[TOSA] Add torch.prim.NumToTensor.Scalar float support

lib/Conversion/TorchToTosa/TorchToTosa.cpp

@@ -3459,13 +3459,27 @@ LogicalResult ConvertAtenOp<PrimNumToTensorScalarOp>::matchAndRewrite(
   // Only supports integer operand type, because for the floating point operand
   // type result tensor has to be of type `f64` which is not supported in the
   // tosa.
-  int64_t initValue;
-  if (!matchPattern(op.getA(), m_TorchConstantInt(&initValue)))
-    return rewriter.notifyMatchFailure(
-        op, "unimplemented: input should be a torch constant int");
+  double doubleValue;
+  auto isFloat = matchPattern(op.getA(), m_TorchConstantFloat(&doubleValue));
+  int64_t intValue;
+  auto isInt = matchPattern(op.getA(), m_TorchConstantInt(&intValue));
+  if (!isFloat && !isInt)
+    return rewriter.notifyMatchFailure(op,
+                                       "Unable to extract the scalar constant");
+  float floatValue = static_cast<float>(doubleValue);
+
+  auto outElemTy = resultType.getElementType();
+  if (outElemTy.isa<mlir::IntegerType>() || outElemTy.isF32()) {
+    DenseElementsAttr constAttr =
+        isInt ? DenseElementsAttr::get(resultType, {intValue})
+              : DenseElementsAttr::get(resultType, {floatValue});
+    rewriter.replaceOpWithNewOp<tosa::ConstOp>(op, resultType, constAttr);
+  } else if (outElemTy.isF64()) {
+    auto resultF32 =
+        tosa::getConstTensor<float>(rewriter, op, floatValue, {}).value();
+    rewriter.replaceOpWithNewOp<tosa::CastOp>(op, resultType, resultF32);
+  }
 
-  DenseElementsAttr constAttr = DenseElementsAttr::get(resultType, {initValue});
-  rewriter.replaceOpWithNewOp<tosa::ConstOp>(op, resultType, constAttr);
   return success();
 }