File size: 6,633 Bytes
ad8b997 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 | program(1.3)
[buildInfo = dict<string, string>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1.2"}, {"coremltools-component-torch", "2.14.1"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "9.0"}})]
{
func main<ios18>(tensor<fp32, [1, 1, 1, 50]> mask1, tensor<fp32, [1, 1, 1, 25]> mask2, tensor<fp32, [1, 128, 100]> mel) {
tensor<fp32, [13, 1024]> m_pos = const()[name = string("m_pos"), val = tensor<fp32, [13, 1024]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(64)))];
tensor<fp32, [480]> m_conv1_bias = const()[name = string("m_conv1_bias"), val = tensor<fp32, [480]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(53376)))];
tensor<fp32, [480, 1, 3, 3]> m_conv1_weight = const()[name = string("m_conv1_weight"), val = tensor<fp32, [480, 1, 3, 3]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(55360)))];
tensor<fp32, [480]> m_conv2_bias = const()[name = string("m_conv2_bias"), val = tensor<fp32, [480]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72704)))];
tensor<fp32, [480, 480, 3, 3]> m_conv2_weight = const()[name = string("m_conv2_weight"), val = tensor<fp32, [480, 480, 3, 3]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(74688)))];
tensor<fp32, [480]> m_conv3_bias = const()[name = string("m_conv3_bias"), val = tensor<fp32, [480]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(8369152)))];
tensor<fp32, [480, 480, 3, 3]> m_conv3_weight = const()[name = string("m_conv3_weight"), val = tensor<fp32, [480, 480, 3, 3]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(8371136)))];
tensor<fp32, [1024, 7680]> m_conv_out_weight = const()[name = string("m_conv_out_weight"), val = tensor<fp32, [1024, 7680]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(16665600)))];
tensor<int32, [4]> var_18 = const()[name = string("op_18"), val = tensor<int32, [4]>([1, 1, 128, 100])];
tensor<fp32, [1, 1, 128, 100]> input_1 = reshape(shape = var_18, x = mel)[name = string("input_1")];
string var_31_pad_type_0 = const()[name = string("op_31_pad_type_0"), val = string("custom")];
tensor<int32, [4]> var_31_pad_0 = const()[name = string("op_31_pad_0"), val = tensor<int32, [4]>([1, 1, 1, 1])];
tensor<int32, [2]> var_31_strides_0 = const()[name = string("op_31_strides_0"), val = tensor<int32, [2]>([2, 2])];
tensor<int32, [2]> var_31_dilations_0 = const()[name = string("op_31_dilations_0"), val = tensor<int32, [2]>([1, 1])];
int32 var_31_groups_0 = const()[name = string("op_31_groups_0"), val = int32(1)];
tensor<fp32, [1, 480, 64, 50]> var_31 = conv(bias = m_conv1_bias, dilations = var_31_dilations_0, groups = var_31_groups_0, pad = var_31_pad_0, pad_type = var_31_pad_type_0, strides = var_31_strides_0, weight = m_conv1_weight, x = input_1)[name = string("op_31")];
string var_33_mode_0 = const()[name = string("op_33_mode_0"), val = string("EXACT")];
tensor<fp32, [1, 480, 64, 50]> var_33 = gelu(mode = var_33_mode_0, x = var_31)[name = string("op_33")];
tensor<fp32, [1, 480, 64, 50]> input_3 = mul(x = var_33, y = mask1)[name = string("input_3")];
string var_46_pad_type_0 = const()[name = string("op_46_pad_type_0"), val = string("custom")];
tensor<int32, [4]> var_46_pad_0 = const()[name = string("op_46_pad_0"), val = tensor<int32, [4]>([1, 1, 1, 1])];
tensor<int32, [2]> var_46_strides_0 = const()[name = string("op_46_strides_0"), val = tensor<int32, [2]>([2, 2])];
tensor<int32, [2]> var_46_dilations_0 = const()[name = string("op_46_dilations_0"), val = tensor<int32, [2]>([1, 1])];
int32 var_46_groups_0 = const()[name = string("op_46_groups_0"), val = int32(1)];
tensor<fp32, [1, 480, 32, 25]> var_46 = conv(bias = m_conv2_bias, dilations = var_46_dilations_0, groups = var_46_groups_0, pad = var_46_pad_0, pad_type = var_46_pad_type_0, strides = var_46_strides_0, weight = m_conv2_weight, x = input_3)[name = string("op_46")];
string var_48_mode_0 = const()[name = string("op_48_mode_0"), val = string("EXACT")];
tensor<fp32, [1, 480, 32, 25]> var_48 = gelu(mode = var_48_mode_0, x = var_46)[name = string("op_48")];
tensor<fp32, [1, 480, 32, 25]> input_5 = mul(x = var_48, y = mask2)[name = string("input_5")];
string var_61_pad_type_0 = const()[name = string("op_61_pad_type_0"), val = string("custom")];
tensor<int32, [4]> var_61_pad_0 = const()[name = string("op_61_pad_0"), val = tensor<int32, [4]>([1, 1, 1, 1])];
tensor<int32, [2]> var_61_strides_0 = const()[name = string("op_61_strides_0"), val = tensor<int32, [2]>([2, 2])];
tensor<int32, [2]> var_61_dilations_0 = const()[name = string("op_61_dilations_0"), val = tensor<int32, [2]>([1, 1])];
int32 var_61_groups_0 = const()[name = string("op_61_groups_0"), val = int32(1)];
tensor<fp32, [1, 480, 16, 13]> var_61 = conv(bias = m_conv3_bias, dilations = var_61_dilations_0, groups = var_61_groups_0, pad = var_61_pad_0, pad_type = var_61_pad_type_0, strides = var_61_strides_0, weight = m_conv3_weight, x = input_5)[name = string("op_61")];
string x_mode_0 = const()[name = string("x_mode_0"), val = string("EXACT")];
tensor<fp32, [1, 480, 16, 13]> x = gelu(mode = x_mode_0, x = var_61)[name = string("x")];
tensor<int32, [4]> var_68 = const()[name = string("op_68"), val = tensor<int32, [4]>([0, 3, 1, 2])];
tensor<int32, [3]> var_73 = const()[name = string("op_73"), val = tensor<int32, [3]>([1, 13, 7680])];
tensor<fp32, [1, 13, 480, 16]> var_69 = transpose(perm = var_68, x = x)[name = string("transpose_0")];
tensor<fp32, [1, 13, 7680]> input = reshape(shape = var_73, x = var_69)[name = string("input")];
tensor<fp32, [1024]> linear_0_bias_0 = const()[name = string("linear_0_bias_0"), val = tensor<fp32, [1024]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(48122944)))];
tensor<fp32, [1, 13, 1024]> var_77 = linear(bias = linear_0_bias_0, weight = m_conv_out_weight, x = input)[name = string("linear_0")];
tensor<fp32, [1, 13, 1024]> embeddings = add(x = var_77, y = m_pos)[name = string("op_79")];
} -> (embeddings);
} |