diff --git a/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp b/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp
--- a/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp
+++ b/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp
@@ -165,6 +165,12 @@
 
   LogicalResult matchAndRewrite(tosa::TransposeOp op,
                                 PatternRewriter &rewriter) const override {
+    auto outputType = op.getType().cast<ShapedType>();
+    ArrayRef<int64_t> outputShape = outputType.getShape();
+    // TOSA supports quantized types.
+    if (!outputType.getElementType().isIntOrIndexOrFloat())
+      return failure();
+
     DenseElementsAttr inputValues;
     if (!matchPattern(op.input1(), m_Constant(&inputValues)))
       return failure();
@@ -184,9 +190,6 @@
     ArrayRef<int64_t> inputShape = inputType.getShape();
     int64_t numElements = inputType.getNumElements();
 
-    auto outputType = op.getType().cast<ShapedType>();
-    ArrayRef<int64_t> outputShape = outputType.getShape();
-
     SmallVector<Attribute, 4> outputValues;
     outputValues.resize(numElements);
 
diff --git a/mlir/test/Dialect/Tosa/canonicalize.mlir b/mlir/test/Dialect/Tosa/canonicalize.mlir
--- a/mlir/test/Dialect/Tosa/canonicalize.mlir
+++ b/mlir/test/Dialect/Tosa/canonicalize.mlir
@@ -314,3 +314,14 @@
   %1 = "tosa.transpose"(%input, %perms) : (tensor<2x3xf32>, tensor<2xi32>) -> tensor<3x2xf32>
   return %1, %input : tensor<3x2xf32>, tensor<2x3xf32>
 }
+
+// -----
+
+// CHECK-LABEL: @transpose_nofold_quantized_types
+func @transpose_nofold_quantized_types() -> tensor<1x1x16x1x!quant.uniform<i8<-127:127>:f32:3, {1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,2.100000e+00,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01}>> {
+  %perms = "tosa.const"() {value = dense<[1, 2, 3, 0]> : tensor<4xi32>} : () -> tensor<4xi32>
+  %input = "tosa.const"() {value = dense<[[[[-127, 127, 127, -127, -127, -127, -127, -127, -127, 127, 127, 127, 127, 127, -127, 127]]]]> : tensor<1x1x1x16xi8>} : () -> tensor<1x1x1x16xi8>
+  // CHECK: tosa.transpose
+  %0 = "tosa.transpose"(%input, %perms) : (tensor<1x1x1x16xi8>, tensor<4xi32>) -> tensor<1x1x16x1x!quant.uniform<i8<-127:127>:f32:3, {1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,2.100000e+00,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01}>>
+  return %0: tensor<1x1x16x1x!quant.uniform<i8<-127:127>:f32:3, {1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,2.100000e+00,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01,1.000000e-01}>>
+}