hillclimber 0.1.6__cp314-cp314t-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl

plumed/_lib/include/plumed/gridtools/KDE.h

@@ -0,0 +1,455 @@
+/* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+   Copyright (c) 2012-2017 The plumed team
+   (see the PEOPLE file at the root of the distribution for a list of names)
+
+   See http://www.plumed.org for more information.
+
+   This file is part of plumed, version 2.
+
+   plumed is free software: you can redistribute it and/or modify
+   it under the terms of the GNU Lesser General Public License as published by
+   the Free Software Foundation, either version 3 of the License, or
+   (at your option) any later version.
+
+   plumed is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public License
+   along with plumed.  If not, see <http://www.gnu.org/licenses/>.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
+#ifndef __PLUMED_gridtools_KDE_h
+#define __PLUMED_gridtools_KDE_h
+
+#include "ActionWithGrid.h"
+#include "SumOfKernels.h"
+#include "../core/PlumedMain.h"
+#include "../core/ActionSet.h"
+#include "../core/PbcAction.h"
+#include "../tools/HistogramBead.h"
+#include "../core/ParallelTaskManager.h"
+#include "../tools/SwitchingFunction.h"
+#include "../tools/Matrix.h"
+
+namespace PLMD {
+namespace gridtools {
+
+template <class K, class P, class G>
+class KDEHelper {
+public:
+  G g;
+  bool fixed_width;
+  std::size_t maxkernels;
+  SumOfKernels<K, P> kernelsum;
+  std::vector<unsigned> nneigh;
+  std::vector<std::size_t> nkernels_per_point;
+  std::vector<std::size_t> kernels_for_gridpoint;
+  static void registerKeywords( Keywords& keys );
+  static void read( KDEHelper<K,P,G>& func,
+                    ActionWithArguments* action,
+                    const std::vector<Value*>& args,
+                    GridCoordinatesObject& gridobject,
+                    std::vector<std::size_t>& shape,
+                    function::FunctionOptions& options );
+  static void readKernelParameters( std::string& value, ActionWithArguments* action, const std::string& outlab, bool rerequestargs );
+  static void addArgument( const std::string& value, ActionWithArguments* action );
+  static void setupGridBounds( KDEHelper<K,P,G>& func, const Tensor& box, GridCoordinatesObject& gridobject, const std::vector<Value*>& args, Value* myval );
+  static void transferParamsToKernel( const std::vector<double>& argval, KDEHelper<K,P,G>& func, GridCoordinatesObject& gridobject, bool updateNeighborsOnEachKernel, std::size_t nkernels, unsigned kval, K& kp );
+  static void transferKernels( KDEHelper<K,P,G>& func, const std::vector<Value*>& args, GridCoordinatesObject& gridobject );
+};
+
+template <class K, class P, class G>
+void KDEHelper<K,P,G>::registerKeywords( Keywords& keys ) {
+  SumOfKernels<K,P>::registerKeywords( keys );
+  G::registerKeywords( keys );
+}
+
+template <class K, class P, class G>
+void KDEHelper<K,P,G>::read( KDEHelper<K,P,G>& func,
+                             ActionWithArguments* action,
+                             const std::vector<Value*>& args,
+                             GridCoordinatesObject& gridobject,
+                             std::vector<std::size_t>& shape,
+                             function::FunctionOptions& options ) {
+  func.nneigh.resize( args.size() );
+  SumOfKernels<K,P>::read( func.kernelsum, action, args, options );
+  G::readBandwidthAndHeight( func.kernelsum.params, action );
+  for(unsigned i=1; i<action->getNumberOfArguments(); ++i) {
+    if( (action->getPntrToArgument(0))->getNumberOfValues()==0 && (action->getPntrToArgument(i))->isConstant() ) {
+      continue;
+    }
+
+    if( (action->getPntrToArgument(0))->getNumberOfValues()!=1 || (action->getPntrToArgument(0))->getNumberOfValues()!=1 ) {
+      if( (action->getPntrToArgument(0))->getRank()!=(action->getPntrToArgument(i))->getRank() ) {
+        action->error("mismatch between ranks of input actions");
+      }
+      for(unsigned j=0; j<(action->getPntrToArgument(0))->getRank(); ++j) {
+        if( (action->getPntrToArgument(0))->getShape()[j]!=(action->getPntrToArgument(i))->getShape()[j] ) {
+          action->error("mismatch between shapes of input actions");
+        }
+      }
+    }
+  }
+  G::readGridParameters( func.g, action, gridobject, shape );
+}
+
+template <class K, class P, class G>
+void KDEHelper<K,P,G>::setupGridBounds( KDEHelper<K,P,G>& func, const Tensor& box, GridCoordinatesObject& gridobject, const std::vector<Value*>& args, Value* myval ) {
+  // Setup the grid boundaries on first step
+  G::setupGridBounds( func.g, box, gridobject, args, myval );
+  // Check if the bandwidth changes during the simulation
+  func.fixed_width = false;
+  if( K::bandwidthIsConstant( gridobject.getDimension(), args ) && K::bandwidthsAllSame( gridobject.getDimension(), args ) ) {
+    K myk;
+    std::vector<double> myargs( args.size() );
+    for(unsigned j=0; j<args.size(); ++j) {
+      myargs[j] = args[j]->get(0);
+    }
+    K::setKernelAndCheckHeight( myk, gridobject.getDimension(), myargs );
+    G::getDiscreteSupport( func.g, func.kernelsum.params, myk, func.nneigh, gridobject );
+    func.fixed_width = true;
+  }
+  if( gridobject.getGridType()=="fibonacci" ) {
+    return;
+  }
+  // Set the periodicity of the parameters
+  for(unsigned i=0; i<gridobject.getDimension(); ++i) {
+    P::setArgumentDomain( i, func.kernelsum.params, gridobject.getGridSpacing()[i], gridobject.isPeriodic(i), gridobject.getMin()[i], gridobject.getMax()[i] );
+  }
+}
+
+template <class K, class P, class G>
+void KDEHelper<K,P,G>::readKernelParameters( std::string& value, ActionWithArguments* action, const std::string& outlab, bool rerequestargs ) {
+  std::size_t dot = value.find_first_of('.');
+  ActionWithValue* av = action->plumed.getActionSet().selectWithLabel<ActionWithValue*>( value.substr(0,dot) );
+  if( !av ) {
+    std::string matstr, vals = "VALUES=" + value;
+    if( (action->getPntrToArgument(0))->getRank()==2 ) {
+      std::string nr, nc;
+      Tools::convert( (action->getPntrToArgument(0))->getShape()[0], nr );
+      Tools::convert( (action->getPntrToArgument(0))->getShape()[1], nc );
+      matstr = " NROWS=" + nr + " NCOLS=" + nc;
+    }
+    for(unsigned i=1; i<(action->getPntrToArgument(0))->getNumberOfValues(); ++i) {
+      vals += "," + value;
+    }
+    action->plumed.readInputWords( Tools::getWords(action->getLabel() + outlab + ": CONSTANT " + vals + matstr ), false );
+    value = action->getLabel() + outlab;
+  } else {
+    Value* myval;
+    if( dot!=std::string::npos ) {
+      myval = av->copyOutput( value );
+    } else {
+      if( av->getNumberOfComponents()>1 ) {
+        action->error("problem reading argument " + value );
+      }
+      myval = av->copyOutput(0);
+    }
+    if( myval->getRank()==0 ) {
+      std::string nvals;
+      if( (action->getPntrToArgument(0))->getRank()==2 ) {
+        std::string nr, nc;
+        Tools::convert( (action->getPntrToArgument(0))->getShape()[0], nr );
+        Tools::convert( (action->getPntrToArgument(0))->getShape()[1], nc );
+        nvals = nr + "," + nc;
+      } else {
+        Tools::convert( (action->getPntrToArgument(0))->getNumberOfValues(), nvals );
+      }
+      action->plumed.readInputWords( Tools::getWords(action->getLabel() + outlab + "_ones: ONES SIZE=" + nvals ), false );
+      action->plumed.readInputWords( Tools::getWords(action->getLabel() + outlab + ": CUSTOM ARG=" + action->getLabel() + outlab + "_ones," + value ), false );
+      value = action->getLabel() + outlab;
+    }
+  }
+  if( !rerequestargs ) {
+    return;
+  }
+  KDEHelper<K,P,G>::addArgument( value, action );
+}
+
+template <class K, class P, class G>
+void KDEHelper<K,P,G>::addArgument( const std::string& value, ActionWithArguments* action ) {
+  std::size_t dot = value.find_first_of(".");
+  ActionWithValue* av = action->plumed.getActionSet().selectWithLabel<ActionWithValue*>( value.substr(0,dot) );
+  plumed_assert( av );
+  std::vector<Value*> args( action->getArguments() );
+  args.push_back( av->copyOutput(0) );
+  action->requestArguments( args );
+}
+
+template <class K, class P, class G>
+void KDEHelper<K,P,G>::transferParamsToKernel( const std::vector<double>& argval, KDEHelper<K,P,G>& func, GridCoordinatesObject& gridobject, bool updateNeighborsOnEachKernel, std::size_t nkernels, unsigned kval, K& kp ) {
+  // This sets the kernel parameters for the Kth kernel and checks that we want
+  // to consider it
+  if( !K::setKernelAndCheckHeight( kp, gridobject.getDimension(), argval ) ) {
+    return;
+  }
+  // If the widths of each kernel are not all the same then get the discrete support
+  if( updateNeighborsOnEachKernel ) {
+    G::getDiscreteSupport( func.g, func.kernelsum.params, kp, func.nneigh, gridobject );
+  }
+
+  // Now get the grid points for this particular kernel
+  unsigned num_neigh;
+  std::vector<unsigned> neighbors;
+  G::getNeighbors( func.kernelsum.params, kp, gridobject, func.nneigh, num_neigh, neighbors );
+
+  // And transfer the neighbor information to the holders
+  for(unsigned j=0; j<num_neigh; ++j) {
+    func.kernels_for_gridpoint[ neighbors[j]*nkernels + func.nkernels_per_point[neighbors[j]] ] = kval;
+    func.nkernels_per_point[ neighbors[j] ]++;
+  }
+}
+
+template <class K, class P, class G>
+void KDEHelper<K,P,G>::transferKernels( KDEHelper<K,P,G>& func, const std::vector<Value*>& args, GridCoordinatesObject& gridobject ) {
+  // Resize the kernel sum if we need to
+  // Number of kernels is determined based on sparsity pattern of matrix input as matrix of heights
+  std::size_t nkernels = args[args.size()-1]->getNumberOfStoredValues();
+  if( func.kernelsum.kernelParams.size()!=nkernels ) {
+    func.kernelsum.kernelParams.resize( nkernels );
+  }
+  // And resize the grid counters if we need to
+  std::size_t ngp = gridobject.getNumberOfPoints();
+  if( func.nkernels_per_point.size()!=ngp ) {
+    func.nkernels_per_point.resize( ngp );
+    func.kernels_for_gridpoint.resize( ngp*nkernels );
+  }
+  std::fill( func.nkernels_per_point.begin(), func.nkernels_per_point.end(), 0 );
+
+  bool updateNeighborsOnEachKernel = !func.fixed_width;
+  if( !func.fixed_width && K::bandwidthsAllSame( gridobject.getDimension(), args ) ) {
+    G::getDiscreteSupport( func.g, func.kernelsum.params, func.kernelsum.kernelParams[0], func.nneigh, gridobject );
+    updateNeighborsOnEachKernel = false;
+  }
+
+  std::vector<double> argval( args.size() );
+  if( args[args.size()-1]->getRank()==2 ) {
+    const unsigned nc    = args[args.size()-1]->getShape()[1];
+    const unsigned nrows = args[args.size()-1]->getShape()[0];
+    const unsigned ncs   = args[args.size()-1]->getNumberOfColumns();
+    for(unsigned i=0; i<nrows; ++i) {
+      unsigned ncols = args[args.size()-1]->getRowLength(i);
+      for(unsigned k=0; k<args.size(); ++k) {
+        plumed_massert( args[k]->isConstant() || ncols==args[k]->getRowLength(i), "all input matrices must have same sparsity pattern" );
+      }
+      for(unsigned j=0; j<ncols; ++j) {
+        unsigned jind = args[args.size()-1]->getRowIndex( i, j );
+        for(unsigned k=0; k<args.size(); ++k) {
+          if( jind==args[k]->getRowIndex( i, j ) ) {
+            argval[k] = args[k]->get( i*ncs+ j, false );
+          } else {
+            argval[k] = args[k]->get( i*nc + jind );
+          }
+        }
+        KDEHelper<K,P,G>::transferParamsToKernel( argval, func, gridobject, updateNeighborsOnEachKernel, nkernels, i*ncs+j, func.kernelsum.kernelParams[i*ncs+j] );
+      }
+    }
+  } else {
+    for(unsigned i=0; i<nkernels; ++i) {
+      // Transfer the kernel parameters to local vector of doubles
+      for(unsigned j=0; j<args.size(); ++j) {
+        argval[j] = args[j]->get(i,false);
+      }
+      KDEHelper<K,P,G>::transferParamsToKernel( argval, func, gridobject, updateNeighborsOnEachKernel, nkernels, i, func.kernelsum.kernelParams[i] );
+    }
+  }
+  // Get the maximum number of kernels for any given grid point (used for resizing derivatives)
+  func.maxkernels = 0;
+  for(unsigned i=0; i<ngp; ++i) {
+    if( func.nkernels_per_point[i]>func.maxkernels ) {
+      func.maxkernels = func.nkernels_per_point[i];
+    }
+  }
+}
+
+template <class K, class P, class G>
+class KDE : public ActionWithGrid {
+public:
+  using input_type = KDEHelper<K, P, G>;
+  using PTM = ParallelTaskManager<KDE<K,P,G>>;
+private:
+  bool firststep;
+/// The parallel task manager
+  PTM taskmanager;
+  GridCoordinatesObject gridobject;
+public:
+  static void registerKeywords( Keywords& keys );
+  explicit KDE(const ActionOptions&ao);
+  std::vector<std::string> getGridCoordinateNames() const override ;
+  const GridCoordinatesObject& getGridCoordinatesObject() const override ;
+  unsigned getNumberOfDerivatives() override;
+  int checkTaskIsActive( const unsigned& itask ) const override ;
+  void prepare() override ;
+  void calculate() override ;
+  void getInputData( std::vector<double>& inputdata ) const override ;
+  static void performTask( std::size_t task_index,
+                           const KDEHelper<K, P, G>& actiondata,
+                           ParallelActionsInput& input,
+                           ParallelActionsOutput& output );
+  void applyNonZeroRankForces( std::vector<double>& outforces ) override ;
+  static int getNumberOfValuesPerTask( std::size_t task_index,
+                                       const KDEHelper<K, P, G>& actiondata );
+  static void getForceIndices( std::size_t task_index,
+                               std::size_t colno,
+                               std::size_t ntotal_force,
+                               const KDEHelper<K, P, G>& actiondata,
+                               const ParallelActionsInput& input,
+                               ForceIndexHolder force_indices );
+};
+
+template <class K, class P, class G>
+void KDE<K,P,G>::registerKeywords( Keywords& keys ) {
+  ActionWithGrid::registerKeywords( keys );
+  keys.addInputKeyword("compulsory","ARG","scalar/vector/matrix","the label for the value that should be used to construct the histogram");
+  KDEHelper<K,P,G>::registerKeywords( keys );
+  // Keywords for spherical KDE
+  keys.add("hidden","MASKED_INPUT_ALLOWED","turns on that you are allowed to use masked inputs ");
+  keys.setValueDescription("grid","a function on a grid that was obtained by doing a Kernel Density Estimation using the input arguments");
+  if( keys.getDisplayName()!="SPHERICAL_KDE" ) {
+    keys.setDisplayName("KDE");
+  }
+  PTM::registerKeywords( keys );
+}
+
+template <class K, class P, class G>
+KDE<K,P,G>::KDE(const ActionOptions&ao):
+  Action(ao),
+  ActionWithGrid(ao),
+  firststep(true),
+  taskmanager(this) {
+
+  std::vector<std::size_t> shape( getNumberOfArguments() );
+  unsigned numberOfKernels=getPntrToArgument(0)->getNumberOfValues();
+  for(unsigned i=1; i<shape.size(); ++i) {
+    if( numberOfKernels!=getPntrToArgument(i)->getNumberOfValues() ) {
+      error("mismatch between numbers of values in input arguments");
+    }
+  }
+
+  function::FunctionOptions foptions;
+  KDEHelper<K,P,G>::read( taskmanager.getActionInput(), this, getArguments(), gridobject, shape, foptions );
+  addValueWithDerivatives( shape );
+  setNotPeriodic();
+  getPntrToComponent(0)->setDerivativeIsZeroWhenValueIsZero();
+}
+
+template <class K, class P, class G>
+unsigned KDE<K,P,G>::getNumberOfDerivatives() {
+  return gridobject.getDimension();
+}
+
+template <class K, class P, class G>
+std::vector<std::string> KDE<K,P,G>::getGridCoordinateNames() const {
+  std::vector<std::string> names( gridobject.getDimension() );
+  for(unsigned i=0; i<names.size(); ++i) {
+    names[i] = getPntrToArgument(i)->getName();
+  }
+  return names;
+}
+
+template <class K, class P, class G>
+const GridCoordinatesObject& KDE<K,P,G>::getGridCoordinatesObject() const {
+  return gridobject;
+}
+
+template <class K, class P, class G>
+int KDE<K,P,G>::checkTaskIsActive( const unsigned& itask ) const {
+  if( taskmanager.getActionInput().nkernels_per_point[itask]>0 ) {
+    return 1;
+  }
+  return -1;
+}
+
+template <class K, class P, class G>
+void KDE<K,P,G>::prepare() {
+  ActionWithVector::prepare();
+  std::size_t nkernels = getPntrToArgument(0)->getNumberOfValues();
+  for(unsigned i=1; i<getNumberOfArguments(); ++i) {
+    Value* myarg = getPntrToArgument(i);
+    if( myarg->getNumberOfValues()!=nkernels ) {
+      if( myarg->isConstant() && myarg->getNumberOfValues()==1 ) {
+        myarg->reshapeConstantValue( getPntrToArgument(0)->getShape() );
+      } else {
+        plumed_merror("found mismatched numbers of arguments in input");
+      }
+    }
+  }
+}
+
+template <class K, class P, class G>
+void KDE<K,P,G>::calculate() {
+  if( firststep ) {
+    PbcAction* bv = plumed.getActionSet().template selectWithLabel<PbcAction*>("Box");
+    KDEHelper<K,P,G>::setupGridBounds( taskmanager.getActionInput(), bv->getPbc().getBox(), gridobject, getArguments(), getPntrToComponent(0) );
+    firststep=false;
+  }
+  KDEHelper<K,P,G>::transferKernels( taskmanager.getActionInput(), getArguments(), gridobject );
+  taskmanager.setupParallelTaskManager( getNumberOfArguments()*taskmanager.getActionInput().maxkernels, getNumberOfForceDerivatives() );
+  taskmanager.runAllTasks();
+}
+
+template <class K, class P, class G>
+void KDE<K,P,G>::getInputData( std::vector<double>& inputdata ) const {
+  std::size_t ndim = gridobject.getDimension();
+  std::size_t nstored = getConstPntrToComponent(0)->getNumberOfStoredValues();
+  std::vector<double> pos( ndim );
+  if( inputdata.size()!=nstored*ndim ) {
+    inputdata.resize( ndim*nstored );
+  }
+
+  for(unsigned i=0; i<nstored; ++i) {
+    gridobject.getGridPointCoordinates( i, pos );
+    for(unsigned j=0; j<ndim; ++j) {
+      inputdata[ i*ndim + j ] = pos[j];
+    }
+  }
+}
+
+template <class K, class P, class G>
+void KDE<K,P,G>::performTask( std::size_t task_index,
+                              const KDEHelper<K, P, G>& actiondata,
+                              ParallelActionsInput& input,
+                              ParallelActionsOutput& output ) {
+  std::size_t ndim = actiondata.nneigh.size();
+  SumOfKernels<K,P>::calc( View<const std::size_t>( actiondata.kernels_for_gridpoint.data() + task_index*input.argstarts[1], actiondata.nkernels_per_point[task_index] ),
+                           actiondata.kernelsum,
+                           View<const double>(input.inputdata + task_index*actiondata.nneigh.size(),ndim),
+                           View<double>(output.values.data(), 1),
+                           View<double>(output.values.data()+1, ndim),
+                           View<double>(output.derivatives.data(),actiondata.maxkernels*input.nargs) );
+
+}
+
+template <class K, class P, class G>
+void KDE<K,P,G>::applyNonZeroRankForces( std::vector<double>& outforces ) {
+  taskmanager.applyForces( outforces );
+}
+
+template <class K, class P, class G>
+int KDE<K,P,G>::getNumberOfValuesPerTask( std::size_t task_index,
+    const KDEHelper<K, P, G>& actiondata ) {
+  return 1;
+}
+
+template <class K, class P, class G>
+void KDE<K,P,G>::getForceIndices( std::size_t task_index,
+                                  std::size_t colno,
+                                  std::size_t ntotal_force,
+                                  const KDEHelper<K, P, G>& actiondata,
+                                  const ParallelActionsInput& input,
+                                  ForceIndexHolder force_indices ) {
+  force_indices.threadsafe_derivatives_end[0] = 0;
+  std::size_t nparams = K::getNumberOfParameters( actiondata.kernelsum.kernelParams[0] );
+  View<const std::size_t> kernellist( actiondata.kernels_for_gridpoint.data() + task_index*input.argstarts[1], actiondata.nkernels_per_point[task_index] );
+  for(unsigned i=0; i<kernellist.size(); ++i) {
+    for(unsigned j=0; j<nparams; ++j) {
+      force_indices.indices[0][i*nparams+j] = input.argstarts[j] + kernellist[i];
+    }
+  }
+  force_indices.tot_indices[0] = kernellist.size()*nparams;
+}
+
+}
+}
+#endif

plumed/_lib/include/plumed/gridtools/RDF.h

@@ -0,0 +1,40 @@
+/* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+   Copyright (c) 2013-2020 The plumed team
+   (see the PEOPLE file at the root of the distribution for a list of names)
+
+   See http://www.plumed.org for more information.
+
+   This file is part of plumed, version 2.
+
+   plumed is free software: you can redistribute it and/or modify
+   it under the terms of the GNU Lesser General Public License as published by
+   the Free Software Foundation, either version 3 of the License, or
+   (at your option) any later version.
+
+   plumed is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public License
+   along with plumed.  If not, see <http://www.gnu.org/licenses/>.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
+#ifndef __PLUMED_gridtools_RDF_h
+#define __PLUMED_gridtools_RDF_h
+
+#include "../core/ActionShortcut.h"
+
+namespace PLMD {
+namespace gridtools {
+
+class RDF : public ActionShortcut {
+public:
+  static void createX2ReferenceObject( const std::string& lab,  const std::string& grid_setup, const bool& calc_dens, const bool& no_average, ActionShortcut* action );
+  static void registerKeywords( Keywords& keys );
+  static void getDistanceMatrixShape( const std::string& lab, ActionShortcut* action, std::vector<std::string>& shape_str );
+  explicit RDF(const ActionOptions&ao);
+};
+
+}
+}
+#endif