Add Module execute API.

Summary: #8363

Reviewed By: bsoyluoglu

Differential Revision: D71919658

Author: shoumikhin

extension/apple/ExecuTorch/Exported/ExecuTorchModule.h

@@ -115,6 +115,21 @@ __attribute__((deprecated("This API is experimental.")))
  */
 - (nullable NSSet<NSString *> *)methodNames:(NSError **)error;
 
+/**
+ * Executes a specific method with the provided input values.
+ *
+ * The method is loaded on demand if not already loaded.
+ *
+ * @param methodName A string representing the method name.
+ * @param values An NSArray of ExecuTorchValue objects representing the inputs.
+ * @param error A pointer to an NSError pointer that is set if an error occurs.
+ * @return An NSArray of ExecuTorchValue objects representing the outputs, or nil in case of an error.
+ */
+- (nullable NSArray<ExecuTorchValue *> *)executeMethod:(NSString *)methodName
+                                            withInputs:(NSArray<ExecuTorchValue *> *)values
+                                                 error:(NSError **)error
+    NS_SWIFT_NAME(execute(_:_:));
+
 + (instancetype)new NS_UNAVAILABLE;
 - (instancetype)init NS_UNAVAILABLE;
 

extension/apple/ExecuTorch/Exported/ExecuTorchModule.mm

@@ -16,6 +16,27 @@
 using namespace executorch::extension;
 using namespace executorch::runtime;
 
+static inline EValue toEValue(ExecuTorchValue *value) {
+  if (value.isTensor) {
+    auto *nativeTensorPtr = value.tensorValue.nativeInstance;
+    ET_CHECK(nativeTensorPtr);
+    auto nativeTensor = *reinterpret_cast<TensorPtr *>(nativeTensorPtr);
+    ET_CHECK(nativeTensor);
+    return *nativeTensor;
+  }
+  ET_CHECK_MSG(false, "Unsupported ExecuTorchValue type");
+  return EValue();
+}
+
+static inline ExecuTorchValue *toExecuTorchValue(EValue value) {
+  if (value.isTensor()) {
+    auto nativeInstance = make_tensor_ptr(value.toTensor());
+    return [ExecuTorchValue valueWithTensor:[[ExecuTorchTensor alloc] initWithNativeInstance:&nativeInstance]];
+  }
+  ET_CHECK_MSG(false, "Unsupported EValue type");
+  return [ExecuTorchValue new];
+}
+
 @implementation ExecuTorchModule {
   std::unique_ptr<Module> _module;
 }
@@ -94,4 +115,28 @@ - (BOOL)isMethodLoaded:(NSString *)methodName {
   return methods;
 }
 
+- (nullable NSArray<ExecuTorchValue *> *)executeMethod:(NSString *)methodName
+                                            withInputs:(NSArray<ExecuTorchValue *> *)values
+                                                 error:(NSError **)error {
+  std::vector<EValue> inputs;
+  inputs.reserve(values.count);
+  for (ExecuTorchValue *value in values) {
+    inputs.push_back(toEValue(value));
+  }
+  const auto result = _module->execute(methodName.UTF8String, inputs);
+  if (!result.ok()) {
+    if (error) {
+      *error = [NSError errorWithDomain:ExecuTorchErrorDomain
+                                   code:(NSInteger)result.error()
+                               userInfo:nil];
+    }
+    return nil;
+  }
+  NSMutableArray<ExecuTorchValue *> *outputs = [NSMutableArray arrayWithCapacity:result->size()];
+  for (const auto &value : *result) {
+    [outputs addObject:toExecuTorchValue(value)];
+  }
+  return outputs;
+}
+
 @end

extension/apple/ExecuTorch/__tests__/ModuleTest.swift

@@ -51,4 +51,25 @@ class ModuleTest: XCTestCase {
     XCTAssertNoThrow(methodNames = try module.methodNames())
     XCTAssertEqual(methodNames, Set(["forward"]))
   }
+
+  func testExecute() {
+    let bundle = Bundle(for: type(of: self))
+    guard let modelPath = bundle.path(forResource: "add", ofType: "pte") else {
+      XCTFail("Couldn't find the model file")
+      return
+    }
+    let module = Module(filePath: modelPath)
+    var inputData: [Float] = [1.0]
+    let inputTensor = inputData.withUnsafeMutableBytes {
+      Tensor(bytesNoCopy: $0.baseAddress!, shape:[1], dataType: .float)
+    }
+    let inputs = [Value(inputTensor), Value(inputTensor)]
+    var outputs: [Value]?
+    XCTAssertNoThrow(outputs = try module.execute("forward", inputs))
+    var outputData: [Float] = [2.0]
+    let outputTensor = outputData.withUnsafeMutableBytes {
+      Tensor(bytesNoCopy: $0.baseAddress!, shape:[1], dataType: .float, shapeDynamism: .static)
+    }
+    XCTAssertEqual(outputs?[0].tensor, outputTensor)
+  }
 }