Upload 42 files

conformer_streaming.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bdae7c868374f23bd22a0a9dc48b8e98473762558feba961b9f88cbcb880b8b
+size 243

conformer_streaming.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5216b17be13bf211fe2c3399f87be509502e46ea959eb360793d9ddd9e7e0124
+size 624

conformer_streaming.mlmodelc/metadata.json

@@ -0,0 +1,167 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "shortDescription" : "ConformerStreaming",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 512 × 16)",
+        "shortDescription" : "",
+        "shape" : "[1, 512, 16]",
+        "name" : "encoder",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "encoder_length",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 17 × 1 × 70 × 512)",
+        "shortDescription" : "",
+        "shape" : "[17, 1, 70, 512]",
+        "name" : "new_cache_channel",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 17 × 1 × 512 × 8)",
+        "shortDescription" : "",
+        "shape" : "[17, 1, 512, 8]",
+        "name" : "new_cache_time",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "new_cache_len",
+        "type" : "MultiArray"
+      }
+    ],
+    "storagePrecision" : "Float16",
+    "modelParameters" : [
+
+    ],
+    "author" : "Fluid Inference",
+    "specificationVersion" : 8,
+    "mlProgramOperationTypeHistogram" : {
+      "Ios17.reshape" : 102,
+      "Ios17.logicalAnd" : 3,
+      "Ios16.softmax" : 17,
+      "Ios17.matmul" : 51,
+      "Ios17.transpose" : 155,
+      "Split" : 17,
+      "Ios17.expandDims" : 5,
+      "Select" : 51,
+      "Ios17.add" : 122,
+      "Tile" : 1,
+      "Ios17.sliceByIndex" : 139,
+      "Ios16.sigmoid" : 17,
+      "Pad" : 17,
+      "Ios17.logicalNot" : 2,
+      "Ios17.layerNorm" : 102,
+      "Ios17.less" : 1,
+      "Ios17.sub" : 1,
+      "Ios17.clip" : 2,
+      "Ios17.cast" : 10,
+      "Ios16.silu" : 51,
+      "Ios17.linear" : 136,
+      "Ios17.concat" : 51,
+      "Ios17.greaterEqual" : 1,
+      "Ios17.conv" : 51,
+      "Stack" : 2,
+      "Ios17.mul" : 69
+    },
+    "computePrecision" : "Mixed (Float16, Float32, Int32)",
+    "isUpdatable" : "0",
+    "stateSchema" : [
+
+    ],
+    "availability" : {
+      "macOS" : "14.0",
+      "tvOS" : "17.0",
+      "visionOS" : "1.0",
+      "watchOS" : "10.0",
+      "iOS" : "17.0",
+      "macCatalyst" : "17.0"
+    },
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 18 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 18, 512]",
+        "name" : "pre_encoded",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "pre_encoded_length",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 17 × 1 × 70 × 512)",
+        "shortDescription" : "",
+        "shape" : "[17, 1, 70, 512]",
+        "name" : "cache_last_channel",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 17 × 1 × 512 × 8)",
+        "shortDescription" : "",
+        "shape" : "[17, 1, 512, 8]",
+        "name" : "cache_last_time",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "cache_last_channel_len",
+        "type" : "MultiArray"
+      }
+    ],
+    "userDefinedMetadata" : {
+      "com.github.apple.coremltools.source_dialect" : "TorchScript",
+      "com.github.apple.coremltools.source" : "torch==2.4.0",
+      "com.github.apple.coremltools.version" : "8.3.0"
+    },
+    "generatedClassName" : "conformer_streaming",
+    "method" : "predict"
+  }
+]

conformer_streaming.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc803a8f01e659ff3a72c33142841d16a9ce35761e9e3b59732d5f2144a280f9
+size 208407616

conformer_streaming.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd73072f0d5ec05510531ad65278caeae1eca3db0ad8e0272eb328934b89217a
+size 513705

conformer_streaming.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc803a8f01e659ff3a72c33142841d16a9ce35761e9e3b59732d5f2144a280f9
+size 208407616

conformer_streaming.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "CD5C49FD-C6DE-43DF-8727-686BAB1E82D8": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        },
+        "EA8A367B-5B03-442B-9E15-5678C18CE731": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        }
+    },
+    "rootModelIdentifier": "EA8A367B-5B03-442B-9E15-5678C18CE731"
+}

decoder.mlmodelc/analytics/coremldata.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a74ae596105c77f3b467d3cd86621a5dadaf34de07782e0e787ddc821244c88f
+oid sha256:1e739d0ede0b1992e22d740c0f7af587e8f3cf95fa3717e1757b0c8cfe1ca94f
 size 243

decoder.mlmodelc/coremldata.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8f2bebfa8881714759c1c2f15eab02962590d72d2edc50a3b316231b71c2ed10
+oid sha256:8ab259f1638ef9995a342c1a36d1768f9b0f6152a0fc9c832a7b1e972ea2095a
 size 458

decoder.mlmodelc/metadata.json

@@ -111,7 +111,7 @@
     "userDefinedMetadata" : {
       "com.github.apple.coremltools.source_dialect" : "TorchScript",
       "com.github.apple.coremltools.version" : "8.3.0",
-      "com.github.apple.coremltools.source" : "torch==2.8.0"
+      "com.github.apple.coremltools.source" : "torch==2.4.0"
     },
     "generatedClassName" : "decoder",
     "method" : "predict"

decoder.mlmodelc/model.mil

@@ -1,5 +1,5 @@
 program(1.0)
-[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3500.14.1"}, {"coremlc-version", "3500.32.1"}, {"coremltools-component-torch", "2.8.0"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "8.3.0"}})]
+[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3500.14.1"}, {"coremlc-version", "3500.32.1"}, {"coremltools-component-torch", "2.4.0"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "8.3.0"}})]
 {
     func main<ios17>(tensor<fp32, [1, 1, 640]> c_in, tensor<fp32, [1, 1, 640]> h_in, tensor<int32, [1]> target_length, tensor<int32, [1, 1]> targets) {
             tensor<int32, []> var_15 = const()[name = tensor<string, []>("op_15"), val = tensor<int32, []>(1)];

decoder.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cb554462b9b0167b04b863f08714ba28403e22775a9a5c246b8d1ae54a5dba55
+oid sha256:c1244c8436e6a6be286408bbeb406e289de1fc97d91b542b1ec4446baefff1d3
 size 7265

decoder.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "9125BEB5-F2CB-435E-9229-CC76BD95DF16": {
-            "author": "com.apple.CoreML",
-            "description": "CoreML Model Weights",
-            "name": "weights",
-            "path": "com.apple.CoreML/weights"
-        },
-        "C6EF3FD9-9BAA-4C33-A4E1-39278E0AFAA1": {
+        "1ED1A279-6262-48EA-A0AB-BDC71BF0E685": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "23073A42-CE49-4FD6-804B-0202E5F2F5FE": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
         }
     },
-    "rootModelIdentifier": "C6EF3FD9-9BAA-4C33-A4E1-39278E0AFAA1"
+    "rootModelIdentifier": "1ED1A279-6262-48EA-A0AB-BDC71BF0E685"
 }

joint_decision.mlmodelc/analytics/coremldata.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f48b187ef4c7a01beb197b4ac280f74858d90999fe6392ec7b7d58de554396b7
+oid sha256:bf591aad6f49236974c15ec2c458e3f6149f4b14e4e86ec7c793e36b6168a9c8
 size 243

joint_decision.mlmodelc/coremldata.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b62ea3fa42ff5d5fc5ac4949f2df1a6392f46a58a35a7c032be21a1aaf64b50a
+oid sha256:b5ba480215cdc9d07c66536da11e3219973373d5292f722f0d876168c809302d
 size 467

joint_decision.mlmodelc/metadata.json

@@ -102,9 +102,9 @@
       }
     ],
     "userDefinedMetadata" : {
-      "com.github.apple.coremltools.version" : "8.3.0",
       "com.github.apple.coremltools.source_dialect" : "TorchScript",
-      "com.github.apple.coremltools.source" : "torch==2.8.0"
+      "com.github.apple.coremltools.version" : "8.3.0",
+      "com.github.apple.coremltools.source" : "torch==2.4.0"
     },
     "generatedClassName" : "joint_decision",
     "method" : "predict"

joint_decision.mlmodelc/model.mil

@@ -1,5 +1,5 @@
 program(1.0)
-[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3500.14.1"}, {"coremlc-version", "3500.32.1"}, {"coremltools-component-torch", "2.8.0"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "8.3.0"}})]
+[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3500.14.1"}, {"coremlc-version", "3500.32.1"}, {"coremltools-component-torch", "2.4.0"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "8.3.0"}})]
 {
     func main<ios17>(tensor<fp32, [1, 640, 1]> decoder_step, tensor<fp32, [1, 512, 1]> encoder_step) {
             tensor<int32, [3]> input_1_perm_0 = const()[name = tensor<string, []>("input_1_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];

joint_decision.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2c69ef246a51488ce5f43fe659396440309042d47185a9d8047ed2f9bdef93bc
+oid sha256:056030ae3bdaa34798599fa85d65161620e37b57736bad36e66bcc325a0697a6
 size 8659

joint_decision.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "51CAC0ED-5C70-4C3C-95BB-9912AEDB5233": {
+        "0CF23592-7D60-49DD-9E5B-9AFE5264788C": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
         },
-        "E2870F2D-A319-45D9-BC41-1051355BD8A4": {
+        "9E9496C6-3C38-4F1C-A99D-D6110DDE81CB": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "E2870F2D-A319-45D9-BC41-1051355BD8A4"
+    "rootModelIdentifier": "9E9496C6-3C38-4F1C-A99D-D6110DDE81CB"
 }

metadata.json

@@ -6,7 +6,7 @@
   "mel_dim": 128,
   "hidden_dim": 512,
   "num_layers": 17,
-  "mel_frames_per_chunk": 493,
+  "mel_frames_per_chunk": 1005,
   "vocab_size": 1026,
   "blank_id": 1026,
   "decoder_hidden": 640,

pre_encode.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b2d216925c6a5984440394c017bf480966f7208197209d7d0420022033f6ecd
+size 243

pre_encode.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21af7edc583391f2ad383e01381b448c2c0507f2ff2e8f33ca4dd2eb17b136b6
+size 458

pre_encode.mlmodelc/metadata.json

@@ -0,0 +1,115 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "shortDescription" : "PreEncode",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 18 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 18, 512]",
+        "name" : "pre_encoded",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "pre_encoded_len",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 9 × 128)",
+        "shortDescription" : "",
+        "shape" : "[1, 9, 128]",
+        "name" : "new_pre_cache",
+        "type" : "MultiArray"
+      }
+    ],
+    "storagePrecision" : "Float32",
+    "modelParameters" : [
+
+    ],
+    "author" : "Fluid Inference",
+    "specificationVersion" : 8,
+    "mlProgramOperationTypeHistogram" : {
+      "Identity" : 1,
+      "Ios17.mul" : 3,
+      "Ios17.linear" : 1,
+      "Ios17.floor" : 3,
+      "Ios17.transpose" : 2,
+      "Ios17.conv" : 5,
+      "Ios17.concat" : 1,
+      "Ios17.add" : 4,
+      "Ios17.sliceByIndex" : 1,
+      "Ios16.relu" : 3,
+      "Ios17.expandDims" : 1,
+      "Ios17.cast" : 2,
+      "Ios17.reshape" : 1,
+      "Pad" : 3
+    },
+    "computePrecision" : "Mixed (Float32, Int32)",
+    "isUpdatable" : "0",
+    "stateSchema" : [
+
+    ],
+    "availability" : {
+      "macOS" : "14.0",
+      "tvOS" : "17.0",
+      "visionOS" : "1.0",
+      "watchOS" : "10.0",
+      "iOS" : "17.0",
+      "macCatalyst" : "17.0"
+    },
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 128 × 128)",
+        "shortDescription" : "",
+        "shape" : "[1, 128, 128]",
+        "name" : "mel",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "mel_length",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 9 × 128)",
+        "shortDescription" : "",
+        "shape" : "[1, 9, 128]",
+        "name" : "pre_cache",
+        "type" : "MultiArray"
+      }
+    ],
+    "userDefinedMetadata" : {
+      "com.github.apple.coremltools.source_dialect" : "TorchScript",
+      "com.github.apple.coremltools.version" : "8.3.0",
+      "com.github.apple.coremltools.source" : "torch==2.4.0"
+    },
+    "generatedClassName" : "pre_encode",
+    "method" : "predict"
+  }
+]

pre_encode.mlmodelc/model.mil

@@ -0,0 +1,101 @@
+program(1.0)
+[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3500.14.1"}, {"coremlc-version", "3500.32.1"}, {"coremltools-component-torch", "2.4.0"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "8.3.0"}})]
+{
+    func main<ios17>(tensor<fp32, [1, 128, 128]> mel, tensor<int32, [1]> mel_length, tensor<fp32, [1, 9, 128]> pre_cache) {
+            tensor<fp32, [256]> pre_encode_conv_0_bias = const()[name = tensor<string, []>("pre_encode_conv_0_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(64)))];
+            tensor<fp32, [256, 1, 3, 3]> pre_encode_conv_0_weight = const()[name = tensor<string, []>("pre_encode_conv_0_weight"), val = tensor<fp32, [256, 1, 3, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1152)))];
+            tensor<fp32, [256]> pre_encode_conv_2_bias = const()[name = tensor<string, []>("pre_encode_conv_2_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(10432)))];
+            tensor<fp32, [256, 1, 3, 3]> pre_encode_conv_2_weight = const()[name = tensor<string, []>("pre_encode_conv_2_weight"), val = tensor<fp32, [256, 1, 3, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(11520)))];
+            tensor<fp32, [256]> pre_encode_conv_3_bias = const()[name = tensor<string, []>("pre_encode_conv_3_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(20800)))];
+            tensor<fp32, [256, 256, 1, 1]> pre_encode_conv_3_weight = const()[name = tensor<string, []>("pre_encode_conv_3_weight"), val = tensor<fp32, [256, 256, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(21888)))];
+            tensor<fp32, [256]> pre_encode_conv_5_bias = const()[name = tensor<string, []>("pre_encode_conv_5_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(284096)))];
+            tensor<fp32, [256, 1, 3, 3]> pre_encode_conv_5_weight = const()[name = tensor<string, []>("pre_encode_conv_5_weight"), val = tensor<fp32, [256, 1, 3, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(285184)))];
+            tensor<fp32, [256]> pre_encode_conv_6_bias = const()[name = tensor<string, []>("pre_encode_conv_6_bias"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(294464)))];
+            tensor<fp32, [256, 256, 1, 1]> pre_encode_conv_6_weight = const()[name = tensor<string, []>("pre_encode_conv_6_weight"), val = tensor<fp32, [256, 256, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(295552)))];
+            tensor<fp32, [512]> pre_encode_out_bias = const()[name = tensor<string, []>("pre_encode_out_bias"), val = tensor<fp32, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(557760)))];
+            tensor<fp32, [512, 4352]> pre_encode_out_weight = const()[name = tensor<string, []>("pre_encode_out_weight"), val = tensor<fp32, [512, 4352]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(559872)))];
+            tensor<int32, [3]> mel_perm_0 = const()[name = tensor<string, []>("mel_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
+            tensor<int32, []> var_9 = const()[name = tensor<string, []>("op_9"), val = tensor<int32, []>(1)];
+            tensor<bool, []> var_10_interleave_0 = const()[name = tensor<string, []>("op_10_interleave_0"), val = tensor<bool, []>(false)];
+            tensor<fp32, [1, 128, 128]> mel_1 = transpose(perm = mel_perm_0, x = mel)[name = tensor<string, []>("transpose_1")];
+            tensor<fp32, [1, 137, 128]> var_10 = concat(axis = var_9, interleave = var_10_interleave_0, values = (pre_cache, mel_1))[name = tensor<string, []>("op_10")];
+            tensor<int32, []> var_12 = const()[name = tensor<string, []>("op_12"), val = tensor<int32, []>(9)];
+            tensor<int32, [1]> var_13 = add(x = mel_length, y = var_12)[name = tensor<string, []>("op_13")];
+            tensor<string, []> cast_0_dtype_0 = const()[name = tensor<string, []>("cast_0_dtype_0"), val = tensor<string, []>("fp32")];
+            tensor<fp32, []> _inversed_31_y_0 = const()[name = tensor<string, []>("_inversed_31_y_0"), val = tensor<fp32, []>(0x1p-1)];
+            tensor<fp32, [1]> cast_0 = cast(dtype = cast_0_dtype_0, x = var_13)[name = tensor<string, []>("cast_12")];
+            tensor<fp32, [1]> _inversed_31 = mul(x = cast_0, y = _inversed_31_y_0)[name = tensor<string, []>("_inversed_31")];
+            tensor<fp32, []> var_32 = const()[name = tensor<string, []>("op_32"), val = tensor<fp32, []>(0x1p+0)];
+            tensor<fp32, [1]> lengths_3 = add(x = _inversed_31, y = var_32)[name = tensor<string, []>("lengths_3")];
+            tensor<fp32, [1]> lengths_5 = floor(x = lengths_3)[name = tensor<string, []>("lengths_5")];
+            tensor<fp32, []> _inversed_39_y_0 = const()[name = tensor<string, []>("_inversed_39_y_0"), val = tensor<fp32, []>(0x1p-1)];
+            tensor<fp32, [1]> _inversed_39 = mul(x = lengths_5, y = _inversed_39_y_0)[name = tensor<string, []>("_inversed_39")];
+            tensor<fp32, []> var_40 = const()[name = tensor<string, []>("op_40"), val = tensor<fp32, []>(0x1p+0)];
+            tensor<fp32, [1]> lengths_9 = add(x = _inversed_39, y = var_40)[name = tensor<string, []>("lengths_9")];
+            tensor<fp32, [1]> lengths_11 = floor(x = lengths_9)[name = tensor<string, []>("lengths_11")];
+            tensor<fp32, []> _inversed_47_y_0 = const()[name = tensor<string, []>("_inversed_47_y_0"), val = tensor<fp32, []>(0x1p-1)];
+            tensor<fp32, [1]> _inversed_47 = mul(x = lengths_11, y = _inversed_47_y_0)[name = tensor<string, []>("_inversed_47")];
+            tensor<fp32, []> var_48 = const()[name = tensor<string, []>("op_48"), val = tensor<fp32, []>(0x1p+0)];
+            tensor<fp32, [1]> lengths_15 = add(x = _inversed_47, y = var_48)[name = tensor<string, []>("lengths_15")];
+            tensor<fp32, [1]> lengths = floor(x = lengths_15)[name = tensor<string, []>("lengths")];
+            tensor<string, []> cast_9_dtype_0 = const()[name = tensor<string, []>("cast_9_dtype_0"), val = tensor<string, []>("int32")];
+            tensor<int32, [1]> input_1_axes_0 = const()[name = tensor<string, []>("input_1_axes_0"), val = tensor<int32, [1]>([1])];
+            tensor<fp32, [1, 1, 137, 128]> input_1 = expand_dims(axes = input_1_axes_0, x = var_10)[name = tensor<string, []>("input_1")];
+            tensor<fp32, []> const_0 = const()[name = tensor<string, []>("const_0"), val = tensor<fp32, []>(0x0p+0)];
+            tensor<int32, [8]> input_3_pad_0 = const()[name = tensor<string, []>("input_3_pad_0"), val = tensor<int32, [8]>([0, 0, 0, 0, 2, 1, 2, 1])];
+            tensor<string, []> input_3_mode_0 = const()[name = tensor<string, []>("input_3_mode_0"), val = tensor<string, []>("constant")];
+            tensor<fp32, [1, 1, 140, 131]> input_3 = pad(constant_val = const_0, mode = input_3_mode_0, pad = input_3_pad_0, x = input_1)[name = tensor<string, []>("input_3")];
+            tensor<string, []> input_5_pad_type_0 = const()[name = tensor<string, []>("input_5_pad_type_0"), val = tensor<string, []>("valid")];
+            tensor<int32, [2]> input_5_strides_0 = const()[name = tensor<string, []>("input_5_strides_0"), val = tensor<int32, [2]>([2, 2])];
+            tensor<int32, [4]> input_5_pad_0 = const()[name = tensor<string, []>("input_5_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> input_5_dilations_0 = const()[name = tensor<string, []>("input_5_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, []> input_5_groups_0 = const()[name = tensor<string, []>("input_5_groups_0"), val = tensor<int32, []>(1)];
+            tensor<fp32, [1, 256, 69, 65]> input_5 = conv(bias = pre_encode_conv_0_bias, dilations = input_5_dilations_0, groups = input_5_groups_0, pad = input_5_pad_0, pad_type = input_5_pad_type_0, strides = input_5_strides_0, weight = pre_encode_conv_0_weight, x = input_3)[name = tensor<string, []>("input_5")];
+            tensor<fp32, [1, 256, 69, 65]> input_7 = relu(x = input_5)[name = tensor<string, []>("input_7")];
+            tensor<fp32, []> const_1 = const()[name = tensor<string, []>("const_1"), val = tensor<fp32, []>(0x0p+0)];
+            tensor<int32, [8]> input_9_pad_0 = const()[name = tensor<string, []>("input_9_pad_0"), val = tensor<int32, [8]>([0, 0, 0, 0, 2, 1, 2, 1])];
+            tensor<string, []> input_9_mode_0 = const()[name = tensor<string, []>("input_9_mode_0"), val = tensor<string, []>("constant")];
+            tensor<fp32, [1, 256, 72, 68]> input_9 = pad(constant_val = const_1, mode = input_9_mode_0, pad = input_9_pad_0, x = input_7)[name = tensor<string, []>("input_9")];
+            tensor<string, []> input_11_pad_type_0 = const()[name = tensor<string, []>("input_11_pad_type_0"), val = tensor<string, []>("valid")];
+            tensor<int32, [2]> input_11_strides_0 = const()[name = tensor<string, []>("input_11_strides_0"), val = tensor<int32, [2]>([2, 2])];
+            tensor<int32, []> input_11_groups_0 = const()[name = tensor<string, []>("input_11_groups_0"), val = tensor<int32, []>(256)];
+            tensor<int32, [4]> input_11_pad_0 = const()[name = tensor<string, []>("input_11_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> input_11_dilations_0 = const()[name = tensor<string, []>("input_11_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<fp32, [1, 256, 35, 33]> input_11 = conv(bias = pre_encode_conv_2_bias, dilations = input_11_dilations_0, groups = input_11_groups_0, pad = input_11_pad_0, pad_type = input_11_pad_type_0, strides = input_11_strides_0, weight = pre_encode_conv_2_weight, x = input_9)[name = tensor<string, []>("input_11")];
+            tensor<string, []> input_13_pad_type_0 = const()[name = tensor<string, []>("input_13_pad_type_0"), val = tensor<string, []>("valid")];
+            tensor<int32, [2]> input_13_strides_0 = const()[name = tensor<string, []>("input_13_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> input_13_pad_0 = const()[name = tensor<string, []>("input_13_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> input_13_dilations_0 = const()[name = tensor<string, []>("input_13_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, []> input_13_groups_0 = const()[name = tensor<string, []>("input_13_groups_0"), val = tensor<int32, []>(1)];
+            tensor<fp32, [1, 256, 35, 33]> input_13 = conv(bias = pre_encode_conv_3_bias, dilations = input_13_dilations_0, groups = input_13_groups_0, pad = input_13_pad_0, pad_type = input_13_pad_type_0, strides = input_13_strides_0, weight = pre_encode_conv_3_weight, x = input_11)[name = tensor<string, []>("input_13")];
+            tensor<fp32, [1, 256, 35, 33]> input_15 = relu(x = input_13)[name = tensor<string, []>("input_15")];
+            tensor<fp32, []> const_2 = const()[name = tensor<string, []>("const_2"), val = tensor<fp32, []>(0x0p+0)];
+            tensor<int32, [8]> input_17_pad_0 = const()[name = tensor<string, []>("input_17_pad_0"), val = tensor<int32, [8]>([0, 0, 0, 0, 2, 1, 2, 1])];
+            tensor<string, []> input_17_mode_0 = const()[name = tensor<string, []>("input_17_mode_0"), val = tensor<string, []>("constant")];
+            tensor<fp32, [1, 256, 38, 36]> input_17 = pad(constant_val = const_2, mode = input_17_mode_0, pad = input_17_pad_0, x = input_15)[name = tensor<string, []>("input_17")];
+            tensor<string, []> input_19_pad_type_0 = const()[name = tensor<string, []>("input_19_pad_type_0"), val = tensor<string, []>("valid")];
+            tensor<int32, [2]> input_19_strides_0 = const()[name = tensor<string, []>("input_19_strides_0"), val = tensor<int32, [2]>([2, 2])];
+            tensor<int32, []> input_19_groups_0 = const()[name = tensor<string, []>("input_19_groups_0"), val = tensor<int32, []>(256)];
+            tensor<int32, [4]> input_19_pad_0 = const()[name = tensor<string, []>("input_19_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> input_19_dilations_0 = const()[name = tensor<string, []>("input_19_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<fp32, [1, 256, 18, 17]> input_19 = conv(bias = pre_encode_conv_5_bias, dilations = input_19_dilations_0, groups = input_19_groups_0, pad = input_19_pad_0, pad_type = input_19_pad_type_0, strides = input_19_strides_0, weight = pre_encode_conv_5_weight, x = input_17)[name = tensor<string, []>("input_19")];
+            tensor<string, []> input_21_pad_type_0 = const()[name = tensor<string, []>("input_21_pad_type_0"), val = tensor<string, []>("valid")];
+            tensor<int32, [2]> input_21_strides_0 = const()[name = tensor<string, []>("input_21_strides_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, [4]> input_21_pad_0 = const()[name = tensor<string, []>("input_21_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
+            tensor<int32, [2]> input_21_dilations_0 = const()[name = tensor<string, []>("input_21_dilations_0"), val = tensor<int32, [2]>([1, 1])];
+            tensor<int32, []> input_21_groups_0 = const()[name = tensor<string, []>("input_21_groups_0"), val = tensor<int32, []>(1)];
+            tensor<fp32, [1, 256, 18, 17]> input_21 = conv(bias = pre_encode_conv_6_bias, dilations = input_21_dilations_0, groups = input_21_groups_0, pad = input_21_pad_0, pad_type = input_21_pad_type_0, strides = input_21_strides_0, weight = pre_encode_conv_6_weight, x = input_19)[name = tensor<string, []>("input_21")];
+            tensor<fp32, [1, 256, 18, 17]> x = relu(x = input_21)[name = tensor<string, []>("x")];
+            tensor<int32, [4]> var_104_perm_0 = const()[name = tensor<string, []>("op_104_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
+            tensor<int32, [3]> var_105 = const()[name = tensor<string, []>("op_105"), val = tensor<int32, [3]>([1, 18, -1])];
+            tensor<fp32, [1, 18, 256, 17]> var_104 = transpose(perm = var_104_perm_0, x = x)[name = tensor<string, []>("transpose_0")];
+            tensor<fp32, [1, 18, 4352]> input = reshape(shape = var_105, x = var_104)[name = tensor<string, []>("input")];
+            tensor<fp32, [1, 18, 512]> pre_encoded = linear(bias = pre_encode_out_bias, weight = pre_encode_out_weight, x = input)[name = tensor<string, []>("linear_0")];
+            tensor<int32, [3]> var_122_begin_0 = const()[name = tensor<string, []>("op_122_begin_0"), val = tensor<int32, [3]>([0, 119, 0])];
+            tensor<int32, [3]> var_122_end_0 = const()[name = tensor<string, []>("op_122_end_0"), val = tensor<int32, [3]>([1, 128, 128])];
+            tensor<bool, [3]> var_122_end_mask_0 = const()[name = tensor<string, []>("op_122_end_mask_0"), val = tensor<bool, [3]>([true, true, true])];
+            tensor<fp32, [1, 9, 128]> var_122 = slice_by_index(begin = var_122_begin_0, end = var_122_end_0, end_mask = var_122_end_mask_0, x = mel_1)[name = tensor<string, []>("op_122")];
+            tensor<fp32, [1, 9, 128]> new_pre_cache = identity(x = var_122)[name = tensor<string, []>("op_127")];
+            tensor<int32, [1]> pre_encoded_len = cast(dtype = cast_9_dtype_0, x = lengths)[name = tensor<string, []>("cast_11")];
+        } -> (pre_encoded, pre_encoded_len, new_pre_cache);
+}

pre_encode.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:842054a63f229efc5c0bd938c05a044631797a2b856ad1aef27aba0db3177d0e
+size 9472832

pre_encode.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:038662380cc73aa6d4f177c394dea5567781a03bb8498868ebe6f58d4a95f98c
+size 13455

pre_encode.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:842054a63f229efc5c0bd938c05a044631797a2b856ad1aef27aba0db3177d0e
+size 9472832

pre_encode.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "21700551-3429-4543-9ED4-0EFEF9372D61": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        },
+        "292518D4-FD04-458B-8541-CE9A213C40B8": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        }
+    },
+    "rootModelIdentifier": "292518D4-FD04-458B-8541-CE9A213C40B8"
+}