Hugging Face's logo

FluidInference
/
kokoro-82m-coreml

Text-to-Speech
Core ML
English
Japanese
Chinese
Model card Files
xet
 Community
3
alexwengg commited on
Commit
9db203c
·
verified ·
1 Parent(s): c032f6d

Add KokoroProsody_v2 (fp32 compute; fixes quiet onset for long utterances, FluidAudio #947)

Browse files
Files changed (12) hide show
  1. ANE-ja/KokoroProsody_v2.mlmodelc/analytics/coremldata.bin +3 -0
  2. ANE-ja/KokoroProsody_v2.mlmodelc/coremldata.bin +3 -0
  3. ANE-ja/KokoroProsody_v2.mlmodelc/model.mil +398 -0
  4. ANE-ja/KokoroProsody_v2.mlmodelc/weights/weight.bin +3 -0
  5. ANE-zh/KokoroProsody_v2.mlmodelc/analytics/coremldata.bin +3 -0
  6. ANE-zh/KokoroProsody_v2.mlmodelc/coremldata.bin +3 -0
  7. ANE-zh/KokoroProsody_v2.mlmodelc/model.mil +398 -0
  8. ANE-zh/KokoroProsody_v2.mlmodelc/weights/weight.bin +3 -0
  9. ANE/KokoroProsody_v2.mlmodelc/analytics/coremldata.bin +3 -0
  10. ANE/KokoroProsody_v2.mlmodelc/coremldata.bin +3 -0
  11. ANE/KokoroProsody_v2.mlmodelc/model.mil +398 -0
  12. ANE/KokoroProsody_v2.mlmodelc/weights/weight.bin +3 -0
ANE-ja/KokoroProsody_v2.mlmodelc/analytics/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:804d7d2e9a8fed345cedb57da61bd27fdd95db22a070cfa7d42ccd4f8e80a443
3
+ size 243
ANE-ja/KokoroProsody_v2.mlmodelc/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db8776a4896f690fd254cdb68a4551c168411c9b3fb98219630215bc3c357986
3
+ size 421
ANE-ja/KokoroProsody_v2.mlmodelc/model.mil ADDED
@@ -0,0 +1,398 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ program(1.0)
2
+ [buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3600.16.1"}, {"coremlc-version", "3600.25.2"}})]
3
+ {
4
+ func main<ios17>(tensor<fp16, [1, 640, ?]> en, tensor<fp16, [1, 128]> style_s) [FlexibleShapeInformation = tuple<tuple<tensor<string, []>, dict<tensor<string, []>, tensor<int32, [?]>>>, tuple<tensor<string, []>, dict<tensor<string, []>, list<tensor<int32, [2]>, ?>>>>((("DefaultShapes", {{"en", [1, 640, 120]}}), ("RangeDims", {{"en", [[1, 1], [640, 640], [1, 2000]]}})))] {
5
+ tensor<string, []> cast_0_dtype_0 = const()[name = tensor<string, []>("cast_0_dtype_0"), val = tensor<string, []>("fp32")];
6
+ tensor<string, []> cast_1_dtype_0 = const()[name = tensor<string, []>("cast_1_dtype_0"), val = tensor<string, []>("fp32")];
7
+ tensor<fp32, [1024]> F0_0_norm1_fc_bias = const()[name = tensor<string, []>("F0_0_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(64)))];
8
+ tensor<fp32, [1024, 128]> F0_0_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4224))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(135360))), name = tensor<string, []>("F0_0_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
9
+ tensor<fp32, [512]> F0_0_norm1_norm_bias = const()[name = tensor<string, []>("F0_0_norm1_norm_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(136448)))];
10
+ tensor<fp32, [512]> F0_0_norm1_norm_weight = const()[name = tensor<string, []>("F0_0_norm1_norm_weight"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(138560)))];
11
+ tensor<fp32, [512]> F0_0_conv1_bias = const()[name = tensor<string, []>("F0_0_conv1_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(140672)))];
12
+ tensor<fp32, [1024]> F0_0_norm2_fc_bias = const()[name = tensor<string, []>("F0_0_norm2_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(142784)))];
13
+ tensor<fp32, [1024, 128]> F0_0_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(146944))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(278080))), name = tensor<string, []>("F0_0_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
14
+ tensor<fp32, [512]> F0_0_conv2_bias = const()[name = tensor<string, []>("F0_0_conv2_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(279168)))];
15
+ tensor<fp32, [1024]> F0_1_norm1_fc_bias = const()[name = tensor<string, []>("F0_1_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(281280)))];
16
+ tensor<fp32, [1024, 128]> F0_1_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(285440))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(416576))), name = tensor<string, []>("F0_1_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
17
+ tensor<fp32, [512]> F0_1_pool_bias = const()[name = tensor<string, []>("F0_1_pool_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(417664)))];
18
+ tensor<fp32, [256]> F0_1_conv1_bias = const()[name = tensor<string, []>("F0_1_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(419776)))];
19
+ tensor<fp32, [512]> F0_1_norm2_fc_bias = const()[name = tensor<string, []>("F0_1_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(420864)))];
20
+ tensor<fp32, [512, 128]> F0_1_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(422976))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(488576))), name = tensor<string, []>("F0_1_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
21
+ tensor<fp32, [256]> F0_1_norm2_norm_bias = const()[name = tensor<string, []>("F0_1_norm2_norm_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(489664)))];
22
+ tensor<fp32, [256]> F0_1_norm2_norm_weight = const()[name = tensor<string, []>("F0_1_norm2_norm_weight"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(490752)))];
23
+ tensor<fp32, [256]> F0_1_conv2_bias = const()[name = tensor<string, []>("F0_1_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(491840)))];
24
+ tensor<fp32, [512]> F0_2_norm1_fc_bias = const()[name = tensor<string, []>("F0_2_norm1_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(492928)))];
25
+ tensor<fp32, [512, 128]> F0_2_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(495040))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(560640))), name = tensor<string, []>("F0_2_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
26
+ tensor<fp32, [256]> F0_2_conv1_bias = const()[name = tensor<string, []>("F0_2_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(561728)))];
27
+ tensor<fp32, [512]> F0_2_norm2_fc_bias = const()[name = tensor<string, []>("F0_2_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(562816)))];
28
+ tensor<fp32, [512, 128]> F0_2_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(564928))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(630528))), name = tensor<string, []>("F0_2_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
29
+ tensor<fp32, [256]> F0_2_conv2_bias = const()[name = tensor<string, []>("F0_2_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(631616)))];
30
+ tensor<fp32, [1]> F0_proj_bias = const()[name = tensor<string, []>("F0_proj_bias"), val = tensor<fp32, [1]>([0x1.edcbf4p-3])];
31
+ tensor<fp32, [1, 256, 1]> F0_proj_weight = const()[name = tensor<string, []>("F0_proj_weight"), val = tensor<fp32, [1, 256, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(632704)))];
32
+ tensor<fp32, [1024]> N_0_norm1_fc_bias = const()[name = tensor<string, []>("N_0_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(633792)))];
33
+ tensor<fp32, [1024, 128]> N_0_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(637952))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(769088))), name = tensor<string, []>("N_0_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
34
+ tensor<fp32, [512]> N_0_conv1_bias = const()[name = tensor<string, []>("N_0_conv1_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(770176)))];
35
+ tensor<fp32, [1024]> N_0_norm2_fc_bias = const()[name = tensor<string, []>("N_0_norm2_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(772288)))];
36
+ tensor<fp32, [1024, 128]> N_0_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(776448))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(907584))), name = tensor<string, []>("N_0_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
37
+ tensor<fp32, [512]> N_0_conv2_bias = const()[name = tensor<string, []>("N_0_conv2_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(908672)))];
38
+ tensor<fp32, [1024]> N_1_norm1_fc_bias = const()[name = tensor<string, []>("N_1_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(910784)))];
39
+ tensor<fp32, [1024, 128]> N_1_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(914944))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1046080))), name = tensor<string, []>("N_1_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
40
+ tensor<fp32, [512]> N_1_pool_bias = const()[name = tensor<string, []>("N_1_pool_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1047168)))];
41
+ tensor<fp32, [256]> N_1_conv1_bias = const()[name = tensor<string, []>("N_1_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1049280)))];
42
+ tensor<fp32, [512]> N_1_norm2_fc_bias = const()[name = tensor<string, []>("N_1_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1050368)))];
43
+ tensor<fp32, [512, 128]> N_1_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1052480))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1118080))), name = tensor<string, []>("N_1_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
44
+ tensor<fp32, [256]> N_1_conv2_bias = const()[name = tensor<string, []>("N_1_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1119168)))];
45
+ tensor<fp32, [512]> N_2_norm1_fc_bias = const()[name = tensor<string, []>("N_2_norm1_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1120256)))];
46
+ tensor<fp32, [512, 128]> N_2_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1122368))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1187968))), name = tensor<string, []>("N_2_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
47
+ tensor<fp32, [256]> N_2_conv1_bias = const()[name = tensor<string, []>("N_2_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1189056)))];
48
+ tensor<fp32, [512]> N_2_norm2_fc_bias = const()[name = tensor<string, []>("N_2_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1190144)))];
49
+ tensor<fp32, [512, 128]> N_2_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1192256))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1257856))), name = tensor<string, []>("N_2_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
50
+ tensor<fp32, [256]> N_2_conv2_bias = const()[name = tensor<string, []>("N_2_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1258944)))];
51
+ tensor<fp32, [1]> N_proj_bias = const()[name = tensor<string, []>("N_proj_bias"), val = tensor<fp32, [1]>([0x1.13144ap-4])];
52
+ tensor<fp32, [1, 256, 1]> N_proj_weight = const()[name = tensor<string, []>("N_proj_weight"), val = tensor<fp32, [1, 256, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1260032)))];
53
+ tensor<int32, [3]> transpose_0_perm_0 = const()[name = tensor<string, []>("transpose_0_perm_0"), val = tensor<int32, [3]>([-1, 0, -2])];
54
+ tensor<fp32, [1024]> add_0 = const()[name = tensor<string, []>("add_0"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1261120)))];
55
+ tensor<fp32, [1024]> add_1 = const()[name = tensor<string, []>("add_1"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1265280)))];
56
+ tensor<fp32, [1024, 640]> concat_4_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [655360]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1269440))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1924864))), name = tensor<string, []>("concat_4_palettized"), shape = tensor<uint32, [2]>([1024, 640])];
57
+ tensor<fp32, [1024, 256]> concat_5_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [262144]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1925952))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2188160))), name = tensor<string, []>("concat_5_palettized"), shape = tensor<uint32, [2]>([1024, 256])];
58
+ tensor<fp32, [1024, 640]> concat_6_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [655360]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2189248))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2844672))), name = tensor<string, []>("concat_6_palettized"), shape = tensor<uint32, [2]>([1024, 640])];
59
+ tensor<fp32, [1024, 256]> concat_7_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [262144]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2845760))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3107968))), name = tensor<string, []>("concat_7_palettized"), shape = tensor<uint32, [2]>([1024, 256])];
60
+ tensor<fp32, [1, 512]> x_batch_first_lstm_h0_reshaped = const()[name = tensor<string, []>("x_batch_first_lstm_h0_reshaped"), val = tensor<fp32, [1, 512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3109056)))];
61
+ tensor<string, []> x_batch_first_direction_0 = const()[name = tensor<string, []>("x_batch_first_direction_0"), val = tensor<string, []>("bidirectional")];
62
+ tensor<bool, []> x_batch_first_output_sequence_0 = const()[name = tensor<string, []>("x_batch_first_output_sequence_0"), val = tensor<bool, []>(true)];
63
+ tensor<string, []> x_batch_first_recurrent_activation_0 = const()[name = tensor<string, []>("x_batch_first_recurrent_activation_0"), val = tensor<string, []>("sigmoid")];
64
+ tensor<string, []> x_batch_first_cell_activation_0 = const()[name = tensor<string, []>("x_batch_first_cell_activation_0"), val = tensor<string, []>("tanh")];
65
+ tensor<string, []> x_batch_first_activation_0 = const()[name = tensor<string, []>("x_batch_first_activation_0"), val = tensor<string, []>("tanh")];
66
+ tensor<fp32, [1, 640, ?]> cast_0 = cast(dtype = cast_0_dtype_0, x = en)[name = tensor<string, []>("cast_1")];
67
+ tensor<fp32, [?, 1, 640]> transpose_0 = transpose(perm = transpose_0_perm_0, x = cast_0)[name = tensor<string, []>("transpose_2")];
68
+ tensor<fp32, [?, 1, 512]> x_batch_first_0, tensor<fp32, [1, 512]> x_batch_first_1, tensor<fp32, [1, 512]> x_batch_first_2 = lstm(activation = x_batch_first_activation_0, bias = add_0, bias_back = add_1, cell_activation = x_batch_first_cell_activation_0, direction = x_batch_first_direction_0, initial_c = x_batch_first_lstm_h0_reshaped, initial_h = x_batch_first_lstm_h0_reshaped, output_sequence = x_batch_first_output_sequence_0, recurrent_activation = x_batch_first_recurrent_activation_0, weight_hh = concat_5_palettized, weight_hh_back = concat_7_palettized, weight_ih = concat_4_palettized, weight_ih_back = concat_6_palettized, x = transpose_0)[name = tensor<string, []>("x_batch_first")];
69
+ tensor<fp32, []> var_53 = const()[name = tensor<string, []>("op_53"), val = tensor<fp32, []>(0x1.99999ap-3)];
70
+ tensor<fp32, []> var_56 = const()[name = tensor<string, []>("op_56"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
71
+ tensor<fp32, [1, 128]> cast_1 = cast(dtype = cast_1_dtype_0, x = style_s)[name = tensor<string, []>("cast_0")];
72
+ tensor<fp32, [1, 1024]> h_1 = linear(bias = F0_0_norm1_fc_bias, weight = F0_0_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_0")];
73
+ tensor<int32, [3]> var_79 = const()[name = tensor<string, []>("op_79"), val = tensor<int32, [3]>([1, 1024, 1])];
74
+ tensor<fp32, [1, 1024, 1]> h_3 = reshape(shape = var_79, x = h_1)[name = tensor<string, []>("h_3")];
75
+ tensor<int32, [2]> var_81_split_sizes_0 = const()[name = tensor<string, []>("op_81_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
76
+ tensor<int32, []> var_81_axis_0 = const()[name = tensor<string, []>("op_81_axis_0"), val = tensor<int32, []>(1)];
77
+ tensor<fp32, [1, 512, 1]> var_81_0, tensor<fp32, [1, 512, 1]> var_81_1 = split(axis = var_81_axis_0, split_sizes = var_81_split_sizes_0, x = h_3)[name = tensor<string, []>("op_81")];
78
+ tensor<fp32, []> var_83_promoted = const()[name = tensor<string, []>("op_83_promoted"), val = tensor<fp32, []>(0x1p+0)];
79
+ tensor<fp32, [1, 512, 1]> var_84 = add(x = var_81_0, y = var_83_promoted)[name = tensor<string, []>("op_84")];
80
+ tensor<int32, [3]> transpose_0_perm_0_1 = const()[name = tensor<string, []>("transpose_0_perm_0_1"), val = tensor<int32, [3]>([1, 2, 0])];
81
+ tensor<fp32, [1, 512, ?]> transpose_0_1 = transpose(perm = transpose_0_perm_0_1, x = x_batch_first_0)[name = tensor<string, []>("transpose_1")];
82
+ tensor<fp32, [1, 512, ?]> var_87 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_56, gamma = F0_0_norm1_norm_weight, x = transpose_0_1)[name = tensor<string, []>("op_87")];
83
+ tensor<fp32, [1, 512, ?]> var_88 = mul(x = var_84, y = var_87)[name = tensor<string, []>("op_88")];
84
+ tensor<fp32, [1, 512, ?]> input_5 = add(x = var_88, y = var_81_1)[name = tensor<string, []>("input_5")];
85
+ tensor<fp32, [1, 512, ?]> input_7 = leaky_relu(alpha = var_53, x = input_5)[name = tensor<string, []>("input_7")];
86
+ tensor<fp32, [512, 512, 3]> weight_3_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3111168))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3897664))), name = tensor<string, []>("weight_3_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
87
+ tensor<string, []> input_9_pad_type_0 = const()[name = tensor<string, []>("input_9_pad_type_0"), val = tensor<string, []>("custom")];
88
+ tensor<int32, [2]> input_9_pad_0 = const()[name = tensor<string, []>("input_9_pad_0"), val = tensor<int32, [2]>([1, 1])];
89
+ tensor<int32, [1]> input_9_strides_0 = const()[name = tensor<string, []>("input_9_strides_0"), val = tensor<int32, [1]>([1])];
90
+ tensor<int32, [1]> input_9_dilations_0 = const()[name = tensor<string, []>("input_9_dilations_0"), val = tensor<int32, [1]>([1])];
91
+ tensor<int32, []> input_9_groups_0 = const()[name = tensor<string, []>("input_9_groups_0"), val = tensor<int32, []>(1)];
92
+ tensor<fp32, [1, 512, ?]> input_9 = conv(bias = F0_0_conv1_bias, dilations = input_9_dilations_0, groups = input_9_groups_0, pad = input_9_pad_0, pad_type = input_9_pad_type_0, strides = input_9_strides_0, weight = weight_3_palettized, x = input_7)[name = tensor<string, []>("input_9")];
93
+ tensor<fp32, [1, 1024]> h_5 = linear(bias = F0_0_norm2_fc_bias, weight = F0_0_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_1")];
94
+ tensor<int32, [3]> var_105 = const()[name = tensor<string, []>("op_105"), val = tensor<int32, [3]>([1, 1024, 1])];
95
+ tensor<fp32, [1, 1024, 1]> h_7 = reshape(shape = var_105, x = h_5)[name = tensor<string, []>("h_7")];
96
+ tensor<int32, [2]> var_107_split_sizes_0 = const()[name = tensor<string, []>("op_107_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
97
+ tensor<int32, []> var_107_axis_0 = const()[name = tensor<string, []>("op_107_axis_0"), val = tensor<int32, []>(1)];
98
+ tensor<fp32, [1, 512, 1]> var_107_0, tensor<fp32, [1, 512, 1]> var_107_1 = split(axis = var_107_axis_0, split_sizes = var_107_split_sizes_0, x = h_7)[name = tensor<string, []>("op_107")];
99
+ tensor<fp32, []> var_109_promoted = const()[name = tensor<string, []>("op_109_promoted"), val = tensor<fp32, []>(0x1p+0)];
100
+ tensor<fp32, [1, 512, 1]> var_110 = add(x = var_107_0, y = var_109_promoted)[name = tensor<string, []>("op_110")];
101
+ tensor<fp32, [1, 512, ?]> var_113 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_56, gamma = F0_0_norm1_norm_weight, x = input_9)[name = tensor<string, []>("op_113")];
102
+ tensor<fp32, [1, 512, ?]> var_114 = mul(x = var_110, y = var_113)[name = tensor<string, []>("op_114")];
103
+ tensor<fp32, [1, 512, ?]> input_11 = add(x = var_114, y = var_107_1)[name = tensor<string, []>("input_11")];
104
+ tensor<fp32, [1, 512, ?]> input_13 = leaky_relu(alpha = var_53, x = input_11)[name = tensor<string, []>("input_13")];
105
+ tensor<fp32, [512, 512, 3]> weight_7_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3898752))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4685248))), name = tensor<string, []>("weight_7_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
106
+ tensor<string, []> out_1_pad_type_0 = const()[name = tensor<string, []>("out_1_pad_type_0"), val = tensor<string, []>("custom")];
107
+ tensor<int32, [2]> out_1_pad_0 = const()[name = tensor<string, []>("out_1_pad_0"), val = tensor<int32, [2]>([1, 1])];
108
+ tensor<int32, [1]> out_1_strides_0 = const()[name = tensor<string, []>("out_1_strides_0"), val = tensor<int32, [1]>([1])];
109
+ tensor<int32, [1]> out_1_dilations_0 = const()[name = tensor<string, []>("out_1_dilations_0"), val = tensor<int32, [1]>([1])];
110
+ tensor<int32, []> out_1_groups_0 = const()[name = tensor<string, []>("out_1_groups_0"), val = tensor<int32, []>(1)];
111
+ tensor<fp32, [1, 512, ?]> out_1 = conv(bias = F0_0_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_13)[name = tensor<string, []>("out_1")];
112
+ tensor<fp32, [1, 512, ?]> var_124 = add(x = out_1, y = transpose_0_1)[name = tensor<string, []>("op_124")];
113
+ tensor<fp32, []> var_125 = const()[name = tensor<string, []>("op_125"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
114
+ tensor<fp32, [1, 512, ?]> input_15 = mul(x = var_124, y = var_125)[name = tensor<string, []>("input_15")];
115
+ tensor<fp32, []> var_130 = const()[name = tensor<string, []>("op_130"), val = tensor<fp32, []>(0x1.99999ap-3)];
116
+ tensor<fp32, []> var_134 = const()[name = tensor<string, []>("op_134"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
117
+ tensor<fp32, [1, 1024]> h_9 = linear(bias = F0_1_norm1_fc_bias, weight = F0_1_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_2")];
118
+ tensor<int32, [3]> var_166 = const()[name = tensor<string, []>("op_166"), val = tensor<int32, [3]>([1, 1024, 1])];
119
+ tensor<fp32, [1, 1024, 1]> h_11 = reshape(shape = var_166, x = h_9)[name = tensor<string, []>("h_11")];
120
+ tensor<int32, [2]> var_168_split_sizes_0 = const()[name = tensor<string, []>("op_168_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
121
+ tensor<int32, []> var_168_axis_0 = const()[name = tensor<string, []>("op_168_axis_0"), val = tensor<int32, []>(1)];
122
+ tensor<fp32, [1, 512, 1]> var_168_0, tensor<fp32, [1, 512, 1]> var_168_1 = split(axis = var_168_axis_0, split_sizes = var_168_split_sizes_0, x = h_11)[name = tensor<string, []>("op_168")];
123
+ tensor<fp32, []> var_170_promoted = const()[name = tensor<string, []>("op_170_promoted"), val = tensor<fp32, []>(0x1p+0)];
124
+ tensor<fp32, [1, 512, 1]> var_171 = add(x = var_168_0, y = var_170_promoted)[name = tensor<string, []>("op_171")];
125
+ tensor<fp32, [1, 512, ?]> var_174 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_134, gamma = F0_0_norm1_norm_weight, x = input_15)[name = tensor<string, []>("op_174")];
126
+ tensor<fp32, [1, 512, ?]> var_175 = mul(x = var_171, y = var_174)[name = tensor<string, []>("op_175")];
127
+ tensor<fp32, [1, 512, ?]> input_17 = add(x = var_175, y = var_168_1)[name = tensor<string, []>("input_17")];
128
+ tensor<fp32, [1, 512, ?]> input_19 = leaky_relu(alpha = var_130, x = input_17)[name = tensor<string, []>("input_19")];
129
+ tensor<fp32, [512, 1, 3]> var_178 = const()[name = tensor<string, []>("op_178"), val = tensor<fp32, [512, 1, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4686336)))];
130
+ tensor<string, []> conv_transpose_0_pad_type_0 = const()[name = tensor<string, []>("conv_transpose_0_pad_type_0"), val = tensor<string, []>("custom")];
131
+ tensor<int32, [2]> conv_transpose_0_pad_0 = const()[name = tensor<string, []>("conv_transpose_0_pad_0"), val = tensor<int32, [2]>([0, 0])];
132
+ tensor<int32, [1]> conv_transpose_0_strides_0 = const()[name = tensor<string, []>("conv_transpose_0_strides_0"), val = tensor<int32, [1]>([2])];
133
+ tensor<int32, []> conv_transpose_0_groups_0 = const()[name = tensor<string, []>("conv_transpose_0_groups_0"), val = tensor<int32, []>(512)];
134
+ tensor<int32, [1]> conv_transpose_0_dilations_0 = const()[name = tensor<string, []>("conv_transpose_0_dilations_0"), val = tensor<int32, [1]>([1])];
135
+ tensor<fp32, [1, 512, ?]> conv_transpose_0 = conv_transpose(bias = F0_1_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 = var_178, x = input_19)[name = tensor<string, []>("conv_transpose_0")];
136
+ tensor<int32, [3]> input_21_begin_0 = const()[name = tensor<string, []>("input_21_begin_0"), val = tensor<int32, [3]>([0, 0, 1])];
137
+ tensor<int32, [3]> input_21_end_0 = const()[name = tensor<string, []>("input_21_end_0"), val = tensor<int32, [3]>([0, 0, 0])];
138
+ tensor<bool, [3]> input_21_begin_mask_0 = const()[name = tensor<string, []>("input_21_begin_mask_0"), val = tensor<bool, [3]>([true, true, false])];
139
+ tensor<bool, [3]> input_21_end_mask_0 = const()[name = tensor<string, []>("input_21_end_mask_0"), val = tensor<bool, [3]>([true, true, true])];
140
+ tensor<fp32, [1, 512, ?]> input_21 = slice_by_index(begin = input_21_begin_0, begin_mask = input_21_begin_mask_0, end = input_21_end_0, end_mask = input_21_end_mask_0, x = conv_transpose_0)[name = tensor<string, []>("input_21")];
141
+ tensor<fp32, [256, 512, 3]> weight_11_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [393216]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4692544))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5085824))), name = tensor<string, []>("weight_11_palettized"), shape = tensor<uint32, [3]>([256, 512, 3])];
142
+ tensor<string, []> input_23_pad_type_0 = const()[name = tensor<string, []>("input_23_pad_type_0"), val = tensor<string, []>("custom")];
143
+ tensor<int32, [2]> input_23_pad_0 = const()[name = tensor<string, []>("input_23_pad_0"), val = tensor<int32, [2]>([1, 1])];
144
+ tensor<int32, [1]> input_23_strides_0 = const()[name = tensor<string, []>("input_23_strides_0"), val = tensor<int32, [1]>([1])];
145
+ tensor<int32, [1]> input_23_dilations_0 = const()[name = tensor<string, []>("input_23_dilations_0"), val = tensor<int32, [1]>([1])];
146
+ tensor<int32, []> input_23_groups_0 = const()[name = tensor<string, []>("input_23_groups_0"), val = tensor<int32, []>(1)];
147
+ tensor<fp32, [1, 256, ?]> input_23 = conv(bias = F0_1_conv1_bias, dilations = input_23_dilations_0, groups = input_23_groups_0, pad = input_23_pad_0, pad_type = input_23_pad_type_0, strides = input_23_strides_0, weight = weight_11_palettized, x = input_21)[name = tensor<string, []>("input_23")];
148
+ tensor<fp32, [1, 512]> h_13 = linear(bias = F0_1_norm2_fc_bias, weight = F0_1_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_3")];
149
+ tensor<int32, [3]> var_199 = const()[name = tensor<string, []>("op_199"), val = tensor<int32, [3]>([1, 512, 1])];
150
+ tensor<fp32, [1, 512, 1]> h_15 = reshape(shape = var_199, x = h_13)[name = tensor<string, []>("h_15")];
151
+ tensor<int32, [2]> var_201_split_sizes_0 = const()[name = tensor<string, []>("op_201_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
152
+ tensor<int32, []> var_201_axis_0 = const()[name = tensor<string, []>("op_201_axis_0"), val = tensor<int32, []>(1)];
153
+ tensor<fp32, [1, 256, 1]> var_201_0, tensor<fp32, [1, 256, 1]> var_201_1 = split(axis = var_201_axis_0, split_sizes = var_201_split_sizes_0, x = h_15)[name = tensor<string, []>("op_201")];
154
+ tensor<fp32, []> var_203_promoted = const()[name = tensor<string, []>("op_203_promoted"), val = tensor<fp32, []>(0x1p+0)];
155
+ tensor<fp32, [1, 256, 1]> var_204 = add(x = var_201_0, y = var_203_promoted)[name = tensor<string, []>("op_204")];
156
+ tensor<fp32, [1, 256, ?]> var_207 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_134, gamma = F0_1_norm2_norm_weight, x = input_23)[name = tensor<string, []>("op_207")];
157
+ tensor<fp32, [1, 256, ?]> var_208 = mul(x = var_204, y = var_207)[name = tensor<string, []>("op_208")];
158
+ tensor<fp32, [1, 256, ?]> input_25 = add(x = var_208, y = var_201_1)[name = tensor<string, []>("input_25")];
159
+ tensor<fp32, [1, 256, ?]> input_27 = leaky_relu(alpha = var_130, x = input_25)[name = tensor<string, []>("input_27")];
160
+ tensor<fp32, [256, 256, 3]> weight_15_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5086912))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5283584))), name = tensor<string, []>("weight_15_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
161
+ tensor<string, []> out_3_pad_type_0 = const()[name = tensor<string, []>("out_3_pad_type_0"), val = tensor<string, []>("custom")];
162
+ tensor<int32, [2]> out_3_pad_0 = const()[name = tensor<string, []>("out_3_pad_0"), val = tensor<int32, [2]>([1, 1])];
163
+ tensor<int32, [1]> out_3_strides_0 = const()[name = tensor<string, []>("out_3_strides_0"), val = tensor<int32, [1]>([1])];
164
+ tensor<int32, [1]> out_3_dilations_0 = const()[name = tensor<string, []>("out_3_dilations_0"), val = tensor<int32, [1]>([1])];
165
+ tensor<int32, []> out_3_groups_0 = const()[name = tensor<string, []>("out_3_groups_0"), val = tensor<int32, []>(1)];
166
+ tensor<fp32, [1, 256, ?]> out_3 = conv(bias = F0_1_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_27)[name = tensor<string, []>("out_3")];
167
+ tensor<int32, [1]> expand_dims_0_axes_0 = const()[name = tensor<string, []>("expand_dims_0_axes_0"), val = tensor<int32, [1]>([3])];
168
+ tensor<fp32, [1, 512, ?, 1]> expand_dims_0 = expand_dims(axes = expand_dims_0_axes_0, x = input_15)[name = tensor<string, []>("expand_dims_0")];
169
+ tensor<int32, []> upsample_nearest_neighbor_0_scale_factor_height_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_0_scale_factor_height_0"), val = tensor<int32, []>(2)];
170
+ tensor<int32, []> upsample_nearest_neighbor_0_scale_factor_width_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_0_scale_factor_width_0"), val = tensor<int32, []>(1)];
171
+ tensor<fp32, [1, 512, ?, 1]> 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<string, []>("upsample_nearest_neighbor_0")];
172
+ tensor<int32, [1]> input_29_axes_0 = const()[name = tensor<string, []>("input_29_axes_0"), val = tensor<int32, [1]>([3])];
173
+ tensor<fp32, [1, 512, ?]> input_29 = squeeze(axes = input_29_axes_0, x = upsample_nearest_neighbor_0)[name = tensor<string, []>("input_29")];
174
+ tensor<fp32, [256, 512, 1]> weight_17_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5284672))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5415808))), name = tensor<string, []>("weight_17_palettized"), shape = tensor<uint32, [3]>([256, 512, 1])];
175
+ tensor<string, []> var_225_pad_type_0 = const()[name = tensor<string, []>("op_225_pad_type_0"), val = tensor<string, []>("valid")];
176
+ tensor<int32, [1]> var_225_strides_0 = const()[name = tensor<string, []>("op_225_strides_0"), val = tensor<int32, [1]>([1])];
177
+ tensor<int32, [2]> var_225_pad_0 = const()[name = tensor<string, []>("op_225_pad_0"), val = tensor<int32, [2]>([0, 0])];
178
+ tensor<int32, [1]> var_225_dilations_0 = const()[name = tensor<string, []>("op_225_dilations_0"), val = tensor<int32, [1]>([1])];
179
+ tensor<int32, []> var_225_groups_0 = const()[name = tensor<string, []>("op_225_groups_0"), val = tensor<int32, []>(1)];
180
+ tensor<fp32, [1, 256, ?]> var_225 = conv(dilations = var_225_dilations_0, groups = var_225_groups_0, pad = var_225_pad_0, pad_type = var_225_pad_type_0, strides = var_225_strides_0, weight = weight_17_palettized, x = input_29)[name = tensor<string, []>("op_225")];
181
+ tensor<fp32, [1, 256, ?]> var_226 = add(x = out_3, y = var_225)[name = tensor<string, []>("op_226")];
182
+ tensor<fp32, []> var_227 = const()[name = tensor<string, []>("op_227"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
183
+ tensor<fp32, [1, 256, ?]> input_31 = mul(x = var_226, y = var_227)[name = tensor<string, []>("input_31")];
184
+ tensor<fp32, []> var_230 = const()[name = tensor<string, []>("op_230"), val = tensor<fp32, []>(0x1.99999ap-3)];
185
+ tensor<fp32, []> var_233 = const()[name = tensor<string, []>("op_233"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
186
+ tensor<fp32, [1, 512]> h_17 = linear(bias = F0_2_norm1_fc_bias, weight = F0_2_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_4")];
187
+ tensor<int32, [3]> var_256 = const()[name = tensor<string, []>("op_256"), val = tensor<int32, [3]>([1, 512, 1])];
188
+ tensor<fp32, [1, 512, 1]> h_19 = reshape(shape = var_256, x = h_17)[name = tensor<string, []>("h_19")];
189
+ tensor<int32, [2]> var_258_split_sizes_0 = const()[name = tensor<string, []>("op_258_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
190
+ tensor<int32, []> var_258_axis_0 = const()[name = tensor<string, []>("op_258_axis_0"), val = tensor<int32, []>(1)];
191
+ tensor<fp32, [1, 256, 1]> var_258_0, tensor<fp32, [1, 256, 1]> var_258_1 = split(axis = var_258_axis_0, split_sizes = var_258_split_sizes_0, x = h_19)[name = tensor<string, []>("op_258")];
192
+ tensor<fp32, []> var_260_promoted = const()[name = tensor<string, []>("op_260_promoted"), val = tensor<fp32, []>(0x1p+0)];
193
+ tensor<fp32, [1, 256, 1]> var_261 = add(x = var_258_0, y = var_260_promoted)[name = tensor<string, []>("op_261")];
194
+ tensor<fp32, [1, 256, ?]> var_264 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_233, gamma = F0_1_norm2_norm_weight, x = input_31)[name = tensor<string, []>("op_264")];
195
+ tensor<fp32, [1, 256, ?]> var_265 = mul(x = var_261, y = var_264)[name = tensor<string, []>("op_265")];
196
+ tensor<fp32, [1, 256, ?]> input_33 = add(x = var_265, y = var_258_1)[name = tensor<string, []>("input_33")];
197
+ tensor<fp32, [1, 256, ?]> input_35 = leaky_relu(alpha = var_230, x = input_33)[name = tensor<string, []>("input_35")];
198
+ tensor<fp32, [256, 256, 3]> weight_21_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5416896))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5613568))), name = tensor<string, []>("weight_21_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
199
+ tensor<string, []> input_37_pad_type_0 = const()[name = tensor<string, []>("input_37_pad_type_0"), val = tensor<string, []>("custom")];
200
+ tensor<int32, [2]> input_37_pad_0 = const()[name = tensor<string, []>("input_37_pad_0"), val = tensor<int32, [2]>([1, 1])];
201
+ tensor<int32, [1]> input_37_strides_0 = const()[name = tensor<string, []>("input_37_strides_0"), val = tensor<int32, [1]>([1])];
202
+ tensor<int32, [1]> input_37_dilations_0 = const()[name = tensor<string, []>("input_37_dilations_0"), val = tensor<int32, [1]>([1])];
203
+ tensor<int32, []> input_37_groups_0 = const()[name = tensor<string, []>("input_37_groups_0"), val = tensor<int32, []>(1)];
204
+ tensor<fp32, [1, 256, ?]> input_37 = conv(bias = F0_2_conv1_bias, dilations = input_37_dilations_0, groups = input_37_groups_0, pad = input_37_pad_0, pad_type = input_37_pad_type_0, strides = input_37_strides_0, weight = weight_21_palettized, x = input_35)[name = tensor<string, []>("input_37")];
205
+ tensor<fp32, [1, 512]> h_21 = linear(bias = F0_2_norm2_fc_bias, weight = F0_2_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_5")];
206
+ tensor<int32, [3]> var_282 = const()[name = tensor<string, []>("op_282"), val = tensor<int32, [3]>([1, 512, 1])];
207
+ tensor<fp32, [1, 512, 1]> h_23 = reshape(shape = var_282, x = h_21)[name = tensor<string, []>("h_23")];
208
+ tensor<int32, [2]> var_284_split_sizes_0 = const()[name = tensor<string, []>("op_284_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
209
+ tensor<int32, []> var_284_axis_0 = const()[name = tensor<string, []>("op_284_axis_0"), val = tensor<int32, []>(1)];
210
+ tensor<fp32, [1, 256, 1]> var_284_0, tensor<fp32, [1, 256, 1]> var_284_1 = split(axis = var_284_axis_0, split_sizes = var_284_split_sizes_0, x = h_23)[name = tensor<string, []>("op_284")];
211
+ tensor<fp32, []> var_286_promoted = const()[name = tensor<string, []>("op_286_promoted"), val = tensor<fp32, []>(0x1p+0)];
212
+ tensor<fp32, [1, 256, 1]> var_287 = add(x = var_284_0, y = var_286_promoted)[name = tensor<string, []>("op_287")];
213
+ tensor<fp32, [1, 256, ?]> var_290 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_233, gamma = F0_1_norm2_norm_weight, x = input_37)[name = tensor<string, []>("op_290")];
214
+ tensor<fp32, [1, 256, ?]> var_291 = mul(x = var_287, y = var_290)[name = tensor<string, []>("op_291")];
215
+ tensor<fp32, [1, 256, ?]> input_39 = add(x = var_291, y = var_284_1)[name = tensor<string, []>("input_39")];
216
+ tensor<fp32, [1, 256, ?]> input_41 = leaky_relu(alpha = var_230, x = input_39)[name = tensor<string, []>("input_41")];
217
+ tensor<fp32, [256, 256, 3]> weight_25_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5614656))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5811328))), name = tensor<string, []>("weight_25_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
218
+ tensor<string, []> out_5_pad_type_0 = const()[name = tensor<string, []>("out_5_pad_type_0"), val = tensor<string, []>("custom")];
219
+ tensor<int32, [2]> out_5_pad_0 = const()[name = tensor<string, []>("out_5_pad_0"), val = tensor<int32, [2]>([1, 1])];
220
+ tensor<int32, [1]> out_5_strides_0 = const()[name = tensor<string, []>("out_5_strides_0"), val = tensor<int32, [1]>([1])];
221
+ tensor<int32, [1]> out_5_dilations_0 = const()[name = tensor<string, []>("out_5_dilations_0"), val = tensor<int32, [1]>([1])];
222
+ tensor<int32, []> out_5_groups_0 = const()[name = tensor<string, []>("out_5_groups_0"), val = tensor<int32, []>(1)];
223
+ tensor<fp32, [1, 256, ?]> out_5 = conv(bias = F0_2_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_25_palettized, x = input_41)[name = tensor<string, []>("out_5")];
224
+ tensor<fp32, [1, 256, ?]> var_301 = add(x = out_5, y = input_31)[name = tensor<string, []>("op_301")];
225
+ tensor<fp32, []> var_302 = const()[name = tensor<string, []>("op_302"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
226
+ tensor<fp32, [1, 256, ?]> input_43 = mul(x = var_301, y = var_302)[name = tensor<string, []>("input_43")];
227
+ tensor<string, []> var_314_pad_type_0 = const()[name = tensor<string, []>("op_314_pad_type_0"), val = tensor<string, []>("valid")];
228
+ tensor<int32, [1]> var_314_strides_0 = const()[name = tensor<string, []>("op_314_strides_0"), val = tensor<int32, [1]>([1])];
229
+ tensor<int32, [2]> var_314_pad_0 = const()[name = tensor<string, []>("op_314_pad_0"), val = tensor<int32, [2]>([0, 0])];
230
+ tensor<int32, [1]> var_314_dilations_0 = const()[name = tensor<string, []>("op_314_dilations_0"), val = tensor<int32, [1]>([1])];
231
+ tensor<int32, []> var_314_groups_0 = const()[name = tensor<string, []>("op_314_groups_0"), val = tensor<int32, []>(1)];
232
+ tensor<fp32, [1, 1, ?]> var_314 = conv(bias = F0_proj_bias, dilations = var_314_dilations_0, groups = var_314_groups_0, pad = var_314_pad_0, pad_type = var_314_pad_type_0, strides = var_314_strides_0, weight = F0_proj_weight, x = input_43)[name = tensor<string, []>("op_314")];
233
+ tensor<int32, [1]> var_316_axes_0 = const()[name = tensor<string, []>("op_316_axes_0"), val = tensor<int32, [1]>([1])];
234
+ tensor<fp32, [1, ?]> F0 = squeeze(axes = var_316_axes_0, x = var_314)[name = tensor<string, []>("op_316")];
235
+ tensor<fp32, []> var_321 = const()[name = tensor<string, []>("op_321"), val = tensor<fp32, []>(0x1.99999ap-3)];
236
+ tensor<fp32, []> var_324 = const()[name = tensor<string, []>("op_324"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
237
+ tensor<fp32, [1, 1024]> h_25 = linear(bias = N_0_norm1_fc_bias, weight = N_0_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_6")];
238
+ tensor<int32, [3]> var_347 = const()[name = tensor<string, []>("op_347"), val = tensor<int32, [3]>([1, 1024, 1])];
239
+ tensor<fp32, [1, 1024, 1]> h_27 = reshape(shape = var_347, x = h_25)[name = tensor<string, []>("h_27")];
240
+ tensor<int32, [2]> var_349_split_sizes_0 = const()[name = tensor<string, []>("op_349_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
241
+ tensor<int32, []> var_349_axis_0 = const()[name = tensor<string, []>("op_349_axis_0"), val = tensor<int32, []>(1)];
242
+ tensor<fp32, [1, 512, 1]> var_349_0, tensor<fp32, [1, 512, 1]> var_349_1 = split(axis = var_349_axis_0, split_sizes = var_349_split_sizes_0, x = h_27)[name = tensor<string, []>("op_349")];
243
+ tensor<fp32, []> var_351_promoted = const()[name = tensor<string, []>("op_351_promoted"), val = tensor<fp32, []>(0x1p+0)];
244
+ tensor<fp32, [1, 512, 1]> var_352 = add(x = var_349_0, y = var_351_promoted)[name = tensor<string, []>("op_352")];
245
+ tensor<fp32, [1, 512, ?]> var_356 = mul(x = var_352, y = var_87)[name = tensor<string, []>("op_356")];
246
+ tensor<fp32, [1, 512, ?]> input_47 = add(x = var_356, y = var_349_1)[name = tensor<string, []>("input_47")];
247
+ tensor<fp32, [1, 512, ?]> input_49 = leaky_relu(alpha = var_321, x = input_47)[name = tensor<string, []>("input_49")];
248
+ tensor<fp32, [512, 512, 3]> weight_31_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5812416))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(6598912))), name = tensor<string, []>("weight_31_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
249
+ tensor<string, []> input_51_pad_type_0 = const()[name = tensor<string, []>("input_51_pad_type_0"), val = tensor<string, []>("custom")];
250
+ tensor<int32, [2]> input_51_pad_0 = const()[name = tensor<string, []>("input_51_pad_0"), val = tensor<int32, [2]>([1, 1])];
251
+ tensor<int32, [1]> input_51_strides_0 = const()[name = tensor<string, []>("input_51_strides_0"), val = tensor<int32, [1]>([1])];
252
+ tensor<int32, [1]> input_51_dilations_0 = const()[name = tensor<string, []>("input_51_dilations_0"), val = tensor<int32, [1]>([1])];
253
+ tensor<int32, []> input_51_groups_0 = const()[name = tensor<string, []>("input_51_groups_0"), val = tensor<int32, []>(1)];
254
+ tensor<fp32, [1, 512, ?]> input_51 = conv(bias = N_0_conv1_bias, dilations = input_51_dilations_0, groups = input_51_groups_0, pad = input_51_pad_0, pad_type = input_51_pad_type_0, strides = input_51_strides_0, weight = weight_31_palettized, x = input_49)[name = tensor<string, []>("input_51")];
255
+ tensor<fp32, [1, 1024]> h_29 = linear(bias = N_0_norm2_fc_bias, weight = N_0_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_7")];
256
+ tensor<int32, [3]> var_373 = const()[name = tensor<string, []>("op_373"), val = tensor<int32, [3]>([1, 1024, 1])];
257
+ tensor<fp32, [1, 1024, 1]> h_31 = reshape(shape = var_373, x = h_29)[name = tensor<string, []>("h_31")];
258
+ tensor<int32, [2]> var_375_split_sizes_0 = const()[name = tensor<string, []>("op_375_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
259
+ tensor<int32, []> var_375_axis_0 = const()[name = tensor<string, []>("op_375_axis_0"), val = tensor<int32, []>(1)];
260
+ tensor<fp32, [1, 512, 1]> var_375_0, tensor<fp32, [1, 512, 1]> var_375_1 = split(axis = var_375_axis_0, split_sizes = var_375_split_sizes_0, x = h_31)[name = tensor<string, []>("op_375")];
261
+ tensor<fp32, []> var_377_promoted = const()[name = tensor<string, []>("op_377_promoted"), val = tensor<fp32, []>(0x1p+0)];
262
+ tensor<fp32, [1, 512, 1]> var_378 = add(x = var_375_0, y = var_377_promoted)[name = tensor<string, []>("op_378")];
263
+ tensor<fp32, [1, 512, ?]> var_381 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_324, gamma = F0_0_norm1_norm_weight, x = input_51)[name = tensor<string, []>("op_381")];
264
+ tensor<fp32, [1, 512, ?]> var_382 = mul(x = var_378, y = var_381)[name = tensor<string, []>("op_382")];
265
+ tensor<fp32, [1, 512, ?]> input_53 = add(x = var_382, y = var_375_1)[name = tensor<string, []>("input_53")];
266
+ tensor<fp32, [1, 512, ?]> input_55 = leaky_relu(alpha = var_321, x = input_53)[name = tensor<string, []>("input_55")];
267
+ tensor<fp32, [512, 512, 3]> weight_35_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(6600000))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7386496))), name = tensor<string, []>("weight_35_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
268
+ tensor<string, []> out_7_pad_type_0 = const()[name = tensor<string, []>("out_7_pad_type_0"), val = tensor<string, []>("custom")];
269
+ tensor<int32, [2]> out_7_pad_0 = const()[name = tensor<string, []>("out_7_pad_0"), val = tensor<int32, [2]>([1, 1])];
270
+ tensor<int32, [1]> out_7_strides_0 = const()[name = tensor<string, []>("out_7_strides_0"), val = tensor<int32, [1]>([1])];
271
+ tensor<int32, [1]> out_7_dilations_0 = const()[name = tensor<string, []>("out_7_dilations_0"), val = tensor<int32, [1]>([1])];
272
+ tensor<int32, []> out_7_groups_0 = const()[name = tensor<string, []>("out_7_groups_0"), val = tensor<int32, []>(1)];
273
+ tensor<fp32, [1, 512, ?]> out_7 = conv(bias = N_0_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_35_palettized, x = input_55)[name = tensor<string, []>("out_7")];
274
+ tensor<fp32, [1, 512, ?]> var_392 = add(x = out_7, y = transpose_0_1)[name = tensor<string, []>("op_392")];
275
+ tensor<fp32, []> var_393 = const()[name = tensor<string, []>("op_393"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
276
+ tensor<fp32, [1, 512, ?]> input_57 = mul(x = var_392, y = var_393)[name = tensor<string, []>("input_57")];
277
+ tensor<fp32, []> var_398 = const()[name = tensor<string, []>("op_398"), val = tensor<fp32, []>(0x1.99999ap-3)];
278
+ tensor<fp32, []> var_402 = const()[name = tensor<string, []>("op_402"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
279
+ tensor<fp32, [1, 1024]> h_33 = linear(bias = N_1_norm1_fc_bias, weight = N_1_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_8")];
280
+ tensor<int32, [3]> var_434 = const()[name = tensor<string, []>("op_434"), val = tensor<int32, [3]>([1, 1024, 1])];
281
+ tensor<fp32, [1, 1024, 1]> h_35 = reshape(shape = var_434, x = h_33)[name = tensor<string, []>("h_35")];
282
+ tensor<int32, [2]> var_436_split_sizes_0 = const()[name = tensor<string, []>("op_436_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
283
+ tensor<int32, []> var_436_axis_0 = const()[name = tensor<string, []>("op_436_axis_0"), val = tensor<int32, []>(1)];
284
+ tensor<fp32, [1, 512, 1]> var_436_0, tensor<fp32, [1, 512, 1]> var_436_1 = split(axis = var_436_axis_0, split_sizes = var_436_split_sizes_0, x = h_35)[name = tensor<string, []>("op_436")];
285
+ tensor<fp32, []> var_438_promoted = const()[name = tensor<string, []>("op_438_promoted"), val = tensor<fp32, []>(0x1p+0)];
286
+ tensor<fp32, [1, 512, 1]> var_439 = add(x = var_436_0, y = var_438_promoted)[name = tensor<string, []>("op_439")];
287
+ tensor<fp32, [1, 512, ?]> var_442 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_402, gamma = F0_0_norm1_norm_weight, x = input_57)[name = tensor<string, []>("op_442")];
288
+ tensor<fp32, [1, 512, ?]> var_443 = mul(x = var_439, y = var_442)[name = tensor<string, []>("op_443")];
289
+ tensor<fp32, [1, 512, ?]> input_59 = add(x = var_443, y = var_436_1)[name = tensor<string, []>("input_59")];
290
+ tensor<fp32, [1, 512, ?]> input_61 = leaky_relu(alpha = var_398, x = input_59)[name = tensor<string, []>("input_61")];
291
+ tensor<fp32, [512, 1, 3]> var_446 = const()[name = tensor<string, []>("op_446"), val = tensor<fp32, [512, 1, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7387584)))];
292
+ tensor<string, []> conv_transpose_1_pad_type_0 = const()[name = tensor<string, []>("conv_transpose_1_pad_type_0"), val = tensor<string, []>("custom")];
293
+ tensor<int32, [2]> conv_transpose_1_pad_0 = const()[name = tensor<string, []>("conv_transpose_1_pad_0"), val = tensor<int32, [2]>([0, 0])];
294
+ tensor<int32, [1]> conv_transpose_1_strides_0 = const()[name = tensor<string, []>("conv_transpose_1_strides_0"), val = tensor<int32, [1]>([2])];
295
+ tensor<int32, []> conv_transpose_1_groups_0 = const()[name = tensor<string, []>("conv_transpose_1_groups_0"), val = tensor<int32, []>(512)];
296
+ tensor<int32, [1]> conv_transpose_1_dilations_0 = const()[name = tensor<string, []>("conv_transpose_1_dilations_0"), val = tensor<int32, [1]>([1])];
297
+ tensor<fp32, [1, 512, ?]> conv_transpose_1 = conv_transpose(bias = N_1_pool_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 = var_446, x = input_61)[name = tensor<string, []>("conv_transpose_1")];
298
+ tensor<int32, [3]> input_63_begin_0 = const()[name = tensor<string, []>("input_63_begin_0"), val = tensor<int32, [3]>([0, 0, 1])];
299
+ tensor<int32, [3]> input_63_end_0 = const()[name = tensor<string, []>("input_63_end_0"), val = tensor<int32, [3]>([0, 0, 0])];
300
+ tensor<bool, [3]> input_63_begin_mask_0 = const()[name = tensor<string, []>("input_63_begin_mask_0"), val = tensor<bool, [3]>([true, true, false])];
301
+ tensor<bool, [3]> input_63_end_mask_0 = const()[name = tensor<string, []>("input_63_end_mask_0"), val = tensor<bool, [3]>([true, true, true])];
302
+ tensor<fp32, [1, 512, ?]> input_63 = slice_by_index(begin = input_63_begin_0, begin_mask = input_63_begin_mask_0, end = input_63_end_0, end_mask = input_63_end_mask_0, x = conv_transpose_1)[name = tensor<string, []>("input_63")];
303
+ tensor<fp32, [256, 512, 3]> weight_39_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [393216]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7393792))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7787072))), name = tensor<string, []>("weight_39_palettized"), shape = tensor<uint32, [3]>([256, 512, 3])];
304
+ tensor<string, []> input_65_pad_type_0 = const()[name = tensor<string, []>("input_65_pad_type_0"), val = tensor<string, []>("custom")];
305
+ tensor<int32, [2]> input_65_pad_0 = const()[name = tensor<string, []>("input_65_pad_0"), val = tensor<int32, [2]>([1, 1])];
306
+ tensor<int32, [1]> input_65_strides_0 = const()[name = tensor<string, []>("input_65_strides_0"), val = tensor<int32, [1]>([1])];
307
+ tensor<int32, [1]> input_65_dilations_0 = const()[name = tensor<string, []>("input_65_dilations_0"), val = tensor<int32, [1]>([1])];
308
+ tensor<int32, []> input_65_groups_0 = const()[name = tensor<string, []>("input_65_groups_0"), val = tensor<int32, []>(1)];
309
+ tensor<fp32, [1, 256, ?]> input_65 = conv(bias = N_1_conv1_bias, 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_39_palettized, x = input_63)[name = tensor<string, []>("input_65")];
310
+ tensor<fp32, [1, 512]> h_37 = linear(bias = N_1_norm2_fc_bias, weight = N_1_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_9")];
311
+ tensor<int32, [3]> var_467 = const()[name = tensor<string, []>("op_467"), val = tensor<int32, [3]>([1, 512, 1])];
312
+ tensor<fp32, [1, 512, 1]> h_39 = reshape(shape = var_467, x = h_37)[name = tensor<string, []>("h_39")];
313
+ tensor<int32, [2]> var_469_split_sizes_0 = const()[name = tensor<string, []>("op_469_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
314
+ tensor<int32, []> var_469_axis_0 = const()[name = tensor<string, []>("op_469_axis_0"), val = tensor<int32, []>(1)];
315
+ tensor<fp32, [1, 256, 1]> var_469_0, tensor<fp32, [1, 256, 1]> var_469_1 = split(axis = var_469_axis_0, split_sizes = var_469_split_sizes_0, x = h_39)[name = tensor<string, []>("op_469")];
316
+ tensor<fp32, []> var_471_promoted = const()[name = tensor<string, []>("op_471_promoted"), val = tensor<fp32, []>(0x1p+0)];
317
+ tensor<fp32, [1, 256, 1]> var_472 = add(x = var_469_0, y = var_471_promoted)[name = tensor<string, []>("op_472")];
318
+ tensor<fp32, [1, 256, ?]> var_475 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_402, gamma = F0_1_norm2_norm_weight, x = input_65)[name = tensor<string, []>("op_475")];
319
+ tensor<fp32, [1, 256, ?]> var_476 = mul(x = var_472, y = var_475)[name = tensor<string, []>("op_476")];
320
+ tensor<fp32, [1, 256, ?]> input_67 = add(x = var_476, y = var_469_1)[name = tensor<string, []>("input_67")];
321
+ tensor<fp32, [1, 256, ?]> input_69 = leaky_relu(alpha = var_398, x = input_67)[name = tensor<string, []>("input_69")];
322
+ tensor<fp32, [256, 256, 3]> weight_43_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7788160))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7984832))), name = tensor<string, []>("weight_43_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
323
+ tensor<string, []> out_9_pad_type_0 = const()[name = tensor<string, []>("out_9_pad_type_0"), val = tensor<string, []>("custom")];
324
+ tensor<int32, [2]> out_9_pad_0 = const()[name = tensor<string, []>("out_9_pad_0"), val = tensor<int32, [2]>([1, 1])];
325
+ tensor<int32, [1]> out_9_strides_0 = const()[name = tensor<string, []>("out_9_strides_0"), val = tensor<int32, [1]>([1])];
326
+ tensor<int32, [1]> out_9_dilations_0 = const()[name = tensor<string, []>("out_9_dilations_0"), val = tensor<int32, [1]>([1])];
327
+ tensor<int32, []> out_9_groups_0 = const()[name = tensor<string, []>("out_9_groups_0"), val = tensor<int32, []>(1)];
328
+ tensor<fp32, [1, 256, ?]> out_9 = conv(bias = N_1_conv2_bias, 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_43_palettized, x = input_69)[name = tensor<string, []>("out_9")];
329
+ tensor<int32, [1]> expand_dims_1_axes_0 = const()[name = tensor<string, []>("expand_dims_1_axes_0"), val = tensor<int32, [1]>([3])];
330
+ tensor<fp32, [1, 512, ?, 1]> expand_dims_1 = expand_dims(axes = expand_dims_1_axes_0, x = input_57)[name = tensor<string, []>("expand_dims_1")];
331
+ tensor<int32, []> upsample_nearest_neighbor_1_scale_factor_height_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_1_scale_factor_height_0"), val = tensor<int32, []>(2)];
332
+ tensor<int32, []> upsample_nearest_neighbor_1_scale_factor_width_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_1_scale_factor_width_0"), val = tensor<int32, []>(1)];
333
+ tensor<fp32, [1, 512, ?, 1]> upsample_nearest_neighbor_1 = 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)[name = tensor<string, []>("upsample_nearest_neighbor_1")];
334
+ tensor<int32, [1]> input_71_axes_0 = const()[name = tensor<string, []>("input_71_axes_0"), val = tensor<int32, [1]>([3])];
335
+ tensor<fp32, [1, 512, ?]> input_71 = squeeze(axes = input_71_axes_0, x = upsample_nearest_neighbor_1)[name = tensor<string, []>("input_71")];
336
+ tensor<fp32, [256, 512, 1]> weight_45_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7985920))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8117056))), name = tensor<string, []>("weight_45_palettized"), shape = tensor<uint32, [3]>([256, 512, 1])];
337
+ tensor<string, []> var_493_pad_type_0 = const()[name = tensor<string, []>("op_493_pad_type_0"), val = tensor<string, []>("valid")];
338
+ tensor<int32, [1]> var_493_strides_0 = const()[name = tensor<string, []>("op_493_strides_0"), val = tensor<int32, [1]>([1])];
339
+ tensor<int32, [2]> var_493_pad_0 = const()[name = tensor<string, []>("op_493_pad_0"), val = tensor<int32, [2]>([0, 0])];
340
+ tensor<int32, [1]> var_493_dilations_0 = const()[name = tensor<string, []>("op_493_dilations_0"), val = tensor<int32, [1]>([1])];
341
+ tensor<int32, []> var_493_groups_0 = const()[name = tensor<string, []>("op_493_groups_0"), val = tensor<int32, []>(1)];
342
+ tensor<fp32, [1, 256, ?]> var_493 = conv(dilations = var_493_dilations_0, groups = var_493_groups_0, pad = var_493_pad_0, pad_type = var_493_pad_type_0, strides = var_493_strides_0, weight = weight_45_palettized, x = input_71)[name = tensor<string, []>("op_493")];
343
+ tensor<fp32, [1, 256, ?]> var_494 = add(x = out_9, y = var_493)[name = tensor<string, []>("op_494")];
344
+ tensor<fp32, []> var_495 = const()[name = tensor<string, []>("op_495"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
345
+ tensor<fp32, [1, 256, ?]> input_73 = mul(x = var_494, y = var_495)[name = tensor<string, []>("input_73")];
346
+ tensor<fp32, []> var_498 = const()[name = tensor<string, []>("op_498"), val = tensor<fp32, []>(0x1.99999ap-3)];
347
+ tensor<fp32, []> var_501 = const()[name = tensor<string, []>("op_501"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
348
+ tensor<fp32, [1, 512]> h_41 = linear(bias = N_2_norm1_fc_bias, weight = N_2_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_10")];
349
+ tensor<int32, [3]> var_524 = const()[name = tensor<string, []>("op_524"), val = tensor<int32, [3]>([1, 512, 1])];
350
+ tensor<fp32, [1, 512, 1]> h_43 = reshape(shape = var_524, x = h_41)[name = tensor<string, []>("h_43")];
351
+ tensor<int32, [2]> var_526_split_sizes_0 = const()[name = tensor<string, []>("op_526_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
352
+ tensor<int32, []> var_526_axis_0 = const()[name = tensor<string, []>("op_526_axis_0"), val = tensor<int32, []>(1)];
353
+ tensor<fp32, [1, 256, 1]> var_526_0, tensor<fp32, [1, 256, 1]> var_526_1 = split(axis = var_526_axis_0, split_sizes = var_526_split_sizes_0, x = h_43)[name = tensor<string, []>("op_526")];
354
+ tensor<fp32, []> var_528_promoted = const()[name = tensor<string, []>("op_528_promoted"), val = tensor<fp32, []>(0x1p+0)];
355
+ tensor<fp32, [1, 256, 1]> var_529 = add(x = var_526_0, y = var_528_promoted)[name = tensor<string, []>("op_529")];
356
+ tensor<fp32, [1, 256, ?]> var_532 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_501, gamma = F0_1_norm2_norm_weight, x = input_73)[name = tensor<string, []>("op_532")];
357
+ tensor<fp32, [1, 256, ?]> var_533 = mul(x = var_529, y = var_532)[name = tensor<string, []>("op_533")];
358
+ tensor<fp32, [1, 256, ?]> input_75 = add(x = var_533, y = var_526_1)[name = tensor<string, []>("input_75")];
359
+ tensor<fp32, [1, 256, ?]> input_77 = leaky_relu(alpha = var_498, x = input_75)[name = tensor<string, []>("input_77")];
360
+ tensor<fp32, [256, 256, 3]> weight_49_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8118144))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8314816))), name = tensor<string, []>("weight_49_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
361
+ tensor<string, []> input_79_pad_type_0 = const()[name = tensor<string, []>("input_79_pad_type_0"), val = tensor<string, []>("custom")];
362
+ tensor<int32, [2]> input_79_pad_0 = const()[name = tensor<string, []>("input_79_pad_0"), val = tensor<int32, [2]>([1, 1])];
363
+ tensor<int32, [1]> input_79_strides_0 = const()[name = tensor<string, []>("input_79_strides_0"), val = tensor<int32, [1]>([1])];
364
+ tensor<int32, [1]> input_79_dilations_0 = const()[name = tensor<string, []>("input_79_dilations_0"), val = tensor<int32, [1]>([1])];
365
+ tensor<int32, []> input_79_groups_0 = const()[name = tensor<string, []>("input_79_groups_0"), val = tensor<int32, []>(1)];
366
+ tensor<fp32, [1, 256, ?]> input_79 = conv(bias = N_2_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_49_palettized, x = input_77)[name = tensor<string, []>("input_79")];
367
+ tensor<fp32, [1, 512]> h_45 = linear(bias = N_2_norm2_fc_bias, weight = N_2_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_11")];
368
+ tensor<int32, [3]> var_550 = const()[name = tensor<string, []>("op_550"), val = tensor<int32, [3]>([1, 512, 1])];
369
+ tensor<fp32, [1, 512, 1]> h = reshape(shape = var_550, x = h_45)[name = tensor<string, []>("h")];
370
+ tensor<int32, [2]> var_552_split_sizes_0 = const()[name = tensor<string, []>("op_552_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
371
+ tensor<int32, []> var_552_axis_0 = const()[name = tensor<string, []>("op_552_axis_0"), val = tensor<int32, []>(1)];
372
+ tensor<fp32, [1, 256, 1]> var_552_0, tensor<fp32, [1, 256, 1]> var_552_1 = split(axis = var_552_axis_0, split_sizes = var_552_split_sizes_0, x = h)[name = tensor<string, []>("op_552")];
373
+ tensor<fp32, []> var_554_promoted = const()[name = tensor<string, []>("op_554_promoted"), val = tensor<fp32, []>(0x1p+0)];
374
+ tensor<fp32, [1, 256, 1]> var_555 = add(x = var_552_0, y = var_554_promoted)[name = tensor<string, []>("op_555")];
375
+ tensor<fp32, [1, 256, ?]> var_558 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_501, gamma = F0_1_norm2_norm_weight, x = input_79)[name = tensor<string, []>("op_558")];
376
+ tensor<fp32, [1, 256, ?]> var_559 = mul(x = var_555, y = var_558)[name = tensor<string, []>("op_559")];
377
+ tensor<fp32, [1, 256, ?]> input_81 = add(x = var_559, y = var_552_1)[name = tensor<string, []>("input_81")];
378
+ tensor<fp32, [1, 256, ?]> input_83 = leaky_relu(alpha = var_498, x = input_81)[name = tensor<string, []>("input_83")];
379
+ tensor<fp32, [256, 256, 3]> weight_53_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8315904))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8512576))), name = tensor<string, []>("weight_53_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
380
+ tensor<string, []> out_pad_type_0 = const()[name = tensor<string, []>("out_pad_type_0"), val = tensor<string, []>("custom")];
381
+ tensor<int32, [2]> out_pad_0 = const()[name = tensor<string, []>("out_pad_0"), val = tensor<int32, [2]>([1, 1])];
382
+ tensor<int32, [1]> out_strides_0 = const()[name = tensor<string, []>("out_strides_0"), val = tensor<int32, [1]>([1])];
383
+ tensor<int32, [1]> out_dilations_0 = const()[name = tensor<string, []>("out_dilations_0"), val = tensor<int32, [1]>([1])];
384
+ tensor<int32, []> out_groups_0 = const()[name = tensor<string, []>("out_groups_0"), val = tensor<int32, []>(1)];
385
+ tensor<fp32, [1, 256, ?]> out = conv(bias = N_2_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_53_palettized, x = input_83)[name = tensor<string, []>("out")];
386
+ tensor<fp32, [1, 256, ?]> var_569 = add(x = out, y = input_73)[name = tensor<string, []>("op_569")];
387
+ tensor<fp32, []> var_570 = const()[name = tensor<string, []>("op_570"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
388
+ tensor<fp32, [1, 256, ?]> input = mul(x = var_569, y = var_570)[name = tensor<string, []>("input")];
389
+ tensor<string, []> var_582_pad_type_0 = const()[name = tensor<string, []>("op_582_pad_type_0"), val = tensor<string, []>("valid")];
390
+ tensor<int32, [1]> var_582_strides_0 = const()[name = tensor<string, []>("op_582_strides_0"), val = tensor<int32, [1]>([1])];
391
+ tensor<int32, [2]> var_582_pad_0 = const()[name = tensor<string, []>("op_582_pad_0"), val = tensor<int32, [2]>([0, 0])];
392
+ tensor<int32, [1]> var_582_dilations_0 = const()[name = tensor<string, []>("op_582_dilations_0"), val = tensor<int32, [1]>([1])];
393
+ tensor<int32, []> var_582_groups_0 = const()[name = tensor<string, []>("op_582_groups_0"), val = tensor<int32, []>(1)];
394
+ tensor<fp32, [1, 1, ?]> var_582 = conv(bias = N_proj_bias, dilations = var_582_dilations_0, groups = var_582_groups_0, pad = var_582_pad_0, pad_type = var_582_pad_type_0, strides = var_582_strides_0, weight = N_proj_weight, x = input)[name = tensor<string, []>("op_582")];
395
+ tensor<int32, [1]> var_584_axes_0 = const()[name = tensor<string, []>("op_584_axes_0"), val = tensor<int32, [1]>([1])];
396
+ tensor<fp32, [1, ?]> N = squeeze(axes = var_584_axes_0, x = var_582)[name = tensor<string, []>("op_584")];
397
+ } -> (F0, N);
398
+ }
ANE-ja/KokoroProsody_v2.mlmodelc/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70eea4cd2523992fa1fea840e31ab6f4fdd69c1598e220985a2ae6ea5da1e944
3
+ size 8513664
ANE-zh/KokoroProsody_v2.mlmodelc/analytics/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:623c7ad069ba684c8d58d545de23e72456d66deeb5ce84c59ae2f70ccbaa37a2
3
+ size 243
ANE-zh/KokoroProsody_v2.mlmodelc/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ae6eb6035b2e24124753a6fe42a66972723da55630d966def4dc3fc655f3e1ff
3
+ size 422
ANE-zh/KokoroProsody_v2.mlmodelc/model.mil ADDED
@@ -0,0 +1,398 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ program(1.0)
2
+ [buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3600.16.1"}, {"coremlc-version", "3600.25.2"}})]
3
+ {
4
+ func main<ios17>(tensor<fp16, [1, 640, ?]> en, tensor<fp16, [1, 128]> style_s) [FlexibleShapeInformation = tuple<tuple<tensor<string, []>, dict<tensor<string, []>, tensor<int32, [?]>>>, tuple<tensor<string, []>, dict<tensor<string, []>, list<tensor<int32, [2]>, ?>>>>((("DefaultShapes", {{"en", [1, 640, 133]}}), ("RangeDims", {{"en", [[1, 1], [640, 640], [1, 2000]]}})))] {
5
+ tensor<string, []> cast_0_dtype_0 = const()[name = tensor<string, []>("cast_0_dtype_0"), val = tensor<string, []>("fp32")];
6
+ tensor<string, []> cast_1_dtype_0 = const()[name = tensor<string, []>("cast_1_dtype_0"), val = tensor<string, []>("fp32")];
7
+ tensor<fp32, [1024]> F0_0_norm1_fc_bias = const()[name = tensor<string, []>("F0_0_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(64)))];
8
+ tensor<fp32, [1024, 128]> F0_0_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4224))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(135360))), name = tensor<string, []>("F0_0_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
9
+ tensor<fp32, [512]> F0_0_norm1_norm_bias = const()[name = tensor<string, []>("F0_0_norm1_norm_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(136448)))];
10
+ tensor<fp32, [512]> F0_0_norm1_norm_weight = const()[name = tensor<string, []>("F0_0_norm1_norm_weight"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(138560)))];
11
+ tensor<fp32, [512]> F0_0_conv1_bias = const()[name = tensor<string, []>("F0_0_conv1_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(140672)))];
12
+ tensor<fp32, [1024]> F0_0_norm2_fc_bias = const()[name = tensor<string, []>("F0_0_norm2_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(142784)))];
13
+ tensor<fp32, [1024, 128]> F0_0_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(146944))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(278080))), name = tensor<string, []>("F0_0_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
14
+ tensor<fp32, [512]> F0_0_conv2_bias = const()[name = tensor<string, []>("F0_0_conv2_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(279168)))];
15
+ tensor<fp32, [1024]> F0_1_norm1_fc_bias = const()[name = tensor<string, []>("F0_1_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(281280)))];
16
+ tensor<fp32, [1024, 128]> F0_1_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(285440))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(416576))), name = tensor<string, []>("F0_1_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
17
+ tensor<fp32, [512]> F0_1_pool_bias = const()[name = tensor<string, []>("F0_1_pool_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(417664)))];
18
+ tensor<fp32, [256]> F0_1_conv1_bias = const()[name = tensor<string, []>("F0_1_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(419776)))];
19
+ tensor<fp32, [512]> F0_1_norm2_fc_bias = const()[name = tensor<string, []>("F0_1_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(420864)))];
20
+ tensor<fp32, [512, 128]> F0_1_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(422976))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(488576))), name = tensor<string, []>("F0_1_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
21
+ tensor<fp32, [256]> F0_1_norm2_norm_bias = const()[name = tensor<string, []>("F0_1_norm2_norm_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(489664)))];
22
+ tensor<fp32, [256]> F0_1_norm2_norm_weight = const()[name = tensor<string, []>("F0_1_norm2_norm_weight"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(490752)))];
23
+ tensor<fp32, [256]> F0_1_conv2_bias = const()[name = tensor<string, []>("F0_1_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(491840)))];
24
+ tensor<fp32, [512]> F0_2_norm1_fc_bias = const()[name = tensor<string, []>("F0_2_norm1_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(492928)))];
25
+ tensor<fp32, [512, 128]> F0_2_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(495040))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(560640))), name = tensor<string, []>("F0_2_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
26
+ tensor<fp32, [256]> F0_2_conv1_bias = const()[name = tensor<string, []>("F0_2_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(561728)))];
27
+ tensor<fp32, [512]> F0_2_norm2_fc_bias = const()[name = tensor<string, []>("F0_2_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(562816)))];
28
+ tensor<fp32, [512, 128]> F0_2_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(564928))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(630528))), name = tensor<string, []>("F0_2_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
29
+ tensor<fp32, [256]> F0_2_conv2_bias = const()[name = tensor<string, []>("F0_2_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(631616)))];
30
+ tensor<fp32, [1]> F0_proj_bias = const()[name = tensor<string, []>("F0_proj_bias"), val = tensor<fp32, [1]>([0x1.f095bcp-3])];
31
+ tensor<fp32, [1, 256, 1]> F0_proj_weight = const()[name = tensor<string, []>("F0_proj_weight"), val = tensor<fp32, [1, 256, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(632704)))];
32
+ tensor<fp32, [1024]> N_0_norm1_fc_bias = const()[name = tensor<string, []>("N_0_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(633792)))];
33
+ tensor<fp32, [1024, 128]> N_0_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(637952))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(769088))), name = tensor<string, []>("N_0_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
34
+ tensor<fp32, [512]> N_0_conv1_bias = const()[name = tensor<string, []>("N_0_conv1_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(770176)))];
35
+ tensor<fp32, [1024]> N_0_norm2_fc_bias = const()[name = tensor<string, []>("N_0_norm2_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(772288)))];
36
+ tensor<fp32, [1024, 128]> N_0_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(776448))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(907584))), name = tensor<string, []>("N_0_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
37
+ tensor<fp32, [512]> N_0_conv2_bias = const()[name = tensor<string, []>("N_0_conv2_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(908672)))];
38
+ tensor<fp32, [1024]> N_1_norm1_fc_bias = const()[name = tensor<string, []>("N_1_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(910784)))];
39
+ tensor<fp32, [1024, 128]> N_1_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(914944))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1046080))), name = tensor<string, []>("N_1_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
40
+ tensor<fp32, [512]> N_1_pool_bias = const()[name = tensor<string, []>("N_1_pool_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1047168)))];
41
+ tensor<fp32, [256]> N_1_conv1_bias = const()[name = tensor<string, []>("N_1_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1049280)))];
42
+ tensor<fp32, [512]> N_1_norm2_fc_bias = const()[name = tensor<string, []>("N_1_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1050368)))];
43
+ tensor<fp32, [512, 128]> N_1_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1052480))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1118080))), name = tensor<string, []>("N_1_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
44
+ tensor<fp32, [256]> N_1_conv2_bias = const()[name = tensor<string, []>("N_1_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1119168)))];
45
+ tensor<fp32, [512]> N_2_norm1_fc_bias = const()[name = tensor<string, []>("N_2_norm1_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1120256)))];
46
+ tensor<fp32, [512, 128]> N_2_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1122368))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1187968))), name = tensor<string, []>("N_2_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
47
+ tensor<fp32, [256]> N_2_conv1_bias = const()[name = tensor<string, []>("N_2_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1189056)))];
48
+ tensor<fp32, [512]> N_2_norm2_fc_bias = const()[name = tensor<string, []>("N_2_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1190144)))];
49
+ tensor<fp32, [512, 128]> N_2_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1192256))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1257856))), name = tensor<string, []>("N_2_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
50
+ tensor<fp32, [256]> N_2_conv2_bias = const()[name = tensor<string, []>("N_2_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1258944)))];
51
+ tensor<fp32, [1]> N_proj_bias = const()[name = tensor<string, []>("N_proj_bias"), val = tensor<fp32, [1]>([0x1.3f131ep-5])];
52
+ tensor<fp32, [1, 256, 1]> N_proj_weight = const()[name = tensor<string, []>("N_proj_weight"), val = tensor<fp32, [1, 256, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1260032)))];
53
+ tensor<int32, [3]> transpose_0_perm_0 = const()[name = tensor<string, []>("transpose_0_perm_0"), val = tensor<int32, [3]>([-1, 0, -2])];
54
+ tensor<fp32, [1024]> add_0 = const()[name = tensor<string, []>("add_0"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1261120)))];
55
+ tensor<fp32, [1024]> add_1 = const()[name = tensor<string, []>("add_1"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1265280)))];
56
+ tensor<fp32, [1024, 640]> concat_4_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [655360]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1269440))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1924864))), name = tensor<string, []>("concat_4_palettized"), shape = tensor<uint32, [2]>([1024, 640])];
57
+ tensor<fp32, [1024, 256]> concat_5_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [262144]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1925952))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2188160))), name = tensor<string, []>("concat_5_palettized"), shape = tensor<uint32, [2]>([1024, 256])];
58
+ tensor<fp32, [1024, 640]> concat_6_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [655360]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2189248))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2844672))), name = tensor<string, []>("concat_6_palettized"), shape = tensor<uint32, [2]>([1024, 640])];
59
+ tensor<fp32, [1024, 256]> concat_7_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [262144]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2845760))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3107968))), name = tensor<string, []>("concat_7_palettized"), shape = tensor<uint32, [2]>([1024, 256])];
60
+ tensor<fp32, [1, 512]> x_batch_first_lstm_h0_reshaped = const()[name = tensor<string, []>("x_batch_first_lstm_h0_reshaped"), val = tensor<fp32, [1, 512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3109056)))];
61
+ tensor<string, []> x_batch_first_direction_0 = const()[name = tensor<string, []>("x_batch_first_direction_0"), val = tensor<string, []>("bidirectional")];
62
+ tensor<bool, []> x_batch_first_output_sequence_0 = const()[name = tensor<string, []>("x_batch_first_output_sequence_0"), val = tensor<bool, []>(true)];
63
+ tensor<string, []> x_batch_first_recurrent_activation_0 = const()[name = tensor<string, []>("x_batch_first_recurrent_activation_0"), val = tensor<string, []>("sigmoid")];
64
+ tensor<string, []> x_batch_first_cell_activation_0 = const()[name = tensor<string, []>("x_batch_first_cell_activation_0"), val = tensor<string, []>("tanh")];
65
+ tensor<string, []> x_batch_first_activation_0 = const()[name = tensor<string, []>("x_batch_first_activation_0"), val = tensor<string, []>("tanh")];
66
+ tensor<fp32, [1, 640, ?]> cast_0 = cast(dtype = cast_0_dtype_0, x = en)[name = tensor<string, []>("cast_1")];
67
+ tensor<fp32, [?, 1, 640]> transpose_0 = transpose(perm = transpose_0_perm_0, x = cast_0)[name = tensor<string, []>("transpose_2")];
68
+ tensor<fp32, [?, 1, 512]> x_batch_first_0, tensor<fp32, [1, 512]> x_batch_first_1, tensor<fp32, [1, 512]> x_batch_first_2 = lstm(activation = x_batch_first_activation_0, bias = add_0, bias_back = add_1, cell_activation = x_batch_first_cell_activation_0, direction = x_batch_first_direction_0, initial_c = x_batch_first_lstm_h0_reshaped, initial_h = x_batch_first_lstm_h0_reshaped, output_sequence = x_batch_first_output_sequence_0, recurrent_activation = x_batch_first_recurrent_activation_0, weight_hh = concat_5_palettized, weight_hh_back = concat_7_palettized, weight_ih = concat_4_palettized, weight_ih_back = concat_6_palettized, x = transpose_0)[name = tensor<string, []>("x_batch_first")];
69
+ tensor<fp32, []> var_53 = const()[name = tensor<string, []>("op_53"), val = tensor<fp32, []>(0x1.99999ap-3)];
70
+ tensor<fp32, []> var_56 = const()[name = tensor<string, []>("op_56"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
71
+ tensor<fp32, [1, 128]> cast_1 = cast(dtype = cast_1_dtype_0, x = style_s)[name = tensor<string, []>("cast_0")];
72
+ tensor<fp32, [1, 1024]> h_1 = linear(bias = F0_0_norm1_fc_bias, weight = F0_0_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_0")];
73
+ tensor<int32, [3]> var_79 = const()[name = tensor<string, []>("op_79"), val = tensor<int32, [3]>([1, 1024, 1])];
74
+ tensor<fp32, [1, 1024, 1]> h_3 = reshape(shape = var_79, x = h_1)[name = tensor<string, []>("h_3")];
75
+ tensor<int32, [2]> var_81_split_sizes_0 = const()[name = tensor<string, []>("op_81_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
76
+ tensor<int32, []> var_81_axis_0 = const()[name = tensor<string, []>("op_81_axis_0"), val = tensor<int32, []>(1)];
77
+ tensor<fp32, [1, 512, 1]> var_81_0, tensor<fp32, [1, 512, 1]> var_81_1 = split(axis = var_81_axis_0, split_sizes = var_81_split_sizes_0, x = h_3)[name = tensor<string, []>("op_81")];
78
+ tensor<fp32, []> var_83_promoted = const()[name = tensor<string, []>("op_83_promoted"), val = tensor<fp32, []>(0x1p+0)];
79
+ tensor<fp32, [1, 512, 1]> var_84 = add(x = var_81_0, y = var_83_promoted)[name = tensor<string, []>("op_84")];
80
+ tensor<int32, [3]> transpose_0_perm_0_1 = const()[name = tensor<string, []>("transpose_0_perm_0_1"), val = tensor<int32, [3]>([1, 2, 0])];
81
+ tensor<fp32, [1, 512, ?]> transpose_0_1 = transpose(perm = transpose_0_perm_0_1, x = x_batch_first_0)[name = tensor<string, []>("transpose_1")];
82
+ tensor<fp32, [1, 512, ?]> var_87 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_56, gamma = F0_0_norm1_norm_weight, x = transpose_0_1)[name = tensor<string, []>("op_87")];
83
+ tensor<fp32, [1, 512, ?]> var_88 = mul(x = var_84, y = var_87)[name = tensor<string, []>("op_88")];
84
+ tensor<fp32, [1, 512, ?]> input_5 = add(x = var_88, y = var_81_1)[name = tensor<string, []>("input_5")];
85
+ tensor<fp32, [1, 512, ?]> input_7 = leaky_relu(alpha = var_53, x = input_5)[name = tensor<string, []>("input_7")];
86
+ tensor<fp32, [512, 512, 3]> weight_3_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3111168))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3897664))), name = tensor<string, []>("weight_3_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
87
+ tensor<string, []> input_9_pad_type_0 = const()[name = tensor<string, []>("input_9_pad_type_0"), val = tensor<string, []>("custom")];
88
+ tensor<int32, [2]> input_9_pad_0 = const()[name = tensor<string, []>("input_9_pad_0"), val = tensor<int32, [2]>([1, 1])];
89
+ tensor<int32, [1]> input_9_strides_0 = const()[name = tensor<string, []>("input_9_strides_0"), val = tensor<int32, [1]>([1])];
90
+ tensor<int32, [1]> input_9_dilations_0 = const()[name = tensor<string, []>("input_9_dilations_0"), val = tensor<int32, [1]>([1])];
91
+ tensor<int32, []> input_9_groups_0 = const()[name = tensor<string, []>("input_9_groups_0"), val = tensor<int32, []>(1)];
92
+ tensor<fp32, [1, 512, ?]> input_9 = conv(bias = F0_0_conv1_bias, dilations = input_9_dilations_0, groups = input_9_groups_0, pad = input_9_pad_0, pad_type = input_9_pad_type_0, strides = input_9_strides_0, weight = weight_3_palettized, x = input_7)[name = tensor<string, []>("input_9")];
93
+ tensor<fp32, [1, 1024]> h_5 = linear(bias = F0_0_norm2_fc_bias, weight = F0_0_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_1")];
94
+ tensor<int32, [3]> var_105 = const()[name = tensor<string, []>("op_105"), val = tensor<int32, [3]>([1, 1024, 1])];
95
+ tensor<fp32, [1, 1024, 1]> h_7 = reshape(shape = var_105, x = h_5)[name = tensor<string, []>("h_7")];
96
+ tensor<int32, [2]> var_107_split_sizes_0 = const()[name = tensor<string, []>("op_107_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
97
+ tensor<int32, []> var_107_axis_0 = const()[name = tensor<string, []>("op_107_axis_0"), val = tensor<int32, []>(1)];
98
+ tensor<fp32, [1, 512, 1]> var_107_0, tensor<fp32, [1, 512, 1]> var_107_1 = split(axis = var_107_axis_0, split_sizes = var_107_split_sizes_0, x = h_7)[name = tensor<string, []>("op_107")];
99
+ tensor<fp32, []> var_109_promoted = const()[name = tensor<string, []>("op_109_promoted"), val = tensor<fp32, []>(0x1p+0)];
100
+ tensor<fp32, [1, 512, 1]> var_110 = add(x = var_107_0, y = var_109_promoted)[name = tensor<string, []>("op_110")];
101
+ tensor<fp32, [1, 512, ?]> var_113 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_56, gamma = F0_0_norm1_norm_weight, x = input_9)[name = tensor<string, []>("op_113")];
102
+ tensor<fp32, [1, 512, ?]> var_114 = mul(x = var_110, y = var_113)[name = tensor<string, []>("op_114")];
103
+ tensor<fp32, [1, 512, ?]> input_11 = add(x = var_114, y = var_107_1)[name = tensor<string, []>("input_11")];
104
+ tensor<fp32, [1, 512, ?]> input_13 = leaky_relu(alpha = var_53, x = input_11)[name = tensor<string, []>("input_13")];
105
+ tensor<fp32, [512, 512, 3]> weight_7_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3898752))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4685248))), name = tensor<string, []>("weight_7_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
106
+ tensor<string, []> out_1_pad_type_0 = const()[name = tensor<string, []>("out_1_pad_type_0"), val = tensor<string, []>("custom")];
107
+ tensor<int32, [2]> out_1_pad_0 = const()[name = tensor<string, []>("out_1_pad_0"), val = tensor<int32, [2]>([1, 1])];
108
+ tensor<int32, [1]> out_1_strides_0 = const()[name = tensor<string, []>("out_1_strides_0"), val = tensor<int32, [1]>([1])];
109
+ tensor<int32, [1]> out_1_dilations_0 = const()[name = tensor<string, []>("out_1_dilations_0"), val = tensor<int32, [1]>([1])];
110
+ tensor<int32, []> out_1_groups_0 = const()[name = tensor<string, []>("out_1_groups_0"), val = tensor<int32, []>(1)];
111
+ tensor<fp32, [1, 512, ?]> out_1 = conv(bias = F0_0_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_13)[name = tensor<string, []>("out_1")];
112
+ tensor<fp32, [1, 512, ?]> var_124 = add(x = out_1, y = transpose_0_1)[name = tensor<string, []>("op_124")];
113
+ tensor<fp32, []> var_125 = const()[name = tensor<string, []>("op_125"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
114
+ tensor<fp32, [1, 512, ?]> input_15 = mul(x = var_124, y = var_125)[name = tensor<string, []>("input_15")];
115
+ tensor<fp32, []> var_130 = const()[name = tensor<string, []>("op_130"), val = tensor<fp32, []>(0x1.99999ap-3)];
116
+ tensor<fp32, []> var_134 = const()[name = tensor<string, []>("op_134"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
117
+ tensor<fp32, [1, 1024]> h_9 = linear(bias = F0_1_norm1_fc_bias, weight = F0_1_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_2")];
118
+ tensor<int32, [3]> var_166 = const()[name = tensor<string, []>("op_166"), val = tensor<int32, [3]>([1, 1024, 1])];
119
+ tensor<fp32, [1, 1024, 1]> h_11 = reshape(shape = var_166, x = h_9)[name = tensor<string, []>("h_11")];
120
+ tensor<int32, [2]> var_168_split_sizes_0 = const()[name = tensor<string, []>("op_168_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
121
+ tensor<int32, []> var_168_axis_0 = const()[name = tensor<string, []>("op_168_axis_0"), val = tensor<int32, []>(1)];
122
+ tensor<fp32, [1, 512, 1]> var_168_0, tensor<fp32, [1, 512, 1]> var_168_1 = split(axis = var_168_axis_0, split_sizes = var_168_split_sizes_0, x = h_11)[name = tensor<string, []>("op_168")];
123
+ tensor<fp32, []> var_170_promoted = const()[name = tensor<string, []>("op_170_promoted"), val = tensor<fp32, []>(0x1p+0)];
124
+ tensor<fp32, [1, 512, 1]> var_171 = add(x = var_168_0, y = var_170_promoted)[name = tensor<string, []>("op_171")];
125
+ tensor<fp32, [1, 512, ?]> var_174 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_134, gamma = F0_0_norm1_norm_weight, x = input_15)[name = tensor<string, []>("op_174")];
126
+ tensor<fp32, [1, 512, ?]> var_175 = mul(x = var_171, y = var_174)[name = tensor<string, []>("op_175")];
127
+ tensor<fp32, [1, 512, ?]> input_17 = add(x = var_175, y = var_168_1)[name = tensor<string, []>("input_17")];
128
+ tensor<fp32, [1, 512, ?]> input_19 = leaky_relu(alpha = var_130, x = input_17)[name = tensor<string, []>("input_19")];
129
+ tensor<fp32, [512, 1, 3]> var_178 = const()[name = tensor<string, []>("op_178"), val = tensor<fp32, [512, 1, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4686336)))];
130
+ tensor<string, []> conv_transpose_0_pad_type_0 = const()[name = tensor<string, []>("conv_transpose_0_pad_type_0"), val = tensor<string, []>("custom")];
131
+ tensor<int32, [2]> conv_transpose_0_pad_0 = const()[name = tensor<string, []>("conv_transpose_0_pad_0"), val = tensor<int32, [2]>([0, 0])];
132
+ tensor<int32, [1]> conv_transpose_0_strides_0 = const()[name = tensor<string, []>("conv_transpose_0_strides_0"), val = tensor<int32, [1]>([2])];
133
+ tensor<int32, []> conv_transpose_0_groups_0 = const()[name = tensor<string, []>("conv_transpose_0_groups_0"), val = tensor<int32, []>(512)];
134
+ tensor<int32, [1]> conv_transpose_0_dilations_0 = const()[name = tensor<string, []>("conv_transpose_0_dilations_0"), val = tensor<int32, [1]>([1])];
135
+ tensor<fp32, [1, 512, ?]> conv_transpose_0 = conv_transpose(bias = F0_1_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 = var_178, x = input_19)[name = tensor<string, []>("conv_transpose_0")];
136
+ tensor<int32, [3]> input_21_begin_0 = const()[name = tensor<string, []>("input_21_begin_0"), val = tensor<int32, [3]>([0, 0, 1])];
137
+ tensor<int32, [3]> input_21_end_0 = const()[name = tensor<string, []>("input_21_end_0"), val = tensor<int32, [3]>([0, 0, 0])];
138
+ tensor<bool, [3]> input_21_begin_mask_0 = const()[name = tensor<string, []>("input_21_begin_mask_0"), val = tensor<bool, [3]>([true, true, false])];
139
+ tensor<bool, [3]> input_21_end_mask_0 = const()[name = tensor<string, []>("input_21_end_mask_0"), val = tensor<bool, [3]>([true, true, true])];
140
+ tensor<fp32, [1, 512, ?]> input_21 = slice_by_index(begin = input_21_begin_0, begin_mask = input_21_begin_mask_0, end = input_21_end_0, end_mask = input_21_end_mask_0, x = conv_transpose_0)[name = tensor<string, []>("input_21")];
141
+ tensor<fp32, [256, 512, 3]> weight_11_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [393216]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4692544))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5085824))), name = tensor<string, []>("weight_11_palettized"), shape = tensor<uint32, [3]>([256, 512, 3])];
142
+ tensor<string, []> input_23_pad_type_0 = const()[name = tensor<string, []>("input_23_pad_type_0"), val = tensor<string, []>("custom")];
143
+ tensor<int32, [2]> input_23_pad_0 = const()[name = tensor<string, []>("input_23_pad_0"), val = tensor<int32, [2]>([1, 1])];
144
+ tensor<int32, [1]> input_23_strides_0 = const()[name = tensor<string, []>("input_23_strides_0"), val = tensor<int32, [1]>([1])];
145
+ tensor<int32, [1]> input_23_dilations_0 = const()[name = tensor<string, []>("input_23_dilations_0"), val = tensor<int32, [1]>([1])];
146
+ tensor<int32, []> input_23_groups_0 = const()[name = tensor<string, []>("input_23_groups_0"), val = tensor<int32, []>(1)];
147
+ tensor<fp32, [1, 256, ?]> input_23 = conv(bias = F0_1_conv1_bias, dilations = input_23_dilations_0, groups = input_23_groups_0, pad = input_23_pad_0, pad_type = input_23_pad_type_0, strides = input_23_strides_0, weight = weight_11_palettized, x = input_21)[name = tensor<string, []>("input_23")];
148
+ tensor<fp32, [1, 512]> h_13 = linear(bias = F0_1_norm2_fc_bias, weight = F0_1_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_3")];
149
+ tensor<int32, [3]> var_199 = const()[name = tensor<string, []>("op_199"), val = tensor<int32, [3]>([1, 512, 1])];
150
+ tensor<fp32, [1, 512, 1]> h_15 = reshape(shape = var_199, x = h_13)[name = tensor<string, []>("h_15")];
151
+ tensor<int32, [2]> var_201_split_sizes_0 = const()[name = tensor<string, []>("op_201_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
152
+ tensor<int32, []> var_201_axis_0 = const()[name = tensor<string, []>("op_201_axis_0"), val = tensor<int32, []>(1)];
153
+ tensor<fp32, [1, 256, 1]> var_201_0, tensor<fp32, [1, 256, 1]> var_201_1 = split(axis = var_201_axis_0, split_sizes = var_201_split_sizes_0, x = h_15)[name = tensor<string, []>("op_201")];
154
+ tensor<fp32, []> var_203_promoted = const()[name = tensor<string, []>("op_203_promoted"), val = tensor<fp32, []>(0x1p+0)];
155
+ tensor<fp32, [1, 256, 1]> var_204 = add(x = var_201_0, y = var_203_promoted)[name = tensor<string, []>("op_204")];
156
+ tensor<fp32, [1, 256, ?]> var_207 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_134, gamma = F0_1_norm2_norm_weight, x = input_23)[name = tensor<string, []>("op_207")];
157
+ tensor<fp32, [1, 256, ?]> var_208 = mul(x = var_204, y = var_207)[name = tensor<string, []>("op_208")];
158
+ tensor<fp32, [1, 256, ?]> input_25 = add(x = var_208, y = var_201_1)[name = tensor<string, []>("input_25")];
159
+ tensor<fp32, [1, 256, ?]> input_27 = leaky_relu(alpha = var_130, x = input_25)[name = tensor<string, []>("input_27")];
160
+ tensor<fp32, [256, 256, 3]> weight_15_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5086912))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5283584))), name = tensor<string, []>("weight_15_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
161
+ tensor<string, []> out_3_pad_type_0 = const()[name = tensor<string, []>("out_3_pad_type_0"), val = tensor<string, []>("custom")];
162
+ tensor<int32, [2]> out_3_pad_0 = const()[name = tensor<string, []>("out_3_pad_0"), val = tensor<int32, [2]>([1, 1])];
163
+ tensor<int32, [1]> out_3_strides_0 = const()[name = tensor<string, []>("out_3_strides_0"), val = tensor<int32, [1]>([1])];
164
+ tensor<int32, [1]> out_3_dilations_0 = const()[name = tensor<string, []>("out_3_dilations_0"), val = tensor<int32, [1]>([1])];
165
+ tensor<int32, []> out_3_groups_0 = const()[name = tensor<string, []>("out_3_groups_0"), val = tensor<int32, []>(1)];
166
+ tensor<fp32, [1, 256, ?]> out_3 = conv(bias = F0_1_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_27)[name = tensor<string, []>("out_3")];
167
+ tensor<int32, [1]> expand_dims_0_axes_0 = const()[name = tensor<string, []>("expand_dims_0_axes_0"), val = tensor<int32, [1]>([3])];
168
+ tensor<fp32, [1, 512, ?, 1]> expand_dims_0 = expand_dims(axes = expand_dims_0_axes_0, x = input_15)[name = tensor<string, []>("expand_dims_0")];
169
+ tensor<int32, []> upsample_nearest_neighbor_0_scale_factor_height_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_0_scale_factor_height_0"), val = tensor<int32, []>(2)];
170
+ tensor<int32, []> upsample_nearest_neighbor_0_scale_factor_width_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_0_scale_factor_width_0"), val = tensor<int32, []>(1)];
171
+ tensor<fp32, [1, 512, ?, 1]> 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<string, []>("upsample_nearest_neighbor_0")];
172
+ tensor<int32, [1]> input_29_axes_0 = const()[name = tensor<string, []>("input_29_axes_0"), val = tensor<int32, [1]>([3])];
173
+ tensor<fp32, [1, 512, ?]> input_29 = squeeze(axes = input_29_axes_0, x = upsample_nearest_neighbor_0)[name = tensor<string, []>("input_29")];
174
+ tensor<fp32, [256, 512, 1]> weight_17_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5284672))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5415808))), name = tensor<string, []>("weight_17_palettized"), shape = tensor<uint32, [3]>([256, 512, 1])];
175
+ tensor<string, []> var_225_pad_type_0 = const()[name = tensor<string, []>("op_225_pad_type_0"), val = tensor<string, []>("valid")];
176
+ tensor<int32, [1]> var_225_strides_0 = const()[name = tensor<string, []>("op_225_strides_0"), val = tensor<int32, [1]>([1])];
177
+ tensor<int32, [2]> var_225_pad_0 = const()[name = tensor<string, []>("op_225_pad_0"), val = tensor<int32, [2]>([0, 0])];
178
+ tensor<int32, [1]> var_225_dilations_0 = const()[name = tensor<string, []>("op_225_dilations_0"), val = tensor<int32, [1]>([1])];
179
+ tensor<int32, []> var_225_groups_0 = const()[name = tensor<string, []>("op_225_groups_0"), val = tensor<int32, []>(1)];
180
+ tensor<fp32, [1, 256, ?]> var_225 = conv(dilations = var_225_dilations_0, groups = var_225_groups_0, pad = var_225_pad_0, pad_type = var_225_pad_type_0, strides = var_225_strides_0, weight = weight_17_palettized, x = input_29)[name = tensor<string, []>("op_225")];
181
+ tensor<fp32, [1, 256, ?]> var_226 = add(x = out_3, y = var_225)[name = tensor<string, []>("op_226")];
182
+ tensor<fp32, []> var_227 = const()[name = tensor<string, []>("op_227"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
183
+ tensor<fp32, [1, 256, ?]> input_31 = mul(x = var_226, y = var_227)[name = tensor<string, []>("input_31")];
184
+ tensor<fp32, []> var_230 = const()[name = tensor<string, []>("op_230"), val = tensor<fp32, []>(0x1.99999ap-3)];
185
+ tensor<fp32, []> var_233 = const()[name = tensor<string, []>("op_233"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
186
+ tensor<fp32, [1, 512]> h_17 = linear(bias = F0_2_norm1_fc_bias, weight = F0_2_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_4")];
187
+ tensor<int32, [3]> var_256 = const()[name = tensor<string, []>("op_256"), val = tensor<int32, [3]>([1, 512, 1])];
188
+ tensor<fp32, [1, 512, 1]> h_19 = reshape(shape = var_256, x = h_17)[name = tensor<string, []>("h_19")];
189
+ tensor<int32, [2]> var_258_split_sizes_0 = const()[name = tensor<string, []>("op_258_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
190
+ tensor<int32, []> var_258_axis_0 = const()[name = tensor<string, []>("op_258_axis_0"), val = tensor<int32, []>(1)];
191
+ tensor<fp32, [1, 256, 1]> var_258_0, tensor<fp32, [1, 256, 1]> var_258_1 = split(axis = var_258_axis_0, split_sizes = var_258_split_sizes_0, x = h_19)[name = tensor<string, []>("op_258")];
192
+ tensor<fp32, []> var_260_promoted = const()[name = tensor<string, []>("op_260_promoted"), val = tensor<fp32, []>(0x1p+0)];
193
+ tensor<fp32, [1, 256, 1]> var_261 = add(x = var_258_0, y = var_260_promoted)[name = tensor<string, []>("op_261")];
194
+ tensor<fp32, [1, 256, ?]> var_264 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_233, gamma = F0_1_norm2_norm_weight, x = input_31)[name = tensor<string, []>("op_264")];
195
+ tensor<fp32, [1, 256, ?]> var_265 = mul(x = var_261, y = var_264)[name = tensor<string, []>("op_265")];
196
+ tensor<fp32, [1, 256, ?]> input_33 = add(x = var_265, y = var_258_1)[name = tensor<string, []>("input_33")];
197
+ tensor<fp32, [1, 256, ?]> input_35 = leaky_relu(alpha = var_230, x = input_33)[name = tensor<string, []>("input_35")];
198
+ tensor<fp32, [256, 256, 3]> weight_21_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5416896))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5613568))), name = tensor<string, []>("weight_21_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
199
+ tensor<string, []> input_37_pad_type_0 = const()[name = tensor<string, []>("input_37_pad_type_0"), val = tensor<string, []>("custom")];
200
+ tensor<int32, [2]> input_37_pad_0 = const()[name = tensor<string, []>("input_37_pad_0"), val = tensor<int32, [2]>([1, 1])];
201
+ tensor<int32, [1]> input_37_strides_0 = const()[name = tensor<string, []>("input_37_strides_0"), val = tensor<int32, [1]>([1])];
202
+ tensor<int32, [1]> input_37_dilations_0 = const()[name = tensor<string, []>("input_37_dilations_0"), val = tensor<int32, [1]>([1])];
203
+ tensor<int32, []> input_37_groups_0 = const()[name = tensor<string, []>("input_37_groups_0"), val = tensor<int32, []>(1)];
204
+ tensor<fp32, [1, 256, ?]> input_37 = conv(bias = F0_2_conv1_bias, dilations = input_37_dilations_0, groups = input_37_groups_0, pad = input_37_pad_0, pad_type = input_37_pad_type_0, strides = input_37_strides_0, weight = weight_21_palettized, x = input_35)[name = tensor<string, []>("input_37")];
205
+ tensor<fp32, [1, 512]> h_21 = linear(bias = F0_2_norm2_fc_bias, weight = F0_2_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_5")];
206
+ tensor<int32, [3]> var_282 = const()[name = tensor<string, []>("op_282"), val = tensor<int32, [3]>([1, 512, 1])];
207
+ tensor<fp32, [1, 512, 1]> h_23 = reshape(shape = var_282, x = h_21)[name = tensor<string, []>("h_23")];
208
+ tensor<int32, [2]> var_284_split_sizes_0 = const()[name = tensor<string, []>("op_284_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
209
+ tensor<int32, []> var_284_axis_0 = const()[name = tensor<string, []>("op_284_axis_0"), val = tensor<int32, []>(1)];
210
+ tensor<fp32, [1, 256, 1]> var_284_0, tensor<fp32, [1, 256, 1]> var_284_1 = split(axis = var_284_axis_0, split_sizes = var_284_split_sizes_0, x = h_23)[name = tensor<string, []>("op_284")];
211
+ tensor<fp32, []> var_286_promoted = const()[name = tensor<string, []>("op_286_promoted"), val = tensor<fp32, []>(0x1p+0)];
212
+ tensor<fp32, [1, 256, 1]> var_287 = add(x = var_284_0, y = var_286_promoted)[name = tensor<string, []>("op_287")];
213
+ tensor<fp32, [1, 256, ?]> var_290 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_233, gamma = F0_1_norm2_norm_weight, x = input_37)[name = tensor<string, []>("op_290")];
214
+ tensor<fp32, [1, 256, ?]> var_291 = mul(x = var_287, y = var_290)[name = tensor<string, []>("op_291")];
215
+ tensor<fp32, [1, 256, ?]> input_39 = add(x = var_291, y = var_284_1)[name = tensor<string, []>("input_39")];
216
+ tensor<fp32, [1, 256, ?]> input_41 = leaky_relu(alpha = var_230, x = input_39)[name = tensor<string, []>("input_41")];
217
+ tensor<fp32, [256, 256, 3]> weight_25_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5614656))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5811328))), name = tensor<string, []>("weight_25_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
218
+ tensor<string, []> out_5_pad_type_0 = const()[name = tensor<string, []>("out_5_pad_type_0"), val = tensor<string, []>("custom")];
219
+ tensor<int32, [2]> out_5_pad_0 = const()[name = tensor<string, []>("out_5_pad_0"), val = tensor<int32, [2]>([1, 1])];
220
+ tensor<int32, [1]> out_5_strides_0 = const()[name = tensor<string, []>("out_5_strides_0"), val = tensor<int32, [1]>([1])];
221
+ tensor<int32, [1]> out_5_dilations_0 = const()[name = tensor<string, []>("out_5_dilations_0"), val = tensor<int32, [1]>([1])];
222
+ tensor<int32, []> out_5_groups_0 = const()[name = tensor<string, []>("out_5_groups_0"), val = tensor<int32, []>(1)];
223
+ tensor<fp32, [1, 256, ?]> out_5 = conv(bias = F0_2_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_25_palettized, x = input_41)[name = tensor<string, []>("out_5")];
224
+ tensor<fp32, [1, 256, ?]> var_301 = add(x = out_5, y = input_31)[name = tensor<string, []>("op_301")];
225
+ tensor<fp32, []> var_302 = const()[name = tensor<string, []>("op_302"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
226
+ tensor<fp32, [1, 256, ?]> input_43 = mul(x = var_301, y = var_302)[name = tensor<string, []>("input_43")];
227
+ tensor<string, []> var_314_pad_type_0 = const()[name = tensor<string, []>("op_314_pad_type_0"), val = tensor<string, []>("valid")];
228
+ tensor<int32, [1]> var_314_strides_0 = const()[name = tensor<string, []>("op_314_strides_0"), val = tensor<int32, [1]>([1])];
229
+ tensor<int32, [2]> var_314_pad_0 = const()[name = tensor<string, []>("op_314_pad_0"), val = tensor<int32, [2]>([0, 0])];
230
+ tensor<int32, [1]> var_314_dilations_0 = const()[name = tensor<string, []>("op_314_dilations_0"), val = tensor<int32, [1]>([1])];
231
+ tensor<int32, []> var_314_groups_0 = const()[name = tensor<string, []>("op_314_groups_0"), val = tensor<int32, []>(1)];
232
+ tensor<fp32, [1, 1, ?]> var_314 = conv(bias = F0_proj_bias, dilations = var_314_dilations_0, groups = var_314_groups_0, pad = var_314_pad_0, pad_type = var_314_pad_type_0, strides = var_314_strides_0, weight = F0_proj_weight, x = input_43)[name = tensor<string, []>("op_314")];
233
+ tensor<int32, [1]> var_316_axes_0 = const()[name = tensor<string, []>("op_316_axes_0"), val = tensor<int32, [1]>([1])];
234
+ tensor<fp32, [1, ?]> F0 = squeeze(axes = var_316_axes_0, x = var_314)[name = tensor<string, []>("op_316")];
235
+ tensor<fp32, []> var_321 = const()[name = tensor<string, []>("op_321"), val = tensor<fp32, []>(0x1.99999ap-3)];
236
+ tensor<fp32, []> var_324 = const()[name = tensor<string, []>("op_324"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
237
+ tensor<fp32, [1, 1024]> h_25 = linear(bias = N_0_norm1_fc_bias, weight = N_0_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_6")];
238
+ tensor<int32, [3]> var_347 = const()[name = tensor<string, []>("op_347"), val = tensor<int32, [3]>([1, 1024, 1])];
239
+ tensor<fp32, [1, 1024, 1]> h_27 = reshape(shape = var_347, x = h_25)[name = tensor<string, []>("h_27")];
240
+ tensor<int32, [2]> var_349_split_sizes_0 = const()[name = tensor<string, []>("op_349_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
241
+ tensor<int32, []> var_349_axis_0 = const()[name = tensor<string, []>("op_349_axis_0"), val = tensor<int32, []>(1)];
242
+ tensor<fp32, [1, 512, 1]> var_349_0, tensor<fp32, [1, 512, 1]> var_349_1 = split(axis = var_349_axis_0, split_sizes = var_349_split_sizes_0, x = h_27)[name = tensor<string, []>("op_349")];
243
+ tensor<fp32, []> var_351_promoted = const()[name = tensor<string, []>("op_351_promoted"), val = tensor<fp32, []>(0x1p+0)];
244
+ tensor<fp32, [1, 512, 1]> var_352 = add(x = var_349_0, y = var_351_promoted)[name = tensor<string, []>("op_352")];
245
+ tensor<fp32, [1, 512, ?]> var_356 = mul(x = var_352, y = var_87)[name = tensor<string, []>("op_356")];
246
+ tensor<fp32, [1, 512, ?]> input_47 = add(x = var_356, y = var_349_1)[name = tensor<string, []>("input_47")];
247
+ tensor<fp32, [1, 512, ?]> input_49 = leaky_relu(alpha = var_321, x = input_47)[name = tensor<string, []>("input_49")];
248
+ tensor<fp32, [512, 512, 3]> weight_31_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5812416))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(6598912))), name = tensor<string, []>("weight_31_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
249
+ tensor<string, []> input_51_pad_type_0 = const()[name = tensor<string, []>("input_51_pad_type_0"), val = tensor<string, []>("custom")];
250
+ tensor<int32, [2]> input_51_pad_0 = const()[name = tensor<string, []>("input_51_pad_0"), val = tensor<int32, [2]>([1, 1])];
251
+ tensor<int32, [1]> input_51_strides_0 = const()[name = tensor<string, []>("input_51_strides_0"), val = tensor<int32, [1]>([1])];
252
+ tensor<int32, [1]> input_51_dilations_0 = const()[name = tensor<string, []>("input_51_dilations_0"), val = tensor<int32, [1]>([1])];
253
+ tensor<int32, []> input_51_groups_0 = const()[name = tensor<string, []>("input_51_groups_0"), val = tensor<int32, []>(1)];
254
+ tensor<fp32, [1, 512, ?]> input_51 = conv(bias = N_0_conv1_bias, dilations = input_51_dilations_0, groups = input_51_groups_0, pad = input_51_pad_0, pad_type = input_51_pad_type_0, strides = input_51_strides_0, weight = weight_31_palettized, x = input_49)[name = tensor<string, []>("input_51")];
255
+ tensor<fp32, [1, 1024]> h_29 = linear(bias = N_0_norm2_fc_bias, weight = N_0_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_7")];
256
+ tensor<int32, [3]> var_373 = const()[name = tensor<string, []>("op_373"), val = tensor<int32, [3]>([1, 1024, 1])];
257
+ tensor<fp32, [1, 1024, 1]> h_31 = reshape(shape = var_373, x = h_29)[name = tensor<string, []>("h_31")];
258
+ tensor<int32, [2]> var_375_split_sizes_0 = const()[name = tensor<string, []>("op_375_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
259
+ tensor<int32, []> var_375_axis_0 = const()[name = tensor<string, []>("op_375_axis_0"), val = tensor<int32, []>(1)];
260
+ tensor<fp32, [1, 512, 1]> var_375_0, tensor<fp32, [1, 512, 1]> var_375_1 = split(axis = var_375_axis_0, split_sizes = var_375_split_sizes_0, x = h_31)[name = tensor<string, []>("op_375")];
261
+ tensor<fp32, []> var_377_promoted = const()[name = tensor<string, []>("op_377_promoted"), val = tensor<fp32, []>(0x1p+0)];
262
+ tensor<fp32, [1, 512, 1]> var_378 = add(x = var_375_0, y = var_377_promoted)[name = tensor<string, []>("op_378")];
263
+ tensor<fp32, [1, 512, ?]> var_381 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_324, gamma = F0_0_norm1_norm_weight, x = input_51)[name = tensor<string, []>("op_381")];
264
+ tensor<fp32, [1, 512, ?]> var_382 = mul(x = var_378, y = var_381)[name = tensor<string, []>("op_382")];
265
+ tensor<fp32, [1, 512, ?]> input_53 = add(x = var_382, y = var_375_1)[name = tensor<string, []>("input_53")];
266
+ tensor<fp32, [1, 512, ?]> input_55 = leaky_relu(alpha = var_321, x = input_53)[name = tensor<string, []>("input_55")];
267
+ tensor<fp32, [512, 512, 3]> weight_35_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(6600000))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7386496))), name = tensor<string, []>("weight_35_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
268
+ tensor<string, []> out_7_pad_type_0 = const()[name = tensor<string, []>("out_7_pad_type_0"), val = tensor<string, []>("custom")];
269
+ tensor<int32, [2]> out_7_pad_0 = const()[name = tensor<string, []>("out_7_pad_0"), val = tensor<int32, [2]>([1, 1])];
270
+ tensor<int32, [1]> out_7_strides_0 = const()[name = tensor<string, []>("out_7_strides_0"), val = tensor<int32, [1]>([1])];
271
+ tensor<int32, [1]> out_7_dilations_0 = const()[name = tensor<string, []>("out_7_dilations_0"), val = tensor<int32, [1]>([1])];
272
+ tensor<int32, []> out_7_groups_0 = const()[name = tensor<string, []>("out_7_groups_0"), val = tensor<int32, []>(1)];
273
+ tensor<fp32, [1, 512, ?]> out_7 = conv(bias = N_0_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_35_palettized, x = input_55)[name = tensor<string, []>("out_7")];
274
+ tensor<fp32, [1, 512, ?]> var_392 = add(x = out_7, y = transpose_0_1)[name = tensor<string, []>("op_392")];
275
+ tensor<fp32, []> var_393 = const()[name = tensor<string, []>("op_393"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
276
+ tensor<fp32, [1, 512, ?]> input_57 = mul(x = var_392, y = var_393)[name = tensor<string, []>("input_57")];
277
+ tensor<fp32, []> var_398 = const()[name = tensor<string, []>("op_398"), val = tensor<fp32, []>(0x1.99999ap-3)];
278
+ tensor<fp32, []> var_402 = const()[name = tensor<string, []>("op_402"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
279
+ tensor<fp32, [1, 1024]> h_33 = linear(bias = N_1_norm1_fc_bias, weight = N_1_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_8")];
280
+ tensor<int32, [3]> var_434 = const()[name = tensor<string, []>("op_434"), val = tensor<int32, [3]>([1, 1024, 1])];
281
+ tensor<fp32, [1, 1024, 1]> h_35 = reshape(shape = var_434, x = h_33)[name = tensor<string, []>("h_35")];
282
+ tensor<int32, [2]> var_436_split_sizes_0 = const()[name = tensor<string, []>("op_436_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
283
+ tensor<int32, []> var_436_axis_0 = const()[name = tensor<string, []>("op_436_axis_0"), val = tensor<int32, []>(1)];
284
+ tensor<fp32, [1, 512, 1]> var_436_0, tensor<fp32, [1, 512, 1]> var_436_1 = split(axis = var_436_axis_0, split_sizes = var_436_split_sizes_0, x = h_35)[name = tensor<string, []>("op_436")];
285
+ tensor<fp32, []> var_438_promoted = const()[name = tensor<string, []>("op_438_promoted"), val = tensor<fp32, []>(0x1p+0)];
286
+ tensor<fp32, [1, 512, 1]> var_439 = add(x = var_436_0, y = var_438_promoted)[name = tensor<string, []>("op_439")];
287
+ tensor<fp32, [1, 512, ?]> var_442 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_402, gamma = F0_0_norm1_norm_weight, x = input_57)[name = tensor<string, []>("op_442")];
288
+ tensor<fp32, [1, 512, ?]> var_443 = mul(x = var_439, y = var_442)[name = tensor<string, []>("op_443")];
289
+ tensor<fp32, [1, 512, ?]> input_59 = add(x = var_443, y = var_436_1)[name = tensor<string, []>("input_59")];
290
+ tensor<fp32, [1, 512, ?]> input_61 = leaky_relu(alpha = var_398, x = input_59)[name = tensor<string, []>("input_61")];
291
+ tensor<fp32, [512, 1, 3]> var_446 = const()[name = tensor<string, []>("op_446"), val = tensor<fp32, [512, 1, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7387584)))];
292
+ tensor<string, []> conv_transpose_1_pad_type_0 = const()[name = tensor<string, []>("conv_transpose_1_pad_type_0"), val = tensor<string, []>("custom")];
293
+ tensor<int32, [2]> conv_transpose_1_pad_0 = const()[name = tensor<string, []>("conv_transpose_1_pad_0"), val = tensor<int32, [2]>([0, 0])];
294
+ tensor<int32, [1]> conv_transpose_1_strides_0 = const()[name = tensor<string, []>("conv_transpose_1_strides_0"), val = tensor<int32, [1]>([2])];
295
+ tensor<int32, []> conv_transpose_1_groups_0 = const()[name = tensor<string, []>("conv_transpose_1_groups_0"), val = tensor<int32, []>(512)];
296
+ tensor<int32, [1]> conv_transpose_1_dilations_0 = const()[name = tensor<string, []>("conv_transpose_1_dilations_0"), val = tensor<int32, [1]>([1])];
297
+ tensor<fp32, [1, 512, ?]> conv_transpose_1 = conv_transpose(bias = N_1_pool_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 = var_446, x = input_61)[name = tensor<string, []>("conv_transpose_1")];
298
+ tensor<int32, [3]> input_63_begin_0 = const()[name = tensor<string, []>("input_63_begin_0"), val = tensor<int32, [3]>([0, 0, 1])];
299
+ tensor<int32, [3]> input_63_end_0 = const()[name = tensor<string, []>("input_63_end_0"), val = tensor<int32, [3]>([0, 0, 0])];
300
+ tensor<bool, [3]> input_63_begin_mask_0 = const()[name = tensor<string, []>("input_63_begin_mask_0"), val = tensor<bool, [3]>([true, true, false])];
301
+ tensor<bool, [3]> input_63_end_mask_0 = const()[name = tensor<string, []>("input_63_end_mask_0"), val = tensor<bool, [3]>([true, true, true])];
302
+ tensor<fp32, [1, 512, ?]> input_63 = slice_by_index(begin = input_63_begin_0, begin_mask = input_63_begin_mask_0, end = input_63_end_0, end_mask = input_63_end_mask_0, x = conv_transpose_1)[name = tensor<string, []>("input_63")];
303
+ tensor<fp32, [256, 512, 3]> weight_39_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [393216]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7393792))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7787072))), name = tensor<string, []>("weight_39_palettized"), shape = tensor<uint32, [3]>([256, 512, 3])];
304
+ tensor<string, []> input_65_pad_type_0 = const()[name = tensor<string, []>("input_65_pad_type_0"), val = tensor<string, []>("custom")];
305
+ tensor<int32, [2]> input_65_pad_0 = const()[name = tensor<string, []>("input_65_pad_0"), val = tensor<int32, [2]>([1, 1])];
306
+ tensor<int32, [1]> input_65_strides_0 = const()[name = tensor<string, []>("input_65_strides_0"), val = tensor<int32, [1]>([1])];
307
+ tensor<int32, [1]> input_65_dilations_0 = const()[name = tensor<string, []>("input_65_dilations_0"), val = tensor<int32, [1]>([1])];
308
+ tensor<int32, []> input_65_groups_0 = const()[name = tensor<string, []>("input_65_groups_0"), val = tensor<int32, []>(1)];
309
+ tensor<fp32, [1, 256, ?]> input_65 = conv(bias = N_1_conv1_bias, 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_39_palettized, x = input_63)[name = tensor<string, []>("input_65")];
310
+ tensor<fp32, [1, 512]> h_37 = linear(bias = N_1_norm2_fc_bias, weight = N_1_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_9")];
311
+ tensor<int32, [3]> var_467 = const()[name = tensor<string, []>("op_467"), val = tensor<int32, [3]>([1, 512, 1])];
312
+ tensor<fp32, [1, 512, 1]> h_39 = reshape(shape = var_467, x = h_37)[name = tensor<string, []>("h_39")];
313
+ tensor<int32, [2]> var_469_split_sizes_0 = const()[name = tensor<string, []>("op_469_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
314
+ tensor<int32, []> var_469_axis_0 = const()[name = tensor<string, []>("op_469_axis_0"), val = tensor<int32, []>(1)];
315
+ tensor<fp32, [1, 256, 1]> var_469_0, tensor<fp32, [1, 256, 1]> var_469_1 = split(axis = var_469_axis_0, split_sizes = var_469_split_sizes_0, x = h_39)[name = tensor<string, []>("op_469")];
316
+ tensor<fp32, []> var_471_promoted = const()[name = tensor<string, []>("op_471_promoted"), val = tensor<fp32, []>(0x1p+0)];
317
+ tensor<fp32, [1, 256, 1]> var_472 = add(x = var_469_0, y = var_471_promoted)[name = tensor<string, []>("op_472")];
318
+ tensor<fp32, [1, 256, ?]> var_475 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_402, gamma = F0_1_norm2_norm_weight, x = input_65)[name = tensor<string, []>("op_475")];
319
+ tensor<fp32, [1, 256, ?]> var_476 = mul(x = var_472, y = var_475)[name = tensor<string, []>("op_476")];
320
+ tensor<fp32, [1, 256, ?]> input_67 = add(x = var_476, y = var_469_1)[name = tensor<string, []>("input_67")];
321
+ tensor<fp32, [1, 256, ?]> input_69 = leaky_relu(alpha = var_398, x = input_67)[name = tensor<string, []>("input_69")];
322
+ tensor<fp32, [256, 256, 3]> weight_43_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7788160))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7984832))), name = tensor<string, []>("weight_43_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
323
+ tensor<string, []> out_9_pad_type_0 = const()[name = tensor<string, []>("out_9_pad_type_0"), val = tensor<string, []>("custom")];
324
+ tensor<int32, [2]> out_9_pad_0 = const()[name = tensor<string, []>("out_9_pad_0"), val = tensor<int32, [2]>([1, 1])];
325
+ tensor<int32, [1]> out_9_strides_0 = const()[name = tensor<string, []>("out_9_strides_0"), val = tensor<int32, [1]>([1])];
326
+ tensor<int32, [1]> out_9_dilations_0 = const()[name = tensor<string, []>("out_9_dilations_0"), val = tensor<int32, [1]>([1])];
327
+ tensor<int32, []> out_9_groups_0 = const()[name = tensor<string, []>("out_9_groups_0"), val = tensor<int32, []>(1)];
328
+ tensor<fp32, [1, 256, ?]> out_9 = conv(bias = N_1_conv2_bias, 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_43_palettized, x = input_69)[name = tensor<string, []>("out_9")];
329
+ tensor<int32, [1]> expand_dims_1_axes_0 = const()[name = tensor<string, []>("expand_dims_1_axes_0"), val = tensor<int32, [1]>([3])];
330
+ tensor<fp32, [1, 512, ?, 1]> expand_dims_1 = expand_dims(axes = expand_dims_1_axes_0, x = input_57)[name = tensor<string, []>("expand_dims_1")];
331
+ tensor<int32, []> upsample_nearest_neighbor_1_scale_factor_height_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_1_scale_factor_height_0"), val = tensor<int32, []>(2)];
332
+ tensor<int32, []> upsample_nearest_neighbor_1_scale_factor_width_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_1_scale_factor_width_0"), val = tensor<int32, []>(1)];
333
+ tensor<fp32, [1, 512, ?, 1]> upsample_nearest_neighbor_1 = 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)[name = tensor<string, []>("upsample_nearest_neighbor_1")];
334
+ tensor<int32, [1]> input_71_axes_0 = const()[name = tensor<string, []>("input_71_axes_0"), val = tensor<int32, [1]>([3])];
335
+ tensor<fp32, [1, 512, ?]> input_71 = squeeze(axes = input_71_axes_0, x = upsample_nearest_neighbor_1)[name = tensor<string, []>("input_71")];
336
+ tensor<fp32, [256, 512, 1]> weight_45_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7985920))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8117056))), name = tensor<string, []>("weight_45_palettized"), shape = tensor<uint32, [3]>([256, 512, 1])];
337
+ tensor<string, []> var_493_pad_type_0 = const()[name = tensor<string, []>("op_493_pad_type_0"), val = tensor<string, []>("valid")];
338
+ tensor<int32, [1]> var_493_strides_0 = const()[name = tensor<string, []>("op_493_strides_0"), val = tensor<int32, [1]>([1])];
339
+ tensor<int32, [2]> var_493_pad_0 = const()[name = tensor<string, []>("op_493_pad_0"), val = tensor<int32, [2]>([0, 0])];
340
+ tensor<int32, [1]> var_493_dilations_0 = const()[name = tensor<string, []>("op_493_dilations_0"), val = tensor<int32, [1]>([1])];
341
+ tensor<int32, []> var_493_groups_0 = const()[name = tensor<string, []>("op_493_groups_0"), val = tensor<int32, []>(1)];
342
+ tensor<fp32, [1, 256, ?]> var_493 = conv(dilations = var_493_dilations_0, groups = var_493_groups_0, pad = var_493_pad_0, pad_type = var_493_pad_type_0, strides = var_493_strides_0, weight = weight_45_palettized, x = input_71)[name = tensor<string, []>("op_493")];
343
+ tensor<fp32, [1, 256, ?]> var_494 = add(x = out_9, y = var_493)[name = tensor<string, []>("op_494")];
344
+ tensor<fp32, []> var_495 = const()[name = tensor<string, []>("op_495"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
345
+ tensor<fp32, [1, 256, ?]> input_73 = mul(x = var_494, y = var_495)[name = tensor<string, []>("input_73")];
346
+ tensor<fp32, []> var_498 = const()[name = tensor<string, []>("op_498"), val = tensor<fp32, []>(0x1.99999ap-3)];
347
+ tensor<fp32, []> var_501 = const()[name = tensor<string, []>("op_501"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
348
+ tensor<fp32, [1, 512]> h_41 = linear(bias = N_2_norm1_fc_bias, weight = N_2_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_10")];
349
+ tensor<int32, [3]> var_524 = const()[name = tensor<string, []>("op_524"), val = tensor<int32, [3]>([1, 512, 1])];
350
+ tensor<fp32, [1, 512, 1]> h_43 = reshape(shape = var_524, x = h_41)[name = tensor<string, []>("h_43")];
351
+ tensor<int32, [2]> var_526_split_sizes_0 = const()[name = tensor<string, []>("op_526_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
352
+ tensor<int32, []> var_526_axis_0 = const()[name = tensor<string, []>("op_526_axis_0"), val = tensor<int32, []>(1)];
353
+ tensor<fp32, [1, 256, 1]> var_526_0, tensor<fp32, [1, 256, 1]> var_526_1 = split(axis = var_526_axis_0, split_sizes = var_526_split_sizes_0, x = h_43)[name = tensor<string, []>("op_526")];
354
+ tensor<fp32, []> var_528_promoted = const()[name = tensor<string, []>("op_528_promoted"), val = tensor<fp32, []>(0x1p+0)];
355
+ tensor<fp32, [1, 256, 1]> var_529 = add(x = var_526_0, y = var_528_promoted)[name = tensor<string, []>("op_529")];
356
+ tensor<fp32, [1, 256, ?]> var_532 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_501, gamma = F0_1_norm2_norm_weight, x = input_73)[name = tensor<string, []>("op_532")];
357
+ tensor<fp32, [1, 256, ?]> var_533 = mul(x = var_529, y = var_532)[name = tensor<string, []>("op_533")];
358
+ tensor<fp32, [1, 256, ?]> input_75 = add(x = var_533, y = var_526_1)[name = tensor<string, []>("input_75")];
359
+ tensor<fp32, [1, 256, ?]> input_77 = leaky_relu(alpha = var_498, x = input_75)[name = tensor<string, []>("input_77")];
360
+ tensor<fp32, [256, 256, 3]> weight_49_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8118144))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8314816))), name = tensor<string, []>("weight_49_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
361
+ tensor<string, []> input_79_pad_type_0 = const()[name = tensor<string, []>("input_79_pad_type_0"), val = tensor<string, []>("custom")];
362
+ tensor<int32, [2]> input_79_pad_0 = const()[name = tensor<string, []>("input_79_pad_0"), val = tensor<int32, [2]>([1, 1])];
363
+ tensor<int32, [1]> input_79_strides_0 = const()[name = tensor<string, []>("input_79_strides_0"), val = tensor<int32, [1]>([1])];
364
+ tensor<int32, [1]> input_79_dilations_0 = const()[name = tensor<string, []>("input_79_dilations_0"), val = tensor<int32, [1]>([1])];
365
+ tensor<int32, []> input_79_groups_0 = const()[name = tensor<string, []>("input_79_groups_0"), val = tensor<int32, []>(1)];
366
+ tensor<fp32, [1, 256, ?]> input_79 = conv(bias = N_2_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_49_palettized, x = input_77)[name = tensor<string, []>("input_79")];
367
+ tensor<fp32, [1, 512]> h_45 = linear(bias = N_2_norm2_fc_bias, weight = N_2_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_11")];
368
+ tensor<int32, [3]> var_550 = const()[name = tensor<string, []>("op_550"), val = tensor<int32, [3]>([1, 512, 1])];
369
+ tensor<fp32, [1, 512, 1]> h = reshape(shape = var_550, x = h_45)[name = tensor<string, []>("h")];
370
+ tensor<int32, [2]> var_552_split_sizes_0 = const()[name = tensor<string, []>("op_552_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
371
+ tensor<int32, []> var_552_axis_0 = const()[name = tensor<string, []>("op_552_axis_0"), val = tensor<int32, []>(1)];
372
+ tensor<fp32, [1, 256, 1]> var_552_0, tensor<fp32, [1, 256, 1]> var_552_1 = split(axis = var_552_axis_0, split_sizes = var_552_split_sizes_0, x = h)[name = tensor<string, []>("op_552")];
373
+ tensor<fp32, []> var_554_promoted = const()[name = tensor<string, []>("op_554_promoted"), val = tensor<fp32, []>(0x1p+0)];
374
+ tensor<fp32, [1, 256, 1]> var_555 = add(x = var_552_0, y = var_554_promoted)[name = tensor<string, []>("op_555")];
375
+ tensor<fp32, [1, 256, ?]> var_558 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_501, gamma = F0_1_norm2_norm_weight, x = input_79)[name = tensor<string, []>("op_558")];
376
+ tensor<fp32, [1, 256, ?]> var_559 = mul(x = var_555, y = var_558)[name = tensor<string, []>("op_559")];
377
+ tensor<fp32, [1, 256, ?]> input_81 = add(x = var_559, y = var_552_1)[name = tensor<string, []>("input_81")];
378
+ tensor<fp32, [1, 256, ?]> input_83 = leaky_relu(alpha = var_498, x = input_81)[name = tensor<string, []>("input_83")];
379
+ tensor<fp32, [256, 256, 3]> weight_53_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8315904))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8512576))), name = tensor<string, []>("weight_53_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
380
+ tensor<string, []> out_pad_type_0 = const()[name = tensor<string, []>("out_pad_type_0"), val = tensor<string, []>("custom")];
381
+ tensor<int32, [2]> out_pad_0 = const()[name = tensor<string, []>("out_pad_0"), val = tensor<int32, [2]>([1, 1])];
382
+ tensor<int32, [1]> out_strides_0 = const()[name = tensor<string, []>("out_strides_0"), val = tensor<int32, [1]>([1])];
383
+ tensor<int32, [1]> out_dilations_0 = const()[name = tensor<string, []>("out_dilations_0"), val = tensor<int32, [1]>([1])];
384
+ tensor<int32, []> out_groups_0 = const()[name = tensor<string, []>("out_groups_0"), val = tensor<int32, []>(1)];
385
+ tensor<fp32, [1, 256, ?]> out = conv(bias = N_2_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_53_palettized, x = input_83)[name = tensor<string, []>("out")];
386
+ tensor<fp32, [1, 256, ?]> var_569 = add(x = out, y = input_73)[name = tensor<string, []>("op_569")];
387
+ tensor<fp32, []> var_570 = const()[name = tensor<string, []>("op_570"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
388
+ tensor<fp32, [1, 256, ?]> input = mul(x = var_569, y = var_570)[name = tensor<string, []>("input")];
389
+ tensor<string, []> var_582_pad_type_0 = const()[name = tensor<string, []>("op_582_pad_type_0"), val = tensor<string, []>("valid")];
390
+ tensor<int32, [1]> var_582_strides_0 = const()[name = tensor<string, []>("op_582_strides_0"), val = tensor<int32, [1]>([1])];
391
+ tensor<int32, [2]> var_582_pad_0 = const()[name = tensor<string, []>("op_582_pad_0"), val = tensor<int32, [2]>([0, 0])];
392
+ tensor<int32, [1]> var_582_dilations_0 = const()[name = tensor<string, []>("op_582_dilations_0"), val = tensor<int32, [1]>([1])];
393
+ tensor<int32, []> var_582_groups_0 = const()[name = tensor<string, []>("op_582_groups_0"), val = tensor<int32, []>(1)];
394
+ tensor<fp32, [1, 1, ?]> var_582 = conv(bias = N_proj_bias, dilations = var_582_dilations_0, groups = var_582_groups_0, pad = var_582_pad_0, pad_type = var_582_pad_type_0, strides = var_582_strides_0, weight = N_proj_weight, x = input)[name = tensor<string, []>("op_582")];
395
+ tensor<int32, [1]> var_584_axes_0 = const()[name = tensor<string, []>("op_584_axes_0"), val = tensor<int32, [1]>([1])];
396
+ tensor<fp32, [1, ?]> N = squeeze(axes = var_584_axes_0, x = var_582)[name = tensor<string, []>("op_584")];
397
+ } -> (F0, N);
398
+ }
ANE-zh/KokoroProsody_v2.mlmodelc/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9835a7d16162985bbac5cde1ad206b1092e935ffc958146a4d137ab83eeae82f
3
+ size 8513664
ANE/KokoroProsody_v2.mlmodelc/analytics/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:804d7d2e9a8fed345cedb57da61bd27fdd95db22a070cfa7d42ccd4f8e80a443
3
+ size 243
ANE/KokoroProsody_v2.mlmodelc/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db8776a4896f690fd254cdb68a4551c168411c9b3fb98219630215bc3c357986
3
+ size 421
ANE/KokoroProsody_v2.mlmodelc/model.mil ADDED
@@ -0,0 +1,398 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ program(1.0)
2
+ [buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3600.16.1"}, {"coremlc-version", "3600.25.2"}})]
3
+ {
4
+ func main<ios17>(tensor<fp16, [1, 640, ?]> en, tensor<fp16, [1, 128]> style_s) [FlexibleShapeInformation = tuple<tuple<tensor<string, []>, dict<tensor<string, []>, tensor<int32, [?]>>>, tuple<tensor<string, []>, dict<tensor<string, []>, list<tensor<int32, [2]>, ?>>>>((("DefaultShapes", {{"en", [1, 640, 120]}}), ("RangeDims", {{"en", [[1, 1], [640, 640], [1, 2000]]}})))] {
5
+ tensor<string, []> cast_0_dtype_0 = const()[name = tensor<string, []>("cast_0_dtype_0"), val = tensor<string, []>("fp32")];
6
+ tensor<string, []> cast_1_dtype_0 = const()[name = tensor<string, []>("cast_1_dtype_0"), val = tensor<string, []>("fp32")];
7
+ tensor<fp32, [1024]> F0_0_norm1_fc_bias = const()[name = tensor<string, []>("F0_0_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(64)))];
8
+ tensor<fp32, [1024, 128]> F0_0_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4224))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(135360))), name = tensor<string, []>("F0_0_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
9
+ tensor<fp32, [512]> F0_0_norm1_norm_bias = const()[name = tensor<string, []>("F0_0_norm1_norm_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(136448)))];
10
+ tensor<fp32, [512]> F0_0_norm1_norm_weight = const()[name = tensor<string, []>("F0_0_norm1_norm_weight"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(138560)))];
11
+ tensor<fp32, [512]> F0_0_conv1_bias = const()[name = tensor<string, []>("F0_0_conv1_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(140672)))];
12
+ tensor<fp32, [1024]> F0_0_norm2_fc_bias = const()[name = tensor<string, []>("F0_0_norm2_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(142784)))];
13
+ tensor<fp32, [1024, 128]> F0_0_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(146944))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(278080))), name = tensor<string, []>("F0_0_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
14
+ tensor<fp32, [512]> F0_0_conv2_bias = const()[name = tensor<string, []>("F0_0_conv2_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(279168)))];
15
+ tensor<fp32, [1024]> F0_1_norm1_fc_bias = const()[name = tensor<string, []>("F0_1_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(281280)))];
16
+ tensor<fp32, [1024, 128]> F0_1_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(285440))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(416576))), name = tensor<string, []>("F0_1_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
17
+ tensor<fp32, [512]> F0_1_pool_bias = const()[name = tensor<string, []>("F0_1_pool_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(417664)))];
18
+ tensor<fp32, [256]> F0_1_conv1_bias = const()[name = tensor<string, []>("F0_1_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(419776)))];
19
+ tensor<fp32, [512]> F0_1_norm2_fc_bias = const()[name = tensor<string, []>("F0_1_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(420864)))];
20
+ tensor<fp32, [512, 128]> F0_1_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(422976))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(488576))), name = tensor<string, []>("F0_1_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
21
+ tensor<fp32, [256]> F0_1_norm2_norm_bias = const()[name = tensor<string, []>("F0_1_norm2_norm_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(489664)))];
22
+ tensor<fp32, [256]> F0_1_norm2_norm_weight = const()[name = tensor<string, []>("F0_1_norm2_norm_weight"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(490752)))];
23
+ tensor<fp32, [256]> F0_1_conv2_bias = const()[name = tensor<string, []>("F0_1_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(491840)))];
24
+ tensor<fp32, [512]> F0_2_norm1_fc_bias = const()[name = tensor<string, []>("F0_2_norm1_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(492928)))];
25
+ tensor<fp32, [512, 128]> F0_2_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(495040))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(560640))), name = tensor<string, []>("F0_2_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
26
+ tensor<fp32, [256]> F0_2_conv1_bias = const()[name = tensor<string, []>("F0_2_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(561728)))];
27
+ tensor<fp32, [512]> F0_2_norm2_fc_bias = const()[name = tensor<string, []>("F0_2_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(562816)))];
28
+ tensor<fp32, [512, 128]> F0_2_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(564928))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(630528))), name = tensor<string, []>("F0_2_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
29
+ tensor<fp32, [256]> F0_2_conv2_bias = const()[name = tensor<string, []>("F0_2_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(631616)))];
30
+ tensor<fp32, [1]> F0_proj_bias = const()[name = tensor<string, []>("F0_proj_bias"), val = tensor<fp32, [1]>([0x1.edcbf4p-3])];
31
+ tensor<fp32, [1, 256, 1]> F0_proj_weight = const()[name = tensor<string, []>("F0_proj_weight"), val = tensor<fp32, [1, 256, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(632704)))];
32
+ tensor<fp32, [1024]> N_0_norm1_fc_bias = const()[name = tensor<string, []>("N_0_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(633792)))];
33
+ tensor<fp32, [1024, 128]> N_0_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(637952))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(769088))), name = tensor<string, []>("N_0_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
34
+ tensor<fp32, [512]> N_0_conv1_bias = const()[name = tensor<string, []>("N_0_conv1_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(770176)))];
35
+ tensor<fp32, [1024]> N_0_norm2_fc_bias = const()[name = tensor<string, []>("N_0_norm2_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(772288)))];
36
+ tensor<fp32, [1024, 128]> N_0_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(776448))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(907584))), name = tensor<string, []>("N_0_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
37
+ tensor<fp32, [512]> N_0_conv2_bias = const()[name = tensor<string, []>("N_0_conv2_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(908672)))];
38
+ tensor<fp32, [1024]> N_1_norm1_fc_bias = const()[name = tensor<string, []>("N_1_norm1_fc_bias"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(910784)))];
39
+ tensor<fp32, [1024, 128]> N_1_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(914944))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1046080))), name = tensor<string, []>("N_1_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([1024, 128])];
40
+ tensor<fp32, [512]> N_1_pool_bias = const()[name = tensor<string, []>("N_1_pool_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1047168)))];
41
+ tensor<fp32, [256]> N_1_conv1_bias = const()[name = tensor<string, []>("N_1_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1049280)))];
42
+ tensor<fp32, [512]> N_1_norm2_fc_bias = const()[name = tensor<string, []>("N_1_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1050368)))];
43
+ tensor<fp32, [512, 128]> N_1_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1052480))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1118080))), name = tensor<string, []>("N_1_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
44
+ tensor<fp32, [256]> N_1_conv2_bias = const()[name = tensor<string, []>("N_1_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1119168)))];
45
+ tensor<fp32, [512]> N_2_norm1_fc_bias = const()[name = tensor<string, []>("N_2_norm1_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1120256)))];
46
+ tensor<fp32, [512, 128]> N_2_norm1_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1122368))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1187968))), name = tensor<string, []>("N_2_norm1_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
47
+ tensor<fp32, [256]> N_2_conv1_bias = const()[name = tensor<string, []>("N_2_conv1_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1189056)))];
48
+ tensor<fp32, [512]> N_2_norm2_fc_bias = const()[name = tensor<string, []>("N_2_norm2_fc_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1190144)))];
49
+ tensor<fp32, [512, 128]> N_2_norm2_fc_weight_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [65536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1192256))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1257856))), name = tensor<string, []>("N_2_norm2_fc_weight_palettized"), shape = tensor<uint32, [2]>([512, 128])];
50
+ tensor<fp32, [256]> N_2_conv2_bias = const()[name = tensor<string, []>("N_2_conv2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1258944)))];
51
+ tensor<fp32, [1]> N_proj_bias = const()[name = tensor<string, []>("N_proj_bias"), val = tensor<fp32, [1]>([0x1.13144ap-4])];
52
+ tensor<fp32, [1, 256, 1]> N_proj_weight = const()[name = tensor<string, []>("N_proj_weight"), val = tensor<fp32, [1, 256, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1260032)))];
53
+ tensor<int32, [3]> transpose_0_perm_0 = const()[name = tensor<string, []>("transpose_0_perm_0"), val = tensor<int32, [3]>([-1, 0, -2])];
54
+ tensor<fp32, [1024]> add_0 = const()[name = tensor<string, []>("add_0"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1261120)))];
55
+ tensor<fp32, [1024]> add_1 = const()[name = tensor<string, []>("add_1"), val = tensor<fp32, [1024]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1265280)))];
56
+ tensor<fp32, [1024, 640]> concat_4_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [655360]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1269440))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1924864))), name = tensor<string, []>("concat_4_palettized"), shape = tensor<uint32, [2]>([1024, 640])];
57
+ tensor<fp32, [1024, 256]> concat_5_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [262144]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1925952))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2188160))), name = tensor<string, []>("concat_5_palettized"), shape = tensor<uint32, [2]>([1024, 256])];
58
+ tensor<fp32, [1024, 640]> concat_6_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [655360]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2189248))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2844672))), name = tensor<string, []>("concat_6_palettized"), shape = tensor<uint32, [2]>([1024, 640])];
59
+ tensor<fp32, [1024, 256]> concat_7_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [262144]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(2845760))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3107968))), name = tensor<string, []>("concat_7_palettized"), shape = tensor<uint32, [2]>([1024, 256])];
60
+ tensor<fp32, [1, 512]> x_batch_first_lstm_h0_reshaped = const()[name = tensor<string, []>("x_batch_first_lstm_h0_reshaped"), val = tensor<fp32, [1, 512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3109056)))];
61
+ tensor<string, []> x_batch_first_direction_0 = const()[name = tensor<string, []>("x_batch_first_direction_0"), val = tensor<string, []>("bidirectional")];
62
+ tensor<bool, []> x_batch_first_output_sequence_0 = const()[name = tensor<string, []>("x_batch_first_output_sequence_0"), val = tensor<bool, []>(true)];
63
+ tensor<string, []> x_batch_first_recurrent_activation_0 = const()[name = tensor<string, []>("x_batch_first_recurrent_activation_0"), val = tensor<string, []>("sigmoid")];
64
+ tensor<string, []> x_batch_first_cell_activation_0 = const()[name = tensor<string, []>("x_batch_first_cell_activation_0"), val = tensor<string, []>("tanh")];
65
+ tensor<string, []> x_batch_first_activation_0 = const()[name = tensor<string, []>("x_batch_first_activation_0"), val = tensor<string, []>("tanh")];
66
+ tensor<fp32, [1, 640, ?]> cast_0 = cast(dtype = cast_0_dtype_0, x = en)[name = tensor<string, []>("cast_1")];
67
+ tensor<fp32, [?, 1, 640]> transpose_0 = transpose(perm = transpose_0_perm_0, x = cast_0)[name = tensor<string, []>("transpose_2")];
68
+ tensor<fp32, [?, 1, 512]> x_batch_first_0, tensor<fp32, [1, 512]> x_batch_first_1, tensor<fp32, [1, 512]> x_batch_first_2 = lstm(activation = x_batch_first_activation_0, bias = add_0, bias_back = add_1, cell_activation = x_batch_first_cell_activation_0, direction = x_batch_first_direction_0, initial_c = x_batch_first_lstm_h0_reshaped, initial_h = x_batch_first_lstm_h0_reshaped, output_sequence = x_batch_first_output_sequence_0, recurrent_activation = x_batch_first_recurrent_activation_0, weight_hh = concat_5_palettized, weight_hh_back = concat_7_palettized, weight_ih = concat_4_palettized, weight_ih_back = concat_6_palettized, x = transpose_0)[name = tensor<string, []>("x_batch_first")];
69
+ tensor<fp32, []> var_53 = const()[name = tensor<string, []>("op_53"), val = tensor<fp32, []>(0x1.99999ap-3)];
70
+ tensor<fp32, []> var_56 = const()[name = tensor<string, []>("op_56"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
71
+ tensor<fp32, [1, 128]> cast_1 = cast(dtype = cast_1_dtype_0, x = style_s)[name = tensor<string, []>("cast_0")];
72
+ tensor<fp32, [1, 1024]> h_1 = linear(bias = F0_0_norm1_fc_bias, weight = F0_0_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_0")];
73
+ tensor<int32, [3]> var_79 = const()[name = tensor<string, []>("op_79"), val = tensor<int32, [3]>([1, 1024, 1])];
74
+ tensor<fp32, [1, 1024, 1]> h_3 = reshape(shape = var_79, x = h_1)[name = tensor<string, []>("h_3")];
75
+ tensor<int32, [2]> var_81_split_sizes_0 = const()[name = tensor<string, []>("op_81_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
76
+ tensor<int32, []> var_81_axis_0 = const()[name = tensor<string, []>("op_81_axis_0"), val = tensor<int32, []>(1)];
77
+ tensor<fp32, [1, 512, 1]> var_81_0, tensor<fp32, [1, 512, 1]> var_81_1 = split(axis = var_81_axis_0, split_sizes = var_81_split_sizes_0, x = h_3)[name = tensor<string, []>("op_81")];
78
+ tensor<fp32, []> var_83_promoted = const()[name = tensor<string, []>("op_83_promoted"), val = tensor<fp32, []>(0x1p+0)];
79
+ tensor<fp32, [1, 512, 1]> var_84 = add(x = var_81_0, y = var_83_promoted)[name = tensor<string, []>("op_84")];
80
+ tensor<int32, [3]> transpose_0_perm_0_1 = const()[name = tensor<string, []>("transpose_0_perm_0_1"), val = tensor<int32, [3]>([1, 2, 0])];
81
+ tensor<fp32, [1, 512, ?]> transpose_0_1 = transpose(perm = transpose_0_perm_0_1, x = x_batch_first_0)[name = tensor<string, []>("transpose_1")];
82
+ tensor<fp32, [1, 512, ?]> var_87 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_56, gamma = F0_0_norm1_norm_weight, x = transpose_0_1)[name = tensor<string, []>("op_87")];
83
+ tensor<fp32, [1, 512, ?]> var_88 = mul(x = var_84, y = var_87)[name = tensor<string, []>("op_88")];
84
+ tensor<fp32, [1, 512, ?]> input_5 = add(x = var_88, y = var_81_1)[name = tensor<string, []>("input_5")];
85
+ tensor<fp32, [1, 512, ?]> input_7 = leaky_relu(alpha = var_53, x = input_5)[name = tensor<string, []>("input_7")];
86
+ tensor<fp32, [512, 512, 3]> weight_3_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3111168))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3897664))), name = tensor<string, []>("weight_3_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
87
+ tensor<string, []> input_9_pad_type_0 = const()[name = tensor<string, []>("input_9_pad_type_0"), val = tensor<string, []>("custom")];
88
+ tensor<int32, [2]> input_9_pad_0 = const()[name = tensor<string, []>("input_9_pad_0"), val = tensor<int32, [2]>([1, 1])];
89
+ tensor<int32, [1]> input_9_strides_0 = const()[name = tensor<string, []>("input_9_strides_0"), val = tensor<int32, [1]>([1])];
90
+ tensor<int32, [1]> input_9_dilations_0 = const()[name = tensor<string, []>("input_9_dilations_0"), val = tensor<int32, [1]>([1])];
91
+ tensor<int32, []> input_9_groups_0 = const()[name = tensor<string, []>("input_9_groups_0"), val = tensor<int32, []>(1)];
92
+ tensor<fp32, [1, 512, ?]> input_9 = conv(bias = F0_0_conv1_bias, dilations = input_9_dilations_0, groups = input_9_groups_0, pad = input_9_pad_0, pad_type = input_9_pad_type_0, strides = input_9_strides_0, weight = weight_3_palettized, x = input_7)[name = tensor<string, []>("input_9")];
93
+ tensor<fp32, [1, 1024]> h_5 = linear(bias = F0_0_norm2_fc_bias, weight = F0_0_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_1")];
94
+ tensor<int32, [3]> var_105 = const()[name = tensor<string, []>("op_105"), val = tensor<int32, [3]>([1, 1024, 1])];
95
+ tensor<fp32, [1, 1024, 1]> h_7 = reshape(shape = var_105, x = h_5)[name = tensor<string, []>("h_7")];
96
+ tensor<int32, [2]> var_107_split_sizes_0 = const()[name = tensor<string, []>("op_107_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
97
+ tensor<int32, []> var_107_axis_0 = const()[name = tensor<string, []>("op_107_axis_0"), val = tensor<int32, []>(1)];
98
+ tensor<fp32, [1, 512, 1]> var_107_0, tensor<fp32, [1, 512, 1]> var_107_1 = split(axis = var_107_axis_0, split_sizes = var_107_split_sizes_0, x = h_7)[name = tensor<string, []>("op_107")];
99
+ tensor<fp32, []> var_109_promoted = const()[name = tensor<string, []>("op_109_promoted"), val = tensor<fp32, []>(0x1p+0)];
100
+ tensor<fp32, [1, 512, 1]> var_110 = add(x = var_107_0, y = var_109_promoted)[name = tensor<string, []>("op_110")];
101
+ tensor<fp32, [1, 512, ?]> var_113 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_56, gamma = F0_0_norm1_norm_weight, x = input_9)[name = tensor<string, []>("op_113")];
102
+ tensor<fp32, [1, 512, ?]> var_114 = mul(x = var_110, y = var_113)[name = tensor<string, []>("op_114")];
103
+ tensor<fp32, [1, 512, ?]> input_11 = add(x = var_114, y = var_107_1)[name = tensor<string, []>("input_11")];
104
+ tensor<fp32, [1, 512, ?]> input_13 = leaky_relu(alpha = var_53, x = input_11)[name = tensor<string, []>("input_13")];
105
+ tensor<fp32, [512, 512, 3]> weight_7_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3898752))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4685248))), name = tensor<string, []>("weight_7_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
106
+ tensor<string, []> out_1_pad_type_0 = const()[name = tensor<string, []>("out_1_pad_type_0"), val = tensor<string, []>("custom")];
107
+ tensor<int32, [2]> out_1_pad_0 = const()[name = tensor<string, []>("out_1_pad_0"), val = tensor<int32, [2]>([1, 1])];
108
+ tensor<int32, [1]> out_1_strides_0 = const()[name = tensor<string, []>("out_1_strides_0"), val = tensor<int32, [1]>([1])];
109
+ tensor<int32, [1]> out_1_dilations_0 = const()[name = tensor<string, []>("out_1_dilations_0"), val = tensor<int32, [1]>([1])];
110
+ tensor<int32, []> out_1_groups_0 = const()[name = tensor<string, []>("out_1_groups_0"), val = tensor<int32, []>(1)];
111
+ tensor<fp32, [1, 512, ?]> out_1 = conv(bias = F0_0_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_13)[name = tensor<string, []>("out_1")];
112
+ tensor<fp32, [1, 512, ?]> var_124 = add(x = out_1, y = transpose_0_1)[name = tensor<string, []>("op_124")];
113
+ tensor<fp32, []> var_125 = const()[name = tensor<string, []>("op_125"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
114
+ tensor<fp32, [1, 512, ?]> input_15 = mul(x = var_124, y = var_125)[name = tensor<string, []>("input_15")];
115
+ tensor<fp32, []> var_130 = const()[name = tensor<string, []>("op_130"), val = tensor<fp32, []>(0x1.99999ap-3)];
116
+ tensor<fp32, []> var_134 = const()[name = tensor<string, []>("op_134"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
117
+ tensor<fp32, [1, 1024]> h_9 = linear(bias = F0_1_norm1_fc_bias, weight = F0_1_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_2")];
118
+ tensor<int32, [3]> var_166 = const()[name = tensor<string, []>("op_166"), val = tensor<int32, [3]>([1, 1024, 1])];
119
+ tensor<fp32, [1, 1024, 1]> h_11 = reshape(shape = var_166, x = h_9)[name = tensor<string, []>("h_11")];
120
+ tensor<int32, [2]> var_168_split_sizes_0 = const()[name = tensor<string, []>("op_168_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
121
+ tensor<int32, []> var_168_axis_0 = const()[name = tensor<string, []>("op_168_axis_0"), val = tensor<int32, []>(1)];
122
+ tensor<fp32, [1, 512, 1]> var_168_0, tensor<fp32, [1, 512, 1]> var_168_1 = split(axis = var_168_axis_0, split_sizes = var_168_split_sizes_0, x = h_11)[name = tensor<string, []>("op_168")];
123
+ tensor<fp32, []> var_170_promoted = const()[name = tensor<string, []>("op_170_promoted"), val = tensor<fp32, []>(0x1p+0)];
124
+ tensor<fp32, [1, 512, 1]> var_171 = add(x = var_168_0, y = var_170_promoted)[name = tensor<string, []>("op_171")];
125
+ tensor<fp32, [1, 512, ?]> var_174 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_134, gamma = F0_0_norm1_norm_weight, x = input_15)[name = tensor<string, []>("op_174")];
126
+ tensor<fp32, [1, 512, ?]> var_175 = mul(x = var_171, y = var_174)[name = tensor<string, []>("op_175")];
127
+ tensor<fp32, [1, 512, ?]> input_17 = add(x = var_175, y = var_168_1)[name = tensor<string, []>("input_17")];
128
+ tensor<fp32, [1, 512, ?]> input_19 = leaky_relu(alpha = var_130, x = input_17)[name = tensor<string, []>("input_19")];
129
+ tensor<fp32, [512, 1, 3]> var_178 = const()[name = tensor<string, []>("op_178"), val = tensor<fp32, [512, 1, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4686336)))];
130
+ tensor<string, []> conv_transpose_0_pad_type_0 = const()[name = tensor<string, []>("conv_transpose_0_pad_type_0"), val = tensor<string, []>("custom")];
131
+ tensor<int32, [2]> conv_transpose_0_pad_0 = const()[name = tensor<string, []>("conv_transpose_0_pad_0"), val = tensor<int32, [2]>([0, 0])];
132
+ tensor<int32, [1]> conv_transpose_0_strides_0 = const()[name = tensor<string, []>("conv_transpose_0_strides_0"), val = tensor<int32, [1]>([2])];
133
+ tensor<int32, []> conv_transpose_0_groups_0 = const()[name = tensor<string, []>("conv_transpose_0_groups_0"), val = tensor<int32, []>(512)];
134
+ tensor<int32, [1]> conv_transpose_0_dilations_0 = const()[name = tensor<string, []>("conv_transpose_0_dilations_0"), val = tensor<int32, [1]>([1])];
135
+ tensor<fp32, [1, 512, ?]> conv_transpose_0 = conv_transpose(bias = F0_1_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 = var_178, x = input_19)[name = tensor<string, []>("conv_transpose_0")];
136
+ tensor<int32, [3]> input_21_begin_0 = const()[name = tensor<string, []>("input_21_begin_0"), val = tensor<int32, [3]>([0, 0, 1])];
137
+ tensor<int32, [3]> input_21_end_0 = const()[name = tensor<string, []>("input_21_end_0"), val = tensor<int32, [3]>([0, 0, 0])];
138
+ tensor<bool, [3]> input_21_begin_mask_0 = const()[name = tensor<string, []>("input_21_begin_mask_0"), val = tensor<bool, [3]>([true, true, false])];
139
+ tensor<bool, [3]> input_21_end_mask_0 = const()[name = tensor<string, []>("input_21_end_mask_0"), val = tensor<bool, [3]>([true, true, true])];
140
+ tensor<fp32, [1, 512, ?]> input_21 = slice_by_index(begin = input_21_begin_0, begin_mask = input_21_begin_mask_0, end = input_21_end_0, end_mask = input_21_end_mask_0, x = conv_transpose_0)[name = tensor<string, []>("input_21")];
141
+ tensor<fp32, [256, 512, 3]> weight_11_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [393216]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4692544))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5085824))), name = tensor<string, []>("weight_11_palettized"), shape = tensor<uint32, [3]>([256, 512, 3])];
142
+ tensor<string, []> input_23_pad_type_0 = const()[name = tensor<string, []>("input_23_pad_type_0"), val = tensor<string, []>("custom")];
143
+ tensor<int32, [2]> input_23_pad_0 = const()[name = tensor<string, []>("input_23_pad_0"), val = tensor<int32, [2]>([1, 1])];
144
+ tensor<int32, [1]> input_23_strides_0 = const()[name = tensor<string, []>("input_23_strides_0"), val = tensor<int32, [1]>([1])];
145
+ tensor<int32, [1]> input_23_dilations_0 = const()[name = tensor<string, []>("input_23_dilations_0"), val = tensor<int32, [1]>([1])];
146
+ tensor<int32, []> input_23_groups_0 = const()[name = tensor<string, []>("input_23_groups_0"), val = tensor<int32, []>(1)];
147
+ tensor<fp32, [1, 256, ?]> input_23 = conv(bias = F0_1_conv1_bias, dilations = input_23_dilations_0, groups = input_23_groups_0, pad = input_23_pad_0, pad_type = input_23_pad_type_0, strides = input_23_strides_0, weight = weight_11_palettized, x = input_21)[name = tensor<string, []>("input_23")];
148
+ tensor<fp32, [1, 512]> h_13 = linear(bias = F0_1_norm2_fc_bias, weight = F0_1_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_3")];
149
+ tensor<int32, [3]> var_199 = const()[name = tensor<string, []>("op_199"), val = tensor<int32, [3]>([1, 512, 1])];
150
+ tensor<fp32, [1, 512, 1]> h_15 = reshape(shape = var_199, x = h_13)[name = tensor<string, []>("h_15")];
151
+ tensor<int32, [2]> var_201_split_sizes_0 = const()[name = tensor<string, []>("op_201_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
152
+ tensor<int32, []> var_201_axis_0 = const()[name = tensor<string, []>("op_201_axis_0"), val = tensor<int32, []>(1)];
153
+ tensor<fp32, [1, 256, 1]> var_201_0, tensor<fp32, [1, 256, 1]> var_201_1 = split(axis = var_201_axis_0, split_sizes = var_201_split_sizes_0, x = h_15)[name = tensor<string, []>("op_201")];
154
+ tensor<fp32, []> var_203_promoted = const()[name = tensor<string, []>("op_203_promoted"), val = tensor<fp32, []>(0x1p+0)];
155
+ tensor<fp32, [1, 256, 1]> var_204 = add(x = var_201_0, y = var_203_promoted)[name = tensor<string, []>("op_204")];
156
+ tensor<fp32, [1, 256, ?]> var_207 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_134, gamma = F0_1_norm2_norm_weight, x = input_23)[name = tensor<string, []>("op_207")];
157
+ tensor<fp32, [1, 256, ?]> var_208 = mul(x = var_204, y = var_207)[name = tensor<string, []>("op_208")];
158
+ tensor<fp32, [1, 256, ?]> input_25 = add(x = var_208, y = var_201_1)[name = tensor<string, []>("input_25")];
159
+ tensor<fp32, [1, 256, ?]> input_27 = leaky_relu(alpha = var_130, x = input_25)[name = tensor<string, []>("input_27")];
160
+ tensor<fp32, [256, 256, 3]> weight_15_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5086912))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5283584))), name = tensor<string, []>("weight_15_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
161
+ tensor<string, []> out_3_pad_type_0 = const()[name = tensor<string, []>("out_3_pad_type_0"), val = tensor<string, []>("custom")];
162
+ tensor<int32, [2]> out_3_pad_0 = const()[name = tensor<string, []>("out_3_pad_0"), val = tensor<int32, [2]>([1, 1])];
163
+ tensor<int32, [1]> out_3_strides_0 = const()[name = tensor<string, []>("out_3_strides_0"), val = tensor<int32, [1]>([1])];
164
+ tensor<int32, [1]> out_3_dilations_0 = const()[name = tensor<string, []>("out_3_dilations_0"), val = tensor<int32, [1]>([1])];
165
+ tensor<int32, []> out_3_groups_0 = const()[name = tensor<string, []>("out_3_groups_0"), val = tensor<int32, []>(1)];
166
+ tensor<fp32, [1, 256, ?]> out_3 = conv(bias = F0_1_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_27)[name = tensor<string, []>("out_3")];
167
+ tensor<int32, [1]> expand_dims_0_axes_0 = const()[name = tensor<string, []>("expand_dims_0_axes_0"), val = tensor<int32, [1]>([3])];
168
+ tensor<fp32, [1, 512, ?, 1]> expand_dims_0 = expand_dims(axes = expand_dims_0_axes_0, x = input_15)[name = tensor<string, []>("expand_dims_0")];
169
+ tensor<int32, []> upsample_nearest_neighbor_0_scale_factor_height_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_0_scale_factor_height_0"), val = tensor<int32, []>(2)];
170
+ tensor<int32, []> upsample_nearest_neighbor_0_scale_factor_width_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_0_scale_factor_width_0"), val = tensor<int32, []>(1)];
171
+ tensor<fp32, [1, 512, ?, 1]> 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<string, []>("upsample_nearest_neighbor_0")];
172
+ tensor<int32, [1]> input_29_axes_0 = const()[name = tensor<string, []>("input_29_axes_0"), val = tensor<int32, [1]>([3])];
173
+ tensor<fp32, [1, 512, ?]> input_29 = squeeze(axes = input_29_axes_0, x = upsample_nearest_neighbor_0)[name = tensor<string, []>("input_29")];
174
+ tensor<fp32, [256, 512, 1]> weight_17_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5284672))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5415808))), name = tensor<string, []>("weight_17_palettized"), shape = tensor<uint32, [3]>([256, 512, 1])];
175
+ tensor<string, []> var_225_pad_type_0 = const()[name = tensor<string, []>("op_225_pad_type_0"), val = tensor<string, []>("valid")];
176
+ tensor<int32, [1]> var_225_strides_0 = const()[name = tensor<string, []>("op_225_strides_0"), val = tensor<int32, [1]>([1])];
177
+ tensor<int32, [2]> var_225_pad_0 = const()[name = tensor<string, []>("op_225_pad_0"), val = tensor<int32, [2]>([0, 0])];
178
+ tensor<int32, [1]> var_225_dilations_0 = const()[name = tensor<string, []>("op_225_dilations_0"), val = tensor<int32, [1]>([1])];
179
+ tensor<int32, []> var_225_groups_0 = const()[name = tensor<string, []>("op_225_groups_0"), val = tensor<int32, []>(1)];
180
+ tensor<fp32, [1, 256, ?]> var_225 = conv(dilations = var_225_dilations_0, groups = var_225_groups_0, pad = var_225_pad_0, pad_type = var_225_pad_type_0, strides = var_225_strides_0, weight = weight_17_palettized, x = input_29)[name = tensor<string, []>("op_225")];
181
+ tensor<fp32, [1, 256, ?]> var_226 = add(x = out_3, y = var_225)[name = tensor<string, []>("op_226")];
182
+ tensor<fp32, []> var_227 = const()[name = tensor<string, []>("op_227"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
183
+ tensor<fp32, [1, 256, ?]> input_31 = mul(x = var_226, y = var_227)[name = tensor<string, []>("input_31")];
184
+ tensor<fp32, []> var_230 = const()[name = tensor<string, []>("op_230"), val = tensor<fp32, []>(0x1.99999ap-3)];
185
+ tensor<fp32, []> var_233 = const()[name = tensor<string, []>("op_233"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
186
+ tensor<fp32, [1, 512]> h_17 = linear(bias = F0_2_norm1_fc_bias, weight = F0_2_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_4")];
187
+ tensor<int32, [3]> var_256 = const()[name = tensor<string, []>("op_256"), val = tensor<int32, [3]>([1, 512, 1])];
188
+ tensor<fp32, [1, 512, 1]> h_19 = reshape(shape = var_256, x = h_17)[name = tensor<string, []>("h_19")];
189
+ tensor<int32, [2]> var_258_split_sizes_0 = const()[name = tensor<string, []>("op_258_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
190
+ tensor<int32, []> var_258_axis_0 = const()[name = tensor<string, []>("op_258_axis_0"), val = tensor<int32, []>(1)];
191
+ tensor<fp32, [1, 256, 1]> var_258_0, tensor<fp32, [1, 256, 1]> var_258_1 = split(axis = var_258_axis_0, split_sizes = var_258_split_sizes_0, x = h_19)[name = tensor<string, []>("op_258")];
192
+ tensor<fp32, []> var_260_promoted = const()[name = tensor<string, []>("op_260_promoted"), val = tensor<fp32, []>(0x1p+0)];
193
+ tensor<fp32, [1, 256, 1]> var_261 = add(x = var_258_0, y = var_260_promoted)[name = tensor<string, []>("op_261")];
194
+ tensor<fp32, [1, 256, ?]> var_264 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_233, gamma = F0_1_norm2_norm_weight, x = input_31)[name = tensor<string, []>("op_264")];
195
+ tensor<fp32, [1, 256, ?]> var_265 = mul(x = var_261, y = var_264)[name = tensor<string, []>("op_265")];
196
+ tensor<fp32, [1, 256, ?]> input_33 = add(x = var_265, y = var_258_1)[name = tensor<string, []>("input_33")];
197
+ tensor<fp32, [1, 256, ?]> input_35 = leaky_relu(alpha = var_230, x = input_33)[name = tensor<string, []>("input_35")];
198
+ tensor<fp32, [256, 256, 3]> weight_21_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5416896))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5613568))), name = tensor<string, []>("weight_21_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
199
+ tensor<string, []> input_37_pad_type_0 = const()[name = tensor<string, []>("input_37_pad_type_0"), val = tensor<string, []>("custom")];
200
+ tensor<int32, [2]> input_37_pad_0 = const()[name = tensor<string, []>("input_37_pad_0"), val = tensor<int32, [2]>([1, 1])];
201
+ tensor<int32, [1]> input_37_strides_0 = const()[name = tensor<string, []>("input_37_strides_0"), val = tensor<int32, [1]>([1])];
202
+ tensor<int32, [1]> input_37_dilations_0 = const()[name = tensor<string, []>("input_37_dilations_0"), val = tensor<int32, [1]>([1])];
203
+ tensor<int32, []> input_37_groups_0 = const()[name = tensor<string, []>("input_37_groups_0"), val = tensor<int32, []>(1)];
204
+ tensor<fp32, [1, 256, ?]> input_37 = conv(bias = F0_2_conv1_bias, dilations = input_37_dilations_0, groups = input_37_groups_0, pad = input_37_pad_0, pad_type = input_37_pad_type_0, strides = input_37_strides_0, weight = weight_21_palettized, x = input_35)[name = tensor<string, []>("input_37")];
205
+ tensor<fp32, [1, 512]> h_21 = linear(bias = F0_2_norm2_fc_bias, weight = F0_2_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_5")];
206
+ tensor<int32, [3]> var_282 = const()[name = tensor<string, []>("op_282"), val = tensor<int32, [3]>([1, 512, 1])];
207
+ tensor<fp32, [1, 512, 1]> h_23 = reshape(shape = var_282, x = h_21)[name = tensor<string, []>("h_23")];
208
+ tensor<int32, [2]> var_284_split_sizes_0 = const()[name = tensor<string, []>("op_284_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
209
+ tensor<int32, []> var_284_axis_0 = const()[name = tensor<string, []>("op_284_axis_0"), val = tensor<int32, []>(1)];
210
+ tensor<fp32, [1, 256, 1]> var_284_0, tensor<fp32, [1, 256, 1]> var_284_1 = split(axis = var_284_axis_0, split_sizes = var_284_split_sizes_0, x = h_23)[name = tensor<string, []>("op_284")];
211
+ tensor<fp32, []> var_286_promoted = const()[name = tensor<string, []>("op_286_promoted"), val = tensor<fp32, []>(0x1p+0)];
212
+ tensor<fp32, [1, 256, 1]> var_287 = add(x = var_284_0, y = var_286_promoted)[name = tensor<string, []>("op_287")];
213
+ tensor<fp32, [1, 256, ?]> var_290 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_233, gamma = F0_1_norm2_norm_weight, x = input_37)[name = tensor<string, []>("op_290")];
214
+ tensor<fp32, [1, 256, ?]> var_291 = mul(x = var_287, y = var_290)[name = tensor<string, []>("op_291")];
215
+ tensor<fp32, [1, 256, ?]> input_39 = add(x = var_291, y = var_284_1)[name = tensor<string, []>("input_39")];
216
+ tensor<fp32, [1, 256, ?]> input_41 = leaky_relu(alpha = var_230, x = input_39)[name = tensor<string, []>("input_41")];
217
+ tensor<fp32, [256, 256, 3]> weight_25_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5614656))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5811328))), name = tensor<string, []>("weight_25_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
218
+ tensor<string, []> out_5_pad_type_0 = const()[name = tensor<string, []>("out_5_pad_type_0"), val = tensor<string, []>("custom")];
219
+ tensor<int32, [2]> out_5_pad_0 = const()[name = tensor<string, []>("out_5_pad_0"), val = tensor<int32, [2]>([1, 1])];
220
+ tensor<int32, [1]> out_5_strides_0 = const()[name = tensor<string, []>("out_5_strides_0"), val = tensor<int32, [1]>([1])];
221
+ tensor<int32, [1]> out_5_dilations_0 = const()[name = tensor<string, []>("out_5_dilations_0"), val = tensor<int32, [1]>([1])];
222
+ tensor<int32, []> out_5_groups_0 = const()[name = tensor<string, []>("out_5_groups_0"), val = tensor<int32, []>(1)];
223
+ tensor<fp32, [1, 256, ?]> out_5 = conv(bias = F0_2_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_25_palettized, x = input_41)[name = tensor<string, []>("out_5")];
224
+ tensor<fp32, [1, 256, ?]> var_301 = add(x = out_5, y = input_31)[name = tensor<string, []>("op_301")];
225
+ tensor<fp32, []> var_302 = const()[name = tensor<string, []>("op_302"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
226
+ tensor<fp32, [1, 256, ?]> input_43 = mul(x = var_301, y = var_302)[name = tensor<string, []>("input_43")];
227
+ tensor<string, []> var_314_pad_type_0 = const()[name = tensor<string, []>("op_314_pad_type_0"), val = tensor<string, []>("valid")];
228
+ tensor<int32, [1]> var_314_strides_0 = const()[name = tensor<string, []>("op_314_strides_0"), val = tensor<int32, [1]>([1])];
229
+ tensor<int32, [2]> var_314_pad_0 = const()[name = tensor<string, []>("op_314_pad_0"), val = tensor<int32, [2]>([0, 0])];
230
+ tensor<int32, [1]> var_314_dilations_0 = const()[name = tensor<string, []>("op_314_dilations_0"), val = tensor<int32, [1]>([1])];
231
+ tensor<int32, []> var_314_groups_0 = const()[name = tensor<string, []>("op_314_groups_0"), val = tensor<int32, []>(1)];
232
+ tensor<fp32, [1, 1, ?]> var_314 = conv(bias = F0_proj_bias, dilations = var_314_dilations_0, groups = var_314_groups_0, pad = var_314_pad_0, pad_type = var_314_pad_type_0, strides = var_314_strides_0, weight = F0_proj_weight, x = input_43)[name = tensor<string, []>("op_314")];
233
+ tensor<int32, [1]> var_316_axes_0 = const()[name = tensor<string, []>("op_316_axes_0"), val = tensor<int32, [1]>([1])];
234
+ tensor<fp32, [1, ?]> F0 = squeeze(axes = var_316_axes_0, x = var_314)[name = tensor<string, []>("op_316")];
235
+ tensor<fp32, []> var_321 = const()[name = tensor<string, []>("op_321"), val = tensor<fp32, []>(0x1.99999ap-3)];
236
+ tensor<fp32, []> var_324 = const()[name = tensor<string, []>("op_324"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
237
+ tensor<fp32, [1, 1024]> h_25 = linear(bias = N_0_norm1_fc_bias, weight = N_0_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_6")];
238
+ tensor<int32, [3]> var_347 = const()[name = tensor<string, []>("op_347"), val = tensor<int32, [3]>([1, 1024, 1])];
239
+ tensor<fp32, [1, 1024, 1]> h_27 = reshape(shape = var_347, x = h_25)[name = tensor<string, []>("h_27")];
240
+ tensor<int32, [2]> var_349_split_sizes_0 = const()[name = tensor<string, []>("op_349_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
241
+ tensor<int32, []> var_349_axis_0 = const()[name = tensor<string, []>("op_349_axis_0"), val = tensor<int32, []>(1)];
242
+ tensor<fp32, [1, 512, 1]> var_349_0, tensor<fp32, [1, 512, 1]> var_349_1 = split(axis = var_349_axis_0, split_sizes = var_349_split_sizes_0, x = h_27)[name = tensor<string, []>("op_349")];
243
+ tensor<fp32, []> var_351_promoted = const()[name = tensor<string, []>("op_351_promoted"), val = tensor<fp32, []>(0x1p+0)];
244
+ tensor<fp32, [1, 512, 1]> var_352 = add(x = var_349_0, y = var_351_promoted)[name = tensor<string, []>("op_352")];
245
+ tensor<fp32, [1, 512, ?]> var_356 = mul(x = var_352, y = var_87)[name = tensor<string, []>("op_356")];
246
+ tensor<fp32, [1, 512, ?]> input_47 = add(x = var_356, y = var_349_1)[name = tensor<string, []>("input_47")];
247
+ tensor<fp32, [1, 512, ?]> input_49 = leaky_relu(alpha = var_321, x = input_47)[name = tensor<string, []>("input_49")];
248
+ tensor<fp32, [512, 512, 3]> weight_31_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(5812416))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(6598912))), name = tensor<string, []>("weight_31_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
249
+ tensor<string, []> input_51_pad_type_0 = const()[name = tensor<string, []>("input_51_pad_type_0"), val = tensor<string, []>("custom")];
250
+ tensor<int32, [2]> input_51_pad_0 = const()[name = tensor<string, []>("input_51_pad_0"), val = tensor<int32, [2]>([1, 1])];
251
+ tensor<int32, [1]> input_51_strides_0 = const()[name = tensor<string, []>("input_51_strides_0"), val = tensor<int32, [1]>([1])];
252
+ tensor<int32, [1]> input_51_dilations_0 = const()[name = tensor<string, []>("input_51_dilations_0"), val = tensor<int32, [1]>([1])];
253
+ tensor<int32, []> input_51_groups_0 = const()[name = tensor<string, []>("input_51_groups_0"), val = tensor<int32, []>(1)];
254
+ tensor<fp32, [1, 512, ?]> input_51 = conv(bias = N_0_conv1_bias, dilations = input_51_dilations_0, groups = input_51_groups_0, pad = input_51_pad_0, pad_type = input_51_pad_type_0, strides = input_51_strides_0, weight = weight_31_palettized, x = input_49)[name = tensor<string, []>("input_51")];
255
+ tensor<fp32, [1, 1024]> h_29 = linear(bias = N_0_norm2_fc_bias, weight = N_0_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_7")];
256
+ tensor<int32, [3]> var_373 = const()[name = tensor<string, []>("op_373"), val = tensor<int32, [3]>([1, 1024, 1])];
257
+ tensor<fp32, [1, 1024, 1]> h_31 = reshape(shape = var_373, x = h_29)[name = tensor<string, []>("h_31")];
258
+ tensor<int32, [2]> var_375_split_sizes_0 = const()[name = tensor<string, []>("op_375_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
259
+ tensor<int32, []> var_375_axis_0 = const()[name = tensor<string, []>("op_375_axis_0"), val = tensor<int32, []>(1)];
260
+ tensor<fp32, [1, 512, 1]> var_375_0, tensor<fp32, [1, 512, 1]> var_375_1 = split(axis = var_375_axis_0, split_sizes = var_375_split_sizes_0, x = h_31)[name = tensor<string, []>("op_375")];
261
+ tensor<fp32, []> var_377_promoted = const()[name = tensor<string, []>("op_377_promoted"), val = tensor<fp32, []>(0x1p+0)];
262
+ tensor<fp32, [1, 512, 1]> var_378 = add(x = var_375_0, y = var_377_promoted)[name = tensor<string, []>("op_378")];
263
+ tensor<fp32, [1, 512, ?]> var_381 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_324, gamma = F0_0_norm1_norm_weight, x = input_51)[name = tensor<string, []>("op_381")];
264
+ tensor<fp32, [1, 512, ?]> var_382 = mul(x = var_378, y = var_381)[name = tensor<string, []>("op_382")];
265
+ tensor<fp32, [1, 512, ?]> input_53 = add(x = var_382, y = var_375_1)[name = tensor<string, []>("input_53")];
266
+ tensor<fp32, [1, 512, ?]> input_55 = leaky_relu(alpha = var_321, x = input_53)[name = tensor<string, []>("input_55")];
267
+ tensor<fp32, [512, 512, 3]> weight_35_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [786432]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(6600000))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7386496))), name = tensor<string, []>("weight_35_palettized"), shape = tensor<uint32, [3]>([512, 512, 3])];
268
+ tensor<string, []> out_7_pad_type_0 = const()[name = tensor<string, []>("out_7_pad_type_0"), val = tensor<string, []>("custom")];
269
+ tensor<int32, [2]> out_7_pad_0 = const()[name = tensor<string, []>("out_7_pad_0"), val = tensor<int32, [2]>([1, 1])];
270
+ tensor<int32, [1]> out_7_strides_0 = const()[name = tensor<string, []>("out_7_strides_0"), val = tensor<int32, [1]>([1])];
271
+ tensor<int32, [1]> out_7_dilations_0 = const()[name = tensor<string, []>("out_7_dilations_0"), val = tensor<int32, [1]>([1])];
272
+ tensor<int32, []> out_7_groups_0 = const()[name = tensor<string, []>("out_7_groups_0"), val = tensor<int32, []>(1)];
273
+ tensor<fp32, [1, 512, ?]> out_7 = conv(bias = N_0_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_35_palettized, x = input_55)[name = tensor<string, []>("out_7")];
274
+ tensor<fp32, [1, 512, ?]> var_392 = add(x = out_7, y = transpose_0_1)[name = tensor<string, []>("op_392")];
275
+ tensor<fp32, []> var_393 = const()[name = tensor<string, []>("op_393"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
276
+ tensor<fp32, [1, 512, ?]> input_57 = mul(x = var_392, y = var_393)[name = tensor<string, []>("input_57")];
277
+ tensor<fp32, []> var_398 = const()[name = tensor<string, []>("op_398"), val = tensor<fp32, []>(0x1.99999ap-3)];
278
+ tensor<fp32, []> var_402 = const()[name = tensor<string, []>("op_402"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
279
+ tensor<fp32, [1, 1024]> h_33 = linear(bias = N_1_norm1_fc_bias, weight = N_1_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_8")];
280
+ tensor<int32, [3]> var_434 = const()[name = tensor<string, []>("op_434"), val = tensor<int32, [3]>([1, 1024, 1])];
281
+ tensor<fp32, [1, 1024, 1]> h_35 = reshape(shape = var_434, x = h_33)[name = tensor<string, []>("h_35")];
282
+ tensor<int32, [2]> var_436_split_sizes_0 = const()[name = tensor<string, []>("op_436_split_sizes_0"), val = tensor<int32, [2]>([512, 512])];
283
+ tensor<int32, []> var_436_axis_0 = const()[name = tensor<string, []>("op_436_axis_0"), val = tensor<int32, []>(1)];
284
+ tensor<fp32, [1, 512, 1]> var_436_0, tensor<fp32, [1, 512, 1]> var_436_1 = split(axis = var_436_axis_0, split_sizes = var_436_split_sizes_0, x = h_35)[name = tensor<string, []>("op_436")];
285
+ tensor<fp32, []> var_438_promoted = const()[name = tensor<string, []>("op_438_promoted"), val = tensor<fp32, []>(0x1p+0)];
286
+ tensor<fp32, [1, 512, 1]> var_439 = add(x = var_436_0, y = var_438_promoted)[name = tensor<string, []>("op_439")];
287
+ tensor<fp32, [1, 512, ?]> var_442 = instance_norm(beta = F0_0_norm1_norm_bias, epsilon = var_402, gamma = F0_0_norm1_norm_weight, x = input_57)[name = tensor<string, []>("op_442")];
288
+ tensor<fp32, [1, 512, ?]> var_443 = mul(x = var_439, y = var_442)[name = tensor<string, []>("op_443")];
289
+ tensor<fp32, [1, 512, ?]> input_59 = add(x = var_443, y = var_436_1)[name = tensor<string, []>("input_59")];
290
+ tensor<fp32, [1, 512, ?]> input_61 = leaky_relu(alpha = var_398, x = input_59)[name = tensor<string, []>("input_61")];
291
+ tensor<fp32, [512, 1, 3]> var_446 = const()[name = tensor<string, []>("op_446"), val = tensor<fp32, [512, 1, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7387584)))];
292
+ tensor<string, []> conv_transpose_1_pad_type_0 = const()[name = tensor<string, []>("conv_transpose_1_pad_type_0"), val = tensor<string, []>("custom")];
293
+ tensor<int32, [2]> conv_transpose_1_pad_0 = const()[name = tensor<string, []>("conv_transpose_1_pad_0"), val = tensor<int32, [2]>([0, 0])];
294
+ tensor<int32, [1]> conv_transpose_1_strides_0 = const()[name = tensor<string, []>("conv_transpose_1_strides_0"), val = tensor<int32, [1]>([2])];
295
+ tensor<int32, []> conv_transpose_1_groups_0 = const()[name = tensor<string, []>("conv_transpose_1_groups_0"), val = tensor<int32, []>(512)];
296
+ tensor<int32, [1]> conv_transpose_1_dilations_0 = const()[name = tensor<string, []>("conv_transpose_1_dilations_0"), val = tensor<int32, [1]>([1])];
297
+ tensor<fp32, [1, 512, ?]> conv_transpose_1 = conv_transpose(bias = N_1_pool_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 = var_446, x = input_61)[name = tensor<string, []>("conv_transpose_1")];
298
+ tensor<int32, [3]> input_63_begin_0 = const()[name = tensor<string, []>("input_63_begin_0"), val = tensor<int32, [3]>([0, 0, 1])];
299
+ tensor<int32, [3]> input_63_end_0 = const()[name = tensor<string, []>("input_63_end_0"), val = tensor<int32, [3]>([0, 0, 0])];
300
+ tensor<bool, [3]> input_63_begin_mask_0 = const()[name = tensor<string, []>("input_63_begin_mask_0"), val = tensor<bool, [3]>([true, true, false])];
301
+ tensor<bool, [3]> input_63_end_mask_0 = const()[name = tensor<string, []>("input_63_end_mask_0"), val = tensor<bool, [3]>([true, true, true])];
302
+ tensor<fp32, [1, 512, ?]> input_63 = slice_by_index(begin = input_63_begin_0, begin_mask = input_63_begin_mask_0, end = input_63_end_0, end_mask = input_63_end_mask_0, x = conv_transpose_1)[name = tensor<string, []>("input_63")];
303
+ tensor<fp32, [256, 512, 3]> weight_39_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [393216]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7393792))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7787072))), name = tensor<string, []>("weight_39_palettized"), shape = tensor<uint32, [3]>([256, 512, 3])];
304
+ tensor<string, []> input_65_pad_type_0 = const()[name = tensor<string, []>("input_65_pad_type_0"), val = tensor<string, []>("custom")];
305
+ tensor<int32, [2]> input_65_pad_0 = const()[name = tensor<string, []>("input_65_pad_0"), val = tensor<int32, [2]>([1, 1])];
306
+ tensor<int32, [1]> input_65_strides_0 = const()[name = tensor<string, []>("input_65_strides_0"), val = tensor<int32, [1]>([1])];
307
+ tensor<int32, [1]> input_65_dilations_0 = const()[name = tensor<string, []>("input_65_dilations_0"), val = tensor<int32, [1]>([1])];
308
+ tensor<int32, []> input_65_groups_0 = const()[name = tensor<string, []>("input_65_groups_0"), val = tensor<int32, []>(1)];
309
+ tensor<fp32, [1, 256, ?]> input_65 = conv(bias = N_1_conv1_bias, 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_39_palettized, x = input_63)[name = tensor<string, []>("input_65")];
310
+ tensor<fp32, [1, 512]> h_37 = linear(bias = N_1_norm2_fc_bias, weight = N_1_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_9")];
311
+ tensor<int32, [3]> var_467 = const()[name = tensor<string, []>("op_467"), val = tensor<int32, [3]>([1, 512, 1])];
312
+ tensor<fp32, [1, 512, 1]> h_39 = reshape(shape = var_467, x = h_37)[name = tensor<string, []>("h_39")];
313
+ tensor<int32, [2]> var_469_split_sizes_0 = const()[name = tensor<string, []>("op_469_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
314
+ tensor<int32, []> var_469_axis_0 = const()[name = tensor<string, []>("op_469_axis_0"), val = tensor<int32, []>(1)];
315
+ tensor<fp32, [1, 256, 1]> var_469_0, tensor<fp32, [1, 256, 1]> var_469_1 = split(axis = var_469_axis_0, split_sizes = var_469_split_sizes_0, x = h_39)[name = tensor<string, []>("op_469")];
316
+ tensor<fp32, []> var_471_promoted = const()[name = tensor<string, []>("op_471_promoted"), val = tensor<fp32, []>(0x1p+0)];
317
+ tensor<fp32, [1, 256, 1]> var_472 = add(x = var_469_0, y = var_471_promoted)[name = tensor<string, []>("op_472")];
318
+ tensor<fp32, [1, 256, ?]> var_475 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_402, gamma = F0_1_norm2_norm_weight, x = input_65)[name = tensor<string, []>("op_475")];
319
+ tensor<fp32, [1, 256, ?]> var_476 = mul(x = var_472, y = var_475)[name = tensor<string, []>("op_476")];
320
+ tensor<fp32, [1, 256, ?]> input_67 = add(x = var_476, y = var_469_1)[name = tensor<string, []>("input_67")];
321
+ tensor<fp32, [1, 256, ?]> input_69 = leaky_relu(alpha = var_398, x = input_67)[name = tensor<string, []>("input_69")];
322
+ tensor<fp32, [256, 256, 3]> weight_43_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7788160))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7984832))), name = tensor<string, []>("weight_43_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
323
+ tensor<string, []> out_9_pad_type_0 = const()[name = tensor<string, []>("out_9_pad_type_0"), val = tensor<string, []>("custom")];
324
+ tensor<int32, [2]> out_9_pad_0 = const()[name = tensor<string, []>("out_9_pad_0"), val = tensor<int32, [2]>([1, 1])];
325
+ tensor<int32, [1]> out_9_strides_0 = const()[name = tensor<string, []>("out_9_strides_0"), val = tensor<int32, [1]>([1])];
326
+ tensor<int32, [1]> out_9_dilations_0 = const()[name = tensor<string, []>("out_9_dilations_0"), val = tensor<int32, [1]>([1])];
327
+ tensor<int32, []> out_9_groups_0 = const()[name = tensor<string, []>("out_9_groups_0"), val = tensor<int32, []>(1)];
328
+ tensor<fp32, [1, 256, ?]> out_9 = conv(bias = N_1_conv2_bias, 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_43_palettized, x = input_69)[name = tensor<string, []>("out_9")];
329
+ tensor<int32, [1]> expand_dims_1_axes_0 = const()[name = tensor<string, []>("expand_dims_1_axes_0"), val = tensor<int32, [1]>([3])];
330
+ tensor<fp32, [1, 512, ?, 1]> expand_dims_1 = expand_dims(axes = expand_dims_1_axes_0, x = input_57)[name = tensor<string, []>("expand_dims_1")];
331
+ tensor<int32, []> upsample_nearest_neighbor_1_scale_factor_height_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_1_scale_factor_height_0"), val = tensor<int32, []>(2)];
332
+ tensor<int32, []> upsample_nearest_neighbor_1_scale_factor_width_0 = const()[name = tensor<string, []>("upsample_nearest_neighbor_1_scale_factor_width_0"), val = tensor<int32, []>(1)];
333
+ tensor<fp32, [1, 512, ?, 1]> upsample_nearest_neighbor_1 = 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)[name = tensor<string, []>("upsample_nearest_neighbor_1")];
334
+ tensor<int32, [1]> input_71_axes_0 = const()[name = tensor<string, []>("input_71_axes_0"), val = tensor<int32, [1]>([3])];
335
+ tensor<fp32, [1, 512, ?]> input_71 = squeeze(axes = input_71_axes_0, x = upsample_nearest_neighbor_1)[name = tensor<string, []>("input_71")];
336
+ tensor<fp32, [256, 512, 1]> weight_45_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [131072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7985920))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8117056))), name = tensor<string, []>("weight_45_palettized"), shape = tensor<uint32, [3]>([256, 512, 1])];
337
+ tensor<string, []> var_493_pad_type_0 = const()[name = tensor<string, []>("op_493_pad_type_0"), val = tensor<string, []>("valid")];
338
+ tensor<int32, [1]> var_493_strides_0 = const()[name = tensor<string, []>("op_493_strides_0"), val = tensor<int32, [1]>([1])];
339
+ tensor<int32, [2]> var_493_pad_0 = const()[name = tensor<string, []>("op_493_pad_0"), val = tensor<int32, [2]>([0, 0])];
340
+ tensor<int32, [1]> var_493_dilations_0 = const()[name = tensor<string, []>("op_493_dilations_0"), val = tensor<int32, [1]>([1])];
341
+ tensor<int32, []> var_493_groups_0 = const()[name = tensor<string, []>("op_493_groups_0"), val = tensor<int32, []>(1)];
342
+ tensor<fp32, [1, 256, ?]> var_493 = conv(dilations = var_493_dilations_0, groups = var_493_groups_0, pad = var_493_pad_0, pad_type = var_493_pad_type_0, strides = var_493_strides_0, weight = weight_45_palettized, x = input_71)[name = tensor<string, []>("op_493")];
343
+ tensor<fp32, [1, 256, ?]> var_494 = add(x = out_9, y = var_493)[name = tensor<string, []>("op_494")];
344
+ tensor<fp32, []> var_495 = const()[name = tensor<string, []>("op_495"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
345
+ tensor<fp32, [1, 256, ?]> input_73 = mul(x = var_494, y = var_495)[name = tensor<string, []>("input_73")];
346
+ tensor<fp32, []> var_498 = const()[name = tensor<string, []>("op_498"), val = tensor<fp32, []>(0x1.99999ap-3)];
347
+ tensor<fp32, []> var_501 = const()[name = tensor<string, []>("op_501"), val = tensor<fp32, []>(0x1.4f8b58p-17)];
348
+ tensor<fp32, [1, 512]> h_41 = linear(bias = N_2_norm1_fc_bias, weight = N_2_norm1_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_10")];
349
+ tensor<int32, [3]> var_524 = const()[name = tensor<string, []>("op_524"), val = tensor<int32, [3]>([1, 512, 1])];
350
+ tensor<fp32, [1, 512, 1]> h_43 = reshape(shape = var_524, x = h_41)[name = tensor<string, []>("h_43")];
351
+ tensor<int32, [2]> var_526_split_sizes_0 = const()[name = tensor<string, []>("op_526_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
352
+ tensor<int32, []> var_526_axis_0 = const()[name = tensor<string, []>("op_526_axis_0"), val = tensor<int32, []>(1)];
353
+ tensor<fp32, [1, 256, 1]> var_526_0, tensor<fp32, [1, 256, 1]> var_526_1 = split(axis = var_526_axis_0, split_sizes = var_526_split_sizes_0, x = h_43)[name = tensor<string, []>("op_526")];
354
+ tensor<fp32, []> var_528_promoted = const()[name = tensor<string, []>("op_528_promoted"), val = tensor<fp32, []>(0x1p+0)];
355
+ tensor<fp32, [1, 256, 1]> var_529 = add(x = var_526_0, y = var_528_promoted)[name = tensor<string, []>("op_529")];
356
+ tensor<fp32, [1, 256, ?]> var_532 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_501, gamma = F0_1_norm2_norm_weight, x = input_73)[name = tensor<string, []>("op_532")];
357
+ tensor<fp32, [1, 256, ?]> var_533 = mul(x = var_529, y = var_532)[name = tensor<string, []>("op_533")];
358
+ tensor<fp32, [1, 256, ?]> input_75 = add(x = var_533, y = var_526_1)[name = tensor<string, []>("input_75")];
359
+ tensor<fp32, [1, 256, ?]> input_77 = leaky_relu(alpha = var_498, x = input_75)[name = tensor<string, []>("input_77")];
360
+ tensor<fp32, [256, 256, 3]> weight_49_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8118144))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8314816))), name = tensor<string, []>("weight_49_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
361
+ tensor<string, []> input_79_pad_type_0 = const()[name = tensor<string, []>("input_79_pad_type_0"), val = tensor<string, []>("custom")];
362
+ tensor<int32, [2]> input_79_pad_0 = const()[name = tensor<string, []>("input_79_pad_0"), val = tensor<int32, [2]>([1, 1])];
363
+ tensor<int32, [1]> input_79_strides_0 = const()[name = tensor<string, []>("input_79_strides_0"), val = tensor<int32, [1]>([1])];
364
+ tensor<int32, [1]> input_79_dilations_0 = const()[name = tensor<string, []>("input_79_dilations_0"), val = tensor<int32, [1]>([1])];
365
+ tensor<int32, []> input_79_groups_0 = const()[name = tensor<string, []>("input_79_groups_0"), val = tensor<int32, []>(1)];
366
+ tensor<fp32, [1, 256, ?]> input_79 = conv(bias = N_2_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_49_palettized, x = input_77)[name = tensor<string, []>("input_79")];
367
+ tensor<fp32, [1, 512]> h_45 = linear(bias = N_2_norm2_fc_bias, weight = N_2_norm2_fc_weight_palettized, x = cast_1)[name = tensor<string, []>("linear_11")];
368
+ tensor<int32, [3]> var_550 = const()[name = tensor<string, []>("op_550"), val = tensor<int32, [3]>([1, 512, 1])];
369
+ tensor<fp32, [1, 512, 1]> h = reshape(shape = var_550, x = h_45)[name = tensor<string, []>("h")];
370
+ tensor<int32, [2]> var_552_split_sizes_0 = const()[name = tensor<string, []>("op_552_split_sizes_0"), val = tensor<int32, [2]>([256, 256])];
371
+ tensor<int32, []> var_552_axis_0 = const()[name = tensor<string, []>("op_552_axis_0"), val = tensor<int32, []>(1)];
372
+ tensor<fp32, [1, 256, 1]> var_552_0, tensor<fp32, [1, 256, 1]> var_552_1 = split(axis = var_552_axis_0, split_sizes = var_552_split_sizes_0, x = h)[name = tensor<string, []>("op_552")];
373
+ tensor<fp32, []> var_554_promoted = const()[name = tensor<string, []>("op_554_promoted"), val = tensor<fp32, []>(0x1p+0)];
374
+ tensor<fp32, [1, 256, 1]> var_555 = add(x = var_552_0, y = var_554_promoted)[name = tensor<string, []>("op_555")];
375
+ tensor<fp32, [1, 256, ?]> var_558 = instance_norm(beta = F0_1_norm2_norm_bias, epsilon = var_501, gamma = F0_1_norm2_norm_weight, x = input_79)[name = tensor<string, []>("op_558")];
376
+ tensor<fp32, [1, 256, ?]> var_559 = mul(x = var_555, y = var_558)[name = tensor<string, []>("op_559")];
377
+ tensor<fp32, [1, 256, ?]> input_81 = add(x = var_559, y = var_552_1)[name = tensor<string, []>("input_81")];
378
+ tensor<fp32, [1, 256, ?]> input_83 = leaky_relu(alpha = var_498, x = input_81)[name = tensor<string, []>("input_83")];
379
+ tensor<fp32, [256, 256, 3]> weight_53_palettized = constexpr_lut_to_dense()[indices = tensor<uint8, [196608]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8315904))), lut = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8512576))), name = tensor<string, []>("weight_53_palettized"), shape = tensor<uint32, [3]>([256, 256, 3])];
380
+ tensor<string, []> out_pad_type_0 = const()[name = tensor<string, []>("out_pad_type_0"), val = tensor<string, []>("custom")];
381
+ tensor<int32, [2]> out_pad_0 = const()[name = tensor<string, []>("out_pad_0"), val = tensor<int32, [2]>([1, 1])];
382
+ tensor<int32, [1]> out_strides_0 = const()[name = tensor<string, []>("out_strides_0"), val = tensor<int32, [1]>([1])];
383
+ tensor<int32, [1]> out_dilations_0 = const()[name = tensor<string, []>("out_dilations_0"), val = tensor<int32, [1]>([1])];
384
+ tensor<int32, []> out_groups_0 = const()[name = tensor<string, []>("out_groups_0"), val = tensor<int32, []>(1)];
385
+ tensor<fp32, [1, 256, ?]> out = conv(bias = N_2_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_53_palettized, x = input_83)[name = tensor<string, []>("out")];
386
+ tensor<fp32, [1, 256, ?]> var_569 = add(x = out, y = input_73)[name = tensor<string, []>("op_569")];
387
+ tensor<fp32, []> var_570 = const()[name = tensor<string, []>("op_570"), val = tensor<fp32, []>(0x1.6a09e6p-1)];
388
+ tensor<fp32, [1, 256, ?]> input = mul(x = var_569, y = var_570)[name = tensor<string, []>("input")];
389
+ tensor<string, []> var_582_pad_type_0 = const()[name = tensor<string, []>("op_582_pad_type_0"), val = tensor<string, []>("valid")];
390
+ tensor<int32, [1]> var_582_strides_0 = const()[name = tensor<string, []>("op_582_strides_0"), val = tensor<int32, [1]>([1])];
391
+ tensor<int32, [2]> var_582_pad_0 = const()[name = tensor<string, []>("op_582_pad_0"), val = tensor<int32, [2]>([0, 0])];
392
+ tensor<int32, [1]> var_582_dilations_0 = const()[name = tensor<string, []>("op_582_dilations_0"), val = tensor<int32, [1]>([1])];
393
+ tensor<int32, []> var_582_groups_0 = const()[name = tensor<string, []>("op_582_groups_0"), val = tensor<int32, []>(1)];
394
+ tensor<fp32, [1, 1, ?]> var_582 = conv(bias = N_proj_bias, dilations = var_582_dilations_0, groups = var_582_groups_0, pad = var_582_pad_0, pad_type = var_582_pad_type_0, strides = var_582_strides_0, weight = N_proj_weight, x = input)[name = tensor<string, []>("op_582")];
395
+ tensor<int32, [1]> var_584_axes_0 = const()[name = tensor<string, []>("op_584_axes_0"), val = tensor<int32, [1]>([1])];
396
+ tensor<fp32, [1, ?]> N = squeeze(axes = var_584_axes_0, x = var_582)[name = tensor<string, []>("op_584")];
397
+ } -> (F0, N);
398
+ }
ANE/KokoroProsody_v2.mlmodelc/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70eea4cd2523992fa1fea840e31ab6f4fdd69c1598e220985a2ae6ea5da1e944
3
+ size 8513664