com.github.asus4.onnxruntime 0.4.1 → 0.4.2

package/Plugins/iOS~/onnxruntime.xcframework/ios-arm64_x86_64-simulator/onnxruntime.framework/Headers/onnxruntime_run_options_config_keys.h

@@ -43,6 +43,9 @@ static const char* const kOrtRunOptionsConfigQnnPerfModePostRun = "qnn.htp_perf_
 // Set RPC control latency for QNN HTP backend
 static const char* const kOrtRunOptionsConfigQnnRpcControlLatency = "qnn.rpc_control_latency";
 
+// Set QNN Lora Config File for apply Lora in QNN context binary
+static const char* const kOrtRunOptionsConfigQnnLoraConfig = "qnn.lora_config";
+
 // Set graph annotation id for CUDA EP. Use with enable_cuda_graph=true.
 // The value should be an integer. If the value is not set, the default value is 0 and
 // ORT session only captures one cuda graph before another capture is requested.

package/Plugins/iOS~/onnxruntime.xcframework/ios-arm64_x86_64-simulator/onnxruntime.framework/Headers/onnxruntime_session_options_config_keys.h

@@ -10,10 +10,10 @@
  * "[Area][.[SubArea1].[SubArea2]...].[Keyname]"
  * Such as "ep.cuda.use_arena"
  * The Config Key cannot be empty
- * The maximum length of the Config Key is 128
+ * The maximum length of the Config Key is 1024
  *
  * The string format of a SessionOptions Config Value is defined individually for each Config.
- * The maximum length of the Config Value is 1024
+ * The maximum length of the Config Value is 2048
  */
 
 // Key for disable PrePacking,
@@ -315,11 +315,19 @@ static const char* const kOrtSessionOptionEpContextEmbedMode = "ep.context_embed
 // in case user need to merge/connect multiple EPContext nodes in one model
 static const char* const kOrtSessionOptionEpContextNodeNamePrefix = "ep.context_node_name_prefix";
 
-// Share EP related resources across EPs
+// Share EP related resources across sessions
 static const char* const kOrtSessionOptionShareEpContexts = "ep.share_ep_contexts";
 
-// Use this config when dumping EP context model with an external initializers file
-// All initializers will be inside the external data file if specified, otherwise all in Onnx file
+// Stop to share EP related resources across sessions from then on
+static const char* const kOrtSessionOptionStopShareEpContexts = "ep.stop_share_ep_contexts";
+
+// Used only for context model generation.
+// This configuration is used when some nodes are partitioned on the CPU EP and those nodes have external initializers.
+// When generating the EP context model, the new model should not rely on the old external data file used by the source ONNX model.
+// Use this setting when dumping the EP context model with an external initializers file.
+// If specified, all initializers will be placed inside the external data file.
+// Otherwise, all initializers will be embedded inside the generated ONNX file.
+// By default, this option is not set, meaning all initializers will be included within the ONNX file.
 static const char* const kOrtSessionOptionsEpContextModelExternalInitializersFileName =
     "ep.context_model_external_initializers_file_name";
 
@@ -340,3 +348,19 @@ static const char* const kOrtSessionOptionsQDQMatMulNBitsAccuracyLevel = "sessio
 // “Default”: OS determines the scheduling priority and processor performance to service this workload. [Default]
 // “Efficient”: OS treats this workload is efficiency oriented with low scheduling priority and efficient processor performance.
 static const char* const kOrtEpDynamicOptionsWorkloadType = "ep.dynamic.workload_type";
+
+// Disables model compilation during session initialization.
+//
+// If this option is set to "1", inference session creation will fail with error code ORT_MODEL_REQUIRES_COMPILATION
+// if compilation is required to run the model on any Execution Provider added to the session.
+// Only the following kinds of models are valid when this option is set to "1":
+//   - Pre-compiled models that have EPContext nodes for the compiling Execution Providers in the session.
+//   - Non-compiled models that run only on non-compiling Execution Providers, like CPU EP.
+//
+// See \href https://onnxruntime.ai/docs/execution-providers/EP-Context-Design.html for details about
+// compiled models with EPContext nodes.
+//
+// Option values:
+// - "0": EP compile is not disabled. [DEFAULT]
+// - "1": EP compile is disabled.
+static const char* const kOrtSessionOptionsDisableModelCompile = "session.disable_model_compile";

package/Plugins/iOS~/onnxruntime.xcframework/ios-arm64_x86_64-simulator/onnxruntime.framework/Info.plist

@@ -9,9 +9,9 @@
 	<key>CFBundleIdentifier</key>
 	<string>com.microsoft.onnxruntime</string>
 	<key>CFBundleVersion</key>
-	<string>1.21.0</string>
+	<string>1.22.1</string>
 	<key>CFBundleShortVersionString</key>
-	<string>1.21.0</string>
+	<string>1.22.1</string>
 	<key>CFBundleSignature</key>
 	<string>????</string>
 	<key>CFBundlePackageType</key>

package/README.md

@@ -23,8 +23,9 @@ Pre-built ONNX Runtime libraries for Unity.
 ## Tested environment
 
 - Unity: 6000.0.43f1 (LTS)
-- ONNX Runtime: [1.21.0](https://github.com/microsoft/onnxruntime/releases/tag/v1.21.0)
+- ONNX Runtime: [1.22.1](https://github.com/microsoft/onnxruntime/releases/tag/v1.22.1)
 - ONNX Runtime Extensions: [0.14.0](https://github.com/microsoft/onnxruntime-extensions/releases/tag/v0.14.0)
+- ONNX Runtime GenAI: [0.8.3](https://github.com/microsoft/onnxruntime-genai/releases/tag/v0.8.3)
 
 ### Execution Providers & Extensions
 
@@ -70,10 +71,10 @@ Pre-built libraries are available on [NPM](https://www.npmjs.com/package/com.git
     }
   ]
   "dependencies": {
-    "com.github.asus4.onnxruntime": "0.4.1",
-    "com.github.asus4.onnxruntime.unity": "0.4.1",
-    "com.github.asus4.onnxruntime.win-x64-gpu": "0.4.1",
-    "com.github.asus4.onnxruntime-extensions": "0.4.1",
+    "com.github.asus4.onnxruntime": "0.4.2",
+    "com.github.asus4.onnxruntime.unity": "0.4.2",
+    "com.github.asus4.onnxruntime.win-x64-gpu": "0.4.2",
+    "com.github.asus4.onnxruntime-extensions": "0.4.2",
     ... other dependencies
   }
 ```

package/Runtime/ManagedProjections.shared.cs

@@ -25,7 +25,7 @@ namespace Microsoft.ML.OnnxRuntime
         /// </summary>
         /// <param name="namedOnnxValue"></param>
         /// <param name="metadata"></param>
-        /// <returns>OrtValye created accoding to the metadata</returns>
+        /// <returns>OrtValue created according to the metadata</returns>
         internal static OrtValue CreateProjection(NamedOnnxValue namedOnnxValue, NodeMetadata metadata)
         {
             OrtValue result;
@@ -191,4 +191,3 @@ namespace Microsoft.ML.OnnxRuntime
         }
     }
 }
-

package/Runtime/NativeMethods.shared.cs

@@ -15,6 +15,7 @@
 
 using System;
 using System.Runtime.InteropServices;
+using static Microsoft.ML.OnnxRuntime.NativeMethods;
 
 namespace Microsoft.ML.OnnxRuntime
 {
@@ -337,6 +338,16 @@ namespace Microsoft.ML.OnnxRuntime
         public IntPtr CreateLoraAdapterFromArray;
         public IntPtr ReleaseLoraAdapter;
         public IntPtr RunOptionsAddActiveLoraAdapter;
+        public IntPtr SetEpDynamicOptions;
+        public IntPtr ReleaseValueInfo;
+        public IntPtr ReleaseNode;
+        public IntPtr ReleaseGraph;
+        public IntPtr ReleaseModel;
+        public IntPtr GetValueInfoName;
+        public IntPtr GetValueInfoTypeInfo;
+        public IntPtr GetModelEditorApi;
+        public IntPtr CreateTensorWithDataAndDeleterAsOrtValue;
+        public IntPtr SessionOptionsSetLoadCancellationFlag;
     }
 
     internal static class NativeMethods
@@ -416,6 +427,7 @@ namespace Microsoft.ML.OnnxRuntime
             OrtReleaseSessionOptions = (DOrtReleaseSessionOptions)Marshal.GetDelegateForFunctionPointer(api_.ReleaseSessionOptions, typeof(DOrtReleaseSessionOptions));
             OrtCloneSessionOptions = (DOrtCloneSessionOptions)Marshal.GetDelegateForFunctionPointer(api_.CloneSessionOptions, typeof(DOrtCloneSessionOptions));
             OrtSetSessionExecutionMode = (DOrtSetSessionExecutionMode)Marshal.GetDelegateForFunctionPointer(api_.SetSessionExecutionMode, typeof(DOrtSetSessionExecutionMode));
+            OrtSessionOptionsSetLoadCancellationFlag = (DOrtSessionOptionsSetLoadCancellationFlag)Marshal.GetDelegateForFunctionPointer(api_.SessionOptionsSetLoadCancellationFlag, typeof(DOrtSessionOptionsSetLoadCancellationFlag));
             OrtSetOptimizedModelFilePath = (DOrtSetOptimizedModelFilePath)Marshal.GetDelegateForFunctionPointer(api_.SetOptimizedModelFilePath, typeof(DOrtSetOptimizedModelFilePath));
             OrtEnableProfiling = (DOrtEnableProfiling)Marshal.GetDelegateForFunctionPointer(api_.EnableProfiling, typeof(DOrtEnableProfiling));
             OrtDisableProfiling = (DOrtDisableProfiling)Marshal.GetDelegateForFunctionPointer(api_.DisableProfiling, typeof(DOrtDisableProfiling));
@@ -859,7 +871,7 @@ namespace Microsoft.ML.OnnxRuntime
         /// Creates an instance of OrtSession with provided parameters
         /// </summary>
         /// <param name="environment">Native OrtEnv instance</param>
-        /// <param name="modelData">Byte array correspoonding to the model</param>
+        /// <param name="modelData">Byte array corresponding to the model</param>
         /// <param name="modelSize">Size of the model in bytes</param>
         /// <param name="sessionOptions">Native SessionOptions instance</param>
         /// <param name="prepackedWeightsContainer">Native OrtPrepackedWeightsContainer instance</param>
@@ -1037,6 +1049,12 @@ namespace Microsoft.ML.OnnxRuntime
                                                                             ExecutionMode execution_mode);
         public static DOrtSetSessionExecutionMode OrtSetSessionExecutionMode;
 
+        [UnmanagedFunctionPointer(CallingConvention.Winapi)]
+        public delegate IntPtr /*(OrtStatus*)*/ DOrtSessionOptionsSetLoadCancellationFlag(IntPtr /*(OrtSessionOptions*)*/ options,
+                                                                        bool value);
+        public static DOrtSessionOptionsSetLoadCancellationFlag OrtSessionOptionsSetLoadCancellationFlag;
+
+
         [UnmanagedFunctionPointer(CallingConvention.Winapi)]
         public delegate IntPtr /*(OrtStatus*)*/ DOrtSetOptimizedModelFilePath(IntPtr /* OrtSessionOptions* */ options, byte[] optimizedModelFilepath);
         public static DOrtSetOptimizedModelFilePath OrtSetOptimizedModelFilePath;
@@ -1270,7 +1288,7 @@ namespace Microsoft.ML.OnnxRuntime
         /// </summary>
         /// <param name="options">Native SessionOptions instance</param>
         /// <param name="name">Name of the initializer</param>
-        /// <param name="ortValue">Native OrtValue instnce</param>
+        /// <param name="ortValue">Native OrtValue instance</param>
         [UnmanagedFunctionPointer(CallingConvention.Winapi)]
         public delegate IntPtr /*(OrtStatus*)*/ DOrtAddInitializer(IntPtr /*(OrtSessionOptions*)*/ options,
                                                                    byte[] /*(const char*)*/ name,

package/Runtime/OrtValue.shared.cs

@@ -2,7 +2,7 @@
 // #define NET8_0_OR_GREATER
 #endif // UNITY_6000_0_OR_NEWER
 
-// Copyright (c) Microsoft Corporation. All rights reserved.
+// Copyright (c) Microsoft Corporation. All rights reserved.
 // Licensed under the MIT License.
 
 using Microsoft.ML.OnnxRuntime.Tensors;

package/Runtime/SessionOptions.shared.cs

@@ -814,6 +814,16 @@ namespace Microsoft.ML.OnnxRuntime
         }
         private ExecutionMode _executionMode = ExecutionMode.ORT_SEQUENTIAL;
 
+        /// <summary>
+        /// Sets the load cancellation flag for the session. Default is set to false.
+        /// Provides an opportunity for the user to cancel model loading.
+        /// </summary>
+        /// <param name="value">true to request cancellation, false to proceed</param>
+        public void SetLoadCancellationFlag(bool value)
+        {
+            NativeApiStatus.VerifySuccess(NativeMethods.OrtSessionOptionsSetLoadCancellationFlag(handle, value));
+        }
+
         #endregion
 
         #region Private Methods

package/Runtime/Training/NativeTrainingMethods.shared.cs

@@ -76,7 +76,7 @@ namespace Microsoft.ML.OnnxRuntime
                 DOrtGetApi OrtGetApi = (DOrtGetApi)Marshal.GetDelegateForFunctionPointer(NativeMethods.OrtGetApiBase().GetApi, typeof(DOrtGetApi));
 #endif
 
-                const uint ORT_API_VERSION = 21;
+                const uint ORT_API_VERSION = 22;
 #if NETSTANDARD2_0
                 IntPtr ortApiPtr = OrtGetApi(ORT_API_VERSION);
                 api_ = (OrtApi)Marshal.PtrToStructure(ortApiPtr, typeof(OrtApi));

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "com.github.asus4.onnxruntime",
-  "version": "0.4.1",
+  "version": "0.4.2",
   "displayName": "ONNX Runtime",
   "description": "ONNX Runtime Plugin for Unity",
   "keywords": [