//__________________________________________________________________
void _HYSequencePane::CleanUpSequenceNames (void)
{
bool doSomething = false;
_List namesl;
_AVLList names (&namesl);
for (long k=0; k<speciesIndex.lLength; k++)
{
_String * thisString = (_String*)rowHeaders (speciesIndex.lData[k]);
if (!thisString->IsValidIdentifier(false))
{
BufferToConsole ("Changed ");
StringToConsole(*thisString);
thisString->ConvertToAnIdent(false);
BufferToConsole (" to ");
StringToConsole(*thisString);
NLToConsole();
doSomething = true;
}
_String * testString = new _String (*thisString);
if (!testString)
checkPointer (testString);
long tryThisSuffix = 2;
while (names.Find (testString)>=0)
{
*testString = *thisString & '_' & tryThisSuffix;
tryThisSuffix++;
}
if (tryThisSuffix>2)
{
BufferToConsole ("Changed ");
StringToConsole(*thisString);
BufferToConsole (" to ");
StringToConsole(*testString);
BufferToConsole (" to avoid duplicate identifiers\n");
doSomething = true;
thisString->CopyDynamicString (testString,true);
}
else
DeleteObject (testString);
names.Insert(thisString);
thisString->nInstances++;
}
if (doSomething)
{
SetHeaders (nil,true);
_MarkForUpdate();
}
}
//__________________________________________________________________________________
void SetStatusLineUser (_String s)
{
setvbuf(stdout, NULL, _IONBF, 0);
BufferToConsole("\33[2K\r");
StringToConsole(s);
needExtraNL = true;
}
bool ExpressionCalculator (void)
{
_String data (StringFromConsole(false));
#ifndef __UNIX__
if (terminateExecution) {
return false;
}
BufferToConsole (">");
StringToConsole (data);
BufferToConsole ("\n");
#endif
if (data.sLength == 4) {
_String checkForExit (data);
checkForExit.LoCase();
if (checkForExit == _String ("exit")) {
return false;
}
}
_Formula lhs,
rhs;
_FormulaParsingContext fpc;
long retCode = Parse(&lhs, data, fpc, nil);
if (!terminateExecution) {
if (retCode == HY_FORMULA_EXPRESSION) {
_PMathObj formRes = lhs.Compute();
if (!formRes) {
BufferToConsole ("NULL\n");
} else {
_String * objValue = (_String*)formRes->toStr();
StringToConsole (*objValue);
//BufferToConsole ("\n");
DeleteObject (objValue);
}
} else {
BufferToConsole ("NO RETURN VALUE");
}
}
NLToConsole();
terminateExecution = false;
return true;
}
_String ReturnDialogInput(bool dispPath)
{
if (!dispPath) {
NLToConsole ();
StringToConsole (dialogPrompt);
BufferToConsole (":");
} else {
NLToConsole ();
if (pathNames.lLength) {
StringToConsole(*(_String*)pathNames(pathNames.lLength-1));
} else {
StringToConsole (baseDirectory);
}
StringToConsole (dialogPrompt);
BufferToConsole (":");
}
return StringFromConsole();
}
bool ExpressionCalculator (_String data)
{
//Checking for exit
#ifndef __HYPHYQT__
if (data.sLength == 4) {
_String checkForExit (data);
checkForExit.LoCase();
if (checkForExit == _String ("exit")) {
return false;
}
}
#endif
_Formula lhs,
rhs;
_String errMsg;
_FormulaParsingContext fpc (&errMsg, nil);
long retCode = Parse(&lhs, data, fpc, nil);
if (retCode != HY_FORMULA_FAILED) {
if (retCode == HY_FORMULA_EXPRESSION) {
_PMathObj formRes = lhs.Compute(0,nil,nil,&errMsg);
if (errMsg.sLength) {
WarnError(errMsg);
} else {
_String * objValue = (_String*)formRes->toStr();
StringToConsole(*objValue);
DeleteObject(objValue);
}
} else {
BufferToConsole ("NO RETURN VALUE");
}
} else {
WarnError(errMsg);
}
return true;
}
void TrainModelNN (_String* model, _String* matrix)
{
_String errMsg;
long modelIdx = modelNames.Find(model);
_Parameter verbI;
checkParameter (VerbosityLevelString, verbI, 0.0);
char buffer [128];
if (modelIdx < 0) {
errMsg = *model & " did not refer to an existring model";
} else {
_Variable* boundsMatrix = FetchVar (LocateVarByName (*matrix));
if (boundsMatrix && (boundsMatrix->ObjectClass() == MATRIX)) {
_Matrix * bmatrix = (_Matrix*) boundsMatrix->GetValue ();
if (bmatrix->IsAStringMatrix() && (bmatrix->GetVDim () == 3)) {
_Variable* modelMatrix = LocateVar (modelMatrixIndices.lData[modelIdx]);
_SimpleList modelVariableList;
{
_AVLList mvla (&modelVariableList);
modelMatrix->ScanForVariables (mvla, true);
mvla.ReorderList();
}
if (bmatrix->GetHDim () == modelVariableList.lLength) {
// now map model variables to bounds matrix
_SimpleList variableMap;
_String *myName;
for (long k = 0; k < modelVariableList.lLength; k++) {
myName = ((_FString*)bmatrix->GetFormula(k,0)->Compute())->theString;
long vID = LocateVarByName (*myName);
if (vID < 0) {
break;
}
vID = variableNames.GetXtra (vID);
vID = modelVariableList.Find(vID);
if (vID < 0) {
break;
}
variableMap << vID;
}
if (variableMap.lLength == modelVariableList.lLength) {
_Matrix vBounds (variableMap.lLength,2, false, true);
long k2 = 0;
for (; k2 < variableMap.lLength; k2++) {
_Parameter lb = ((_FString*)bmatrix->GetFormula(k2,1)->Compute())->theString->toNum(),
ub = ((_FString*)bmatrix->GetFormula(k2,2)->Compute())->theString->toNum();
if ( ub>lb || k2) {
vBounds.Store (k2,0,lb);
vBounds.Store (k2,1,ub);
if (ub<=lb && vBounds (k2-1,0) <= vBounds (k2-1,1) && (!CheckEqual(vBounds (k2-1,0),0.0) || !CheckEqual(vBounds (k2-1,1),1.0))) {
break;
}
}
}
if (k2 == modelVariableList.lLength) {
// set up the sampling now
_String fName = ProcessLiteralArgument (&ModelNNFile,nil);
FILE* nnFile = doFileOpen (fName.getStr(), "w");
if (nnFile) {
_Matrix* modelMatrix = (_Matrix*) LocateVar(modelMatrixIndices.lData[modelIdx])->GetValue();
_Parameter mainSteps,
checkSteps,
errorTerm,
loopMax,
hiddenNodes,
absError,
nn1,
nn2;
long fullDimension = modelMatrix->GetHDim() * modelMatrix->GetVDim();
checkParameter (ModelNNTrainingSteps, mainSteps, 10000.0);
checkParameter (ModelNNVerificationSample, checkSteps, 500.0);
checkParameter (ModelNNPrecision, errorTerm, 0.01);
checkParameter (ModelNNTrainingSteps, loopMax, 10);
checkParameter (ModelNNHiddenNodes, hiddenNodes, 5);
checkParameter (ModelNNLearningRate, nn1, .3);
checkParameter (ModelNNPersistenceRate, nn2, .1);
Net** matrixNet = new Net* [fullDimension] ;
for (long i = 0; i < fullDimension; i++) {
checkPointer (matrixNet [i] = new Net (variableMap.lLength,(long)hiddenNodes,1,errorTerm,nn1,nn2,100,200,true));
//matrixNet[i]->verbose = true;
}
checkPointer (matrixNet);
_List tIn,
tOut;
FILE* varSamples = doFileOpen ("variableSamples.out", "w");
fprintf (varSamples, "%s" ,LocateVar(modelVariableList.lData[0])->GetName()->getStr());
for (long vc = 1; vc < modelVariableList.lLength; vc++) {
fprintf (varSamples, ",%s" ,LocateVar(modelVariableList.lData[variableMap.lData[vc]])->GetName()->getStr());
}
fprintf (varSamples, "\n");
for (long itCount = 0; itCount < loopMax; itCount ++) {
if (verbI > 5) {
snprintf (buffer, sizeof(buffer), "\nNeural Network Pass %ld. Building a training set...\n", itCount);
BufferToConsole (buffer);
}
while (tIn.countitems() < mainSteps) {
NNMatrixSampler (0, vBounds, modelVariableList, variableMap, modelMatrix, tIn, tOut);
}
_Matrix inData (mainSteps, variableMap.lLength, false, true);
_Parameter *md = inData.theData;
for (long matrixC = 0; matrixC < mainSteps; matrixC++) {
_Parameter * ed = ((_Matrix*)tIn (matrixC))->theData;
fprintf (varSamples, "\n%g",*ed);
*md = *ed;
ed++;
md++;
for (long entryC = 1; entryC < variableMap.lLength; entryC++, ed++, md++) {
*md = *ed;
fprintf (varSamples, ",%g", *md);
}
}
tIn.Clear();
if (verbI > 5) {
BufferToConsole ( "Done Building Training Set. Training...\n");
}
long lastDone = 0;
for (long cellCount = 0; cellCount < fullDimension; cellCount++) {
Net* thisCell = matrixNet[cellCount];
_Matrix outVector (mainSteps, 1, false, true);
for (long oc = 0; oc < mainSteps; oc++) {
outVector.theData[oc] = ((_Matrix*)tOut(oc))->theData[cellCount];
}
thisCell->studyAll (inData.theData, outVector.theData, mainSteps);
long nowDone = (cellCount+1)*100./fullDimension;
if (nowDone > lastDone) {
snprintf (buffer, sizeof(buffer),"%ld%% done\n", lastDone = nowDone);
BufferToConsole (buffer);
}
}
tOut.Clear();
if (verbI > 5) {
BufferToConsole ( "Done Training. Resampling...\n");
}
_PMathObj tObj = _Constant(0).Time();
_Parameter time1 = tObj->Value(),
time2;
while (tIn.countitems() < checkSteps) {
NNMatrixSampler (0, vBounds, modelVariableList, variableMap, modelMatrix, tIn, tOut);
}
absError = 0.0;
DeleteObject (tObj);
tObj = _Constant(0).Time();
time2 = tObj->Value();
if (verbI > 5) {
snprintf (buffer, sizeof(buffer),"Done Resampling in %g seconds. Computing Error...\n", time2-time1);
BufferToConsole (buffer);
}
_Parameter maxValT,
maxValE;
for (long verCount = 0; verCount < checkSteps; verCount++) {
_Parameter* inData = ((_Matrix*)tIn(verCount))->theData,
* outData = ((_Matrix*)tOut(verCount))->theData;
for (long cellCount = 0; cellCount < fullDimension; cellCount++) {
Net *thisCell = matrixNet[cellCount];
_Parameter estVal = thisCell->eval(inData)[0],
trueVal = outData[cellCount],
localError;
localError = estVal-trueVal;
if (localError < 0) {
localError = -localError;
}
if (absError < localError) {
maxValT = trueVal;
maxValE = estVal;
absError = localError;
}
}
}
DeleteObject (tObj);
tObj = _Constant(0).Time();
time1 = tObj->Value();
DeleteObject (tObj);
if (verbI > 5) {
snprintf (buffer, sizeof(buffer), "Done Error Checking in %g seconds. Got max abs error %g on the pair %g %g\n", time1-time2, absError, maxValT, maxValE);
BufferToConsole (buffer);
}
if (absError <= errorTerm) {
break;
}
}
if (absError > errorTerm) {
ReportWarning (_String("Couldn't achive desired precision in TrainModelNN. Achieved error of ") & absError);
}
fclose (varSamples);
fprintf (nnFile,"{{\n\"%s\"", LocateVar(modelVariableList.lData[0])->GetName()->getStr());
_Matrix newBounds (modelVariableList.lLength, 2, false, true);
if (vBounds(0,0)>vBounds(0,1)) {
newBounds.Store (variableMap.lData[0],0,0.);
newBounds.Store (variableMap.lData[0],1,1.);
} else {
newBounds.Store (variableMap.lData[0],0,vBounds(0,0));
newBounds.Store (variableMap.lData[0],1,vBounds(0,1));
}
for (long varCounter = 1; varCounter < modelVariableList.lLength; varCounter ++) {
fprintf (nnFile,",\n\"%s\"", LocateVar(modelVariableList.lData[varCounter])->GetName()->getStr());
if (vBounds(varCounter,0)>vBounds(varCounter,1)) {
newBounds.Store (variableMap.lData[varCounter],0,0.);
newBounds.Store (variableMap.lData[varCounter],1,1.);
} else {
newBounds.Store (variableMap.lData[varCounter],0,vBounds(varCounter,0));
newBounds.Store (variableMap.lData[varCounter],1,vBounds(varCounter,1));
}
}
fprintf (nnFile,"\n}}\n");
newBounds.toFileStr (nnFile);
for (long i2 = 0; i2 < fullDimension; i2++) {
matrixNet[i2]->save(nnFile);
delete matrixNet [i2];
}
fclose (nnFile);
delete matrixNet;
} else {
errMsg = _String ("Failed to open ") & fName & " for writing";
}
} else {
errMsg = _String ("Invalid variable bounds in row ") & (k2+1) & " of the bounds matrix";
}
} else {
errMsg = *myName & " was not one of the model parameters";
}
} else {
errMsg = *matrix & " must be a have the same number of rows as the number of model parameters";
}
} else {
errMsg = *matrix & " must be a string matrix with 3 columns";
}
} else {
errMsg = *matrix & " was not the identifier of a valid matrix variable";
}
}
if (errMsg.sLength) {
errMsg = errMsg & _String(" in call to TrainModelNN.");
WarnError (errMsg);
}
}
void NLToConsole (void)
{
BufferToConsole ("\n");
}
void StringToConsole (_String & s, _SimpleList *)
{
BufferToConsole ((const char*)s.sData);
}
int main( int argc, char *argv[] )
{
#ifdef __HYPHYMPI__
int rank,
size;
MPI_Init (&argc, &argv);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &size);
_hy_mpi_node_rank = rank;
setParameter (mpiNodeID, (_Parameter)rank);
setParameter (mpiNodeCount, (_Parameter)size);
if (rank == 0)
#endif
gtk_init (&argc, &argv);
/* set up globals */
char curWd[4096];
getcwd (curWd,4096);
_String baseDir (curWd);
baseDir=baseDir&'/';
pathNames&& &baseDir;
baseDirectory = baseDir;
for (long i=1; i<argc;i++)
{
_String thisArg (argv[i]);
if (thisArg.beginswith ("BASEPATH="))
{
baseDirectory = thisArg.Cut(9,-1);
if (baseDirectory.sLength)
{
if (baseDirectory.sData[baseDirectory.sLength-1]!='/')
baseDirectory = baseDirectory&"/";
}
}
else
if (thisArg.beginswith ("USEPATH="))
{
_String baseArgDir (thisArg,8,-1);
errorFileName = baseArgDir & errorFileName;
messageFileName = baseArgDir & messageFileName;
pathNames.Delete (0);
pathNames&& &baseDir;
}
else
if (thisArg.beginswith ("CPU="))
{
#ifdef __MP__
_String cpus = thisArg.Cut(4,-1);
systemCPUCount = cpus.toNum();
if (systemCPUCount<1)
systemCPUCount = 1;
#ifdef __MP2__
pthread_setconcurrency (systemCPUCount+1);
#endif
#endif
}
#ifdef __HYPHYMPI__
else
if (thisArg == _String("MPIOPTIMIZER"))
{
mpiParallelOptimizer = true;
setParameter (mpiNodeCount, 0.0);
}
else
if (thisArg == _String("MPIPARTITIONS"))
{
mpiPartitionOptimizer = true;
setParameter (mpiNodeCount, 0.0);
}
#endif
}
#ifdef __HYPHYMPI__
if (rank == 0)
#endif
{
baseDir = baseDirectory & "GTKResources";
_List scanRes;
ScanDirectoryForFileNames(baseDir,scanRes,false);
if (scanRes.lLength == 0)
{
GtkWidget * noRez = gtk_message_dialog_new (NULL, GTK_DIALOG_MODAL, GTK_MESSAGE_ERROR, GTK_BUTTONS_OK, "HYPHY_GTK was unable to find a required GTKResources directory in %s. Please use BASEPATH= command line option to specify where the installation directory of HyPhy can be found.", baseDirectory.sData);
gtk_dialog_run (GTK_DIALOG (noRez));
gtk_widget_destroy (noRez);
return 1;
}
_String rcPath = baseDir & "/theme/theme.rc";
//printf ("Loading res files from %s\n", rcPath.sData);
gtk_rc_parse (rcPath.sData);
}
GlobalStartup();
#ifdef __HYPHYMPI__
if (rank == 0)
{
#endif
GdkDisplay * defDisplay = gdk_screen_get_display (gdk_screen_get_default());
hSizeCursor = gdk_cursor_new_for_display (defDisplay,GDK_SB_H_DOUBLE_ARROW);
pickUpCursor = gdk_cursor_new_for_display (defDisplay,GDK_TARGET);
dropOffCursor = gdk_cursor_new_for_display (defDisplay,GDK_TCROSS);
screenPContext = gdk_pango_context_get_for_screen (gdk_screen_get_default());
tablePDMenuIcon = (GdkPixbuf*)ProcureIconResource(4020);
/*{
GdkScreen * defD = gdk_screen_get_default();
fontConversionFactor = 72.27 / (gdk_screen_get_height (defD) *25.4 / gdk_screen_get_height_mm(defD));
printf ("Pango conversion factor computed at: %g\n", fontConversionFactor);
}*/
ReadInTemplateFiles ();
hyphyConsoleWindow = new _HYConsoleWindow ("HYPHY Console");
ReadPreferences ();
SetStatusLine ("None","Idle","00:00:00");
while (gtk_events_pending())
gtk_main_iteration();
SetPreferences ();
ReadGeneticCodes ();
ReadModelTemplates ();
ReadTreeProcessors ();
MoveConsoleWindow (consolePositionRectangle);
StringToConsole (hyphyCiteString);
hyphyConsoleWindow->BringToFront();
#ifdef __HYPHYMPI__
{
char statBuffer[1024];
sprintf (statBuffer,"MPI version of HyPhy running on %d nodes (a master and %d compute nodes) in %s mode\n",
size,
size-1,
mpiPartitionOptimizer?"partition":(mpiParallelOptimizer?"rate heterogeneity":"normal"));
BufferToConsole (statBuffer);
}
#endif
g_timeout_add (100,GlobalQueueTimer,nil);
g_timeout_add (1000,progressTimerFunction,nil);
gtk_main ();
WritePreferences();
#ifdef __HYPHYMPI__
}
else // slave node
{
if (mpiParallelOptimizer || mpiPartitionOptimizer)
mpiOptimizerLoop (rank, size);
else
mpiNormalLoop (rank, size, baseDir);
}
#endif
GlobalShutdown();
return 0;
}
void StringToConsole (_String & s)
{
BufferToConsole ((const char*)s.sData);
}
