Upload 21 files

.gitattributes

@@ -41,3 +41,6 @@ test/ambient5.mp3 filter=lfs diff=lfs merge=lfs -text
 test/human1.mp3 filter=lfs diff=lfs merge=lfs -text
 test/human3.mp3 filter=lfs diff=lfs merge=lfs -text
 test/human5.mp3 filter=lfs diff=lfs merge=lfs -text
+silero_encoder.mlmodelc/model.espresso.weights filter=lfs diff=lfs merge=lfs -text
+silero_rnn_decoder.mlmodelc/model.espresso.weights filter=lfs diff=lfs merge=lfs -text
+silero_stft.mlmodelc/model.espresso.weights filter=lfs diff=lfs merge=lfs -text

silero_encoder.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aae01b0c105877fb1dceefde19f53230c046ff7e64aa1add8d47fdadbd9c7715
+size 480

silero_encoder.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b819a1ebc472dd0ad4525930241dcea33b39d39f81e7b9193e5d10d5d0783d9
+size 329

silero_encoder.mlmodelc/model.espresso.net

@@ -0,0 +1,152 @@
+{
+  "storage" : "model.espresso.weights",
+  "analyses" : {
+
+  },
+  "properties" : {
+
+  },
+  "format_version" : 200,
+  "metadata_in_weights" : [
+
+  ],
+  "layers" : [
+    {
+      "size_of_axes" : 1,
+      "bottom" : "stft_features",
+      "axes_0" : -2,
+      "weights" : {
+
+      },
+      "nd_axis" : 0,
+      "debug_info" : "expand_dims_0",
+      "top" : "expand_dims_0",
+      "type" : "expand_dims",
+      "name" : "expand_dims_0"
+    },
+    {
+      "pad_r" : 1,
+      "fused_relu" : 0,
+      "fused_tanh" : 0,
+      "debug_info" : "conv_0",
+      "pad_fill_mode" : 0,
+      "pad_b" : 0,
+      "pad_l" : 1,
+      "top" : "conv_0",
+      "blob_weights" : 3,
+      "K" : 201,
+      "blob_biases" : 1,
+      "name" : "conv_0",
+      "has_batch_norm" : 0,
+      "type" : "convolution",
+      "n_groups" : 1,
+      "pad_t" : 0,
+      "has_biases" : 1,
+      "C" : 64,
+      "bottom" : "expand_dims_0",
+      "weights" : {
+
+      },
+      "Nx" : 3,
+      "pad_mode" : 0,
+      "pad_value" : 0,
+      "Ny" : 1,
+      "n_parallel" : 1
+    },
+    {
+      "size_of_axes" : 1,
+      "bottom" : "conv_0",
+      "axes_0" : -2,
+      "weights" : {
+
+      },
+      "nd_axis" : 0,
+      "debug_info" : "input.1",
+      "version" : 1,
+      "top" : "input.1",
+      "type" : "squeeze",
+      "name" : "input.1"
+    },
+    {
+      "bottom" : "input.1",
+      "weights" : {
+
+      },
+      "mode" : 0,
+      "debug_info" : "input.3",
+      "top" : "input.3",
+      "type" : "activation",
+      "name" : "input.3"
+    },
+    {
+      "size_of_axes" : 1,
+      "bottom" : "input.3",
+      "axes_0" : -2,
+      "weights" : {
+
+      },
+      "nd_axis" : 0,
+      "debug_info" : "expand_dims_2",
+      "top" : "expand_dims_2",
+      "type" : "expand_dims",
+      "name" : "expand_dims_2"
+    },
+    {
+      "pad_r" : 1,
+      "fused_relu" : 0,
+      "fused_tanh" : 0,
+      "debug_info" : "conv_1",
+      "pad_fill_mode" : 0,
+      "pad_b" : 0,
+      "pad_l" : 1,
+      "top" : "conv_1",
+      "blob_weights" : 7,
+      "K" : 64,
+      "blob_biases" : 5,
+      "name" : "conv_1",
+      "has_batch_norm" : 0,
+      "type" : "convolution",
+      "n_groups" : 1,
+      "pad_t" : 0,
+      "has_biases" : 1,
+      "C" : 128,
+      "bottom" : "expand_dims_2",
+      "weights" : {
+
+      },
+      "Nx" : 3,
+      "pad_mode" : 0,
+      "pad_value" : 0,
+      "Ny" : 1,
+      "n_parallel" : 1
+    },
+    {
+      "size_of_axes" : 1,
+      "bottom" : "conv_1",
+      "axes_0" : -2,
+      "weights" : {
+
+      },
+      "nd_axis" : 0,
+      "debug_info" : "input",
+      "version" : 1,
+      "top" : "input",
+      "type" : "squeeze",
+      "name" : "input"
+    },
+    {
+      "bottom" : "input",
+      "weights" : {
+
+      },
+      "mode" : 0,
+      "debug_info" : "27",
+      "top" : "var_27",
+      "type" : "activation",
+      "name" : "27",
+      "attributes" : {
+        "is_output" : 1
+      }
+    }
+  ]
+}

silero_encoder.mlmodelc/model.espresso.shape

@@ -0,0 +1,67 @@
+{
+  "layer_shapes" : {
+    "conv_1" : {
+      "k" : 128,
+      "w" : 4,
+      "n" : 1,
+      "_rank" : 4,
+      "h" : 1
+    },
+    "conv_0" : {
+      "k" : 64,
+      "w" : 4,
+      "n" : 1,
+      "_rank" : 4,
+      "h" : 1
+    },
+    "stft_features" : {
+      "k" : 1,
+      "w" : 4,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 201
+    },
+    "expand_dims_0" : {
+      "k" : 201,
+      "w" : 4,
+      "n" : 1,
+      "_rank" : 4,
+      "h" : 1
+    },
+    "input" : {
+      "k" : 1,
+      "w" : 4,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 128
+    },
+    "input.3" : {
+      "k" : 1,
+      "w" : 4,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 64
+    },
+    "var_27" : {
+      "k" : 1,
+      "w" : 4,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 128
+    },
+    "input.1" : {
+      "k" : 1,
+      "w" : 4,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 64
+    },
+    "expand_dims_2" : {
+      "k" : 64,
+      "w" : 4,
+      "n" : 1,
+      "_rank" : 4,
+      "h" : 1
+    }
+  }
+}

silero_encoder.mlmodelc/model.espresso.weights

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9006e7a1398b3811c21355bc1e6a92eae81aceaece8b86291cfc5cceb0bff0cd
+size 253632

silero_encoder.mlmodelc/model/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5a5fd42d16a20302798ef6ed309979b43003d2320d9f0e8ea9831a92759fb4b
+size 64

silero_encoder.mlmodelc/neural_network_optionals/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5931bd536c4550294a212c6fc8d3c1bee75da817980ba0b671c8603dafb36238
+size 40

silero_rnn_decoder.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4202f14c95389e9186aca8aae853bba786204d21e87810d02e7388448ad652ec
+size 485

silero_rnn_decoder.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29a1e42eb1180e9d008197e11841396e63b3f2774a5c8aab33a5598cace9b86d
+size 376

silero_rnn_decoder.mlmodelc/model.espresso.net

@@ -0,0 +1,247 @@
+{
+  "storage" : "model.espresso.weights",
+  "analyses" : {
+
+  },
+  "properties" : {
+
+  },
+  "format_version" : 200,
+  "metadata_in_weights" : [
+
+  ],
+  "layers" : [
+    {
+      "size_of_axes" : 1,
+      "bottom" : "h_in",
+      "axes_0" : 0,
+      "weights" : {
+
+      },
+      "nd_axis" : 0,
+      "debug_info" : "15_batch_first_lstm_h0_squeeze",
+      "version" : 1,
+      "top" : "15_batch_first_lstm_h0_squeeze",
+      "type" : "squeeze",
+      "name" : "15_batch_first_lstm_h0_squeeze"
+    },
+    {
+      "size_of_axes" : 1,
+      "bottom" : "c_in",
+      "axes_0" : 0,
+      "weights" : {
+
+      },
+      "nd_axis" : 0,
+      "debug_info" : "15_batch_first_lstm_c0_squeeze",
+      "version" : 1,
+      "top" : "15_batch_first_lstm_c0_squeeze",
+      "type" : "squeeze",
+      "name" : "15_batch_first_lstm_c0_squeeze"
+    },
+    {
+      "axis_h" : 2,
+      "axis_w" : 0,
+      "bottom" : "encoder_features",
+      "axis_k" : 1,
+      "axis_n" : 3,
+      "axis_seq" : 4,
+      "weights" : {
+
+      },
+      "debug_info" : "transpose_1",
+      "top" : "transpose_1",
+      "type" : "transpose",
+      "name" : "transpose_1"
+    },
+    {
+      "size_of_axes" : 2,
+      "bottom" : "transpose_1",
+      "axes_0" : 3,
+      "axes_1" : 4,
+      "weights" : {
+
+      },
+      "nd_axis" : 0,
+      "debug_info" : "transpose_1_expanded",
+      "top" : "transpose_1_expanded",
+      "type" : "expand_dims",
+      "name" : "transpose_1_expanded"
+    },
+    {
+      "size_of_axes" : 3,
+      "bottom" : "15_batch_first_lstm_h0_squeeze",
+      "axes_0" : 0,
+      "axes_1" : 3,
+      "axes_2" : 4,
+      "nd_axis" : 0,
+      "weights" : {
+
+      },
+      "debug_info" : "15_batch_first_lstm_h0_squeeze_expanded",
+      "top" : "15_batch_first_lstm_h0_squeeze_expanded",
+      "type" : "expand_dims",
+      "name" : "15_batch_first_lstm_h0_squeeze_expanded"
+    },
+    {
+      "size_of_axes" : 3,
+      "bottom" : "15_batch_first_lstm_c0_squeeze",
+      "axes_0" : 0,
+      "axes_1" : 3,
+      "axes_2" : 4,
+      "nd_axis" : 0,
+      "weights" : {
+
+      },
+      "debug_info" : "15_batch_first_lstm_c0_squeeze_expanded2",
+      "top" : "15_batch_first_lstm_c0_squeeze_expanded2",
+      "type" : "expand_dims",
+      "name" : "15_batch_first_lstm_c0_squeeze_expanded2"
+    },
+    {
+      "input_size" : 128,
+      "arch" : 1,
+      "quantization_scale_h" : 1,
+      "quantization_scale_x" : 1,
+      "hidden_size" : 128,
+      "nonlinearity_alpha" : 1,
+      "nonlinearity_bidir_type" : 0,
+      "lstm_forget_bias" : 0,
+      "coupled_input_forget" : 0,
+      "quantization_simple_mode" : false,
+      "bias_vector_flag" : 1,
+      "mode" : 0,
+      "cell_clip_threshold" : 500,
+      "type" : "rnn_arch",
+      "internal_nonlinearity_second_beta" : 1,
+      "internal_nonlinearity_alpha" : 1,
+      "internal_nonlinearity_beta" : 1,
+      "forget_bias" : 0,
+      "weights" : {
+        "W_h" : 1,
+        "W_x" : 3,
+        "b" : 5
+      },
+      "lstm_has_peepholes" : 0,
+      "beta" : 1,
+      "quantization_mode" : 0,
+      "name" : "15_batch_first",
+      "quantization_scale_h_reverse" : 0,
+      "top" : "15_batch_first_0_5d,15_batch_first_1_5d,15_batch_first_2_5d",
+      "input_bias" : false,
+      "alpha" : 1,
+      "quantization_scale_x_reverse" : 0,
+      "max_batch_size" : 1,
+      "reset_after" : false,
+      "debug_info" : "15_batch_first",
+      "hidden_bias" : false,
+      "nonlinearity_bidir_alpha" : 1,
+      "nonlinearity_beta" : 1,
+      "internal_nonlinearity_second_alpha" : 1,
+      "encoder_bidirectional_mode" : 0,
+      "internal_nonlinearity_type" : 2,
+      "bottom" : "transpose_1_expanded,15_batch_first_lstm_h0_squeeze_expanded,15_batch_first_lstm_c0_squeeze_expanded2",
+      "nonlinearity_type" : 0,
+      "nonlinearity_bidir_beta" : 1,
+      "encoder_output_sequence" : 1,
+      "internal_nonlinearity_second_type" : 0
+    },
+    {
+      "size_of_axes" : 2,
+      "bottom" : "15_batch_first_0_5d",
+      "axes_0" : 3,
+      "axes_1" : 4,
+      "weights" : {
+
+      },
+      "nd_axis" : 0,
+      "debug_info" : "15_batch_first_0",
+      "version" : 1,
+      "top" : "15_batch_first_0",
+      "type" : "squeeze",
+      "name" : "15_batch_first_0"
+    },
+    {
+      "name" : "15_batch_first_1",
+      "version" : 1,
+      "axes_2" : 4,
+      "axes_1" : 3,
+      "type" : "squeeze",
+      "axes_0" : 0,
+      "nd_axis" : 0,
+      "bottom" : "15_batch_first_1_5d",
+      "debug_info" : "15_batch_first_1",
+      "size_of_axes" : 3,
+      "weights" : {
+
+      },
+      "top" : "15_batch_first_1"
+    },
+    {
+      "name" : "15_batch_first_2",
+      "version" : 1,
+      "axes_2" : 4,
+      "axes_1" : 3,
+      "type" : "squeeze",
+      "axes_0" : 0,
+      "nd_axis" : 0,
+      "bottom" : "15_batch_first_2_5d",
+      "debug_info" : "15_batch_first_2",
+      "size_of_axes" : 3,
+      "weights" : {
+
+      },
+      "top" : "15_batch_first_2"
+    },
+    {
+      "bottom" : "15_batch_first_1",
+      "size_of_axes" : 1,
+      "axes_0" : 0,
+      "weights" : {
+
+      },
+      "nd_axis" : 0,
+      "debug_info" : "16",
+      "top" : "h_out",
+      "type" : "expand_dims",
+      "name" : "16",
+      "attributes" : {
+        "is_output" : 1
+      }
+    },
+    {
+      "bottom" : "15_batch_first_2",
+      "size_of_axes" : 1,
+      "axes_0" : 0,
+      "weights" : {
+
+      },
+      "nd_axis" : 0,
+      "debug_info" : "17",
+      "top" : "c_out",
+      "type" : "expand_dims",
+      "name" : "17",
+      "attributes" : {
+        "is_output" : 1
+      }
+    },
+    {
+      "axis_seq" : 4,
+      "name" : "transpose_0",
+      "axis_n" : 3,
+      "axis_h" : 2,
+      "type" : "transpose",
+      "attributes" : {
+        "is_output" : 1
+      },
+      "bottom" : "15_batch_first_0",
+      "axis_w" : 0,
+      "axis_k" : 1,
+      "debug_info" : "transpose_0",
+      "weights" : {
+
+      },
+      "top" : "rnn_out"
+    }
+  ]
+}

silero_rnn_decoder.mlmodelc/model.espresso.shape

@@ -0,0 +1,132 @@
+{
+  "layer_shapes" : {
+    "15_batch_first_0" : {
+      "k" : 4,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "transpose_1" : {
+      "k" : 4,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "15_batch_first_lstm_h0_squeeze" : {
+      "k" : 1,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 2,
+      "h" : 1
+    },
+    "h_out" : {
+      "k" : 1,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "h_in" : {
+      "k" : 1,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "transpose_1_expanded" : {
+      "k" : 128,
+      "w" : 1,
+      "n" : 1,
+      "seq" : 4,
+      "_rank" : 5,
+      "h" : 1
+    },
+    "rnn_out" : {
+      "k" : 1,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 4
+    },
+    "15_batch_first_1" : {
+      "k" : 1,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 2,
+      "h" : 1
+    },
+    "15_batch_first_lstm_h0_squeeze_expanded" : {
+      "k" : 128,
+      "w" : 1,
+      "n" : 1,
+      "_rank" : 5,
+      "h" : 1
+    },
+    "c_out" : {
+      "k" : 1,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "15_batch_first_0_5d" : {
+      "k" : 128,
+      "w" : 1,
+      "n" : 1,
+      "seq" : 4,
+      "_rank" : 5,
+      "h" : 1
+    },
+    "15_batch_first_1_5d" : {
+      "k" : 128,
+      "w" : 1,
+      "n" : 1,
+      "_rank" : 5,
+      "h" : 1
+    },
+    "c_in" : {
+      "k" : 1,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "15_batch_first_lstm_c0_squeeze" : {
+      "k" : 1,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 2,
+      "h" : 1
+    },
+    "15_batch_first_2_5d" : {
+      "k" : 128,
+      "w" : 1,
+      "n" : 1,
+      "_rank" : 5,
+      "h" : 1
+    },
+    "encoder_features" : {
+      "k" : 1,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 4
+    },
+    "15_batch_first_2" : {
+      "k" : 1,
+      "w" : 128,
+      "n" : 1,
+      "_rank" : 2,
+      "h" : 1
+    },
+    "15_batch_first_lstm_c0_squeeze_expanded2" : {
+      "k" : 128,
+      "w" : 1,
+      "n" : 1,
+      "_rank" : 5,
+      "h" : 1
+    }
+  }
+}

silero_rnn_decoder.mlmodelc/model.espresso.weights

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fab772578fc4193827ee214c2012e39570072aec06753f02778c4256f20eaa8b
+size 526464

silero_rnn_decoder.mlmodelc/model/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5a5fd42d16a20302798ef6ed309979b43003d2320d9f0e8ea9831a92759fb4b
+size 64

silero_rnn_decoder.mlmodelc/neural_network_optionals/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5931bd536c4550294a212c6fc8d3c1bee75da817980ba0b671c8603dafb36238
+size 40

silero_stft.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:697661bb3385166304dcaf62d7d03652d316211768f2c91cbb567a7532d42936
+size 478

silero_stft.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92f8611827e189effbe6249666325694e96f1899bddd55f0881d8482b8c623b9
+size 326

silero_stft.mlmodelc/model.espresso.net

@@ -0,0 +1,253 @@
+{
+  "storage" : "model.espresso.weights",
+  "analyses" : {
+
+  },
+  "properties" : {
+
+  },
+  "format_version" : 200,
+  "metadata_in_weights" : [
+
+  ],
+  "layers" : [
+    {
+      "bottom" : "audio_input",
+      "end" : 400,
+      "start" : 0,
+      "weights" : {
+
+      },
+      "axis" : 0,
+      "debug_info" : "37",
+      "top" : "37",
+      "type" : "slice",
+      "name" : "37"
+    },
+    {
+      "name" : "reshape_2",
+      "weights" : {
+
+      },
+      "dst_w" : -1,
+      "version" : 1,
+      "dst_n" : 1,
+      "dst_nd_rank" : 1,
+      "type" : "reshape",
+      "dst_h" : 1,
+      "mode" : 0,
+      "dynamic_shape" : false,
+      "bottom" : "37",
+      "debug_info" : "reshape_2",
+      "dst_seq" : 1,
+      "dst_k" : 1,
+      "top" : "reshape_2"
+    },
+    {
+      "top" : "reshape_3",
+      "w" : 400,
+      "h" : 1,
+      "name" : "reshape_3",
+      "nd_rank" : 1,
+      "type" : "load_constant",
+      "k" : 1,
+      "bottom" : "",
+      "debug_info" : "reshape_3",
+      "n" : 1,
+      "weights" : {
+
+      },
+      "constant_blob" : 1
+    },
+    {
+      "top" : "range_1d_0",
+      "w" : 400,
+      "h" : 1,
+      "name" : "range_1d_0",
+      "nd_rank" : 1,
+      "type" : "load_constant",
+      "k" : 1,
+      "bottom" : "",
+      "debug_info" : "range_1d_0",
+      "n" : 1,
+      "weights" : {
+
+      },
+      "constant_blob" : 3
+    },
+    {
+      "bottom" : "reshape_3,range_1d_0,reshape_2",
+      "validate_indices" : true,
+      "allow_negative_indices" : true,
+      "mode" : 0,
+      "weights" : {
+
+      },
+      "debug_info" : "scatter_0",
+      "axis" : 0,
+      "top" : "scatter_0",
+      "type" : "scatter_nd",
+      "name" : "scatter_0",
+      "accumulation_mode" : 0
+    },
+    {
+      "name" : "reshape_4",
+      "weights" : {
+
+      },
+      "dst_w" : 400,
+      "version" : 1,
+      "dst_n" : 1,
+      "dst_nd_rank" : 3,
+      "type" : "reshape",
+      "dst_h" : 1,
+      "mode" : 0,
+      "dynamic_shape" : false,
+      "bottom" : "scatter_0",
+      "debug_info" : "reshape_4",
+      "dst_seq" : 1,
+      "dst_k" : 1,
+      "top" : "reshape_4"
+    },
+    {
+      "top" : "56",
+      "w" : 400,
+      "h" : 1,
+      "name" : "56",
+      "nd_rank" : 3,
+      "type" : "load_constant",
+      "k" : 1,
+      "bottom" : "",
+      "debug_info" : "56",
+      "n" : 1,
+      "weights" : {
+
+      },
+      "constant_blob" : 5
+    },
+    {
+      "bottom" : "reshape_4,56",
+      "alpha" : 1,
+      "operation" : 1,
+      "weights" : {
+
+      },
+      "fused_relu" : 0,
+      "debug_info" : "windowed_frames",
+      "top" : "windowed_frames",
+      "type" : "elementwise",
+      "name" : "windowed_frames",
+      "beta" : 0
+    },
+    {
+      "nB" : 400,
+      "top" : "59",
+      "has_biases" : 1,
+      "weights" : {
+
+      },
+      "nC" : 201,
+      "blob_weights" : 9,
+      "type" : "inner_product",
+      "has_relu" : 0,
+      "bottom" : "windowed_frames",
+      "blob_biases" : 7,
+      "has_tanh" : 0,
+      "debug_info" : "59",
+      "name" : "59",
+      "has_prelu" : 0
+    },
+    {
+      "nB" : 400,
+      "top" : "61",
+      "has_biases" : 1,
+      "weights" : {
+
+      },
+      "nC" : 201,
+      "blob_weights" : 13,
+      "type" : "inner_product",
+      "has_relu" : 0,
+      "bottom" : "windowed_frames",
+      "blob_biases" : 11,
+      "has_tanh" : 0,
+      "debug_info" : "61",
+      "name" : "61",
+      "has_prelu" : 0
+    },
+    {
+      "bottom" : "59",
+      "alpha" : 2,
+      "operation" : 13,
+      "weights" : {
+
+      },
+      "fused_relu" : 0,
+      "debug_info" : "63",
+      "top" : "63",
+      "type" : "elementwise",
+      "name" : "63",
+      "beta" : 0
+    },
+    {
+      "bottom" : "61",
+      "alpha" : 2,
+      "operation" : 13,
+      "weights" : {
+
+      },
+      "fused_relu" : 0,
+      "debug_info" : "65",
+      "top" : "65",
+      "type" : "elementwise",
+      "name" : "65",
+      "beta" : 0
+    },
+    {
+      "bottom" : "63,65",
+      "alpha" : 1,
+      "operation" : 0,
+      "weights" : {
+
+      },
+      "fused_relu" : 0,
+      "debug_info" : "67",
+      "top" : "67",
+      "type" : "elementwise",
+      "name" : "67",
+      "beta" : 0
+    },
+    {
+      "bottom" : "67",
+      "alpha" : 1,
+      "operation" : 11,
+      "weights" : {
+
+      },
+      "fused_relu" : 0,
+      "debug_info" : "magnitude",
+      "top" : "magnitude",
+      "type" : "elementwise",
+      "name" : "magnitude",
+      "beta" : 0
+    },
+    {
+      "axis_seq" : 4,
+      "name" : "transpose_2",
+      "axis_n" : 3,
+      "axis_h" : 0,
+      "type" : "transpose",
+      "attributes" : {
+        "is_output" : 1
+      },
+      "bottom" : "magnitude",
+      "axis_w" : 1,
+      "axis_k" : 2,
+      "debug_info" : "transpose_2",
+      "weights" : {
+
+      },
+      "top" : "var_71"
+    }
+  ]
+}

silero_stft.mlmodelc/model.espresso.shape

@@ -0,0 +1,116 @@
+{
+  "layer_shapes" : {
+    "63" : {
+      "k" : 1,
+      "w" : 201,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "reshape_3" : {
+      "k" : 1,
+      "w" : 400,
+      "n" : 1,
+      "_rank" : 1,
+      "h" : 1
+    },
+    "windowed_frames" : {
+      "k" : 1,
+      "w" : 400,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "37" : {
+      "k" : 1,
+      "w" : 400,
+      "n" : 1,
+      "_rank" : 2,
+      "h" : 1
+    },
+    "reshape_4" : {
+      "k" : 1,
+      "w" : 400,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "var_71" : {
+      "k" : 1,
+      "w" : 1,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 201
+    },
+    "audio_input" : {
+      "k" : 1,
+      "w" : 512,
+      "n" : 1,
+      "_rank" : 2,
+      "h" : 1
+    },
+    "59" : {
+      "k" : 1,
+      "w" : 201,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "65" : {
+      "k" : 1,
+      "w" : 201,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "61" : {
+      "k" : 1,
+      "w" : 201,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "magnitude" : {
+      "k" : 1,
+      "w" : 201,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "reshape_2" : {
+      "k" : 1,
+      "w" : 400,
+      "n" : 1,
+      "_rank" : 1,
+      "h" : 1
+    },
+    "56" : {
+      "k" : 1,
+      "w" : 400,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "67" : {
+      "k" : 1,
+      "w" : 201,
+      "n" : 1,
+      "_rank" : 3,
+      "h" : 1
+    },
+    "scatter_0" : {
+      "k" : 1,
+      "w" : 400,
+      "n" : 1,
+      "_rank" : 1,
+      "h" : 1
+    },
+    "range_1d_0" : {
+      "k" : 1,
+      "w" : 400,
+      "n" : 1,
+      "_rank" : 1,
+      "h" : 1
+    }
+  }
+}

silero_stft.mlmodelc/model.espresso.weights

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61e94a7b7ab94eef111adf1a03189de47e6f43dd3fffdffccdd253b0f6094f7b
+size 654784

silero_stft.mlmodelc/model/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5a5fd42d16a20302798ef6ed309979b43003d2320d9f0e8ea9831a92759fb4b
+size 64

silero_stft.mlmodelc/neural_network_optionals/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5931bd536c4550294a212c6fc8d3c1bee75da817980ba0b671c8603dafb36238
+size 40