Hugging Face's logo

schismaudio
/
deepfilternet3

Core ML
audio
music-source-separation
schism
Model card Files
xet
Community
zkeown commited on
Commit
f6d9ccf
·
verified ·
1 Parent(s): 3a81232

Add DeepFilterNet3_ERBDecoder.mlmodelc

Browse files
Files changed (5) hide show
  1. v1.0.0/DeepFilterNet3_ERBDecoder.mlmodelc/analytics/coremldata.bin +3 -0
  2. v1.0.0/DeepFilterNet3_ERBDecoder.mlmodelc/coremldata.bin +3 -0
  3. v1.0.0/DeepFilterNet3_ERBDecoder.mlmodelc/metadata.json +122 -0
  4. v1.0.0/DeepFilterNet3_ERBDecoder.mlmodelc/model.mil +295 -0
  5. v1.0.0/DeepFilterNet3_ERBDecoder.mlmodelc/weights/weight.bin +3 -0
v1.0.0/DeepFilterNet3_ERBDecoder.mlmodelc/analytics/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91d1f93d546ac2b81d7e389eebd2935e7f946f577eb3ef1d2f8e08dfbf962322
3
+ size 243
v1.0.0/DeepFilterNet3_ERBDecoder.mlmodelc/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c5a9ff38afaf9520cdc96e0052f48571b138015d40a306e28bb591ca10af1fc
3
+ size 454
v1.0.0/DeepFilterNet3_ERBDecoder.mlmodelc/metadata.json ADDED
@@ -0,0 +1,122 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "metadataOutputVersion" : "3.0",
4
+ "storagePrecision" : "Float32",
5
+ "outputSchema" : [
6
+ {
7
+ "hasShapeFlexibility" : "0",
8
+ "isOptional" : "0",
9
+ "dataType" : "Float32",
10
+ "formattedType" : "MultiArray (Float32 1 × 1 × 100 × 32)",
11
+ "shortDescription" : "",
12
+ "shape" : "[1, 1, 100, 32]",
13
+ "name" : "erb_gains",
14
+ "type" : "MultiArray"
15
+ }
16
+ ],
17
+ "modelParameters" : [
18
+
19
+ ],
20
+ "specificationVersion" : 8,
21
+ "mlProgramOperationTypeHistogram" : {
22
+ "Ios17.reshape" : 9,
23
+ "Ios17.scatter" : 2,
24
+ "Ios17.matmul" : 2,
25
+ "Ios17.transpose" : 7,
26
+ "Crop" : 2,
27
+ "Select" : 4,
28
+ "Ios17.expandDims" : 2,
29
+ "Ios17.add" : 18,
30
+ "Ios16.sigmoid" : 4,
31
+ "Ios17.sliceByIndex" : 2,
32
+ "Ios17.convTranspose" : 2,
33
+ "Ios17.squeeze" : 4,
34
+ "Ios17.gather" : 4,
35
+ "WhileLoop" : 2,
36
+ "Ios17.less" : 2,
37
+ "Ios17.sub" : 2,
38
+ "Ios17.conv" : 9,
39
+ "Ios16.relu" : 9,
40
+ "Ios17.tanh" : 2,
41
+ "Ios17.linear" : 12,
42
+ "Ios17.greaterEqual" : 2,
43
+ "Ios17.mul" : 6
44
+ },
45
+ "computePrecision" : "Mixed (Float32, Int32)",
46
+ "isUpdatable" : "0",
47
+ "stateSchema" : [
48
+
49
+ ],
50
+ "availability" : {
51
+ "macOS" : "14.0",
52
+ "tvOS" : "17.0",
53
+ "visionOS" : "1.0",
54
+ "watchOS" : "10.0",
55
+ "iOS" : "17.0",
56
+ "macCatalyst" : "17.0"
57
+ },
58
+ "modelType" : {
59
+ "name" : "MLModelType_mlProgram"
60
+ },
61
+ "userDefinedMetadata" : {
62
+ "com.github.apple.coremltools.conversion_date" : "2026-02-12",
63
+ "com.github.apple.coremltools.source" : "torch==2.10.0",
64
+ "com.github.apple.coremltools.version" : "9.0",
65
+ "com.github.apple.coremltools.source_dialect" : "TorchScript"
66
+ },
67
+ "inputSchema" : [
68
+ {
69
+ "hasShapeFlexibility" : "0",
70
+ "isOptional" : "0",
71
+ "dataType" : "Float32",
72
+ "formattedType" : "MultiArray (Float32 1 × 100 × 512)",
73
+ "shortDescription" : "",
74
+ "shape" : "[1, 100, 512]",
75
+ "name" : "emb",
76
+ "type" : "MultiArray"
77
+ },
78
+ {
79
+ "hasShapeFlexibility" : "0",
80
+ "isOptional" : "0",
81
+ "dataType" : "Float32",
82
+ "formattedType" : "MultiArray (Float32 1 × 64 × 100 × 8)",
83
+ "shortDescription" : "",
84
+ "shape" : "[1, 64, 100, 8]",
85
+ "name" : "e3",
86
+ "type" : "MultiArray"
87
+ },
88
+ {
89
+ "hasShapeFlexibility" : "0",
90
+ "isOptional" : "0",
91
+ "dataType" : "Float32",
92
+ "formattedType" : "MultiArray (Float32 1 × 64 × 100 × 8)",
93
+ "shortDescription" : "",
94
+ "shape" : "[1, 64, 100, 8]",
95
+ "name" : "e2",
96
+ "type" : "MultiArray"
97
+ },
98
+ {
99
+ "hasShapeFlexibility" : "0",
100
+ "isOptional" : "0",
101
+ "dataType" : "Float32",
102
+ "formattedType" : "MultiArray (Float32 1 × 64 × 100 × 16)",
103
+ "shortDescription" : "",
104
+ "shape" : "[1, 64, 100, 16]",
105
+ "name" : "e1",
106
+ "type" : "MultiArray"
107
+ },
108
+ {
109
+ "hasShapeFlexibility" : "0",
110
+ "isOptional" : "0",
111
+ "dataType" : "Float32",
112
+ "formattedType" : "MultiArray (Float32 1 × 64 × 100 × 32)",
113
+ "shortDescription" : "",
114
+ "shape" : "[1, 64, 100, 32]",
115
+ "name" : "e0",
116
+ "type" : "MultiArray"
117
+ }
118
+ ],
119
+ "generatedClassName" : "DeepFilterNet3_ERBDecoder",
120
+ "method" : "predict"
121
+ }
122
+ ]
v1.0.0/DeepFilterNet3_ERBDecoder.mlmodelc/model.mil ADDED
@@ -0,0 +1,295 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ program(1.0)
2
+ [buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3510.2.1"}, {"coremlc-version", "3500.32.1"}, {"coremltools-component-torch", "2.10.0"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "9.0"}})]
3
+ {
4
+ func main<ios17>(tensor<fp32, [1, 64, 100, 32]> e0, tensor<fp32, [1, 64, 100, 16]> e1, tensor<fp32, [1, 64, 100, 8]> e2, tensor<fp32, [1, 64, 100, 8]> e3, tensor<fp32, [1, 100, 512]> emb) {
5
+ tensor<fp32, [16, 32, 16]> decoder_emb_gru_linear_in_0_weight = const()[name = tensor<string, []>("decoder_emb_gru_linear_in_0_weight"), val = tensor<fp32, [16, 32, 16]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(64)))];
6
+ tensor<fp32, [16, 16, 32]> decoder_emb_gru_linear_out_0_weight = const()[name = tensor<string, []>("decoder_emb_gru_linear_out_0_weight"), val = tensor<fp32, [16, 16, 32]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(32896)))];
7
+ tensor<fp32, [64, 1, 1, 3]> decoder_convt3_0_weight = const()[name = tensor<string, []>("decoder_convt3_0_weight"), val = tensor<fp32, [64, 1, 1, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(65728)))];
8
+ tensor<fp32, [64, 1, 1, 3]> decoder_convt2_0_weight = const()[name = tensor<string, []>("decoder_convt2_0_weight"), val = tensor<fp32, [64, 1, 1, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(66560)))];
9
+ tensor<fp32, [64, 1, 1, 3]> decoder_convt1_0_weight = const()[name = tensor<string, []>("decoder_convt1_0_weight"), val = tensor<fp32, [64, 1, 1, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(67392)))];
10
+ tensor<int32, [4]> var_63 = const()[name = tensor<string, []>("op_63"), val = tensor<int32, [4]>([1, 100, 16, 32])];
11
+ tensor<fp32, [1, 100, 16, 32]> var_64 = reshape(shape = var_63, x = emb)[name = tensor<string, []>("op_64")];
12
+ tensor<int32, [4]> transpose_0_perm_0 = const()[name = tensor<string, []>("transpose_0_perm_0"), val = tensor<int32, [4]>([2, 0, 1, 3])];
13
+ tensor<int32, [3]> concat_4 = const()[name = tensor<string, []>("concat_4"), val = tensor<int32, [3]>([16, 100, 32])];
14
+ tensor<fp32, [16, 1, 100, 32]> transpose_0 = transpose(perm = transpose_0_perm_0, x = var_64)[name = tensor<string, []>("transpose_11")];
15
+ tensor<fp32, [16, 100, 32]> reshape_0 = reshape(shape = concat_4, x = transpose_0)[name = tensor<string, []>("reshape_0")];
16
+ tensor<bool, []> matmul_0_transpose_x_0 = const()[name = tensor<string, []>("matmul_0_transpose_x_0"), val = tensor<bool, []>(false)];
17
+ tensor<bool, []> matmul_0_transpose_y_0 = const()[name = tensor<string, []>("matmul_0_transpose_y_0"), val = tensor<bool, []>(false)];
18
+ tensor<fp32, [16, 100, 16]> matmul_0 = matmul(transpose_x = matmul_0_transpose_x_0, transpose_y = matmul_0_transpose_y_0, x = reshape_0, y = decoder_emb_gru_linear_in_0_weight)[name = tensor<string, []>("matmul_0")];
19
+ tensor<int32, [4]> concat_9 = const()[name = tensor<string, []>("concat_9"), val = tensor<int32, [4]>([16, 1, 100, 16])];
20
+ tensor<fp32, [16, 1, 100, 16]> reshape_2 = reshape(shape = concat_9, x = matmul_0)[name = tensor<string, []>("reshape_2")];
21
+ tensor<int32, [4]> x_1_perm_0 = const()[name = tensor<string, []>("x_1_perm_0"), val = tensor<int32, [4]>([1, 2, 0, 3])];
22
+ tensor<int32, [3]> concat_10 = const()[name = tensor<string, []>("concat_10"), val = tensor<int32, [3]>([1, 100, 256])];
23
+ tensor<fp32, [1, 100, 16, 16]> x_1 = transpose(perm = x_1_perm_0, x = reshape_2)[name = tensor<string, []>("transpose_10")];
24
+ tensor<fp32, [1, 100, 256]> input_1 = reshape(shape = concat_10, x = x_1)[name = tensor<string, []>("input_1")];
25
+ tensor<fp32, [1, 100, 256]> input_3 = relu(x = input_1)[name = tensor<string, []>("input_3")];
26
+ tensor<int32, [3]> transpose_2_perm_0 = const()[name = tensor<string, []>("transpose_2_perm_0"), val = tensor<int32, [3]>([1, 0, 2])];
27
+ tensor<int32, [1]> slice_by_index_10 = const()[name = tensor<string, []>("slice_by_index_10"), val = tensor<int32, [1]>([100])];
28
+ tensor<fp32, [101, 1, 256]> concat_12 = const()[name = tensor<string, []>("concat_12"), val = tensor<fp32, [101, 1, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(68224)))];
29
+ tensor<int32, [1]> while_loop_0_loop_vars0_0 = const()[name = tensor<string, []>("while_loop_0_loop_vars0_0"), val = tensor<int32, [1]>([0])];
30
+ tensor<fp32, [100, 1, 256]> transpose_2 = transpose(perm = transpose_2_perm_0, x = input_3)[name = tensor<string, []>("transpose_9")];
31
+ tensor<int32, [1]> while_loop_0_0, tensor<fp32, [101, 1, 256]> while_loop_0_1 = while_loop(loop_vars = (while_loop_0_loop_vars0_0, concat_12))[name = tensor<string, []>("while_loop_0")]
32
+ (tensor<int32, [1]> while_loop_0_loop_vars0_0_x0_1_1_1_0, tensor<fp32, [101, 1, 256]> concat_12_x0_1_1_1_0) {
33
+ tensor<bool, [1]> less_1 = less(x = while_loop_0_loop_vars0_0_x0_1_1_1_0, y = slice_by_index_10)[name = tensor<string, []>("less_1")];
34
+ } -> (less_1)
35
+ (tensor<int32, [1]> while_loop_0_loop_vars0_0_x0_1_1_1_1, tensor<fp32, [101, 1, 256]> concat_12_x0_1_1_1_1) {
36
+ tensor<int32, []> gather_2_batch_dims_0 = const()[name = tensor<string, []>("gather_2_batch_dims_0"), val = tensor<int32, []>(0)];
37
+ tensor<bool, []> gather_2_validate_indices_0 = const()[name = tensor<string, []>("gather_2_validate_indices_0"), val = tensor<bool, []>(false)];
38
+ tensor<int32, []> greater_equal_0_y_0 = const()[name = tensor<string, []>("greater_equal_0_y_0"), val = tensor<int32, []>(0)];
39
+ tensor<bool, [1]> greater_equal_0 = greater_equal(x = while_loop_0_loop_vars0_0_x0_1_1_1_1, y = greater_equal_0_y_0)[name = tensor<string, []>("greater_equal_0")];
40
+ tensor<int32, []> slice_by_index_26 = const()[name = tensor<string, []>("slice_by_index_26"), val = tensor<int32, []>(100)];
41
+ tensor<int32, [1]> add_20 = add(x = while_loop_0_loop_vars0_0_x0_1_1_1_1, y = slice_by_index_26)[name = tensor<string, []>("add_20")];
42
+ tensor<int32, [1]> select_0 = select(a = while_loop_0_loop_vars0_0_x0_1_1_1_1, b = add_20, cond = greater_equal_0)[name = tensor<string, []>("select_0")];
43
+ tensor<int32, []> gather_2_axis_1 = const()[name = tensor<string, []>("gather_2_axis_1"), val = tensor<int32, []>(0)];
44
+ tensor<fp32, [1, 1, 256]> gather_2 = gather(axis = gather_2_axis_1, batch_dims = gather_2_batch_dims_0, indices = select_0, validate_indices = gather_2_validate_indices_0, x = transpose_2)[name = tensor<string, []>("gather_2")];
45
+ tensor<int32, []> gather_3_batch_dims_0 = const()[name = tensor<string, []>("gather_3_batch_dims_0"), val = tensor<int32, []>(0)];
46
+ tensor<bool, []> gather_3_validate_indices_0 = const()[name = tensor<string, []>("gather_3_validate_indices_0"), val = tensor<bool, []>(false)];
47
+ tensor<int32, []> slice_by_index_27 = const()[name = tensor<string, []>("slice_by_index_27"), val = tensor<int32, []>(101)];
48
+ tensor<int32, [1]> add_21 = add(x = while_loop_0_loop_vars0_0_x0_1_1_1_1, y = slice_by_index_27)[name = tensor<string, []>("add_21")];
49
+ tensor<int32, [1]> select_1 = select(a = while_loop_0_loop_vars0_0_x0_1_1_1_1, b = add_21, cond = greater_equal_0)[name = tensor<string, []>("select_1")];
50
+ tensor<int32, []> gather_3_axis_1 = const()[name = tensor<string, []>("gather_3_axis_1"), val = tensor<int32, []>(0)];
51
+ tensor<fp32, [1, 1, 256]> gather_3 = gather(axis = gather_3_axis_1, batch_dims = gather_3_batch_dims_0, indices = select_1, validate_indices = gather_3_validate_indices_0, x = concat_12_x0_1_1_1_1)[name = tensor<string, []>("gather_3")];
52
+ tensor<int32, [1]> squeeze_2_axes_0 = const()[name = tensor<string, []>("squeeze_2_axes_0"), val = tensor<int32, [1]>([0])];
53
+ tensor<fp32, [1, 256]> squeeze_2 = squeeze(axes = squeeze_2_axes_0, x = gather_2)[name = tensor<string, []>("squeeze_2")];
54
+ tensor<int32, [1]> squeeze_3_axes_0 = const()[name = tensor<string, []>("squeeze_3_axes_0"), val = tensor<int32, [1]>([0])];
55
+ tensor<fp32, [1, 256]> squeeze_3 = squeeze(axes = squeeze_3_axes_0, x = gather_3)[name = tensor<string, []>("squeeze_3")];
56
+ tensor<fp32, [256, 256]> linear_6_weight_0 = const()[name = tensor<string, []>("linear_6_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(171712)))];
57
+ tensor<fp32, [256]> linear_6_bias_0 = const()[name = tensor<string, []>("linear_6_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(433920)))];
58
+ tensor<fp32, [1, 256]> linear_6 = linear(bias = linear_6_bias_0, weight = linear_6_weight_0, x = squeeze_2)[name = tensor<string, []>("linear_6")];
59
+ tensor<fp32, [256, 256]> linear_7_weight_0 = const()[name = tensor<string, []>("linear_7_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(435008)))];
60
+ tensor<fp32, [256]> linear_7_bias_0 = const()[name = tensor<string, []>("linear_7_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(697216)))];
61
+ tensor<fp32, [1, 256]> linear_7 = linear(bias = linear_7_bias_0, weight = linear_7_weight_0, x = squeeze_3)[name = tensor<string, []>("linear_7")];
62
+ tensor<fp32, [1, 256]> add_5 = add(x = linear_6, y = linear_7)[name = tensor<string, []>("add_5")];
63
+ tensor<fp32, [1, 256]> sigmoid_2 = sigmoid(x = add_5)[name = tensor<string, []>("sigmoid_2")];
64
+ tensor<fp32, [256, 256]> linear_8_weight_0 = const()[name = tensor<string, []>("linear_8_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(698304)))];
65
+ tensor<fp32, [256]> linear_8_bias_0 = const()[name = tensor<string, []>("linear_8_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(960512)))];
66
+ tensor<fp32, [1, 256]> linear_8 = linear(bias = linear_8_bias_0, weight = linear_8_weight_0, x = squeeze_2)[name = tensor<string, []>("linear_8")];
67
+ tensor<fp32, [256, 256]> linear_9_weight_0 = const()[name = tensor<string, []>("linear_9_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(961600)))];
68
+ tensor<fp32, [256]> linear_9_bias_0 = const()[name = tensor<string, []>("linear_9_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1223808)))];
69
+ tensor<fp32, [1, 256]> linear_9 = linear(bias = linear_9_bias_0, weight = linear_9_weight_0, x = squeeze_3)[name = tensor<string, []>("linear_9")];
70
+ tensor<fp32, [1, 256]> add_6 = add(x = linear_8, y = linear_9)[name = tensor<string, []>("add_6")];
71
+ tensor<fp32, [1, 256]> sigmoid_3 = sigmoid(x = add_6)[name = tensor<string, []>("sigmoid_3")];
72
+ tensor<fp32, [256, 256]> linear_10_weight_0 = const()[name = tensor<string, []>("linear_10_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1224896)))];
73
+ tensor<fp32, [256]> linear_10_bias_0 = const()[name = tensor<string, []>("linear_10_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1487104)))];
74
+ tensor<fp32, [1, 256]> linear_10 = linear(bias = linear_10_bias_0, weight = linear_10_weight_0, x = squeeze_2)[name = tensor<string, []>("linear_10")];
75
+ tensor<fp32, [256, 256]> linear_11_weight_0 = const()[name = tensor<string, []>("linear_11_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1488192)))];
76
+ tensor<fp32, [256]> linear_11_bias_0 = const()[name = tensor<string, []>("linear_11_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1750400)))];
77
+ tensor<fp32, [1, 256]> linear_11 = linear(bias = linear_11_bias_0, weight = linear_11_weight_0, x = squeeze_3)[name = tensor<string, []>("linear_11")];
78
+ tensor<fp32, [1, 256]> mul_3 = mul(x = sigmoid_2, y = linear_11)[name = tensor<string, []>("mul_3")];
79
+ tensor<fp32, [1, 256]> add_7 = add(x = linear_10, y = mul_3)[name = tensor<string, []>("add_7")];
80
+ tensor<fp32, [1, 256]> tanh_1 = tanh(x = add_7)[name = tensor<string, []>("tanh_1")];
81
+ tensor<fp32, []> sub_1_x_0 = const()[name = tensor<string, []>("sub_1_x_0"), val = tensor<fp32, []>(0x1p+0)];
82
+ tensor<fp32, [1, 256]> sub_1 = sub(x = sub_1_x_0, y = sigmoid_3)[name = tensor<string, []>("sub_1")];
83
+ tensor<fp32, [1, 256]> mul_4 = mul(x = sub_1, y = tanh_1)[name = tensor<string, []>("mul_4")];
84
+ tensor<fp32, [1, 256]> mul_5 = mul(x = sigmoid_3, y = squeeze_3)[name = tensor<string, []>("mul_5")];
85
+ tensor<fp32, [1, 256]> add_8 = add(x = mul_4, y = mul_5)[name = tensor<string, []>("add_8")];
86
+ tensor<int32, []> add_9_y_0 = const()[name = tensor<string, []>("add_9_y_0"), val = tensor<int32, []>(1)];
87
+ tensor<int32, [1]> add_9 = add(x = while_loop_0_loop_vars0_0_x0_1_1_1_1, y = add_9_y_0)[name = tensor<string, []>("add_9")];
88
+ tensor<int32, [1]> expand_dims_1_axes_0 = const()[name = tensor<string, []>("expand_dims_1_axes_0"), val = tensor<int32, [1]>([0])];
89
+ tensor<fp32, [1, 1, 256]> expand_dims_1 = expand_dims(axes = expand_dims_1_axes_0, x = add_8)[name = tensor<string, []>("expand_dims_1")];
90
+ tensor<int32, []> scatter_1_axis_0 = const()[name = tensor<string, []>("scatter_1_axis_0"), val = tensor<int32, []>(0)];
91
+ tensor<string, []> scatter_1_mode_0 = const()[name = tensor<string, []>("scatter_1_mode_0"), val = tensor<string, []>("add")];
92
+ tensor<bool, []> scatter_1_validate_indices_0 = const()[name = tensor<string, []>("scatter_1_validate_indices_0"), val = tensor<bool, []>(false)];
93
+ tensor<fp32, [101, 1, 256]> scatter_1 = scatter(axis = scatter_1_axis_0, data = concat_12_x0_1_1_1_1, indices = add_9, mode = scatter_1_mode_0, updates = expand_dims_1, validate_indices = scatter_1_validate_indices_0)[name = tensor<string, []>("scatter_1")];
94
+ } -> (add_9, scatter_1);
95
+ tensor<int32, [3]> x_3_layer_0_tmp_begin_0 = const()[name = tensor<string, []>("x_3_layer_0_tmp_begin_0"), val = tensor<int32, [3]>([1, 0, 0])];
96
+ tensor<int32, [3]> x_3_layer_0_tmp_end_0 = const()[name = tensor<string, []>("x_3_layer_0_tmp_end_0"), val = tensor<int32, [3]>([0, 0, 0])];
97
+ tensor<bool, [3]> x_3_layer_0_tmp_begin_mask_0 = const()[name = tensor<string, []>("x_3_layer_0_tmp_begin_mask_0"), val = tensor<bool, [3]>([false, true, true])];
98
+ tensor<bool, [3]> x_3_layer_0_tmp_end_mask_0 = const()[name = tensor<string, []>("x_3_layer_0_tmp_end_mask_0"), val = tensor<bool, [3]>([true, true, true])];
99
+ tensor<fp32, [100, 1, 256]> x_3_layer_0_tmp = slice_by_index(begin = x_3_layer_0_tmp_begin_0, begin_mask = x_3_layer_0_tmp_begin_mask_0, end = x_3_layer_0_tmp_end_0, end_mask = x_3_layer_0_tmp_end_mask_0, x = while_loop_0_1)[name = tensor<string, []>("x_3_layer_0_tmp")];
100
+ tensor<int32, [1]> slice_by_index_13 = const()[name = tensor<string, []>("slice_by_index_13"), val = tensor<int32, [1]>([100])];
101
+ tensor<int32, [1]> while_loop_1_loop_vars0_0 = const()[name = tensor<string, []>("while_loop_1_loop_vars0_0"), val = tensor<int32, [1]>([0])];
102
+ tensor<int32, [1]> while_loop_1_0, tensor<fp32, [101, 1, 256]> while_loop_1_1 = while_loop(loop_vars = (while_loop_1_loop_vars0_0, concat_12))[name = tensor<string, []>("while_loop_1")]
103
+ (tensor<int32, [1]> while_loop_1_loop_vars0_0_x0_1_1_1_0, tensor<fp32, [101, 1, 256]> concat_14_x0_1_1_1_0) {
104
+ tensor<bool, [1]> less_3 = less(x = while_loop_1_loop_vars0_0_x0_1_1_1_0, y = slice_by_index_13)[name = tensor<string, []>("less_3")];
105
+ } -> (less_3)
106
+ (tensor<int32, [1]> while_loop_1_loop_vars0_0_x0_1_1_1_1, tensor<fp32, [101, 1, 256]> concat_14_x0_1_1_1_1) {
107
+ tensor<int32, []> gather_6_batch_dims_0 = const()[name = tensor<string, []>("gather_6_batch_dims_0"), val = tensor<int32, []>(0)];
108
+ tensor<bool, []> gather_6_validate_indices_0 = const()[name = tensor<string, []>("gather_6_validate_indices_0"), val = tensor<bool, []>(false)];
109
+ tensor<int32, []> greater_equal_2_y_0 = const()[name = tensor<string, []>("greater_equal_2_y_0"), val = tensor<int32, []>(0)];
110
+ tensor<bool, [1]> greater_equal_2 = greater_equal(x = while_loop_1_loop_vars0_0_x0_1_1_1_1, y = greater_equal_2_y_0)[name = tensor<string, []>("greater_equal_2")];
111
+ tensor<int32, []> slice_by_index_28 = const()[name = tensor<string, []>("slice_by_index_28"), val = tensor<int32, []>(100)];
112
+ tensor<int32, [1]> add_22 = add(x = while_loop_1_loop_vars0_0_x0_1_1_1_1, y = slice_by_index_28)[name = tensor<string, []>("add_22")];
113
+ tensor<int32, [1]> select_2 = select(a = while_loop_1_loop_vars0_0_x0_1_1_1_1, b = add_22, cond = greater_equal_2)[name = tensor<string, []>("select_2")];
114
+ tensor<int32, []> gather_6_axis_1 = const()[name = tensor<string, []>("gather_6_axis_1"), val = tensor<int32, []>(0)];
115
+ tensor<fp32, [1, 1, 256]> gather_6 = gather(axis = gather_6_axis_1, batch_dims = gather_6_batch_dims_0, indices = select_2, validate_indices = gather_6_validate_indices_0, x = x_3_layer_0_tmp)[name = tensor<string, []>("gather_6")];
116
+ tensor<int32, []> gather_7_batch_dims_0 = const()[name = tensor<string, []>("gather_7_batch_dims_0"), val = tensor<int32, []>(0)];
117
+ tensor<bool, []> gather_7_validate_indices_0 = const()[name = tensor<string, []>("gather_7_validate_indices_0"), val = tensor<bool, []>(false)];
118
+ tensor<int32, []> slice_by_index_29 = const()[name = tensor<string, []>("slice_by_index_29"), val = tensor<int32, []>(101)];
119
+ tensor<int32, [1]> add_23 = add(x = while_loop_1_loop_vars0_0_x0_1_1_1_1, y = slice_by_index_29)[name = tensor<string, []>("add_23")];
120
+ tensor<int32, [1]> select_3 = select(a = while_loop_1_loop_vars0_0_x0_1_1_1_1, b = add_23, cond = greater_equal_2)[name = tensor<string, []>("select_3")];
121
+ tensor<int32, []> gather_7_axis_1 = const()[name = tensor<string, []>("gather_7_axis_1"), val = tensor<int32, []>(0)];
122
+ tensor<fp32, [1, 1, 256]> gather_7 = gather(axis = gather_7_axis_1, batch_dims = gather_7_batch_dims_0, indices = select_3, validate_indices = gather_7_validate_indices_0, x = concat_14_x0_1_1_1_1)[name = tensor<string, []>("gather_7")];
123
+ tensor<int32, [1]> squeeze_6_axes_0 = const()[name = tensor<string, []>("squeeze_6_axes_0"), val = tensor<int32, [1]>([0])];
124
+ tensor<fp32, [1, 256]> squeeze_6 = squeeze(axes = squeeze_6_axes_0, x = gather_6)[name = tensor<string, []>("squeeze_6")];
125
+ tensor<int32, [1]> squeeze_7_axes_0 = const()[name = tensor<string, []>("squeeze_7_axes_0"), val = tensor<int32, [1]>([0])];
126
+ tensor<fp32, [1, 256]> squeeze_7 = squeeze(axes = squeeze_7_axes_0, x = gather_7)[name = tensor<string, []>("squeeze_7")];
127
+ tensor<fp32, [256, 256]> linear_18_weight_0 = const()[name = tensor<string, []>("linear_18_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1751488)))];
128
+ tensor<fp32, [256]> linear_18_bias_0 = const()[name = tensor<string, []>("linear_18_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2013696)))];
129
+ tensor<fp32, [1, 256]> linear_18 = linear(bias = linear_18_bias_0, weight = linear_18_weight_0, x = squeeze_6)[name = tensor<string, []>("linear_18")];
130
+ tensor<fp32, [256, 256]> linear_19_weight_0 = const()[name = tensor<string, []>("linear_19_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2014784)))];
131
+ tensor<fp32, [256]> linear_19_bias_0 = const()[name = tensor<string, []>("linear_19_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2276992)))];
132
+ tensor<fp32, [1, 256]> linear_19 = linear(bias = linear_19_bias_0, weight = linear_19_weight_0, x = squeeze_7)[name = tensor<string, []>("linear_19")];
133
+ tensor<fp32, [1, 256]> add_15 = add(x = linear_18, y = linear_19)[name = tensor<string, []>("add_15")];
134
+ tensor<fp32, [1, 256]> sigmoid_6 = sigmoid(x = add_15)[name = tensor<string, []>("sigmoid_6")];
135
+ tensor<fp32, [256, 256]> linear_20_weight_0 = const()[name = tensor<string, []>("linear_20_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2278080)))];
136
+ tensor<fp32, [256]> linear_20_bias_0 = const()[name = tensor<string, []>("linear_20_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2540288)))];
137
+ tensor<fp32, [1, 256]> linear_20 = linear(bias = linear_20_bias_0, weight = linear_20_weight_0, x = squeeze_6)[name = tensor<string, []>("linear_20")];
138
+ tensor<fp32, [256, 256]> linear_21_weight_0 = const()[name = tensor<string, []>("linear_21_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2541376)))];
139
+ tensor<fp32, [256]> linear_21_bias_0 = const()[name = tensor<string, []>("linear_21_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2803584)))];
140
+ tensor<fp32, [1, 256]> linear_21 = linear(bias = linear_21_bias_0, weight = linear_21_weight_0, x = squeeze_7)[name = tensor<string, []>("linear_21")];
141
+ tensor<fp32, [1, 256]> add_16 = add(x = linear_20, y = linear_21)[name = tensor<string, []>("add_16")];
142
+ tensor<fp32, [1, 256]> sigmoid_7 = sigmoid(x = add_16)[name = tensor<string, []>("sigmoid_7")];
143
+ tensor<fp32, [256, 256]> linear_22_weight_0 = const()[name = tensor<string, []>("linear_22_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2804672)))];
144
+ tensor<fp32, [256]> linear_22_bias_0 = const()[name = tensor<string, []>("linear_22_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3066880)))];
145
+ tensor<fp32, [1, 256]> linear_22 = linear(bias = linear_22_bias_0, weight = linear_22_weight_0, x = squeeze_6)[name = tensor<string, []>("linear_22")];
146
+ tensor<fp32, [256, 256]> linear_23_weight_0 = const()[name = tensor<string, []>("linear_23_weight_0"), val = tensor<fp32, [256, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3067968)))];
147
+ tensor<fp32, [256]> linear_23_bias_0 = const()[name = tensor<string, []>("linear_23_bias_0"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3330176)))];
148
+ tensor<fp32, [1, 256]> linear_23 = linear(bias = linear_23_bias_0, weight = linear_23_weight_0, x = squeeze_7)[name = tensor<string, []>("linear_23")];
149
+ tensor<fp32, [1, 256]> mul_9 = mul(x = sigmoid_6, y = linear_23)[name = tensor<string, []>("mul_9")];
150
+ tensor<fp32, [1, 256]> add_17 = add(x = linear_22, y = mul_9)[name = tensor<string, []>("add_17")];
151
+ tensor<fp32, [1, 256]> tanh_3 = tanh(x = add_17)[name = tensor<string, []>("tanh_3")];
152
+ tensor<fp32, []> sub_3_x_0 = const()[name = tensor<string, []>("sub_3_x_0"), val = tensor<fp32, []>(0x1p+0)];
153
+ tensor<fp32, [1, 256]> sub_3 = sub(x = sub_3_x_0, y = sigmoid_7)[name = tensor<string, []>("sub_3")];
154
+ tensor<fp32, [1, 256]> mul_10 = mul(x = sub_3, y = tanh_3)[name = tensor<string, []>("mul_10")];
155
+ tensor<fp32, [1, 256]> mul_11 = mul(x = sigmoid_7, y = squeeze_7)[name = tensor<string, []>("mul_11")];
156
+ tensor<fp32, [1, 256]> add_18 = add(x = mul_10, y = mul_11)[name = tensor<string, []>("add_18")];
157
+ tensor<int32, []> add_19_y_0 = const()[name = tensor<string, []>("add_19_y_0"), val = tensor<int32, []>(1)];
158
+ tensor<int32, [1]> add_19 = add(x = while_loop_1_loop_vars0_0_x0_1_1_1_1, y = add_19_y_0)[name = tensor<string, []>("add_19")];
159
+ tensor<int32, [1]> expand_dims_3_axes_0 = const()[name = tensor<string, []>("expand_dims_3_axes_0"), val = tensor<int32, [1]>([0])];
160
+ tensor<fp32, [1, 1, 256]> expand_dims_3 = expand_dims(axes = expand_dims_3_axes_0, x = add_18)[name = tensor<string, []>("expand_dims_3")];
161
+ tensor<int32, []> scatter_3_axis_0 = const()[name = tensor<string, []>("scatter_3_axis_0"), val = tensor<int32, []>(0)];
162
+ tensor<string, []> scatter_3_mode_0 = const()[name = tensor<string, []>("scatter_3_mode_0"), val = tensor<string, []>("add")];
163
+ tensor<bool, []> scatter_3_validate_indices_0 = const()[name = tensor<string, []>("scatter_3_validate_indices_0"), val = tensor<bool, []>(false)];
164
+ tensor<fp32, [101, 1, 256]> scatter_3 = scatter(axis = scatter_3_axis_0, data = concat_14_x0_1_1_1_1, indices = add_19, mode = scatter_3_mode_0, updates = expand_dims_3, validate_indices = scatter_3_validate_indices_0)[name = tensor<string, []>("scatter_3")];
165
+ } -> (add_19, scatter_3);
166
+ tensor<int32, [3]> x_3_tmp_begin_0 = const()[name = tensor<string, []>("x_3_tmp_begin_0"), val = tensor<int32, [3]>([1, 0, 0])];
167
+ tensor<int32, [3]> x_3_tmp_end_0 = const()[name = tensor<string, []>("x_3_tmp_end_0"), val = tensor<int32, [3]>([0, 0, 0])];
168
+ tensor<bool, [3]> x_3_tmp_begin_mask_0 = const()[name = tensor<string, []>("x_3_tmp_begin_mask_0"), val = tensor<bool, [3]>([false, true, true])];
169
+ tensor<bool, [3]> x_3_tmp_end_mask_0 = const()[name = tensor<string, []>("x_3_tmp_end_mask_0"), val = tensor<bool, [3]>([true, true, true])];
170
+ tensor<fp32, [100, 1, 256]> x_3_tmp = slice_by_index(begin = x_3_tmp_begin_0, begin_mask = x_3_tmp_begin_mask_0, end = x_3_tmp_end_0, end_mask = x_3_tmp_end_mask_0, x = while_loop_1_1)[name = tensor<string, []>("x_3_tmp")];
171
+ tensor<int32, [3]> x_3_perm_0 = const()[name = tensor<string, []>("x_3_perm_0"), val = tensor<int32, [3]>([1, 0, 2])];
172
+ tensor<int32, [4]> var_87 = const()[name = tensor<string, []>("op_87"), val = tensor<int32, [4]>([1, 100, 16, 16])];
173
+ tensor<fp32, [1, 100, 256]> x_3 = transpose(perm = x_3_perm_0, x = x_3_tmp)[name = tensor<string, []>("transpose_8")];
174
+ tensor<fp32, [1, 100, 16, 16]> var_88 = reshape(shape = var_87, x = x_3)[name = tensor<string, []>("op_88")];
175
+ tensor<int32, [4]> transpose_3_perm_0 = const()[name = tensor<string, []>("transpose_3_perm_0"), val = tensor<int32, [4]>([2, 0, 1, 3])];
176
+ tensor<int32, [3]> concat_19 = const()[name = tensor<string, []>("concat_19"), val = tensor<int32, [3]>([16, 100, 16])];
177
+ tensor<fp32, [16, 1, 100, 16]> transpose_3 = transpose(perm = transpose_3_perm_0, x = var_88)[name = tensor<string, []>("transpose_7")];
178
+ tensor<fp32, [16, 100, 16]> reshape_3 = reshape(shape = concat_19, x = transpose_3)[name = tensor<string, []>("reshape_3")];
179
+ tensor<bool, []> matmul_1_transpose_x_0 = const()[name = tensor<string, []>("matmul_1_transpose_x_0"), val = tensor<bool, []>(false)];
180
+ tensor<bool, []> matmul_1_transpose_y_0 = const()[name = tensor<string, []>("matmul_1_transpose_y_0"), val = tensor<bool, []>(false)];
181
+ tensor<fp32, [16, 100, 32]> matmul_1 = matmul(transpose_x = matmul_1_transpose_x_0, transpose_y = matmul_1_transpose_y_0, x = reshape_3, y = decoder_emb_gru_linear_out_0_weight)[name = tensor<string, []>("matmul_1")];
182
+ tensor<int32, [4]> concat_24 = const()[name = tensor<string, []>("concat_24"), val = tensor<int32, [4]>([16, 1, 100, 32])];
183
+ tensor<fp32, [16, 1, 100, 32]> reshape_5 = reshape(shape = concat_24, x = matmul_1)[name = tensor<string, []>("reshape_5")];
184
+ tensor<int32, [4]> x_perm_0 = const()[name = tensor<string, []>("x_perm_0"), val = tensor<int32, [4]>([1, 2, 0, 3])];
185
+ tensor<int32, [3]> concat_25 = const()[name = tensor<string, []>("concat_25"), val = tensor<int32, [3]>([1, 100, 512])];
186
+ tensor<fp32, [1, 100, 16, 32]> x = transpose(perm = x_perm_0, x = reshape_5)[name = tensor<string, []>("transpose_6")];
187
+ tensor<fp32, [1, 100, 512]> input_5 = reshape(shape = concat_25, x = x)[name = tensor<string, []>("input_5")];
188
+ tensor<fp32, [1, 100, 512]> var_92 = relu(x = input_5)[name = tensor<string, []>("op_92")];
189
+ tensor<int32, [4]> concat_26 = const()[name = tensor<string, []>("concat_26"), val = tensor<int32, [4]>([1, 100, 8, 64])];
190
+ tensor<fp32, [1, 100, 8, 64]> var_97 = reshape(shape = concat_26, x = var_92)[name = tensor<string, []>("op_97")];
191
+ tensor<int32, [4]> var_102 = const()[name = tensor<string, []>("op_102"), val = tensor<int32, [4]>([0, 3, 1, 2])];
192
+ tensor<string, []> input_7_pad_type_0 = const()[name = tensor<string, []>("input_7_pad_type_0"), val = tensor<string, []>("valid")];
193
+ tensor<int32, []> input_7_groups_0 = const()[name = tensor<string, []>("input_7_groups_0"), val = tensor<int32, []>(64)];
194
+ tensor<int32, [2]> input_7_strides_0 = const()[name = tensor<string, []>("input_7_strides_0"), val = tensor<int32, [2]>([1, 1])];
195
+ tensor<int32, [4]> input_7_pad_0 = const()[name = tensor<string, []>("input_7_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
196
+ tensor<int32, [2]> input_7_dilations_0 = const()[name = tensor<string, []>("input_7_dilations_0"), val = tensor<int32, [2]>([1, 1])];
197
+ tensor<fp32, [64, 1, 1, 1]> const_5 = const()[name = tensor<string, []>("const_5"), val = tensor<fp32, [64, 1, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3331264)))];
198
+ tensor<fp32, [64]> const_6 = const()[name = tensor<string, []>("const_6"), val = tensor<fp32, [64]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3331584)))];
199
+ tensor<fp32, [1, 64, 100, 8]> input_9 = conv(bias = const_6, dilations = input_7_dilations_0, groups = input_7_groups_0, pad = input_7_pad_0, pad_type = input_7_pad_type_0, strides = input_7_strides_0, weight = const_5, x = e3)[name = tensor<string, []>("input_9")];
200
+ tensor<fp32, [1, 64, 100, 8]> var_125 = relu(x = input_9)[name = tensor<string, []>("op_125")];
201
+ tensor<fp32, [1, 64, 100, 8]> emb_out = transpose(perm = var_102, x = var_97)[name = tensor<string, []>("transpose_5")];
202
+ tensor<fp32, [1, 64, 100, 8]> input_11 = add(x = var_125, y = emb_out)[name = tensor<string, []>("input_11")];
203
+ tensor<string, []> input_13_pad_type_0 = const()[name = tensor<string, []>("input_13_pad_type_0"), val = tensor<string, []>("custom")];
204
+ tensor<int32, [4]> input_13_pad_0 = const()[name = tensor<string, []>("input_13_pad_0"), val = tensor<int32, [4]>([0, 0, 1, 1])];
205
+ tensor<int32, []> input_13_groups_0 = const()[name = tensor<string, []>("input_13_groups_0"), val = tensor<int32, []>(64)];
206
+ tensor<int32, [2]> input_13_strides_0 = const()[name = tensor<string, []>("input_13_strides_0"), val = tensor<int32, [2]>([1, 1])];
207
+ tensor<int32, [2]> input_13_dilations_0 = const()[name = tensor<string, []>("input_13_dilations_0"), val = tensor<int32, [2]>([1, 1])];
208
+ tensor<fp32, [1, 64, 100, 8]> input_13 = conv(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 = decoder_convt3_0_weight, x = input_11)[name = tensor<string, []>("input_13")];
209
+ tensor<string, []> input_15_pad_type_0 = const()[name = tensor<string, []>("input_15_pad_type_0"), val = tensor<string, []>("valid")];
210
+ tensor<int32, [2]> input_15_strides_0 = const()[name = tensor<string, []>("input_15_strides_0"), val = tensor<int32, [2]>([1, 1])];
211
+ tensor<int32, [4]> input_15_pad_0 = const()[name = tensor<string, []>("input_15_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
212
+ tensor<int32, [2]> input_15_dilations_0 = const()[name = tensor<string, []>("input_15_dilations_0"), val = tensor<int32, [2]>([1, 1])];
213
+ tensor<int32, []> input_15_groups_0 = const()[name = tensor<string, []>("input_15_groups_0"), val = tensor<int32, []>(1)];
214
+ tensor<fp32, [64, 64, 1, 1]> const_7 = const()[name = tensor<string, []>("const_7"), val = tensor<fp32, [64, 64, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3331904)))];
215
+ tensor<fp32, [64]> const_8 = const()[name = tensor<string, []>("const_8"), val = tensor<fp32, [64]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3348352)))];
216
+ tensor<fp32, [1, 64, 100, 8]> input_17 = conv(bias = const_8, dilations = input_15_dilations_0, groups = input_15_groups_0, pad = input_15_pad_0, pad_type = input_15_pad_type_0, strides = input_15_strides_0, weight = const_7, x = input_13)[name = tensor<string, []>("input_17")];
217
+ tensor<fp32, [1, 64, 100, 8]> e3_1 = relu(x = input_17)[name = tensor<string, []>("e3")];
218
+ tensor<string, []> input_19_pad_type_0 = const()[name = tensor<string, []>("input_19_pad_type_0"), val = tensor<string, []>("valid")];
219
+ tensor<int32, []> input_19_groups_0 = const()[name = tensor<string, []>("input_19_groups_0"), val = tensor<int32, []>(64)];
220
+ tensor<int32, [2]> input_19_strides_0 = const()[name = tensor<string, []>("input_19_strides_0"), val = tensor<int32, [2]>([1, 1])];
221
+ tensor<int32, [4]> input_19_pad_0 = const()[name = tensor<string, []>("input_19_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
222
+ tensor<int32, [2]> input_19_dilations_0 = const()[name = tensor<string, []>("input_19_dilations_0"), val = tensor<int32, [2]>([1, 1])];
223
+ tensor<fp32, [64, 1, 1, 1]> const_9 = const()[name = tensor<string, []>("const_9"), val = tensor<fp32, [64, 1, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3348672)))];
224
+ tensor<fp32, [64]> const_10 = const()[name = tensor<string, []>("const_10"), val = tensor<fp32, [64]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3348992)))];
225
+ tensor<fp32, [1, 64, 100, 8]> input_21 = conv(bias = const_10, dilations = input_19_dilations_0, groups = input_19_groups_0, pad = input_19_pad_0, pad_type = input_19_pad_type_0, strides = input_19_strides_0, weight = const_9, x = e2)[name = tensor<string, []>("input_21")];
226
+ tensor<fp32, [1, 64, 100, 8]> var_178 = relu(x = input_21)[name = tensor<string, []>("op_178")];
227
+ tensor<fp32, [1, 64, 100, 8]> input_23 = add(x = var_178, y = e3_1)[name = tensor<string, []>("input_23")];
228
+ tensor<string, []> conv_transpose_0_pad_type_0 = const()[name = tensor<string, []>("conv_transpose_0_pad_type_0"), val = tensor<string, []>("custom")];
229
+ tensor<int32, [4]> conv_transpose_0_pad_0 = const()[name = tensor<string, []>("conv_transpose_0_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
230
+ tensor<int32, [2]> conv_transpose_0_strides_0 = const()[name = tensor<string, []>("conv_transpose_0_strides_0"), val = tensor<int32, [2]>([1, 2])];
231
+ tensor<int32, []> conv_transpose_0_groups_0 = const()[name = tensor<string, []>("conv_transpose_0_groups_0"), val = tensor<int32, []>(64)];
232
+ tensor<int32, [2]> conv_transpose_0_dilations_0 = const()[name = tensor<string, []>("conv_transpose_0_dilations_0"), val = tensor<int32, [2]>([1, 1])];
233
+ tensor<int32, [4]> conv_transpose_0_has_output_shape_output_shape_0 = const()[name = tensor<string, []>("conv_transpose_0_has_output_shape_output_shape_0"), val = tensor<int32, [4]>([1, 64, 100, 17])];
234
+ tensor<fp32, [1, 64, 100, 17]> conv_transpose_0_has_output_shape = conv_transpose(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 = decoder_convt2_0_weight, x = input_23)[name = tensor<string, []>("conv_transpose_0_has_output_shape")];
235
+ tensor<int32, [2]> input_25_crop_height_0 = const()[name = tensor<string, []>("input_25_crop_height_0"), val = tensor<int32, [2]>([0, 0])];
236
+ tensor<int32, [2]> input_25_crop_width_0 = const()[name = tensor<string, []>("input_25_crop_width_0"), val = tensor<int32, [2]>([1, 0])];
237
+ tensor<fp32, [1, 64, 100, 16]> input_25 = crop(crop_height = input_25_crop_height_0, crop_width = input_25_crop_width_0, x = conv_transpose_0_has_output_shape)[name = tensor<string, []>("input_25")];
238
+ tensor<string, []> input_27_pad_type_0 = const()[name = tensor<string, []>("input_27_pad_type_0"), val = tensor<string, []>("valid")];
239
+ tensor<int32, [2]> input_27_strides_0 = const()[name = tensor<string, []>("input_27_strides_0"), val = tensor<int32, [2]>([1, 1])];
240
+ tensor<int32, [4]> input_27_pad_0 = const()[name = tensor<string, []>("input_27_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
241
+ tensor<int32, [2]> input_27_dilations_0 = const()[name = tensor<string, []>("input_27_dilations_0"), val = tensor<int32, [2]>([1, 1])];
242
+ tensor<int32, []> input_27_groups_0 = const()[name = tensor<string, []>("input_27_groups_0"), val = tensor<int32, []>(1)];
243
+ tensor<fp32, [64, 64, 1, 1]> const_11 = const()[name = tensor<string, []>("const_11"), val = tensor<fp32, [64, 64, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3349312)))];
244
+ tensor<fp32, [64]> const_12 = const()[name = tensor<string, []>("const_12"), val = tensor<fp32, [64]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3365760)))];
245
+ tensor<fp32, [1, 64, 100, 16]> input_29 = conv(bias = const_12, dilations = input_27_dilations_0, groups = input_27_groups_0, pad = input_27_pad_0, pad_type = input_27_pad_type_0, strides = input_27_strides_0, weight = const_11, x = input_25)[name = tensor<string, []>("input_29")];
246
+ tensor<fp32, [1, 64, 100, 16]> e2_1 = relu(x = input_29)[name = tensor<string, []>("e2")];
247
+ tensor<string, []> input_31_pad_type_0 = const()[name = tensor<string, []>("input_31_pad_type_0"), val = tensor<string, []>("valid")];
248
+ tensor<int32, []> input_31_groups_0 = const()[name = tensor<string, []>("input_31_groups_0"), val = tensor<int32, []>(64)];
249
+ tensor<int32, [2]> input_31_strides_0 = const()[name = tensor<string, []>("input_31_strides_0"), val = tensor<int32, [2]>([1, 1])];
250
+ tensor<int32, [4]> input_31_pad_0 = const()[name = tensor<string, []>("input_31_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
251
+ tensor<int32, [2]> input_31_dilations_0 = const()[name = tensor<string, []>("input_31_dilations_0"), val = tensor<int32, [2]>([1, 1])];
252
+ tensor<fp32, [64, 1, 1, 1]> const_13 = const()[name = tensor<string, []>("const_13"), val = tensor<fp32, [64, 1, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3366080)))];
253
+ tensor<fp32, [64]> const_14 = const()[name = tensor<string, []>("const_14"), val = tensor<fp32, [64]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3366400)))];
254
+ tensor<fp32, [1, 64, 100, 16]> input_33 = conv(bias = const_14, dilations = input_31_dilations_0, groups = input_31_groups_0, pad = input_31_pad_0, pad_type = input_31_pad_type_0, strides = input_31_strides_0, weight = const_13, x = e1)[name = tensor<string, []>("input_33")];
255
+ tensor<fp32, [1, 64, 100, 16]> var_232 = relu(x = input_33)[name = tensor<string, []>("op_232")];
256
+ tensor<fp32, [1, 64, 100, 16]> input_35 = add(x = var_232, y = e2_1)[name = tensor<string, []>("input_35")];
257
+ tensor<string, []> conv_transpose_1_pad_type_0 = const()[name = tensor<string, []>("conv_transpose_1_pad_type_0"), val = tensor<string, []>("custom")];
258
+ tensor<int32, [4]> conv_transpose_1_pad_0 = const()[name = tensor<string, []>("conv_transpose_1_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
259
+ tensor<int32, [2]> conv_transpose_1_strides_0 = const()[name = tensor<string, []>("conv_transpose_1_strides_0"), val = tensor<int32, [2]>([1, 2])];
260
+ tensor<int32, []> conv_transpose_1_groups_0 = const()[name = tensor<string, []>("conv_transpose_1_groups_0"), val = tensor<int32, []>(64)];
261
+ tensor<int32, [2]> conv_transpose_1_dilations_0 = const()[name = tensor<string, []>("conv_transpose_1_dilations_0"), val = tensor<int32, [2]>([1, 1])];
262
+ tensor<int32, [4]> conv_transpose_1_has_output_shape_output_shape_0 = const()[name = tensor<string, []>("conv_transpose_1_has_output_shape_output_shape_0"), val = tensor<int32, [4]>([1, 64, 100, 33])];
263
+ tensor<fp32, [1, 64, 100, 33]> conv_transpose_1_has_output_shape = conv_transpose(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 = decoder_convt1_0_weight, x = input_35)[name = tensor<string, []>("conv_transpose_1_has_output_shape")];
264
+ tensor<int32, [2]> input_37_crop_height_0 = const()[name = tensor<string, []>("input_37_crop_height_0"), val = tensor<int32, [2]>([0, 0])];
265
+ tensor<int32, [2]> input_37_crop_width_0 = const()[name = tensor<string, []>("input_37_crop_width_0"), val = tensor<int32, [2]>([1, 0])];
266
+ tensor<fp32, [1, 64, 100, 32]> input_37 = crop(crop_height = input_37_crop_height_0, crop_width = input_37_crop_width_0, x = conv_transpose_1_has_output_shape)[name = tensor<string, []>("input_37")];
267
+ tensor<string, []> input_39_pad_type_0 = const()[name = tensor<string, []>("input_39_pad_type_0"), val = tensor<string, []>("valid")];
268
+ tensor<int32, [2]> input_39_strides_0 = const()[name = tensor<string, []>("input_39_strides_0"), val = tensor<int32, [2]>([1, 1])];
269
+ tensor<int32, [4]> input_39_pad_0 = const()[name = tensor<string, []>("input_39_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
270
+ tensor<int32, [2]> input_39_dilations_0 = const()[name = tensor<string, []>("input_39_dilations_0"), val = tensor<int32, [2]>([1, 1])];
271
+ tensor<int32, []> input_39_groups_0 = const()[name = tensor<string, []>("input_39_groups_0"), val = tensor<int32, []>(1)];
272
+ tensor<fp32, [64, 64, 1, 1]> const_15 = const()[name = tensor<string, []>("const_15"), val = tensor<fp32, [64, 64, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3366720)))];
273
+ tensor<fp32, [64]> const_16 = const()[name = tensor<string, []>("const_16"), val = tensor<fp32, [64]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3383168)))];
274
+ tensor<fp32, [1, 64, 100, 32]> input_41 = conv(bias = const_16, dilations = input_39_dilations_0, groups = input_39_groups_0, pad = input_39_pad_0, pad_type = input_39_pad_type_0, strides = input_39_strides_0, weight = const_15, x = input_37)[name = tensor<string, []>("input_41")];
275
+ tensor<fp32, [1, 64, 100, 32]> e1_1 = relu(x = input_41)[name = tensor<string, []>("e1")];
276
+ tensor<string, []> input_43_pad_type_0 = const()[name = tensor<string, []>("input_43_pad_type_0"), val = tensor<string, []>("valid")];
277
+ tensor<int32, []> input_43_groups_0 = const()[name = tensor<string, []>("input_43_groups_0"), val = tensor<int32, []>(64)];
278
+ tensor<int32, [2]> input_43_strides_0 = const()[name = tensor<string, []>("input_43_strides_0"), val = tensor<int32, [2]>([1, 1])];
279
+ tensor<int32, [4]> input_43_pad_0 = const()[name = tensor<string, []>("input_43_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
280
+ tensor<int32, [2]> input_43_dilations_0 = const()[name = tensor<string, []>("input_43_dilations_0"), val = tensor<int32, [2]>([1, 1])];
281
+ tensor<fp32, [64, 1, 1, 1]> const_17 = const()[name = tensor<string, []>("const_17"), val = tensor<fp32, [64, 1, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3383488)))];
282
+ tensor<fp32, [64]> const_18 = const()[name = tensor<string, []>("const_18"), val = tensor<fp32, [64]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3383808)))];
283
+ tensor<fp32, [1, 64, 100, 32]> input_45 = conv(bias = const_18, dilations = input_43_dilations_0, groups = input_43_groups_0, pad = input_43_pad_0, pad_type = input_43_pad_type_0, strides = input_43_strides_0, weight = const_17, x = e0)[name = tensor<string, []>("input_45")];
284
+ tensor<fp32, [1, 64, 100, 32]> var_286 = relu(x = input_45)[name = tensor<string, []>("op_286")];
285
+ tensor<fp32, [1, 64, 100, 32]> input_47 = add(x = var_286, y = e1_1)[name = tensor<string, []>("input_47")];
286
+ tensor<string, []> input_pad_type_0 = const()[name = tensor<string, []>("input_pad_type_0"), val = tensor<string, []>("custom")];
287
+ tensor<int32, [4]> input_pad_0 = const()[name = tensor<string, []>("input_pad_0"), val = tensor<int32, [4]>([0, 0, 1, 1])];
288
+ tensor<int32, [2]> input_strides_0 = const()[name = tensor<string, []>("input_strides_0"), val = tensor<int32, [2]>([1, 1])];
289
+ tensor<int32, [2]> input_dilations_0 = const()[name = tensor<string, []>("input_dilations_0"), val = tensor<int32, [2]>([1, 1])];
290
+ tensor<int32, []> input_groups_0 = const()[name = tensor<string, []>("input_groups_0"), val = tensor<int32, []>(1)];
291
+ tensor<fp32, [1, 64, 1, 3]> const_19 = const()[name = tensor<string, []>("const_19"), val = tensor<fp32, [1, 64, 1, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3384128)))];
292
+ tensor<fp32, [1]> const_20 = const()[name = tensor<string, []>("const_20"), val = tensor<fp32, [1]>([-0x1.0ebe88p+0])];
293
+ tensor<fp32, [1, 1, 100, 32]> erb_gains = conv(bias = const_20, dilations = input_dilations_0, groups = input_groups_0, pad = input_pad_0, pad_type = input_pad_type_0, strides = input_strides_0, weight = const_19, x = input_47)[name = tensor<string, []>("erb_gains")];
294
+ } -> (erb_gains);
295
+ }
v1.0.0/DeepFilterNet3_ERBDecoder.mlmodelc/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9fca45aa0e894e5c4f40f871942c4c2419961cc751af8e444666462b4ad0de31
3
+ size 3384960