program(1.3)
[buildInfo = dict<string, string>({{"coremlc-component-MIL", "3600.16.1"}, {"coremlc-version", "3600.22.1"}, {"coremltools-component-torch", "2.7.0"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "9.0b1"}})]
{
    func main<ios18>(tensor<int32, [1]> audio_length, tensor<fp32, [1, 240000]> audio_signal) {
            int32 var_9 = const()[name = string("op_9"), val = int32(1)];
            int32 var_10 = const()[name = string("op_10"), val = int32(160)];
            fp32 var_24 = const()[name = string("op_24"), val = fp32(0x0p+0)];
            fp32 var_25 = const()[name = string("op_25"), val = fp32(0x1.4f8b58p-17)];
            int32 var_34 = const()[name = string("op_34"), val = int32(512)];
            tensor<int32, [1]> var_35 = add(x = audio_length, y = var_34)[name = string("op_35")];
            int32 var_36 = const()[name = string("op_36"), val = int32(512)];
            tensor<int32, [1]> var_37 = sub(x = var_35, y = var_36)[name = string("op_37")];
            tensor<int32, [1]> floor_div_0 = floor_div(x = var_37, y = var_10)[name = string("floor_div_0")];
            string var_38_dtype_0 = const()[name = string("op_38_dtype_0"), val = string("fp32")];
            fp32 var_39_promoted = const()[name = string("op_39_promoted"), val = fp32(0x1p+0)];
            tensor<fp32, [1]> var_38 = cast(dtype = var_38_dtype_0, x = floor_div_0)[name = string("cast_12")];
            tensor<fp32, [1]> seq_len_1 = add(x = var_38, y = var_39_promoted)[name = string("seq_len_1")];
            string seq_len_dtype_0 = const()[name = string("seq_len_dtype_0"), val = string("int32")];
            tensor<int32, [2]> var_43_begin_0 = const()[name = string("op_43_begin_0"), val = tensor<int32, [2]>([0, 0])];
            tensor<int32, [2]> var_43_end_0 = const()[name = string("op_43_end_0"), val = tensor<int32, [2]>([1, 1])];
            tensor<bool, [2]> var_43_end_mask_0 = const()[name = string("op_43_end_mask_0"), val = tensor<bool, [2]>([true, false])];
            tensor<bool, [2]> var_43_squeeze_mask_0 = const()[name = string("op_43_squeeze_mask_0"), val = tensor<bool, [2]>([false, true])];
            tensor<fp32, [1]> var_43 = slice_by_index(begin = var_43_begin_0, end = var_43_end_0, end_mask = var_43_end_mask_0, squeeze_mask = var_43_squeeze_mask_0, x = audio_signal)[name = string("op_43")];
            tensor<int32, [1]> var_44_axes_0 = const()[name = string("op_44_axes_0"), val = tensor<int32, [1]>([1])];
            tensor<fp32, [1, 1]> var_44 = expand_dims(axes = var_44_axes_0, x = var_43)[name = string("op_44")];
            tensor<int32, [2]> var_46_begin_0 = const()[name = string("op_46_begin_0"), val = tensor<int32, [2]>([0, 1])];
            tensor<int32, [2]> var_46_end_0 = const()[name = string("op_46_end_0"), val = tensor<int32, [2]>([1, 240000])];
            tensor<bool, [2]> var_46_end_mask_0 = const()[name = string("op_46_end_mask_0"), val = tensor<bool, [2]>([true, true])];
            tensor<fp32, [1, 239999]> var_46 = slice_by_index(begin = var_46_begin_0, end = var_46_end_0, end_mask = var_46_end_mask_0, x = audio_signal)[name = string("op_46")];
            tensor<int32, [2]> var_48_begin_0 = const()[name = string("op_48_begin_0"), val = tensor<int32, [2]>([0, 0])];
            tensor<int32, [2]> var_48_end_0 = const()[name = string("op_48_end_0"), val = tensor<int32, [2]>([1, 239999])];
            tensor<bool, [2]> var_48_end_mask_0 = const()[name = string("op_48_end_mask_0"), val = tensor<bool, [2]>([true, false])];
            tensor<fp32, [1, 239999]> var_48 = slice_by_index(begin = var_48_begin_0, end = var_48_end_0, end_mask = var_48_end_mask_0, x = audio_signal)[name = string("op_48")];
            fp32 var_49 = const()[name = string("op_49"), val = fp32(0x1.f0a3d8p-1)];
            tensor<fp32, [1, 239999]> var_50 = mul(x = var_48, y = var_49)[name = string("op_50")];
            tensor<fp32, [1, 239999]> var_51 = sub(x = var_46, y = var_50)[name = string("op_51")];
            bool input_1_interleave_0 = const()[name = string("input_1_interleave_0"), val = bool(false)];
            tensor<fp32, [1, 240000]> input_1 = concat(axis = var_9, interleave = input_1_interleave_0, values = (var_44, var_51))[name = string("input_1")];
            tensor<int32, [3]> var_57 = const()[name = string("op_57"), val = tensor<int32, [3]>([1, 1, 240000])];
            tensor<fp32, [1, 1, 240000]> input_3 = reshape(shape = var_57, x = input_1)[name = string("input_3")];
            fp32 const_3 = const()[name = string("const_3"), val = fp32(0x0p+0)];
            tensor<int32, [6]> input_5_pad_0 = const()[name = string("input_5_pad_0"), val = tensor<int32, [6]>([0, 0, 0, 0, 256, 256])];
            string input_5_mode_0 = const()[name = string("input_5_mode_0"), val = string("reflect")];
            tensor<fp32, [1, 1, 240512]> input_5 = pad(constant_val = const_3, mode = input_5_mode_0, pad = input_5_pad_0, x = input_3)[name = string("input_5")];
            tensor<int32, [2]> var_63 = const()[name = string("op_63"), val = tensor<int32, [2]>([1, 240512])];
            tensor<fp32, [1, 240512]> input = reshape(shape = var_63, x = input_5)[name = string("input")];
            tensor<fp32, [257, 1, 512]> expand_dims_3 = const()[name = string("expand_dims_3"), val = tensor<fp32, [257, 1, 512]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(64)))];
            tensor<fp32, [257, 1, 512]> expand_dims_4 = const()[name = string("expand_dims_4"), val = tensor<fp32, [257, 1, 512]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(526464)))];
            tensor<int32, [1]> expand_dims_5 = const()[name = string("expand_dims_5"), val = tensor<int32, [1]>([160])];
            tensor<int32, [1]> expand_dims_6_axes_0 = const()[name = string("expand_dims_6_axes_0"), val = tensor<int32, [1]>([1])];
            tensor<fp32, [1, 1, 240512]> expand_dims_6 = expand_dims(axes = expand_dims_6_axes_0, x = input)[name = string("expand_dims_6")];
            string conv_0_pad_type_0 = const()[name = string("conv_0_pad_type_0"), val = string("valid")];
            tensor<int32, [2]> conv_0_pad_0 = const()[name = string("conv_0_pad_0"), val = tensor<int32, [2]>([0, 0])];
            tensor<int32, [1]> conv_0_dilations_0 = const()[name = string("conv_0_dilations_0"), val = tensor<int32, [1]>([1])];
            int32 conv_0_groups_0 = const()[name = string("conv_0_groups_0"), val = int32(1)];
            tensor<fp32, [1, 257, 1501]> conv_0 = 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_5, weight = expand_dims_3, x = expand_dims_6)[name = string("conv_0")];
            string conv_1_pad_type_0 = const()[name = string("conv_1_pad_type_0"), val = string("valid")];
            tensor<int32, [2]> conv_1_pad_0 = const()[name = string("conv_1_pad_0"), val = tensor<int32, [2]>([0, 0])];
            tensor<int32, [1]> conv_1_dilations_0 = const()[name = string("conv_1_dilations_0"), val = tensor<int32, [1]>([1])];
            int32 conv_1_groups_0 = const()[name = string("conv_1_groups_0"), val = int32(1)];
            tensor<fp32, [1, 257, 1501]> conv_1 = 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_5, weight = expand_dims_4, x = expand_dims_6)[name = string("conv_1")];
            int32 stack_0_axis_0 = const()[name = string("stack_0_axis_0"), val = int32(-1)];
            tensor<fp32, [1, 257, 1501, 2]> stack_0 = stack(axis = stack_0_axis_0, values = (conv_0, conv_1))[name = string("stack_0")];
            fp32 var_17_promoted = const()[name = string("op_17_promoted"), val = fp32(0x1p+1)];
            tensor<fp32, [1, 257, 1501, 2]> var_67 = pow(x = stack_0, y = var_17_promoted)[name = string("op_67")];
            tensor<int32, [1]> var_69_axes_0 = const()[name = string("op_69_axes_0"), val = tensor<int32, [1]>([-1])];
            bool var_69_keep_dims_0 = const()[name = string("op_69_keep_dims_0"), val = bool(false)];
            tensor<fp32, [1, 257, 1501]> var_69 = reduce_sum(axes = var_69_axes_0, keep_dims = var_69_keep_dims_0, x = var_67)[name = string("op_69")];
            tensor<fp32, [1, 257, 1501]> x_9 = identity(x = var_69)[name = string("x_9")];
            tensor<fp32, [1, 128, 257]> const_6 = const()[name = string("const_6"), val = tensor<fp32, [1, 128, 257]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(1052864)))];
            bool x_11_transpose_x_0 = const()[name = string("x_11_transpose_x_0"), val = bool(false)];
            bool x_11_transpose_y_0 = const()[name = string("x_11_transpose_y_0"), val = bool(false)];
            tensor<fp32, [1, 128, 1501]> x_11 = matmul(transpose_x = x_11_transpose_x_0, transpose_y = x_11_transpose_y_0, x = const_6, y = x_9)[name = string("x_11")];
            fp32 var_76 = const()[name = string("op_76"), val = fp32(0x1p-24)];
            tensor<fp32, [1, 128, 1501]> var_77 = add(x = x_11, y = var_76)[name = string("op_77")];
            fp32 x_13_epsilon_0 = const()[name = string("x_13_epsilon_0"), val = fp32(0x1p-149)];
            tensor<fp32, [1, 128, 1501]> x_13 = log(epsilon = x_13_epsilon_0, x = var_77)[name = string("x_13")];
            tensor<int32, [1, 1501]> var_82 = const()[name = string("op_82"), val = tensor<int32, [1, 1501]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(1184512)))];
            tensor<int32, [1]> var_85_axes_0 = const()[name = string("op_85_axes_0"), val = tensor<int32, [1]>([1])];
            tensor<int32, [1]> mel_length = cast(dtype = seq_len_dtype_0, x = seq_len_1)[name = string("cast_11")];
            tensor<int32, [1, 1]> var_85 = expand_dims(axes = var_85_axes_0, x = mel_length)[name = string("op_85")];
            tensor<bool, [1, 1501]> valid_mask = less(x = var_82, y = var_85)[name = string("valid_mask")];
            tensor<int32, [1]> var_87_axes_0 = const()[name = string("op_87_axes_0"), val = tensor<int32, [1]>([1])];
            tensor<bool, [1, 1, 1501]> var_87 = expand_dims(axes = var_87_axes_0, x = valid_mask)[name = string("op_87")];
            tensor<int32, [3]> var_87_after_broadcast_reps_0 = const()[name = string("op_87_after_broadcast_reps_0"), val = tensor<int32, [3]>([1, 128, 1])];
            tensor<bool, [1, 128, 1501]> var_87_after_broadcast = tile(reps = var_87_after_broadcast_reps_0, x = var_87)[name = string("op_87_after_broadcast")];
            tensor<fp32, [1, 128, 1501]> var_24_after_broadcast = const()[name = string("op_24_after_broadcast"), val = tensor<fp32, [1, 128, 1501]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(1190592)))];
            tensor<fp32, [1, 128, 1501]> var_88 = select(a = x_13, b = var_24_after_broadcast, cond = var_87_after_broadcast)[name = string("op_88")];
            tensor<int32, [1]> x_mean_numerator_axes_0 = const()[name = string("x_mean_numerator_axes_0"), val = tensor<int32, [1]>([2])];
            bool x_mean_numerator_keep_dims_0 = const()[name = string("x_mean_numerator_keep_dims_0"), val = bool(false)];
            tensor<fp32, [1, 128]> x_mean_numerator = reduce_sum(axes = x_mean_numerator_axes_0, keep_dims = x_mean_numerator_keep_dims_0, x = var_88)[name = string("x_mean_numerator")];
            string cast_2_dtype_0 = const()[name = string("cast_2_dtype_0"), val = string("fp32")];
            tensor<int32, [1]> x_mean_denominator_axes_0 = const()[name = string("x_mean_denominator_axes_0"), val = tensor<int32, [1]>([1])];
            bool x_mean_denominator_keep_dims_0 = const()[name = string("x_mean_denominator_keep_dims_0"), val = bool(false)];
            tensor<fp32, [1, 1501]> cast_2 = cast(dtype = cast_2_dtype_0, x = valid_mask)[name = string("cast_10")];
            tensor<fp32, [1]> x_mean_denominator = reduce_sum(axes = x_mean_denominator_axes_0, keep_dims = x_mean_denominator_keep_dims_0, x = cast_2)[name = string("x_mean_denominator")];
            tensor<int32, [1]> var_93_axes_0 = const()[name = string("op_93_axes_0"), val = tensor<int32, [1]>([1])];
            tensor<fp32, [1, 1]> var_93 = expand_dims(axes = var_93_axes_0, x = x_mean_denominator)[name = string("op_93")];
            tensor<fp32, [1, 128]> x_mean = real_div(x = x_mean_numerator, y = var_93)[name = string("x_mean")];
            tensor<int32, [1]> var_96_axes_0 = const()[name = string("op_96_axes_0"), val = tensor<int32, [1]>([2])];
            tensor<fp32, [1, 128, 1]> var_96 = expand_dims(axes = var_96_axes_0, x = x_mean)[name = string("op_96")];
            tensor<fp32, [1, 128, 1501]> var_97 = sub(x = x_13, y = var_96)[name = string("op_97")];
            tensor<fp32, [1, 128, 1501]> var_98 = select(a = var_97, b = var_24_after_broadcast, cond = var_87_after_broadcast)[name = string("op_98")];
            fp32 var_17_promoted_1 = const()[name = string("op_17_promoted_1"), val = fp32(0x1p+1)];
            tensor<fp32, [1, 128, 1501]> var_99 = pow(x = var_98, y = var_17_promoted_1)[name = string("op_99")];
            tensor<int32, [1]> var_101_axes_0 = const()[name = string("op_101_axes_0"), val = tensor<int32, [1]>([2])];
            bool var_101_keep_dims_0 = const()[name = string("op_101_keep_dims_0"), val = bool(false)];
            tensor<fp32, [1, 128]> var_101 = reduce_sum(axes = var_101_axes_0, keep_dims = var_101_keep_dims_0, x = var_99)[name = string("op_101")];
            fp32 var_103 = const()[name = string("op_103"), val = fp32(0x1p+0)];
            tensor<fp32, [1, 1]> var_104 = sub(x = var_93, y = var_103)[name = string("op_104")];
            tensor<fp32, [1, 128]> var_105 = real_div(x = var_101, y = var_104)[name = string("op_105")];
            tensor<fp32, [1, 128]> x_std_1 = sqrt(x = var_105)[name = string("x_std_1")];
            tensor<fp32, [1, 128]> x_std = add(x = x_std_1, y = var_25)[name = string("x_std")];
            tensor<int32, [1]> var_110_axes_0 = const()[name = string("op_110_axes_0"), val = tensor<int32, [1]>([2])];
            tensor<fp32, [1, 128, 1]> var_110 = expand_dims(axes = var_110_axes_0, x = x_std)[name = string("op_110")];
            tensor<fp32, [1, 128, 1501]> x = real_div(x = var_97, y = var_110)[name = string("x")];
            tensor<bool, [1, 1501]> mask = greater_equal(x = var_82, y = var_85)[name = string("mask")];
            tensor<int32, [1]> var_119_axes_0 = const()[name = string("op_119_axes_0"), val = tensor<int32, [1]>([1])];
            tensor<bool, [1, 1, 1501]> var_119 = expand_dims(axes = var_119_axes_0, x = mask)[name = string("op_119")];
            tensor<fp32, [1, 128, 1501]> mel = select(a = var_24, b = x, cond = var_119)[name = string("processed_signal")];
        } -> (mel, mel_length);
}