aotrih commited on
Commit
32b0bd4
1 Parent(s): 620fcd0

Support openai_whisper-large-v3-v20240930

Browse files
.gitignore ADDED
@@ -0,0 +1 @@
 
 
1
+ .DS_Store
openai_whisper-large-v3-v20240930/AudioEncoder.mlmodelc/analytics/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:742c7bd747be2f2a27116a6ebe8ab4675babda518d2123f3efd774a890bc8304
3
+ size 243
openai_whisper-large-v3-v20240930/AudioEncoder.mlmodelc/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6345ee37d7bf7e45d4766046fa04822eb9e13bf144469541e9c1a3e0b911d2fe
3
+ size 348
openai_whisper-large-v3-v20240930/AudioEncoder.mlmodelc/metadata.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "metadataOutputVersion" : "3.0",
4
+ "storagePrecision" : "Float16",
5
+ "outputSchema" : [
6
+ {
7
+ "hasShapeFlexibility" : "0",
8
+ "isOptional" : "0",
9
+ "dataType" : "Float16",
10
+ "formattedType" : "MultiArray (Float16 1 × 1280 × 1 × 1500)",
11
+ "shortDescription" : "",
12
+ "shape" : "[1, 1280, 1, 1500]",
13
+ "name" : "encoder_output_embeds",
14
+ "type" : "MultiArray"
15
+ }
16
+ ],
17
+ "modelParameters" : [
18
+
19
+ ],
20
+ "specificationVersion" : 7,
21
+ "mlProgramOperationTypeHistogram" : {
22
+ "Ios16.softmax" : 32,
23
+ "Ios16.add" : 65,
24
+ "Ios16.mul" : 32,
25
+ "Ios16.batchNorm" : 65,
26
+ "Ios16.gelu" : 34,
27
+ "Ios16.reshape" : 128,
28
+ "Ios16.matmul" : 64,
29
+ "Ios16.layerNorm" : 65,
30
+ "Ios16.conv" : 194
31
+ },
32
+ "computePrecision" : "Mixed (Float16, Int32)",
33
+ "isUpdatable" : "0",
34
+ "availability" : {
35
+ "macOS" : "13.0",
36
+ "tvOS" : "16.0",
37
+ "visionOS" : "1.0",
38
+ "watchOS" : "9.0",
39
+ "iOS" : "16.0",
40
+ "macCatalyst" : "16.0"
41
+ },
42
+ "modelType" : {
43
+ "name" : "MLModelType_mlProgram"
44
+ },
45
+ "userDefinedMetadata" : {
46
+ "com.github.apple.coremltools.source_dialect" : "TorchScript",
47
+ "com.github.apple.coremltools.version" : "8.0",
48
+ "com.github.apple.coremltools.source" : "torch==2.4.1"
49
+ },
50
+ "inputSchema" : [
51
+ {
52
+ "hasShapeFlexibility" : "0",
53
+ "isOptional" : "0",
54
+ "dataType" : "Float16",
55
+ "formattedType" : "MultiArray (Float16 1 × 128 × 1 × 3000)",
56
+ "shortDescription" : "",
57
+ "shape" : "[1, 128, 1, 3000]",
58
+ "name" : "melspectrogram_features",
59
+ "type" : "MultiArray"
60
+ }
61
+ ],
62
+ "generatedClassName" : "AudioEncoder",
63
+ "method" : "predict"
64
+ }
65
+ ]
openai_whisper-large-v3-v20240930/AudioEncoder.mlmodelc/model.mil ADDED
The diff for this file is too large to render. See raw diff
 
openai_whisper-large-v3-v20240930/AudioEncoder.mlmodelc/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98daf651a919978e28fe185daf55ce2f70085a8e59fa07fe8a4d08c87d368ae4
3
+ size 1273974400
openai_whisper-large-v3-v20240930/MelSpectrogram.mlmodelc/analytics/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f478e6afa8fdee3711d97b00728f82ee4dcdf2912f7ff033ee0dfcf57bf28bb
3
+ size 243
openai_whisper-large-v3-v20240930/MelSpectrogram.mlmodelc/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a888718e98af679eee42db9e3609627472c32f77e4fdda28f3735960cbf526b3
3
+ size 329
openai_whisper-large-v3-v20240930/MelSpectrogram.mlmodelc/metadata.json ADDED
@@ -0,0 +1,71 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "metadataOutputVersion" : "3.0",
4
+ "storagePrecision" : "Float16",
5
+ "outputSchema" : [
6
+ {
7
+ "hasShapeFlexibility" : "0",
8
+ "isOptional" : "0",
9
+ "dataType" : "Float16",
10
+ "formattedType" : "MultiArray (Float16 1 × 128 × 1 × 3000)",
11
+ "shortDescription" : "",
12
+ "shape" : "[1, 128, 1, 3000]",
13
+ "name" : "melspectrogram_features",
14
+ "type" : "MultiArray"
15
+ }
16
+ ],
17
+ "modelParameters" : [
18
+
19
+ ],
20
+ "specificationVersion" : 8,
21
+ "mlProgramOperationTypeHistogram" : {
22
+ "Ios17.mul" : 2,
23
+ "Ios17.square" : 2,
24
+ "Ios17.sub" : 1,
25
+ "Ios17.matmul" : 1,
26
+ "Ios17.conv" : 2,
27
+ "Ios17.log" : 1,
28
+ "Ios17.add" : 3,
29
+ "Ios17.sliceByIndex" : 1,
30
+ "Ios17.expandDims" : 4,
31
+ "Ios17.maximum" : 1,
32
+ "Ios16.reduceMax" : 1,
33
+ "Ios17.squeeze" : 2,
34
+ "Ios17.reshape" : 2,
35
+ "Identity" : 1,
36
+ "Pad" : 1
37
+ },
38
+ "computePrecision" : "Mixed (Float16, Float32, Int32)",
39
+ "isUpdatable" : "0",
40
+ "availability" : {
41
+ "macOS" : "14.0",
42
+ "tvOS" : "17.0",
43
+ "visionOS" : "1.0",
44
+ "watchOS" : "10.0",
45
+ "iOS" : "17.0",
46
+ "macCatalyst" : "17.0"
47
+ },
48
+ "modelType" : {
49
+ "name" : "MLModelType_mlProgram"
50
+ },
51
+ "userDefinedMetadata" : {
52
+ "com.github.apple.coremltools.source_dialect" : "TorchScript",
53
+ "com.github.apple.coremltools.source" : "torch==2.2.1",
54
+ "com.github.apple.coremltools.version" : "7.1"
55
+ },
56
+ "inputSchema" : [
57
+ {
58
+ "hasShapeFlexibility" : "0",
59
+ "isOptional" : "0",
60
+ "dataType" : "Float16",
61
+ "formattedType" : "MultiArray (Float16 480000)",
62
+ "shortDescription" : "",
63
+ "shape" : "[480000]",
64
+ "name" : "audio",
65
+ "type" : "MultiArray"
66
+ }
67
+ ],
68
+ "generatedClassName" : "MelSpectrogram",
69
+ "method" : "predict"
70
+ }
71
+ ]
openai_whisper-large-v3-v20240930/MelSpectrogram.mlmodelc/model.mil ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ program(1.0)
2
+ [buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3304.5.2"}, {"coremlc-version", "3304.6.2"}, {"coremltools-component-torch", "2.2.1"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "7.1"}})]
3
+ {
4
+ func main<ios17>(tensor<fp16, [480000]> audio) {
5
+ tensor<int32, [3]> var_10 = const()[name = tensor<string, []>("op_10"), val = tensor<int32, [3]>([1, 1, 480000])];
6
+ tensor<fp16, [1, 1, 480000]> input_1_cast_fp16 = reshape(shape = var_10, x = audio)[name = tensor<string, []>("input_1_cast_fp16")];
7
+ tensor<int32, [6]> input_3_pad_0 = const()[name = tensor<string, []>("input_3_pad_0"), val = tensor<int32, [6]>([0, 0, 0, 0, 200, 200])];
8
+ tensor<string, []> input_3_mode_0 = const()[name = tensor<string, []>("input_3_mode_0"), val = tensor<string, []>("reflect")];
9
+ tensor<fp16, []> input_3_constant_val_0_to_fp16 = const()[name = tensor<string, []>("input_3_constant_val_0_to_fp16"), val = tensor<fp16, []>(0x0p+0)];
10
+ tensor<fp16, [1, 1, 480400]> input_3_cast_fp16 = pad(constant_val = input_3_constant_val_0_to_fp16, mode = input_3_mode_0, pad = input_3_pad_0, x = input_1_cast_fp16)[name = tensor<string, []>("input_3_cast_fp16")];
11
+ tensor<int32, [1]> var_22 = const()[name = tensor<string, []>("op_22"), val = tensor<int32, [1]>([480400])];
12
+ tensor<fp16, [480400]> input_cast_fp16 = reshape(shape = var_22, x = input_3_cast_fp16)[name = tensor<string, []>("input_cast_fp16")];
13
+ tensor<int32, [1]> expand_dims_0_axes_0 = const()[name = tensor<string, []>("expand_dims_0_axes_0"), val = tensor<int32, [1]>([0])];
14
+ tensor<fp16, [1, 480400]> expand_dims_0_cast_fp16 = expand_dims(axes = expand_dims_0_axes_0, x = input_cast_fp16)[name = tensor<string, []>("expand_dims_0_cast_fp16")];
15
+ tensor<int32, [1]> expand_dims_3 = const()[name = tensor<string, []>("expand_dims_3"), val = tensor<int32, [1]>([160])];
16
+ tensor<int32, [1]> expand_dims_4_axes_0 = const()[name = tensor<string, []>("expand_dims_4_axes_0"), val = tensor<int32, [1]>([1])];
17
+ tensor<fp16, [1, 1, 480400]> expand_dims_4_cast_fp16 = expand_dims(axes = expand_dims_4_axes_0, x = expand_dims_0_cast_fp16)[name = tensor<string, []>("expand_dims_4_cast_fp16")];
18
+ tensor<string, []> conv_0_pad_type_0 = const()[name = tensor<string, []>("conv_0_pad_type_0"), val = tensor<string, []>("valid")];
19
+ tensor<int32, [2]> conv_0_pad_0 = const()[name = tensor<string, []>("conv_0_pad_0"), val = tensor<int32, [2]>([0, 0])];
20
+ tensor<int32, [1]> conv_0_dilations_0 = const()[name = tensor<string, []>("conv_0_dilations_0"), val = tensor<int32, [1]>([1])];
21
+ tensor<int32, []> conv_0_groups_0 = const()[name = tensor<string, []>("conv_0_groups_0"), val = tensor<int32, []>(1)];
22
+ tensor<fp16, [201, 1, 400]> expand_dims_1_to_fp16 = const()[name = tensor<string, []>("expand_dims_1_to_fp16"), val = tensor<fp16, [201, 1, 400]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(64)))];
23
+ tensor<fp16, [1, 201, 3001]> conv_0_cast_fp16 = conv(dilations = conv_0_dilations_0, groups = conv_0_groups_0, pad = conv_0_pad_0, pad_type = conv_0_pad_type_0, strides = expand_dims_3, weight = expand_dims_1_to_fp16, x = expand_dims_4_cast_fp16)[name = tensor<string, []>("conv_0_cast_fp16")];
24
+ tensor<string, []> conv_1_pad_type_0 = const()[name = tensor<string, []>("conv_1_pad_type_0"), val = tensor<string, []>("valid")];
25
+ tensor<int32, [2]> conv_1_pad_0 = const()[name = tensor<string, []>("conv_1_pad_0"), val = tensor<int32, [2]>([0, 0])];
26
+ tensor<int32, [1]> conv_1_dilations_0 = const()[name = tensor<string, []>("conv_1_dilations_0"), val = tensor<int32, [1]>([1])];
27
+ tensor<int32, []> conv_1_groups_0 = const()[name = tensor<string, []>("conv_1_groups_0"), val = tensor<int32, []>(1)];
28
+ tensor<fp16, [201, 1, 400]> expand_dims_2_to_fp16 = const()[name = tensor<string, []>("expand_dims_2_to_fp16"), val = tensor<fp16, [201, 1, 400]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(160960)))];
29
+ tensor<fp16, [1, 201, 3001]> conv_1_cast_fp16 = conv(dilations = conv_1_dilations_0, groups = conv_1_groups_0, pad = conv_1_pad_0, pad_type = conv_1_pad_type_0, strides = expand_dims_3, weight = expand_dims_2_to_fp16, x = expand_dims_4_cast_fp16)[name = tensor<string, []>("conv_1_cast_fp16")];
30
+ tensor<int32, [1]> squeeze_0_axes_0 = const()[name = tensor<string, []>("squeeze_0_axes_0"), val = tensor<int32, [1]>([0])];
31
+ tensor<fp16, [201, 3001]> squeeze_0_cast_fp16 = squeeze(axes = squeeze_0_axes_0, x = conv_0_cast_fp16)[name = tensor<string, []>("squeeze_0_cast_fp16")];
32
+ tensor<int32, [1]> squeeze_1_axes_0 = const()[name = tensor<string, []>("squeeze_1_axes_0"), val = tensor<int32, [1]>([0])];
33
+ tensor<fp16, [201, 3001]> squeeze_1_cast_fp16 = squeeze(axes = squeeze_1_axes_0, x = conv_1_cast_fp16)[name = tensor<string, []>("squeeze_1_cast_fp16")];
34
+ tensor<fp16, [201, 3001]> square_0_cast_fp16 = square(x = squeeze_0_cast_fp16)[name = tensor<string, []>("square_0_cast_fp16")];
35
+ tensor<fp16, [201, 3001]> square_1_cast_fp16 = square(x = squeeze_1_cast_fp16)[name = tensor<string, []>("square_1_cast_fp16")];
36
+ tensor<fp16, [201, 3001]> add_1_cast_fp16 = add(x = square_0_cast_fp16, y = square_1_cast_fp16)[name = tensor<string, []>("add_1_cast_fp16")];
37
+ tensor<fp16, [201, 3001]> magnitudes_1_cast_fp16 = identity(x = add_1_cast_fp16)[name = tensor<string, []>("magnitudes_1_cast_fp16")];
38
+ tensor<int32, [2]> magnitudes_begin_0 = const()[name = tensor<string, []>("magnitudes_begin_0"), val = tensor<int32, [2]>([0, 0])];
39
+ tensor<int32, [2]> magnitudes_end_0 = const()[name = tensor<string, []>("magnitudes_end_0"), val = tensor<int32, [2]>([201, 3000])];
40
+ tensor<bool, [2]> magnitudes_end_mask_0 = const()[name = tensor<string, []>("magnitudes_end_mask_0"), val = tensor<bool, [2]>([true, false])];
41
+ tensor<fp16, [201, 3000]> magnitudes_cast_fp16 = slice_by_index(begin = magnitudes_begin_0, end = magnitudes_end_0, end_mask = magnitudes_end_mask_0, x = magnitudes_1_cast_fp16)[name = tensor<string, []>("magnitudes_cast_fp16")];
42
+ tensor<bool, []> mel_spec_1_transpose_x_0 = const()[name = tensor<string, []>("mel_spec_1_transpose_x_0"), val = tensor<bool, []>(false)];
43
+ tensor<bool, []> mel_spec_1_transpose_y_0 = const()[name = tensor<string, []>("mel_spec_1_transpose_y_0"), val = tensor<bool, []>(false)];
44
+ tensor<fp16, [128, 201]> mel_filters_to_fp16 = const()[name = tensor<string, []>("mel_filters_to_fp16"), val = tensor<fp16, [128, 201]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(321856)))];
45
+ tensor<fp16, [128, 3000]> mel_spec_1_cast_fp16 = matmul(transpose_x = mel_spec_1_transpose_x_0, transpose_y = mel_spec_1_transpose_y_0, x = mel_filters_to_fp16, y = magnitudes_cast_fp16)[name = tensor<string, []>("mel_spec_1_cast_fp16")];
46
+ tensor<fp16, []> var_41_to_fp16 = const()[name = tensor<string, []>("op_41_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
47
+ tensor<fp16, [128, 3000]> mel_spec_cast_fp16 = add(x = mel_spec_1_cast_fp16, y = var_41_to_fp16)[name = tensor<string, []>("mel_spec_cast_fp16")];
48
+ tensor<fp32, []> log_0_epsilon_0 = const()[name = tensor<string, []>("log_0_epsilon_0"), val = tensor<fp32, []>(0x1p-149)];
49
+ tensor<fp16, [128, 3000]> log_0_cast_fp16 = log(epsilon = log_0_epsilon_0, x = mel_spec_cast_fp16)[name = tensor<string, []>("log_0_cast_fp16")];
50
+ tensor<fp16, []> mul_0_y_0_to_fp16 = const()[name = tensor<string, []>("mul_0_y_0_to_fp16"), val = tensor<fp16, []>(0x1.bccp-2)];
51
+ tensor<fp16, [128, 3000]> mul_0_cast_fp16 = mul(x = log_0_cast_fp16, y = mul_0_y_0_to_fp16)[name = tensor<string, []>("mul_0_cast_fp16")];
52
+ tensor<bool, []> var_44_keep_dims_0 = const()[name = tensor<string, []>("op_44_keep_dims_0"), val = tensor<bool, []>(false)];
53
+ tensor<fp16, []> var_44_cast_fp16 = reduce_max(keep_dims = var_44_keep_dims_0, x = mul_0_cast_fp16)[name = tensor<string, []>("op_44_cast_fp16")];
54
+ tensor<fp16, []> var_46_to_fp16 = const()[name = tensor<string, []>("op_46_to_fp16"), val = tensor<fp16, []>(0x1p+3)];
55
+ tensor<fp16, []> var_47_cast_fp16 = sub(x = var_44_cast_fp16, y = var_46_to_fp16)[name = tensor<string, []>("op_47_cast_fp16")];
56
+ tensor<fp16, [128, 3000]> log_spec_3_cast_fp16 = maximum(x = mul_0_cast_fp16, y = var_47_cast_fp16)[name = tensor<string, []>("log_spec_3_cast_fp16")];
57
+ tensor<fp16, []> var_50_to_fp16 = const()[name = tensor<string, []>("op_50_to_fp16"), val = tensor<fp16, []>(0x1p+2)];
58
+ tensor<fp16, [128, 3000]> var_51_cast_fp16 = add(x = log_spec_3_cast_fp16, y = var_50_to_fp16)[name = tensor<string, []>("op_51_cast_fp16")];
59
+ tensor<fp16, []> _inversed_log_spec_y_0_to_fp16 = const()[name = tensor<string, []>("_inversed_log_spec_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-2)];
60
+ tensor<fp16, [128, 3000]> _inversed_log_spec_cast_fp16 = mul(x = var_51_cast_fp16, y = _inversed_log_spec_y_0_to_fp16)[name = tensor<string, []>("_inversed_log_spec_cast_fp16")];
61
+ tensor<int32, [1]> var_55_axes_0 = const()[name = tensor<string, []>("op_55_axes_0"), val = tensor<int32, [1]>([0])];
62
+ tensor<fp16, [1, 128, 3000]> var_55_cast_fp16 = expand_dims(axes = var_55_axes_0, x = _inversed_log_spec_cast_fp16)[name = tensor<string, []>("op_55_cast_fp16")];
63
+ tensor<int32, [1]> var_62_axes_0 = const()[name = tensor<string, []>("op_62_axes_0"), val = tensor<int32, [1]>([2])];
64
+ tensor<fp16, [1, 128, 1, 3000]> melspectrogram_features = expand_dims(axes = var_62_axes_0, x = var_55_cast_fp16)[name = tensor<string, []>("op_62_cast_fp16")];
65
+ } -> (melspectrogram_features);
66
+ }
openai_whisper-large-v3-v20240930/MelSpectrogram.mlmodelc/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81275398516781f9755514a5ab85db4687374dd611013625f3d4493588783968
3
+ size 373376
openai_whisper-large-v3-v20240930/TextDecoder.mlmodelc/analytics/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931efc5180111dbab2542c200f10b326d760c65a42dcf6c5edef73d803b93255
3
+ size 243
openai_whisper-large-v3-v20240930/TextDecoder.mlmodelc/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3faabaf66930e66956d8291d0ff485fb382496e30a91a7185548b9b898ce90a9
3
+ size 633
openai_whisper-large-v3-v20240930/TextDecoder.mlmodelc/metadata.json ADDED
@@ -0,0 +1,165 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "metadataOutputVersion" : "3.0",
4
+ "storagePrecision" : "Float16",
5
+ "outputSchema" : [
6
+ {
7
+ "hasShapeFlexibility" : "0",
8
+ "isOptional" : "0",
9
+ "dataType" : "Float16",
10
+ "formattedType" : "MultiArray (Float16 1 × 1 × 51866)",
11
+ "shortDescription" : "",
12
+ "shape" : "[1, 1, 51866]",
13
+ "name" : "logits",
14
+ "type" : "MultiArray"
15
+ },
16
+ {
17
+ "hasShapeFlexibility" : "0",
18
+ "isOptional" : "0",
19
+ "dataType" : "Float16",
20
+ "formattedType" : "MultiArray (Float16 1 × 5120 × 1 × 1)",
21
+ "shortDescription" : "",
22
+ "shape" : "[1, 5120, 1, 1]",
23
+ "name" : "key_cache_updates",
24
+ "type" : "MultiArray"
25
+ },
26
+ {
27
+ "hasShapeFlexibility" : "0",
28
+ "isOptional" : "0",
29
+ "dataType" : "Float16",
30
+ "formattedType" : "MultiArray (Float16 1 × 5120 × 1 × 1)",
31
+ "shortDescription" : "",
32
+ "shape" : "[1, 5120, 1, 1]",
33
+ "name" : "value_cache_updates",
34
+ "type" : "MultiArray"
35
+ },
36
+ {
37
+ "hasShapeFlexibility" : "0",
38
+ "isOptional" : "0",
39
+ "dataType" : "Float16",
40
+ "formattedType" : "MultiArray (Float16 1 × 1500)",
41
+ "shortDescription" : "",
42
+ "shape" : "[1, 1500]",
43
+ "name" : "alignment_heads_weights",
44
+ "type" : "MultiArray"
45
+ }
46
+ ],
47
+ "modelParameters" : [
48
+
49
+ ],
50
+ "specificationVersion" : 7,
51
+ "mlProgramOperationTypeHistogram" : {
52
+ "Split" : 2,
53
+ "Concat" : 3,
54
+ "Squeeze" : 1,
55
+ "Ios16.mul" : 24,
56
+ "Ios16.layerNorm" : 13,
57
+ "SliceByIndex" : 12,
58
+ "Ios16.sub" : 1,
59
+ "Transpose" : 1,
60
+ "Ios16.conv" : 40,
61
+ "Ios16.add" : 25,
62
+ "Ios16.linear" : 1,
63
+ "Ios16.matmul" : 16,
64
+ "Ios16.gelu" : 4,
65
+ "Ios16.reduceMean" : 1,
66
+ "ExpandDims" : 6,
67
+ "Ios16.batchNorm" : 13,
68
+ "Ios16.gather" : 2,
69
+ "Ios16.reshape" : 32,
70
+ "Ios16.softmax" : 8
71
+ },
72
+ "computePrecision" : "Mixed (Float16, Int32)",
73
+ "isUpdatable" : "0",
74
+ "availability" : {
75
+ "macOS" : "13.0",
76
+ "tvOS" : "16.0",
77
+ "visionOS" : "1.0",
78
+ "watchOS" : "9.0",
79
+ "iOS" : "16.0",
80
+ "macCatalyst" : "16.0"
81
+ },
82
+ "modelType" : {
83
+ "name" : "MLModelType_mlProgram"
84
+ },
85
+ "userDefinedMetadata" : {
86
+ "com.github.apple.coremltools.source_dialect" : "TorchScript",
87
+ "com.github.apple.coremltools.source" : "torch==2.4.1",
88
+ "com.github.apple.coremltools.version" : "8.0"
89
+ },
90
+ "inputSchema" : [
91
+ {
92
+ "hasShapeFlexibility" : "0",
93
+ "isOptional" : "0",
94
+ "dataType" : "Int32",
95
+ "formattedType" : "MultiArray (Int32 1)",
96
+ "shortDescription" : "",
97
+ "shape" : "[1]",
98
+ "name" : "input_ids",
99
+ "type" : "MultiArray"
100
+ },
101
+ {
102
+ "hasShapeFlexibility" : "0",
103
+ "isOptional" : "0",
104
+ "dataType" : "Int32",
105
+ "formattedType" : "MultiArray (Int32 1)",
106
+ "shortDescription" : "",
107
+ "shape" : "[1]",
108
+ "name" : "cache_length",
109
+ "type" : "MultiArray"
110
+ },
111
+ {
112
+ "hasShapeFlexibility" : "0",
113
+ "isOptional" : "0",
114
+ "dataType" : "Float16",
115
+ "formattedType" : "MultiArray (Float16 1 × 5120 × 1 × 448)",
116
+ "shortDescription" : "",
117
+ "shape" : "[1, 5120, 1, 448]",
118
+ "name" : "key_cache",
119
+ "type" : "MultiArray"
120
+ },
121
+ {
122
+ "hasShapeFlexibility" : "0",
123
+ "isOptional" : "0",
124
+ "dataType" : "Float16",
125
+ "formattedType" : "MultiArray (Float16 1 × 5120 × 1 × 448)",
126
+ "shortDescription" : "",
127
+ "shape" : "[1, 5120, 1, 448]",
128
+ "name" : "value_cache",
129
+ "type" : "MultiArray"
130
+ },
131
+ {
132
+ "hasShapeFlexibility" : "0",
133
+ "isOptional" : "0",
134
+ "dataType" : "Float16",
135
+ "formattedType" : "MultiArray (Float16 1 × 448)",
136
+ "shortDescription" : "",
137
+ "shape" : "[1, 448]",
138
+ "name" : "kv_cache_update_mask",
139
+ "type" : "MultiArray"
140
+ },
141
+ {
142
+ "hasShapeFlexibility" : "0",
143
+ "isOptional" : "0",
144
+ "dataType" : "Float16",
145
+ "formattedType" : "MultiArray (Float16 1 × 1280 × 1 × 1500)",
146
+ "shortDescription" : "",
147
+ "shape" : "[1, 1280, 1, 1500]",
148
+ "name" : "encoder_output_embeds",
149
+ "type" : "MultiArray"
150
+ },
151
+ {
152
+ "hasShapeFlexibility" : "0",
153
+ "isOptional" : "0",
154
+ "dataType" : "Float16",
155
+ "formattedType" : "MultiArray (Float16 1 × 448)",
156
+ "shortDescription" : "",
157
+ "shape" : "[1, 448]",
158
+ "name" : "decoder_key_padding_mask",
159
+ "type" : "MultiArray"
160
+ }
161
+ ],
162
+ "generatedClassName" : "TextDecoder",
163
+ "method" : "predict"
164
+ }
165
+ ]
openai_whisper-large-v3-v20240930/TextDecoder.mlmodelc/model.mil ADDED
The diff for this file is too large to render. See raw diff
 
openai_whisper-large-v3-v20240930/TextDecoder.mlmodelc/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:47b2703aa37448e09cf2f06e45984fabd5ded4c34ba3400cec38a5294af39dc1
3
+ size 343933748
openai_whisper-large-v3-v20240930/config.json ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/raid/yoach/tmp_whisper_turbo",
3
+ "activation_dropout": 0.0,
4
+ "activation_function": "gelu",
5
+ "apply_spec_augment": false,
6
+ "architectures": [
7
+ "WhisperForConditionalGeneration"
8
+ ],
9
+ "attention_dropout": 0.0,
10
+ "begin_suppress_tokens": [
11
+ 220,
12
+ 50256
13
+ ],
14
+ "bos_token_id": 50257,
15
+ "classifier_proj_size": 256,
16
+ "d_model": 1280,
17
+ "decoder_attention_heads": 20,
18
+ "decoder_ffn_dim": 5120,
19
+ "decoder_layerdrop": 0.0,
20
+ "decoder_layers": 4,
21
+ "decoder_start_token_id": 50258,
22
+ "dropout": 0.0,
23
+ "encoder_attention_heads": 20,
24
+ "encoder_ffn_dim": 5120,
25
+ "encoder_layerdrop": 0.0,
26
+ "encoder_layers": 32,
27
+ "eos_token_id": 50257,
28
+ "init_std": 0.02,
29
+ "is_encoder_decoder": true,
30
+ "mask_feature_length": 10,
31
+ "mask_feature_min_masks": 0,
32
+ "mask_feature_prob": 0.0,
33
+ "mask_time_length": 10,
34
+ "mask_time_min_masks": 2,
35
+ "mask_time_prob": 0.05,
36
+ "max_source_positions": 1500,
37
+ "max_target_positions": 448,
38
+ "median_filter_width": 7,
39
+ "model_type": "whisper",
40
+ "num_hidden_layers": 32,
41
+ "num_mel_bins": 128,
42
+ "pad_token_id": 50257,
43
+ "scale_embedding": false,
44
+ "torch_dtype": "float16",
45
+ "transformers_version": "4.46.0.dev0",
46
+ "use_cache": true,
47
+ "use_weighted_layer_sum": false,
48
+ "vocab_size": 51866
49
+ }
openai_whisper-large-v3-v20240930/generation_config.json ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignment_heads": [
3
+ [
4
+ 2,
5
+ 4
6
+ ],
7
+ [
8
+ 2,
9
+ 11
10
+ ],
11
+ [
12
+ 3,
13
+ 3
14
+ ],
15
+ [
16
+ 3,
17
+ 6
18
+ ],
19
+ [
20
+ 3,
21
+ 11
22
+ ],
23
+ [
24
+ 3,
25
+ 14
26
+ ]
27
+ ],
28
+ "begin_suppress_tokens": [
29
+ 220,
30
+ 50257
31
+ ],
32
+ "bos_token_id": 50257,
33
+ "decoder_start_token_id": 50258,
34
+ "eos_token_id": 50257,
35
+ "forced_decoder_ids": [
36
+ [
37
+ 1,
38
+ null
39
+ ],
40
+ [
41
+ 2,
42
+ 50360
43
+ ]
44
+ ],
45
+ "is_multilingual": true,
46
+ "lang_to_id": {
47
+ "<|af|>": 50327,
48
+ "<|am|>": 50334,
49
+ "<|ar|>": 50272,
50
+ "<|as|>": 50350,
51
+ "<|az|>": 50304,
52
+ "<|ba|>": 50355,
53
+ "<|be|>": 50330,
54
+ "<|bg|>": 50292,
55
+ "<|bn|>": 50302,
56
+ "<|bo|>": 50347,
57
+ "<|br|>": 50309,
58
+ "<|bs|>": 50315,
59
+ "<|ca|>": 50270,
60
+ "<|cs|>": 50283,
61
+ "<|cy|>": 50297,
62
+ "<|da|>": 50285,
63
+ "<|de|>": 50261,
64
+ "<|el|>": 50281,
65
+ "<|en|>": 50259,
66
+ "<|es|>": 50262,
67
+ "<|et|>": 50307,
68
+ "<|eu|>": 50310,
69
+ "<|fa|>": 50300,
70
+ "<|fi|>": 50277,
71
+ "<|fo|>": 50338,
72
+ "<|fr|>": 50265,
73
+ "<|gl|>": 50319,
74
+ "<|gu|>": 50333,
75
+ "<|haw|>": 50352,
76
+ "<|ha|>": 50354,
77
+ "<|he|>": 50279,
78
+ "<|hi|>": 50276,
79
+ "<|hr|>": 50291,
80
+ "<|ht|>": 50339,
81
+ "<|hu|>": 50286,
82
+ "<|hy|>": 50312,
83
+ "<|id|>": 50275,
84
+ "<|is|>": 50311,
85
+ "<|it|>": 50274,
86
+ "<|ja|>": 50266,
87
+ "<|jw|>": 50356,
88
+ "<|ka|>": 50329,
89
+ "<|kk|>": 50316,
90
+ "<|km|>": 50323,
91
+ "<|kn|>": 50306,
92
+ "<|ko|>": 50264,
93
+ "<|la|>": 50294,
94
+ "<|lb|>": 50345,
95
+ "<|ln|>": 50353,
96
+ "<|lo|>": 50336,
97
+ "<|lt|>": 50293,
98
+ "<|lv|>": 50301,
99
+ "<|mg|>": 50349,
100
+ "<|mi|>": 50295,
101
+ "<|mk|>": 50308,
102
+ "<|ml|>": 50296,
103
+ "<|mn|>": 50314,
104
+ "<|mr|>": 50320,
105
+ "<|ms|>": 50282,
106
+ "<|mt|>": 50343,
107
+ "<|my|>": 50346,
108
+ "<|ne|>": 50313,
109
+ "<|nl|>": 50271,
110
+ "<|nn|>": 50342,
111
+ "<|no|>": 50288,
112
+ "<|oc|>": 50328,
113
+ "<|pa|>": 50321,
114
+ "<|pl|>": 50269,
115
+ "<|ps|>": 50340,
116
+ "<|pt|>": 50267,
117
+ "<|ro|>": 50284,
118
+ "<|ru|>": 50263,
119
+ "<|sa|>": 50344,
120
+ "<|sd|>": 50332,
121
+ "<|si|>": 50322,
122
+ "<|sk|>": 50298,
123
+ "<|sl|>": 50305,
124
+ "<|sn|>": 50324,
125
+ "<|so|>": 50326,
126
+ "<|sq|>": 50317,
127
+ "<|sr|>": 50303,
128
+ "<|su|>": 50357,
129
+ "<|sv|>": 50273,
130
+ "<|sw|>": 50318,
131
+ "<|ta|>": 50287,
132
+ "<|te|>": 50299,
133
+ "<|tg|>": 50331,
134
+ "<|th|>": 50289,
135
+ "<|tk|>": 50341,
136
+ "<|tl|>": 50348,
137
+ "<|tr|>": 50268,
138
+ "<|tt|>": 50351,
139
+ "<|uk|>": 50280,
140
+ "<|ur|>": 50290,
141
+ "<|uz|>": 50337,
142
+ "<|vi|>": 50278,
143
+ "<|yi|>": 50335,
144
+ "<|yo|>": 50325,
145
+ "<|yue|>": 50358,
146
+ "<|zh|>": 50260
147
+ },
148
+ "max_initial_timestamp_index": 50,
149
+ "max_length": 448,
150
+ "no_timestamps_token_id": 50364,
151
+ "pad_token_id": 50257,
152
+ "prev_sot_token_id": 50362,
153
+ "return_timestamps": false,
154
+ "suppress_tokens": [
155
+ 1,
156
+ 2,
157
+ 7,
158
+ 8,
159
+ 9,
160
+ 10,
161
+ 14,
162
+ 25,
163
+ 26,
164
+ 27,
165
+ 28,
166
+ 29,
167
+ 31,
168
+ 58,
169
+ 59,
170
+ 60,
171
+ 61,
172
+ 62,
173
+ 63,
174
+ 90,
175
+ 91,
176
+ 92,
177
+ 93,
178
+ 359,
179
+ 503,
180
+ 522,
181
+ 542,
182
+ 873,
183
+ 893,
184
+ 902,
185
+ 918,
186
+ 922,
187
+ 931,
188
+ 1350,
189
+ 1853,
190
+ 1982,
191
+ 2460,
192
+ 2627,
193
+ 3246,
194
+ 3253,
195
+ 3268,
196
+ 3536,
197
+ 3846,
198
+ 3961,
199
+ 4183,
200
+ 4667,
201
+ 6585,
202
+ 6647,
203
+ 7273,
204
+ 9061,
205
+ 9383,
206
+ 10428,
207
+ 10929,
208
+ 11938,
209
+ 12033,
210
+ 12331,
211
+ 12562,
212
+ 13793,
213
+ 14157,
214
+ 14635,
215
+ 15265,
216
+ 15618,
217
+ 16553,
218
+ 16604,
219
+ 18362,
220
+ 18956,
221
+ 20075,
222
+ 21675,
223
+ 22520,
224
+ 26130,
225
+ 26161,
226
+ 26435,
227
+ 28279,
228
+ 29464,
229
+ 31650,
230
+ 32302,
231
+ 32470,
232
+ 36865,
233
+ 42863,
234
+ 47425,
235
+ 49870,
236
+ 50254,
237
+ 50258,
238
+ 50359,
239
+ 50360,
240
+ 50361,
241
+ 50362,
242
+ 50363
243
+ ],
244
+ "task_to_id": {
245
+ "transcribe": 50360,
246
+ "translate": 50359
247
+ },
248
+ "transformers_version": "4.46.0.dev0"
249
+ }