50 lines
4.6 KiB
MLIR
50 lines
4.6 KiB
MLIR
// RUN: scalehls-opt -scalehls-tosa-fake-quantize %s | FileCheck %s
|
|
|
|
module {
|
|
// CHECK: func @forward(%arg0: tensor<1x3x32x32xi8>) -> tensor<1x10xi8> {
|
|
func @forward(%arg0: tensor<1x3x32x32xf32>) -> tensor<1x10xf32> {
|
|
|
|
// CHECK: %0 = "tosa.const"() {value = dense<1> : tensor<1x10xi8>} : () -> tensor<1x10xi8>
|
|
// CHECK: %1 = "tosa.const"() {value = dense<2> : tensor<1x64x10xi8>} : () -> tensor<1x64x10xi8>
|
|
%0 = "tosa.const"() {value = dense<0.000000e+00> : tensor<1x10xf32>} : () -> tensor<1x10xf32>
|
|
%1 = "tosa.const"() {value = dense<1.000000e+00> : tensor<1x64x10xf32>} : () -> tensor<1x64x10xf32>
|
|
%2 = "tosa.const"() {value = dense<2.000000e+00> : tensor<64x3x3x64xf32>} : () -> tensor<64x3x3x64xf32>
|
|
%3 = "tosa.const"() {value = dense<3.000000e+00> : tensor<64x3x3x64xf32>} : () -> tensor<64x3x3x64xf32>
|
|
%4 = "tosa.const"() {value = dense<4.000000e+00> : tensor<64x3x3x3xf32>} : () -> tensor<64x3x3x3xf32>
|
|
|
|
// CHECK: %5 = "tosa.const"() {value = dense<[0, 3, 1, 2]> : tensor<4xi32>} : () -> tensor<4xi32>
|
|
%5 = "tosa.const"() {value = dense<[0, 3, 1, 2]> : tensor<4xi32>} : () -> tensor<4xi32>
|
|
%6 = "tosa.const"() {value = dense<[0, 2, 3, 1]> : tensor<4xi32>} : () -> tensor<4xi32>
|
|
%7 = "tosa.const"() {value = dense<5.000000e+00> : tensor<64xf32>} : () -> tensor<64xf32>
|
|
|
|
// CHECK: %8 = "tosa.transpose"(%arg0, %6) : (tensor<1x3x32x32xi8>, tensor<4xi32>) -> tensor<1x32x32x3xi8>
|
|
%8 = "tosa.transpose"(%arg0, %6) : (tensor<1x3x32x32xf32>, tensor<4xi32>) -> tensor<1x32x32x3xf32>
|
|
|
|
// CHECK: %9 = "tosa.conv2d"(%8, %4, %7) {dilation = [1, 1], pad = [1, 1, 1, 1], quantization_info = {input_zp = 0 : i32, weight_zp = 0 : i32}, stride = [1, 1]} : (tensor<1x32x32x3xi8>, tensor<64x3x3x3xi8>, tensor<64xi8>) -> tensor<1x32x32x64xi8>
|
|
%9 = "tosa.conv2d"(%8, %4, %7) {dilation = [1, 1], pad = [1, 1, 1, 1], stride = [1, 1]} : (tensor<1x32x32x3xf32>, tensor<64x3x3x3xf32>, tensor<64xf32>) -> tensor<1x32x32x64xf32>
|
|
%10 = "tosa.transpose"(%9, %5) : (tensor<1x32x32x64xf32>, tensor<4xi32>) -> tensor<1x64x32x32xf32>
|
|
%11 = "tosa.clamp"(%10) {max_fp = 3.40282347E+38 : f32, max_int = 2147483647 : i64, min_fp = 0.000000e+00 : f32, min_int = 0 : i64} : (tensor<1x64x32x32xf32>) -> tensor<1x64x32x32xf32>
|
|
%12 = "tosa.transpose"(%11, %6) : (tensor<1x64x32x32xf32>, tensor<4xi32>) -> tensor<1x32x32x64xf32>
|
|
%13 = "tosa.conv2d"(%12, %3, %7) {dilation = [1, 1], pad = [1, 1, 1, 1], stride = [1, 1]} : (tensor<1x32x32x64xf32>, tensor<64x3x3x64xf32>, tensor<64xf32>) -> tensor<1x32x32x64xf32>
|
|
%14 = "tosa.transpose"(%13, %5) : (tensor<1x32x32x64xf32>, tensor<4xi32>) -> tensor<1x64x32x32xf32>
|
|
%15 = "tosa.clamp"(%14) {max_fp = 3.40282347E+38 : f32, max_int = 2147483647 : i64, min_fp = 0.000000e+00 : f32, min_int = 0 : i64} : (tensor<1x64x32x32xf32>) -> tensor<1x64x32x32xf32>
|
|
%16 = "tosa.transpose"(%15, %6) : (tensor<1x64x32x32xf32>, tensor<4xi32>) -> tensor<1x32x32x64xf32>
|
|
%17 = "tosa.conv2d"(%16, %2, %7) {dilation = [1, 1], pad = [1, 1, 1, 1], stride = [1, 1]} : (tensor<1x32x32x64xf32>, tensor<64x3x3x64xf32>, tensor<64xf32>) -> tensor<1x32x32x64xf32>
|
|
%18 = "tosa.transpose"(%17, %5) : (tensor<1x32x32x64xf32>, tensor<4xi32>) -> tensor<1x64x32x32xf32>
|
|
%19 = "tosa.add"(%18, %11) : (tensor<1x64x32x32xf32>, tensor<1x64x32x32xf32>) -> tensor<1x64x32x32xf32>
|
|
%20 = "tosa.clamp"(%19) {max_fp = 3.40282347E+38 : f32, max_int = 2147483647 : i64, min_fp = 0.000000e+00 : f32, min_int = 0 : i64} : (tensor<1x64x32x32xf32>) -> tensor<1x64x32x32xf32>
|
|
%21 = "tosa.transpose"(%20, %6) : (tensor<1x64x32x32xf32>, tensor<4xi32>) -> tensor<1x32x32x64xf32>
|
|
|
|
// CHECK: %22 = "tosa.avg_pool2d"(%21) {kernel = [32, 32], pad = [0, 0, 0, 0], quantization_info = {input_zp = 0 : i32, output_zp = 0 : i32}, stride = [32, 32]} : (tensor<1x32x32x64xi8>) -> tensor<1x1x1x64xi8>
|
|
%22 = "tosa.avg_pool2d"(%21) {kernel = [32, 32], pad = [0, 0, 0, 0], stride = [32, 32]} : (tensor<1x32x32x64xf32>) -> tensor<1x1x1x64xf32>
|
|
%23 = "tosa.transpose"(%22, %5) : (tensor<1x1x1x64xf32>, tensor<4xi32>) -> tensor<1x64x1x1xf32>
|
|
%24 = "tosa.reshape"(%23) {new_shape = [1, 1, 64]} : (tensor<1x64x1x1xf32>) -> tensor<1x1x64xf32>
|
|
|
|
// CHECK: %25 = "tosa.matmul"(%24, %1) {quantization_info = {a_zp = 0 : i32, b_zp = 0 : i32}} : (tensor<1x1x64xi8>, tensor<1x64x10xi8>) -> tensor<1x1x10xi8>
|
|
%25 = "tosa.matmul"(%24, %1) : (tensor<1x1x64xf32>, tensor<1x64x10xf32>) -> tensor<1x1x10xf32>
|
|
%26 = "tosa.reshape"(%25) {new_shape = [1, 10]} : (tensor<1x1x10xf32>) -> tensor<1x10xf32>
|
|
%27 = "tosa.add"(%26, %0) : (tensor<1x10xf32>, tensor<1x10xf32>) -> tensor<1x10xf32>
|
|
return %27 : tensor<1x10xf32>
|
|
}
|
|
}
|