Upload 5 files

PLDA.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e862420707c6fc86c5720adcbc811286235970cf5e0c5afdb7e10e03c937026
+size 243

PLDA.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d5ddb4cf367ad23fb512587f751cddbe4cbf867f53c5395435e6719a6acfd20
+size 535

PLDA.mlmodelc/metadata.json

@@ -0,0 +1,71 @@
+[
+  {
+    "shortDescription" : "pyannote community-1 PLDA transformation (x-vector whitening + PLDA projection)",
+    "metadataOutputVersion" : "3.0",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 128)",
+        "shortDescription" : "",
+        "shape" : "[1, 128]",
+        "name" : "plda_features",
+        "type" : "MultiArray"
+      }
+    ],
+    "version" : "pyannote-speaker-diarization-community-1",
+    "modelParameters" : [
+
+    ],
+    "author" : "Fluid Inference",
+    "specificationVersion" : 8,
+    "storagePrecision" : "Float32",
+    "license" : "CC-BY-4.0",
+    "mlProgramOperationTypeHistogram" : {
+      "Ios16.reduceSum" : 2,
+      "Ios17.clip" : 2,
+      "Ios17.sqrt" : 2,
+      "Ios17.linear" : 2,
+      "Ios17.realDiv" : 2,
+      "Ios17.mul" : 4,
+      "Ios17.sub" : 3
+    },
+    "computePrecision" : "Mixed (Float32, Int32)",
+    "stateSchema" : [
+
+    ],
+    "isUpdatable" : "0",
+    "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 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 256]",
+        "name" : "embeddings",
+        "type" : "MultiArray"
+      }
+    ],
+    "userDefinedMetadata" : {
+      "com.github.apple.coremltools.conversion_date" : "2025-10-01",
+      "com.github.apple.coremltools.source" : "torch==2.8.0",
+      "com.github.apple.coremltools.version" : "9.0b1",
+      "com.github.apple.coremltools.source_dialect" : "TorchScript"
+    },
+    "generatedClassName" : "plda_community_1",
+    "method" : "predict"
+  }
+]

PLDA.mlmodelc/model.mil

@@ -0,0 +1,39 @@
+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", "9.0b1"}})]
+{
+    func main<ios17>(tensor<fp32, [1, 256]> embeddings) {
+            tensor<fp32, [128, 128]> plda_tr = const()[name = tensor<string, []>("plda_tr"), val = tensor<fp32, [128, 128]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(64)))];
+            tensor<fp32, [128]> mu = const()[name = tensor<string, []>("mu"), val = tensor<fp32, [128]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(65664)))];
+            tensor<fp32, []> lda_dim_scale = const()[name = tensor<string, []>("lda_dim_scale"), val = tensor<fp32, []>(0x1.6a09e6p+3)];
+            tensor<fp32, [128]> mean2 = const()[name = tensor<string, []>("mean2"), val = tensor<fp32, [128]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(66240)))];
+            tensor<fp32, []> lda_scale = const()[name = tensor<string, []>("lda_scale"), val = tensor<fp32, []>(0x1p+4)];
+            tensor<fp32, [256]> mean1 = const()[name = tensor<string, []>("mean1"), val = tensor<fp32, [256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(66816)))];
+            tensor<fp32, [1, 256]> x_1 = sub(x = embeddings, y = mean1)[name = tensor<string, []>("x_1")];
+            tensor<fp32, [1, 256]> var_11 = mul(x = x_1, y = x_1)[name = tensor<string, []>("op_11")];
+            tensor<int32, [1]> var_16_axes_0 = const()[name = tensor<string, []>("op_16_axes_0"), val = tensor<int32, [1]>([-1])];
+            tensor<bool, []> var_16_keep_dims_0 = const()[name = tensor<string, []>("op_16_keep_dims_0"), val = tensor<bool, []>(true)];
+            tensor<fp32, [1, 1]> var_16 = reduce_sum(axes = var_16_axes_0, keep_dims = var_16_keep_dims_0, x = var_11)[name = tensor<string, []>("op_16")];
+            tensor<fp32, []> var_17 = const()[name = tensor<string, []>("op_17"), val = tensor<fp32, []>(0x1.197998p-40)];
+            tensor<fp32, []> const_0 = const()[name = tensor<string, []>("const_0"), val = tensor<fp32, []>(0x1.fffffep+127)];
+            tensor<fp32, [1, 1]> clip_0 = clip(alpha = var_17, beta = const_0, x = var_16)[name = tensor<string, []>("clip_0")];
+            tensor<fp32, [1, 1]> norm_1 = sqrt(x = clip_0)[name = tensor<string, []>("norm_1")];
+            tensor<fp32, [1, 256]> normalized1 = real_div(x = x_1, y = norm_1)[name = tensor<string, []>("normalized1")];
+            tensor<fp32, [128, 256]> transpose_0 = const()[name = tensor<string, []>("transpose_0"), val = tensor<fp32, [128, 256]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(67904)))];
+            tensor<fp32, [128]> var_22_bias_0 = const()[name = tensor<string, []>("op_22_bias_0"), val = tensor<fp32, [128]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(199040)))];
+            tensor<fp32, [1, 128]> var_22 = linear(bias = var_22_bias_0, weight = transpose_0, x = normalized1)[name = tensor<string, []>("op_22")];
+            tensor<fp32, [1, 128]> projected = mul(x = var_22, y = lda_scale)[name = tensor<string, []>("projected")];
+            tensor<fp32, [1, 128]> x = sub(x = projected, y = mean2)[name = tensor<string, []>("x")];
+            tensor<fp32, [1, 128]> var_26 = mul(x = x, y = x)[name = tensor<string, []>("op_26")];
+            tensor<int32, [1]> var_31_axes_0 = const()[name = tensor<string, []>("op_31_axes_0"), val = tensor<int32, [1]>([-1])];
+            tensor<bool, []> var_31_keep_dims_0 = const()[name = tensor<string, []>("op_31_keep_dims_0"), val = tensor<bool, []>(true)];
+            tensor<fp32, [1, 1]> var_31 = reduce_sum(axes = var_31_axes_0, keep_dims = var_31_keep_dims_0, x = var_26)[name = tensor<string, []>("op_31")];
+            tensor<fp32, []> var_32 = const()[name = tensor<string, []>("op_32"), val = tensor<fp32, []>(0x1.197998p-40)];
+            tensor<fp32, []> const_1 = const()[name = tensor<string, []>("const_1"), val = tensor<fp32, []>(0x1.fffffep+127)];
+            tensor<fp32, [1, 1]> clip_1 = clip(alpha = var_32, beta = const_1, x = var_31)[name = tensor<string, []>("clip_1")];
+            tensor<fp32, [1, 1]> norm = sqrt(x = clip_1)[name = tensor<string, []>("norm")];
+            tensor<fp32, [1, 128]> var_36 = real_div(x = x, y = norm)[name = tensor<string, []>("op_36")];
+            tensor<fp32, [1, 128]> normalized2 = mul(x = var_36, y = lda_dim_scale)[name = tensor<string, []>("normalized2")];
+            tensor<fp32, [1, 128]> plda_centered = sub(x = normalized2, y = mu)[name = tensor<string, []>("plda_centered")];
+            tensor<fp32, [1, 128]> plda_features = linear(bias = var_22_bias_0, weight = plda_tr, x = plda_centered)[name = tensor<string, []>("op_41")];
+        } -> (plda_features);
+}

PLDA.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:566c14f27af4ef1a4bdfb8ea875adeedd7026a85e1026659f48a3d305d51de0c
+size 199616