-
-
Notifications
You must be signed in to change notification settings - Fork 44
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
1 parent
9e231e2
commit b6ad50b
Showing
2 changed files
with
402 additions
and
44 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,336 @@ | ||
| //+------------------------------------------------------------------+ | ||
| //| CNN.mqh | | ||
| //| Copyright 2023, Omega Joctan | | ||
| //| https://www.mql5.com/en/users/omegajoctan | | ||
| //+------------------------------------------------------------------+ | ||
| #property copyright "Copyright 2023, Omega Joctan" | ||
| #property link "https://www.mql5.com/en/users/omegajoctan" | ||
| //+------------------------------------------------------------------+ | ||
| //| Class for loading the convolution Neural network class in onnx | | ||
| //| format, tested on keras built model. \ | ||
| //+------------------------------------------------------------------+ | ||
|
|
||
| #include <MALE5\Tensors.mqh> | ||
| #define UNDEFINED_REPLACE 1 | ||
|
|
||
| class CConvNet | ||
| { | ||
|
|
||
| protected: | ||
|
|
||
| bool initialized; | ||
| long onnx_handle; | ||
| void PrintTypeInfo(const long num,const string layer,const OnnxTypeInfo& type_info); | ||
| long inputs[], outputs[]; | ||
|
|
||
| void replace(long &arr[]) { for (uint i=0; i<arr.Size(); i++) if (arr[i] <= -1) arr[i] = UNDEFINED_REPLACE; } | ||
| string ConvertTime(double seconds); | ||
|
|
||
| public: | ||
| CConvNet(void); | ||
| ~CConvNet(void); | ||
|
|
||
| bool Init(const uchar &onnx_buff[], ulong flags=ONNX_DEFAULT); | ||
| bool Init(string onnx_filename, uint flags=ONNX_DEFAULT); | ||
|
|
||
| virtual int predict_bin(const matrix &x, const vector &classes_in_data); //predicts classes | useful for live trading | ||
| virtual vector predict_bin(C3DTensor &x_tensor, const vector &classes_in_data); //predict classes for the entire data | useful for training purposes | ||
| virtual vector predict_proba(const matrix &x); //Predicts probabilities | Useful for live trading | ||
|
|
||
| double predict(const matrix &x); //for a regression problem | ||
| }; | ||
| //+------------------------------------------------------------------+ | ||
| //| | | ||
| //+------------------------------------------------------------------+ | ||
| CConvNet::CConvNet(void) | ||
| { | ||
|
|
||
| } | ||
| //+------------------------------------------------------------------+ | ||
| //| | | ||
| //+------------------------------------------------------------------+ | ||
| CConvNet::~CConvNet(void) | ||
| { | ||
|
|
||
| } | ||
| //+------------------------------------------------------------------+ | ||
| //| | | ||
| //+------------------------------------------------------------------+ | ||
|
|
||
| bool CConvNet::Init(string onnx_filename, uint flags=ONNX_DEFAULT) | ||
| { | ||
| Print("Initilaizing ONNX model..."); | ||
|
|
||
| onnx_handle = OnnxCreate(onnx_filename, flags); | ||
|
|
||
| //--- since not all sizes defined in the input tensor we must set them explicitly | ||
| //--- first index - batch size, second index - series size, third index - number of series (only Close) | ||
|
|
||
| OnnxTypeInfo type_info; //Getting onnx information for Reference In case you forgot what the loaded ONNX is all about | ||
|
|
||
| long input_count=OnnxGetInputCount(onnx_handle); | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| Print("model has ",input_count," input(s)"); | ||
|
|
||
| for(long i=0; i<input_count; i++) | ||
| { | ||
| string input_name=OnnxGetInputName(onnx_handle,i); | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| Print(i," input name is ",input_name); | ||
|
|
||
| if(OnnxGetInputTypeInfo(onnx_handle,i,type_info)) | ||
| { | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| PrintTypeInfo(i,"input",type_info); | ||
| ArrayCopy(inputs, type_info.tensor.dimensions); | ||
| } | ||
| } | ||
|
|
||
| long output_count=OnnxGetOutputCount(onnx_handle); | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| Print("model has ",output_count," output(s)"); | ||
| for(long i=0; i<output_count; i++) | ||
| { | ||
| string output_name=OnnxGetOutputName(onnx_handle,i); | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| Print(i," output name is ",output_name); | ||
|
|
||
| if(OnnxGetOutputTypeInfo(onnx_handle,i,type_info)) | ||
| { | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| PrintTypeInfo(i,"output",type_info); | ||
| ArrayCopy(outputs, type_info.tensor.dimensions); | ||
| } | ||
| } | ||
|
|
||
| //--- | ||
|
|
||
| replace(inputs); | ||
| replace(outputs); | ||
|
|
||
| if (!OnnxSetInputShape(onnx_handle, 0, inputs)) //Giving the Onnx handle the input shape | ||
| { | ||
| printf("Failed to set the input shape Err=%d",GetLastError()); | ||
| return false; | ||
| } | ||
|
|
||
| if (!OnnxSetOutputShape(onnx_handle, 0, outputs)) //giving the onnx handle the output shape | ||
| { | ||
| printf("Failed to set the Output shape Err=%d",GetLastError()); | ||
| return false; | ||
| } | ||
|
|
||
| initialized = true; | ||
| Print("ONNX model Initialized"); | ||
|
|
||
| return true; | ||
| } | ||
| //+------------------------------------------------------------------+ | ||
| //| | | ||
| //+------------------------------------------------------------------+ | ||
| bool CConvNet::Init(const uchar &onnx_buff[], ulong flags=ONNX_DEFAULT) | ||
| { | ||
| Print("Initilaizing ONNX model..."); | ||
|
|
||
| onnx_handle = OnnxCreateFromBuffer(onnx_buff, flags); //creating onnx handle buffer | rUN DEGUG MODE during debug mode | ||
|
|
||
|
|
||
| //--- since not all sizes defined in the input tensor we must set them explicitly | ||
| //--- first index - batch size, second index - series size, third index - number of series (only Close) | ||
|
|
||
| OnnxTypeInfo type_info; //Getting onnx information for Reference In case you forgot what the loaded ONNX is all about | ||
|
|
||
| long input_count=OnnxGetInputCount(onnx_handle); | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| Print("model has ",input_count," input(s)"); | ||
|
|
||
| for(long i=0; i<input_count; i++) | ||
| { | ||
| string input_name=OnnxGetInputName(onnx_handle,i); | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| Print(i," input name is ",input_name); | ||
|
|
||
| if(OnnxGetInputTypeInfo(onnx_handle,i,type_info)) | ||
| { | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| PrintTypeInfo(i,"input",type_info); | ||
| ArrayCopy(inputs, type_info.tensor.dimensions); | ||
| } | ||
| } | ||
|
|
||
| long output_count=OnnxGetOutputCount(onnx_handle); | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| Print("model has ",output_count," output(s)"); | ||
| for(long i=0; i<output_count; i++) | ||
| { | ||
| string output_name=OnnxGetOutputName(onnx_handle,i); | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| Print(i," output name is ",output_name); | ||
|
|
||
| if(OnnxGetOutputTypeInfo(onnx_handle,i,type_info)) | ||
| { | ||
| if (MQLInfoInteger(MQL_DEBUG)) | ||
| PrintTypeInfo(i,"output",type_info); | ||
| ArrayCopy(outputs, type_info.tensor.dimensions); | ||
| } | ||
| } | ||
|
|
||
| //--- | ||
|
|
||
| replace(inputs); | ||
| replace(outputs); | ||
|
|
||
| if (!OnnxSetInputShape(onnx_handle, 0, inputs)) //Giving the Onnx handle the input shape | ||
| { | ||
| printf("Failed to set the input shape Err=%d",GetLastError()); | ||
| return false; | ||
| } | ||
|
|
||
| if (!OnnxSetOutputShape(onnx_handle, 0, outputs)) //giving the onnx handle the output shape | ||
| { | ||
| printf("Failed to set the Output shape Err=%d",GetLastError()); | ||
| return false; | ||
| } | ||
|
|
||
| initialized = true; | ||
|
|
||
| Print("ONNX model Initialized"); | ||
| return true; | ||
| } | ||
| //+------------------------------------------------------------------+ | ||
| //| | | ||
| //+------------------------------------------------------------------+ | ||
| void CConvNet::PrintTypeInfo(const long num,const string layer,const OnnxTypeInfo& type_info) | ||
| { | ||
| Print(" type ",EnumToString(type_info.type)); | ||
| Print(" data type ",EnumToString(type_info.type)); | ||
|
|
||
| if(type_info.tensor.dimensions.Size()>0) | ||
| { | ||
| bool dim_defined=(type_info.tensor.dimensions[0]>0); | ||
| string dimensions=IntegerToString(type_info.tensor.dimensions[0]); | ||
|
|
||
|
|
||
| for(long n=1; n<type_info.tensor.dimensions.Size(); n++) | ||
| { | ||
| if(type_info.tensor.dimensions[n]<=0) | ||
| dim_defined=false; | ||
| dimensions+=", "; | ||
| dimensions+=IntegerToString(type_info.tensor.dimensions[n]); | ||
| } | ||
| Print(" shape [",dimensions,"]"); | ||
| //--- not all dimensions defined | ||
| if(!dim_defined) | ||
| PrintFormat(" %I64d %s shape must be defined explicitly before model inference",num,layer); | ||
| //--- reduce shape | ||
| uint reduced=0; | ||
| long dims[]; | ||
| for(long n=0; n<type_info.tensor.dimensions.Size(); n++) | ||
| { | ||
| long dimension=type_info.tensor.dimensions[n]; | ||
| //--- replace undefined dimension | ||
| if(dimension<=0) | ||
| dimension=UNDEFINED_REPLACE; | ||
| //--- 1 can be reduced | ||
| if(dimension>1) | ||
| { | ||
| ArrayResize(dims,reduced+1); | ||
| dims[reduced++]=dimension; | ||
| } | ||
| } | ||
| //--- all dimensions assumed 1 | ||
| if(reduced==0) | ||
| { | ||
| ArrayResize(dims,1); | ||
| dims[reduced++]=1; | ||
| } | ||
| //--- shape was reduced | ||
| if(reduced<type_info.tensor.dimensions.Size()) | ||
| { | ||
| dimensions=IntegerToString(dims[0]); | ||
| for(long n=1; n<dims.Size(); n++) | ||
| { | ||
| dimensions+=", "; | ||
| dimensions+=IntegerToString(dims[n]); | ||
| } | ||
| string sentence=""; | ||
| if(!dim_defined) | ||
| sentence=" if undefined dimension set to "+(string)UNDEFINED_REPLACE; | ||
| PrintFormat(" shape of %s data can be reduced to [%s]%s",layer,dimensions,sentence); | ||
| } | ||
| } | ||
| else | ||
| PrintFormat("no dimensions defined for %I64d %s",num,layer); | ||
| } | ||
| //+------------------------------------------------------------------+ | ||
| //| | | ||
| //+------------------------------------------------------------------+ | ||
| int CConvNet::predict_bin(const matrix &x, const vector &classes_in_data) | ||
| { | ||
| if (!this.initialized) | ||
| { | ||
| printf("%s The model is not initialized yet to make predictions | call Init function first",__FUNCTION__); | ||
| return 0; | ||
| } | ||
|
|
||
| //--- | ||
|
|
||
| matrixf x_float; | ||
| x_float.Assign(x); | ||
|
|
||
| vector output_data(this.outputs[outputs.Size()-1]); | ||
|
|
||
| if (!OnnxRun(onnx_handle, ONNX_DATA_TYPE_FLOAT, x_float, output_data)) | ||
| { | ||
| printf("Failed to get predictions from Onnx err %d",GetLastError()); | ||
| return false; | ||
| } | ||
|
|
||
| return (int)classes_in_data[output_data.ArgMax()]; //Return the class with highest probability | ||
| } | ||
| //+------------------------------------------------------------------+ | ||
| //| | | ||
| //+------------------------------------------------------------------+ | ||
| vector CConvNet::predict_proba(const matrix &x) | ||
| { | ||
|
|
||
| if (!this.initialized) | ||
| { | ||
| printf("%s The model is not initialized yet to make predictions | call Init function first",__FUNCTION__); | ||
| vector empty = {}; | ||
| return empty; | ||
| } | ||
|
|
||
| //--- | ||
|
|
||
| matrixf x_float; | ||
| x_float.Assign(x); | ||
|
|
||
| vector proba(this.outputs[outputs.Size()-1]); | ||
|
|
||
| if (!OnnxRun(onnx_handle, ONNX_DATA_TYPE_FLOAT, x_float, proba)) | ||
| { | ||
| printf("Failed to get predictions from Onnx err %d",GetLastError()); | ||
| return proba; | ||
| } | ||
|
|
||
| return proba; //Return the class with highest probability | ||
| } | ||
| //+------------------------------------------------------------------+ | ||
| //| When given a matrix for timeseries data collected it provides | | ||
| //| a scalar binary value which is a prediction | | ||
| //+------------------------------------------------------------------+ | ||
| vector CConvNet::predict_bin(C3DTensor ×eries_tensor, const vector &classes_in_data) | ||
| { | ||
| vector preds(timeseries_tensor.SIZE); | ||
| for (uint i=0; i<timeseries_tensor.SIZE; i++) | ||
| { | ||
| matrix x = timeseries_tensor.Get(i); | ||
| preds[i] = predict_bin(x, classes_in_data); | ||
| } | ||
|
|
||
| return preds; | ||
| } | ||
| //+------------------------------------------------------------------+ | ||
| //| | | ||
| //+------------------------------------------------------------------+ |
Oops, something went wrong.