_123_groups_0, pad = input_123_pad_0, pad_type = input_123_pad_type_0, strides = input_123_strides_0, weight = weight_51_palettized, x = input_121)[name = tensor("input_123")]; + tensor h_69 = linear(bias = generator_resblocks_0_adain2_0_fc_bias, weight = generator_resblocks_0_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_18")]; + tensor var_1104 = const()[name = tensor("op_1104"), val = tensor([1, 512, 1])]; + tensor h_71 = reshape(shape = var_1104, x = h_69)[name = tensor("h_71")]; + tensor var_1106_split_sizes_0 = const()[name = tensor("op_1106_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1106_axis_0 = const()[name = tensor("op_1106_axis_0"), val = tensor(1)]; + tensor var_1106_0, tensor var_1106_1 = split(axis = var_1106_axis_0, split_sizes = var_1106_split_sizes_0, x = h_71)[name = tensor("op_1106")]; + tensor var_1108_promoted = const()[name = tensor("op_1108_promoted"), val = tensor(0x1p+0)]; + tensor var_1109 = add(x = var_1106_0, y = var_1108_promoted)[name = tensor("op_1109")]; + tensor var_1110 = instance_norm(epsilon = var_987, x = input_123)[name = tensor("op_1110")]; + tensor var_1111 = mul(x = var_1109, y = var_1110)[name = tensor("op_1111")]; + tensor xt_21 = add(x = var_1111, y = var_1106_1)[name = tensor("xt_21")]; + tensor var_1114 = const()[name = tensor("op_1114"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42554624)))]; + tensor var_1115 = mul(x = xt_21, y = var_1114)[name = tensor("op_1115")]; + tensor cv_15 = cos(x = var_1115)[name = tensor("cv_15")]; + tensor var_1117 = const()[name = tensor("op_1117"), val = tensor(-0x1p-1)]; + tensor var_1118 = mul(x = cv_15, y = var_1117)[name = tensor("op_1118")]; + tensor var_1119 = const()[name = tensor("op_1119"), val = tensor(0x1p-1)]; + tensor var_1120 = add(x = var_1118, y = var_1119)[name = tensor("op_1120")]; + tensor var_1121 = const()[name = tensor("op_1121"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42555712)))]; + tensor var_1124 = mul(x = var_1120, y = var_1121)[name = tensor("op_1124")]; + tensor input_125 = add(x = xt_21, y = var_1124)[name = tensor("input_125")]; + tensor weight_53_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42556800))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42753472))), name = tensor("weight_53_palettized"), shape = tensor([256, 256, 3])]; + tensor xt_23_pad_type_0 = const()[name = tensor("xt_23_pad_type_0"), val = tensor("custom")]; + tensor xt_23_pad_0 = const()[name = tensor("xt_23_pad_0"), val = tensor([1, 1])]; + tensor xt_23_strides_0 = const()[name = tensor("xt_23_strides_0"), val = tensor([1])]; + tensor xt_23_dilations_0 = const()[name = tensor("xt_23_dilations_0"), val = tensor([1])]; + tensor xt_23_groups_0 = const()[name = tensor("xt_23_groups_0"), val = tensor(1)]; + tensor xt_23 = conv(bias = generator_resblocks_0_convs2_0_bias, dilations = xt_23_dilations_0, groups = xt_23_groups_0, pad = xt_23_pad_0, pad_type = xt_23_pad_type_0, strides = xt_23_strides_0, weight = weight_53_palettized, x = input_125)[name = tensor("xt_23")]; + tensor input_127 = add(x = xt_23, y = input_119)[name = tensor("input_127")]; + tensor h_73 = linear(bias = generator_resblocks_0_adain1_1_fc_bias, weight = generator_resblocks_0_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_19")]; + tensor var_1140 = const()[name = tensor("op_1140"), val = tensor([1, 512, 1])]; + tensor h_75 = reshape(shape = var_1140, x = h_73)[name = tensor("h_75")]; + tensor var_1142_split_sizes_0 = const()[name = tensor("op_1142_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1142_axis_0 = const()[name = tensor("op_1142_axis_0"), val = tensor(1)]; + tensor var_1142_0, tensor var_1142_1 = split(axis = var_1142_axis_0, split_sizes = var_1142_split_sizes_0, x = h_75)[name = tensor("op_1142")]; + tensor var_1144_promoted = const()[name = tensor("op_1144_promoted"), val = tensor(0x1p+0)]; + tensor var_1145 = add(x = var_1142_0, y = var_1144_promoted)[name = tensor("op_1145")]; + tensor var_1146 = instance_norm(epsilon = var_987, x = input_127)[name = tensor("op_1146")]; + tensor var_1147 = mul(x = var_1145, y = var_1146)[name = tensor("op_1147")]; + tensor xt_25 = add(x = var_1147, y = var_1142_1)[name = tensor("xt_25")]; + tensor var_1150 = const()[name = tensor("op_1150"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42754560)))]; + tensor var_1151 = mul(x = xt_25, y = var_1150)[name = tensor("op_1151")]; + tensor cv_17 = cos(x = var_1151)[name = tensor("cv_17")]; + tensor var_1153 = const()[name = tensor("op_1153"), val = tensor(-0x1p-1)]; + tensor var_1154 = mul(x = cv_17, y = var_1153)[name = tensor("op_1154")]; + tensor var_1155 = const()[name = tensor("op_1155"), val = tensor(0x1p-1)]; + tensor var_1156 = add(x = var_1154, y = var_1155)[name = tensor("op_1156")]; + tensor var_1157 = const()[name = tensor("op_1157"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42755648)))]; + tensor var_1160 = mul(x = var_1156, y = var_1157)[name = tensor("op_1160")]; + tensor input_129 = add(x = xt_25, y = var_1160)[name = tensor("input_129")]; + tensor weight_55_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42756736))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42953408))), name = tensor("weight_55_palettized"), shape = tensor([256, 256, 3])]; + tensor input_131_pad_type_0 = const()[name = tensor("input_131_pad_type_0"), val = tensor("custom")]; + tensor input_131_pad_0 = const()[name = tensor("input_131_pad_0"), val = tensor([3, 3])]; + tensor input_131_dilations_0 = const()[name = tensor("input_131_dilations_0"), val = tensor([3])]; + tensor input_131_strides_0 = const()[name = tensor("input_131_strides_0"), val = tensor([1])]; + tensor input_131_groups_0 = const()[name = tensor("input_131_groups_0"), val = tensor(1)]; + tensor input_131 = conv(bias = generator_resblocks_0_convs1_1_bias, dilations = input_131_dilations_0, groups = input_131_groups_0, pad = input_131_pad_0, pad_type = input_131_pad_type_0, strides = input_131_strides_0, weight = weight_55_palettized, x = input_129)[name = tensor("input_131")]; + tensor h_77 = linear(bias = generator_resblocks_0_adain2_1_fc_bias, weight = generator_resblocks_0_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_20")]; + tensor var_1175 = const()[name = tensor("op_1175"), val = tensor([1, 512, 1])]; + tensor h_79 = reshape(shape = var_1175, x = h_77)[name = tensor("h_79")]; + tensor var_1177_split_sizes_0 = const()[name = tensor("op_1177_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1177_axis_0 = const()[name = tensor("op_1177_axis_0"), val = tensor(1)]; + tensor var_1177_0, tensor var_1177_1 = split(axis = var_1177_axis_0, split_sizes = var_1177_split_sizes_0, x = h_79)[name = tensor("op_1177")]; + tensor var_1179_promoted = const()[name = tensor("op_1179_promoted"), val = tensor(0x1p+0)]; + tensor var_1180 = add(x = var_1177_0, y = var_1179_promoted)[name = tensor("op_1180")]; + tensor var_1181 = instance_norm(epsilon = var_987, x = input_131)[name = tensor("op_1181")]; + tensor var_1182 = mul(x = var_1180, y = var_1181)[name = tensor("op_1182")]; + tensor xt_27 = add(x = var_1182, y = var_1177_1)[name = tensor("xt_27")]; + tensor var_1185 = const()[name = tensor("op_1185"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42954496)))]; + tensor var_1186 = mul(x = xt_27, y = var_1185)[name = tensor("op_1186")]; + tensor cv_19 = cos(x = var_1186)[name = tensor("cv_19")]; + tensor var_1188 = const()[name = tensor("op_1188"), val = tensor(-0x1p-1)]; + tensor var_1189 = mul(x = cv_19, y = var_1188)[name = tensor("op_1189")]; + tensor var_1190 = const()[name = tensor("op_1190"), val = tensor(0x1p-1)]; + tensor var_1191 = add(x = var_1189, y = var_1190)[name = tensor("op_1191")]; + tensor var_1192 = const()[name = tensor("op_1192"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42955584)))]; + tensor var_1195 = mul(x = var_1191, y = var_1192)[name = tensor("op_1195")]; + tensor input_133 = add(x = xt_27, y = var_1195)[name = tensor("input_133")]; + tensor weight_57_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(42956672))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43153344))), name = tensor("weight_57_palettized"), shape = tensor([256, 256, 3])]; + tensor xt_29_pad_type_0 = const()[name = tensor("xt_29_pad_type_0"), val = tensor("custom")]; + tensor xt_29_pad_0 = const()[name = tensor("xt_29_pad_0"), val = tensor([1, 1])]; + tensor xt_29_strides_0 = const()[name = tensor("xt_29_strides_0"), val = tensor([1])]; + tensor xt_29_dilations_0 = const()[name = tensor("xt_29_dilations_0"), val = tensor([1])]; + tensor xt_29_groups_0 = const()[name = tensor("xt_29_groups_0"), val = tensor(1)]; + tensor xt_29 = conv(bias = generator_resblocks_0_convs2_1_bias, dilations = xt_29_dilations_0, groups = xt_29_groups_0, pad = xt_29_pad_0, pad_type = xt_29_pad_type_0, strides = xt_29_strides_0, weight = weight_57_palettized, x = input_133)[name = tensor("xt_29")]; + tensor input_135 = add(x = xt_29, y = input_127)[name = tensor("input_135")]; + tensor h_81 = linear(bias = generator_resblocks_0_adain1_2_fc_bias, weight = generator_resblocks_0_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_21")]; + tensor var_1211 = const()[name = tensor("op_1211"), val = tensor([1, 512, 1])]; + tensor h_83 = reshape(shape = var_1211, x = h_81)[name = tensor("h_83")]; + tensor var_1213_split_sizes_0 = const()[name = tensor("op_1213_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1213_axis_0 = const()[name = tensor("op_1213_axis_0"), val = tensor(1)]; + tensor var_1213_0, tensor var_1213_1 = split(axis = var_1213_axis_0, split_sizes = var_1213_split_sizes_0, x = h_83)[name = tensor("op_1213")]; + tensor var_1215_promoted = const()[name = tensor("op_1215_promoted"), val = tensor(0x1p+0)]; + tensor var_1216 = add(x = var_1213_0, y = var_1215_promoted)[name = tensor("op_1216")]; + tensor var_1217 = instance_norm(epsilon = var_987, x = input_135)[name = tensor("op_1217")]; + tensor var_1218 = mul(x = var_1216, y = var_1217)[name = tensor("op_1218")]; + tensor xt_31 = add(x = var_1218, y = var_1213_1)[name = tensor("xt_31")]; + tensor var_1221 = const()[name = tensor("op_1221"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43154432)))]; + tensor var_1222 = mul(x = xt_31, y = var_1221)[name = tensor("op_1222")]; + tensor cv_21 = cos(x = var_1222)[name = tensor("cv_21")]; + tensor var_1224 = const()[name = tensor("op_1224"), val = tensor(-0x1p-1)]; + tensor var_1225 = mul(x = cv_21, y = var_1224)[name = tensor("op_1225")]; + tensor var_1226 = const()[name = tensor("op_1226"), val = tensor(0x1p-1)]; + tensor var_1227 = add(x = var_1225, y = var_1226)[name = tensor("op_1227")]; + tensor var_1228 = const()[name = tensor("op_1228"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43155520)))]; + tensor var_1231 = mul(x = var_1227, y = var_1228)[name = tensor("op_1231")]; + tensor input_137 = add(x = xt_31, y = var_1231)[name = tensor("input_137")]; + tensor weight_59_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43156608))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43353280))), name = tensor("weight_59_palettized"), shape = tensor([256, 256, 3])]; + tensor input_139_pad_type_0 = const()[name = tensor("input_139_pad_type_0"), val = tensor("custom")]; + tensor input_139_pad_0 = const()[name = tensor("input_139_pad_0"), val = tensor([5, 5])]; + tensor input_139_dilations_0 = const()[name = tensor("input_139_dilations_0"), val = tensor([5])]; + tensor input_139_strides_0 = const()[name = tensor("input_139_strides_0"), val = tensor([1])]; + tensor input_139_groups_0 = const()[name = tensor("input_139_groups_0"), val = tensor(1)]; + tensor input_139 = conv(bias = generator_resblocks_0_convs1_2_bias, dilations = input_139_dilations_0, groups = input_139_groups_0, pad = input_139_pad_0, pad_type = input_139_pad_type_0, strides = input_139_strides_0, weight = weight_59_palettized, x = input_137)[name = tensor("input_139")]; + tensor h_85 = linear(bias = generator_resblocks_0_adain2_2_fc_bias, weight = generator_resblocks_0_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_22")]; + tensor var_1246 = const()[name = tensor("op_1246"), val = tensor([1, 512, 1])]; + tensor h_87 = reshape(shape = var_1246, x = h_85)[name = tensor("h_87")]; + tensor var_1248_split_sizes_0 = const()[name = tensor("op_1248_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1248_axis_0 = const()[name = tensor("op_1248_axis_0"), val = tensor(1)]; + tensor var_1248_0, tensor var_1248_1 = split(axis = var_1248_axis_0, split_sizes = var_1248_split_sizes_0, x = h_87)[name = tensor("op_1248")]; + tensor var_1250_promoted = const()[name = tensor("op_1250_promoted"), val = tensor(0x1p+0)]; + tensor var_1251 = add(x = var_1248_0, y = var_1250_promoted)[name = tensor("op_1251")]; + tensor var_1252 = instance_norm(epsilon = var_987, x = input_139)[name = tensor("op_1252")]; + tensor var_1253 = mul(x = var_1251, y = var_1252)[name = tensor("op_1253")]; + tensor xt_33 = add(x = var_1253, y = var_1248_1)[name = tensor("xt_33")]; + tensor var_1256 = const()[name = tensor("op_1256"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43354368)))]; + tensor var_1257 = mul(x = xt_33, y = var_1256)[name = tensor("op_1257")]; + tensor cv_23 = cos(x = var_1257)[name = tensor("cv_23")]; + tensor var_1259 = const()[name = tensor("op_1259"), val = tensor(-0x1p-1)]; + tensor var_1260 = mul(x = cv_23, y = var_1259)[name = tensor("op_1260")]; + tensor var_1261 = const()[name = tensor("op_1261"), val = tensor(0x1p-1)]; + tensor var_1262 = add(x = var_1260, y = var_1261)[name = tensor("op_1262")]; + tensor var_1263 = const()[name = tensor("op_1263"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43355456)))]; + tensor var_1266 = mul(x = var_1262, y = var_1263)[name = tensor("op_1266")]; + tensor input_141 = add(x = xt_33, y = var_1266)[name = tensor("input_141")]; + tensor weight_61_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43356544))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43553216))), name = tensor("weight_61_palettized"), shape = tensor([256, 256, 3])]; + tensor xt_35_pad_type_0 = const()[name = tensor("xt_35_pad_type_0"), val = tensor("custom")]; + tensor xt_35_pad_0 = const()[name = tensor("xt_35_pad_0"), val = tensor([1, 1])]; + tensor xt_35_strides_0 = const()[name = tensor("xt_35_strides_0"), val = tensor([1])]; + tensor xt_35_dilations_0 = const()[name = tensor("xt_35_dilations_0"), val = tensor([1])]; + tensor xt_35_groups_0 = const()[name = tensor("xt_35_groups_0"), val = tensor(1)]; + tensor xt_35 = conv(bias = generator_resblocks_0_convs2_2_bias, dilations = xt_35_dilations_0, groups = xt_35_groups_0, pad = xt_35_pad_0, pad_type = xt_35_pad_type_0, strides = xt_35_strides_0, weight = weight_61_palettized, x = input_141)[name = tensor("xt_35")]; + tensor xs_1 = add(x = xt_35, y = input_135)[name = tensor("xs_1")]; + tensor var_1283 = const()[name = tensor("op_1283"), val = tensor(0x1.4f8b58p-17)]; + tensor h_89 = linear(bias = generator_resblocks_1_adain1_0_fc_bias, weight = generator_resblocks_1_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_23")]; + tensor var_1365 = const()[name = tensor("op_1365"), val = tensor([1, 512, 1])]; + tensor h_91 = reshape(shape = var_1365, x = h_89)[name = tensor("h_91")]; + tensor var_1367_split_sizes_0 = const()[name = tensor("op_1367_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1367_axis_0 = const()[name = tensor("op_1367_axis_0"), val = tensor(1)]; + tensor var_1367_0, tensor var_1367_1 = split(axis = var_1367_axis_0, split_sizes = var_1367_split_sizes_0, x = h_91)[name = tensor("op_1367")]; + tensor var_1369_promoted = const()[name = tensor("op_1369_promoted"), val = tensor(0x1p+0)]; + tensor var_1370 = add(x = var_1367_0, y = var_1369_promoted)[name = tensor("op_1370")]; + tensor var_1372 = mul(x = var_1370, y = var_1075)[name = tensor("op_1372")]; + tensor xt_37 = add(x = var_1372, y = var_1367_1)[name = tensor("xt_37")]; + tensor var_1375 = const()[name = tensor("op_1375"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43554304)))]; + tensor var_1376 = mul(x = xt_37, y = var_1375)[name = tensor("op_1376")]; + tensor cv_25 = cos(x = var_1376)[name = tensor("cv_25")]; + tensor var_1378 = const()[name = tensor("op_1378"), val = tensor(-0x1p-1)]; + tensor var_1379 = mul(x = cv_25, y = var_1378)[name = tensor("op_1379")]; + tensor var_1380 = const()[name = tensor("op_1380"), val = tensor(0x1p-1)]; + tensor var_1381 = add(x = var_1379, y = var_1380)[name = tensor("op_1381")]; + tensor var_1382 = const()[name = tensor("op_1382"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43555392)))]; + tensor var_1385 = mul(x = var_1381, y = var_1382)[name = tensor("op_1385")]; + tensor input_143 = add(x = xt_37, y = var_1385)[name = tensor("input_143")]; + tensor weight_63_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43556480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44015296))), name = tensor("weight_63_palettized"), shape = tensor([256, 256, 7])]; + tensor input_145_pad_type_0 = const()[name = tensor("input_145_pad_type_0"), val = tensor("custom")]; + tensor input_145_pad_0 = const()[name = tensor("input_145_pad_0"), val = tensor([3, 3])]; + tensor input_145_strides_0 = const()[name = tensor("input_145_strides_0"), val = tensor([1])]; + tensor input_145_dilations_0 = const()[name = tensor("input_145_dilations_0"), val = tensor([1])]; + tensor input_145_groups_0 = const()[name = tensor("input_145_groups_0"), val = tensor(1)]; + tensor input_145 = conv(bias = generator_resblocks_1_convs1_0_bias, dilations = input_145_dilations_0, groups = input_145_groups_0, pad = input_145_pad_0, pad_type = input_145_pad_type_0, strides = input_145_strides_0, weight = weight_63_palettized, x = input_143)[name = tensor("input_145")]; + tensor h_93 = linear(bias = generator_resblocks_1_adain2_0_fc_bias, weight = generator_resblocks_1_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_24")]; + tensor var_1400 = const()[name = tensor("op_1400"), val = tensor([1, 512, 1])]; + tensor h_95 = reshape(shape = var_1400, x = h_93)[name = tensor("h_95")]; + tensor var_1402_split_sizes_0 = const()[name = tensor("op_1402_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1402_axis_0 = const()[name = tensor("op_1402_axis_0"), val = tensor(1)]; + tensor var_1402_0, tensor var_1402_1 = split(axis = var_1402_axis_0, split_sizes = var_1402_split_sizes_0, x = h_95)[name = tensor("op_1402")]; + tensor var_1404_promoted = const()[name = tensor("op_1404_promoted"), val = tensor(0x1p+0)]; + tensor var_1405 = add(x = var_1402_0, y = var_1404_promoted)[name = tensor("op_1405")]; + tensor var_1406 = instance_norm(epsilon = var_1283, x = input_145)[name = tensor("op_1406")]; + tensor var_1407 = mul(x = var_1405, y = var_1406)[name = tensor("op_1407")]; + tensor xt_39 = add(x = var_1407, y = var_1402_1)[name = tensor("xt_39")]; + tensor var_1410 = const()[name = tensor("op_1410"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44016384)))]; + tensor var_1411 = mul(x = xt_39, y = var_1410)[name = tensor("op_1411")]; + tensor cv_27 = cos(x = var_1411)[name = tensor("cv_27")]; + tensor var_1413 = const()[name = tensor("op_1413"), val = tensor(-0x1p-1)]; + tensor var_1414 = mul(x = cv_27, y = var_1413)[name = tensor("op_1414")]; + tensor var_1415 = const()[name = tensor("op_1415"), val = tensor(0x1p-1)]; + tensor var_1416 = add(x = var_1414, y = var_1415)[name = tensor("op_1416")]; + tensor var_1417 = const()[name = tensor("op_1417"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44017472)))]; + tensor var_1420 = mul(x = var_1416, y = var_1417)[name = tensor("op_1420")]; + tensor input_147 = add(x = xt_39, y = var_1420)[name = tensor("input_147")]; + tensor weight_65_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44018560))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44477376))), name = tensor("weight_65_palettized"), shape = tensor([256, 256, 7])]; + tensor xt_41_pad_type_0 = const()[name = tensor("xt_41_pad_type_0"), val = tensor("custom")]; + tensor xt_41_pad_0 = const()[name = tensor("xt_41_pad_0"), val = tensor([3, 3])]; + tensor xt_41_strides_0 = const()[name = tensor("xt_41_strides_0"), val = tensor([1])]; + tensor xt_41_dilations_0 = const()[name = tensor("xt_41_dilations_0"), val = tensor([1])]; + tensor xt_41_groups_0 = const()[name = tensor("xt_41_groups_0"), val = tensor(1)]; + tensor xt_41 = conv(bias = generator_resblocks_1_convs2_0_bias, dilations = xt_41_dilations_0, groups = xt_41_groups_0, pad = xt_41_pad_0, pad_type = xt_41_pad_type_0, strides = xt_41_strides_0, weight = weight_65_palettized, x = input_147)[name = tensor("xt_41")]; + tensor input_149 = add(x = xt_41, y = input_119)[name = tensor("input_149")]; + tensor h_97 = linear(bias = generator_resblocks_1_adain1_1_fc_bias, weight = generator_resblocks_1_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_25")]; + tensor var_1436 = const()[name = tensor("op_1436"), val = tensor([1, 512, 1])]; + tensor h_99 = reshape(shape = var_1436, x = h_97)[name = tensor("h_99")]; + tensor var_1438_split_sizes_0 = const()[name = tensor("op_1438_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1438_axis_0 = const()[name = tensor("op_1438_axis_0"), val = tensor(1)]; + tensor var_1438_0, tensor var_1438_1 = split(axis = var_1438_axis_0, split_sizes = var_1438_split_sizes_0, x = h_99)[name = tensor("op_1438")]; + tensor var_1440_promoted = const()[name = tensor("op_1440_promoted"), val = tensor(0x1p+0)]; + tensor var_1441 = add(x = var_1438_0, y = var_1440_promoted)[name = tensor("op_1441")]; + tensor var_1442 = instance_norm(epsilon = var_1283, x = input_149)[name = tensor("op_1442")]; + tensor var_1443 = mul(x = var_1441, y = var_1442)[name = tensor("op_1443")]; + tensor xt_43 = add(x = var_1443, y = var_1438_1)[name = tensor("xt_43")]; + tensor var_1446 = const()[name = tensor("op_1446"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44478464)))]; + tensor var_1447 = mul(x = xt_43, y = var_1446)[name = tensor("op_1447")]; + tensor cv_29 = cos(x = var_1447)[name = tensor("cv_29")]; + tensor var_1449 = const()[name = tensor("op_1449"), val = tensor(-0x1p-1)]; + tensor var_1450 = mul(x = cv_29, y = var_1449)[name = tensor("op_1450")]; + tensor var_1451 = const()[name = tensor("op_1451"), val = tensor(0x1p-1)]; + tensor var_1452 = add(x = var_1450, y = var_1451)[name = tensor("op_1452")]; + tensor var_1453 = const()[name = tensor("op_1453"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44479552)))]; + tensor var_1456 = mul(x = var_1452, y = var_1453)[name = tensor("op_1456")]; + tensor input_151 = add(x = xt_43, y = var_1456)[name = tensor("input_151")]; + tensor weight_67_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44480640))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44939456))), name = tensor("weight_67_palettized"), shape = tensor([256, 256, 7])]; + tensor input_153_pad_type_0 = const()[name = tensor("input_153_pad_type_0"), val = tensor("custom")]; + tensor input_153_pad_0 = const()[name = tensor("input_153_pad_0"), val = tensor([9, 9])]; + tensor input_153_dilations_0 = const()[name = tensor("input_153_dilations_0"), val = tensor([3])]; + tensor input_153_strides_0 = const()[name = tensor("input_153_strides_0"), val = tensor([1])]; + tensor input_153_groups_0 = const()[name = tensor("input_153_groups_0"), val = tensor(1)]; + tensor input_153 = conv(bias = generator_resblocks_1_convs1_1_bias, dilations = input_153_dilations_0, groups = input_153_groups_0, pad = input_153_pad_0, pad_type = input_153_pad_type_0, strides = input_153_strides_0, weight = weight_67_palettized, x = input_151)[name = tensor("input_153")]; + tensor h_101 = linear(bias = generator_resblocks_1_adain2_1_fc_bias, weight = generator_resblocks_1_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_26")]; + tensor var_1471 = const()[name = tensor("op_1471"), val = tensor([1, 512, 1])]; + tensor h_103 = reshape(shape = var_1471, x = h_101)[name = tensor("h_103")]; + tensor var_1473_split_sizes_0 = const()[name = tensor("op_1473_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1473_axis_0 = const()[name = tensor("op_1473_axis_0"), val = tensor(1)]; + tensor var_1473_0, tensor var_1473_1 = split(axis = var_1473_axis_0, split_sizes = var_1473_split_sizes_0, x = h_103)[name = tensor("op_1473")]; + tensor var_1475_promoted = const()[name = tensor("op_1475_promoted"), val = tensor(0x1p+0)]; + tensor var_1476 = add(x = var_1473_0, y = var_1475_promoted)[name = tensor("op_1476")]; + tensor var_1477 = instance_norm(epsilon = var_1283, x = input_153)[name = tensor("op_1477")]; + tensor var_1478 = mul(x = var_1476, y = var_1477)[name = tensor("op_1478")]; + tensor xt_45 = add(x = var_1478, y = var_1473_1)[name = tensor("xt_45")]; + tensor var_1481 = const()[name = tensor("op_1481"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44940544)))]; + tensor var_1482 = mul(x = xt_45, y = var_1481)[name = tensor("op_1482")]; + tensor cv_31 = cos(x = var_1482)[name = tensor("cv_31")]; + tensor var_1484 = const()[name = tensor("op_1484"), val = tensor(-0x1p-1)]; + tensor var_1485 = mul(x = cv_31, y = var_1484)[name = tensor("op_1485")]; + tensor var_1486 = const()[name = tensor("op_1486"), val = tensor(0x1p-1)]; + tensor var_1487 = add(x = var_1485, y = var_1486)[name = tensor("op_1487")]; + tensor var_1488 = const()[name = tensor("op_1488"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44941632)))]; + tensor var_1491 = mul(x = var_1487, y = var_1488)[name = tensor("op_1491")]; + tensor input_155 = add(x = xt_45, y = var_1491)[name = tensor("input_155")]; + tensor weight_69_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44942720))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45401536))), name = tensor("weight_69_palettized"), shape = tensor([256, 256, 7])]; + tensor xt_47_pad_type_0 = const()[name = tensor("xt_47_pad_type_0"), val = tensor("custom")]; + tensor xt_47_pad_0 = const()[name = tensor("xt_47_pad_0"), val = tensor([3, 3])]; + tensor xt_47_strides_0 = const()[name = tensor("xt_47_strides_0"), val = tensor([1])]; + tensor xt_47_dilations_0 = const()[name = tensor("xt_47_dilations_0"), val = tensor([1])]; + tensor xt_47_groups_0 = const()[name = tensor("xt_47_groups_0"), val = tensor(1)]; + tensor xt_47 = conv(bias = generator_resblocks_1_convs2_1_bias, dilations = xt_47_dilations_0, groups = xt_47_groups_0, pad = xt_47_pad_0, pad_type = xt_47_pad_type_0, strides = xt_47_strides_0, weight = weight_69_palettized, x = input_155)[name = tensor("xt_47")]; + tensor input_157 = add(x = xt_47, y = input_149)[name = tensor("input_157")]; + tensor h_105 = linear(bias = generator_resblocks_1_adain1_2_fc_bias, weight = generator_resblocks_1_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_27")]; + tensor var_1507 = const()[name = tensor("op_1507"), val = tensor([1, 512, 1])]; + tensor h_107 = reshape(shape = var_1507, x = h_105)[name = tensor("h_107")]; + tensor var_1509_split_sizes_0 = const()[name = tensor("op_1509_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1509_axis_0 = const()[name = tensor("op_1509_axis_0"), val = tensor(1)]; + tensor var_1509_0, tensor var_1509_1 = split(axis = var_1509_axis_0, split_sizes = var_1509_split_sizes_0, x = h_107)[name = tensor("op_1509")]; + tensor var_1511_promoted = const()[name = tensor("op_1511_promoted"), val = tensor(0x1p+0)]; + tensor var_1512 = add(x = var_1509_0, y = var_1511_promoted)[name = tensor("op_1512")]; + tensor var_1513 = instance_norm(epsilon = var_1283, x = input_157)[name = tensor("op_1513")]; + tensor var_1514 = mul(x = var_1512, y = var_1513)[name = tensor("op_1514")]; + tensor xt_49 = add(x = var_1514, y = var_1509_1)[name = tensor("xt_49")]; + tensor var_1517 = const()[name = tensor("op_1517"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45402624)))]; + tensor var_1518 = mul(x = xt_49, y = var_1517)[name = tensor("op_1518")]; + tensor cv_33 = cos(x = var_1518)[name = tensor("cv_33")]; + tensor var_1520 = const()[name = tensor("op_1520"), val = tensor(-0x1p-1)]; + tensor var_1521 = mul(x = cv_33, y = var_1520)[name = tensor("op_1521")]; + tensor var_1522 = const()[name = tensor("op_1522"), val = tensor(0x1p-1)]; + tensor var_1523 = add(x = var_1521, y = var_1522)[name = tensor("op_1523")]; + tensor var_1524 = const()[name = tensor("op_1524"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45403712)))]; + tensor var_1527 = mul(x = var_1523, y = var_1524)[name = tensor("op_1527")]; + tensor input_159 = add(x = xt_49, y = var_1527)[name = tensor("input_159")]; + tensor weight_71_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45404800))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45863616))), name = tensor("weight_71_palettized"), shape = tensor([256, 256, 7])]; + tensor input_161_pad_type_0 = const()[name = tensor("input_161_pad_type_0"), val = tensor("custom")]; + tensor input_161_pad_0 = const()[name = tensor("input_161_pad_0"), val = tensor([15, 15])]; + tensor input_161_dilations_0 = const()[name = tensor("input_161_dilations_0"), val = tensor([5])]; + tensor input_161_strides_0 = const()[name = tensor("input_161_strides_0"), val = tensor([1])]; + tensor input_161_groups_0 = const()[name = tensor("input_161_groups_0"), val = tensor(1)]; + tensor input_161 = conv(bias = generator_resblocks_1_convs1_2_bias, dilations = input_161_dilations_0, groups = input_161_groups_0, pad = input_161_pad_0, pad_type = input_161_pad_type_0, strides = input_161_strides_0, weight = weight_71_palettized, x = input_159)[name = tensor("input_161")]; + tensor h_109 = linear(bias = generator_resblocks_1_adain2_2_fc_bias, weight = generator_resblocks_1_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_28")]; + tensor var_1542 = const()[name = tensor("op_1542"), val = tensor([1, 512, 1])]; + tensor h_111 = reshape(shape = var_1542, x = h_109)[name = tensor("h_111")]; + tensor var_1544_split_sizes_0 = const()[name = tensor("op_1544_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1544_axis_0 = const()[name = tensor("op_1544_axis_0"), val = tensor(1)]; + tensor var_1544_0, tensor var_1544_1 = split(axis = var_1544_axis_0, split_sizes = var_1544_split_sizes_0, x = h_111)[name = tensor("op_1544")]; + tensor var_1546_promoted = const()[name = tensor("op_1546_promoted"), val = tensor(0x1p+0)]; + tensor var_1547 = add(x = var_1544_0, y = var_1546_promoted)[name = tensor("op_1547")]; + tensor var_1548 = instance_norm(epsilon = var_1283, x = input_161)[name = tensor("op_1548")]; + tensor var_1549 = mul(x = var_1547, y = var_1548)[name = tensor("op_1549")]; + tensor xt_51 = add(x = var_1549, y = var_1544_1)[name = tensor("xt_51")]; + tensor var_1552 = const()[name = tensor("op_1552"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45864704)))]; + tensor var_1553 = mul(x = xt_51, y = var_1552)[name = tensor("op_1553")]; + tensor cv_35 = cos(x = var_1553)[name = tensor("cv_35")]; + tensor var_1555 = const()[name = tensor("op_1555"), val = tensor(-0x1p-1)]; + tensor var_1556 = mul(x = cv_35, y = var_1555)[name = tensor("op_1556")]; + tensor var_1557 = const()[name = tensor("op_1557"), val = tensor(0x1p-1)]; + tensor var_1558 = add(x = var_1556, y = var_1557)[name = tensor("op_1558")]; + tensor var_1559 = const()[name = tensor("op_1559"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45865792)))]; + tensor var_1562 = mul(x = var_1558, y = var_1559)[name = tensor("op_1562")]; + tensor input_163 = add(x = xt_51, y = var_1562)[name = tensor("input_163")]; + tensor weight_73_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45866880))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(46325696))), name = tensor("weight_73_palettized"), shape = tensor([256, 256, 7])]; + tensor xt_53_pad_type_0 = const()[name = tensor("xt_53_pad_type_0"), val = tensor("custom")]; + tensor xt_53_pad_0 = const()[name = tensor("xt_53_pad_0"), val = tensor([3, 3])]; + tensor xt_53_strides_0 = const()[name = tensor("xt_53_strides_0"), val = tensor([1])]; + tensor xt_53_dilations_0 = const()[name = tensor("xt_53_dilations_0"), val = tensor([1])]; + tensor xt_53_groups_0 = const()[name = tensor("xt_53_groups_0"), val = tensor(1)]; + tensor xt_53 = conv(bias = generator_resblocks_1_convs2_2_bias, dilations = xt_53_dilations_0, groups = xt_53_groups_0, pad = xt_53_pad_0, pad_type = xt_53_pad_type_0, strides = xt_53_strides_0, weight = weight_73_palettized, x = input_163)[name = tensor("xt_53")]; + tensor var_1571 = add(x = xt_53, y = input_157)[name = tensor("op_1571")]; + tensor xs_3 = add(x = xs_1, y = var_1571)[name = tensor("xs_3")]; + tensor var_1581 = const()[name = tensor("op_1581"), val = tensor(0x1.4f8b58p-17)]; + tensor h_113 = linear(bias = generator_resblocks_2_adain1_0_fc_bias, weight = generator_resblocks_2_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_29")]; + tensor var_1663 = const()[name = tensor("op_1663"), val = tensor([1, 512, 1])]; + tensor h_115 = reshape(shape = var_1663, x = h_113)[name = tensor("h_115")]; + tensor var_1665_split_sizes_0 = const()[name = tensor("op_1665_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1665_axis_0 = const()[name = tensor("op_1665_axis_0"), val = tensor(1)]; + tensor var_1665_0, tensor var_1665_1 = split(axis = var_1665_axis_0, split_sizes = var_1665_split_sizes_0, x = h_115)[name = tensor("op_1665")]; + tensor var_1667_promoted = const()[name = tensor("op_1667_promoted"), val = tensor(0x1p+0)]; + tensor var_1668 = add(x = var_1665_0, y = var_1667_promoted)[name = tensor("op_1668")]; + tensor var_1670 = mul(x = var_1668, y = var_1075)[name = tensor("op_1670")]; + tensor xt_55 = add(x = var_1670, y = var_1665_1)[name = tensor("xt_55")]; + tensor var_1673 = const()[name = tensor("op_1673"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(46326784)))]; + tensor var_1674 = mul(x = xt_55, y = var_1673)[name = tensor("op_1674")]; + tensor cv_37 = cos(x = var_1674)[name = tensor("cv_37")]; + tensor var_1676 = const()[name = tensor("op_1676"), val = tensor(-0x1p-1)]; + tensor var_1677 = mul(x = cv_37, y = var_1676)[name = tensor("op_1677")]; + tensor var_1678 = const()[name = tensor("op_1678"), val = tensor(0x1p-1)]; + tensor var_1679 = add(x = var_1677, y = var_1678)[name = tensor("op_1679")]; + tensor var_1680 = const()[name = tensor("op_1680"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(46327872)))]; + tensor var_1683 = mul(x = var_1679, y = var_1680)[name = tensor("op_1683")]; + tensor input_165 = add(x = xt_55, y = var_1683)[name = tensor("input_165")]; + tensor weight_75_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(46328960))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(47049920))), name = tensor("weight_75_palettized"), shape = tensor([256, 256, 11])]; + tensor input_167_pad_type_0 = const()[name = tensor("input_167_pad_type_0"), val = tensor("custom")]; + tensor input_167_pad_0 = const()[name = tensor("input_167_pad_0"), val = tensor([5, 5])]; + tensor input_167_strides_0 = const()[name = tensor("input_167_strides_0"), val = tensor([1])]; + tensor input_167_dilations_0 = const()[name = tensor("input_167_dilations_0"), val = tensor([1])]; + tensor input_167_groups_0 = const()[name = tensor("input_167_groups_0"), val = tensor(1)]; + tensor input_167 = conv(bias = generator_resblocks_2_convs1_0_bias, dilations = input_167_dilations_0, groups = input_167_groups_0, pad = input_167_pad_0, pad_type = input_167_pad_type_0, strides = input_167_strides_0, weight = weight_75_palettized, x = input_165)[name = tensor("input_167")]; + tensor h_117 = linear(bias = generator_resblocks_2_adain2_0_fc_bias, weight = generator_resblocks_2_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_30")]; + tensor var_1698 = const()[name = tensor("op_1698"), val = tensor([1, 512, 1])]; + tensor h_119 = reshape(shape = var_1698, x = h_117)[name = tensor("h_119")]; + tensor var_1700_split_sizes_0 = const()[name = tensor("op_1700_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1700_axis_0 = const()[name = tensor("op_1700_axis_0"), val = tensor(1)]; + tensor var_1700_0, tensor var_1700_1 = split(axis = var_1700_axis_0, split_sizes = var_1700_split_sizes_0, x = h_119)[name = tensor("op_1700")]; + tensor var_1702_promoted = const()[name = tensor("op_1702_promoted"), val = tensor(0x1p+0)]; + tensor var_1703 = add(x = var_1700_0, y = var_1702_promoted)[name = tensor("op_1703")]; + tensor var_1704 = instance_norm(epsilon = var_1581, x = input_167)[name = tensor("op_1704")]; + tensor var_1705 = mul(x = var_1703, y = var_1704)[name = tensor("op_1705")]; + tensor xt_57 = add(x = var_1705, y = var_1700_1)[name = tensor("xt_57")]; + tensor var_1708 = const()[name = tensor("op_1708"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(47051008)))]; + tensor var_1709 = mul(x = xt_57, y = var_1708)[name = tensor("op_1709")]; + tensor cv_39 = cos(x = var_1709)[name = tensor("cv_39")]; + tensor var_1711 = const()[name = tensor("op_1711"), val = tensor(-0x1p-1)]; + tensor var_1712 = mul(x = cv_39, y = var_1711)[name = tensor("op_1712")]; + tensor var_1713 = const()[name = tensor("op_1713"), val = tensor(0x1p-1)]; + tensor var_1714 = add(x = var_1712, y = var_1713)[name = tensor("op_1714")]; + tensor var_1715 = const()[name = tensor("op_1715"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(47052096)))]; + tensor var_1718 = mul(x = var_1714, y = var_1715)[name = tensor("op_1718")]; + tensor input_169 = add(x = xt_57, y = var_1718)[name = tensor("input_169")]; + tensor weight_77_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(47053184))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(47774144))), name = tensor("weight_77_palettized"), shape = tensor([256, 256, 11])]; + tensor xt_59_pad_type_0 = const()[name = tensor("xt_59_pad_type_0"), val = tensor("custom")]; + tensor xt_59_pad_0 = const()[name = tensor("xt_59_pad_0"), val = tensor([5, 5])]; + tensor xt_59_strides_0 = const()[name = tensor("xt_59_strides_0"), val = tensor([1])]; + tensor xt_59_dilations_0 = const()[name = tensor("xt_59_dilations_0"), val = tensor([1])]; + tensor xt_59_groups_0 = const()[name = tensor("xt_59_groups_0"), val = tensor(1)]; + tensor xt_59 = conv(bias = generator_resblocks_2_convs2_0_bias, dilations = xt_59_dilations_0, groups = xt_59_groups_0, pad = xt_59_pad_0, pad_type = xt_59_pad_type_0, strides = xt_59_strides_0, weight = weight_77_palettized, x = input_169)[name = tensor("xt_59")]; + tensor input_171 = add(x = xt_59, y = input_119)[name = tensor("input_171")]; + tensor h_121 = linear(bias = generator_resblocks_2_adain1_1_fc_bias, weight = generator_resblocks_2_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_31")]; + tensor var_1734 = const()[name = tensor("op_1734"), val = tensor([1, 512, 1])]; + tensor h_123 = reshape(shape = var_1734, x = h_121)[name = tensor("h_123")]; + tensor var_1736_split_sizes_0 = const()[name = tensor("op_1736_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1736_axis_0 = const()[name = tensor("op_1736_axis_0"), val = tensor(1)]; + tensor var_1736_0, tensor var_1736_1 = split(axis = var_1736_axis_0, split_sizes = var_1736_split_sizes_0, x = h_123)[name = tensor("op_1736")]; + tensor var_1738_promoted = const()[name = tensor("op_1738_promoted"), val = tensor(0x1p+0)]; + tensor var_1739 = add(x = var_1736_0, y = var_1738_promoted)[name = tensor("op_1739")]; + tensor var_1740 = instance_norm(epsilon = var_1581, x = input_171)[name = tensor("op_1740")]; + tensor var_1741 = mul(x = var_1739, y = var_1740)[name = tensor("op_1741")]; + tensor xt_61 = add(x = var_1741, y = var_1736_1)[name = tensor("xt_61")]; + tensor var_1744 = const()[name = tensor("op_1744"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(47775232)))]; + tensor var_1745 = mul(x = xt_61, y = var_1744)[name = tensor("op_1745")]; + tensor cv_41 = cos(x = var_1745)[name = tensor("cv_41")]; + tensor var_1747 = const()[name = tensor("op_1747"), val = tensor(-0x1p-1)]; + tensor var_1748 = mul(x = cv_41, y = var_1747)[name = tensor("op_1748")]; + tensor var_1749 = const()[name = tensor("op_1749"), val = tensor(0x1p-1)]; + tensor var_1750 = add(x = var_1748, y = var_1749)[name = tensor("op_1750")]; + tensor var_1751 = const()[name = tensor("op_1751"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(47776320)))]; + tensor var_1754 = mul(x = var_1750, y = var_1751)[name = tensor("op_1754")]; + tensor input_173 = add(x = xt_61, y = var_1754)[name = tensor("input_173")]; + tensor weight_79_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(47777408))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(48498368))), name = tensor("weight_79_palettized"), shape = tensor([256, 256, 11])]; + tensor input_175_pad_type_0 = const()[name = tensor("input_175_pad_type_0"), val = tensor("custom")]; + tensor input_175_pad_0 = const()[name = tensor("input_175_pad_0"), val = tensor([15, 15])]; + tensor input_175_dilations_0 = const()[name = tensor("input_175_dilations_0"), val = tensor([3])]; + tensor input_175_strides_0 = const()[name = tensor("input_175_strides_0"), val = tensor([1])]; + tensor input_175_groups_0 = const()[name = tensor("input_175_groups_0"), val = tensor(1)]; + tensor input_175 = conv(bias = generator_resblocks_2_convs1_1_bias, dilations = input_175_dilations_0, groups = input_175_groups_0, pad = input_175_pad_0, pad_type = input_175_pad_type_0, strides = input_175_strides_0, weight = weight_79_palettized, x = input_173)[name = tensor("input_175")]; + tensor h_125 = linear(bias = generator_resblocks_2_adain2_1_fc_bias, weight = generator_resblocks_2_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_32")]; + tensor var_1769 = const()[name = tensor("op_1769"), val = tensor([1, 512, 1])]; + tensor h_127 = reshape(shape = var_1769, x = h_125)[name = tensor("h_127")]; + tensor var_1771_split_sizes_0 = const()[name = tensor("op_1771_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1771_axis_0 = const()[name = tensor("op_1771_axis_0"), val = tensor(1)]; + tensor var_1771_0, tensor var_1771_1 = split(axis = var_1771_axis_0, split_sizes = var_1771_split_sizes_0, x = h_127)[name = tensor("op_1771")]; + tensor var_1773_promoted = const()[name = tensor("op_1773_promoted"), val = tensor(0x1p+0)]; + tensor var_1774 = add(x = var_1771_0, y = var_1773_promoted)[name = tensor("op_1774")]; + tensor var_1775 = instance_norm(epsilon = var_1581, x = input_175)[name = tensor("op_1775")]; + tensor var_1776 = mul(x = var_1774, y = var_1775)[name = tensor("op_1776")]; + tensor xt_63 = add(x = var_1776, y = var_1771_1)[name = tensor("xt_63")]; + tensor var_1779 = const()[name = tensor("op_1779"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(48499456)))]; + tensor var_1780 = mul(x = xt_63, y = var_1779)[name = tensor("op_1780")]; + tensor cv_43 = cos(x = var_1780)[name = tensor("cv_43")]; + tensor var_1782 = const()[name = tensor("op_1782"), val = tensor(-0x1p-1)]; + tensor var_1783 = mul(x = cv_43, y = var_1782)[name = tensor("op_1783")]; + tensor var_1784 = const()[name = tensor("op_1784"), val = tensor(0x1p-1)]; + tensor var_1785 = add(x = var_1783, y = var_1784)[name = tensor("op_1785")]; + tensor var_1786 = const()[name = tensor("op_1786"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(48500544)))]; + tensor var_1789 = mul(x = var_1785, y = var_1786)[name = tensor("op_1789")]; + tensor input_177 = add(x = xt_63, y = var_1789)[name = tensor("input_177")]; + tensor weight_81_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(48501632))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(49222592))), name = tensor("weight_81_palettized"), shape = tensor([256, 256, 11])]; + tensor xt_65_pad_type_0 = const()[name = tensor("xt_65_pad_type_0"), val = tensor("custom")]; + tensor xt_65_pad_0 = const()[name = tensor("xt_65_pad_0"), val = tensor([5, 5])]; + tensor xt_65_strides_0 = const()[name = tensor("xt_65_strides_0"), val = tensor([1])]; + tensor xt_65_dilations_0 = const()[name = tensor("xt_65_dilations_0"), val = tensor([1])]; + tensor xt_65_groups_0 = const()[name = tensor("xt_65_groups_0"), val = tensor(1)]; + tensor xt_65 = conv(bias = generator_resblocks_2_convs2_1_bias, dilations = xt_65_dilations_0, groups = xt_65_groups_0, pad = xt_65_pad_0, pad_type = xt_65_pad_type_0, strides = xt_65_strides_0, weight = weight_81_palettized, x = input_177)[name = tensor("xt_65")]; + tensor input_179 = add(x = xt_65, y = input_171)[name = tensor("input_179")]; + tensor h_129 = linear(bias = generator_resblocks_2_adain1_2_fc_bias, weight = generator_resblocks_2_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_33")]; + tensor var_1805 = const()[name = tensor("op_1805"), val = tensor([1, 512, 1])]; + tensor h_131 = reshape(shape = var_1805, x = h_129)[name = tensor("h_131")]; + tensor var_1807_split_sizes_0 = const()[name = tensor("op_1807_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1807_axis_0 = const()[name = tensor("op_1807_axis_0"), val = tensor(1)]; + tensor var_1807_0, tensor var_1807_1 = split(axis = var_1807_axis_0, split_sizes = var_1807_split_sizes_0, x = h_131)[name = tensor("op_1807")]; + tensor var_1809_promoted = const()[name = tensor("op_1809_promoted"), val = tensor(0x1p+0)]; + tensor var_1810 = add(x = var_1807_0, y = var_1809_promoted)[name = tensor("op_1810")]; + tensor var_1811 = instance_norm(epsilon = var_1581, x = input_179)[name = tensor("op_1811")]; + tensor var_1812 = mul(x = var_1810, y = var_1811)[name = tensor("op_1812")]; + tensor xt_67 = add(x = var_1812, y = var_1807_1)[name = tensor("xt_67")]; + tensor var_1815 = const()[name = tensor("op_1815"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(49223680)))]; + tensor var_1816 = mul(x = xt_67, y = var_1815)[name = tensor("op_1816")]; + tensor cv_45 = cos(x = var_1816)[name = tensor("cv_45")]; + tensor var_1818 = const()[name = tensor("op_1818"), val = tensor(-0x1p-1)]; + tensor var_1819 = mul(x = cv_45, y = var_1818)[name = tensor("op_1819")]; + tensor var_1820 = const()[name = tensor("op_1820"), val = tensor(0x1p-1)]; + tensor var_1821 = add(x = var_1819, y = var_1820)[name = tensor("op_1821")]; + tensor var_1822 = const()[name = tensor("op_1822"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(49224768)))]; + tensor var_1825 = mul(x = var_1821, y = var_1822)[name = tensor("op_1825")]; + tensor input_181 = add(x = xt_67, y = var_1825)[name = tensor("input_181")]; + tensor weight_83_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(49225856))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(49946816))), name = tensor("weight_83_palettized"), shape = tensor([256, 256, 11])]; + tensor input_183_pad_type_0 = const()[name = tensor("input_183_pad_type_0"), val = tensor("custom")]; + tensor input_183_pad_0 = const()[name = tensor("input_183_pad_0"), val = tensor([25, 25])]; + tensor input_183_dilations_0 = const()[name = tensor("input_183_dilations_0"), val = tensor([5])]; + tensor input_183_strides_0 = const()[name = tensor("input_183_strides_0"), val = tensor([1])]; + tensor input_183_groups_0 = const()[name = tensor("input_183_groups_0"), val = tensor(1)]; + tensor input_183 = conv(bias = generator_resblocks_2_convs1_2_bias, dilations = input_183_dilations_0, groups = input_183_groups_0, pad = input_183_pad_0, pad_type = input_183_pad_type_0, strides = input_183_strides_0, weight = weight_83_palettized, x = input_181)[name = tensor("input_183")]; + tensor h_133 = linear(bias = generator_resblocks_2_adain2_2_fc_bias, weight = generator_resblocks_2_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_34")]; + tensor var_1840 = const()[name = tensor("op_1840"), val = tensor([1, 512, 1])]; + tensor h_135 = reshape(shape = var_1840, x = h_133)[name = tensor("h_135")]; + tensor var_1842_split_sizes_0 = const()[name = tensor("op_1842_split_sizes_0"), val = tensor([256, 256])]; + tensor var_1842_axis_0 = const()[name = tensor("op_1842_axis_0"), val = tensor(1)]; + tensor var_1842_0, tensor var_1842_1 = split(axis = var_1842_axis_0, split_sizes = var_1842_split_sizes_0, x = h_135)[name = tensor("op_1842")]; + tensor var_1844_promoted = const()[name = tensor("op_1844_promoted"), val = tensor(0x1p+0)]; + tensor var_1845 = add(x = var_1842_0, y = var_1844_promoted)[name = tensor("op_1845")]; + tensor var_1846 = instance_norm(epsilon = var_1581, x = input_183)[name = tensor("op_1846")]; + tensor var_1847 = mul(x = var_1845, y = var_1846)[name = tensor("op_1847")]; + tensor xt_69 = add(x = var_1847, y = var_1842_1)[name = tensor("xt_69")]; + tensor var_1850 = const()[name = tensor("op_1850"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(49947904)))]; + tensor var_1851 = mul(x = xt_69, y = var_1850)[name = tensor("op_1851")]; + tensor cv_47 = cos(x = var_1851)[name = tensor("cv_47")]; + tensor var_1853 = const()[name = tensor("op_1853"), val = tensor(-0x1p-1)]; + tensor var_1854 = mul(x = cv_47, y = var_1853)[name = tensor("op_1854")]; + tensor var_1855 = const()[name = tensor("op_1855"), val = tensor(0x1p-1)]; + tensor var_1856 = add(x = var_1854, y = var_1855)[name = tensor("op_1856")]; + tensor var_1857 = const()[name = tensor("op_1857"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(49948992)))]; + tensor var_1860 = mul(x = var_1856, y = var_1857)[name = tensor("op_1860")]; + tensor input_185 = add(x = xt_69, y = var_1860)[name = tensor("input_185")]; + tensor weight_85_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(49950080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50671040))), name = tensor("weight_85_palettized"), shape = tensor([256, 256, 11])]; + tensor xt_71_pad_type_0 = const()[name = tensor("xt_71_pad_type_0"), val = tensor("custom")]; + tensor xt_71_pad_0 = const()[name = tensor("xt_71_pad_0"), val = tensor([5, 5])]; + tensor xt_71_strides_0 = const()[name = tensor("xt_71_strides_0"), val = tensor([1])]; + tensor xt_71_dilations_0 = const()[name = tensor("xt_71_dilations_0"), val = tensor([1])]; + tensor xt_71_groups_0 = const()[name = tensor("xt_71_groups_0"), val = tensor(1)]; + tensor xt_71 = conv(bias = generator_resblocks_2_convs2_2_bias, dilations = xt_71_dilations_0, groups = xt_71_groups_0, pad = xt_71_pad_0, pad_type = xt_71_pad_type_0, strides = xt_71_strides_0, weight = weight_85_palettized, x = input_185)[name = tensor("xt_71")]; + tensor var_1869 = add(x = xt_71, y = input_179)[name = tensor("op_1869")]; + tensor xs_5 = add(x = xs_3, y = var_1869)[name = tensor("xs_5")]; + tensor _inversed_z_3_y_0 = const()[name = tensor("_inversed_z_3_y_0"), val = tensor(0x1.555556p-2)]; + tensor _inversed_z_3 = mul(x = xs_5, y = _inversed_z_3_y_0)[name = tensor("_inversed_z_3")]; + tensor var_1874 = mul(x = _inversed_z_3, y = generator_alphas_1)[name = tensor("op_1874")]; + tensor sv_3 = sin(x = var_1874)[name = tensor("sv_3")]; + tensor var_1876 = mul(x = sv_3, y = sv_3)[name = tensor("op_1876")]; + tensor _inversed_1877_y_0 = const()[name = tensor("_inversed_1877_y_0"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50672128)))]; + tensor _inversed_1877 = mul(x = var_1876, y = _inversed_1877_y_0)[name = tensor("_inversed_1877")]; + tensor input_211 = add(x = _inversed_z_3, y = _inversed_1877)[name = tensor("input_211")]; + tensor input_187_pad_type_0 = const()[name = tensor("input_187_pad_type_0"), val = tensor("custom")]; + tensor input_187_pad_0 = const()[name = tensor("input_187_pad_0"), val = tensor([3, 3])]; + tensor input_187_strides_0 = const()[name = tensor("input_187_strides_0"), val = tensor([6])]; + tensor input_187_dilations_0 = const()[name = tensor("input_187_dilations_0"), val = tensor([1])]; + tensor input_187_groups_0 = const()[name = tensor("input_187_groups_0"), val = tensor(1)]; + tensor input_187 = conv(bias = generator_noise_convs_1_bias, dilations = input_187_dilations_0, groups = input_187_groups_0, pad = input_187_pad_0, pad_type = input_187_pad_type_0, strides = input_187_strides_0, weight = generator_noise_convs_1_weight, x = input_91)[name = tensor("input_187")]; + tensor var_1900 = const()[name = tensor("op_1900"), val = tensor(0x1.4f8b58p-17)]; + tensor h_137 = linear(bias = generator_noise_res_1_adain1_0_fc_bias, weight = generator_noise_res_1_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_35")]; + tensor var_1982 = const()[name = tensor("op_1982"), val = tensor([1, 256, 1])]; + tensor h_139 = reshape(shape = var_1982, x = h_137)[name = tensor("h_139")]; + tensor var_1984_split_sizes_0 = const()[name = tensor("op_1984_split_sizes_0"), val = tensor([128, 128])]; + tensor var_1984_axis_0 = const()[name = tensor("op_1984_axis_0"), val = tensor(1)]; + tensor var_1984_0, tensor var_1984_1 = split(axis = var_1984_axis_0, split_sizes = var_1984_split_sizes_0, x = h_139)[name = tensor("op_1984")]; + tensor var_1986_promoted = const()[name = tensor("op_1986_promoted"), val = tensor(0x1p+0)]; + tensor var_1987 = add(x = var_1984_0, y = var_1986_promoted)[name = tensor("op_1987")]; + tensor var_1988 = instance_norm(epsilon = var_1900, x = input_187)[name = tensor("op_1988")]; + tensor var_1989 = mul(x = var_1987, y = var_1988)[name = tensor("op_1989")]; + tensor xt_73 = add(x = var_1989, y = var_1984_1)[name = tensor("xt_73")]; + tensor var_1992 = const()[name = tensor("op_1992"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50673216)))]; + tensor var_1993 = mul(x = xt_73, y = var_1992)[name = tensor("op_1993")]; + tensor cv_49 = cos(x = var_1993)[name = tensor("cv_49")]; + tensor var_1995 = const()[name = tensor("op_1995"), val = tensor(-0x1p-1)]; + tensor var_1996 = mul(x = cv_49, y = var_1995)[name = tensor("op_1996")]; + tensor var_1997 = const()[name = tensor("op_1997"), val = tensor(0x1p-1)]; + tensor var_1998 = add(x = var_1996, y = var_1997)[name = tensor("op_1998")]; + tensor var_1999 = const()[name = tensor("op_1999"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50673792)))]; + tensor var_2002 = mul(x = var_1998, y = var_1999)[name = tensor("op_2002")]; + tensor input_189 = add(x = xt_73, y = var_2002)[name = tensor("input_189")]; + tensor weight_89_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50674368))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50789120))), name = tensor("weight_89_palettized"), shape = tensor([128, 128, 7])]; + tensor input_191_pad_type_0 = const()[name = tensor("input_191_pad_type_0"), val = tensor("custom")]; + tensor input_191_pad_0 = const()[name = tensor("input_191_pad_0"), val = tensor([3, 3])]; + tensor input_191_strides_0 = const()[name = tensor("input_191_strides_0"), val = tensor([1])]; + tensor input_191_dilations_0 = const()[name = tensor("input_191_dilations_0"), val = tensor([1])]; + tensor input_191_groups_0 = const()[name = tensor("input_191_groups_0"), val = tensor(1)]; + tensor input_191 = conv(bias = generator_noise_res_1_convs1_0_bias, dilations = input_191_dilations_0, groups = input_191_groups_0, pad = input_191_pad_0, pad_type = input_191_pad_type_0, strides = input_191_strides_0, weight = weight_89_palettized, x = input_189)[name = tensor("input_191")]; + tensor h_141 = linear(bias = generator_noise_res_1_adain2_0_fc_bias, weight = generator_noise_res_1_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_36")]; + tensor var_2017 = const()[name = tensor("op_2017"), val = tensor([1, 256, 1])]; + tensor h_143 = reshape(shape = var_2017, x = h_141)[name = tensor("h_143")]; + tensor var_2019_split_sizes_0 = const()[name = tensor("op_2019_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2019_axis_0 = const()[name = tensor("op_2019_axis_0"), val = tensor(1)]; + tensor var_2019_0, tensor var_2019_1 = split(axis = var_2019_axis_0, split_sizes = var_2019_split_sizes_0, x = h_143)[name = tensor("op_2019")]; + tensor var_2021_promoted = const()[name = tensor("op_2021_promoted"), val = tensor(0x1p+0)]; + tensor var_2022 = add(x = var_2019_0, y = var_2021_promoted)[name = tensor("op_2022")]; + tensor var_2023 = instance_norm(epsilon = var_1900, x = input_191)[name = tensor("op_2023")]; + tensor var_2024 = mul(x = var_2022, y = var_2023)[name = tensor("op_2024")]; + tensor xt_75 = add(x = var_2024, y = var_2019_1)[name = tensor("xt_75")]; + tensor var_2027 = const()[name = tensor("op_2027"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50790208)))]; + tensor var_2028 = mul(x = xt_75, y = var_2027)[name = tensor("op_2028")]; + tensor cv_51 = cos(x = var_2028)[name = tensor("cv_51")]; + tensor var_2030 = const()[name = tensor("op_2030"), val = tensor(-0x1p-1)]; + tensor var_2031 = mul(x = cv_51, y = var_2030)[name = tensor("op_2031")]; + tensor var_2032 = const()[name = tensor("op_2032"), val = tensor(0x1p-1)]; + tensor var_2033 = add(x = var_2031, y = var_2032)[name = tensor("op_2033")]; + tensor var_2034 = const()[name = tensor("op_2034"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50790784)))]; + tensor var_2037 = mul(x = var_2033, y = var_2034)[name = tensor("op_2037")]; + tensor input_193 = add(x = xt_75, y = var_2037)[name = tensor("input_193")]; + tensor weight_91_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50791360))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50906112))), name = tensor("weight_91_palettized"), shape = tensor([128, 128, 7])]; + tensor xt_77_pad_type_0 = const()[name = tensor("xt_77_pad_type_0"), val = tensor("custom")]; + tensor xt_77_pad_0 = const()[name = tensor("xt_77_pad_0"), val = tensor([3, 3])]; + tensor xt_77_strides_0 = const()[name = tensor("xt_77_strides_0"), val = tensor([1])]; + tensor xt_77_dilations_0 = const()[name = tensor("xt_77_dilations_0"), val = tensor([1])]; + tensor xt_77_groups_0 = const()[name = tensor("xt_77_groups_0"), val = tensor(1)]; + tensor xt_77 = conv(bias = generator_noise_res_1_convs2_0_bias, dilations = xt_77_dilations_0, groups = xt_77_groups_0, pad = xt_77_pad_0, pad_type = xt_77_pad_type_0, strides = xt_77_strides_0, weight = weight_91_palettized, x = input_193)[name = tensor("xt_77")]; + tensor input_195 = add(x = xt_77, y = input_187)[name = tensor("input_195")]; + tensor h_145 = linear(bias = generator_noise_res_1_adain1_1_fc_bias, weight = generator_noise_res_1_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_37")]; + tensor var_2053 = const()[name = tensor("op_2053"), val = tensor([1, 256, 1])]; + tensor h_147 = reshape(shape = var_2053, x = h_145)[name = tensor("h_147")]; + tensor var_2055_split_sizes_0 = const()[name = tensor("op_2055_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2055_axis_0 = const()[name = tensor("op_2055_axis_0"), val = tensor(1)]; + tensor var_2055_0, tensor var_2055_1 = split(axis = var_2055_axis_0, split_sizes = var_2055_split_sizes_0, x = h_147)[name = tensor("op_2055")]; + tensor var_2057_promoted = const()[name = tensor("op_2057_promoted"), val = tensor(0x1p+0)]; + tensor var_2058 = add(x = var_2055_0, y = var_2057_promoted)[name = tensor("op_2058")]; + tensor var_2059 = instance_norm(epsilon = var_1900, x = input_195)[name = tensor("op_2059")]; + tensor var_2060 = mul(x = var_2058, y = var_2059)[name = tensor("op_2060")]; + tensor xt_79 = add(x = var_2060, y = var_2055_1)[name = tensor("xt_79")]; + tensor var_2063 = const()[name = tensor("op_2063"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50907200)))]; + tensor var_2064 = mul(x = xt_79, y = var_2063)[name = tensor("op_2064")]; + tensor cv_53 = cos(x = var_2064)[name = tensor("cv_53")]; + tensor var_2066 = const()[name = tensor("op_2066"), val = tensor(-0x1p-1)]; + tensor var_2067 = mul(x = cv_53, y = var_2066)[name = tensor("op_2067")]; + tensor var_2068 = const()[name = tensor("op_2068"), val = tensor(0x1p-1)]; + tensor var_2069 = add(x = var_2067, y = var_2068)[name = tensor("op_2069")]; + tensor var_2070 = const()[name = tensor("op_2070"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50907776)))]; + tensor var_2073 = mul(x = var_2069, y = var_2070)[name = tensor("op_2073")]; + tensor input_197 = add(x = xt_79, y = var_2073)[name = tensor("input_197")]; + tensor weight_93_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(50908352))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51023104))), name = tensor("weight_93_palettized"), shape = tensor([128, 128, 7])]; + tensor input_199_pad_type_0 = const()[name = tensor("input_199_pad_type_0"), val = tensor("custom")]; + tensor input_199_pad_0 = const()[name = tensor("input_199_pad_0"), val = tensor([9, 9])]; + tensor input_199_dilations_0 = const()[name = tensor("input_199_dilations_0"), val = tensor([3])]; + tensor input_199_strides_0 = const()[name = tensor("input_199_strides_0"), val = tensor([1])]; + tensor input_199_groups_0 = const()[name = tensor("input_199_groups_0"), val = tensor(1)]; + tensor input_199 = conv(bias = generator_noise_res_1_convs1_1_bias, dilations = input_199_dilations_0, groups = input_199_groups_0, pad = input_199_pad_0, pad_type = input_199_pad_type_0, strides = input_199_strides_0, weight = weight_93_palettized, x = input_197)[name = tensor("input_199")]; + tensor h_149 = linear(bias = generator_noise_res_1_adain2_1_fc_bias, weight = generator_noise_res_1_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_38")]; + tensor var_2088 = const()[name = tensor("op_2088"), val = tensor([1, 256, 1])]; + tensor h_151 = reshape(shape = var_2088, x = h_149)[name = tensor("h_151")]; + tensor var_2090_split_sizes_0 = const()[name = tensor("op_2090_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2090_axis_0 = const()[name = tensor("op_2090_axis_0"), val = tensor(1)]; + tensor var_2090_0, tensor var_2090_1 = split(axis = var_2090_axis_0, split_sizes = var_2090_split_sizes_0, x = h_151)[name = tensor("op_2090")]; + tensor var_2092_promoted = const()[name = tensor("op_2092_promoted"), val = tensor(0x1p+0)]; + tensor var_2093 = add(x = var_2090_0, y = var_2092_promoted)[name = tensor("op_2093")]; + tensor var_2094 = instance_norm(epsilon = var_1900, x = input_199)[name = tensor("op_2094")]; + tensor var_2095 = mul(x = var_2093, y = var_2094)[name = tensor("op_2095")]; + tensor xt_81 = add(x = var_2095, y = var_2090_1)[name = tensor("xt_81")]; + tensor var_2098 = const()[name = tensor("op_2098"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51024192)))]; + tensor var_2099 = mul(x = xt_81, y = var_2098)[name = tensor("op_2099")]; + tensor cv_55 = cos(x = var_2099)[name = tensor("cv_55")]; + tensor var_2101 = const()[name = tensor("op_2101"), val = tensor(-0x1p-1)]; + tensor var_2102 = mul(x = cv_55, y = var_2101)[name = tensor("op_2102")]; + tensor var_2103 = const()[name = tensor("op_2103"), val = tensor(0x1p-1)]; + tensor var_2104 = add(x = var_2102, y = var_2103)[name = tensor("op_2104")]; + tensor var_2105 = const()[name = tensor("op_2105"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51024768)))]; + tensor var_2108 = mul(x = var_2104, y = var_2105)[name = tensor("op_2108")]; + tensor input_201 = add(x = xt_81, y = var_2108)[name = tensor("input_201")]; + tensor weight_95_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51025344))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51140096))), name = tensor("weight_95_palettized"), shape = tensor([128, 128, 7])]; + tensor xt_83_pad_type_0 = const()[name = tensor("xt_83_pad_type_0"), val = tensor("custom")]; + tensor xt_83_pad_0 = const()[name = tensor("xt_83_pad_0"), val = tensor([3, 3])]; + tensor xt_83_strides_0 = const()[name = tensor("xt_83_strides_0"), val = tensor([1])]; + tensor xt_83_dilations_0 = const()[name = tensor("xt_83_dilations_0"), val = tensor([1])]; + tensor xt_83_groups_0 = const()[name = tensor("xt_83_groups_0"), val = tensor(1)]; + tensor xt_83 = conv(bias = generator_noise_res_1_convs2_1_bias, dilations = xt_83_dilations_0, groups = xt_83_groups_0, pad = xt_83_pad_0, pad_type = xt_83_pad_type_0, strides = xt_83_strides_0, weight = weight_95_palettized, x = input_201)[name = tensor("xt_83")]; + tensor input_203 = add(x = xt_83, y = input_195)[name = tensor("input_203")]; + tensor h_153 = linear(bias = generator_noise_res_1_adain1_2_fc_bias, weight = generator_noise_res_1_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_39")]; + tensor var_2124 = const()[name = tensor("op_2124"), val = tensor([1, 256, 1])]; + tensor h_155 = reshape(shape = var_2124, x = h_153)[name = tensor("h_155")]; + tensor var_2126_split_sizes_0 = const()[name = tensor("op_2126_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2126_axis_0 = const()[name = tensor("op_2126_axis_0"), val = tensor(1)]; + tensor var_2126_0, tensor var_2126_1 = split(axis = var_2126_axis_0, split_sizes = var_2126_split_sizes_0, x = h_155)[name = tensor("op_2126")]; + tensor var_2128_promoted = const()[name = tensor("op_2128_promoted"), val = tensor(0x1p+0)]; + tensor var_2129 = add(x = var_2126_0, y = var_2128_promoted)[name = tensor("op_2129")]; + tensor var_2130 = instance_norm(epsilon = var_1900, x = input_203)[name = tensor("op_2130")]; + tensor var_2131 = mul(x = var_2129, y = var_2130)[name = tensor("op_2131")]; + tensor xt_85 = add(x = var_2131, y = var_2126_1)[name = tensor("xt_85")]; + tensor var_2134 = const()[name = tensor("op_2134"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51141184)))]; + tensor var_2135 = mul(x = xt_85, y = var_2134)[name = tensor("op_2135")]; + tensor cv_57 = cos(x = var_2135)[name = tensor("cv_57")]; + tensor var_2137 = const()[name = tensor("op_2137"), val = tensor(-0x1p-1)]; + tensor var_2138 = mul(x = cv_57, y = var_2137)[name = tensor("op_2138")]; + tensor var_2139 = const()[name = tensor("op_2139"), val = tensor(0x1p-1)]; + tensor var_2140 = add(x = var_2138, y = var_2139)[name = tensor("op_2140")]; + tensor var_2141 = const()[name = tensor("op_2141"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51141760)))]; + tensor var_2144 = mul(x = var_2140, y = var_2141)[name = tensor("op_2144")]; + tensor input_205 = add(x = xt_85, y = var_2144)[name = tensor("input_205")]; + tensor weight_97_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51142336))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51257088))), name = tensor("weight_97_palettized"), shape = tensor([128, 128, 7])]; + tensor input_207_pad_type_0 = const()[name = tensor("input_207_pad_type_0"), val = tensor("custom")]; + tensor input_207_pad_0 = const()[name = tensor("input_207_pad_0"), val = tensor([15, 15])]; + tensor input_207_dilations_0 = const()[name = tensor("input_207_dilations_0"), val = tensor([5])]; + tensor input_207_strides_0 = const()[name = tensor("input_207_strides_0"), val = tensor([1])]; + tensor input_207_groups_0 = const()[name = tensor("input_207_groups_0"), val = tensor(1)]; + tensor input_207 = conv(bias = generator_noise_res_1_convs1_2_bias, dilations = input_207_dilations_0, groups = input_207_groups_0, pad = input_207_pad_0, pad_type = input_207_pad_type_0, strides = input_207_strides_0, weight = weight_97_palettized, x = input_205)[name = tensor("input_207")]; + tensor h_157 = linear(bias = generator_noise_res_1_adain2_2_fc_bias, weight = generator_noise_res_1_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_40")]; + tensor var_2159 = const()[name = tensor("op_2159"), val = tensor([1, 256, 1])]; + tensor h_159 = reshape(shape = var_2159, x = h_157)[name = tensor("h_159")]; + tensor var_2161_split_sizes_0 = const()[name = tensor("op_2161_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2161_axis_0 = const()[name = tensor("op_2161_axis_0"), val = tensor(1)]; + tensor var_2161_0, tensor var_2161_1 = split(axis = var_2161_axis_0, split_sizes = var_2161_split_sizes_0, x = h_159)[name = tensor("op_2161")]; + tensor var_2163_promoted = const()[name = tensor("op_2163_promoted"), val = tensor(0x1p+0)]; + tensor var_2164 = add(x = var_2161_0, y = var_2163_promoted)[name = tensor("op_2164")]; + tensor var_2165 = instance_norm(epsilon = var_1900, x = input_207)[name = tensor("op_2165")]; + tensor var_2166 = mul(x = var_2164, y = var_2165)[name = tensor("op_2166")]; + tensor xt_87 = add(x = var_2166, y = var_2161_1)[name = tensor("xt_87")]; + tensor var_2169 = const()[name = tensor("op_2169"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51258176)))]; + tensor var_2170 = mul(x = xt_87, y = var_2169)[name = tensor("op_2170")]; + tensor cv_59 = cos(x = var_2170)[name = tensor("cv_59")]; + tensor var_2172 = const()[name = tensor("op_2172"), val = tensor(-0x1p-1)]; + tensor var_2173 = mul(x = cv_59, y = var_2172)[name = tensor("op_2173")]; + tensor var_2174 = const()[name = tensor("op_2174"), val = tensor(0x1p-1)]; + tensor var_2175 = add(x = var_2173, y = var_2174)[name = tensor("op_2175")]; + tensor var_2176 = const()[name = tensor("op_2176"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51258752)))]; + tensor var_2179 = mul(x = var_2175, y = var_2176)[name = tensor("op_2179")]; + tensor input_209 = add(x = xt_87, y = var_2179)[name = tensor("input_209")]; + tensor weight_99_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51259328))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51374080))), name = tensor("weight_99_palettized"), shape = tensor([128, 128, 7])]; + tensor xt_89_pad_type_0 = const()[name = tensor("xt_89_pad_type_0"), val = tensor("custom")]; + tensor xt_89_pad_0 = const()[name = tensor("xt_89_pad_0"), val = tensor([3, 3])]; + tensor xt_89_strides_0 = const()[name = tensor("xt_89_strides_0"), val = tensor([1])]; + tensor xt_89_dilations_0 = const()[name = tensor("xt_89_dilations_0"), val = tensor([1])]; + tensor xt_89_groups_0 = const()[name = tensor("xt_89_groups_0"), val = tensor(1)]; + tensor xt_89 = conv(bias = generator_noise_res_1_convs2_2_bias, dilations = xt_89_dilations_0, groups = xt_89_groups_0, pad = xt_89_pad_0, pad_type = xt_89_pad_type_0, strides = xt_89_strides_0, weight = weight_99_palettized, x = input_209)[name = tensor("xt_89")]; + tensor x_source_3 = add(x = xt_89, y = input_203)[name = tensor("x_source_3")]; + tensor op_2190_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51375168))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51702912))), name = tensor("op_2190_palettized"), shape = tensor([256, 128, 10])]; + tensor conv_transpose_1_pad_type_0 = const()[name = tensor("conv_transpose_1_pad_type_0"), val = tensor("custom")]; + tensor conv_transpose_1_pad_0 = const()[name = tensor("conv_transpose_1_pad_0"), val = tensor([0, 0])]; + tensor conv_transpose_1_strides_0 = const()[name = tensor("conv_transpose_1_strides_0"), val = tensor([5])]; + tensor conv_transpose_1_dilations_0 = const()[name = tensor("conv_transpose_1_dilations_0"), val = tensor([1])]; + tensor conv_transpose_1_groups_0 = const()[name = tensor("conv_transpose_1_groups_0"), val = tensor(1)]; + tensor conv_transpose_1 = conv_transpose(bias = generator_ups_1_bias, dilations = conv_transpose_1_dilations_0, groups = conv_transpose_1_groups_0, pad = conv_transpose_1_pad_0, pad_type = conv_transpose_1_pad_type_0, strides = conv_transpose_1_strides_0, weight = op_2190_palettized, x = input_211)[name = tensor("conv_transpose_1")]; + tensor x_11_begin_0 = const()[name = tensor("x_11_begin_0"), val = tensor([0, 0, 3])]; + tensor x_11_end_0 = const()[name = tensor("x_11_end_0"), val = tensor([0, 0, -2])]; + tensor x_11_begin_mask_0 = const()[name = tensor("x_11_begin_mask_0"), val = tensor([true, true, false])]; + tensor x_11_end_mask_0 = const()[name = tensor("x_11_end_mask_0"), val = tensor([true, true, false])]; + tensor x_11 = slice_by_index(begin = x_11_begin_0, begin_mask = x_11_begin_mask_0, end = x_11_end_0, end_mask = x_11_end_mask_0, x = conv_transpose_1)[name = tensor("x_11")]; + tensor input_213 = add(x = x_11, y = x_source_3)[name = tensor("input_213")]; + tensor var_2209 = const()[name = tensor("op_2209"), val = tensor(0x1.4f8b58p-17)]; + tensor h_161 = linear(bias = generator_resblocks_3_adain1_0_fc_bias, weight = generator_resblocks_3_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_41")]; + tensor var_2291 = const()[name = tensor("op_2291"), val = tensor([1, 256, 1])]; + tensor h_163 = reshape(shape = var_2291, x = h_161)[name = tensor("h_163")]; + tensor var_2293_split_sizes_0 = const()[name = tensor("op_2293_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2293_axis_0 = const()[name = tensor("op_2293_axis_0"), val = tensor(1)]; + tensor var_2293_0, tensor var_2293_1 = split(axis = var_2293_axis_0, split_sizes = var_2293_split_sizes_0, x = h_163)[name = tensor("op_2293")]; + tensor var_2295_promoted = const()[name = tensor("op_2295_promoted"), val = tensor(0x1p+0)]; + tensor var_2296 = add(x = var_2293_0, y = var_2295_promoted)[name = tensor("op_2296")]; + tensor var_2297 = instance_norm(epsilon = var_2209, x = input_213)[name = tensor("op_2297")]; + tensor var_2298 = mul(x = var_2296, y = var_2297)[name = tensor("op_2298")]; + tensor xt_91 = add(x = var_2298, y = var_2293_1)[name = tensor("xt_91")]; + tensor var_2301 = const()[name = tensor("op_2301"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51704000)))]; + tensor var_2302 = mul(x = xt_91, y = var_2301)[name = tensor("op_2302")]; + tensor cv_61 = cos(x = var_2302)[name = tensor("cv_61")]; + tensor var_2304 = const()[name = tensor("op_2304"), val = tensor(-0x1p-1)]; + tensor var_2305 = mul(x = cv_61, y = var_2304)[name = tensor("op_2305")]; + tensor var_2306 = const()[name = tensor("op_2306"), val = tensor(0x1p-1)]; + tensor var_2307 = add(x = var_2305, y = var_2306)[name = tensor("op_2307")]; + tensor var_2308 = const()[name = tensor("op_2308"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51704576)))]; + tensor var_2311 = mul(x = var_2307, y = var_2308)[name = tensor("op_2311")]; + tensor input_215 = add(x = xt_91, y = var_2311)[name = tensor("input_215")]; + tensor weight_101_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51705152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51754368))), name = tensor("weight_101_palettized"), shape = tensor([128, 128, 3])]; + tensor input_217_pad_type_0 = const()[name = tensor("input_217_pad_type_0"), val = tensor("custom")]; + tensor input_217_pad_0 = const()[name = tensor("input_217_pad_0"), val = tensor([1, 1])]; + tensor input_217_strides_0 = const()[name = tensor("input_217_strides_0"), val = tensor([1])]; + tensor input_217_dilations_0 = const()[name = tensor("input_217_dilations_0"), val = tensor([1])]; + tensor input_217_groups_0 = const()[name = tensor("input_217_groups_0"), val = tensor(1)]; + tensor input_217 = conv(bias = generator_resblocks_3_convs1_0_bias, dilations = input_217_dilations_0, groups = input_217_groups_0, pad = input_217_pad_0, pad_type = input_217_pad_type_0, strides = input_217_strides_0, weight = weight_101_palettized, x = input_215)[name = tensor("input_217")]; + tensor h_165 = linear(bias = generator_resblocks_3_adain2_0_fc_bias, weight = generator_resblocks_3_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_42")]; + tensor var_2326 = const()[name = tensor("op_2326"), val = tensor([1, 256, 1])]; + tensor h_167 = reshape(shape = var_2326, x = h_165)[name = tensor("h_167")]; + tensor var_2328_split_sizes_0 = const()[name = tensor("op_2328_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2328_axis_0 = const()[name = tensor("op_2328_axis_0"), val = tensor(1)]; + tensor var_2328_0, tensor var_2328_1 = split(axis = var_2328_axis_0, split_sizes = var_2328_split_sizes_0, x = h_167)[name = tensor("op_2328")]; + tensor var_2330_promoted = const()[name = tensor("op_2330_promoted"), val = tensor(0x1p+0)]; + tensor var_2331 = add(x = var_2328_0, y = var_2330_promoted)[name = tensor("op_2331")]; + tensor var_2332 = instance_norm(epsilon = var_2209, x = input_217)[name = tensor("op_2332")]; + tensor var_2333 = mul(x = var_2331, y = var_2332)[name = tensor("op_2333")]; + tensor xt_93 = add(x = var_2333, y = var_2328_1)[name = tensor("xt_93")]; + tensor var_2336 = const()[name = tensor("op_2336"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51755456)))]; + tensor var_2337 = mul(x = xt_93, y = var_2336)[name = tensor("op_2337")]; + tensor cv_63 = cos(x = var_2337)[name = tensor("cv_63")]; + tensor var_2339 = const()[name = tensor("op_2339"), val = tensor(-0x1p-1)]; + tensor var_2340 = mul(x = cv_63, y = var_2339)[name = tensor("op_2340")]; + tensor var_2341 = const()[name = tensor("op_2341"), val = tensor(0x1p-1)]; + tensor var_2342 = add(x = var_2340, y = var_2341)[name = tensor("op_2342")]; + tensor var_2343 = const()[name = tensor("op_2343"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51756032)))]; + tensor var_2346 = mul(x = var_2342, y = var_2343)[name = tensor("op_2346")]; + tensor input_219 = add(x = xt_93, y = var_2346)[name = tensor("input_219")]; + tensor weight_103_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51756608))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51805824))), name = tensor("weight_103_palettized"), shape = tensor([128, 128, 3])]; + tensor xt_95_pad_type_0 = const()[name = tensor("xt_95_pad_type_0"), val = tensor("custom")]; + tensor xt_95_pad_0 = const()[name = tensor("xt_95_pad_0"), val = tensor([1, 1])]; + tensor xt_95_strides_0 = const()[name = tensor("xt_95_strides_0"), val = tensor([1])]; + tensor xt_95_dilations_0 = const()[name = tensor("xt_95_dilations_0"), val = tensor([1])]; + tensor xt_95_groups_0 = const()[name = tensor("xt_95_groups_0"), val = tensor(1)]; + tensor xt_95 = conv(bias = generator_resblocks_3_convs2_0_bias, dilations = xt_95_dilations_0, groups = xt_95_groups_0, pad = xt_95_pad_0, pad_type = xt_95_pad_type_0, strides = xt_95_strides_0, weight = weight_103_palettized, x = input_219)[name = tensor("xt_95")]; + tensor input_221 = add(x = xt_95, y = input_213)[name = tensor("input_221")]; + tensor h_169 = linear(bias = generator_resblocks_3_adain1_1_fc_bias, weight = generator_resblocks_3_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_43")]; + tensor var_2362 = const()[name = tensor("op_2362"), val = tensor([1, 256, 1])]; + tensor h_171 = reshape(shape = var_2362, x = h_169)[name = tensor("h_171")]; + tensor var_2364_split_sizes_0 = const()[name = tensor("op_2364_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2364_axis_0 = const()[name = tensor("op_2364_axis_0"), val = tensor(1)]; + tensor var_2364_0, tensor var_2364_1 = split(axis = var_2364_axis_0, split_sizes = var_2364_split_sizes_0, x = h_171)[name = tensor("op_2364")]; + tensor var_2366_promoted = const()[name = tensor("op_2366_promoted"), val = tensor(0x1p+0)]; + tensor var_2367 = add(x = var_2364_0, y = var_2366_promoted)[name = tensor("op_2367")]; + tensor var_2368 = instance_norm(epsilon = var_2209, x = input_221)[name = tensor("op_2368")]; + tensor var_2369 = mul(x = var_2367, y = var_2368)[name = tensor("op_2369")]; + tensor xt_97 = add(x = var_2369, y = var_2364_1)[name = tensor("xt_97")]; + tensor var_2372 = const()[name = tensor("op_2372"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51806912)))]; + tensor var_2373 = mul(x = xt_97, y = var_2372)[name = tensor("op_2373")]; + tensor cv_65 = cos(x = var_2373)[name = tensor("cv_65")]; + tensor var_2375 = const()[name = tensor("op_2375"), val = tensor(-0x1p-1)]; + tensor var_2376 = mul(x = cv_65, y = var_2375)[name = tensor("op_2376")]; + tensor var_2377 = const()[name = tensor("op_2377"), val = tensor(0x1p-1)]; + tensor var_2378 = add(x = var_2376, y = var_2377)[name = tensor("op_2378")]; + tensor var_2379 = const()[name = tensor("op_2379"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51807488)))]; + tensor var_2382 = mul(x = var_2378, y = var_2379)[name = tensor("op_2382")]; + tensor input_223 = add(x = xt_97, y = var_2382)[name = tensor("input_223")]; + tensor weight_105_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51808064))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51857280))), name = tensor("weight_105_palettized"), shape = tensor([128, 128, 3])]; + tensor input_225_pad_type_0 = const()[name = tensor("input_225_pad_type_0"), val = tensor("custom")]; + tensor input_225_pad_0 = const()[name = tensor("input_225_pad_0"), val = tensor([3, 3])]; + tensor input_225_dilations_0 = const()[name = tensor("input_225_dilations_0"), val = tensor([3])]; + tensor input_225_strides_0 = const()[name = tensor("input_225_strides_0"), val = tensor([1])]; + tensor input_225_groups_0 = const()[name = tensor("input_225_groups_0"), val = tensor(1)]; + tensor input_225 = conv(bias = generator_resblocks_3_convs1_1_bias, dilations = input_225_dilations_0, groups = input_225_groups_0, pad = input_225_pad_0, pad_type = input_225_pad_type_0, strides = input_225_strides_0, weight = weight_105_palettized, x = input_223)[name = tensor("input_225")]; + tensor h_173 = linear(bias = generator_resblocks_3_adain2_1_fc_bias, weight = generator_resblocks_3_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_44")]; + tensor var_2397 = const()[name = tensor("op_2397"), val = tensor([1, 256, 1])]; + tensor h_175 = reshape(shape = var_2397, x = h_173)[name = tensor("h_175")]; + tensor var_2399_split_sizes_0 = const()[name = tensor("op_2399_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2399_axis_0 = const()[name = tensor("op_2399_axis_0"), val = tensor(1)]; + tensor var_2399_0, tensor var_2399_1 = split(axis = var_2399_axis_0, split_sizes = var_2399_split_sizes_0, x = h_175)[name = tensor("op_2399")]; + tensor var_2401_promoted = const()[name = tensor("op_2401_promoted"), val = tensor(0x1p+0)]; + tensor var_2402 = add(x = var_2399_0, y = var_2401_promoted)[name = tensor("op_2402")]; + tensor var_2403 = instance_norm(epsilon = var_2209, x = input_225)[name = tensor("op_2403")]; + tensor var_2404 = mul(x = var_2402, y = var_2403)[name = tensor("op_2404")]; + tensor xt_99 = add(x = var_2404, y = var_2399_1)[name = tensor("xt_99")]; + tensor var_2407 = const()[name = tensor("op_2407"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51858368)))]; + tensor var_2408 = mul(x = xt_99, y = var_2407)[name = tensor("op_2408")]; + tensor cv_67 = cos(x = var_2408)[name = tensor("cv_67")]; + tensor var_2410 = const()[name = tensor("op_2410"), val = tensor(-0x1p-1)]; + tensor var_2411 = mul(x = cv_67, y = var_2410)[name = tensor("op_2411")]; + tensor var_2412 = const()[name = tensor("op_2412"), val = tensor(0x1p-1)]; + tensor var_2413 = add(x = var_2411, y = var_2412)[name = tensor("op_2413")]; + tensor var_2414 = const()[name = tensor("op_2414"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51858944)))]; + tensor var_2417 = mul(x = var_2413, y = var_2414)[name = tensor("op_2417")]; + tensor input_227 = add(x = xt_99, y = var_2417)[name = tensor("input_227")]; + tensor weight_107_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51859520))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51908736))), name = tensor("weight_107_palettized"), shape = tensor([128, 128, 3])]; + tensor xt_101_pad_type_0 = const()[name = tensor("xt_101_pad_type_0"), val = tensor("custom")]; + tensor xt_101_pad_0 = const()[name = tensor("xt_101_pad_0"), val = tensor([1, 1])]; + tensor xt_101_strides_0 = const()[name = tensor("xt_101_strides_0"), val = tensor([1])]; + tensor xt_101_dilations_0 = const()[name = tensor("xt_101_dilations_0"), val = tensor([1])]; + tensor xt_101_groups_0 = const()[name = tensor("xt_101_groups_0"), val = tensor(1)]; + tensor xt_101 = conv(bias = generator_resblocks_3_convs2_1_bias, dilations = xt_101_dilations_0, groups = xt_101_groups_0, pad = xt_101_pad_0, pad_type = xt_101_pad_type_0, strides = xt_101_strides_0, weight = weight_107_palettized, x = input_227)[name = tensor("xt_101")]; + tensor input_229 = add(x = xt_101, y = input_221)[name = tensor("input_229")]; + tensor h_177 = linear(bias = generator_resblocks_3_adain1_2_fc_bias, weight = generator_resblocks_3_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_45")]; + tensor var_2433 = const()[name = tensor("op_2433"), val = tensor([1, 256, 1])]; + tensor h_179 = reshape(shape = var_2433, x = h_177)[name = tensor("h_179")]; + tensor var_2435_split_sizes_0 = const()[name = tensor("op_2435_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2435_axis_0 = const()[name = tensor("op_2435_axis_0"), val = tensor(1)]; + tensor var_2435_0, tensor var_2435_1 = split(axis = var_2435_axis_0, split_sizes = var_2435_split_sizes_0, x = h_179)[name = tensor("op_2435")]; + tensor var_2437_promoted = const()[name = tensor("op_2437_promoted"), val = tensor(0x1p+0)]; + tensor var_2438 = add(x = var_2435_0, y = var_2437_promoted)[name = tensor("op_2438")]; + tensor var_2439 = instance_norm(epsilon = var_2209, x = input_229)[name = tensor("op_2439")]; + tensor var_2440 = mul(x = var_2438, y = var_2439)[name = tensor("op_2440")]; + tensor xt_103 = add(x = var_2440, y = var_2435_1)[name = tensor("xt_103")]; + tensor var_2443 = const()[name = tensor("op_2443"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51909824)))]; + tensor var_2444 = mul(x = xt_103, y = var_2443)[name = tensor("op_2444")]; + tensor cv_69 = cos(x = var_2444)[name = tensor("cv_69")]; + tensor var_2446 = const()[name = tensor("op_2446"), val = tensor(-0x1p-1)]; + tensor var_2447 = mul(x = cv_69, y = var_2446)[name = tensor("op_2447")]; + tensor var_2448 = const()[name = tensor("op_2448"), val = tensor(0x1p-1)]; + tensor var_2449 = add(x = var_2447, y = var_2448)[name = tensor("op_2449")]; + tensor var_2450 = const()[name = tensor("op_2450"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51910400)))]; + tensor var_2453 = mul(x = var_2449, y = var_2450)[name = tensor("op_2453")]; + tensor input_231 = add(x = xt_103, y = var_2453)[name = tensor("input_231")]; + tensor weight_109_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51910976))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51960192))), name = tensor("weight_109_palettized"), shape = tensor([128, 128, 3])]; + tensor input_233_pad_type_0 = const()[name = tensor("input_233_pad_type_0"), val = tensor("custom")]; + tensor input_233_pad_0 = const()[name = tensor("input_233_pad_0"), val = tensor([5, 5])]; + tensor input_233_dilations_0 = const()[name = tensor("input_233_dilations_0"), val = tensor([5])]; + tensor input_233_strides_0 = const()[name = tensor("input_233_strides_0"), val = tensor([1])]; + tensor input_233_groups_0 = const()[name = tensor("input_233_groups_0"), val = tensor(1)]; + tensor input_233 = conv(bias = generator_resblocks_3_convs1_2_bias, dilations = input_233_dilations_0, groups = input_233_groups_0, pad = input_233_pad_0, pad_type = input_233_pad_type_0, strides = input_233_strides_0, weight = weight_109_palettized, x = input_231)[name = tensor("input_233")]; + tensor h_181 = linear(bias = generator_resblocks_3_adain2_2_fc_bias, weight = generator_resblocks_3_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_46")]; + tensor var_2468 = const()[name = tensor("op_2468"), val = tensor([1, 256, 1])]; + tensor h_183 = reshape(shape = var_2468, x = h_181)[name = tensor("h_183")]; + tensor var_2470_split_sizes_0 = const()[name = tensor("op_2470_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2470_axis_0 = const()[name = tensor("op_2470_axis_0"), val = tensor(1)]; + tensor var_2470_0, tensor var_2470_1 = split(axis = var_2470_axis_0, split_sizes = var_2470_split_sizes_0, x = h_183)[name = tensor("op_2470")]; + tensor var_2472_promoted = const()[name = tensor("op_2472_promoted"), val = tensor(0x1p+0)]; + tensor var_2473 = add(x = var_2470_0, y = var_2472_promoted)[name = tensor("op_2473")]; + tensor var_2474 = instance_norm(epsilon = var_2209, x = input_233)[name = tensor("op_2474")]; + tensor var_2475 = mul(x = var_2473, y = var_2474)[name = tensor("op_2475")]; + tensor xt_105 = add(x = var_2475, y = var_2470_1)[name = tensor("xt_105")]; + tensor var_2478 = const()[name = tensor("op_2478"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51961280)))]; + tensor var_2479 = mul(x = xt_105, y = var_2478)[name = tensor("op_2479")]; + tensor cv_71 = cos(x = var_2479)[name = tensor("cv_71")]; + tensor var_2481 = const()[name = tensor("op_2481"), val = tensor(-0x1p-1)]; + tensor var_2482 = mul(x = cv_71, y = var_2481)[name = tensor("op_2482")]; + tensor var_2483 = const()[name = tensor("op_2483"), val = tensor(0x1p-1)]; + tensor var_2484 = add(x = var_2482, y = var_2483)[name = tensor("op_2484")]; + tensor var_2485 = const()[name = tensor("op_2485"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51961856)))]; + tensor var_2488 = mul(x = var_2484, y = var_2485)[name = tensor("op_2488")]; + tensor input_235 = add(x = xt_105, y = var_2488)[name = tensor("input_235")]; + tensor weight_111_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51962432))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52011648))), name = tensor("weight_111_palettized"), shape = tensor([128, 128, 3])]; + tensor xt_107_pad_type_0 = const()[name = tensor("xt_107_pad_type_0"), val = tensor("custom")]; + tensor xt_107_pad_0 = const()[name = tensor("xt_107_pad_0"), val = tensor([1, 1])]; + tensor xt_107_strides_0 = const()[name = tensor("xt_107_strides_0"), val = tensor([1])]; + tensor xt_107_dilations_0 = const()[name = tensor("xt_107_dilations_0"), val = tensor([1])]; + tensor xt_107_groups_0 = const()[name = tensor("xt_107_groups_0"), val = tensor(1)]; + tensor xt_107 = conv(bias = generator_resblocks_3_convs2_2_bias, dilations = xt_107_dilations_0, groups = xt_107_groups_0, pad = xt_107_pad_0, pad_type = xt_107_pad_type_0, strides = xt_107_strides_0, weight = weight_111_palettized, x = input_235)[name = tensor("xt_107")]; + tensor xs_7 = add(x = xt_107, y = input_229)[name = tensor("xs_7")]; + tensor var_2505 = const()[name = tensor("op_2505"), val = tensor(0x1.4f8b58p-17)]; + tensor h_185 = linear(bias = generator_resblocks_4_adain1_0_fc_bias, weight = generator_resblocks_4_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_47")]; + tensor var_2587 = const()[name = tensor("op_2587"), val = tensor([1, 256, 1])]; + tensor h_187 = reshape(shape = var_2587, x = h_185)[name = tensor("h_187")]; + tensor var_2589_split_sizes_0 = const()[name = tensor("op_2589_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2589_axis_0 = const()[name = tensor("op_2589_axis_0"), val = tensor(1)]; + tensor var_2589_0, tensor var_2589_1 = split(axis = var_2589_axis_0, split_sizes = var_2589_split_sizes_0, x = h_187)[name = tensor("op_2589")]; + tensor var_2591_promoted = const()[name = tensor("op_2591_promoted"), val = tensor(0x1p+0)]; + tensor var_2592 = add(x = var_2589_0, y = var_2591_promoted)[name = tensor("op_2592")]; + tensor var_2594 = mul(x = var_2592, y = var_2297)[name = tensor("op_2594")]; + tensor xt_109 = add(x = var_2594, y = var_2589_1)[name = tensor("xt_109")]; + tensor var_2597 = const()[name = tensor("op_2597"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52012736)))]; + tensor var_2598 = mul(x = xt_109, y = var_2597)[name = tensor("op_2598")]; + tensor cv_73 = cos(x = var_2598)[name = tensor("cv_73")]; + tensor var_2600 = const()[name = tensor("op_2600"), val = tensor(-0x1p-1)]; + tensor var_2601 = mul(x = cv_73, y = var_2600)[name = tensor("op_2601")]; + tensor var_2602 = const()[name = tensor("op_2602"), val = tensor(0x1p-1)]; + tensor var_2603 = add(x = var_2601, y = var_2602)[name = tensor("op_2603")]; + tensor var_2604 = const()[name = tensor("op_2604"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52013312)))]; + tensor var_2607 = mul(x = var_2603, y = var_2604)[name = tensor("op_2607")]; + tensor input_237 = add(x = xt_109, y = var_2607)[name = tensor("input_237")]; + tensor weight_113_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52013888))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52128640))), name = tensor("weight_113_palettized"), shape = tensor([128, 128, 7])]; + tensor input_239_pad_type_0 = const()[name = tensor("input_239_pad_type_0"), val = tensor("custom")]; + tensor input_239_pad_0 = const()[name = tensor("input_239_pad_0"), val = tensor([3, 3])]; + tensor input_239_strides_0 = const()[name = tensor("input_239_strides_0"), val = tensor([1])]; + tensor input_239_dilations_0 = const()[name = tensor("input_239_dilations_0"), val = tensor([1])]; + tensor input_239_groups_0 = const()[name = tensor("input_239_groups_0"), val = tensor(1)]; + tensor input_239 = conv(bias = generator_resblocks_4_convs1_0_bias, dilations = input_239_dilations_0, groups = input_239_groups_0, pad = input_239_pad_0, pad_type = input_239_pad_type_0, strides = input_239_strides_0, weight = weight_113_palettized, x = input_237)[name = tensor("input_239")]; + tensor h_189 = linear(bias = generator_resblocks_4_adain2_0_fc_bias, weight = generator_resblocks_4_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_48")]; + tensor var_2622 = const()[name = tensor("op_2622"), val = tensor([1, 256, 1])]; + tensor h_191 = reshape(shape = var_2622, x = h_189)[name = tensor("h_191")]; + tensor var_2624_split_sizes_0 = const()[name = tensor("op_2624_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2624_axis_0 = const()[name = tensor("op_2624_axis_0"), val = tensor(1)]; + tensor var_2624_0, tensor var_2624_1 = split(axis = var_2624_axis_0, split_sizes = var_2624_split_sizes_0, x = h_191)[name = tensor("op_2624")]; + tensor var_2626_promoted = const()[name = tensor("op_2626_promoted"), val = tensor(0x1p+0)]; + tensor var_2627 = add(x = var_2624_0, y = var_2626_promoted)[name = tensor("op_2627")]; + tensor var_2628 = instance_norm(epsilon = var_2505, x = input_239)[name = tensor("op_2628")]; + tensor var_2629 = mul(x = var_2627, y = var_2628)[name = tensor("op_2629")]; + tensor xt_111 = add(x = var_2629, y = var_2624_1)[name = tensor("xt_111")]; + tensor var_2632 = const()[name = tensor("op_2632"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52129728)))]; + tensor var_2633 = mul(x = xt_111, y = var_2632)[name = tensor("op_2633")]; + tensor cv_75 = cos(x = var_2633)[name = tensor("cv_75")]; + tensor var_2635 = const()[name = tensor("op_2635"), val = tensor(-0x1p-1)]; + tensor var_2636 = mul(x = cv_75, y = var_2635)[name = tensor("op_2636")]; + tensor var_2637 = const()[name = tensor("op_2637"), val = tensor(0x1p-1)]; + tensor var_2638 = add(x = var_2636, y = var_2637)[name = tensor("op_2638")]; + tensor var_2639 = const()[name = tensor("op_2639"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52130304)))]; + tensor var_2642 = mul(x = var_2638, y = var_2639)[name = tensor("op_2642")]; + tensor input_241 = add(x = xt_111, y = var_2642)[name = tensor("input_241")]; + tensor weight_115_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52130880))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52245632))), name = tensor("weight_115_palettized"), shape = tensor([128, 128, 7])]; + tensor xt_113_pad_type_0 = const()[name = tensor("xt_113_pad_type_0"), val = tensor("custom")]; + tensor xt_113_pad_0 = const()[name = tensor("xt_113_pad_0"), val = tensor([3, 3])]; + tensor xt_113_strides_0 = const()[name = tensor("xt_113_strides_0"), val = tensor([1])]; + tensor xt_113_dilations_0 = const()[name = tensor("xt_113_dilations_0"), val = tensor([1])]; + tensor xt_113_groups_0 = const()[name = tensor("xt_113_groups_0"), val = tensor(1)]; + tensor xt_113 = conv(bias = generator_resblocks_4_convs2_0_bias, dilations = xt_113_dilations_0, groups = xt_113_groups_0, pad = xt_113_pad_0, pad_type = xt_113_pad_type_0, strides = xt_113_strides_0, weight = weight_115_palettized, x = input_241)[name = tensor("xt_113")]; + tensor input_243 = add(x = xt_113, y = input_213)[name = tensor("input_243")]; + tensor h_193 = linear(bias = generator_resblocks_4_adain1_1_fc_bias, weight = generator_resblocks_4_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_49")]; + tensor var_2658 = const()[name = tensor("op_2658"), val = tensor([1, 256, 1])]; + tensor h_195 = reshape(shape = var_2658, x = h_193)[name = tensor("h_195")]; + tensor var_2660_split_sizes_0 = const()[name = tensor("op_2660_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2660_axis_0 = const()[name = tensor("op_2660_axis_0"), val = tensor(1)]; + tensor var_2660_0, tensor var_2660_1 = split(axis = var_2660_axis_0, split_sizes = var_2660_split_sizes_0, x = h_195)[name = tensor("op_2660")]; + tensor var_2662_promoted = const()[name = tensor("op_2662_promoted"), val = tensor(0x1p+0)]; + tensor var_2663 = add(x = var_2660_0, y = var_2662_promoted)[name = tensor("op_2663")]; + tensor var_2664 = instance_norm(epsilon = var_2505, x = input_243)[name = tensor("op_2664")]; + tensor var_2665 = mul(x = var_2663, y = var_2664)[name = tensor("op_2665")]; + tensor xt_115 = add(x = var_2665, y = var_2660_1)[name = tensor("xt_115")]; + tensor var_2668 = const()[name = tensor("op_2668"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52246720)))]; + tensor var_2669 = mul(x = xt_115, y = var_2668)[name = tensor("op_2669")]; + tensor cv_77 = cos(x = var_2669)[name = tensor("cv_77")]; + tensor var_2671 = const()[name = tensor("op_2671"), val = tensor(-0x1p-1)]; + tensor var_2672 = mul(x = cv_77, y = var_2671)[name = tensor("op_2672")]; + tensor var_2673 = const()[name = tensor("op_2673"), val = tensor(0x1p-1)]; + tensor var_2674 = add(x = var_2672, y = var_2673)[name = tensor("op_2674")]; + tensor var_2675 = const()[name = tensor("op_2675"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52247296)))]; + tensor var_2678 = mul(x = var_2674, y = var_2675)[name = tensor("op_2678")]; + tensor input_245 = add(x = xt_115, y = var_2678)[name = tensor("input_245")]; + tensor weight_117_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52247872))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52362624))), name = tensor("weight_117_palettized"), shape = tensor([128, 128, 7])]; + tensor input_247_pad_type_0 = const()[name = tensor("input_247_pad_type_0"), val = tensor("custom")]; + tensor input_247_pad_0 = const()[name = tensor("input_247_pad_0"), val = tensor([9, 9])]; + tensor input_247_dilations_0 = const()[name = tensor("input_247_dilations_0"), val = tensor([3])]; + tensor input_247_strides_0 = const()[name = tensor("input_247_strides_0"), val = tensor([1])]; + tensor input_247_groups_0 = const()[name = tensor("input_247_groups_0"), val = tensor(1)]; + tensor input_247 = conv(bias = generator_resblocks_4_convs1_1_bias, dilations = input_247_dilations_0, groups = input_247_groups_0, pad = input_247_pad_0, pad_type = input_247_pad_type_0, strides = input_247_strides_0, weight = weight_117_palettized, x = input_245)[name = tensor("input_247")]; + tensor h_197 = linear(bias = generator_resblocks_4_adain2_1_fc_bias, weight = generator_resblocks_4_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_50")]; + tensor var_2693 = const()[name = tensor("op_2693"), val = tensor([1, 256, 1])]; + tensor h_199 = reshape(shape = var_2693, x = h_197)[name = tensor("h_199")]; + tensor var_2695_split_sizes_0 = const()[name = tensor("op_2695_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2695_axis_0 = const()[name = tensor("op_2695_axis_0"), val = tensor(1)]; + tensor var_2695_0, tensor var_2695_1 = split(axis = var_2695_axis_0, split_sizes = var_2695_split_sizes_0, x = h_199)[name = tensor("op_2695")]; + tensor var_2697_promoted = const()[name = tensor("op_2697_promoted"), val = tensor(0x1p+0)]; + tensor var_2698 = add(x = var_2695_0, y = var_2697_promoted)[name = tensor("op_2698")]; + tensor var_2699 = instance_norm(epsilon = var_2505, x = input_247)[name = tensor("op_2699")]; + tensor var_2700 = mul(x = var_2698, y = var_2699)[name = tensor("op_2700")]; + tensor xt_117 = add(x = var_2700, y = var_2695_1)[name = tensor("xt_117")]; + tensor var_2703 = const()[name = tensor("op_2703"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52363712)))]; + tensor var_2704 = mul(x = xt_117, y = var_2703)[name = tensor("op_2704")]; + tensor cv_79 = cos(x = var_2704)[name = tensor("cv_79")]; + tensor var_2706 = const()[name = tensor("op_2706"), val = tensor(-0x1p-1)]; + tensor var_2707 = mul(x = cv_79, y = var_2706)[name = tensor("op_2707")]; + tensor var_2708 = const()[name = tensor("op_2708"), val = tensor(0x1p-1)]; + tensor var_2709 = add(x = var_2707, y = var_2708)[name = tensor("op_2709")]; + tensor var_2710 = const()[name = tensor("op_2710"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52364288)))]; + tensor var_2713 = mul(x = var_2709, y = var_2710)[name = tensor("op_2713")]; + tensor input_249 = add(x = xt_117, y = var_2713)[name = tensor("input_249")]; + tensor weight_119_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52364864))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52479616))), name = tensor("weight_119_palettized"), shape = tensor([128, 128, 7])]; + tensor xt_119_pad_type_0 = const()[name = tensor("xt_119_pad_type_0"), val = tensor("custom")]; + tensor xt_119_pad_0 = const()[name = tensor("xt_119_pad_0"), val = tensor([3, 3])]; + tensor xt_119_strides_0 = const()[name = tensor("xt_119_strides_0"), val = tensor([1])]; + tensor xt_119_dilations_0 = const()[name = tensor("xt_119_dilations_0"), val = tensor([1])]; + tensor xt_119_groups_0 = const()[name = tensor("xt_119_groups_0"), val = tensor(1)]; + tensor xt_119 = conv(bias = generator_resblocks_4_convs2_1_bias, dilations = xt_119_dilations_0, groups = xt_119_groups_0, pad = xt_119_pad_0, pad_type = xt_119_pad_type_0, strides = xt_119_strides_0, weight = weight_119_palettized, x = input_249)[name = tensor("xt_119")]; + tensor input_251 = add(x = xt_119, y = input_243)[name = tensor("input_251")]; + tensor h_201 = linear(bias = generator_resblocks_4_adain1_2_fc_bias, weight = generator_resblocks_4_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_51")]; + tensor var_2729 = const()[name = tensor("op_2729"), val = tensor([1, 256, 1])]; + tensor h_203 = reshape(shape = var_2729, x = h_201)[name = tensor("h_203")]; + tensor var_2731_split_sizes_0 = const()[name = tensor("op_2731_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2731_axis_0 = const()[name = tensor("op_2731_axis_0"), val = tensor(1)]; + tensor var_2731_0, tensor var_2731_1 = split(axis = var_2731_axis_0, split_sizes = var_2731_split_sizes_0, x = h_203)[name = tensor("op_2731")]; + tensor var_2733_promoted = const()[name = tensor("op_2733_promoted"), val = tensor(0x1p+0)]; + tensor var_2734 = add(x = var_2731_0, y = var_2733_promoted)[name = tensor("op_2734")]; + tensor var_2735 = instance_norm(epsilon = var_2505, x = input_251)[name = tensor("op_2735")]; + tensor var_2736 = mul(x = var_2734, y = var_2735)[name = tensor("op_2736")]; + tensor xt_121 = add(x = var_2736, y = var_2731_1)[name = tensor("xt_121")]; + tensor var_2739 = const()[name = tensor("op_2739"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52480704)))]; + tensor var_2740 = mul(x = xt_121, y = var_2739)[name = tensor("op_2740")]; + tensor cv_81 = cos(x = var_2740)[name = tensor("cv_81")]; + tensor var_2742 = const()[name = tensor("op_2742"), val = tensor(-0x1p-1)]; + tensor var_2743 = mul(x = cv_81, y = var_2742)[name = tensor("op_2743")]; + tensor var_2744 = const()[name = tensor("op_2744"), val = tensor(0x1p-1)]; + tensor var_2745 = add(x = var_2743, y = var_2744)[name = tensor("op_2745")]; + tensor var_2746 = const()[name = tensor("op_2746"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52481280)))]; + tensor var_2749 = mul(x = var_2745, y = var_2746)[name = tensor("op_2749")]; + tensor input_253 = add(x = xt_121, y = var_2749)[name = tensor("input_253")]; + tensor weight_121_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52481856))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52596608))), name = tensor("weight_121_palettized"), shape = tensor([128, 128, 7])]; + tensor input_255_pad_type_0 = const()[name = tensor("input_255_pad_type_0"), val = tensor("custom")]; + tensor input_255_pad_0 = const()[name = tensor("input_255_pad_0"), val = tensor([15, 15])]; + tensor input_255_dilations_0 = const()[name = tensor("input_255_dilations_0"), val = tensor([5])]; + tensor input_255_strides_0 = const()[name = tensor("input_255_strides_0"), val = tensor([1])]; + tensor input_255_groups_0 = const()[name = tensor("input_255_groups_0"), val = tensor(1)]; + tensor input_255 = conv(bias = generator_resblocks_4_convs1_2_bias, dilations = input_255_dilations_0, groups = input_255_groups_0, pad = input_255_pad_0, pad_type = input_255_pad_type_0, strides = input_255_strides_0, weight = weight_121_palettized, x = input_253)[name = tensor("input_255")]; + tensor h_205 = linear(bias = generator_resblocks_4_adain2_2_fc_bias, weight = generator_resblocks_4_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_52")]; + tensor var_2764 = const()[name = tensor("op_2764"), val = tensor([1, 256, 1])]; + tensor h_207 = reshape(shape = var_2764, x = h_205)[name = tensor("h_207")]; + tensor var_2766_split_sizes_0 = const()[name = tensor("op_2766_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2766_axis_0 = const()[name = tensor("op_2766_axis_0"), val = tensor(1)]; + tensor var_2766_0, tensor var_2766_1 = split(axis = var_2766_axis_0, split_sizes = var_2766_split_sizes_0, x = h_207)[name = tensor("op_2766")]; + tensor var_2768_promoted = const()[name = tensor("op_2768_promoted"), val = tensor(0x1p+0)]; + tensor var_2769 = add(x = var_2766_0, y = var_2768_promoted)[name = tensor("op_2769")]; + tensor var_2770 = instance_norm(epsilon = var_2505, x = input_255)[name = tensor("op_2770")]; + tensor var_2771 = mul(x = var_2769, y = var_2770)[name = tensor("op_2771")]; + tensor xt_123 = add(x = var_2771, y = var_2766_1)[name = tensor("xt_123")]; + tensor var_2774 = const()[name = tensor("op_2774"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52597696)))]; + tensor var_2775 = mul(x = xt_123, y = var_2774)[name = tensor("op_2775")]; + tensor cv_83 = cos(x = var_2775)[name = tensor("cv_83")]; + tensor var_2777 = const()[name = tensor("op_2777"), val = tensor(-0x1p-1)]; + tensor var_2778 = mul(x = cv_83, y = var_2777)[name = tensor("op_2778")]; + tensor var_2779 = const()[name = tensor("op_2779"), val = tensor(0x1p-1)]; + tensor var_2780 = add(x = var_2778, y = var_2779)[name = tensor("op_2780")]; + tensor var_2781 = const()[name = tensor("op_2781"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52598272)))]; + tensor var_2784 = mul(x = var_2780, y = var_2781)[name = tensor("op_2784")]; + tensor input_257 = add(x = xt_123, y = var_2784)[name = tensor("input_257")]; + tensor weight_123_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52598848))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52713600))), name = tensor("weight_123_palettized"), shape = tensor([128, 128, 7])]; + tensor xt_125_pad_type_0 = const()[name = tensor("xt_125_pad_type_0"), val = tensor("custom")]; + tensor xt_125_pad_0 = const()[name = tensor("xt_125_pad_0"), val = tensor([3, 3])]; + tensor xt_125_strides_0 = const()[name = tensor("xt_125_strides_0"), val = tensor([1])]; + tensor xt_125_dilations_0 = const()[name = tensor("xt_125_dilations_0"), val = tensor([1])]; + tensor xt_125_groups_0 = const()[name = tensor("xt_125_groups_0"), val = tensor(1)]; + tensor xt_125 = conv(bias = generator_resblocks_4_convs2_2_bias, dilations = xt_125_dilations_0, groups = xt_125_groups_0, pad = xt_125_pad_0, pad_type = xt_125_pad_type_0, strides = xt_125_strides_0, weight = weight_123_palettized, x = input_257)[name = tensor("xt_125")]; + tensor var_2793 = add(x = xt_125, y = input_251)[name = tensor("op_2793")]; + tensor xs_9 = add(x = xs_7, y = var_2793)[name = tensor("xs_9")]; + tensor var_2803 = const()[name = tensor("op_2803"), val = tensor(0x1.4f8b58p-17)]; + tensor h_209 = linear(bias = generator_resblocks_5_adain1_0_fc_bias, weight = generator_resblocks_5_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_53")]; + tensor var_2885 = const()[name = tensor("op_2885"), val = tensor([1, 256, 1])]; + tensor h_211 = reshape(shape = var_2885, x = h_209)[name = tensor("h_211")]; + tensor var_2887_split_sizes_0 = const()[name = tensor("op_2887_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2887_axis_0 = const()[name = tensor("op_2887_axis_0"), val = tensor(1)]; + tensor var_2887_0, tensor var_2887_1 = split(axis = var_2887_axis_0, split_sizes = var_2887_split_sizes_0, x = h_211)[name = tensor("op_2887")]; + tensor var_2889_promoted = const()[name = tensor("op_2889_promoted"), val = tensor(0x1p+0)]; + tensor var_2890 = add(x = var_2887_0, y = var_2889_promoted)[name = tensor("op_2890")]; + tensor var_2892 = mul(x = var_2890, y = var_2297)[name = tensor("op_2892")]; + tensor xt_127 = add(x = var_2892, y = var_2887_1)[name = tensor("xt_127")]; + tensor var_2895 = const()[name = tensor("op_2895"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52714688)))]; + tensor var_2896 = mul(x = xt_127, y = var_2895)[name = tensor("op_2896")]; + tensor cv_85 = cos(x = var_2896)[name = tensor("cv_85")]; + tensor var_2898 = const()[name = tensor("op_2898"), val = tensor(-0x1p-1)]; + tensor var_2899 = mul(x = cv_85, y = var_2898)[name = tensor("op_2899")]; + tensor var_2900 = const()[name = tensor("op_2900"), val = tensor(0x1p-1)]; + tensor var_2901 = add(x = var_2899, y = var_2900)[name = tensor("op_2901")]; + tensor var_2902 = const()[name = tensor("op_2902"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52715264)))]; + tensor var_2905 = mul(x = var_2901, y = var_2902)[name = tensor("op_2905")]; + tensor input_259 = add(x = xt_127, y = var_2905)[name = tensor("input_259")]; + tensor weight_125_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52715840))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52896128))), name = tensor("weight_125_palettized"), shape = tensor([128, 128, 11])]; + tensor input_261_pad_type_0 = const()[name = tensor("input_261_pad_type_0"), val = tensor("custom")]; + tensor input_261_pad_0 = const()[name = tensor("input_261_pad_0"), val = tensor([5, 5])]; + tensor input_261_strides_0 = const()[name = tensor("input_261_strides_0"), val = tensor([1])]; + tensor input_261_dilations_0 = const()[name = tensor("input_261_dilations_0"), val = tensor([1])]; + tensor input_261_groups_0 = const()[name = tensor("input_261_groups_0"), val = tensor(1)]; + tensor input_261 = conv(bias = generator_resblocks_5_convs1_0_bias, dilations = input_261_dilations_0, groups = input_261_groups_0, pad = input_261_pad_0, pad_type = input_261_pad_type_0, strides = input_261_strides_0, weight = weight_125_palettized, x = input_259)[name = tensor("input_261")]; + tensor h_213 = linear(bias = generator_resblocks_5_adain2_0_fc_bias, weight = generator_resblocks_5_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_54")]; + tensor var_2920 = const()[name = tensor("op_2920"), val = tensor([1, 256, 1])]; + tensor h_215 = reshape(shape = var_2920, x = h_213)[name = tensor("h_215")]; + tensor var_2922_split_sizes_0 = const()[name = tensor("op_2922_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2922_axis_0 = const()[name = tensor("op_2922_axis_0"), val = tensor(1)]; + tensor var_2922_0, tensor var_2922_1 = split(axis = var_2922_axis_0, split_sizes = var_2922_split_sizes_0, x = h_215)[name = tensor("op_2922")]; + tensor var_2924_promoted = const()[name = tensor("op_2924_promoted"), val = tensor(0x1p+0)]; + tensor var_2925 = add(x = var_2922_0, y = var_2924_promoted)[name = tensor("op_2925")]; + tensor var_2926 = instance_norm(epsilon = var_2803, x = input_261)[name = tensor("op_2926")]; + tensor var_2927 = mul(x = var_2925, y = var_2926)[name = tensor("op_2927")]; + tensor xt_129 = add(x = var_2927, y = var_2922_1)[name = tensor("xt_129")]; + tensor var_2930 = const()[name = tensor("op_2930"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52897216)))]; + tensor var_2931 = mul(x = xt_129, y = var_2930)[name = tensor("op_2931")]; + tensor cv_87 = cos(x = var_2931)[name = tensor("cv_87")]; + tensor var_2933 = const()[name = tensor("op_2933"), val = tensor(-0x1p-1)]; + tensor var_2934 = mul(x = cv_87, y = var_2933)[name = tensor("op_2934")]; + tensor var_2935 = const()[name = tensor("op_2935"), val = tensor(0x1p-1)]; + tensor var_2936 = add(x = var_2934, y = var_2935)[name = tensor("op_2936")]; + tensor var_2937 = const()[name = tensor("op_2937"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52897792)))]; + tensor var_2940 = mul(x = var_2936, y = var_2937)[name = tensor("op_2940")]; + tensor input_263 = add(x = xt_129, y = var_2940)[name = tensor("input_263")]; + tensor weight_127_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52898368))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53078656))), name = tensor("weight_127_palettized"), shape = tensor([128, 128, 11])]; + tensor xt_131_pad_type_0 = const()[name = tensor("xt_131_pad_type_0"), val = tensor("custom")]; + tensor xt_131_pad_0 = const()[name = tensor("xt_131_pad_0"), val = tensor([5, 5])]; + tensor xt_131_strides_0 = const()[name = tensor("xt_131_strides_0"), val = tensor([1])]; + tensor xt_131_dilations_0 = const()[name = tensor("xt_131_dilations_0"), val = tensor([1])]; + tensor xt_131_groups_0 = const()[name = tensor("xt_131_groups_0"), val = tensor(1)]; + tensor xt_131 = conv(bias = generator_resblocks_5_convs2_0_bias, dilations = xt_131_dilations_0, groups = xt_131_groups_0, pad = xt_131_pad_0, pad_type = xt_131_pad_type_0, strides = xt_131_strides_0, weight = weight_127_palettized, x = input_263)[name = tensor("xt_131")]; + tensor input_265 = add(x = xt_131, y = input_213)[name = tensor("input_265")]; + tensor h_217 = linear(bias = generator_resblocks_5_adain1_1_fc_bias, weight = generator_resblocks_5_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_55")]; + tensor var_2956 = const()[name = tensor("op_2956"), val = tensor([1, 256, 1])]; + tensor h_219 = reshape(shape = var_2956, x = h_217)[name = tensor("h_219")]; + tensor var_2958_split_sizes_0 = const()[name = tensor("op_2958_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2958_axis_0 = const()[name = tensor("op_2958_axis_0"), val = tensor(1)]; + tensor var_2958_0, tensor var_2958_1 = split(axis = var_2958_axis_0, split_sizes = var_2958_split_sizes_0, x = h_219)[name = tensor("op_2958")]; + tensor var_2960_promoted = const()[name = tensor("op_2960_promoted"), val = tensor(0x1p+0)]; + tensor var_2961 = add(x = var_2958_0, y = var_2960_promoted)[name = tensor("op_2961")]; + tensor var_2962 = instance_norm(epsilon = var_2803, x = input_265)[name = tensor("op_2962")]; + tensor var_2963 = mul(x = var_2961, y = var_2962)[name = tensor("op_2963")]; + tensor xt_133 = add(x = var_2963, y = var_2958_1)[name = tensor("xt_133")]; + tensor var_2966 = const()[name = tensor("op_2966"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53079744)))]; + tensor var_2967 = mul(x = xt_133, y = var_2966)[name = tensor("op_2967")]; + tensor cv_89 = cos(x = var_2967)[name = tensor("cv_89")]; + tensor var_2969 = const()[name = tensor("op_2969"), val = tensor(-0x1p-1)]; + tensor var_2970 = mul(x = cv_89, y = var_2969)[name = tensor("op_2970")]; + tensor var_2971 = const()[name = tensor("op_2971"), val = tensor(0x1p-1)]; + tensor var_2972 = add(x = var_2970, y = var_2971)[name = tensor("op_2972")]; + tensor var_2973 = const()[name = tensor("op_2973"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53080320)))]; + tensor var_2976 = mul(x = var_2972, y = var_2973)[name = tensor("op_2976")]; + tensor input_267 = add(x = xt_133, y = var_2976)[name = tensor("input_267")]; + tensor weight_129_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53080896))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53261184))), name = tensor("weight_129_palettized"), shape = tensor([128, 128, 11])]; + tensor input_269_pad_type_0 = const()[name = tensor("input_269_pad_type_0"), val = tensor("custom")]; + tensor input_269_pad_0 = const()[name = tensor("input_269_pad_0"), val = tensor([15, 15])]; + tensor input_269_dilations_0 = const()[name = tensor("input_269_dilations_0"), val = tensor([3])]; + tensor input_269_strides_0 = const()[name = tensor("input_269_strides_0"), val = tensor([1])]; + tensor input_269_groups_0 = const()[name = tensor("input_269_groups_0"), val = tensor(1)]; + tensor input_269 = conv(bias = generator_resblocks_5_convs1_1_bias, dilations = input_269_dilations_0, groups = input_269_groups_0, pad = input_269_pad_0, pad_type = input_269_pad_type_0, strides = input_269_strides_0, weight = weight_129_palettized, x = input_267)[name = tensor("input_269")]; + tensor h_221 = linear(bias = generator_resblocks_5_adain2_1_fc_bias, weight = generator_resblocks_5_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_56")]; + tensor var_2991 = const()[name = tensor("op_2991"), val = tensor([1, 256, 1])]; + tensor h_223 = reshape(shape = var_2991, x = h_221)[name = tensor("h_223")]; + tensor var_2993_split_sizes_0 = const()[name = tensor("op_2993_split_sizes_0"), val = tensor([128, 128])]; + tensor var_2993_axis_0 = const()[name = tensor("op_2993_axis_0"), val = tensor(1)]; + tensor var_2993_0, tensor var_2993_1 = split(axis = var_2993_axis_0, split_sizes = var_2993_split_sizes_0, x = h_223)[name = tensor("op_2993")]; + tensor var_2995_promoted = const()[name = tensor("op_2995_promoted"), val = tensor(0x1p+0)]; + tensor var_2996 = add(x = var_2993_0, y = var_2995_promoted)[name = tensor("op_2996")]; + tensor var_2997 = instance_norm(epsilon = var_2803, x = input_269)[name = tensor("op_2997")]; + tensor var_2998 = mul(x = var_2996, y = var_2997)[name = tensor("op_2998")]; + tensor xt_135 = add(x = var_2998, y = var_2993_1)[name = tensor("xt_135")]; + tensor var_3001 = const()[name = tensor("op_3001"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53262272)))]; + tensor var_3002 = mul(x = xt_135, y = var_3001)[name = tensor("op_3002")]; + tensor cv_91 = cos(x = var_3002)[name = tensor("cv_91")]; + tensor var_3004 = const()[name = tensor("op_3004"), val = tensor(-0x1p-1)]; + tensor var_3005 = mul(x = cv_91, y = var_3004)[name = tensor("op_3005")]; + tensor var_3006 = const()[name = tensor("op_3006"), val = tensor(0x1p-1)]; + tensor var_3007 = add(x = var_3005, y = var_3006)[name = tensor("op_3007")]; + tensor var_3008 = const()[name = tensor("op_3008"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53262848)))]; + tensor var_3011 = mul(x = var_3007, y = var_3008)[name = tensor("op_3011")]; + tensor input_271 = add(x = xt_135, y = var_3011)[name = tensor("input_271")]; + tensor weight_131_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53263424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53443712))), name = tensor("weight_131_palettized"), shape = tensor([128, 128, 11])]; + tensor xt_137_pad_type_0 = const()[name = tensor("xt_137_pad_type_0"), val = tensor("custom")]; + tensor xt_137_pad_0 = const()[name = tensor("xt_137_pad_0"), val = tensor([5, 5])]; + tensor xt_137_strides_0 = const()[name = tensor("xt_137_strides_0"), val = tensor([1])]; + tensor xt_137_dilations_0 = const()[name = tensor("xt_137_dilations_0"), val = tensor([1])]; + tensor xt_137_groups_0 = const()[name = tensor("xt_137_groups_0"), val = tensor(1)]; + tensor xt_137 = conv(bias = generator_resblocks_5_convs2_1_bias, dilations = xt_137_dilations_0, groups = xt_137_groups_0, pad = xt_137_pad_0, pad_type = xt_137_pad_type_0, strides = xt_137_strides_0, weight = weight_131_palettized, x = input_271)[name = tensor("xt_137")]; + tensor input_273 = add(x = xt_137, y = input_265)[name = tensor("input_273")]; + tensor h_225 = linear(bias = generator_resblocks_5_adain1_2_fc_bias, weight = generator_resblocks_5_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_57")]; + tensor var_3027 = const()[name = tensor("op_3027"), val = tensor([1, 256, 1])]; + tensor h_227 = reshape(shape = var_3027, x = h_225)[name = tensor("h_227")]; + tensor var_3029_split_sizes_0 = const()[name = tensor("op_3029_split_sizes_0"), val = tensor([128, 128])]; + tensor var_3029_axis_0 = const()[name = tensor("op_3029_axis_0"), val = tensor(1)]; + tensor var_3029_0, tensor var_3029_1 = split(axis = var_3029_axis_0, split_sizes = var_3029_split_sizes_0, x = h_227)[name = tensor("op_3029")]; + tensor var_3031_promoted = const()[name = tensor("op_3031_promoted"), val = tensor(0x1p+0)]; + tensor var_3032 = add(x = var_3029_0, y = var_3031_promoted)[name = tensor("op_3032")]; + tensor var_3033 = instance_norm(epsilon = var_2803, x = input_273)[name = tensor("op_3033")]; + tensor var_3034 = mul(x = var_3032, y = var_3033)[name = tensor("op_3034")]; + tensor xt_139 = add(x = var_3034, y = var_3029_1)[name = tensor("xt_139")]; + tensor var_3037 = const()[name = tensor("op_3037"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53444800)))]; + tensor var_3038 = mul(x = xt_139, y = var_3037)[name = tensor("op_3038")]; + tensor cv_93 = cos(x = var_3038)[name = tensor("cv_93")]; + tensor var_3040 = const()[name = tensor("op_3040"), val = tensor(-0x1p-1)]; + tensor var_3041 = mul(x = cv_93, y = var_3040)[name = tensor("op_3041")]; + tensor var_3042 = const()[name = tensor("op_3042"), val = tensor(0x1p-1)]; + tensor var_3043 = add(x = var_3041, y = var_3042)[name = tensor("op_3043")]; + tensor var_3044 = const()[name = tensor("op_3044"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53445376)))]; + tensor var_3047 = mul(x = var_3043, y = var_3044)[name = tensor("op_3047")]; + tensor input_275 = add(x = xt_139, y = var_3047)[name = tensor("input_275")]; + tensor weight_133_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53445952))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53626240))), name = tensor("weight_133_palettized"), shape = tensor([128, 128, 11])]; + tensor input_277_pad_type_0 = const()[name = tensor("input_277_pad_type_0"), val = tensor("custom")]; + tensor input_277_pad_0 = const()[name = tensor("input_277_pad_0"), val = tensor([25, 25])]; + tensor input_277_dilations_0 = const()[name = tensor("input_277_dilations_0"), val = tensor([5])]; + tensor input_277_strides_0 = const()[name = tensor("input_277_strides_0"), val = tensor([1])]; + tensor input_277_groups_0 = const()[name = tensor("input_277_groups_0"), val = tensor(1)]; + tensor input_277 = conv(bias = generator_resblocks_5_convs1_2_bias, dilations = input_277_dilations_0, groups = input_277_groups_0, pad = input_277_pad_0, pad_type = input_277_pad_type_0, strides = input_277_strides_0, weight = weight_133_palettized, x = input_275)[name = tensor("input_277")]; + tensor h_229 = linear(bias = generator_resblocks_5_adain2_2_fc_bias, weight = generator_resblocks_5_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_58")]; + tensor var_3062 = const()[name = tensor("op_3062"), val = tensor([1, 256, 1])]; + tensor h_231 = reshape(shape = var_3062, x = h_229)[name = tensor("h_231")]; + tensor var_3064_split_sizes_0 = const()[name = tensor("op_3064_split_sizes_0"), val = tensor([128, 128])]; + tensor var_3064_axis_0 = const()[name = tensor("op_3064_axis_0"), val = tensor(1)]; + tensor var_3064_0, tensor var_3064_1 = split(axis = var_3064_axis_0, split_sizes = var_3064_split_sizes_0, x = h_231)[name = tensor("op_3064")]; + tensor var_3066_promoted = const()[name = tensor("op_3066_promoted"), val = tensor(0x1p+0)]; + tensor var_3067 = add(x = var_3064_0, y = var_3066_promoted)[name = tensor("op_3067")]; + tensor var_3068 = instance_norm(epsilon = var_2803, x = input_277)[name = tensor("op_3068")]; + tensor var_3069 = mul(x = var_3067, y = var_3068)[name = tensor("op_3069")]; + tensor xt_141 = add(x = var_3069, y = var_3064_1)[name = tensor("xt_141")]; + tensor var_3072 = const()[name = tensor("op_3072"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53627328)))]; + tensor var_3073 = mul(x = xt_141, y = var_3072)[name = tensor("op_3073")]; + tensor cv_95 = cos(x = var_3073)[name = tensor("cv_95")]; + tensor var_3075 = const()[name = tensor("op_3075"), val = tensor(-0x1p-1)]; + tensor var_3076 = mul(x = cv_95, y = var_3075)[name = tensor("op_3076")]; + tensor var_3077 = const()[name = tensor("op_3077"), val = tensor(0x1p-1)]; + tensor var_3078 = add(x = var_3076, y = var_3077)[name = tensor("op_3078")]; + tensor var_3079 = const()[name = tensor("op_3079"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53627904)))]; + tensor var_3082 = mul(x = var_3078, y = var_3079)[name = tensor("op_3082")]; + tensor input_279 = add(x = xt_141, y = var_3082)[name = tensor("input_279")]; + tensor weight_135_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53628480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53808768))), name = tensor("weight_135_palettized"), shape = tensor([128, 128, 11])]; + tensor xt_143_pad_type_0 = const()[name = tensor("xt_143_pad_type_0"), val = tensor("custom")]; + tensor xt_143_pad_0 = const()[name = tensor("xt_143_pad_0"), val = tensor([5, 5])]; + tensor xt_143_strides_0 = const()[name = tensor("xt_143_strides_0"), val = tensor([1])]; + tensor xt_143_dilations_0 = const()[name = tensor("xt_143_dilations_0"), val = tensor([1])]; + tensor xt_143_groups_0 = const()[name = tensor("xt_143_groups_0"), val = tensor(1)]; + tensor xt_143 = conv(bias = generator_resblocks_5_convs2_2_bias, dilations = xt_143_dilations_0, groups = xt_143_groups_0, pad = xt_143_pad_0, pad_type = xt_143_pad_type_0, strides = xt_143_strides_0, weight = weight_135_palettized, x = input_279)[name = tensor("xt_143")]; + tensor var_3091 = add(x = xt_143, y = input_273)[name = tensor("op_3091")]; + tensor xs_11 = add(x = xs_9, y = var_3091)[name = tensor("xs_11")]; + tensor _inversed_z_5_y_0 = const()[name = tensor("_inversed_z_5_y_0"), val = tensor(0x1.555556p-2)]; + tensor _inversed_z_5 = mul(x = xs_11, y = _inversed_z_5_y_0)[name = tensor("_inversed_z_5")]; + tensor var_3096 = mul(x = _inversed_z_5, y = generator_alphas_2)[name = tensor("op_3096")]; + tensor sv_5 = sin(x = var_3096)[name = tensor("sv_5")]; + tensor var_3098 = mul(x = sv_5, y = sv_5)[name = tensor("op_3098")]; + tensor _inversed_3099_y_0 = const()[name = tensor("_inversed_3099_y_0"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53809856)))]; + tensor _inversed_3099 = mul(x = var_3098, y = _inversed_3099_y_0)[name = tensor("_inversed_3099")]; + tensor input_305 = add(x = _inversed_z_5, y = _inversed_3099)[name = tensor("input_305")]; + tensor input_281_pad_type_0 = const()[name = tensor("input_281_pad_type_0"), val = tensor("custom")]; + tensor input_281_pad_0 = const()[name = tensor("input_281_pad_0"), val = tensor([1, 1])]; + tensor input_281_strides_0 = const()[name = tensor("input_281_strides_0"), val = tensor([2])]; + tensor input_281_dilations_0 = const()[name = tensor("input_281_dilations_0"), val = tensor([1])]; + tensor input_281_groups_0 = const()[name = tensor("input_281_groups_0"), val = tensor(1)]; + tensor input_281 = conv(bias = generator_noise_convs_2_bias, dilations = input_281_dilations_0, groups = input_281_groups_0, pad = input_281_pad_0, pad_type = input_281_pad_type_0, strides = input_281_strides_0, weight = generator_noise_convs_2_weight, x = input_91)[name = tensor("input_281")]; + tensor var_3121 = const()[name = tensor("op_3121"), val = tensor(0x1.4f8b58p-17)]; + tensor h_233 = linear(bias = generator_noise_res_2_adain1_0_fc_bias, weight = generator_noise_res_2_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_59")]; + tensor var_3203 = const()[name = tensor("op_3203"), val = tensor([1, 128, 1])]; + tensor h_235 = reshape(shape = var_3203, x = h_233)[name = tensor("h_235")]; + tensor var_3205_split_sizes_0 = const()[name = tensor("op_3205_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3205_axis_0 = const()[name = tensor("op_3205_axis_0"), val = tensor(1)]; + tensor var_3205_0, tensor var_3205_1 = split(axis = var_3205_axis_0, split_sizes = var_3205_split_sizes_0, x = h_235)[name = tensor("op_3205")]; + tensor var_3207_promoted = const()[name = tensor("op_3207_promoted"), val = tensor(0x1p+0)]; + tensor var_3208 = add(x = var_3205_0, y = var_3207_promoted)[name = tensor("op_3208")]; + tensor var_3209 = instance_norm(epsilon = var_3121, x = input_281)[name = tensor("op_3209")]; + tensor var_3210 = mul(x = var_3208, y = var_3209)[name = tensor("op_3210")]; + tensor xt_145 = add(x = var_3210, y = var_3205_1)[name = tensor("xt_145")]; + tensor var_3213 = const()[name = tensor("op_3213"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53810432)))]; + tensor var_3214 = mul(x = xt_145, y = var_3213)[name = tensor("op_3214")]; + tensor cv_97 = cos(x = var_3214)[name = tensor("cv_97")]; + tensor var_3216 = const()[name = tensor("op_3216"), val = tensor(-0x1p-1)]; + tensor var_3217 = mul(x = cv_97, y = var_3216)[name = tensor("op_3217")]; + tensor var_3218 = const()[name = tensor("op_3218"), val = tensor(0x1p-1)]; + tensor var_3219 = add(x = var_3217, y = var_3218)[name = tensor("op_3219")]; + tensor var_3220 = const()[name = tensor("op_3220"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53810752)))]; + tensor var_3223 = mul(x = var_3219, y = var_3220)[name = tensor("op_3223")]; + tensor input_283 = add(x = xt_145, y = var_3223)[name = tensor("input_283")]; + tensor weight_139_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53811072))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53839808))), name = tensor("weight_139_palettized"), shape = tensor([64, 64, 7])]; + tensor input_285_pad_type_0 = const()[name = tensor("input_285_pad_type_0"), val = tensor("custom")]; + tensor input_285_pad_0 = const()[name = tensor("input_285_pad_0"), val = tensor([3, 3])]; + tensor input_285_strides_0 = const()[name = tensor("input_285_strides_0"), val = tensor([1])]; + tensor input_285_dilations_0 = const()[name = tensor("input_285_dilations_0"), val = tensor([1])]; + tensor input_285_groups_0 = const()[name = tensor("input_285_groups_0"), val = tensor(1)]; + tensor input_285 = conv(bias = generator_noise_res_2_convs1_0_bias, dilations = input_285_dilations_0, groups = input_285_groups_0, pad = input_285_pad_0, pad_type = input_285_pad_type_0, strides = input_285_strides_0, weight = weight_139_palettized, x = input_283)[name = tensor("input_285")]; + tensor h_237 = linear(bias = generator_noise_res_2_adain2_0_fc_bias, weight = generator_noise_res_2_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_60")]; + tensor var_3238 = const()[name = tensor("op_3238"), val = tensor([1, 128, 1])]; + tensor h_239 = reshape(shape = var_3238, x = h_237)[name = tensor("h_239")]; + tensor var_3240_split_sizes_0 = const()[name = tensor("op_3240_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3240_axis_0 = const()[name = tensor("op_3240_axis_0"), val = tensor(1)]; + tensor var_3240_0, tensor var_3240_1 = split(axis = var_3240_axis_0, split_sizes = var_3240_split_sizes_0, x = h_239)[name = tensor("op_3240")]; + tensor var_3242_promoted = const()[name = tensor("op_3242_promoted"), val = tensor(0x1p+0)]; + tensor var_3243 = add(x = var_3240_0, y = var_3242_promoted)[name = tensor("op_3243")]; + tensor var_3244 = instance_norm(epsilon = var_3121, x = input_285)[name = tensor("op_3244")]; + tensor var_3245 = mul(x = var_3243, y = var_3244)[name = tensor("op_3245")]; + tensor xt_147 = add(x = var_3245, y = var_3240_1)[name = tensor("xt_147")]; + tensor var_3248 = const()[name = tensor("op_3248"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53840896)))]; + tensor var_3249 = mul(x = xt_147, y = var_3248)[name = tensor("op_3249")]; + tensor cv_99 = cos(x = var_3249)[name = tensor("cv_99")]; + tensor var_3251 = const()[name = tensor("op_3251"), val = tensor(-0x1p-1)]; + tensor var_3252 = mul(x = cv_99, y = var_3251)[name = tensor("op_3252")]; + tensor var_3253 = const()[name = tensor("op_3253"), val = tensor(0x1p-1)]; + tensor var_3254 = add(x = var_3252, y = var_3253)[name = tensor("op_3254")]; + tensor var_3255 = const()[name = tensor("op_3255"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53841216)))]; + tensor var_3258 = mul(x = var_3254, y = var_3255)[name = tensor("op_3258")]; + tensor input_287 = add(x = xt_147, y = var_3258)[name = tensor("input_287")]; + tensor weight_141_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53841536))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53870272))), name = tensor("weight_141_palettized"), shape = tensor([64, 64, 7])]; + tensor xt_149_pad_type_0 = const()[name = tensor("xt_149_pad_type_0"), val = tensor("custom")]; + tensor xt_149_pad_0 = const()[name = tensor("xt_149_pad_0"), val = tensor([3, 3])]; + tensor xt_149_strides_0 = const()[name = tensor("xt_149_strides_0"), val = tensor([1])]; + tensor xt_149_dilations_0 = const()[name = tensor("xt_149_dilations_0"), val = tensor([1])]; + tensor xt_149_groups_0 = const()[name = tensor("xt_149_groups_0"), val = tensor(1)]; + tensor xt_149 = conv(bias = generator_noise_res_2_convs2_0_bias, dilations = xt_149_dilations_0, groups = xt_149_groups_0, pad = xt_149_pad_0, pad_type = xt_149_pad_type_0, strides = xt_149_strides_0, weight = weight_141_palettized, x = input_287)[name = tensor("xt_149")]; + tensor input_289 = add(x = xt_149, y = input_281)[name = tensor("input_289")]; + tensor h_241 = linear(bias = generator_noise_res_2_adain1_1_fc_bias, weight = generator_noise_res_2_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_61")]; + tensor var_3274 = const()[name = tensor("op_3274"), val = tensor([1, 128, 1])]; + tensor h_243 = reshape(shape = var_3274, x = h_241)[name = tensor("h_243")]; + tensor var_3276_split_sizes_0 = const()[name = tensor("op_3276_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3276_axis_0 = const()[name = tensor("op_3276_axis_0"), val = tensor(1)]; + tensor var_3276_0, tensor var_3276_1 = split(axis = var_3276_axis_0, split_sizes = var_3276_split_sizes_0, x = h_243)[name = tensor("op_3276")]; + tensor var_3278_promoted = const()[name = tensor("op_3278_promoted"), val = tensor(0x1p+0)]; + tensor var_3279 = add(x = var_3276_0, y = var_3278_promoted)[name = tensor("op_3279")]; + tensor var_3280 = instance_norm(epsilon = var_3121, x = input_289)[name = tensor("op_3280")]; + tensor var_3281 = mul(x = var_3279, y = var_3280)[name = tensor("op_3281")]; + tensor xt_151 = add(x = var_3281, y = var_3276_1)[name = tensor("xt_151")]; + tensor var_3284 = const()[name = tensor("op_3284"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53871360)))]; + tensor var_3285 = mul(x = xt_151, y = var_3284)[name = tensor("op_3285")]; + tensor cv_101 = cos(x = var_3285)[name = tensor("cv_101")]; + tensor var_3287 = const()[name = tensor("op_3287"), val = tensor(-0x1p-1)]; + tensor var_3288 = mul(x = cv_101, y = var_3287)[name = tensor("op_3288")]; + tensor var_3289 = const()[name = tensor("op_3289"), val = tensor(0x1p-1)]; + tensor var_3290 = add(x = var_3288, y = var_3289)[name = tensor("op_3290")]; + tensor var_3291 = const()[name = tensor("op_3291"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53871680)))]; + tensor var_3294 = mul(x = var_3290, y = var_3291)[name = tensor("op_3294")]; + tensor input_291 = add(x = xt_151, y = var_3294)[name = tensor("input_291")]; + tensor weight_143_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53872000))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53900736))), name = tensor("weight_143_palettized"), shape = tensor([64, 64, 7])]; + tensor input_293_pad_type_0 = const()[name = tensor("input_293_pad_type_0"), val = tensor("custom")]; + tensor input_293_pad_0 = const()[name = tensor("input_293_pad_0"), val = tensor([9, 9])]; + tensor input_293_dilations_0 = const()[name = tensor("input_293_dilations_0"), val = tensor([3])]; + tensor input_293_strides_0 = const()[name = tensor("input_293_strides_0"), val = tensor([1])]; + tensor input_293_groups_0 = const()[name = tensor("input_293_groups_0"), val = tensor(1)]; + tensor input_293 = conv(bias = generator_noise_res_2_convs1_1_bias, dilations = input_293_dilations_0, groups = input_293_groups_0, pad = input_293_pad_0, pad_type = input_293_pad_type_0, strides = input_293_strides_0, weight = weight_143_palettized, x = input_291)[name = tensor("input_293")]; + tensor h_245 = linear(bias = generator_noise_res_2_adain2_1_fc_bias, weight = generator_noise_res_2_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_62")]; + tensor var_3309 = const()[name = tensor("op_3309"), val = tensor([1, 128, 1])]; + tensor h_247 = reshape(shape = var_3309, x = h_245)[name = tensor("h_247")]; + tensor var_3311_split_sizes_0 = const()[name = tensor("op_3311_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3311_axis_0 = const()[name = tensor("op_3311_axis_0"), val = tensor(1)]; + tensor var_3311_0, tensor var_3311_1 = split(axis = var_3311_axis_0, split_sizes = var_3311_split_sizes_0, x = h_247)[name = tensor("op_3311")]; + tensor var_3313_promoted = const()[name = tensor("op_3313_promoted"), val = tensor(0x1p+0)]; + tensor var_3314 = add(x = var_3311_0, y = var_3313_promoted)[name = tensor("op_3314")]; + tensor var_3315 = instance_norm(epsilon = var_3121, x = input_293)[name = tensor("op_3315")]; + tensor var_3316 = mul(x = var_3314, y = var_3315)[name = tensor("op_3316")]; + tensor xt_153 = add(x = var_3316, y = var_3311_1)[name = tensor("xt_153")]; + tensor var_3319 = const()[name = tensor("op_3319"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53901824)))]; + tensor var_3320 = mul(x = xt_153, y = var_3319)[name = tensor("op_3320")]; + tensor cv_103 = cos(x = var_3320)[name = tensor("cv_103")]; + tensor var_3322 = const()[name = tensor("op_3322"), val = tensor(-0x1p-1)]; + tensor var_3323 = mul(x = cv_103, y = var_3322)[name = tensor("op_3323")]; + tensor var_3324 = const()[name = tensor("op_3324"), val = tensor(0x1p-1)]; + tensor var_3325 = add(x = var_3323, y = var_3324)[name = tensor("op_3325")]; + tensor var_3326 = const()[name = tensor("op_3326"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53902144)))]; + tensor var_3329 = mul(x = var_3325, y = var_3326)[name = tensor("op_3329")]; + tensor input_295 = add(x = xt_153, y = var_3329)[name = tensor("input_295")]; + tensor weight_145_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53902464))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53931200))), name = tensor("weight_145_palettized"), shape = tensor([64, 64, 7])]; + tensor xt_155_pad_type_0 = const()[name = tensor("xt_155_pad_type_0"), val = tensor("custom")]; + tensor xt_155_pad_0 = const()[name = tensor("xt_155_pad_0"), val = tensor([3, 3])]; + tensor xt_155_strides_0 = const()[name = tensor("xt_155_strides_0"), val = tensor([1])]; + tensor xt_155_dilations_0 = const()[name = tensor("xt_155_dilations_0"), val = tensor([1])]; + tensor xt_155_groups_0 = const()[name = tensor("xt_155_groups_0"), val = tensor(1)]; + tensor xt_155 = conv(bias = generator_noise_res_2_convs2_1_bias, dilations = xt_155_dilations_0, groups = xt_155_groups_0, pad = xt_155_pad_0, pad_type = xt_155_pad_type_0, strides = xt_155_strides_0, weight = weight_145_palettized, x = input_295)[name = tensor("xt_155")]; + tensor input_297 = add(x = xt_155, y = input_289)[name = tensor("input_297")]; + tensor h_249 = linear(bias = generator_noise_res_2_adain1_2_fc_bias, weight = generator_noise_res_2_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_63")]; + tensor var_3345 = const()[name = tensor("op_3345"), val = tensor([1, 128, 1])]; + tensor h_251 = reshape(shape = var_3345, x = h_249)[name = tensor("h_251")]; + tensor var_3347_split_sizes_0 = const()[name = tensor("op_3347_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3347_axis_0 = const()[name = tensor("op_3347_axis_0"), val = tensor(1)]; + tensor var_3347_0, tensor var_3347_1 = split(axis = var_3347_axis_0, split_sizes = var_3347_split_sizes_0, x = h_251)[name = tensor("op_3347")]; + tensor var_3349_promoted = const()[name = tensor("op_3349_promoted"), val = tensor(0x1p+0)]; + tensor var_3350 = add(x = var_3347_0, y = var_3349_promoted)[name = tensor("op_3350")]; + tensor var_3351 = instance_norm(epsilon = var_3121, x = input_297)[name = tensor("op_3351")]; + tensor var_3352 = mul(x = var_3350, y = var_3351)[name = tensor("op_3352")]; + tensor xt_157 = add(x = var_3352, y = var_3347_1)[name = tensor("xt_157")]; + tensor var_3355 = const()[name = tensor("op_3355"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53932288)))]; + tensor var_3356 = mul(x = xt_157, y = var_3355)[name = tensor("op_3356")]; + tensor cv_105 = cos(x = var_3356)[name = tensor("cv_105")]; + tensor var_3358 = const()[name = tensor("op_3358"), val = tensor(-0x1p-1)]; + tensor var_3359 = mul(x = cv_105, y = var_3358)[name = tensor("op_3359")]; + tensor var_3360 = const()[name = tensor("op_3360"), val = tensor(0x1p-1)]; + tensor var_3361 = add(x = var_3359, y = var_3360)[name = tensor("op_3361")]; + tensor var_3362 = const()[name = tensor("op_3362"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53932608)))]; + tensor var_3365 = mul(x = var_3361, y = var_3362)[name = tensor("op_3365")]; + tensor input_299 = add(x = xt_157, y = var_3365)[name = tensor("input_299")]; + tensor weight_147_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53932928))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53961664))), name = tensor("weight_147_palettized"), shape = tensor([64, 64, 7])]; + tensor input_301_pad_type_0 = const()[name = tensor("input_301_pad_type_0"), val = tensor("custom")]; + tensor input_301_pad_0 = const()[name = tensor("input_301_pad_0"), val = tensor([15, 15])]; + tensor input_301_dilations_0 = const()[name = tensor("input_301_dilations_0"), val = tensor([5])]; + tensor input_301_strides_0 = const()[name = tensor("input_301_strides_0"), val = tensor([1])]; + tensor input_301_groups_0 = const()[name = tensor("input_301_groups_0"), val = tensor(1)]; + tensor input_301 = conv(bias = generator_noise_res_2_convs1_2_bias, dilations = input_301_dilations_0, groups = input_301_groups_0, pad = input_301_pad_0, pad_type = input_301_pad_type_0, strides = input_301_strides_0, weight = weight_147_palettized, x = input_299)[name = tensor("input_301")]; + tensor h_253 = linear(bias = generator_noise_res_2_adain2_2_fc_bias, weight = generator_noise_res_2_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_64")]; + tensor var_3380 = const()[name = tensor("op_3380"), val = tensor([1, 128, 1])]; + tensor h_255 = reshape(shape = var_3380, x = h_253)[name = tensor("h_255")]; + tensor var_3382_split_sizes_0 = const()[name = tensor("op_3382_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3382_axis_0 = const()[name = tensor("op_3382_axis_0"), val = tensor(1)]; + tensor var_3382_0, tensor var_3382_1 = split(axis = var_3382_axis_0, split_sizes = var_3382_split_sizes_0, x = h_255)[name = tensor("op_3382")]; + tensor var_3384_promoted = const()[name = tensor("op_3384_promoted"), val = tensor(0x1p+0)]; + tensor var_3385 = add(x = var_3382_0, y = var_3384_promoted)[name = tensor("op_3385")]; + tensor var_3386 = instance_norm(epsilon = var_3121, x = input_301)[name = tensor("op_3386")]; + tensor var_3387 = mul(x = var_3385, y = var_3386)[name = tensor("op_3387")]; + tensor xt_159 = add(x = var_3387, y = var_3382_1)[name = tensor("xt_159")]; + tensor var_3390 = const()[name = tensor("op_3390"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53962752)))]; + tensor var_3391 = mul(x = xt_159, y = var_3390)[name = tensor("op_3391")]; + tensor cv_107 = cos(x = var_3391)[name = tensor("cv_107")]; + tensor var_3393 = const()[name = tensor("op_3393"), val = tensor(-0x1p-1)]; + tensor var_3394 = mul(x = cv_107, y = var_3393)[name = tensor("op_3394")]; + tensor var_3395 = const()[name = tensor("op_3395"), val = tensor(0x1p-1)]; + tensor var_3396 = add(x = var_3394, y = var_3395)[name = tensor("op_3396")]; + tensor var_3397 = const()[name = tensor("op_3397"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53963072)))]; + tensor var_3400 = mul(x = var_3396, y = var_3397)[name = tensor("op_3400")]; + tensor input_303 = add(x = xt_159, y = var_3400)[name = tensor("input_303")]; + tensor weight_149_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53963392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53992128))), name = tensor("weight_149_palettized"), shape = tensor([64, 64, 7])]; + tensor xt_161_pad_type_0 = const()[name = tensor("xt_161_pad_type_0"), val = tensor("custom")]; + tensor xt_161_pad_0 = const()[name = tensor("xt_161_pad_0"), val = tensor([3, 3])]; + tensor xt_161_strides_0 = const()[name = tensor("xt_161_strides_0"), val = tensor([1])]; + tensor xt_161_dilations_0 = const()[name = tensor("xt_161_dilations_0"), val = tensor([1])]; + tensor xt_161_groups_0 = const()[name = tensor("xt_161_groups_0"), val = tensor(1)]; + tensor xt_161 = conv(bias = generator_noise_res_2_convs2_2_bias, dilations = xt_161_dilations_0, groups = xt_161_groups_0, pad = xt_161_pad_0, pad_type = xt_161_pad_type_0, strides = xt_161_strides_0, weight = weight_149_palettized, x = input_303)[name = tensor("xt_161")]; + tensor x_source_5 = add(x = xt_161, y = input_297)[name = tensor("x_source_5")]; + tensor op_3411_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(53993216))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54042432))), name = tensor("op_3411_palettized"), shape = tensor([128, 64, 6])]; + tensor conv_transpose_2_pad_type_0 = const()[name = tensor("conv_transpose_2_pad_type_0"), val = tensor("custom")]; + tensor conv_transpose_2_pad_0 = const()[name = tensor("conv_transpose_2_pad_0"), val = tensor([0, 0])]; + tensor conv_transpose_2_strides_0 = const()[name = tensor("conv_transpose_2_strides_0"), val = tensor([3])]; + tensor conv_transpose_2_dilations_0 = const()[name = tensor("conv_transpose_2_dilations_0"), val = tensor([1])]; + tensor conv_transpose_2_groups_0 = const()[name = tensor("conv_transpose_2_groups_0"), val = tensor(1)]; + tensor conv_transpose_2 = conv_transpose(bias = generator_ups_2_bias, dilations = conv_transpose_2_dilations_0, groups = conv_transpose_2_groups_0, pad = conv_transpose_2_pad_0, pad_type = conv_transpose_2_pad_type_0, strides = conv_transpose_2_strides_0, weight = op_3411_palettized, x = input_305)[name = tensor("conv_transpose_2")]; + tensor x_13_begin_0 = const()[name = tensor("x_13_begin_0"), val = tensor([0, 0, 2])]; + tensor x_13_end_0 = const()[name = tensor("x_13_end_0"), val = tensor([0, 0, -1])]; + tensor x_13_begin_mask_0 = const()[name = tensor("x_13_begin_mask_0"), val = tensor([true, true, false])]; + tensor x_13_end_mask_0 = const()[name = tensor("x_13_end_mask_0"), val = tensor([true, true, false])]; + tensor x_13 = slice_by_index(begin = x_13_begin_0, begin_mask = x_13_begin_mask_0, end = x_13_end_0, end_mask = x_13_end_mask_0, x = conv_transpose_2)[name = tensor("x_13")]; + tensor input_307 = add(x = x_13, y = x_source_5)[name = tensor("input_307")]; + tensor var_3430 = const()[name = tensor("op_3430"), val = tensor(0x1.4f8b58p-17)]; + tensor h_257 = linear(bias = generator_resblocks_6_adain1_0_fc_bias, weight = generator_resblocks_6_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_65")]; + tensor var_3512 = const()[name = tensor("op_3512"), val = tensor([1, 128, 1])]; + tensor h_259 = reshape(shape = var_3512, x = h_257)[name = tensor("h_259")]; + tensor var_3514_split_sizes_0 = const()[name = tensor("op_3514_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3514_axis_0 = const()[name = tensor("op_3514_axis_0"), val = tensor(1)]; + tensor var_3514_0, tensor var_3514_1 = split(axis = var_3514_axis_0, split_sizes = var_3514_split_sizes_0, x = h_259)[name = tensor("op_3514")]; + tensor var_3516_promoted = const()[name = tensor("op_3516_promoted"), val = tensor(0x1p+0)]; + tensor var_3517 = add(x = var_3514_0, y = var_3516_promoted)[name = tensor("op_3517")]; + tensor var_3518 = instance_norm(epsilon = var_3430, x = input_307)[name = tensor("op_3518")]; + tensor var_3519 = mul(x = var_3517, y = var_3518)[name = tensor("op_3519")]; + tensor xt_163 = add(x = var_3519, y = var_3514_1)[name = tensor("xt_163")]; + tensor var_3522 = const()[name = tensor("op_3522"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54043520)))]; + tensor var_3523 = mul(x = xt_163, y = var_3522)[name = tensor("op_3523")]; + tensor cv_109 = cos(x = var_3523)[name = tensor("cv_109")]; + tensor var_3525 = const()[name = tensor("op_3525"), val = tensor(-0x1p-1)]; + tensor var_3526 = mul(x = cv_109, y = var_3525)[name = tensor("op_3526")]; + tensor var_3527 = const()[name = tensor("op_3527"), val = tensor(0x1p-1)]; + tensor var_3528 = add(x = var_3526, y = var_3527)[name = tensor("op_3528")]; + tensor var_3529 = const()[name = tensor("op_3529"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54043840)))]; + tensor var_3532 = mul(x = var_3528, y = var_3529)[name = tensor("op_3532")]; + tensor input_309 = add(x = xt_163, y = var_3532)[name = tensor("input_309")]; + tensor weight_151_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54044160))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54056512))), name = tensor("weight_151_palettized"), shape = tensor([64, 64, 3])]; + tensor input_311_pad_type_0 = const()[name = tensor("input_311_pad_type_0"), val = tensor("custom")]; + tensor input_311_pad_0 = const()[name = tensor("input_311_pad_0"), val = tensor([1, 1])]; + tensor input_311_strides_0 = const()[name = tensor("input_311_strides_0"), val = tensor([1])]; + tensor input_311_dilations_0 = const()[name = tensor("input_311_dilations_0"), val = tensor([1])]; + tensor input_311_groups_0 = const()[name = tensor("input_311_groups_0"), val = tensor(1)]; + tensor input_311 = conv(bias = generator_resblocks_6_convs1_0_bias, dilations = input_311_dilations_0, groups = input_311_groups_0, pad = input_311_pad_0, pad_type = input_311_pad_type_0, strides = input_311_strides_0, weight = weight_151_palettized, x = input_309)[name = tensor("input_311")]; + tensor h_261 = linear(bias = generator_resblocks_6_adain2_0_fc_bias, weight = generator_resblocks_6_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_66")]; + tensor var_3547 = const()[name = tensor("op_3547"), val = tensor([1, 128, 1])]; + tensor h_263 = reshape(shape = var_3547, x = h_261)[name = tensor("h_263")]; + tensor var_3549_split_sizes_0 = const()[name = tensor("op_3549_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3549_axis_0 = const()[name = tensor("op_3549_axis_0"), val = tensor(1)]; + tensor var_3549_0, tensor var_3549_1 = split(axis = var_3549_axis_0, split_sizes = var_3549_split_sizes_0, x = h_263)[name = tensor("op_3549")]; + tensor var_3551_promoted = const()[name = tensor("op_3551_promoted"), val = tensor(0x1p+0)]; + tensor var_3552 = add(x = var_3549_0, y = var_3551_promoted)[name = tensor("op_3552")]; + tensor var_3553 = instance_norm(epsilon = var_3430, x = input_311)[name = tensor("op_3553")]; + tensor var_3554 = mul(x = var_3552, y = var_3553)[name = tensor("op_3554")]; + tensor xt_165 = add(x = var_3554, y = var_3549_1)[name = tensor("xt_165")]; + tensor var_3557 = const()[name = tensor("op_3557"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54057600)))]; + tensor var_3558 = mul(x = xt_165, y = var_3557)[name = tensor("op_3558")]; + tensor cv_111 = cos(x = var_3558)[name = tensor("cv_111")]; + tensor var_3560 = const()[name = tensor("op_3560"), val = tensor(-0x1p-1)]; + tensor var_3561 = mul(x = cv_111, y = var_3560)[name = tensor("op_3561")]; + tensor var_3562 = const()[name = tensor("op_3562"), val = tensor(0x1p-1)]; + tensor var_3563 = add(x = var_3561, y = var_3562)[name = tensor("op_3563")]; + tensor var_3564 = const()[name = tensor("op_3564"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54057920)))]; + tensor var_3567 = mul(x = var_3563, y = var_3564)[name = tensor("op_3567")]; + tensor input_313 = add(x = xt_165, y = var_3567)[name = tensor("input_313")]; + tensor weight_153_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54058240))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54070592))), name = tensor("weight_153_palettized"), shape = tensor([64, 64, 3])]; + tensor xt_167_pad_type_0 = const()[name = tensor("xt_167_pad_type_0"), val = tensor("custom")]; + tensor xt_167_pad_0 = const()[name = tensor("xt_167_pad_0"), val = tensor([1, 1])]; + tensor xt_167_strides_0 = const()[name = tensor("xt_167_strides_0"), val = tensor([1])]; + tensor xt_167_dilations_0 = const()[name = tensor("xt_167_dilations_0"), val = tensor([1])]; + tensor xt_167_groups_0 = const()[name = tensor("xt_167_groups_0"), val = tensor(1)]; + tensor xt_167 = conv(bias = generator_resblocks_6_convs2_0_bias, dilations = xt_167_dilations_0, groups = xt_167_groups_0, pad = xt_167_pad_0, pad_type = xt_167_pad_type_0, strides = xt_167_strides_0, weight = weight_153_palettized, x = input_313)[name = tensor("xt_167")]; + tensor input_315 = add(x = xt_167, y = input_307)[name = tensor("input_315")]; + tensor h_265 = linear(bias = generator_resblocks_6_adain1_1_fc_bias, weight = generator_resblocks_6_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_67")]; + tensor var_3583 = const()[name = tensor("op_3583"), val = tensor([1, 128, 1])]; + tensor h_267 = reshape(shape = var_3583, x = h_265)[name = tensor("h_267")]; + tensor var_3585_split_sizes_0 = const()[name = tensor("op_3585_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3585_axis_0 = const()[name = tensor("op_3585_axis_0"), val = tensor(1)]; + tensor var_3585_0, tensor var_3585_1 = split(axis = var_3585_axis_0, split_sizes = var_3585_split_sizes_0, x = h_267)[name = tensor("op_3585")]; + tensor var_3587_promoted = const()[name = tensor("op_3587_promoted"), val = tensor(0x1p+0)]; + tensor var_3588 = add(x = var_3585_0, y = var_3587_promoted)[name = tensor("op_3588")]; + tensor var_3589 = instance_norm(epsilon = var_3430, x = input_315)[name = tensor("op_3589")]; + tensor var_3590 = mul(x = var_3588, y = var_3589)[name = tensor("op_3590")]; + tensor xt_169 = add(x = var_3590, y = var_3585_1)[name = tensor("xt_169")]; + tensor var_3593 = const()[name = tensor("op_3593"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54071680)))]; + tensor var_3594 = mul(x = xt_169, y = var_3593)[name = tensor("op_3594")]; + tensor cv_113 = cos(x = var_3594)[name = tensor("cv_113")]; + tensor var_3596 = const()[name = tensor("op_3596"), val = tensor(-0x1p-1)]; + tensor var_3597 = mul(x = cv_113, y = var_3596)[name = tensor("op_3597")]; + tensor var_3598 = const()[name = tensor("op_3598"), val = tensor(0x1p-1)]; + tensor var_3599 = add(x = var_3597, y = var_3598)[name = tensor("op_3599")]; + tensor var_3600 = const()[name = tensor("op_3600"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54072000)))]; + tensor var_3603 = mul(x = var_3599, y = var_3600)[name = tensor("op_3603")]; + tensor input_317 = add(x = xt_169, y = var_3603)[name = tensor("input_317")]; + tensor weight_155_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54072320))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54084672))), name = tensor("weight_155_palettized"), shape = tensor([64, 64, 3])]; + tensor input_319_pad_type_0 = const()[name = tensor("input_319_pad_type_0"), val = tensor("custom")]; + tensor input_319_pad_0 = const()[name = tensor("input_319_pad_0"), val = tensor([3, 3])]; + tensor input_319_dilations_0 = const()[name = tensor("input_319_dilations_0"), val = tensor([3])]; + tensor input_319_strides_0 = const()[name = tensor("input_319_strides_0"), val = tensor([1])]; + tensor input_319_groups_0 = const()[name = tensor("input_319_groups_0"), val = tensor(1)]; + tensor input_319 = conv(bias = generator_resblocks_6_convs1_1_bias, dilations = input_319_dilations_0, groups = input_319_groups_0, pad = input_319_pad_0, pad_type = input_319_pad_type_0, strides = input_319_strides_0, weight = weight_155_palettized, x = input_317)[name = tensor("input_319")]; + tensor h_269 = linear(bias = generator_resblocks_6_adain2_1_fc_bias, weight = generator_resblocks_6_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_68")]; + tensor var_3618 = const()[name = tensor("op_3618"), val = tensor([1, 128, 1])]; + tensor h_271 = reshape(shape = var_3618, x = h_269)[name = tensor("h_271")]; + tensor var_3620_split_sizes_0 = const()[name = tensor("op_3620_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3620_axis_0 = const()[name = tensor("op_3620_axis_0"), val = tensor(1)]; + tensor var_3620_0, tensor var_3620_1 = split(axis = var_3620_axis_0, split_sizes = var_3620_split_sizes_0, x = h_271)[name = tensor("op_3620")]; + tensor var_3622_promoted = const()[name = tensor("op_3622_promoted"), val = tensor(0x1p+0)]; + tensor var_3623 = add(x = var_3620_0, y = var_3622_promoted)[name = tensor("op_3623")]; + tensor var_3624 = instance_norm(epsilon = var_3430, x = input_319)[name = tensor("op_3624")]; + tensor var_3625 = mul(x = var_3623, y = var_3624)[name = tensor("op_3625")]; + tensor xt_171 = add(x = var_3625, y = var_3620_1)[name = tensor("xt_171")]; + tensor var_3628 = const()[name = tensor("op_3628"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54085760)))]; + tensor var_3629 = mul(x = xt_171, y = var_3628)[name = tensor("op_3629")]; + tensor cv_115 = cos(x = var_3629)[name = tensor("cv_115")]; + tensor var_3631 = const()[name = tensor("op_3631"), val = tensor(-0x1p-1)]; + tensor var_3632 = mul(x = cv_115, y = var_3631)[name = tensor("op_3632")]; + tensor var_3633 = const()[name = tensor("op_3633"), val = tensor(0x1p-1)]; + tensor var_3634 = add(x = var_3632, y = var_3633)[name = tensor("op_3634")]; + tensor var_3635 = const()[name = tensor("op_3635"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54086080)))]; + tensor var_3638 = mul(x = var_3634, y = var_3635)[name = tensor("op_3638")]; + tensor input_321 = add(x = xt_171, y = var_3638)[name = tensor("input_321")]; + tensor weight_157_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54086400))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54098752))), name = tensor("weight_157_palettized"), shape = tensor([64, 64, 3])]; + tensor xt_173_pad_type_0 = const()[name = tensor("xt_173_pad_type_0"), val = tensor("custom")]; + tensor xt_173_pad_0 = const()[name = tensor("xt_173_pad_0"), val = tensor([1, 1])]; + tensor xt_173_strides_0 = const()[name = tensor("xt_173_strides_0"), val = tensor([1])]; + tensor xt_173_dilations_0 = const()[name = tensor("xt_173_dilations_0"), val = tensor([1])]; + tensor xt_173_groups_0 = const()[name = tensor("xt_173_groups_0"), val = tensor(1)]; + tensor xt_173 = conv(bias = generator_resblocks_6_convs2_1_bias, dilations = xt_173_dilations_0, groups = xt_173_groups_0, pad = xt_173_pad_0, pad_type = xt_173_pad_type_0, strides = xt_173_strides_0, weight = weight_157_palettized, x = input_321)[name = tensor("xt_173")]; + tensor input_323 = add(x = xt_173, y = input_315)[name = tensor("input_323")]; + tensor h_273 = linear(bias = generator_resblocks_6_adain1_2_fc_bias, weight = generator_resblocks_6_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_69")]; + tensor var_3654 = const()[name = tensor("op_3654"), val = tensor([1, 128, 1])]; + tensor h_275 = reshape(shape = var_3654, x = h_273)[name = tensor("h_275")]; + tensor var_3656_split_sizes_0 = const()[name = tensor("op_3656_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3656_axis_0 = const()[name = tensor("op_3656_axis_0"), val = tensor(1)]; + tensor var_3656_0, tensor var_3656_1 = split(axis = var_3656_axis_0, split_sizes = var_3656_split_sizes_0, x = h_275)[name = tensor("op_3656")]; + tensor var_3658_promoted = const()[name = tensor("op_3658_promoted"), val = tensor(0x1p+0)]; + tensor var_3659 = add(x = var_3656_0, y = var_3658_promoted)[name = tensor("op_3659")]; + tensor var_3660 = instance_norm(epsilon = var_3430, x = input_323)[name = tensor("op_3660")]; + tensor var_3661 = mul(x = var_3659, y = var_3660)[name = tensor("op_3661")]; + tensor xt_175 = add(x = var_3661, y = var_3656_1)[name = tensor("xt_175")]; + tensor var_3664 = const()[name = tensor("op_3664"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54099840)))]; + tensor var_3665 = mul(x = xt_175, y = var_3664)[name = tensor("op_3665")]; + tensor cv_117 = cos(x = var_3665)[name = tensor("cv_117")]; + tensor var_3667 = const()[name = tensor("op_3667"), val = tensor(-0x1p-1)]; + tensor var_3668 = mul(x = cv_117, y = var_3667)[name = tensor("op_3668")]; + tensor var_3669 = const()[name = tensor("op_3669"), val = tensor(0x1p-1)]; + tensor var_3670 = add(x = var_3668, y = var_3669)[name = tensor("op_3670")]; + tensor var_3671 = const()[name = tensor("op_3671"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54100160)))]; + tensor var_3674 = mul(x = var_3670, y = var_3671)[name = tensor("op_3674")]; + tensor input_325 = add(x = xt_175, y = var_3674)[name = tensor("input_325")]; + tensor weight_159_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54100480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54112832))), name = tensor("weight_159_palettized"), shape = tensor([64, 64, 3])]; + tensor input_327_pad_type_0 = const()[name = tensor("input_327_pad_type_0"), val = tensor("custom")]; + tensor input_327_pad_0 = const()[name = tensor("input_327_pad_0"), val = tensor([5, 5])]; + tensor input_327_dilations_0 = const()[name = tensor("input_327_dilations_0"), val = tensor([5])]; + tensor input_327_strides_0 = const()[name = tensor("input_327_strides_0"), val = tensor([1])]; + tensor input_327_groups_0 = const()[name = tensor("input_327_groups_0"), val = tensor(1)]; + tensor input_327 = conv(bias = generator_resblocks_6_convs1_2_bias, dilations = input_327_dilations_0, groups = input_327_groups_0, pad = input_327_pad_0, pad_type = input_327_pad_type_0, strides = input_327_strides_0, weight = weight_159_palettized, x = input_325)[name = tensor("input_327")]; + tensor h_277 = linear(bias = generator_resblocks_6_adain2_2_fc_bias, weight = generator_resblocks_6_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_70")]; + tensor var_3689 = const()[name = tensor("op_3689"), val = tensor([1, 128, 1])]; + tensor h_279 = reshape(shape = var_3689, x = h_277)[name = tensor("h_279")]; + tensor var_3691_split_sizes_0 = const()[name = tensor("op_3691_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3691_axis_0 = const()[name = tensor("op_3691_axis_0"), val = tensor(1)]; + tensor var_3691_0, tensor var_3691_1 = split(axis = var_3691_axis_0, split_sizes = var_3691_split_sizes_0, x = h_279)[name = tensor("op_3691")]; + tensor var_3693_promoted = const()[name = tensor("op_3693_promoted"), val = tensor(0x1p+0)]; + tensor var_3694 = add(x = var_3691_0, y = var_3693_promoted)[name = tensor("op_3694")]; + tensor var_3695 = instance_norm(epsilon = var_3430, x = input_327)[name = tensor("op_3695")]; + tensor var_3696 = mul(x = var_3694, y = var_3695)[name = tensor("op_3696")]; + tensor xt_177 = add(x = var_3696, y = var_3691_1)[name = tensor("xt_177")]; + tensor var_3699 = const()[name = tensor("op_3699"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54113920)))]; + tensor var_3700 = mul(x = xt_177, y = var_3699)[name = tensor("op_3700")]; + tensor cv_119 = cos(x = var_3700)[name = tensor("cv_119")]; + tensor var_3702 = const()[name = tensor("op_3702"), val = tensor(-0x1p-1)]; + tensor var_3703 = mul(x = cv_119, y = var_3702)[name = tensor("op_3703")]; + tensor var_3704 = const()[name = tensor("op_3704"), val = tensor(0x1p-1)]; + tensor var_3705 = add(x = var_3703, y = var_3704)[name = tensor("op_3705")]; + tensor var_3706 = const()[name = tensor("op_3706"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54114240)))]; + tensor var_3709 = mul(x = var_3705, y = var_3706)[name = tensor("op_3709")]; + tensor input_329 = add(x = xt_177, y = var_3709)[name = tensor("input_329")]; + tensor weight_161_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54114560))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54126912))), name = tensor("weight_161_palettized"), shape = tensor([64, 64, 3])]; + tensor xt_179_pad_type_0 = const()[name = tensor("xt_179_pad_type_0"), val = tensor("custom")]; + tensor xt_179_pad_0 = const()[name = tensor("xt_179_pad_0"), val = tensor([1, 1])]; + tensor xt_179_strides_0 = const()[name = tensor("xt_179_strides_0"), val = tensor([1])]; + tensor xt_179_dilations_0 = const()[name = tensor("xt_179_dilations_0"), val = tensor([1])]; + tensor xt_179_groups_0 = const()[name = tensor("xt_179_groups_0"), val = tensor(1)]; + tensor xt_179 = conv(bias = generator_resblocks_6_convs2_2_bias, dilations = xt_179_dilations_0, groups = xt_179_groups_0, pad = xt_179_pad_0, pad_type = xt_179_pad_type_0, strides = xt_179_strides_0, weight = weight_161_palettized, x = input_329)[name = tensor("xt_179")]; + tensor xs_13 = add(x = xt_179, y = input_323)[name = tensor("xs_13")]; + tensor var_3726 = const()[name = tensor("op_3726"), val = tensor(0x1.4f8b58p-17)]; + tensor h_281 = linear(bias = generator_resblocks_7_adain1_0_fc_bias, weight = generator_resblocks_7_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_71")]; + tensor var_3808 = const()[name = tensor("op_3808"), val = tensor([1, 128, 1])]; + tensor h_283 = reshape(shape = var_3808, x = h_281)[name = tensor("h_283")]; + tensor var_3810_split_sizes_0 = const()[name = tensor("op_3810_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3810_axis_0 = const()[name = tensor("op_3810_axis_0"), val = tensor(1)]; + tensor var_3810_0, tensor var_3810_1 = split(axis = var_3810_axis_0, split_sizes = var_3810_split_sizes_0, x = h_283)[name = tensor("op_3810")]; + tensor var_3812_promoted = const()[name = tensor("op_3812_promoted"), val = tensor(0x1p+0)]; + tensor var_3813 = add(x = var_3810_0, y = var_3812_promoted)[name = tensor("op_3813")]; + tensor var_3815 = mul(x = var_3813, y = var_3518)[name = tensor("op_3815")]; + tensor xt_181 = add(x = var_3815, y = var_3810_1)[name = tensor("xt_181")]; + tensor var_3818 = const()[name = tensor("op_3818"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54128000)))]; + tensor var_3819 = mul(x = xt_181, y = var_3818)[name = tensor("op_3819")]; + tensor cv_121 = cos(x = var_3819)[name = tensor("cv_121")]; + tensor var_3821 = const()[name = tensor("op_3821"), val = tensor(-0x1p-1)]; + tensor var_3822 = mul(x = cv_121, y = var_3821)[name = tensor("op_3822")]; + tensor var_3823 = const()[name = tensor("op_3823"), val = tensor(0x1p-1)]; + tensor var_3824 = add(x = var_3822, y = var_3823)[name = tensor("op_3824")]; + tensor var_3825 = const()[name = tensor("op_3825"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54128320)))]; + tensor var_3828 = mul(x = var_3824, y = var_3825)[name = tensor("op_3828")]; + tensor input_331 = add(x = xt_181, y = var_3828)[name = tensor("input_331")]; + tensor weight_163_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54128640))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54157376))), name = tensor("weight_163_palettized"), shape = tensor([64, 64, 7])]; + tensor input_333_pad_type_0 = const()[name = tensor("input_333_pad_type_0"), val = tensor("custom")]; + tensor input_333_pad_0 = const()[name = tensor("input_333_pad_0"), val = tensor([3, 3])]; + tensor input_333_strides_0 = const()[name = tensor("input_333_strides_0"), val = tensor([1])]; + tensor input_333_dilations_0 = const()[name = tensor("input_333_dilations_0"), val = tensor([1])]; + tensor input_333_groups_0 = const()[name = tensor("input_333_groups_0"), val = tensor(1)]; + tensor input_333 = conv(bias = generator_resblocks_7_convs1_0_bias, dilations = input_333_dilations_0, groups = input_333_groups_0, pad = input_333_pad_0, pad_type = input_333_pad_type_0, strides = input_333_strides_0, weight = weight_163_palettized, x = input_331)[name = tensor("input_333")]; + tensor h_285 = linear(bias = generator_resblocks_7_adain2_0_fc_bias, weight = generator_resblocks_7_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_72")]; + tensor var_3843 = const()[name = tensor("op_3843"), val = tensor([1, 128, 1])]; + tensor h_287 = reshape(shape = var_3843, x = h_285)[name = tensor("h_287")]; + tensor var_3845_split_sizes_0 = const()[name = tensor("op_3845_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3845_axis_0 = const()[name = tensor("op_3845_axis_0"), val = tensor(1)]; + tensor var_3845_0, tensor var_3845_1 = split(axis = var_3845_axis_0, split_sizes = var_3845_split_sizes_0, x = h_287)[name = tensor("op_3845")]; + tensor var_3847_promoted = const()[name = tensor("op_3847_promoted"), val = tensor(0x1p+0)]; + tensor var_3848 = add(x = var_3845_0, y = var_3847_promoted)[name = tensor("op_3848")]; + tensor var_3849 = instance_norm(epsilon = var_3726, x = input_333)[name = tensor("op_3849")]; + tensor var_3850 = mul(x = var_3848, y = var_3849)[name = tensor("op_3850")]; + tensor xt_183 = add(x = var_3850, y = var_3845_1)[name = tensor("xt_183")]; + tensor var_3853 = const()[name = tensor("op_3853"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54158464)))]; + tensor var_3854 = mul(x = xt_183, y = var_3853)[name = tensor("op_3854")]; + tensor cv_123 = cos(x = var_3854)[name = tensor("cv_123")]; + tensor var_3856 = const()[name = tensor("op_3856"), val = tensor(-0x1p-1)]; + tensor var_3857 = mul(x = cv_123, y = var_3856)[name = tensor("op_3857")]; + tensor var_3858 = const()[name = tensor("op_3858"), val = tensor(0x1p-1)]; + tensor var_3859 = add(x = var_3857, y = var_3858)[name = tensor("op_3859")]; + tensor var_3860 = const()[name = tensor("op_3860"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54158784)))]; + tensor var_3863 = mul(x = var_3859, y = var_3860)[name = tensor("op_3863")]; + tensor input_335 = add(x = xt_183, y = var_3863)[name = tensor("input_335")]; + tensor weight_165_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54159104))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54187840))), name = tensor("weight_165_palettized"), shape = tensor([64, 64, 7])]; + tensor xt_185_pad_type_0 = const()[name = tensor("xt_185_pad_type_0"), val = tensor("custom")]; + tensor xt_185_pad_0 = const()[name = tensor("xt_185_pad_0"), val = tensor([3, 3])]; + tensor xt_185_strides_0 = const()[name = tensor("xt_185_strides_0"), val = tensor([1])]; + tensor xt_185_dilations_0 = const()[name = tensor("xt_185_dilations_0"), val = tensor([1])]; + tensor xt_185_groups_0 = const()[name = tensor("xt_185_groups_0"), val = tensor(1)]; + tensor xt_185 = conv(bias = generator_resblocks_7_convs2_0_bias, dilations = xt_185_dilations_0, groups = xt_185_groups_0, pad = xt_185_pad_0, pad_type = xt_185_pad_type_0, strides = xt_185_strides_0, weight = weight_165_palettized, x = input_335)[name = tensor("xt_185")]; + tensor input_337 = add(x = xt_185, y = input_307)[name = tensor("input_337")]; + tensor h_289 = linear(bias = generator_resblocks_7_adain1_1_fc_bias, weight = generator_resblocks_7_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_73")]; + tensor var_3879 = const()[name = tensor("op_3879"), val = tensor([1, 128, 1])]; + tensor h_291 = reshape(shape = var_3879, x = h_289)[name = tensor("h_291")]; + tensor var_3881_split_sizes_0 = const()[name = tensor("op_3881_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3881_axis_0 = const()[name = tensor("op_3881_axis_0"), val = tensor(1)]; + tensor var_3881_0, tensor var_3881_1 = split(axis = var_3881_axis_0, split_sizes = var_3881_split_sizes_0, x = h_291)[name = tensor("op_3881")]; + tensor var_3883_promoted = const()[name = tensor("op_3883_promoted"), val = tensor(0x1p+0)]; + tensor var_3884 = add(x = var_3881_0, y = var_3883_promoted)[name = tensor("op_3884")]; + tensor var_3885 = instance_norm(epsilon = var_3726, x = input_337)[name = tensor("op_3885")]; + tensor var_3886 = mul(x = var_3884, y = var_3885)[name = tensor("op_3886")]; + tensor xt_187 = add(x = var_3886, y = var_3881_1)[name = tensor("xt_187")]; + tensor var_3889 = const()[name = tensor("op_3889"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54188928)))]; + tensor var_3890 = mul(x = xt_187, y = var_3889)[name = tensor("op_3890")]; + tensor cv_125 = cos(x = var_3890)[name = tensor("cv_125")]; + tensor var_3892 = const()[name = tensor("op_3892"), val = tensor(-0x1p-1)]; + tensor var_3893 = mul(x = cv_125, y = var_3892)[name = tensor("op_3893")]; + tensor var_3894 = const()[name = tensor("op_3894"), val = tensor(0x1p-1)]; + tensor var_3895 = add(x = var_3893, y = var_3894)[name = tensor("op_3895")]; + tensor var_3896 = const()[name = tensor("op_3896"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54189248)))]; + tensor var_3899 = mul(x = var_3895, y = var_3896)[name = tensor("op_3899")]; + tensor input_339 = add(x = xt_187, y = var_3899)[name = tensor("input_339")]; + tensor weight_167_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54189568))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54218304))), name = tensor("weight_167_palettized"), shape = tensor([64, 64, 7])]; + tensor input_341_pad_type_0 = const()[name = tensor("input_341_pad_type_0"), val = tensor("custom")]; + tensor input_341_pad_0 = const()[name = tensor("input_341_pad_0"), val = tensor([9, 9])]; + tensor input_341_dilations_0 = const()[name = tensor("input_341_dilations_0"), val = tensor([3])]; + tensor input_341_strides_0 = const()[name = tensor("input_341_strides_0"), val = tensor([1])]; + tensor input_341_groups_0 = const()[name = tensor("input_341_groups_0"), val = tensor(1)]; + tensor input_341 = conv(bias = generator_resblocks_7_convs1_1_bias, dilations = input_341_dilations_0, groups = input_341_groups_0, pad = input_341_pad_0, pad_type = input_341_pad_type_0, strides = input_341_strides_0, weight = weight_167_palettized, x = input_339)[name = tensor("input_341")]; + tensor h_293 = linear(bias = generator_resblocks_7_adain2_1_fc_bias, weight = generator_resblocks_7_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_74")]; + tensor var_3914 = const()[name = tensor("op_3914"), val = tensor([1, 128, 1])]; + tensor h_295 = reshape(shape = var_3914, x = h_293)[name = tensor("h_295")]; + tensor var_3916_split_sizes_0 = const()[name = tensor("op_3916_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3916_axis_0 = const()[name = tensor("op_3916_axis_0"), val = tensor(1)]; + tensor var_3916_0, tensor var_3916_1 = split(axis = var_3916_axis_0, split_sizes = var_3916_split_sizes_0, x = h_295)[name = tensor("op_3916")]; + tensor var_3918_promoted = const()[name = tensor("op_3918_promoted"), val = tensor(0x1p+0)]; + tensor var_3919 = add(x = var_3916_0, y = var_3918_promoted)[name = tensor("op_3919")]; + tensor var_3920 = instance_norm(epsilon = var_3726, x = input_341)[name = tensor("op_3920")]; + tensor var_3921 = mul(x = var_3919, y = var_3920)[name = tensor("op_3921")]; + tensor xt_189 = add(x = var_3921, y = var_3916_1)[name = tensor("xt_189")]; + tensor var_3924 = const()[name = tensor("op_3924"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54219392)))]; + tensor var_3925 = mul(x = xt_189, y = var_3924)[name = tensor("op_3925")]; + tensor cv_127 = cos(x = var_3925)[name = tensor("cv_127")]; + tensor var_3927 = const()[name = tensor("op_3927"), val = tensor(-0x1p-1)]; + tensor var_3928 = mul(x = cv_127, y = var_3927)[name = tensor("op_3928")]; + tensor var_3929 = const()[name = tensor("op_3929"), val = tensor(0x1p-1)]; + tensor var_3930 = add(x = var_3928, y = var_3929)[name = tensor("op_3930")]; + tensor var_3931 = const()[name = tensor("op_3931"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54219712)))]; + tensor var_3934 = mul(x = var_3930, y = var_3931)[name = tensor("op_3934")]; + tensor input_343 = add(x = xt_189, y = var_3934)[name = tensor("input_343")]; + tensor weight_169_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54220032))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54248768))), name = tensor("weight_169_palettized"), shape = tensor([64, 64, 7])]; + tensor xt_191_pad_type_0 = const()[name = tensor("xt_191_pad_type_0"), val = tensor("custom")]; + tensor xt_191_pad_0 = const()[name = tensor("xt_191_pad_0"), val = tensor([3, 3])]; + tensor xt_191_strides_0 = const()[name = tensor("xt_191_strides_0"), val = tensor([1])]; + tensor xt_191_dilations_0 = const()[name = tensor("xt_191_dilations_0"), val = tensor([1])]; + tensor xt_191_groups_0 = const()[name = tensor("xt_191_groups_0"), val = tensor(1)]; + tensor xt_191 = conv(bias = generator_resblocks_7_convs2_1_bias, dilations = xt_191_dilations_0, groups = xt_191_groups_0, pad = xt_191_pad_0, pad_type = xt_191_pad_type_0, strides = xt_191_strides_0, weight = weight_169_palettized, x = input_343)[name = tensor("xt_191")]; + tensor input_345 = add(x = xt_191, y = input_337)[name = tensor("input_345")]; + tensor h_297 = linear(bias = generator_resblocks_7_adain1_2_fc_bias, weight = generator_resblocks_7_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_75")]; + tensor var_3950 = const()[name = tensor("op_3950"), val = tensor([1, 128, 1])]; + tensor h_299 = reshape(shape = var_3950, x = h_297)[name = tensor("h_299")]; + tensor var_3952_split_sizes_0 = const()[name = tensor("op_3952_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3952_axis_0 = const()[name = tensor("op_3952_axis_0"), val = tensor(1)]; + tensor var_3952_0, tensor var_3952_1 = split(axis = var_3952_axis_0, split_sizes = var_3952_split_sizes_0, x = h_299)[name = tensor("op_3952")]; + tensor var_3954_promoted = const()[name = tensor("op_3954_promoted"), val = tensor(0x1p+0)]; + tensor var_3955 = add(x = var_3952_0, y = var_3954_promoted)[name = tensor("op_3955")]; + tensor var_3956 = instance_norm(epsilon = var_3726, x = input_345)[name = tensor("op_3956")]; + tensor var_3957 = mul(x = var_3955, y = var_3956)[name = tensor("op_3957")]; + tensor xt_193 = add(x = var_3957, y = var_3952_1)[name = tensor("xt_193")]; + tensor var_3960 = const()[name = tensor("op_3960"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54249856)))]; + tensor var_3961 = mul(x = xt_193, y = var_3960)[name = tensor("op_3961")]; + tensor cv_129 = cos(x = var_3961)[name = tensor("cv_129")]; + tensor var_3963 = const()[name = tensor("op_3963"), val = tensor(-0x1p-1)]; + tensor var_3964 = mul(x = cv_129, y = var_3963)[name = tensor("op_3964")]; + tensor var_3965 = const()[name = tensor("op_3965"), val = tensor(0x1p-1)]; + tensor var_3966 = add(x = var_3964, y = var_3965)[name = tensor("op_3966")]; + tensor var_3967 = const()[name = tensor("op_3967"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54250176)))]; + tensor var_3970 = mul(x = var_3966, y = var_3967)[name = tensor("op_3970")]; + tensor input_347 = add(x = xt_193, y = var_3970)[name = tensor("input_347")]; + tensor weight_171_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54250496))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54279232))), name = tensor("weight_171_palettized"), shape = tensor([64, 64, 7])]; + tensor input_349_pad_type_0 = const()[name = tensor("input_349_pad_type_0"), val = tensor("custom")]; + tensor input_349_pad_0 = const()[name = tensor("input_349_pad_0"), val = tensor([15, 15])]; + tensor input_349_dilations_0 = const()[name = tensor("input_349_dilations_0"), val = tensor([5])]; + tensor input_349_strides_0 = const()[name = tensor("input_349_strides_0"), val = tensor([1])]; + tensor input_349_groups_0 = const()[name = tensor("input_349_groups_0"), val = tensor(1)]; + tensor input_349 = conv(bias = generator_resblocks_7_convs1_2_bias, dilations = input_349_dilations_0, groups = input_349_groups_0, pad = input_349_pad_0, pad_type = input_349_pad_type_0, strides = input_349_strides_0, weight = weight_171_palettized, x = input_347)[name = tensor("input_349")]; + tensor h_301 = linear(bias = generator_resblocks_7_adain2_2_fc_bias, weight = generator_resblocks_7_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_76")]; + tensor var_3985 = const()[name = tensor("op_3985"), val = tensor([1, 128, 1])]; + tensor h_303 = reshape(shape = var_3985, x = h_301)[name = tensor("h_303")]; + tensor var_3987_split_sizes_0 = const()[name = tensor("op_3987_split_sizes_0"), val = tensor([64, 64])]; + tensor var_3987_axis_0 = const()[name = tensor("op_3987_axis_0"), val = tensor(1)]; + tensor var_3987_0, tensor var_3987_1 = split(axis = var_3987_axis_0, split_sizes = var_3987_split_sizes_0, x = h_303)[name = tensor("op_3987")]; + tensor var_3989_promoted = const()[name = tensor("op_3989_promoted"), val = tensor(0x1p+0)]; + tensor var_3990 = add(x = var_3987_0, y = var_3989_promoted)[name = tensor("op_3990")]; + tensor var_3991 = instance_norm(epsilon = var_3726, x = input_349)[name = tensor("op_3991")]; + tensor var_3992 = mul(x = var_3990, y = var_3991)[name = tensor("op_3992")]; + tensor xt_195 = add(x = var_3992, y = var_3987_1)[name = tensor("xt_195")]; + tensor var_3995 = const()[name = tensor("op_3995"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54280320)))]; + tensor var_3996 = mul(x = xt_195, y = var_3995)[name = tensor("op_3996")]; + tensor cv_131 = cos(x = var_3996)[name = tensor("cv_131")]; + tensor var_3998 = const()[name = tensor("op_3998"), val = tensor(-0x1p-1)]; + tensor var_3999 = mul(x = cv_131, y = var_3998)[name = tensor("op_3999")]; + tensor var_4000 = const()[name = tensor("op_4000"), val = tensor(0x1p-1)]; + tensor var_4001 = add(x = var_3999, y = var_4000)[name = tensor("op_4001")]; + tensor var_4002 = const()[name = tensor("op_4002"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54280640)))]; + tensor var_4005 = mul(x = var_4001, y = var_4002)[name = tensor("op_4005")]; + tensor input_351 = add(x = xt_195, y = var_4005)[name = tensor("input_351")]; + tensor weight_173_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54280960))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54309696))), name = tensor("weight_173_palettized"), shape = tensor([64, 64, 7])]; + tensor xt_197_pad_type_0 = const()[name = tensor("xt_197_pad_type_0"), val = tensor("custom")]; + tensor xt_197_pad_0 = const()[name = tensor("xt_197_pad_0"), val = tensor([3, 3])]; + tensor xt_197_strides_0 = const()[name = tensor("xt_197_strides_0"), val = tensor([1])]; + tensor xt_197_dilations_0 = const()[name = tensor("xt_197_dilations_0"), val = tensor([1])]; + tensor xt_197_groups_0 = const()[name = tensor("xt_197_groups_0"), val = tensor(1)]; + tensor xt_197 = conv(bias = generator_resblocks_7_convs2_2_bias, dilations = xt_197_dilations_0, groups = xt_197_groups_0, pad = xt_197_pad_0, pad_type = xt_197_pad_type_0, strides = xt_197_strides_0, weight = weight_173_palettized, x = input_351)[name = tensor("xt_197")]; + tensor var_4014 = add(x = xt_197, y = input_345)[name = tensor("op_4014")]; + tensor xs_15 = add(x = xs_13, y = var_4014)[name = tensor("xs_15")]; + tensor var_4024 = const()[name = tensor("op_4024"), val = tensor(0x1.4f8b58p-17)]; + tensor h_305 = linear(bias = generator_resblocks_8_adain1_0_fc_bias, weight = generator_resblocks_8_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_77")]; + tensor var_4106 = const()[name = tensor("op_4106"), val = tensor([1, 128, 1])]; + tensor h_307 = reshape(shape = var_4106, x = h_305)[name = tensor("h_307")]; + tensor var_4108_split_sizes_0 = const()[name = tensor("op_4108_split_sizes_0"), val = tensor([64, 64])]; + tensor var_4108_axis_0 = const()[name = tensor("op_4108_axis_0"), val = tensor(1)]; + tensor var_4108_0, tensor var_4108_1 = split(axis = var_4108_axis_0, split_sizes = var_4108_split_sizes_0, x = h_307)[name = tensor("op_4108")]; + tensor var_4110_promoted = const()[name = tensor("op_4110_promoted"), val = tensor(0x1p+0)]; + tensor var_4111 = add(x = var_4108_0, y = var_4110_promoted)[name = tensor("op_4111")]; + tensor var_4113 = mul(x = var_4111, y = var_3518)[name = tensor("op_4113")]; + tensor xt_199 = add(x = var_4113, y = var_4108_1)[name = tensor("xt_199")]; + tensor var_4116 = const()[name = tensor("op_4116"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54310784)))]; + tensor var_4117 = mul(x = xt_199, y = var_4116)[name = tensor("op_4117")]; + tensor cv_133 = cos(x = var_4117)[name = tensor("cv_133")]; + tensor var_4119 = const()[name = tensor("op_4119"), val = tensor(-0x1p-1)]; + tensor var_4120 = mul(x = cv_133, y = var_4119)[name = tensor("op_4120")]; + tensor var_4121 = const()[name = tensor("op_4121"), val = tensor(0x1p-1)]; + tensor var_4122 = add(x = var_4120, y = var_4121)[name = tensor("op_4122")]; + tensor var_4123 = const()[name = tensor("op_4123"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54311104)))]; + tensor var_4126 = mul(x = var_4122, y = var_4123)[name = tensor("op_4126")]; + tensor input_353 = add(x = xt_199, y = var_4126)[name = tensor("input_353")]; + tensor weight_175_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54311424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54356544))), name = tensor("weight_175_palettized"), shape = tensor([64, 64, 11])]; + tensor input_355_pad_type_0 = const()[name = tensor("input_355_pad_type_0"), val = tensor("custom")]; + tensor input_355_pad_0 = const()[name = tensor("input_355_pad_0"), val = tensor([5, 5])]; + tensor input_355_strides_0 = const()[name = tensor("input_355_strides_0"), val = tensor([1])]; + tensor input_355_dilations_0 = const()[name = tensor("input_355_dilations_0"), val = tensor([1])]; + tensor input_355_groups_0 = const()[name = tensor("input_355_groups_0"), val = tensor(1)]; + tensor input_355 = conv(bias = generator_resblocks_8_convs1_0_bias, dilations = input_355_dilations_0, groups = input_355_groups_0, pad = input_355_pad_0, pad_type = input_355_pad_type_0, strides = input_355_strides_0, weight = weight_175_palettized, x = input_353)[name = tensor("input_355")]; + tensor h_309 = linear(bias = generator_resblocks_8_adain2_0_fc_bias, weight = generator_resblocks_8_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_78")]; + tensor var_4141 = const()[name = tensor("op_4141"), val = tensor([1, 128, 1])]; + tensor h_311 = reshape(shape = var_4141, x = h_309)[name = tensor("h_311")]; + tensor var_4143_split_sizes_0 = const()[name = tensor("op_4143_split_sizes_0"), val = tensor([64, 64])]; + tensor var_4143_axis_0 = const()[name = tensor("op_4143_axis_0"), val = tensor(1)]; + tensor var_4143_0, tensor var_4143_1 = split(axis = var_4143_axis_0, split_sizes = var_4143_split_sizes_0, x = h_311)[name = tensor("op_4143")]; + tensor var_4145_promoted = const()[name = tensor("op_4145_promoted"), val = tensor(0x1p+0)]; + tensor var_4146 = add(x = var_4143_0, y = var_4145_promoted)[name = tensor("op_4146")]; + tensor var_4147 = instance_norm(epsilon = var_4024, x = input_355)[name = tensor("op_4147")]; + tensor var_4148 = mul(x = var_4146, y = var_4147)[name = tensor("op_4148")]; + tensor xt_201 = add(x = var_4148, y = var_4143_1)[name = tensor("xt_201")]; + tensor var_4151 = const()[name = tensor("op_4151"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54357632)))]; + tensor var_4152 = mul(x = xt_201, y = var_4151)[name = tensor("op_4152")]; + tensor cv_135 = cos(x = var_4152)[name = tensor("cv_135")]; + tensor var_4154 = const()[name = tensor("op_4154"), val = tensor(-0x1p-1)]; + tensor var_4155 = mul(x = cv_135, y = var_4154)[name = tensor("op_4155")]; + tensor var_4156 = const()[name = tensor("op_4156"), val = tensor(0x1p-1)]; + tensor var_4157 = add(x = var_4155, y = var_4156)[name = tensor("op_4157")]; + tensor var_4158 = const()[name = tensor("op_4158"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54357952)))]; + tensor var_4161 = mul(x = var_4157, y = var_4158)[name = tensor("op_4161")]; + tensor input_357 = add(x = xt_201, y = var_4161)[name = tensor("input_357")]; + tensor weight_177_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54358272))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54403392))), name = tensor("weight_177_palettized"), shape = tensor([64, 64, 11])]; + tensor xt_203_pad_type_0 = const()[name = tensor("xt_203_pad_type_0"), val = tensor("custom")]; + tensor xt_203_pad_0 = const()[name = tensor("xt_203_pad_0"), val = tensor([5, 5])]; + tensor xt_203_strides_0 = const()[name = tensor("xt_203_strides_0"), val = tensor([1])]; + tensor xt_203_dilations_0 = const()[name = tensor("xt_203_dilations_0"), val = tensor([1])]; + tensor xt_203_groups_0 = const()[name = tensor("xt_203_groups_0"), val = tensor(1)]; + tensor xt_203 = conv(bias = generator_resblocks_8_convs2_0_bias, dilations = xt_203_dilations_0, groups = xt_203_groups_0, pad = xt_203_pad_0, pad_type = xt_203_pad_type_0, strides = xt_203_strides_0, weight = weight_177_palettized, x = input_357)[name = tensor("xt_203")]; + tensor input_359 = add(x = xt_203, y = input_307)[name = tensor("input_359")]; + tensor h_313 = linear(bias = generator_resblocks_8_adain1_1_fc_bias, weight = generator_resblocks_8_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_79")]; + tensor var_4177 = const()[name = tensor("op_4177"), val = tensor([1, 128, 1])]; + tensor h_315 = reshape(shape = var_4177, x = h_313)[name = tensor("h_315")]; + tensor var_4179_split_sizes_0 = const()[name = tensor("op_4179_split_sizes_0"), val = tensor([64, 64])]; + tensor var_4179_axis_0 = const()[name = tensor("op_4179_axis_0"), val = tensor(1)]; + tensor var_4179_0, tensor var_4179_1 = split(axis = var_4179_axis_0, split_sizes = var_4179_split_sizes_0, x = h_315)[name = tensor("op_4179")]; + tensor var_4181_promoted = const()[name = tensor("op_4181_promoted"), val = tensor(0x1p+0)]; + tensor var_4182 = add(x = var_4179_0, y = var_4181_promoted)[name = tensor("op_4182")]; + tensor var_4183 = instance_norm(epsilon = var_4024, x = input_359)[name = tensor("op_4183")]; + tensor var_4184 = mul(x = var_4182, y = var_4183)[name = tensor("op_4184")]; + tensor xt_205 = add(x = var_4184, y = var_4179_1)[name = tensor("xt_205")]; + tensor var_4187 = const()[name = tensor("op_4187"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54404480)))]; + tensor var_4188 = mul(x = xt_205, y = var_4187)[name = tensor("op_4188")]; + tensor cv_137 = cos(x = var_4188)[name = tensor("cv_137")]; + tensor var_4190 = const()[name = tensor("op_4190"), val = tensor(-0x1p-1)]; + tensor var_4191 = mul(x = cv_137, y = var_4190)[name = tensor("op_4191")]; + tensor var_4192 = const()[name = tensor("op_4192"), val = tensor(0x1p-1)]; + tensor var_4193 = add(x = var_4191, y = var_4192)[name = tensor("op_4193")]; + tensor var_4194 = const()[name = tensor("op_4194"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54404800)))]; + tensor var_4197 = mul(x = var_4193, y = var_4194)[name = tensor("op_4197")]; + tensor input_361 = add(x = xt_205, y = var_4197)[name = tensor("input_361")]; + tensor weight_179_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54405120))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54450240))), name = tensor("weight_179_palettized"), shape = tensor([64, 64, 11])]; + tensor input_363_pad_type_0 = const()[name = tensor("input_363_pad_type_0"), val = tensor("custom")]; + tensor input_363_pad_0 = const()[name = tensor("input_363_pad_0"), val = tensor([15, 15])]; + tensor input_363_dilations_0 = const()[name = tensor("input_363_dilations_0"), val = tensor([3])]; + tensor input_363_strides_0 = const()[name = tensor("input_363_strides_0"), val = tensor([1])]; + tensor input_363_groups_0 = const()[name = tensor("input_363_groups_0"), val = tensor(1)]; + tensor input_363 = conv(bias = generator_resblocks_8_convs1_1_bias, dilations = input_363_dilations_0, groups = input_363_groups_0, pad = input_363_pad_0, pad_type = input_363_pad_type_0, strides = input_363_strides_0, weight = weight_179_palettized, x = input_361)[name = tensor("input_363")]; + tensor h_317 = linear(bias = generator_resblocks_8_adain2_1_fc_bias, weight = generator_resblocks_8_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_80")]; + tensor var_4212 = const()[name = tensor("op_4212"), val = tensor([1, 128, 1])]; + tensor h_319 = reshape(shape = var_4212, x = h_317)[name = tensor("h_319")]; + tensor var_4214_split_sizes_0 = const()[name = tensor("op_4214_split_sizes_0"), val = tensor([64, 64])]; + tensor var_4214_axis_0 = const()[name = tensor("op_4214_axis_0"), val = tensor(1)]; + tensor var_4214_0, tensor var_4214_1 = split(axis = var_4214_axis_0, split_sizes = var_4214_split_sizes_0, x = h_319)[name = tensor("op_4214")]; + tensor var_4216_promoted = const()[name = tensor("op_4216_promoted"), val = tensor(0x1p+0)]; + tensor var_4217 = add(x = var_4214_0, y = var_4216_promoted)[name = tensor("op_4217")]; + tensor var_4218 = instance_norm(epsilon = var_4024, x = input_363)[name = tensor("op_4218")]; + tensor var_4219 = mul(x = var_4217, y = var_4218)[name = tensor("op_4219")]; + tensor xt_207 = add(x = var_4219, y = var_4214_1)[name = tensor("xt_207")]; + tensor var_4222 = const()[name = tensor("op_4222"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54451328)))]; + tensor var_4223 = mul(x = xt_207, y = var_4222)[name = tensor("op_4223")]; + tensor cv_139 = cos(x = var_4223)[name = tensor("cv_139")]; + tensor var_4225 = const()[name = tensor("op_4225"), val = tensor(-0x1p-1)]; + tensor var_4226 = mul(x = cv_139, y = var_4225)[name = tensor("op_4226")]; + tensor var_4227 = const()[name = tensor("op_4227"), val = tensor(0x1p-1)]; + tensor var_4228 = add(x = var_4226, y = var_4227)[name = tensor("op_4228")]; + tensor var_4229 = const()[name = tensor("op_4229"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54451648)))]; + tensor var_4232 = mul(x = var_4228, y = var_4229)[name = tensor("op_4232")]; + tensor input_365 = add(x = xt_207, y = var_4232)[name = tensor("input_365")]; + tensor weight_181_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54451968))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54497088))), name = tensor("weight_181_palettized"), shape = tensor([64, 64, 11])]; + tensor xt_209_pad_type_0 = const()[name = tensor("xt_209_pad_type_0"), val = tensor("custom")]; + tensor xt_209_pad_0 = const()[name = tensor("xt_209_pad_0"), val = tensor([5, 5])]; + tensor xt_209_strides_0 = const()[name = tensor("xt_209_strides_0"), val = tensor([1])]; + tensor xt_209_dilations_0 = const()[name = tensor("xt_209_dilations_0"), val = tensor([1])]; + tensor xt_209_groups_0 = const()[name = tensor("xt_209_groups_0"), val = tensor(1)]; + tensor xt_209 = conv(bias = generator_resblocks_8_convs2_1_bias, dilations = xt_209_dilations_0, groups = xt_209_groups_0, pad = xt_209_pad_0, pad_type = xt_209_pad_type_0, strides = xt_209_strides_0, weight = weight_181_palettized, x = input_365)[name = tensor("xt_209")]; + tensor input_367 = add(x = xt_209, y = input_359)[name = tensor("input_367")]; + tensor h_321 = linear(bias = generator_resblocks_8_adain1_2_fc_bias, weight = generator_resblocks_8_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_81")]; + tensor var_4248 = const()[name = tensor("op_4248"), val = tensor([1, 128, 1])]; + tensor h_323 = reshape(shape = var_4248, x = h_321)[name = tensor("h_323")]; + tensor var_4250_split_sizes_0 = const()[name = tensor("op_4250_split_sizes_0"), val = tensor([64, 64])]; + tensor var_4250_axis_0 = const()[name = tensor("op_4250_axis_0"), val = tensor(1)]; + tensor var_4250_0, tensor var_4250_1 = split(axis = var_4250_axis_0, split_sizes = var_4250_split_sizes_0, x = h_323)[name = tensor("op_4250")]; + tensor var_4252_promoted = const()[name = tensor("op_4252_promoted"), val = tensor(0x1p+0)]; + tensor var_4253 = add(x = var_4250_0, y = var_4252_promoted)[name = tensor("op_4253")]; + tensor var_4254 = instance_norm(epsilon = var_4024, x = input_367)[name = tensor("op_4254")]; + tensor var_4255 = mul(x = var_4253, y = var_4254)[name = tensor("op_4255")]; + tensor xt_211 = add(x = var_4255, y = var_4250_1)[name = tensor("xt_211")]; + tensor var_4258 = const()[name = tensor("op_4258"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54498176)))]; + tensor var_4259 = mul(x = xt_211, y = var_4258)[name = tensor("op_4259")]; + tensor cv_141 = cos(x = var_4259)[name = tensor("cv_141")]; + tensor var_4261 = const()[name = tensor("op_4261"), val = tensor(-0x1p-1)]; + tensor var_4262 = mul(x = cv_141, y = var_4261)[name = tensor("op_4262")]; + tensor var_4263 = const()[name = tensor("op_4263"), val = tensor(0x1p-1)]; + tensor var_4264 = add(x = var_4262, y = var_4263)[name = tensor("op_4264")]; + tensor var_4265 = const()[name = tensor("op_4265"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54498496)))]; + tensor var_4268 = mul(x = var_4264, y = var_4265)[name = tensor("op_4268")]; + tensor input_369 = add(x = xt_211, y = var_4268)[name = tensor("input_369")]; + tensor weight_183_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54498816))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54543936))), name = tensor("weight_183_palettized"), shape = tensor([64, 64, 11])]; + tensor input_371_pad_type_0 = const()[name = tensor("input_371_pad_type_0"), val = tensor("custom")]; + tensor input_371_pad_0 = const()[name = tensor("input_371_pad_0"), val = tensor([25, 25])]; + tensor input_371_dilations_0 = const()[name = tensor("input_371_dilations_0"), val = tensor([5])]; + tensor input_371_strides_0 = const()[name = tensor("input_371_strides_0"), val = tensor([1])]; + tensor input_371_groups_0 = const()[name = tensor("input_371_groups_0"), val = tensor(1)]; + tensor input_371 = conv(bias = generator_resblocks_8_convs1_2_bias, dilations = input_371_dilations_0, groups = input_371_groups_0, pad = input_371_pad_0, pad_type = input_371_pad_type_0, strides = input_371_strides_0, weight = weight_183_palettized, x = input_369)[name = tensor("input_371")]; + tensor h_325 = linear(bias = generator_resblocks_8_adain2_2_fc_bias, weight = generator_resblocks_8_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_82")]; + tensor var_4283 = const()[name = tensor("op_4283"), val = tensor([1, 128, 1])]; + tensor h_327 = reshape(shape = var_4283, x = h_325)[name = tensor("h_327")]; + tensor var_4285_split_sizes_0 = const()[name = tensor("op_4285_split_sizes_0"), val = tensor([64, 64])]; + tensor var_4285_axis_0 = const()[name = tensor("op_4285_axis_0"), val = tensor(1)]; + tensor var_4285_0, tensor var_4285_1 = split(axis = var_4285_axis_0, split_sizes = var_4285_split_sizes_0, x = h_327)[name = tensor("op_4285")]; + tensor var_4287_promoted = const()[name = tensor("op_4287_promoted"), val = tensor(0x1p+0)]; + tensor var_4288 = add(x = var_4285_0, y = var_4287_promoted)[name = tensor("op_4288")]; + tensor var_4289 = instance_norm(epsilon = var_4024, x = input_371)[name = tensor("op_4289")]; + tensor var_4290 = mul(x = var_4288, y = var_4289)[name = tensor("op_4290")]; + tensor xt_213 = add(x = var_4290, y = var_4285_1)[name = tensor("xt_213")]; + tensor var_4293 = const()[name = tensor("op_4293"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54545024)))]; + tensor var_4294 = mul(x = xt_213, y = var_4293)[name = tensor("op_4294")]; + tensor cv_143 = cos(x = var_4294)[name = tensor("cv_143")]; + tensor var_4296 = const()[name = tensor("op_4296"), val = tensor(-0x1p-1)]; + tensor var_4297 = mul(x = cv_143, y = var_4296)[name = tensor("op_4297")]; + tensor var_4298 = const()[name = tensor("op_4298"), val = tensor(0x1p-1)]; + tensor var_4299 = add(x = var_4297, y = var_4298)[name = tensor("op_4299")]; + tensor var_4300 = const()[name = tensor("op_4300"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54545344)))]; + tensor var_4303 = mul(x = var_4299, y = var_4300)[name = tensor("op_4303")]; + tensor input_373 = add(x = xt_213, y = var_4303)[name = tensor("input_373")]; + tensor weight_185_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54545664))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54590784))), name = tensor("weight_185_palettized"), shape = tensor([64, 64, 11])]; + tensor xt_215_pad_type_0 = const()[name = tensor("xt_215_pad_type_0"), val = tensor("custom")]; + tensor xt_215_pad_0 = const()[name = tensor("xt_215_pad_0"), val = tensor([5, 5])]; + tensor xt_215_strides_0 = const()[name = tensor("xt_215_strides_0"), val = tensor([1])]; + tensor xt_215_dilations_0 = const()[name = tensor("xt_215_dilations_0"), val = tensor([1])]; + tensor xt_215_groups_0 = const()[name = tensor("xt_215_groups_0"), val = tensor(1)]; + tensor xt_215 = conv(bias = generator_resblocks_8_convs2_2_bias, dilations = xt_215_dilations_0, groups = xt_215_groups_0, pad = xt_215_pad_0, pad_type = xt_215_pad_type_0, strides = xt_215_strides_0, weight = weight_185_palettized, x = input_373)[name = tensor("xt_215")]; + tensor var_4312 = add(x = xt_215, y = input_367)[name = tensor("op_4312")]; + tensor xs_17 = add(x = xs_15, y = var_4312)[name = tensor("xs_17")]; + tensor _inversed_z_7_y_0 = const()[name = tensor("_inversed_z_7_y_0"), val = tensor(0x1.555556p-2)]; + tensor _inversed_z_7 = mul(x = xs_17, y = _inversed_z_7_y_0)[name = tensor("_inversed_z_7")]; + tensor var_4317 = mul(x = _inversed_z_7, y = generator_alphas_3)[name = tensor("op_4317")]; + tensor sv_7 = sin(x = var_4317)[name = tensor("sv_7")]; + tensor var_4319 = mul(x = sv_7, y = sv_7)[name = tensor("op_4319")]; + tensor _inversed_4320_y_0 = const()[name = tensor("_inversed_4320_y_0"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54591872)))]; + tensor _inversed_4320 = mul(x = var_4319, y = _inversed_4320_y_0)[name = tensor("_inversed_4320")]; + tensor input_399 = add(x = _inversed_z_7, y = _inversed_4320)[name = tensor("input_399")]; + tensor input_375_pad_type_0 = const()[name = tensor("input_375_pad_type_0"), val = tensor("valid")]; + tensor input_375_strides_0 = const()[name = tensor("input_375_strides_0"), val = tensor([1])]; + tensor input_375_pad_0 = const()[name = tensor("input_375_pad_0"), val = tensor([0, 0])]; + tensor input_375_dilations_0 = const()[name = tensor("input_375_dilations_0"), val = tensor([1])]; + tensor input_375_groups_0 = const()[name = tensor("input_375_groups_0"), val = tensor(1)]; + tensor input_375 = conv(bias = generator_noise_convs_3_bias, dilations = input_375_dilations_0, groups = input_375_groups_0, pad = input_375_pad_0, pad_type = input_375_pad_type_0, strides = input_375_strides_0, weight = generator_noise_convs_3_weight, x = input_91)[name = tensor("input_375")]; + tensor var_4341 = const()[name = tensor("op_4341"), val = tensor(0x1.4f8b58p-17)]; + tensor h_329 = linear(bias = generator_noise_res_3_adain1_0_fc_bias, weight = generator_noise_res_3_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_83")]; + tensor var_4423 = const()[name = tensor("op_4423"), val = tensor([1, 64, 1])]; + tensor h_331 = reshape(shape = var_4423, x = h_329)[name = tensor("h_331")]; + tensor var_4425_split_sizes_0 = const()[name = tensor("op_4425_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4425_axis_0 = const()[name = tensor("op_4425_axis_0"), val = tensor(1)]; + tensor var_4425_0, tensor var_4425_1 = split(axis = var_4425_axis_0, split_sizes = var_4425_split_sizes_0, x = h_331)[name = tensor("op_4425")]; + tensor var_4427_promoted = const()[name = tensor("op_4427_promoted"), val = tensor(0x1p+0)]; + tensor var_4428 = add(x = var_4425_0, y = var_4427_promoted)[name = tensor("op_4428")]; + tensor var_4429 = instance_norm(epsilon = var_4341, x = input_375)[name = tensor("op_4429")]; + tensor var_4430 = mul(x = var_4428, y = var_4429)[name = tensor("op_4430")]; + tensor xt_217 = add(x = var_4430, y = var_4425_1)[name = tensor("xt_217")]; + tensor var_4433 = const()[name = tensor("op_4433"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54592192)))]; + tensor var_4434 = mul(x = xt_217, y = var_4433)[name = tensor("op_4434")]; + tensor cv_145 = cos(x = var_4434)[name = tensor("cv_145")]; + tensor var_4436 = const()[name = tensor("op_4436"), val = tensor(-0x1p-1)]; + tensor var_4437 = mul(x = cv_145, y = var_4436)[name = tensor("op_4437")]; + tensor var_4438 = const()[name = tensor("op_4438"), val = tensor(0x1p-1)]; + tensor var_4439 = add(x = var_4437, y = var_4438)[name = tensor("op_4439")]; + tensor var_4440 = const()[name = tensor("op_4440"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54592384)))]; + tensor var_4443 = mul(x = var_4439, y = var_4440)[name = tensor("op_4443")]; + tensor input_377 = add(x = xt_217, y = var_4443)[name = tensor("input_377")]; + tensor weight_189_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54592576))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54603904))), name = tensor("weight_189_palettized"), shape = tensor([32, 32, 11])]; + tensor input_379_pad_type_0 = const()[name = tensor("input_379_pad_type_0"), val = tensor("custom")]; + tensor input_379_pad_0 = const()[name = tensor("input_379_pad_0"), val = tensor([5, 5])]; + tensor input_379_strides_0 = const()[name = tensor("input_379_strides_0"), val = tensor([1])]; + tensor input_379_dilations_0 = const()[name = tensor("input_379_dilations_0"), val = tensor([1])]; + tensor input_379_groups_0 = const()[name = tensor("input_379_groups_0"), val = tensor(1)]; + tensor input_379 = conv(bias = generator_noise_res_3_convs1_0_bias, dilations = input_379_dilations_0, groups = input_379_groups_0, pad = input_379_pad_0, pad_type = input_379_pad_type_0, strides = input_379_strides_0, weight = weight_189_palettized, x = input_377)[name = tensor("input_379")]; + tensor h_333 = linear(bias = generator_noise_res_3_adain2_0_fc_bias, weight = generator_noise_res_3_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_84")]; + tensor var_4458 = const()[name = tensor("op_4458"), val = tensor([1, 64, 1])]; + tensor h_335 = reshape(shape = var_4458, x = h_333)[name = tensor("h_335")]; + tensor var_4460_split_sizes_0 = const()[name = tensor("op_4460_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4460_axis_0 = const()[name = tensor("op_4460_axis_0"), val = tensor(1)]; + tensor var_4460_0, tensor var_4460_1 = split(axis = var_4460_axis_0, split_sizes = var_4460_split_sizes_0, x = h_335)[name = tensor("op_4460")]; + tensor var_4462_promoted = const()[name = tensor("op_4462_promoted"), val = tensor(0x1p+0)]; + tensor var_4463 = add(x = var_4460_0, y = var_4462_promoted)[name = tensor("op_4463")]; + tensor var_4464 = instance_norm(epsilon = var_4341, x = input_379)[name = tensor("op_4464")]; + tensor var_4465 = mul(x = var_4463, y = var_4464)[name = tensor("op_4465")]; + tensor xt_219 = add(x = var_4465, y = var_4460_1)[name = tensor("xt_219")]; + tensor var_4468 = const()[name = tensor("op_4468"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54604992)))]; + tensor var_4469 = mul(x = xt_219, y = var_4468)[name = tensor("op_4469")]; + tensor cv_147 = cos(x = var_4469)[name = tensor("cv_147")]; + tensor var_4471 = const()[name = tensor("op_4471"), val = tensor(-0x1p-1)]; + tensor var_4472 = mul(x = cv_147, y = var_4471)[name = tensor("op_4472")]; + tensor var_4473 = const()[name = tensor("op_4473"), val = tensor(0x1p-1)]; + tensor var_4474 = add(x = var_4472, y = var_4473)[name = tensor("op_4474")]; + tensor var_4475 = const()[name = tensor("op_4475"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54605184)))]; + tensor var_4478 = mul(x = var_4474, y = var_4475)[name = tensor("op_4478")]; + tensor input_381 = add(x = xt_219, y = var_4478)[name = tensor("input_381")]; + tensor weight_191_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54605376))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54616704))), name = tensor("weight_191_palettized"), shape = tensor([32, 32, 11])]; + tensor xt_221_pad_type_0 = const()[name = tensor("xt_221_pad_type_0"), val = tensor("custom")]; + tensor xt_221_pad_0 = const()[name = tensor("xt_221_pad_0"), val = tensor([5, 5])]; + tensor xt_221_strides_0 = const()[name = tensor("xt_221_strides_0"), val = tensor([1])]; + tensor xt_221_dilations_0 = const()[name = tensor("xt_221_dilations_0"), val = tensor([1])]; + tensor xt_221_groups_0 = const()[name = tensor("xt_221_groups_0"), val = tensor(1)]; + tensor xt_221 = conv(bias = generator_noise_res_3_convs2_0_bias, dilations = xt_221_dilations_0, groups = xt_221_groups_0, pad = xt_221_pad_0, pad_type = xt_221_pad_type_0, strides = xt_221_strides_0, weight = weight_191_palettized, x = input_381)[name = tensor("xt_221")]; + tensor input_383 = add(x = xt_221, y = input_375)[name = tensor("input_383")]; + tensor h_337 = linear(bias = generator_noise_res_3_adain1_1_fc_bias, weight = generator_noise_res_3_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_85")]; + tensor var_4494 = const()[name = tensor("op_4494"), val = tensor([1, 64, 1])]; + tensor h_339 = reshape(shape = var_4494, x = h_337)[name = tensor("h_339")]; + tensor var_4496_split_sizes_0 = const()[name = tensor("op_4496_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4496_axis_0 = const()[name = tensor("op_4496_axis_0"), val = tensor(1)]; + tensor var_4496_0, tensor var_4496_1 = split(axis = var_4496_axis_0, split_sizes = var_4496_split_sizes_0, x = h_339)[name = tensor("op_4496")]; + tensor var_4498_promoted = const()[name = tensor("op_4498_promoted"), val = tensor(0x1p+0)]; + tensor var_4499 = add(x = var_4496_0, y = var_4498_promoted)[name = tensor("op_4499")]; + tensor var_4500 = instance_norm(epsilon = var_4341, x = input_383)[name = tensor("op_4500")]; + tensor var_4501 = mul(x = var_4499, y = var_4500)[name = tensor("op_4501")]; + tensor xt_223 = add(x = var_4501, y = var_4496_1)[name = tensor("xt_223")]; + tensor var_4504 = const()[name = tensor("op_4504"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54617792)))]; + tensor var_4505 = mul(x = xt_223, y = var_4504)[name = tensor("op_4505")]; + tensor cv_149 = cos(x = var_4505)[name = tensor("cv_149")]; + tensor var_4507 = const()[name = tensor("op_4507"), val = tensor(-0x1p-1)]; + tensor var_4508 = mul(x = cv_149, y = var_4507)[name = tensor("op_4508")]; + tensor var_4509 = const()[name = tensor("op_4509"), val = tensor(0x1p-1)]; + tensor var_4510 = add(x = var_4508, y = var_4509)[name = tensor("op_4510")]; + tensor var_4511 = const()[name = tensor("op_4511"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54617984)))]; + tensor var_4514 = mul(x = var_4510, y = var_4511)[name = tensor("op_4514")]; + tensor input_385 = add(x = xt_223, y = var_4514)[name = tensor("input_385")]; + tensor weight_193_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54618176))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54629504))), name = tensor("weight_193_palettized"), shape = tensor([32, 32, 11])]; + tensor input_387_pad_type_0 = const()[name = tensor("input_387_pad_type_0"), val = tensor("custom")]; + tensor input_387_pad_0 = const()[name = tensor("input_387_pad_0"), val = tensor([15, 15])]; + tensor input_387_dilations_0 = const()[name = tensor("input_387_dilations_0"), val = tensor([3])]; + tensor input_387_strides_0 = const()[name = tensor("input_387_strides_0"), val = tensor([1])]; + tensor input_387_groups_0 = const()[name = tensor("input_387_groups_0"), val = tensor(1)]; + tensor input_387 = conv(bias = generator_noise_res_3_convs1_1_bias, dilations = input_387_dilations_0, groups = input_387_groups_0, pad = input_387_pad_0, pad_type = input_387_pad_type_0, strides = input_387_strides_0, weight = weight_193_palettized, x = input_385)[name = tensor("input_387")]; + tensor h_341 = linear(bias = generator_noise_res_3_adain2_1_fc_bias, weight = generator_noise_res_3_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_86")]; + tensor var_4529 = const()[name = tensor("op_4529"), val = tensor([1, 64, 1])]; + tensor h_343 = reshape(shape = var_4529, x = h_341)[name = tensor("h_343")]; + tensor var_4531_split_sizes_0 = const()[name = tensor("op_4531_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4531_axis_0 = const()[name = tensor("op_4531_axis_0"), val = tensor(1)]; + tensor var_4531_0, tensor var_4531_1 = split(axis = var_4531_axis_0, split_sizes = var_4531_split_sizes_0, x = h_343)[name = tensor("op_4531")]; + tensor var_4533_promoted = const()[name = tensor("op_4533_promoted"), val = tensor(0x1p+0)]; + tensor var_4534 = add(x = var_4531_0, y = var_4533_promoted)[name = tensor("op_4534")]; + tensor var_4535 = instance_norm(epsilon = var_4341, x = input_387)[name = tensor("op_4535")]; + tensor var_4536 = mul(x = var_4534, y = var_4535)[name = tensor("op_4536")]; + tensor xt_225 = add(x = var_4536, y = var_4531_1)[name = tensor("xt_225")]; + tensor var_4539 = const()[name = tensor("op_4539"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54630592)))]; + tensor var_4540 = mul(x = xt_225, y = var_4539)[name = tensor("op_4540")]; + tensor cv_151 = cos(x = var_4540)[name = tensor("cv_151")]; + tensor var_4542 = const()[name = tensor("op_4542"), val = tensor(-0x1p-1)]; + tensor var_4543 = mul(x = cv_151, y = var_4542)[name = tensor("op_4543")]; + tensor var_4544 = const()[name = tensor("op_4544"), val = tensor(0x1p-1)]; + tensor var_4545 = add(x = var_4543, y = var_4544)[name = tensor("op_4545")]; + tensor var_4546 = const()[name = tensor("op_4546"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54630784)))]; + tensor var_4549 = mul(x = var_4545, y = var_4546)[name = tensor("op_4549")]; + tensor input_389 = add(x = xt_225, y = var_4549)[name = tensor("input_389")]; + tensor weight_195_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54630976))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54642304))), name = tensor("weight_195_palettized"), shape = tensor([32, 32, 11])]; + tensor xt_227_pad_type_0 = const()[name = tensor("xt_227_pad_type_0"), val = tensor("custom")]; + tensor xt_227_pad_0 = const()[name = tensor("xt_227_pad_0"), val = tensor([5, 5])]; + tensor xt_227_strides_0 = const()[name = tensor("xt_227_strides_0"), val = tensor([1])]; + tensor xt_227_dilations_0 = const()[name = tensor("xt_227_dilations_0"), val = tensor([1])]; + tensor xt_227_groups_0 = const()[name = tensor("xt_227_groups_0"), val = tensor(1)]; + tensor xt_227 = conv(bias = generator_noise_res_3_convs2_1_bias, dilations = xt_227_dilations_0, groups = xt_227_groups_0, pad = xt_227_pad_0, pad_type = xt_227_pad_type_0, strides = xt_227_strides_0, weight = weight_195_palettized, x = input_389)[name = tensor("xt_227")]; + tensor input_391 = add(x = xt_227, y = input_383)[name = tensor("input_391")]; + tensor h_345 = linear(bias = generator_noise_res_3_adain1_2_fc_bias, weight = generator_noise_res_3_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_87")]; + tensor var_4565 = const()[name = tensor("op_4565"), val = tensor([1, 64, 1])]; + tensor h_347 = reshape(shape = var_4565, x = h_345)[name = tensor("h_347")]; + tensor var_4567_split_sizes_0 = const()[name = tensor("op_4567_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4567_axis_0 = const()[name = tensor("op_4567_axis_0"), val = tensor(1)]; + tensor var_4567_0, tensor var_4567_1 = split(axis = var_4567_axis_0, split_sizes = var_4567_split_sizes_0, x = h_347)[name = tensor("op_4567")]; + tensor var_4569_promoted = const()[name = tensor("op_4569_promoted"), val = tensor(0x1p+0)]; + tensor var_4570 = add(x = var_4567_0, y = var_4569_promoted)[name = tensor("op_4570")]; + tensor var_4571 = instance_norm(epsilon = var_4341, x = input_391)[name = tensor("op_4571")]; + tensor var_4572 = mul(x = var_4570, y = var_4571)[name = tensor("op_4572")]; + tensor xt_229 = add(x = var_4572, y = var_4567_1)[name = tensor("xt_229")]; + tensor var_4575 = const()[name = tensor("op_4575"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54643392)))]; + tensor var_4576 = mul(x = xt_229, y = var_4575)[name = tensor("op_4576")]; + tensor cv_153 = cos(x = var_4576)[name = tensor("cv_153")]; + tensor var_4578 = const()[name = tensor("op_4578"), val = tensor(-0x1p-1)]; + tensor var_4579 = mul(x = cv_153, y = var_4578)[name = tensor("op_4579")]; + tensor var_4580 = const()[name = tensor("op_4580"), val = tensor(0x1p-1)]; + tensor var_4581 = add(x = var_4579, y = var_4580)[name = tensor("op_4581")]; + tensor var_4582 = const()[name = tensor("op_4582"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54643584)))]; + tensor var_4585 = mul(x = var_4581, y = var_4582)[name = tensor("op_4585")]; + tensor input_393 = add(x = xt_229, y = var_4585)[name = tensor("input_393")]; + tensor weight_197_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54643776))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54655104))), name = tensor("weight_197_palettized"), shape = tensor([32, 32, 11])]; + tensor input_395_pad_type_0 = const()[name = tensor("input_395_pad_type_0"), val = tensor("custom")]; + tensor input_395_pad_0 = const()[name = tensor("input_395_pad_0"), val = tensor([25, 25])]; + tensor input_395_dilations_0 = const()[name = tensor("input_395_dilations_0"), val = tensor([5])]; + tensor input_395_strides_0 = const()[name = tensor("input_395_strides_0"), val = tensor([1])]; + tensor input_395_groups_0 = const()[name = tensor("input_395_groups_0"), val = tensor(1)]; + tensor input_395 = conv(bias = generator_noise_res_3_convs1_2_bias, dilations = input_395_dilations_0, groups = input_395_groups_0, pad = input_395_pad_0, pad_type = input_395_pad_type_0, strides = input_395_strides_0, weight = weight_197_palettized, x = input_393)[name = tensor("input_395")]; + tensor h_349 = linear(bias = generator_noise_res_3_adain2_2_fc_bias, weight = generator_noise_res_3_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_88")]; + tensor var_4600 = const()[name = tensor("op_4600"), val = tensor([1, 64, 1])]; + tensor h_351 = reshape(shape = var_4600, x = h_349)[name = tensor("h_351")]; + tensor var_4602_split_sizes_0 = const()[name = tensor("op_4602_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4602_axis_0 = const()[name = tensor("op_4602_axis_0"), val = tensor(1)]; + tensor var_4602_0, tensor var_4602_1 = split(axis = var_4602_axis_0, split_sizes = var_4602_split_sizes_0, x = h_351)[name = tensor("op_4602")]; + tensor var_4604_promoted = const()[name = tensor("op_4604_promoted"), val = tensor(0x1p+0)]; + tensor var_4605 = add(x = var_4602_0, y = var_4604_promoted)[name = tensor("op_4605")]; + tensor var_4606 = instance_norm(epsilon = var_4341, x = input_395)[name = tensor("op_4606")]; + tensor var_4607 = mul(x = var_4605, y = var_4606)[name = tensor("op_4607")]; + tensor xt_231 = add(x = var_4607, y = var_4602_1)[name = tensor("xt_231")]; + tensor var_4610 = const()[name = tensor("op_4610"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54656192)))]; + tensor var_4611 = mul(x = xt_231, y = var_4610)[name = tensor("op_4611")]; + tensor cv_155 = cos(x = var_4611)[name = tensor("cv_155")]; + tensor var_4613 = const()[name = tensor("op_4613"), val = tensor(-0x1p-1)]; + tensor var_4614 = mul(x = cv_155, y = var_4613)[name = tensor("op_4614")]; + tensor var_4615 = const()[name = tensor("op_4615"), val = tensor(0x1p-1)]; + tensor var_4616 = add(x = var_4614, y = var_4615)[name = tensor("op_4616")]; + tensor var_4617 = const()[name = tensor("op_4617"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54656384)))]; + tensor var_4620 = mul(x = var_4616, y = var_4617)[name = tensor("op_4620")]; + tensor input_397 = add(x = xt_231, y = var_4620)[name = tensor("input_397")]; + tensor weight_199_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54656576))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54667904))), name = tensor("weight_199_palettized"), shape = tensor([32, 32, 11])]; + tensor xt_233_pad_type_0 = const()[name = tensor("xt_233_pad_type_0"), val = tensor("custom")]; + tensor xt_233_pad_0 = const()[name = tensor("xt_233_pad_0"), val = tensor([5, 5])]; + tensor xt_233_strides_0 = const()[name = tensor("xt_233_strides_0"), val = tensor([1])]; + tensor xt_233_dilations_0 = const()[name = tensor("xt_233_dilations_0"), val = tensor([1])]; + tensor xt_233_groups_0 = const()[name = tensor("xt_233_groups_0"), val = tensor(1)]; + tensor xt_233 = conv(bias = generator_noise_res_3_convs2_2_bias, dilations = xt_233_dilations_0, groups = xt_233_groups_0, pad = xt_233_pad_0, pad_type = xt_233_pad_type_0, strides = xt_233_strides_0, weight = weight_199_palettized, x = input_397)[name = tensor("xt_233")]; + tensor x_source = add(x = xt_233, y = input_391)[name = tensor("x_source")]; + tensor op_4631_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54668992))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54677248))), name = tensor("op_4631_palettized"), shape = tensor([64, 32, 4])]; + tensor x_15_pad_type_0 = const()[name = tensor("x_15_pad_type_0"), val = tensor("custom")]; + tensor x_15_pad_0 = const()[name = tensor("x_15_pad_0"), val = tensor([1, 1])]; + tensor x_15_strides_0 = const()[name = tensor("x_15_strides_0"), val = tensor([2])]; + tensor x_15_dilations_0 = const()[name = tensor("x_15_dilations_0"), val = tensor([1])]; + tensor x_15_groups_0 = const()[name = tensor("x_15_groups_0"), val = tensor(1)]; + tensor x_15 = conv_transpose(bias = generator_ups_3_bias, dilations = x_15_dilations_0, groups = x_15_groups_0, pad = x_15_pad_0, pad_type = x_15_pad_type_0, strides = x_15_strides_0, weight = op_4631_palettized, x = input_399)[name = tensor("x_15")]; + tensor input_401 = add(x = x_15, y = x_source)[name = tensor("input_401")]; + tensor var_4650 = const()[name = tensor("op_4650"), val = tensor(0x1.4f8b58p-17)]; + tensor h_353 = linear(bias = generator_resblocks_9_adain1_0_fc_bias, weight = generator_resblocks_9_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_89")]; + tensor var_4732 = const()[name = tensor("op_4732"), val = tensor([1, 64, 1])]; + tensor h_355 = reshape(shape = var_4732, x = h_353)[name = tensor("h_355")]; + tensor var_4734_split_sizes_0 = const()[name = tensor("op_4734_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4734_axis_0 = const()[name = tensor("op_4734_axis_0"), val = tensor(1)]; + tensor var_4734_0, tensor var_4734_1 = split(axis = var_4734_axis_0, split_sizes = var_4734_split_sizes_0, x = h_355)[name = tensor("op_4734")]; + tensor var_4736_promoted = const()[name = tensor("op_4736_promoted"), val = tensor(0x1p+0)]; + tensor var_4737 = add(x = var_4734_0, y = var_4736_promoted)[name = tensor("op_4737")]; + tensor var_4738 = instance_norm(epsilon = var_4650, x = input_401)[name = tensor("op_4738")]; + tensor var_4739 = mul(x = var_4737, y = var_4738)[name = tensor("op_4739")]; + tensor xt_235 = add(x = var_4739, y = var_4734_1)[name = tensor("xt_235")]; + tensor var_4742 = const()[name = tensor("op_4742"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54678336)))]; + tensor var_4743 = mul(x = xt_235, y = var_4742)[name = tensor("op_4743")]; + tensor cv_157 = cos(x = var_4743)[name = tensor("cv_157")]; + tensor var_4745 = const()[name = tensor("op_4745"), val = tensor(-0x1p-1)]; + tensor var_4746 = mul(x = cv_157, y = var_4745)[name = tensor("op_4746")]; + tensor var_4747 = const()[name = tensor("op_4747"), val = tensor(0x1p-1)]; + tensor var_4748 = add(x = var_4746, y = var_4747)[name = tensor("op_4748")]; + tensor var_4749 = const()[name = tensor("op_4749"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54678528)))]; + tensor var_4752 = mul(x = var_4748, y = var_4749)[name = tensor("op_4752")]; + tensor input_403 = add(x = xt_235, y = var_4752)[name = tensor("input_403")]; + tensor weight_201_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54678720))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54681856))), name = tensor("weight_201_palettized"), shape = tensor([32, 32, 3])]; + tensor input_405_pad_type_0 = const()[name = tensor("input_405_pad_type_0"), val = tensor("custom")]; + tensor input_405_pad_0 = const()[name = tensor("input_405_pad_0"), val = tensor([1, 1])]; + tensor input_405_strides_0 = const()[name = tensor("input_405_strides_0"), val = tensor([1])]; + tensor input_405_dilations_0 = const()[name = tensor("input_405_dilations_0"), val = tensor([1])]; + tensor input_405_groups_0 = const()[name = tensor("input_405_groups_0"), val = tensor(1)]; + tensor input_405 = conv(bias = generator_resblocks_9_convs1_0_bias, dilations = input_405_dilations_0, groups = input_405_groups_0, pad = input_405_pad_0, pad_type = input_405_pad_type_0, strides = input_405_strides_0, weight = weight_201_palettized, x = input_403)[name = tensor("input_405")]; + tensor h_357 = linear(bias = generator_resblocks_9_adain2_0_fc_bias, weight = generator_resblocks_9_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_90")]; + tensor var_4767 = const()[name = tensor("op_4767"), val = tensor([1, 64, 1])]; + tensor h_359 = reshape(shape = var_4767, x = h_357)[name = tensor("h_359")]; + tensor var_4769_split_sizes_0 = const()[name = tensor("op_4769_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4769_axis_0 = const()[name = tensor("op_4769_axis_0"), val = tensor(1)]; + tensor var_4769_0, tensor var_4769_1 = split(axis = var_4769_axis_0, split_sizes = var_4769_split_sizes_0, x = h_359)[name = tensor("op_4769")]; + tensor var_4771_promoted = const()[name = tensor("op_4771_promoted"), val = tensor(0x1p+0)]; + tensor var_4772 = add(x = var_4769_0, y = var_4771_promoted)[name = tensor("op_4772")]; + tensor var_4773 = instance_norm(epsilon = var_4650, x = input_405)[name = tensor("op_4773")]; + tensor var_4774 = mul(x = var_4772, y = var_4773)[name = tensor("op_4774")]; + tensor xt_237 = add(x = var_4774, y = var_4769_1)[name = tensor("xt_237")]; + tensor var_4777 = const()[name = tensor("op_4777"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54682944)))]; + tensor var_4778 = mul(x = xt_237, y = var_4777)[name = tensor("op_4778")]; + tensor cv_159 = cos(x = var_4778)[name = tensor("cv_159")]; + tensor var_4780 = const()[name = tensor("op_4780"), val = tensor(-0x1p-1)]; + tensor var_4781 = mul(x = cv_159, y = var_4780)[name = tensor("op_4781")]; + tensor var_4782 = const()[name = tensor("op_4782"), val = tensor(0x1p-1)]; + tensor var_4783 = add(x = var_4781, y = var_4782)[name = tensor("op_4783")]; + tensor var_4784 = const()[name = tensor("op_4784"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54683136)))]; + tensor var_4787 = mul(x = var_4783, y = var_4784)[name = tensor("op_4787")]; + tensor input_407 = add(x = xt_237, y = var_4787)[name = tensor("input_407")]; + tensor weight_203_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54683328))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54686464))), name = tensor("weight_203_palettized"), shape = tensor([32, 32, 3])]; + tensor xt_239_pad_type_0 = const()[name = tensor("xt_239_pad_type_0"), val = tensor("custom")]; + tensor xt_239_pad_0 = const()[name = tensor("xt_239_pad_0"), val = tensor([1, 1])]; + tensor xt_239_strides_0 = const()[name = tensor("xt_239_strides_0"), val = tensor([1])]; + tensor xt_239_dilations_0 = const()[name = tensor("xt_239_dilations_0"), val = tensor([1])]; + tensor xt_239_groups_0 = const()[name = tensor("xt_239_groups_0"), val = tensor(1)]; + tensor xt_239 = conv(bias = generator_resblocks_9_convs2_0_bias, dilations = xt_239_dilations_0, groups = xt_239_groups_0, pad = xt_239_pad_0, pad_type = xt_239_pad_type_0, strides = xt_239_strides_0, weight = weight_203_palettized, x = input_407)[name = tensor("xt_239")]; + tensor input_409 = add(x = xt_239, y = input_401)[name = tensor("input_409")]; + tensor h_361 = linear(bias = generator_resblocks_9_adain1_1_fc_bias, weight = generator_resblocks_9_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_91")]; + tensor var_4803 = const()[name = tensor("op_4803"), val = tensor([1, 64, 1])]; + tensor h_363 = reshape(shape = var_4803, x = h_361)[name = tensor("h_363")]; + tensor var_4805_split_sizes_0 = const()[name = tensor("op_4805_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4805_axis_0 = const()[name = tensor("op_4805_axis_0"), val = tensor(1)]; + tensor var_4805_0, tensor var_4805_1 = split(axis = var_4805_axis_0, split_sizes = var_4805_split_sizes_0, x = h_363)[name = tensor("op_4805")]; + tensor var_4807_promoted = const()[name = tensor("op_4807_promoted"), val = tensor(0x1p+0)]; + tensor var_4808 = add(x = var_4805_0, y = var_4807_promoted)[name = tensor("op_4808")]; + tensor var_4809 = instance_norm(epsilon = var_4650, x = input_409)[name = tensor("op_4809")]; + tensor var_4810 = mul(x = var_4808, y = var_4809)[name = tensor("op_4810")]; + tensor xt_241 = add(x = var_4810, y = var_4805_1)[name = tensor("xt_241")]; + tensor var_4813 = const()[name = tensor("op_4813"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54687552)))]; + tensor var_4814 = mul(x = xt_241, y = var_4813)[name = tensor("op_4814")]; + tensor cv_161 = cos(x = var_4814)[name = tensor("cv_161")]; + tensor var_4816 = const()[name = tensor("op_4816"), val = tensor(-0x1p-1)]; + tensor var_4817 = mul(x = cv_161, y = var_4816)[name = tensor("op_4817")]; + tensor var_4818 = const()[name = tensor("op_4818"), val = tensor(0x1p-1)]; + tensor var_4819 = add(x = var_4817, y = var_4818)[name = tensor("op_4819")]; + tensor var_4820 = const()[name = tensor("op_4820"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54687744)))]; + tensor var_4823 = mul(x = var_4819, y = var_4820)[name = tensor("op_4823")]; + tensor input_411 = add(x = xt_241, y = var_4823)[name = tensor("input_411")]; + tensor weight_205_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54687936))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54691072))), name = tensor("weight_205_palettized"), shape = tensor([32, 32, 3])]; + tensor input_413_pad_type_0 = const()[name = tensor("input_413_pad_type_0"), val = tensor("custom")]; + tensor input_413_pad_0 = const()[name = tensor("input_413_pad_0"), val = tensor([3, 3])]; + tensor input_413_dilations_0 = const()[name = tensor("input_413_dilations_0"), val = tensor([3])]; + tensor input_413_strides_0 = const()[name = tensor("input_413_strides_0"), val = tensor([1])]; + tensor input_413_groups_0 = const()[name = tensor("input_413_groups_0"), val = tensor(1)]; + tensor input_413 = conv(bias = generator_resblocks_9_convs1_1_bias, dilations = input_413_dilations_0, groups = input_413_groups_0, pad = input_413_pad_0, pad_type = input_413_pad_type_0, strides = input_413_strides_0, weight = weight_205_palettized, x = input_411)[name = tensor("input_413")]; + tensor h_365 = linear(bias = generator_resblocks_9_adain2_1_fc_bias, weight = generator_resblocks_9_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_92")]; + tensor var_4838 = const()[name = tensor("op_4838"), val = tensor([1, 64, 1])]; + tensor h_367 = reshape(shape = var_4838, x = h_365)[name = tensor("h_367")]; + tensor var_4840_split_sizes_0 = const()[name = tensor("op_4840_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4840_axis_0 = const()[name = tensor("op_4840_axis_0"), val = tensor(1)]; + tensor var_4840_0, tensor var_4840_1 = split(axis = var_4840_axis_0, split_sizes = var_4840_split_sizes_0, x = h_367)[name = tensor("op_4840")]; + tensor var_4842_promoted = const()[name = tensor("op_4842_promoted"), val = tensor(0x1p+0)]; + tensor var_4843 = add(x = var_4840_0, y = var_4842_promoted)[name = tensor("op_4843")]; + tensor var_4844 = instance_norm(epsilon = var_4650, x = input_413)[name = tensor("op_4844")]; + tensor var_4845 = mul(x = var_4843, y = var_4844)[name = tensor("op_4845")]; + tensor xt_243 = add(x = var_4845, y = var_4840_1)[name = tensor("xt_243")]; + tensor var_4848 = const()[name = tensor("op_4848"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54692160)))]; + tensor var_4849 = mul(x = xt_243, y = var_4848)[name = tensor("op_4849")]; + tensor cv_163 = cos(x = var_4849)[name = tensor("cv_163")]; + tensor var_4851 = const()[name = tensor("op_4851"), val = tensor(-0x1p-1)]; + tensor var_4852 = mul(x = cv_163, y = var_4851)[name = tensor("op_4852")]; + tensor var_4853 = const()[name = tensor("op_4853"), val = tensor(0x1p-1)]; + tensor var_4854 = add(x = var_4852, y = var_4853)[name = tensor("op_4854")]; + tensor var_4855 = const()[name = tensor("op_4855"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54692352)))]; + tensor var_4858 = mul(x = var_4854, y = var_4855)[name = tensor("op_4858")]; + tensor input_415 = add(x = xt_243, y = var_4858)[name = tensor("input_415")]; + tensor weight_207_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54692544))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54695680))), name = tensor("weight_207_palettized"), shape = tensor([32, 32, 3])]; + tensor xt_245_pad_type_0 = const()[name = tensor("xt_245_pad_type_0"), val = tensor("custom")]; + tensor xt_245_pad_0 = const()[name = tensor("xt_245_pad_0"), val = tensor([1, 1])]; + tensor xt_245_strides_0 = const()[name = tensor("xt_245_strides_0"), val = tensor([1])]; + tensor xt_245_dilations_0 = const()[name = tensor("xt_245_dilations_0"), val = tensor([1])]; + tensor xt_245_groups_0 = const()[name = tensor("xt_245_groups_0"), val = tensor(1)]; + tensor xt_245 = conv(bias = generator_resblocks_9_convs2_1_bias, dilations = xt_245_dilations_0, groups = xt_245_groups_0, pad = xt_245_pad_0, pad_type = xt_245_pad_type_0, strides = xt_245_strides_0, weight = weight_207_palettized, x = input_415)[name = tensor("xt_245")]; + tensor input_417 = add(x = xt_245, y = input_409)[name = tensor("input_417")]; + tensor h_369 = linear(bias = generator_resblocks_9_adain1_2_fc_bias, weight = generator_resblocks_9_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_93")]; + tensor var_4874 = const()[name = tensor("op_4874"), val = tensor([1, 64, 1])]; + tensor h_371 = reshape(shape = var_4874, x = h_369)[name = tensor("h_371")]; + tensor var_4876_split_sizes_0 = const()[name = tensor("op_4876_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4876_axis_0 = const()[name = tensor("op_4876_axis_0"), val = tensor(1)]; + tensor var_4876_0, tensor var_4876_1 = split(axis = var_4876_axis_0, split_sizes = var_4876_split_sizes_0, x = h_371)[name = tensor("op_4876")]; + tensor var_4878_promoted = const()[name = tensor("op_4878_promoted"), val = tensor(0x1p+0)]; + tensor var_4879 = add(x = var_4876_0, y = var_4878_promoted)[name = tensor("op_4879")]; + tensor var_4880 = instance_norm(epsilon = var_4650, x = input_417)[name = tensor("op_4880")]; + tensor var_4881 = mul(x = var_4879, y = var_4880)[name = tensor("op_4881")]; + tensor xt_247 = add(x = var_4881, y = var_4876_1)[name = tensor("xt_247")]; + tensor var_4884 = const()[name = tensor("op_4884"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54696768)))]; + tensor var_4885 = mul(x = xt_247, y = var_4884)[name = tensor("op_4885")]; + tensor cv_165 = cos(x = var_4885)[name = tensor("cv_165")]; + tensor var_4887 = const()[name = tensor("op_4887"), val = tensor(-0x1p-1)]; + tensor var_4888 = mul(x = cv_165, y = var_4887)[name = tensor("op_4888")]; + tensor var_4889 = const()[name = tensor("op_4889"), val = tensor(0x1p-1)]; + tensor var_4890 = add(x = var_4888, y = var_4889)[name = tensor("op_4890")]; + tensor var_4891 = const()[name = tensor("op_4891"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54696960)))]; + tensor var_4894 = mul(x = var_4890, y = var_4891)[name = tensor("op_4894")]; + tensor input_419 = add(x = xt_247, y = var_4894)[name = tensor("input_419")]; + tensor weight_209_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54697152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54700288))), name = tensor("weight_209_palettized"), shape = tensor([32, 32, 3])]; + tensor input_421_pad_type_0 = const()[name = tensor("input_421_pad_type_0"), val = tensor("custom")]; + tensor input_421_pad_0 = const()[name = tensor("input_421_pad_0"), val = tensor([5, 5])]; + tensor input_421_dilations_0 = const()[name = tensor("input_421_dilations_0"), val = tensor([5])]; + tensor input_421_strides_0 = const()[name = tensor("input_421_strides_0"), val = tensor([1])]; + tensor input_421_groups_0 = const()[name = tensor("input_421_groups_0"), val = tensor(1)]; + tensor input_421 = conv(bias = generator_resblocks_9_convs1_2_bias, dilations = input_421_dilations_0, groups = input_421_groups_0, pad = input_421_pad_0, pad_type = input_421_pad_type_0, strides = input_421_strides_0, weight = weight_209_palettized, x = input_419)[name = tensor("input_421")]; + tensor h_373 = linear(bias = generator_resblocks_9_adain2_2_fc_bias, weight = generator_resblocks_9_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_94")]; + tensor var_4909 = const()[name = tensor("op_4909"), val = tensor([1, 64, 1])]; + tensor h_375 = reshape(shape = var_4909, x = h_373)[name = tensor("h_375")]; + tensor var_4911_split_sizes_0 = const()[name = tensor("op_4911_split_sizes_0"), val = tensor([32, 32])]; + tensor var_4911_axis_0 = const()[name = tensor("op_4911_axis_0"), val = tensor(1)]; + tensor var_4911_0, tensor var_4911_1 = split(axis = var_4911_axis_0, split_sizes = var_4911_split_sizes_0, x = h_375)[name = tensor("op_4911")]; + tensor var_4913_promoted = const()[name = tensor("op_4913_promoted"), val = tensor(0x1p+0)]; + tensor var_4914 = add(x = var_4911_0, y = var_4913_promoted)[name = tensor("op_4914")]; + tensor var_4915 = instance_norm(epsilon = var_4650, x = input_421)[name = tensor("op_4915")]; + tensor var_4916 = mul(x = var_4914, y = var_4915)[name = tensor("op_4916")]; + tensor xt_249 = add(x = var_4916, y = var_4911_1)[name = tensor("xt_249")]; + tensor var_4919 = const()[name = tensor("op_4919"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54701376)))]; + tensor var_4920 = mul(x = xt_249, y = var_4919)[name = tensor("op_4920")]; + tensor cv_167 = cos(x = var_4920)[name = tensor("cv_167")]; + tensor var_4922 = const()[name = tensor("op_4922"), val = tensor(-0x1p-1)]; + tensor var_4923 = mul(x = cv_167, y = var_4922)[name = tensor("op_4923")]; + tensor var_4924 = const()[name = tensor("op_4924"), val = tensor(0x1p-1)]; + tensor var_4925 = add(x = var_4923, y = var_4924)[name = tensor("op_4925")]; + tensor var_4926 = const()[name = tensor("op_4926"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54701568)))]; + tensor var_4929 = mul(x = var_4925, y = var_4926)[name = tensor("op_4929")]; + tensor input_423 = add(x = xt_249, y = var_4929)[name = tensor("input_423")]; + tensor weight_211_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54701760))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54704896))), name = tensor("weight_211_palettized"), shape = tensor([32, 32, 3])]; + tensor xt_251_pad_type_0 = const()[name = tensor("xt_251_pad_type_0"), val = tensor("custom")]; + tensor xt_251_pad_0 = const()[name = tensor("xt_251_pad_0"), val = tensor([1, 1])]; + tensor xt_251_strides_0 = const()[name = tensor("xt_251_strides_0"), val = tensor([1])]; + tensor xt_251_dilations_0 = const()[name = tensor("xt_251_dilations_0"), val = tensor([1])]; + tensor xt_251_groups_0 = const()[name = tensor("xt_251_groups_0"), val = tensor(1)]; + tensor xt_251 = conv(bias = generator_resblocks_9_convs2_2_bias, dilations = xt_251_dilations_0, groups = xt_251_groups_0, pad = xt_251_pad_0, pad_type = xt_251_pad_type_0, strides = xt_251_strides_0, weight = weight_211_palettized, x = input_423)[name = tensor("xt_251")]; + tensor xs_19 = add(x = xt_251, y = input_417)[name = tensor("xs_19")]; + tensor var_4946 = const()[name = tensor("op_4946"), val = tensor(0x1.4f8b58p-17)]; + tensor h_377 = linear(bias = generator_resblocks_10_adain1_0_fc_bias, weight = generator_resblocks_10_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_95")]; + tensor var_5028 = const()[name = tensor("op_5028"), val = tensor([1, 64, 1])]; + tensor h_379 = reshape(shape = var_5028, x = h_377)[name = tensor("h_379")]; + tensor var_5030_split_sizes_0 = const()[name = tensor("op_5030_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5030_axis_0 = const()[name = tensor("op_5030_axis_0"), val = tensor(1)]; + tensor var_5030_0, tensor var_5030_1 = split(axis = var_5030_axis_0, split_sizes = var_5030_split_sizes_0, x = h_379)[name = tensor("op_5030")]; + tensor var_5032_promoted = const()[name = tensor("op_5032_promoted"), val = tensor(0x1p+0)]; + tensor var_5033 = add(x = var_5030_0, y = var_5032_promoted)[name = tensor("op_5033")]; + tensor var_5035 = mul(x = var_5033, y = var_4738)[name = tensor("op_5035")]; + tensor xt_253 = add(x = var_5035, y = var_5030_1)[name = tensor("xt_253")]; + tensor var_5038 = const()[name = tensor("op_5038"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54705984)))]; + tensor var_5039 = mul(x = xt_253, y = var_5038)[name = tensor("op_5039")]; + tensor cv_169 = cos(x = var_5039)[name = tensor("cv_169")]; + tensor var_5041 = const()[name = tensor("op_5041"), val = tensor(-0x1p-1)]; + tensor var_5042 = mul(x = cv_169, y = var_5041)[name = tensor("op_5042")]; + tensor var_5043 = const()[name = tensor("op_5043"), val = tensor(0x1p-1)]; + tensor var_5044 = add(x = var_5042, y = var_5043)[name = tensor("op_5044")]; + tensor var_5045 = const()[name = tensor("op_5045"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54706176)))]; + tensor var_5048 = mul(x = var_5044, y = var_5045)[name = tensor("op_5048")]; + tensor input_425 = add(x = xt_253, y = var_5048)[name = tensor("input_425")]; + tensor weight_213_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54706368))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54713600))), name = tensor("weight_213_palettized"), shape = tensor([32, 32, 7])]; + tensor input_427_pad_type_0 = const()[name = tensor("input_427_pad_type_0"), val = tensor("custom")]; + tensor input_427_pad_0 = const()[name = tensor("input_427_pad_0"), val = tensor([3, 3])]; + tensor input_427_strides_0 = const()[name = tensor("input_427_strides_0"), val = tensor([1])]; + tensor input_427_dilations_0 = const()[name = tensor("input_427_dilations_0"), val = tensor([1])]; + tensor input_427_groups_0 = const()[name = tensor("input_427_groups_0"), val = tensor(1)]; + tensor input_427 = conv(bias = generator_resblocks_10_convs1_0_bias, dilations = input_427_dilations_0, groups = input_427_groups_0, pad = input_427_pad_0, pad_type = input_427_pad_type_0, strides = input_427_strides_0, weight = weight_213_palettized, x = input_425)[name = tensor("input_427")]; + tensor h_381 = linear(bias = generator_resblocks_10_adain2_0_fc_bias, weight = generator_resblocks_10_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_96")]; + tensor var_5063 = const()[name = tensor("op_5063"), val = tensor([1, 64, 1])]; + tensor h_383 = reshape(shape = var_5063, x = h_381)[name = tensor("h_383")]; + tensor var_5065_split_sizes_0 = const()[name = tensor("op_5065_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5065_axis_0 = const()[name = tensor("op_5065_axis_0"), val = tensor(1)]; + tensor var_5065_0, tensor var_5065_1 = split(axis = var_5065_axis_0, split_sizes = var_5065_split_sizes_0, x = h_383)[name = tensor("op_5065")]; + tensor var_5067_promoted = const()[name = tensor("op_5067_promoted"), val = tensor(0x1p+0)]; + tensor var_5068 = add(x = var_5065_0, y = var_5067_promoted)[name = tensor("op_5068")]; + tensor var_5069 = instance_norm(epsilon = var_4946, x = input_427)[name = tensor("op_5069")]; + tensor var_5070 = mul(x = var_5068, y = var_5069)[name = tensor("op_5070")]; + tensor xt_255 = add(x = var_5070, y = var_5065_1)[name = tensor("xt_255")]; + tensor var_5073 = const()[name = tensor("op_5073"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54714688)))]; + tensor var_5074 = mul(x = xt_255, y = var_5073)[name = tensor("op_5074")]; + tensor cv_171 = cos(x = var_5074)[name = tensor("cv_171")]; + tensor var_5076 = const()[name = tensor("op_5076"), val = tensor(-0x1p-1)]; + tensor var_5077 = mul(x = cv_171, y = var_5076)[name = tensor("op_5077")]; + tensor var_5078 = const()[name = tensor("op_5078"), val = tensor(0x1p-1)]; + tensor var_5079 = add(x = var_5077, y = var_5078)[name = tensor("op_5079")]; + tensor var_5080 = const()[name = tensor("op_5080"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54714880)))]; + tensor var_5083 = mul(x = var_5079, y = var_5080)[name = tensor("op_5083")]; + tensor input_429 = add(x = xt_255, y = var_5083)[name = tensor("input_429")]; + tensor weight_215_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54715072))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54722304))), name = tensor("weight_215_palettized"), shape = tensor([32, 32, 7])]; + tensor xt_257_pad_type_0 = const()[name = tensor("xt_257_pad_type_0"), val = tensor("custom")]; + tensor xt_257_pad_0 = const()[name = tensor("xt_257_pad_0"), val = tensor([3, 3])]; + tensor xt_257_strides_0 = const()[name = tensor("xt_257_strides_0"), val = tensor([1])]; + tensor xt_257_dilations_0 = const()[name = tensor("xt_257_dilations_0"), val = tensor([1])]; + tensor xt_257_groups_0 = const()[name = tensor("xt_257_groups_0"), val = tensor(1)]; + tensor xt_257 = conv(bias = generator_resblocks_10_convs2_0_bias, dilations = xt_257_dilations_0, groups = xt_257_groups_0, pad = xt_257_pad_0, pad_type = xt_257_pad_type_0, strides = xt_257_strides_0, weight = weight_215_palettized, x = input_429)[name = tensor("xt_257")]; + tensor input_431 = add(x = xt_257, y = input_401)[name = tensor("input_431")]; + tensor h_385 = linear(bias = generator_resblocks_10_adain1_1_fc_bias, weight = generator_resblocks_10_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_97")]; + tensor var_5099 = const()[name = tensor("op_5099"), val = tensor([1, 64, 1])]; + tensor h_387 = reshape(shape = var_5099, x = h_385)[name = tensor("h_387")]; + tensor var_5101_split_sizes_0 = const()[name = tensor("op_5101_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5101_axis_0 = const()[name = tensor("op_5101_axis_0"), val = tensor(1)]; + tensor var_5101_0, tensor var_5101_1 = split(axis = var_5101_axis_0, split_sizes = var_5101_split_sizes_0, x = h_387)[name = tensor("op_5101")]; + tensor var_5103_promoted = const()[name = tensor("op_5103_promoted"), val = tensor(0x1p+0)]; + tensor var_5104 = add(x = var_5101_0, y = var_5103_promoted)[name = tensor("op_5104")]; + tensor var_5105 = instance_norm(epsilon = var_4946, x = input_431)[name = tensor("op_5105")]; + tensor var_5106 = mul(x = var_5104, y = var_5105)[name = tensor("op_5106")]; + tensor xt_259 = add(x = var_5106, y = var_5101_1)[name = tensor("xt_259")]; + tensor var_5109 = const()[name = tensor("op_5109"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54723392)))]; + tensor var_5110 = mul(x = xt_259, y = var_5109)[name = tensor("op_5110")]; + tensor cv_173 = cos(x = var_5110)[name = tensor("cv_173")]; + tensor var_5112 = const()[name = tensor("op_5112"), val = tensor(-0x1p-1)]; + tensor var_5113 = mul(x = cv_173, y = var_5112)[name = tensor("op_5113")]; + tensor var_5114 = const()[name = tensor("op_5114"), val = tensor(0x1p-1)]; + tensor var_5115 = add(x = var_5113, y = var_5114)[name = tensor("op_5115")]; + tensor var_5116 = const()[name = tensor("op_5116"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54723584)))]; + tensor var_5119 = mul(x = var_5115, y = var_5116)[name = tensor("op_5119")]; + tensor input_433 = add(x = xt_259, y = var_5119)[name = tensor("input_433")]; + tensor weight_217_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54723776))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54731008))), name = tensor("weight_217_palettized"), shape = tensor([32, 32, 7])]; + tensor input_435_pad_type_0 = const()[name = tensor("input_435_pad_type_0"), val = tensor("custom")]; + tensor input_435_pad_0 = const()[name = tensor("input_435_pad_0"), val = tensor([9, 9])]; + tensor input_435_dilations_0 = const()[name = tensor("input_435_dilations_0"), val = tensor([3])]; + tensor input_435_strides_0 = const()[name = tensor("input_435_strides_0"), val = tensor([1])]; + tensor input_435_groups_0 = const()[name = tensor("input_435_groups_0"), val = tensor(1)]; + tensor input_435 = conv(bias = generator_resblocks_10_convs1_1_bias, dilations = input_435_dilations_0, groups = input_435_groups_0, pad = input_435_pad_0, pad_type = input_435_pad_type_0, strides = input_435_strides_0, weight = weight_217_palettized, x = input_433)[name = tensor("input_435")]; + tensor h_389 = linear(bias = generator_resblocks_10_adain2_1_fc_bias, weight = generator_resblocks_10_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_98")]; + tensor var_5134 = const()[name = tensor("op_5134"), val = tensor([1, 64, 1])]; + tensor h_391 = reshape(shape = var_5134, x = h_389)[name = tensor("h_391")]; + tensor var_5136_split_sizes_0 = const()[name = tensor("op_5136_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5136_axis_0 = const()[name = tensor("op_5136_axis_0"), val = tensor(1)]; + tensor var_5136_0, tensor var_5136_1 = split(axis = var_5136_axis_0, split_sizes = var_5136_split_sizes_0, x = h_391)[name = tensor("op_5136")]; + tensor var_5138_promoted = const()[name = tensor("op_5138_promoted"), val = tensor(0x1p+0)]; + tensor var_5139 = add(x = var_5136_0, y = var_5138_promoted)[name = tensor("op_5139")]; + tensor var_5140 = instance_norm(epsilon = var_4946, x = input_435)[name = tensor("op_5140")]; + tensor var_5141 = mul(x = var_5139, y = var_5140)[name = tensor("op_5141")]; + tensor xt_261 = add(x = var_5141, y = var_5136_1)[name = tensor("xt_261")]; + tensor var_5144 = const()[name = tensor("op_5144"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54732096)))]; + tensor var_5145 = mul(x = xt_261, y = var_5144)[name = tensor("op_5145")]; + tensor cv_175 = cos(x = var_5145)[name = tensor("cv_175")]; + tensor var_5147 = const()[name = tensor("op_5147"), val = tensor(-0x1p-1)]; + tensor var_5148 = mul(x = cv_175, y = var_5147)[name = tensor("op_5148")]; + tensor var_5149 = const()[name = tensor("op_5149"), val = tensor(0x1p-1)]; + tensor var_5150 = add(x = var_5148, y = var_5149)[name = tensor("op_5150")]; + tensor var_5151 = const()[name = tensor("op_5151"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54732288)))]; + tensor var_5154 = mul(x = var_5150, y = var_5151)[name = tensor("op_5154")]; + tensor input_437 = add(x = xt_261, y = var_5154)[name = tensor("input_437")]; + tensor weight_219_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54732480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54739712))), name = tensor("weight_219_palettized"), shape = tensor([32, 32, 7])]; + tensor xt_263_pad_type_0 = const()[name = tensor("xt_263_pad_type_0"), val = tensor("custom")]; + tensor xt_263_pad_0 = const()[name = tensor("xt_263_pad_0"), val = tensor([3, 3])]; + tensor xt_263_strides_0 = const()[name = tensor("xt_263_strides_0"), val = tensor([1])]; + tensor xt_263_dilations_0 = const()[name = tensor("xt_263_dilations_0"), val = tensor([1])]; + tensor xt_263_groups_0 = const()[name = tensor("xt_263_groups_0"), val = tensor(1)]; + tensor xt_263 = conv(bias = generator_resblocks_10_convs2_1_bias, dilations = xt_263_dilations_0, groups = xt_263_groups_0, pad = xt_263_pad_0, pad_type = xt_263_pad_type_0, strides = xt_263_strides_0, weight = weight_219_palettized, x = input_437)[name = tensor("xt_263")]; + tensor input_439 = add(x = xt_263, y = input_431)[name = tensor("input_439")]; + tensor h_393 = linear(bias = generator_resblocks_10_adain1_2_fc_bias, weight = generator_resblocks_10_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_99")]; + tensor var_5170 = const()[name = tensor("op_5170"), val = tensor([1, 64, 1])]; + tensor h_395 = reshape(shape = var_5170, x = h_393)[name = tensor("h_395")]; + tensor var_5172_split_sizes_0 = const()[name = tensor("op_5172_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5172_axis_0 = const()[name = tensor("op_5172_axis_0"), val = tensor(1)]; + tensor var_5172_0, tensor var_5172_1 = split(axis = var_5172_axis_0, split_sizes = var_5172_split_sizes_0, x = h_395)[name = tensor("op_5172")]; + tensor var_5174_promoted = const()[name = tensor("op_5174_promoted"), val = tensor(0x1p+0)]; + tensor var_5175 = add(x = var_5172_0, y = var_5174_promoted)[name = tensor("op_5175")]; + tensor var_5176 = instance_norm(epsilon = var_4946, x = input_439)[name = tensor("op_5176")]; + tensor var_5177 = mul(x = var_5175, y = var_5176)[name = tensor("op_5177")]; + tensor xt_265 = add(x = var_5177, y = var_5172_1)[name = tensor("xt_265")]; + tensor var_5180 = const()[name = tensor("op_5180"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54740800)))]; + tensor var_5181 = mul(x = xt_265, y = var_5180)[name = tensor("op_5181")]; + tensor cv_177 = cos(x = var_5181)[name = tensor("cv_177")]; + tensor var_5183 = const()[name = tensor("op_5183"), val = tensor(-0x1p-1)]; + tensor var_5184 = mul(x = cv_177, y = var_5183)[name = tensor("op_5184")]; + tensor var_5185 = const()[name = tensor("op_5185"), val = tensor(0x1p-1)]; + tensor var_5186 = add(x = var_5184, y = var_5185)[name = tensor("op_5186")]; + tensor var_5187 = const()[name = tensor("op_5187"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54740992)))]; + tensor var_5190 = mul(x = var_5186, y = var_5187)[name = tensor("op_5190")]; + tensor input_441 = add(x = xt_265, y = var_5190)[name = tensor("input_441")]; + tensor weight_221_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54741184))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54748416))), name = tensor("weight_221_palettized"), shape = tensor([32, 32, 7])]; + tensor input_443_pad_type_0 = const()[name = tensor("input_443_pad_type_0"), val = tensor("custom")]; + tensor input_443_pad_0 = const()[name = tensor("input_443_pad_0"), val = tensor([15, 15])]; + tensor input_443_dilations_0 = const()[name = tensor("input_443_dilations_0"), val = tensor([5])]; + tensor input_443_strides_0 = const()[name = tensor("input_443_strides_0"), val = tensor([1])]; + tensor input_443_groups_0 = const()[name = tensor("input_443_groups_0"), val = tensor(1)]; + tensor input_443 = conv(bias = generator_resblocks_10_convs1_2_bias, dilations = input_443_dilations_0, groups = input_443_groups_0, pad = input_443_pad_0, pad_type = input_443_pad_type_0, strides = input_443_strides_0, weight = weight_221_palettized, x = input_441)[name = tensor("input_443")]; + tensor h_397 = linear(bias = generator_resblocks_10_adain2_2_fc_bias, weight = generator_resblocks_10_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_100")]; + tensor var_5205 = const()[name = tensor("op_5205"), val = tensor([1, 64, 1])]; + tensor h_399 = reshape(shape = var_5205, x = h_397)[name = tensor("h_399")]; + tensor var_5207_split_sizes_0 = const()[name = tensor("op_5207_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5207_axis_0 = const()[name = tensor("op_5207_axis_0"), val = tensor(1)]; + tensor var_5207_0, tensor var_5207_1 = split(axis = var_5207_axis_0, split_sizes = var_5207_split_sizes_0, x = h_399)[name = tensor("op_5207")]; + tensor var_5209_promoted = const()[name = tensor("op_5209_promoted"), val = tensor(0x1p+0)]; + tensor var_5210 = add(x = var_5207_0, y = var_5209_promoted)[name = tensor("op_5210")]; + tensor var_5211 = instance_norm(epsilon = var_4946, x = input_443)[name = tensor("op_5211")]; + tensor var_5212 = mul(x = var_5210, y = var_5211)[name = tensor("op_5212")]; + tensor xt_267 = add(x = var_5212, y = var_5207_1)[name = tensor("xt_267")]; + tensor var_5215 = const()[name = tensor("op_5215"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54749504)))]; + tensor var_5216 = mul(x = xt_267, y = var_5215)[name = tensor("op_5216")]; + tensor cv_179 = cos(x = var_5216)[name = tensor("cv_179")]; + tensor var_5218 = const()[name = tensor("op_5218"), val = tensor(-0x1p-1)]; + tensor var_5219 = mul(x = cv_179, y = var_5218)[name = tensor("op_5219")]; + tensor var_5220 = const()[name = tensor("op_5220"), val = tensor(0x1p-1)]; + tensor var_5221 = add(x = var_5219, y = var_5220)[name = tensor("op_5221")]; + tensor var_5222 = const()[name = tensor("op_5222"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54749696)))]; + tensor var_5225 = mul(x = var_5221, y = var_5222)[name = tensor("op_5225")]; + tensor input_445 = add(x = xt_267, y = var_5225)[name = tensor("input_445")]; + tensor weight_223_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54749888))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54757120))), name = tensor("weight_223_palettized"), shape = tensor([32, 32, 7])]; + tensor xt_269_pad_type_0 = const()[name = tensor("xt_269_pad_type_0"), val = tensor("custom")]; + tensor xt_269_pad_0 = const()[name = tensor("xt_269_pad_0"), val = tensor([3, 3])]; + tensor xt_269_strides_0 = const()[name = tensor("xt_269_strides_0"), val = tensor([1])]; + tensor xt_269_dilations_0 = const()[name = tensor("xt_269_dilations_0"), val = tensor([1])]; + tensor xt_269_groups_0 = const()[name = tensor("xt_269_groups_0"), val = tensor(1)]; + tensor xt_269 = conv(bias = generator_resblocks_10_convs2_2_bias, dilations = xt_269_dilations_0, groups = xt_269_groups_0, pad = xt_269_pad_0, pad_type = xt_269_pad_type_0, strides = xt_269_strides_0, weight = weight_223_palettized, x = input_445)[name = tensor("xt_269")]; + tensor var_5234 = add(x = xt_269, y = input_439)[name = tensor("op_5234")]; + tensor xs_21 = add(x = xs_19, y = var_5234)[name = tensor("xs_21")]; + tensor var_5244 = const()[name = tensor("op_5244"), val = tensor(0x1.4f8b58p-17)]; + tensor h_401 = linear(bias = generator_resblocks_11_adain1_0_fc_bias, weight = generator_resblocks_11_adain1_0_fc_weight_palettized, x = s)[name = tensor("linear_101")]; + tensor var_5326 = const()[name = tensor("op_5326"), val = tensor([1, 64, 1])]; + tensor h_403 = reshape(shape = var_5326, x = h_401)[name = tensor("h_403")]; + tensor var_5328_split_sizes_0 = const()[name = tensor("op_5328_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5328_axis_0 = const()[name = tensor("op_5328_axis_0"), val = tensor(1)]; + tensor var_5328_0, tensor var_5328_1 = split(axis = var_5328_axis_0, split_sizes = var_5328_split_sizes_0, x = h_403)[name = tensor("op_5328")]; + tensor var_5330_promoted = const()[name = tensor("op_5330_promoted"), val = tensor(0x1p+0)]; + tensor var_5331 = add(x = var_5328_0, y = var_5330_promoted)[name = tensor("op_5331")]; + tensor var_5333 = mul(x = var_5331, y = var_4738)[name = tensor("op_5333")]; + tensor xt_271 = add(x = var_5333, y = var_5328_1)[name = tensor("xt_271")]; + tensor var_5336 = const()[name = tensor("op_5336"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54758208)))]; + tensor var_5337 = mul(x = xt_271, y = var_5336)[name = tensor("op_5337")]; + tensor cv_181 = cos(x = var_5337)[name = tensor("cv_181")]; + tensor var_5339 = const()[name = tensor("op_5339"), val = tensor(-0x1p-1)]; + tensor var_5340 = mul(x = cv_181, y = var_5339)[name = tensor("op_5340")]; + tensor var_5341 = const()[name = tensor("op_5341"), val = tensor(0x1p-1)]; + tensor var_5342 = add(x = var_5340, y = var_5341)[name = tensor("op_5342")]; + tensor var_5343 = const()[name = tensor("op_5343"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54758400)))]; + tensor var_5346 = mul(x = var_5342, y = var_5343)[name = tensor("op_5346")]; + tensor input_447 = add(x = xt_271, y = var_5346)[name = tensor("input_447")]; + tensor weight_225_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54758592))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54769920))), name = tensor("weight_225_palettized"), shape = tensor([32, 32, 11])]; + tensor input_449_pad_type_0 = const()[name = tensor("input_449_pad_type_0"), val = tensor("custom")]; + tensor input_449_pad_0 = const()[name = tensor("input_449_pad_0"), val = tensor([5, 5])]; + tensor input_449_strides_0 = const()[name = tensor("input_449_strides_0"), val = tensor([1])]; + tensor input_449_dilations_0 = const()[name = tensor("input_449_dilations_0"), val = tensor([1])]; + tensor input_449_groups_0 = const()[name = tensor("input_449_groups_0"), val = tensor(1)]; + tensor input_449 = conv(bias = generator_resblocks_11_convs1_0_bias, dilations = input_449_dilations_0, groups = input_449_groups_0, pad = input_449_pad_0, pad_type = input_449_pad_type_0, strides = input_449_strides_0, weight = weight_225_palettized, x = input_447)[name = tensor("input_449")]; + tensor h_405 = linear(bias = generator_resblocks_11_adain2_0_fc_bias, weight = generator_resblocks_11_adain2_0_fc_weight_palettized, x = s)[name = tensor("linear_102")]; + tensor var_5361 = const()[name = tensor("op_5361"), val = tensor([1, 64, 1])]; + tensor h_407 = reshape(shape = var_5361, x = h_405)[name = tensor("h_407")]; + tensor var_5363_split_sizes_0 = const()[name = tensor("op_5363_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5363_axis_0 = const()[name = tensor("op_5363_axis_0"), val = tensor(1)]; + tensor var_5363_0, tensor var_5363_1 = split(axis = var_5363_axis_0, split_sizes = var_5363_split_sizes_0, x = h_407)[name = tensor("op_5363")]; + tensor var_5365_promoted = const()[name = tensor("op_5365_promoted"), val = tensor(0x1p+0)]; + tensor var_5366 = add(x = var_5363_0, y = var_5365_promoted)[name = tensor("op_5366")]; + tensor var_5367 = instance_norm(epsilon = var_5244, x = input_449)[name = tensor("op_5367")]; + tensor var_5368 = mul(x = var_5366, y = var_5367)[name = tensor("op_5368")]; + tensor xt_273 = add(x = var_5368, y = var_5363_1)[name = tensor("xt_273")]; + tensor var_5371 = const()[name = tensor("op_5371"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54771008)))]; + tensor var_5372 = mul(x = xt_273, y = var_5371)[name = tensor("op_5372")]; + tensor cv_183 = cos(x = var_5372)[name = tensor("cv_183")]; + tensor var_5374 = const()[name = tensor("op_5374"), val = tensor(-0x1p-1)]; + tensor var_5375 = mul(x = cv_183, y = var_5374)[name = tensor("op_5375")]; + tensor var_5376 = const()[name = tensor("op_5376"), val = tensor(0x1p-1)]; + tensor var_5377 = add(x = var_5375, y = var_5376)[name = tensor("op_5377")]; + tensor var_5378 = const()[name = tensor("op_5378"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54771200)))]; + tensor var_5381 = mul(x = var_5377, y = var_5378)[name = tensor("op_5381")]; + tensor input_451 = add(x = xt_273, y = var_5381)[name = tensor("input_451")]; + tensor weight_227_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54771392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54782720))), name = tensor("weight_227_palettized"), shape = tensor([32, 32, 11])]; + tensor xt_275_pad_type_0 = const()[name = tensor("xt_275_pad_type_0"), val = tensor("custom")]; + tensor xt_275_pad_0 = const()[name = tensor("xt_275_pad_0"), val = tensor([5, 5])]; + tensor xt_275_strides_0 = const()[name = tensor("xt_275_strides_0"), val = tensor([1])]; + tensor xt_275_dilations_0 = const()[name = tensor("xt_275_dilations_0"), val = tensor([1])]; + tensor xt_275_groups_0 = const()[name = tensor("xt_275_groups_0"), val = tensor(1)]; + tensor xt_275 = conv(bias = generator_resblocks_11_convs2_0_bias, dilations = xt_275_dilations_0, groups = xt_275_groups_0, pad = xt_275_pad_0, pad_type = xt_275_pad_type_0, strides = xt_275_strides_0, weight = weight_227_palettized, x = input_451)[name = tensor("xt_275")]; + tensor input_453 = add(x = xt_275, y = input_401)[name = tensor("input_453")]; + tensor h_409 = linear(bias = generator_resblocks_11_adain1_1_fc_bias, weight = generator_resblocks_11_adain1_1_fc_weight_palettized, x = s)[name = tensor("linear_103")]; + tensor var_5397 = const()[name = tensor("op_5397"), val = tensor([1, 64, 1])]; + tensor h_411 = reshape(shape = var_5397, x = h_409)[name = tensor("h_411")]; + tensor var_5399_split_sizes_0 = const()[name = tensor("op_5399_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5399_axis_0 = const()[name = tensor("op_5399_axis_0"), val = tensor(1)]; + tensor var_5399_0, tensor var_5399_1 = split(axis = var_5399_axis_0, split_sizes = var_5399_split_sizes_0, x = h_411)[name = tensor("op_5399")]; + tensor var_5401_promoted = const()[name = tensor("op_5401_promoted"), val = tensor(0x1p+0)]; + tensor var_5402 = add(x = var_5399_0, y = var_5401_promoted)[name = tensor("op_5402")]; + tensor var_5403 = instance_norm(epsilon = var_5244, x = input_453)[name = tensor("op_5403")]; + tensor var_5404 = mul(x = var_5402, y = var_5403)[name = tensor("op_5404")]; + tensor xt_277 = add(x = var_5404, y = var_5399_1)[name = tensor("xt_277")]; + tensor var_5407 = const()[name = tensor("op_5407"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54783808)))]; + tensor var_5408 = mul(x = xt_277, y = var_5407)[name = tensor("op_5408")]; + tensor cv_185 = cos(x = var_5408)[name = tensor("cv_185")]; + tensor var_5410 = const()[name = tensor("op_5410"), val = tensor(-0x1p-1)]; + tensor var_5411 = mul(x = cv_185, y = var_5410)[name = tensor("op_5411")]; + tensor var_5412 = const()[name = tensor("op_5412"), val = tensor(0x1p-1)]; + tensor var_5413 = add(x = var_5411, y = var_5412)[name = tensor("op_5413")]; + tensor var_5414 = const()[name = tensor("op_5414"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54784000)))]; + tensor var_5417 = mul(x = var_5413, y = var_5414)[name = tensor("op_5417")]; + tensor input_455 = add(x = xt_277, y = var_5417)[name = tensor("input_455")]; + tensor weight_229_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54784192))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54795520))), name = tensor("weight_229_palettized"), shape = tensor([32, 32, 11])]; + tensor input_457_pad_type_0 = const()[name = tensor("input_457_pad_type_0"), val = tensor("custom")]; + tensor input_457_pad_0 = const()[name = tensor("input_457_pad_0"), val = tensor([15, 15])]; + tensor input_457_dilations_0 = const()[name = tensor("input_457_dilations_0"), val = tensor([3])]; + tensor input_457_strides_0 = const()[name = tensor("input_457_strides_0"), val = tensor([1])]; + tensor input_457_groups_0 = const()[name = tensor("input_457_groups_0"), val = tensor(1)]; + tensor input_457 = conv(bias = generator_resblocks_11_convs1_1_bias, dilations = input_457_dilations_0, groups = input_457_groups_0, pad = input_457_pad_0, pad_type = input_457_pad_type_0, strides = input_457_strides_0, weight = weight_229_palettized, x = input_455)[name = tensor("input_457")]; + tensor h_413 = linear(bias = generator_resblocks_11_adain2_1_fc_bias, weight = generator_resblocks_11_adain2_1_fc_weight_palettized, x = s)[name = tensor("linear_104")]; + tensor var_5432 = const()[name = tensor("op_5432"), val = tensor([1, 64, 1])]; + tensor h_415 = reshape(shape = var_5432, x = h_413)[name = tensor("h_415")]; + tensor var_5434_split_sizes_0 = const()[name = tensor("op_5434_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5434_axis_0 = const()[name = tensor("op_5434_axis_0"), val = tensor(1)]; + tensor var_5434_0, tensor var_5434_1 = split(axis = var_5434_axis_0, split_sizes = var_5434_split_sizes_0, x = h_415)[name = tensor("op_5434")]; + tensor var_5436_promoted = const()[name = tensor("op_5436_promoted"), val = tensor(0x1p+0)]; + tensor var_5437 = add(x = var_5434_0, y = var_5436_promoted)[name = tensor("op_5437")]; + tensor var_5438 = instance_norm(epsilon = var_5244, x = input_457)[name = tensor("op_5438")]; + tensor var_5439 = mul(x = var_5437, y = var_5438)[name = tensor("op_5439")]; + tensor xt_279 = add(x = var_5439, y = var_5434_1)[name = tensor("xt_279")]; + tensor var_5442 = const()[name = tensor("op_5442"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54796608)))]; + tensor var_5443 = mul(x = xt_279, y = var_5442)[name = tensor("op_5443")]; + tensor cv_187 = cos(x = var_5443)[name = tensor("cv_187")]; + tensor var_5445 = const()[name = tensor("op_5445"), val = tensor(-0x1p-1)]; + tensor var_5446 = mul(x = cv_187, y = var_5445)[name = tensor("op_5446")]; + tensor var_5447 = const()[name = tensor("op_5447"), val = tensor(0x1p-1)]; + tensor var_5448 = add(x = var_5446, y = var_5447)[name = tensor("op_5448")]; + tensor var_5449 = const()[name = tensor("op_5449"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54796800)))]; + tensor var_5452 = mul(x = var_5448, y = var_5449)[name = tensor("op_5452")]; + tensor input_459 = add(x = xt_279, y = var_5452)[name = tensor("input_459")]; + tensor weight_231_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54796992))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54808320))), name = tensor("weight_231_palettized"), shape = tensor([32, 32, 11])]; + tensor xt_281_pad_type_0 = const()[name = tensor("xt_281_pad_type_0"), val = tensor("custom")]; + tensor xt_281_pad_0 = const()[name = tensor("xt_281_pad_0"), val = tensor([5, 5])]; + tensor xt_281_strides_0 = const()[name = tensor("xt_281_strides_0"), val = tensor([1])]; + tensor xt_281_dilations_0 = const()[name = tensor("xt_281_dilations_0"), val = tensor([1])]; + tensor xt_281_groups_0 = const()[name = tensor("xt_281_groups_0"), val = tensor(1)]; + tensor xt_281 = conv(bias = generator_resblocks_11_convs2_1_bias, dilations = xt_281_dilations_0, groups = xt_281_groups_0, pad = xt_281_pad_0, pad_type = xt_281_pad_type_0, strides = xt_281_strides_0, weight = weight_231_palettized, x = input_459)[name = tensor("xt_281")]; + tensor input_461 = add(x = xt_281, y = input_453)[name = tensor("input_461")]; + tensor h_417 = linear(bias = generator_resblocks_11_adain1_2_fc_bias, weight = generator_resblocks_11_adain1_2_fc_weight_palettized, x = s)[name = tensor("linear_105")]; + tensor var_5468 = const()[name = tensor("op_5468"), val = tensor([1, 64, 1])]; + tensor h_419 = reshape(shape = var_5468, x = h_417)[name = tensor("h_419")]; + tensor var_5470_split_sizes_0 = const()[name = tensor("op_5470_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5470_axis_0 = const()[name = tensor("op_5470_axis_0"), val = tensor(1)]; + tensor var_5470_0, tensor var_5470_1 = split(axis = var_5470_axis_0, split_sizes = var_5470_split_sizes_0, x = h_419)[name = tensor("op_5470")]; + tensor var_5472_promoted = const()[name = tensor("op_5472_promoted"), val = tensor(0x1p+0)]; + tensor var_5473 = add(x = var_5470_0, y = var_5472_promoted)[name = tensor("op_5473")]; + tensor var_5474 = instance_norm(epsilon = var_5244, x = input_461)[name = tensor("op_5474")]; + tensor var_5475 = mul(x = var_5473, y = var_5474)[name = tensor("op_5475")]; + tensor xt_283 = add(x = var_5475, y = var_5470_1)[name = tensor("xt_283")]; + tensor var_5478 = const()[name = tensor("op_5478"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54809408)))]; + tensor var_5479 = mul(x = xt_283, y = var_5478)[name = tensor("op_5479")]; + tensor cv_189 = cos(x = var_5479)[name = tensor("cv_189")]; + tensor var_5481 = const()[name = tensor("op_5481"), val = tensor(-0x1p-1)]; + tensor var_5482 = mul(x = cv_189, y = var_5481)[name = tensor("op_5482")]; + tensor var_5483 = const()[name = tensor("op_5483"), val = tensor(0x1p-1)]; + tensor var_5484 = add(x = var_5482, y = var_5483)[name = tensor("op_5484")]; + tensor var_5485 = const()[name = tensor("op_5485"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54809600)))]; + tensor var_5488 = mul(x = var_5484, y = var_5485)[name = tensor("op_5488")]; + tensor input_463 = add(x = xt_283, y = var_5488)[name = tensor("input_463")]; + tensor weight_233_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54809792))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54821120))), name = tensor("weight_233_palettized"), shape = tensor([32, 32, 11])]; + tensor input_465_pad_type_0 = const()[name = tensor("input_465_pad_type_0"), val = tensor("custom")]; + tensor input_465_pad_0 = const()[name = tensor("input_465_pad_0"), val = tensor([25, 25])]; + tensor input_465_dilations_0 = const()[name = tensor("input_465_dilations_0"), val = tensor([5])]; + tensor input_465_strides_0 = const()[name = tensor("input_465_strides_0"), val = tensor([1])]; + tensor input_465_groups_0 = const()[name = tensor("input_465_groups_0"), val = tensor(1)]; + tensor input_465 = conv(bias = generator_resblocks_11_convs1_2_bias, dilations = input_465_dilations_0, groups = input_465_groups_0, pad = input_465_pad_0, pad_type = input_465_pad_type_0, strides = input_465_strides_0, weight = weight_233_palettized, x = input_463)[name = tensor("input_465")]; + tensor h_421 = linear(bias = generator_resblocks_11_adain2_2_fc_bias, weight = generator_resblocks_11_adain2_2_fc_weight_palettized, x = s)[name = tensor("linear_106")]; + tensor var_5503 = const()[name = tensor("op_5503"), val = tensor([1, 64, 1])]; + tensor h = reshape(shape = var_5503, x = h_421)[name = tensor("h")]; + tensor var_5505_split_sizes_0 = const()[name = tensor("op_5505_split_sizes_0"), val = tensor([32, 32])]; + tensor var_5505_axis_0 = const()[name = tensor("op_5505_axis_0"), val = tensor(1)]; + tensor var_5505_0, tensor var_5505_1 = split(axis = var_5505_axis_0, split_sizes = var_5505_split_sizes_0, x = h)[name = tensor("op_5505")]; + tensor var_5507_promoted = const()[name = tensor("op_5507_promoted"), val = tensor(0x1p+0)]; + tensor var_5508 = add(x = var_5505_0, y = var_5507_promoted)[name = tensor("op_5508")]; + tensor var_5509 = instance_norm(epsilon = var_5244, x = input_465)[name = tensor("op_5509")]; + tensor var_5510 = mul(x = var_5508, y = var_5509)[name = tensor("op_5510")]; + tensor xt_285 = add(x = var_5510, y = var_5505_1)[name = tensor("xt_285")]; + tensor var_5513 = const()[name = tensor("op_5513"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54822208)))]; + tensor var_5514 = mul(x = xt_285, y = var_5513)[name = tensor("op_5514")]; + tensor cv = cos(x = var_5514)[name = tensor("cv")]; + tensor var_5516 = const()[name = tensor("op_5516"), val = tensor(-0x1p-1)]; + tensor var_5517 = mul(x = cv, y = var_5516)[name = tensor("op_5517")]; + tensor var_5518 = const()[name = tensor("op_5518"), val = tensor(0x1p-1)]; + tensor var_5519 = add(x = var_5517, y = var_5518)[name = tensor("op_5519")]; + tensor var_5520 = const()[name = tensor("op_5520"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54822400)))]; + tensor var_5523 = mul(x = var_5519, y = var_5520)[name = tensor("op_5523")]; + tensor input_467 = add(x = xt_285, y = var_5523)[name = tensor("input_467")]; + tensor weight_235_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54822592))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54833920))), name = tensor("weight_235_palettized"), shape = tensor([32, 32, 11])]; + tensor xt_pad_type_0 = const()[name = tensor("xt_pad_type_0"), val = tensor("custom")]; + tensor xt_pad_0 = const()[name = tensor("xt_pad_0"), val = tensor([5, 5])]; + tensor xt_strides_0 = const()[name = tensor("xt_strides_0"), val = tensor([1])]; + tensor xt_dilations_0 = const()[name = tensor("xt_dilations_0"), val = tensor([1])]; + tensor xt_groups_0 = const()[name = tensor("xt_groups_0"), val = tensor(1)]; + tensor xt = conv(bias = generator_resblocks_11_convs2_2_bias, dilations = xt_dilations_0, groups = xt_groups_0, pad = xt_pad_0, pad_type = xt_pad_type_0, strides = xt_strides_0, weight = weight_235_palettized, x = input_467)[name = tensor("xt")]; + tensor var_5532 = add(x = xt, y = input_461)[name = tensor("op_5532")]; + tensor xs = add(x = xs_21, y = var_5532)[name = tensor("xs")]; + tensor _inversed_z_y_0 = const()[name = tensor("_inversed_z_y_0"), val = tensor(0x1.555556p-2)]; + tensor _inversed_z = mul(x = xs, y = _inversed_z_y_0)[name = tensor("_inversed_z")]; + tensor var_5537 = mul(x = _inversed_z, y = generator_alphas_4)[name = tensor("op_5537")]; + tensor sv = sin(x = var_5537)[name = tensor("sv")]; + tensor var_5539 = mul(x = sv, y = sv)[name = tensor("op_5539")]; + tensor _inversed_5540_y_0 = const()[name = tensor("_inversed_5540_y_0"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54835008)))]; + tensor _inversed_5540 = mul(x = var_5539, y = _inversed_5540_y_0)[name = tensor("_inversed_5540")]; + tensor input = add(x = _inversed_z, y = _inversed_5540)[name = tensor("input")]; + tensor weight_237 = const()[name = tensor("weight_237"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(54835200)))]; + tensor x_17_pad_type_0 = const()[name = tensor("x_17_pad_type_0"), val = tensor("custom")]; + tensor x_17_pad_0 = const()[name = tensor("x_17_pad_0"), val = tensor([3, 3])]; + tensor x_17_strides_0 = const()[name = tensor("x_17_strides_0"), val = tensor([1])]; + tensor x_17_dilations_0 = const()[name = tensor("x_17_dilations_0"), val = tensor([1])]; + tensor x_17_groups_0 = const()[name = tensor("x_17_groups_0"), val = tensor(1)]; + tensor x_17 = conv(bias = generator_conv_post_bias, dilations = x_17_dilations_0, groups = x_17_groups_0, pad = x_17_pad_0, pad_type = x_17_pad_type_0, strides = x_17_strides_0, weight = weight_237, x = input)[name = tensor("x_17")]; + tensor x = tanh(x = x_17)[name = tensor("x")]; + tensor var_5558_axes_0 = const()[name = tensor("op_5558_axes_0"), val = tensor([1])]; + tensor audio = squeeze(axes = var_5558_axes_0, x = x)[name = tensor("op_5558")]; + } -> (audio); +} \ No newline at end of file diff --git a/ANE/styletts2_ane_vocoder.mlmodelc/weights/weight.bin b/ANE/styletts2_ane_vocoder.mlmodelc/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..0169da027c88a4467c686247d02c88ff81c37fc9 --- /dev/null +++ b/ANE/styletts2_ane_vocoder.mlmodelc/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe39faf7c2a8e3b75f2db31db61a246d1e858ddd56d80340707bc08924821664 +size 54836160 diff --git a/ANE/styletts2_ane_vocoder.mlpackage/Data/com.apple.CoreML/model.mlmodel b/ANE/styletts2_ane_vocoder.mlpackage/Data/com.apple.CoreML/model.mlmodel new file mode 100644 index 0000000000000000000000000000000000000000..acfe9ca8377e914fd34395ab910e767f90978296 --- /dev/null +++ b/ANE/styletts2_ane_vocoder.mlpackage/Data/com.apple.CoreML/model.mlmodel @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ab14ee0e1032fb3786350951c28eccbd3271b0e36e413c8f30a1ff355d6e203f +size 509403 diff --git a/ANE/styletts2_ane_vocoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin b/ANE/styletts2_ane_vocoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin new file mode 100644 index 0000000000000000000000000000000000000000..0169da027c88a4467c686247d02c88ff81c37fc9 --- /dev/null +++ b/ANE/styletts2_ane_vocoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe39faf7c2a8e3b75f2db31db61a246d1e858ddd56d80340707bc08924821664 +size 54836160 diff --git a/ANE/styletts2_ane_vocoder.mlpackage/Manifest.json b/ANE/styletts2_ane_vocoder.mlpackage/Manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..764f3cc2965cada5ce2c612ec374733b59669df6 --- /dev/null +++ b/ANE/styletts2_ane_vocoder.mlpackage/Manifest.json @@ -0,0 +1,18 @@ +{ + "fileFormatVersion": "1.0.0", + "itemInfoEntries": { + "B655CB15-D49A-404B-A0B1-157B99849FE8": { + "author": "com.apple.CoreML", + "description": "CoreML Model Specification", + "name": "model.mlmodel", + "path": "com.apple.CoreML/model.mlmodel" + }, + "D818B36F-C41B-41D8-A441-91A402EF4E22": { + "author": "com.apple.CoreML", + "description": "CoreML Model Weights", + "name": "weights", + "path": "com.apple.CoreML/weights" + } + }, + "rootModelIdentifier": "B655CB15-D49A-404B-A0B1-157B99849FE8" +} diff --git a/voices/ref_s_1221-135767-0014.bin b/voices/ref_s_1221-135767-0014.bin new file mode 100644 index 0000000000000000000000000000000000000000..1b3ed94876b7664c7f7b394a7f80d418ae216879 --- /dev/null +++ b/voices/ref_s_1221-135767-0014.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c8f34b18352203a90acc1620a8e6564cd76a76efaa3ca9f76c2e7f981925195 +size 1024 diff --git a/voices/ref_s_1789_142896_000022_000005.bin b/voices/ref_s_1789_142896_000022_000005.bin new file mode 100644 index 0000000000000000000000000000000000000000..b7e96934309bc4028c0f288ef4fc32da3630dc8f --- /dev/null +++ b/voices/ref_s_1789_142896_000022_000005.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c0ae950fe25ce330dec76e2a312c7e9f08b3770f0c242ec39658c576173b2c53 +size 1024 diff --git a/voices/ref_s_3.bin b/voices/ref_s_3.bin new file mode 100644 index 0000000000000000000000000000000000000000..d56fcce96a651c47d90ede41647e2c319e6d5ee0 --- /dev/null +++ b/voices/ref_s_3.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f5920cc5023efd2d0e65bfb1ce8a3ccaf505d00287401182685622cec62912a +size 1024 diff --git a/voices/ref_s_4.bin b/voices/ref_s_4.bin new file mode 100644 index 0000000000000000000000000000000000000000..a3d8c7fb80048f467310658c3b819e453c0cc9ed --- /dev/null +++ b/voices/ref_s_4.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87b0125f709791b738f176f7d3d66db1278d12fd092ff71ce7603955907e70a7 +size 1024 diff --git a/voices/ref_s_4077-13754-0000.bin b/voices/ref_s_4077-13754-0000.bin new file mode 100644 index 0000000000000000000000000000000000000000..e6d480e1e68c15252281e8d4a4d018ee96e5c69c --- /dev/null +++ b/voices/ref_s_4077-13754-0000.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1bc6b4684b7dd023ab557014cf589931df7b6d6eb0fb45b2c9109e5a0b07b251 +size 1024 diff --git a/voices/ref_s_5.bin b/voices/ref_s_5.bin new file mode 100644 index 0000000000000000000000000000000000000000..c870fe7c839a0912737aa737b7f957202c670fd5 --- /dev/null +++ b/voices/ref_s_5.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:09ac2a9de005df565e72236e4afc6a46facf5e0e9eaf143ca62b1d14e0edd5b3 +size 1024 diff --git a/voices/ref_s_5639-40744-0020.bin b/voices/ref_s_5639-40744-0020.bin new file mode 100644 index 0000000000000000000000000000000000000000..cc60a2ac00e8cba2f9219472e130001308164818 --- /dev/null +++ b/voices/ref_s_5639-40744-0020.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc121e3357ca8e00bd5ef42f48c5b57c441ae0497aede091654ffae9a7fad93a +size 1024 diff --git a/voices/ref_s_696_92939_000016_000006.bin b/voices/ref_s_696_92939_000016_000006.bin new file mode 100644 index 0000000000000000000000000000000000000000..6c1ac242a1de47b69295d3eddc2ca1dd085fa211 --- /dev/null +++ b/voices/ref_s_696_92939_000016_000006.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44a37a22ab2ffb1af297c9b1714e740ae9c481b3ee20afcac4e9fa8ed82185f6 +size 1024 diff --git a/voices/ref_s_908-157963-0027.bin b/voices/ref_s_908-157963-0027.bin new file mode 100644 index 0000000000000000000000000000000000000000..d5c23b695ca9d0d029e613ae1a957b06ceef4af3 --- /dev/null +++ b/voices/ref_s_908-157963-0027.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04026b56d755b3511f40c715465409a0f3fbbd3372de983db5b1b46a532dfc43 +size 1024 diff --git a/voices/ref_s_Gavin.bin b/voices/ref_s_Gavin.bin new file mode 100644 index 0000000000000000000000000000000000000000..9e7975efd76fefaf4d67e804e91fdbbf94f5c05d --- /dev/null +++ b/voices/ref_s_Gavin.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2cd47901c6f8f83da48049fcf376480eb6c63b43e6f8359072f66c8fd6c3a9de +size 1024 diff --git a/voices/ref_s_Nima.bin b/voices/ref_s_Nima.bin new file mode 100644 index 0000000000000000000000000000000000000000..5199d43a58e795ef1a4a000b7b09c75a5854ecaf --- /dev/null +++ b/voices/ref_s_Nima.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bc32e61a64bb8f5031d619902e60cf99113bd6d90abafbd3962b30c45a9aa1d4 +size 1024 diff --git a/voices/ref_s_Vinay.bin b/voices/ref_s_Vinay.bin new file mode 100644 index 0000000000000000000000000000000000000000..d04381552ce0f10753375663adb5d39c1fd92f69 --- /dev/null +++ b/voices/ref_s_Vinay.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3c87703b18df2959e45e1df350e14640fa4be67397f05a716e80c80a8b8f2a13 +size 1024 diff --git a/voices/ref_s_Yinghao.bin b/voices/ref_s_Yinghao.bin new file mode 100644 index 0000000000000000000000000000000000000000..e035971a8e7ff48d80945d6e5619a25ee0e376c8 --- /dev/null +++ b/voices/ref_s_Yinghao.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b227bd8bb8e359150e43acef549720e7a8820c3afffe713c97707020e49e491b +size 1024 diff --git a/voices/ref_s_amused.bin b/voices/ref_s_amused.bin new file mode 100644 index 0000000000000000000000000000000000000000..957e5d1f6bb64ac552e32649152152a894b8e297 --- /dev/null +++ b/voices/ref_s_amused.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:58752cec26e778fc643d2b6b425ac7f6b88f466dfc7c90d5e27146c3e5aef03c +size 1024 diff --git a/voices/ref_s_anger.bin b/voices/ref_s_anger.bin new file mode 100644 index 0000000000000000000000000000000000000000..33e64d543fc9d1cc3ad1adfc9c9b90e50d76a49c --- /dev/null +++ b/voices/ref_s_anger.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bcda68be1f0016b0e8603cc804dfd18af525f533c398aef8221167688190722c +size 1024 diff --git a/voices/ref_s_disgusted.bin b/voices/ref_s_disgusted.bin new file mode 100644 index 0000000000000000000000000000000000000000..152826a96a30fe9a9455d2873a228dbe4d3504e2 --- /dev/null +++ b/voices/ref_s_disgusted.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:865912a9750e5d968b5c08606ea8beaf77f26f0f90a4a8ebf4ad105bf4596e97 +size 1024 diff --git a/voices/ref_s_sleepy.bin b/voices/ref_s_sleepy.bin new file mode 100644 index 0000000000000000000000000000000000000000..bb45b1d2b1149d02275062b43296a2022ea5c242 --- /dev/null +++ b/voices/ref_s_sleepy.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb9e60a4fb5a9fd5a9e615bbef0b9f60d52ed0be3f23d1e0c8f56e8fb7f751f2 +size 1024