Develop Reference

TensorflowSharp - TFException: Attempting to use uninitialized value

I'm attempting to import a model created in Keras/Tensorflow and use it for inference in a Unity project.
I have successfully imported the model and validated by printing names of input and output nodes in the graph. Though, when I try to get the output value from the runner, I get this exception.
TFException: Attempting to use uninitialized value action_W
[[Node: action_W/read = IdentityT=DT_FLOAT, _class=["loc:#action_W"], _device="/job:localhost/replica:0/task:0/cpu:0"]]
TensorFlow.TFStatus.CheckMaybeRaise (TensorFlow.TFStatus incomingStatus, System.Boolean last) (at <6ed6db22f8874deba74ffe3e566039be>:0)
TensorFlow.TFSession.Run (TensorFlow.TFOutput[] inputs, TensorFlow.TFTensor[] inputValues, TensorFlow.TFOutput[] outputs, TensorFlow.TFOperation[] targetOpers, TensorFlow.TFBuffer runMetadata, TensorFlow.TFBuffer runOptions, TensorFlow.TFStatus status) (at <6ed6db22f8874deba74ffe3e566039be>:0)
TensorFlow.TFSession+Runner.Run (TensorFlow.TFStatus status) (at <6ed6db22f8874deba74ffe3e566039be>:0)
RecordArbitraryData.ModelPredict (System.Single[,] input) (at Assets/Scripts/Spells/RecordArbitraryData.cs:230)
RecordArbitraryData.FixedUpdate () (at Assets/Scripts/Spells/RecordArbitraryData.cs:95)
Here are the two functions I use. InstantiateModel is called OnStart() in my unity script. And ModelPredict is called when the user passes an input to the script.
void InstantiateModel(){
string model_name = "simple_as_binary";
//Instantiate Graph
graphModel = Resources.Load (model_name) as TextAsset;
graph = new TFGraph ();
graph.Import (graphModel.bytes);
session = new TFSession (graph);
}
void ModelPredict(float[,] input){
using (graph) {
using (session) {
//Assign input tensors
var runner = session.GetRunner ();
runner.AddInput (graph [input_node_name] [0], input);
//Calculate and access output of graph
runner.Fetch (graph[output_node_name][0]);
Debug.Log ("Output node name: " + graph [output_node_name].Name);
float[,] recurrent_tensor = runner.Run () [0].GetValue () as float[,];
//var results = runner.Run();
//Debug.Log("Prediciton: " + results);
}
}
}
Any help appreciated - TensorflowSharp is very new to me.

I was able to figure out most of my problems. I'm currently at a point where my model is predicting in unity, but only predicting the first of four classes. My guess is it has something to do with the weights not getting initialized correctly from the checkpoint files? Edit: My values weren't being normalized before being passed to neural network.
Preface: Mozilla Firefox works best for displaying tensorboard; it took me a long time to realize that google chrome was causing my graph to be invisible (tensorboard is how I was able to figure out the nodes that needed to be used for input and output).
First Issue: I was renaming a .pb file into a .bytes file. This is incorrect because the model’s weights come from the checkpoint file, and are given to the nodes held in the .pb file. This was causing the uninitialized variables. These variables were used for training, which were removed after using the freeze_graph function.
Second Issue: I was using the file created called ‘checkpoint’, which was throwing an error. I then changed the name of the checkpoint to ‘test’ and used this in the freeze_graph function. When calling the checkpoints file, I was required to use ‘test.ckpt’. I assume this function knows to grab the three files automatically based on the .ckpt? ‘Test’ without ‘.ckpt’ did not work.
Third Issue: when using the freeze_graph function, I needed to export the .pb file in keras/tf with text=False. I tested True and False; True threw an error about “bad wiring”.
Fourth Issue: Tensorboard was very difficult to use without any organization. Using tf.name_scope helped a lot with not only visualization, but making sure I was using/referencing the correct nodes in TensorFlowSharp. In keras I found it helpful to separate the final Dense layer and Activation into their own scopes so I could find the correct output node. The rest of my network was put into a ‘body’ scope, and the sole input layer in ‘input’ scope. The name_scope function prepends ‘scopename/’ to the node name. I don't think it’s necessary, but it helped me.
Fifth Issue: The version of tensorflowsharp released as a unity package is not up to date. This caused an issue with a keras placeholder for ‘keras_training_phase’. In keras, you pass this along with an input like [0] + input. I tried to do the same by creating a new TFTensor(bool), but I was getting an error ‘inaccessible due to its protection level. This was an error with the implicit conversion between bool and TFTensor in my unity TensorFlowSharp version. To fix this I had to use a function found in this stackoverflow solution where the .bytes file is read in, the placeholder for keras_training_phase is found, and is swapped out for a bool constant set to false. This worked for me because my model was pretrained in python, so it may not be a great fix for someone that’s trying to train and test the model. A condition for removing this node with the freeze_graph function would really be useful.
Hope someone finds this useful!

WF4.5- Expression Activity type 'CSharpValue`1' requires compilation in order to run

Background
What I'm trying to do is to have a scoped variable of one of my models in the xaml.
In my workflow project "MyProject.Workflows" I have created model classes, code activities and Xaml files. They are all under same namespace. In another project ("Engine"), I load and execute these workflows.
To load the workflows in the "Engine", I use ActivityXamlServices with ActivityXamlServicesSettings including CompileExpressions = true.
When loading the ActivityXamlServices, I use a XamlXmlReader with XamlXmlReaderSettings where I actually point to the "MyProject.Workflows" dll.
Since Both these projects are in the same solution I actually referred MyProject.Workflows in the "Engine".
Because Earlier, they were in different solutions, So when I tried to do this It gave me It cant find the "MyProject.Workflows" dll even though I point it in the XamlXmlReaderSettings.
Then I tried to load the dll to the app domain and then it worked.But I did not want to deal with App Domains so I decided to get both projects under one solution so I can refer the "MyProject.Workflows" in the "Engine".
Issue:
If I use one of those models inside of the Xaml as an expression like "Assign Activity" the Workflow isn't getting compiled when I try to execute this.
For example if I use this in an "Assign" activity having a scoped variable of type MyObject
Newtonsoft.Json.JsonConvert.DeserializeObject<MyProject.Workflows.Models.MyObject>(inputString);
I will get the below error message when I run the workflow.
NotSupportedException:'Expression Activity type 'CSharpValue`1' requires compilation in order to run. Please ensure that the workflow has been compiled.
If I remove these objects and just deal with strings or ints, it works fine.
Things I found in my research:
I found this was a bug in .Net Framework 4.5. But Im using 4.6
Even though I used CompileExpressions = true , I tried this compile method I found. But did not change a thing.
private static void Compile(DynamicActivity dynamicActivity)
{
TextExpressionCompilerSettings settings = new TextExpressionCompilerSettings
{
Activity = dynamicActivity,
Language = "C#",
ActivityName = dynamicActivity.Name.Split('.').Last() + "_CompiledExpressionRoot",
ActivityNamespace = string.Join(".", dynamicActivity.Name.Split('.').Reverse().Skip(1).Reverse()),
RootNamespace = null,
GenerateAsPartialClass = false,
AlwaysGenerateSource = true,
};
TextExpressionCompilerResults results =
new TextExpressionCompiler(settings).Compile();
if (results.HasErrors)
{
throw new Exception("Compilation failed.");
}
ICompiledExpressionRoot compiledExpressionRoot =
Activator.CreateInstance(results.ResultType,
new object[] { dynamicActivity }) as ICompiledExpressionRoot;
CompiledExpressionInvoker.SetCompiledExpressionRootForImplementation(
dynamicActivity, compiledExpressionRoot);
}
I read that some people faced this problem and they had to actually move the models to a diffrent namespace. I did that too. Didn't fix the problem.
My Xaml file has this entry added at the top.
xmlns:local="clr-namespace:MyProject.Workflows.Models"
Can someone please help me to get through this?

While want to make an instance of SIFTDetector: Get an Attempted to read or write protected memory exception Error

I am trying to use sift algorithm to make the keypoints and descriptors in c# by using opencv library.
fileAddress = dlg.FileName;
cap = new Emgu.CV.Capture(dlg.FileName);
cap.SetCaptureProperty(Emgu.CV.CvEnum.CAP_PROP.CV_CAP_PROP_POS_FRAMES, 3945);
imGray = cap.QueryGrayFrame();
Emgu.CV.Features2D.SIFTDetector siftDet = new Emgu.CV.Features2D.SIFTDetector();
siftDet.DetectKeyPoints(imGray);
MessageBox.Show("test SIFT");
but when it go through the line Emgu.CV.Features2D.SIFTDetector siftDet = new Emgu.CV.Features2D.SIFTDetector(); I face the error:
and view detail is as following:
How could I solve this exception?

The openCv that I was using was version 2.1, then I upgraded the opencv*.* and *.dll files to version 2.4. The problem solved.
I couldn't find any explanation for that, that would be awesome if anybody could even give me an intuitive answer.
For now if you face such problem regarding working with SIFT, try the latest version.

C# code completion with NRefactory 5

I just found out about NRefactory 5 and I would guess, that it is the most suitable solution for my current problem. At the moment I'm developing a little C# scripting application for which I would like to provide code completion. Until recently I've done this using the "Roslyn" project from Microsoft. But as the latest update of this project requires .Net Framework 4.5 I can't use this any more as I would like the app to run under Win XP as well. So I have to switch to another technology here.
My problem is not the compilation stuff. This can be done, with some more effort, by .Net CodeDomProvider as well. The problem ist the code completion stuff. As far as I know, NRefactory 5 provides everything that is required to provide code completion (parser, type system etc.) but I just can't figure out how to use it. I took a look at SharpDevelop source code but they don't use NRefactory 5 for code completion there, they only use it as decompiler. As I couldn't find an example on how to use it for code completion in the net as well I thought that I might find some help here.
The situation is as follows. I have one single file containing the script code. Actually it is not even a file but a string which I get from the editor control (by the way: I'm using AvalonEdit for this. Great editor!) and some assemblies that needs to get referenced. So, no solution files, no project files etc. just one string of source code and the assemblies.
I've taken a look at the Demo that comes with NRefactory 5 and the article on code project and got up with something like this:
var unresolvedTypeSystem = syntaxTree.ToTypeSystem();
IProjectContent pc = new CSharpProjectContent();
// Add parsed files to the type system
pc = pc.AddOrUpdateFiles(unresolvedTypeSystem);
// Add referenced assemblies:
pc = pc.AddAssemblyReferences(new CecilLoader().LoadAssemblyFile(
System.Reflection.Assembly.GetAssembly(typeof(Object)).Location));
My problem is that I have no clue on how to go on. I'm not even sure if it is the right approach to accomplish my goal. How to use the CSharpCompletionEngine? What else is required? etc. You see there are many things that are very unclear at the moment and I hope you can bring some light into this.
Thank you all very much in advance!

I've just compiled and example project that does C# code completion with AvalonEdit and NRefactory.
It can be found on Github here.

Take a look at method ICSharpCode.NRefactory.CSharp.CodeCompletion.CreateEngine. You need to create an instance of CSharpCompletionEngine and pass in the correct document and the resolvers. I managed to get it working for CTRL+Space compltition scenario. However I am having troubles with references to types that are in other namespaces. It looks like CSharpTypeResolveContext does not take into account the using namespace statements - If I resolve the references with CSharpAstResolver, they are resolved OK, but I am unable to correctly use this resolver in code completition scenario...
UPDATE #1:
I've just managed to get the working by obtaining resolver from unresolved fail.
Here is the snippet:
var mb = new DefaultCompletionContextProvider(doc, unresolvedFile);
var resolver3 = unresolvedFile.GetResolver(cmp, loc); // get the resolver from unresolvedFile
var engine = new CSharpCompletionEngine(doc, mb, new CodeCompletionBugTests.TestFactory(resolver3), pctx, resolver3.CurrentTypeResolveContext );
Update #2:
Here is the complete method. It references classes from unit test projects, sou you would need to reference/copy them into your project:
public static IEnumerable<ICompletionData> DoCodeComplete(string editorText, int offset) // not the best way to put in the whole string every time
{
var doc = new ReadOnlyDocument(editorText);
var location = doc.GetLocation(offset);
string parsedText = editorText; // TODO: Why there are different values in test cases?
var syntaxTree = new CSharpParser().Parse(parsedText, "program.cs");
syntaxTree.Freeze();
var unresolvedFile = syntaxTree.ToTypeSystem();
var mb = new DefaultCompletionContextProvider(doc, unresolvedFile);
IProjectContent pctx = new CSharpProjectContent();
var refs = new List<IUnresolvedAssembly> { mscorlib.Value, systemCore.Value, systemAssembly.Value};
pctx = pctx.AddAssemblyReferences(refs);
pctx = pctx.AddOrUpdateFiles(unresolvedFile);
var cmp = pctx.CreateCompilation();
var resolver3 = unresolvedFile.GetResolver(cmp, location);
var engine = new CSharpCompletionEngine(doc, mb, new CodeCompletionBugTests.TestFactory(resolver3), pctx, resolver3.CurrentTypeResolveContext );
engine.EolMarker = Environment.NewLine;
engine.FormattingPolicy = FormattingOptionsFactory.CreateMono();
var data = engine.GetCompletionData(offset, controlSpace: false);
return data;
}
}
Hope it helps,
Matra

NRefactory 5 is being used in SharpDevelop 5. The source code for SharpDevelop 5 is currently available in the newNR branch on github. I would take a look at the CSharpCompletionBinding class which has code to display a completion list window using information from NRefactory's CSharpCompletionEngine.

NUnit is not failing test with dynamic keyword of .Net 4.0

I am using NUnit with Visual Studio Express Edition 2010 for C#, Now, normally test works fine. But whenever I try to use Massive.cs, which is open source api to access database. Test fails from that file only. Now, if I run the application, api is working fine. I have created a different library file to access data base.
I seriously don't understand the error. It is just giving error that object reference is not set to an object. But if I run the code, it works fine. I am using dynamic keyword as shown in link of api above. Does that making problem with NUnit ?
Is there any other way to test in this type of Scenarios?
Here are the further details of the code,
Test class is like this
dynamic item = new Item();
item.Insert(new { Name = "Maggi", Description = "Its 2 Min Nuddles", IsDelete = false });
var items = item.All();
Assert.AreEqual("Maggi", items.FirstOrDefault().Name);
Now, I have put test here. Which gives error like shown in image,
Now if I run code in console application, then code is working fine, code snippet is given below
dynamic item = new Item();
item.Insert(new { Name = "Maggi", Description = "Its 2 Min Nuddles", IsDelete = false });
var result = item.All();
foreach (var i in result)
{
Console.WriteLine(i.Name + i.Description);
}
Console.Read();
Here, code is working and same thing is not working with NUnit Test. Please have a look and help me out. Please let me know if any further information is needed from my side.

Most probable explanation is that you haven't set up your connection string in the test project.
If you are using NUnit, just put it in app.config of your test project.

Solved... There is a issue with NUnit Testing. It was not taking config file pefectly. So, I made two changes. Changes I have done in Project setting.
First change is to change Application Base to bin\debug just give application base as this and then config file to .config to .exe.config and things are up and running. :)