code
stringlengths
31
1.05M
apis
list
extract_api
stringlengths
97
1.91M
import numpy as np import matplotlib as mpl from matplotlib.colors import to_rgb, to_rgba from numpy.testing import assert_array_equal LINE_PROPS = [ "alpha", "color", "linewidth", "linestyle", "xydata", "zorder", ] COLLECTION_PROPS = [ "alpha", "edgecolor", "facecolor", "fill", "hatch", "linestyle", "linewidth", "paths", "zorder", ] BAR_PROPS = [ "alpha", "edgecolor", "facecolor", "fill", "hatch", "height", "linestyle", "linewidth", "xy", "zorder", ] def assert_colors_equal(a, b, check_alpha=True): def handle_array(x): if isinstance(x, np.ndarray): if x.ndim > 1: x = np.unique(x, axis=0).squeeze() if x.ndim > 1: raise ValueError("Color arrays must be 1 dimensional") return x a = handle_array(a) b = handle_array(b) f = to_rgba if check_alpha else to_rgb assert f(a) == f(b) def assert_artists_equal(list1, list2, properties): assert len(list1) == len(list2) for a1, a2 in zip(list1, list2): prop1 = a1.properties() prop2 = a2.properties() for key in properties: v1 = prop1[key] v2 = prop2[key] if key == "paths": for p1, p2 in zip(v1, v2): assert_array_equal(p1.vertices, p2.vertices) assert_array_equal(p1.codes, p2.codes) elif isinstance(v1, np.ndarray): assert_array_equal(v1, v2) elif key == "color": v1 = mpl.colors.to_rgba(v1) v2 = mpl.colors.to_rgba(v2) assert v1 == v2 else: assert v1 == v2 def assert_legends_equal(leg1, leg2): assert leg1.get_title().get_text() == leg2.get_title().get_text() for t1, t2 in zip(leg1.get_texts(), leg2.get_texts()): assert t1.get_text() == t2.get_text() assert_artists_equal( leg1.get_patches(), leg2.get_patches(), BAR_PROPS, ) assert_artists_equal( leg1.get_lines(), leg2.get_lines(), LINE_PROPS, ) def assert_plots_equal(ax1, ax2, labels=True): assert_artists_equal(ax1.patches, ax2.patches, BAR_PROPS) assert_artists_equal(ax1.lines, ax2.lines, LINE_PROPS) poly1 = ax1.findobj(mpl.collections.PolyCollection) poly2 = ax2.findobj(mpl.collections.PolyCollection) assert_artists_equal(poly1, poly2, COLLECTION_PROPS) if labels: assert ax1.get_xlabel() == ax2.get_xlabel() assert ax1.get_ylabel() == ax2.get_ylabel()
[ "matplotlib.colors.to_rgba", "numpy.unique", "numpy.testing.assert_array_equal" ]
[((1359, 1403), 'numpy.testing.assert_array_equal', 'assert_array_equal', (['p1.vertices', 'p2.vertices'], {}), '(p1.vertices, p2.vertices)\n', (1377, 1403), False, 'from numpy.testing import assert_array_equal\n'), ((1424, 1462), 'numpy.testing.assert_array_equal', 'assert_array_equal', (['p1.codes', 'p2.codes'], {}), '(p1.codes, p2.codes)\n', (1442, 1462), False, 'from numpy.testing import assert_array_equal\n'), ((1524, 1550), 'numpy.testing.assert_array_equal', 'assert_array_equal', (['v1', 'v2'], {}), '(v1, v2)\n', (1542, 1550), False, 'from numpy.testing import assert_array_equal\n'), ((723, 743), 'numpy.unique', 'np.unique', (['x'], {'axis': '(0)'}), '(x, axis=0)\n', (732, 743), True, 'import numpy as np\n'), ((1605, 1627), 'matplotlib.colors.to_rgba', 'mpl.colors.to_rgba', (['v1'], {}), '(v1)\n', (1623, 1627), True, 'import matplotlib as mpl\n'), ((1649, 1671), 'matplotlib.colors.to_rgba', 'mpl.colors.to_rgba', (['v2'], {}), '(v2)\n', (1667, 1671), True, 'import matplotlib as mpl\n')]
#!/usr/bin/env python # # Copyright (C) 2009 <NAME> # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. from __future__ import with_statement # for python 2.5 PKG = 'posedetectiondb' # this package name NAME = 'GatherDetectionResults' import roslib; roslib.load_manifest(PKG) import os, sys, time, string, threading from optparse import OptionParser import numpy # nice to be able to explicitly call some functions from numpy import * import rospy import posedetection_msgs.msg from openravepy import * class VisibilityModel(metaclass.AutoReloader): def __init__(self,measurements=None,filename=None,symmetricplane=None,kinbodyfile=None): if measurements is not None: self.rawmeasurements = measurements if symmetricplane is not None: dists = dot(symmetricplane[0:3],transpose(self.rawmeasurements))+symmetricplane[3] self.rawmeasurements = r_[self.rawmeasurements,self.rawmeasurements-dot(reshape(2.0*dists,(dists.shape[0],1)),reshape(symmetricplane[0:3],(1,3)))] self.trimesh = None if kinbodyfile is not None: self.env = Environment() self.orobj = self.env.ReadKinBodyXMLFile(kinbodyfile) if self.orobj is None: raise ValueError('failed to open %s openrave file'%kinbodyfile) self.env.AddKinBody(self.orobj) self.trimesh = self.env.Triangulate(self.orobj) self.measurements = self.rawmeasurements def CreateReducedModel(self,bandwidth=0.04,bandthresh=0.01,neighthresh=0.01,showdata=False,savefile=None): self.measurements,indices = self.Prune(self.rawmeasurements,100,neighthresh**2,1) uniformpoints,dists,pointscale = self.UniformlySampleSpace(bandwidth,bandthresh) if showdata: from enthought.mayavi import mlab mlab.figure(1,fgcolor=(0,0,0), bgcolor=(1,1,1)) src = mlab.pipeline.scalar_field(dists) mlab.pipeline.iso_surface(src,contours=[0.01],opacity=0.1) mlab.pipeline.volume(mlab.pipeline.scalar_field(dists*500)) v = pointscale[0]*self.trimesh.vertices+pointscale[1] mlab.triangular_mesh(v[:,0],v[:,1],v[:,2],self.trimesh.indices,color=(0,0,0.5)) if savefile is None: savefile = self.getfilename() print('saving measurements to %s'%savefile) mkdir_recursive(os.path.split(savefile)[0]) savetxt(savefile,uniformpoints,'%f') def getfilename(self): return os.path.join(self.env.GetHomeDirectory(),'kinbody.'+self.orobj.GetKinematicsGeometryHash(),'visibility.txt') def UniformlySampleSpace(self,bandwidth,bandthresh,delta=0.02): maxradius = sqrt(max(sum(self.measurements**2,1))) nsteps = floor(maxradius/delta) X,Y,Z = mgrid[-nsteps:nsteps,-nsteps:nsteps,-nsteps:nsteps] allpoints = c_[X.flat,Y.flat,Z.flat]*delta/bandwidth sampleinds = flatnonzero(sum(allpoints**2,1)<(maxradius/bandwidth)**2) samplepoints = allpoints[sampleinds,:] kdtree = pyANN.KDTree(self.measurements/bandwidth) sampledists = zeros(samplepoints.shape[0]) goodpoints = [] for i in xrange(samplepoints.shape[0]): neighs,dists,kball = kdtree.kFRSearchArray(samplepoints[i:(i+1),:],5.0**2,32,0.0001) sampledists[i] = sum(exp(-dists[neighs>=0])) uniformpoints = samplepoints[sampledists>bandthresh,:]*bandwidth alldists = zeros(prod(X.shape)) alldists[sampleinds] = sampledists return uniformpoints,reshape(alldists,X.shape),array((1.0/delta,nsteps)) def Prune(self,rawposes, nsize, thresh2, neighsize,giveupiters=100): """rawposes is Nx7""" iter = 1 poses = array(rawposes) indices = range(poses.shape[0]) N = poses.shape[0] nochange=0 while N > nsize: ind = numpy.random.randint(N) g = poses[ind,:] # check g's neighbors d = sum((poses[0:N,:] - tile(g, (N,1)))**2,1) neigh = sum(d < thresh2) if neigh > neighsize: # move to the last pose and resize poses[ind,:] = poses[N-1,:] indices[ind] = indices[N-1] nochange=0 N -= 1 nochange += 1 iter += 1 if iter > 5000 or nochange > giveupiters: break return poses[0:N,:],indices[0:N] class OpenRAVEVisualizer(metaclass.AutoReloader): def __init__(self,kinbodyfile,automaticadd=True,measurementsfilename=None): self.automaticadd = automaticadd self.orenv = Environment() self.orenv.SetViewer('qtcoin') self.orobj = self.orenv.ReadKinBodyXMLFile(kinbodyfile) if self.orobj is None: raise ValueError('failed to open %s openrave file'%kinbodyfile) self.orenv.AddKinBody(self.orobj) self.objab = self.orobj.ComputeAABB() self.Tcamera = None self.lck = threading.Lock() self.camerahandle = None self.measurementhandles = [] self.measurements = [] if measurementsfilename is not None: f = open(measurementsfilename,'r') for l in f.readlines(): m = array([string.atof(s) for s in string.split(l)]) self.drawmeasurement(m) self.measurements.append(m) f.close() self.sub_objdet = rospy.Subscriber("ObjectDetection", posedetection_msgs.msg.ObjectDetection,self.objdetcb, queue_size=1) rospy.init_node(NAME, anonymous=True)#,disable_signals=False) def __del__(self): self.sub_objdet.unregister() self.orenv.Destroy() def objdetcb(self,msg): newcamerahandle = None if len(msg.objects) > 0: q = msg.objects[0].pose.orientation t = msg.objects[0].pose.position with self.lck: self.Tcamera = linalg.inv(matrixFromPose([q.w,q.x,q.y,q.z,t.x,t.y,t.z])) # draw in openrave environment sx = 0.02 sy = 0.02 camlocalpts = array(((sx,sy,0.05),(-sx,sy,0.05),(-sx,-sy,0.05),(sx,-sy,0.05),(sx,sy,0.05),(0,0,0),(-sx,sy,0.05),(0,0,0),(-sx,-sy,0.05),(0,0,0),(sx,-sy,0.05))) campts = dot(self.Tcamera,r_[transpose(camlocalpts),ones((1,camlocalpts.shape[0]))]) self.camerahandle = self.orenv.drawlinestrip(transpose(campts[0:3,:]),2,array([0,0,1])) if self.automaticadd: self.addmeasurement() def addmeasurement(self): self.lck.acquire() Tcamera = self.Tcamera self.lck.release() dist = dot(Tcamera[0:3,2],self.objab.pos()-Tcamera[0:3,3:4]) m = Tcamera[0:3,2]*dist self.drawmeasurement(m) self.measurements.append(m) print('num measurements %d'%len(self.measurements)) def drawmeasurement(self,m): dist = sqrt(sum(m**2)) dir = m/dist p = self.objab.pos()-dist*dir self.measurementhandles.append(self.orenv.drawlinestrip(transpose(c_[p-0.02*dir,p+0.02*dir]),4,array((1,min(1,dist/1.0),0)))) def savemeasurements(self,filename): self.lck.acquire() print('saving measurements to %s'%filename) savetxt(filename,self.measurements,'%f') self.lck.release() if __name__=='__main__': parser = OptionParser(description='Gather object detection transformations and filter and display them.') parser.add_option('--kinbodyfile', action="store",type='string',dest='kinbodyfile', help='OpenRAVE object file that represents the incoming object pose. Updates are show in the openrave window') parser.add_option('-s','--single', action="store_true",dest='single',default=False, help='If set, will wait for user input in order to add a measurement') parser.add_option('-f','--savefile', action="store",dest="filename", help='If specified, will save all recorded measurements to this file at exit time') parser.add_option('-m','--measurements', action="store",dest="measurements",default=None, help='If specified, will start with the current measurements file') (options, args) = parser.parse_args() if not options.kinbodyfile: print('Error: Need to specify an openrave kinbody file') sys.exit(1) visualizer = OpenRAVEVisualizer(options.kinbodyfile,measurementsfilename=options.measurements,automaticadd=not options.single) while True: cmd = raw_input('Enter command (q-quit and save,c-capture): '); if cmd == 'q': break elif cmd == 'c' and options.single: print('adding measurement') visualizer.addmeasurement() else: print('bad command',cmd) if options.filename: visualizer.savemeasurements(options.filename) def test(): "rosrun posedetectiondb GatherDetectionResults.py --kinbodyfile=scenes/cereal_frootloops.kinbody.xml -f test.txt ObjectDetection:=/CerealDetection" import GatherDetectionResults self = GatherDetectionResults.VisibilityModel(measurements=loadtxt('uroncha_raw.txt'),symmetricplane=array([1.0,0,0,0]),kinbodyfile='scenes/uroncha.kinbody.xml') self = GatherDetectionResults.VisibilityModel(measurements=loadtxt('peachjuice_raw.txt'),kinbodyfile='scenes/peachjuice.kinbody.xml') self.CreateReducedModel(bandwidth=0.03,bandthresh=0.02)
[ "enthought.mayavi.mlab.pipeline.scalar_field", "threading.Lock", "rospy.init_node", "string.split", "enthought.mayavi.mlab.figure", "optparse.OptionParser", "string.atof", "roslib.load_manifest", "os.path.split", "numpy.random.randint", "enthought.mayavi.mlab.pipeline.iso_surface", "sys.exit",...
[((746, 771), 'roslib.load_manifest', 'roslib.load_manifest', (['PKG'], {}), '(PKG)\n', (766, 771), False, 'import roslib\n'), ((7884, 7985), 'optparse.OptionParser', 'OptionParser', ([], {'description': '"""Gather object detection transformations and filter and display them."""'}), "(description=\n 'Gather object detection transformations and filter and display them.')\n", (7896, 7985), False, 'from optparse import OptionParser\n'), ((5511, 5527), 'threading.Lock', 'threading.Lock', ([], {}), '()\n', (5525, 5527), False, 'import os, sys, time, string, threading\n'), ((5959, 6067), 'rospy.Subscriber', 'rospy.Subscriber', (['"""ObjectDetection"""', 'posedetection_msgs.msg.ObjectDetection', 'self.objdetcb'], {'queue_size': '(1)'}), "('ObjectDetection', posedetection_msgs.msg.ObjectDetection,\n self.objdetcb, queue_size=1)\n", (5975, 6067), False, 'import rospy\n'), ((6071, 6108), 'rospy.init_node', 'rospy.init_node', (['NAME'], {'anonymous': '(True)'}), '(NAME, anonymous=True)\n', (6086, 6108), False, 'import rospy\n'), ((8981, 8992), 'sys.exit', 'sys.exit', (['(1)'], {}), '(1)\n', (8989, 8992), False, 'import os, sys, time, string, threading\n'), ((2339, 2391), 'enthought.mayavi.mlab.figure', 'mlab.figure', (['(1)'], {'fgcolor': '(0, 0, 0)', 'bgcolor': '(1, 1, 1)'}), '(1, fgcolor=(0, 0, 0), bgcolor=(1, 1, 1))\n', (2350, 2391), False, 'from enthought.mayavi import mlab\n'), ((2405, 2438), 'enthought.mayavi.mlab.pipeline.scalar_field', 'mlab.pipeline.scalar_field', (['dists'], {}), '(dists)\n', (2431, 2438), False, 'from enthought.mayavi import mlab\n'), ((2451, 2511), 'enthought.mayavi.mlab.pipeline.iso_surface', 'mlab.pipeline.iso_surface', (['src'], {'contours': '[0.01]', 'opacity': '(0.1)'}), '(src, contours=[0.01], opacity=0.1)\n', (2476, 2511), False, 'from enthought.mayavi import mlab\n'), ((2660, 2753), 'enthought.mayavi.mlab.triangular_mesh', 'mlab.triangular_mesh', (['v[:, 0]', 'v[:, 1]', 'v[:, 2]', 'self.trimesh.indices'], {'color': '(0, 0, 0.5)'}), '(v[:, 0], v[:, 1], v[:, 2], self.trimesh.indices, color\n =(0, 0, 0.5))\n', (2680, 2753), False, 'from enthought.mayavi import mlab\n'), ((4389, 4412), 'numpy.random.randint', 'numpy.random.randint', (['N'], {}), '(N)\n', (4409, 4412), False, 'import numpy\n'), ((2543, 2582), 'enthought.mayavi.mlab.pipeline.scalar_field', 'mlab.pipeline.scalar_field', (['(dists * 500)'], {}), '(dists * 500)\n', (2569, 2582), False, 'from enthought.mayavi import mlab\n'), ((2887, 2910), 'os.path.split', 'os.path.split', (['savefile'], {}), '(savefile)\n', (2900, 2910), False, 'import os, sys, time, string, threading\n'), ((5784, 5798), 'string.atof', 'string.atof', (['s'], {}), '(s)\n', (5795, 5798), False, 'import os, sys, time, string, threading\n'), ((5808, 5823), 'string.split', 'string.split', (['l'], {}), '(l)\n', (5820, 5823), False, 'import os, sys, time, string, threading\n')]
import numpy as np import tensorflow as tf def encoder_linear(input_shape, latent_dim): """Fully connected linear encoder.""" return tf.keras.Sequential( [ tf.keras.layers.InputLayer(input_shape=input_shape), tf.keras.layers.Flatten(), tf.keras.layers.Dense(latent_dim, activation=None), ], ) def decoder_linear(latent_dim, output_shape): """Fully connected linear decoder.""" return tf.keras.Sequential( [ tf.keras.layers.InputLayer(input_shape=latent_dim), tf.keras.layers.Dense(np.prod(output_shape), activation=None), tf.keras.layers.Reshape(output_shape), ], ) def encoder_fc(input_shape, latent_dim): """Fully connected linear encoder.""" return tf.keras.Sequential( [ tf.keras.layers.InputLayer(input_shape=input_shape), tf.keras.layers.Flatten(), tf.keras.layers.Dense(4096, activation=tf.nn.relu), tf.keras.layers.Dense(1024, activation=tf.nn.relu), tf.keras.layers.Dense(latent_dim, activation=None), ], ) def decoder_fc(latent_dim, output_shape): """Fully connected linear decoder.""" return tf.keras.Sequential( [ tf.keras.layers.InputLayer(input_shape=latent_dim), tf.keras.layers.Dense(1028, activation=tf.nn.relu), tf.keras.layers.Dense(4096, activation=tf.nn.relu), tf.keras.layers.Dense(np.prod(output_shape), activation=None), tf.keras.layers.Reshape(output_shape), ], ) def encoder_conv_64x64(channels, latent_dim): return tf.keras.Sequential( [ tf.keras.layers.InputLayer(input_shape=(64, 64, channels)), # (64, 64, channels) -> (32, 32, 32) tf.keras.layers.Conv2D(32, 5, 2, "same"), tf.keras.layers.BatchNormalization(), tf.keras.layers.LeakyReLU(), # (32, 32, 32) -> (16, 16, 64) tf.keras.layers.Conv2D(64, 3, 2, "same"), tf.keras.layers.BatchNormalization(), tf.keras.layers.LeakyReLU(), # (16, 16, 64) -> (8, 8, 128) tf.keras.layers.Conv2D(128, 3, 2, "same"), tf.keras.layers.BatchNormalization(), tf.keras.layers.LeakyReLU(), # (16, 16, 64) -> (latent_dim, ) tf.keras.layers.Flatten(), tf.keras.layers.Dense(latent_dim, activation=None), ], ) def decoder_conv_64x64(latent_dim, channels): return tf.keras.Sequential( [ tf.keras.layers.InputLayer(input_shape=(latent_dim, )), # (latent_dim, ) -> (8, 8, 128) tf.keras.layers.Dense(8 * 8 * 128, activation=None), tf.keras.layers.Reshape([8, 8, 128]), # (8, 8, 128) -> (16, 16, 64) tf.keras.layers.Conv2DTranspose(64, 3, 2, "same"), tf.keras.layers.BatchNormalization(), tf.keras.layers.LeakyReLU(), # (16, 16, 64) -> (32, 32, 32) tf.keras.layers.Conv2DTranspose(32, 3, 2, "same"), tf.keras.layers.BatchNormalization(), tf.keras.layers.LeakyReLU(), # (32, 32, 32) -> (64, 64, channels) tf.keras.layers.Conv2DTranspose(channels, 5, 2, "same"), ], )
[ "numpy.prod", "tensorflow.keras.layers.Reshape", "tensorflow.keras.layers.Conv2D", "tensorflow.keras.layers.Conv2DTranspose", "tensorflow.keras.layers.LeakyReLU", "tensorflow.keras.layers.BatchNormalization", "tensorflow.keras.layers.Dense", "tensorflow.keras.layers.Flatten", "tensorflow.keras.layer...
[((194, 245), 'tensorflow.keras.layers.InputLayer', 'tf.keras.layers.InputLayer', ([], {'input_shape': 'input_shape'}), '(input_shape=input_shape)\n', (220, 245), True, 'import tensorflow as tf\n'), ((260, 285), 'tensorflow.keras.layers.Flatten', 'tf.keras.layers.Flatten', ([], {}), '()\n', (283, 285), True, 'import tensorflow as tf\n'), ((300, 350), 'tensorflow.keras.layers.Dense', 'tf.keras.layers.Dense', (['latent_dim'], {'activation': 'None'}), '(latent_dim, activation=None)\n', (321, 350), True, 'import tensorflow as tf\n'), ((522, 572), 'tensorflow.keras.layers.InputLayer', 'tf.keras.layers.InputLayer', ([], {'input_shape': 'latent_dim'}), '(input_shape=latent_dim)\n', (548, 572), True, 'import tensorflow as tf\n'), ((663, 700), 'tensorflow.keras.layers.Reshape', 'tf.keras.layers.Reshape', (['output_shape'], {}), '(output_shape)\n', (686, 700), True, 'import tensorflow as tf\n'), ((867, 918), 'tensorflow.keras.layers.InputLayer', 'tf.keras.layers.InputLayer', ([], {'input_shape': 'input_shape'}), '(input_shape=input_shape)\n', (893, 918), True, 'import tensorflow as tf\n'), ((933, 958), 'tensorflow.keras.layers.Flatten', 'tf.keras.layers.Flatten', ([], {}), '()\n', (956, 958), True, 'import tensorflow as tf\n'), ((973, 1023), 'tensorflow.keras.layers.Dense', 'tf.keras.layers.Dense', (['(4096)'], {'activation': 'tf.nn.relu'}), '(4096, activation=tf.nn.relu)\n', (994, 1023), True, 'import tensorflow as tf\n'), ((1038, 1088), 'tensorflow.keras.layers.Dense', 'tf.keras.layers.Dense', (['(1024)'], {'activation': 'tf.nn.relu'}), '(1024, activation=tf.nn.relu)\n', (1059, 1088), True, 'import tensorflow as tf\n'), ((1103, 1153), 'tensorflow.keras.layers.Dense', 'tf.keras.layers.Dense', (['latent_dim'], {'activation': 'None'}), '(latent_dim, activation=None)\n', (1124, 1153), True, 'import tensorflow as tf\n'), ((1321, 1371), 'tensorflow.keras.layers.InputLayer', 'tf.keras.layers.InputLayer', ([], {'input_shape': 'latent_dim'}), '(input_shape=latent_dim)\n', (1347, 1371), True, 'import tensorflow as tf\n'), ((1386, 1436), 'tensorflow.keras.layers.Dense', 'tf.keras.layers.Dense', (['(1028)'], {'activation': 'tf.nn.relu'}), '(1028, activation=tf.nn.relu)\n', (1407, 1436), True, 'import tensorflow as tf\n'), ((1451, 1501), 'tensorflow.keras.layers.Dense', 'tf.keras.layers.Dense', (['(4096)'], {'activation': 'tf.nn.relu'}), '(4096, activation=tf.nn.relu)\n', (1472, 1501), True, 'import tensorflow as tf\n'), ((1592, 1629), 'tensorflow.keras.layers.Reshape', 'tf.keras.layers.Reshape', (['output_shape'], {}), '(output_shape)\n', (1615, 1629), True, 'import tensorflow as tf\n'), ((1758, 1816), 'tensorflow.keras.layers.InputLayer', 'tf.keras.layers.InputLayer', ([], {'input_shape': '(64, 64, channels)'}), '(input_shape=(64, 64, channels))\n', (1784, 1816), True, 'import tensorflow as tf\n'), ((1881, 1921), 'tensorflow.keras.layers.Conv2D', 'tf.keras.layers.Conv2D', (['(32)', '(5)', '(2)', '"""same"""'], {}), "(32, 5, 2, 'same')\n", (1903, 1921), True, 'import tensorflow as tf\n'), ((1936, 1972), 'tensorflow.keras.layers.BatchNormalization', 'tf.keras.layers.BatchNormalization', ([], {}), '()\n', (1970, 1972), True, 'import tensorflow as tf\n'), ((1987, 2014), 'tensorflow.keras.layers.LeakyReLU', 'tf.keras.layers.LeakyReLU', ([], {}), '()\n', (2012, 2014), True, 'import tensorflow as tf\n'), ((2073, 2113), 'tensorflow.keras.layers.Conv2D', 'tf.keras.layers.Conv2D', (['(64)', '(3)', '(2)', '"""same"""'], {}), "(64, 3, 2, 'same')\n", (2095, 2113), True, 'import tensorflow as tf\n'), ((2128, 2164), 'tensorflow.keras.layers.BatchNormalization', 'tf.keras.layers.BatchNormalization', ([], {}), '()\n', (2162, 2164), True, 'import tensorflow as tf\n'), ((2179, 2206), 'tensorflow.keras.layers.LeakyReLU', 'tf.keras.layers.LeakyReLU', ([], {}), '()\n', (2204, 2206), True, 'import tensorflow as tf\n'), ((2264, 2305), 'tensorflow.keras.layers.Conv2D', 'tf.keras.layers.Conv2D', (['(128)', '(3)', '(2)', '"""same"""'], {}), "(128, 3, 2, 'same')\n", (2286, 2305), True, 'import tensorflow as tf\n'), ((2320, 2356), 'tensorflow.keras.layers.BatchNormalization', 'tf.keras.layers.BatchNormalization', ([], {}), '()\n', (2354, 2356), True, 'import tensorflow as tf\n'), ((2371, 2398), 'tensorflow.keras.layers.LeakyReLU', 'tf.keras.layers.LeakyReLU', ([], {}), '()\n', (2396, 2398), True, 'import tensorflow as tf\n'), ((2459, 2484), 'tensorflow.keras.layers.Flatten', 'tf.keras.layers.Flatten', ([], {}), '()\n', (2482, 2484), True, 'import tensorflow as tf\n'), ((2499, 2549), 'tensorflow.keras.layers.Dense', 'tf.keras.layers.Dense', (['latent_dim'], {'activation': 'None'}), '(latent_dim, activation=None)\n', (2520, 2549), True, 'import tensorflow as tf\n'), ((2678, 2731), 'tensorflow.keras.layers.InputLayer', 'tf.keras.layers.InputLayer', ([], {'input_shape': '(latent_dim,)'}), '(input_shape=(latent_dim,))\n', (2704, 2731), True, 'import tensorflow as tf\n'), ((2792, 2843), 'tensorflow.keras.layers.Dense', 'tf.keras.layers.Dense', (['(8 * 8 * 128)'], {'activation': 'None'}), '(8 * 8 * 128, activation=None)\n', (2813, 2843), True, 'import tensorflow as tf\n'), ((2858, 2894), 'tensorflow.keras.layers.Reshape', 'tf.keras.layers.Reshape', (['[8, 8, 128]'], {}), '([8, 8, 128])\n', (2881, 2894), True, 'import tensorflow as tf\n'), ((2952, 3001), 'tensorflow.keras.layers.Conv2DTranspose', 'tf.keras.layers.Conv2DTranspose', (['(64)', '(3)', '(2)', '"""same"""'], {}), "(64, 3, 2, 'same')\n", (2983, 3001), True, 'import tensorflow as tf\n'), ((3016, 3052), 'tensorflow.keras.layers.BatchNormalization', 'tf.keras.layers.BatchNormalization', ([], {}), '()\n', (3050, 3052), True, 'import tensorflow as tf\n'), ((3067, 3094), 'tensorflow.keras.layers.LeakyReLU', 'tf.keras.layers.LeakyReLU', ([], {}), '()\n', (3092, 3094), True, 'import tensorflow as tf\n'), ((3153, 3202), 'tensorflow.keras.layers.Conv2DTranspose', 'tf.keras.layers.Conv2DTranspose', (['(32)', '(3)', '(2)', '"""same"""'], {}), "(32, 3, 2, 'same')\n", (3184, 3202), True, 'import tensorflow as tf\n'), ((3217, 3253), 'tensorflow.keras.layers.BatchNormalization', 'tf.keras.layers.BatchNormalization', ([], {}), '()\n', (3251, 3253), True, 'import tensorflow as tf\n'), ((3268, 3295), 'tensorflow.keras.layers.LeakyReLU', 'tf.keras.layers.LeakyReLU', ([], {}), '()\n', (3293, 3295), True, 'import tensorflow as tf\n'), ((3360, 3415), 'tensorflow.keras.layers.Conv2DTranspose', 'tf.keras.layers.Conv2DTranspose', (['channels', '(5)', '(2)', '"""same"""'], {}), "(channels, 5, 2, 'same')\n", (3391, 3415), True, 'import tensorflow as tf\n'), ((609, 630), 'numpy.prod', 'np.prod', (['output_shape'], {}), '(output_shape)\n', (616, 630), True, 'import numpy as np\n'), ((1538, 1559), 'numpy.prod', 'np.prod', (['output_shape'], {}), '(output_shape)\n', (1545, 1559), True, 'import numpy as np\n')]
#!/usr/bin/env python import argparse import logging import math import sys import time import random import os import json import numpy as np import pickle import six import threading import pdb from tqdm import tqdm import torch import torch.nn as nn from pytorch_pretrained_bert import BertTokenizer, OpenAIGPTTokenizer, GPT2Tokenizer, TransfoXLTokenizer import data_handler as dh from model.mmseq2seq_model import MMSeq2SeqModel from model.multimodal_encoder import MMEncoder from model.lstm_encoder import LSTMEncoder from model.hlstm_encoder import HLSTMEncoder from model.hlstm_decoder import HLSTMDecoder def fetch_batch(dh, data, index, separate_caption, result): result.append(dh.make_batch(data, index, separate_caption=separate_caption, pretrained_elmo=args.pretrained_elmo, pretrained_bert=args.pretrained_bert, bert_tokenizer=bert_tokenizer, pretrained_all=args.pretrained_all, bert_model=args.bert_model, concat_his=args.concat_his)) # Evaluation routine def evaluate(model, data, indices, parallel=False): start_time = time.time() eval_loss = 0. eval_num_words = 0 model.eval() with torch.no_grad(): # fetch the first batch batch = [dh.make_batch(data, indices[0], separate_caption=args.separate_caption, pretrained_elmo=args.pretrained_elmo, pretrained_bert=args.pretrained_bert, bert_tokenizer=bert_tokenizer, pretrained_all=args.pretrained_all, bert_model=args.bert_model, concat_his=args.concat_his)] # evaluation loop it = tqdm(six.moves.range(len(indices)), desc="evaluation", ncols=0) for j in it: #if args.separate_caption: # x_batch, h_batch, q_batch, a_batch_in, a_batch_out, c_batch = batch.pop() #else: # x_batch, h_batch, q_batch, a_batch_in, a_batch_out = batch.pop() b = batch.pop() if j < len(indices)-1: prefetch = threading.Thread(target=fetch_batch, args=([dh, data, indices[j+1], args.separate_caption, batch])) prefetch.start() # propagate for training x = [torch.from_numpy(x) for x in b[0]] if args.concat_his: h = [torch.from_numpy(h_i) for h_i in b[1]] else: h = [[torch.from_numpy(h) for h in hb] for hb in b[1]] q = [torch.from_numpy(q) for q in b[2]] ai = [torch.from_numpy(ai) for ai in b[3]] ao = [torch.from_numpy(ao) for ao in b[4]] if args.separate_caption: c = [torch.from_numpy(c) for c in b[5]] else: c = None if args.pretrained_elmo or args.pretrained_bert: if args.pretrained_all: context_q, context_h, context_ai = b[-3:] else: context_q = b[-1] context_h = None context_ai = None else: context_q = None context_h = None context_ai = None if args.exclude_video: x = None if parallel: _, _, loss = model.module.loss(x, h, q, ai, ao, c, context_q, context_h, context_ai) else: _, _, loss = model.loss(x, h, q, ai, ao, c, context_q, context_h, context_ai) num_words = sum([len(s) for s in ao]) eval_loss += loss.cpu().data.numpy() * num_words eval_num_words += num_words prefetch.join() model.train() wall_time = time.time() - start_time return math.exp(eval_loss/eval_num_words), wall_time ################################## # main if __name__ =="__main__": parser = argparse.ArgumentParser() parser.add_argument('--gpu', '-g', default=0, type=int, help='GPU ID (negative value indicates CPU)') # train, dev and test data parser.add_argument('--vocabfile', default='', type=str, help='Vocabulary file (.json)') parser.add_argument('--dictmap', default='', type=str, help='Dict id-map file (.json)') parser.add_argument('--fea-type', nargs='+', type=str, help='Image feature files (.pkl)') parser.add_argument('--train-path', default='', type=str, help='Path to training feature files') parser.add_argument('--train-set', default='', type=str, help='Filename of train data') parser.add_argument('--valid-path', default='', type=str, help='Path to validation feature files') parser.add_argument('--valid-set', default='', type=str, help='Filename of validation data') parser.add_argument('--include-caption', action='store_true', help='Include caption in the history') parser.add_argument('--separate-caption', default=False, type=bool, help='') parser.add_argument('--exclude-video', action='store_true', help='') parser.add_argument('--pretrained-word-emb', default=None, type=str, help='') parser.add_argument('--pretrained-weights', default=None, type=str, help='') parser.add_argument('--pretrained-elmo', default=False, type=int, help='') parser.add_argument('--elmo-num-outputs', default=1, type=int, help='') parser.add_argument('--finetune-elmo', default=False, type=int, help='') parser.add_argument('--pretrained-bert', default=False, type=int, help='') parser.add_argument('--bert-model', default='bert-base-uncased', type=str, help='') parser.add_argument('--finetune-bert', default=False, type=int, help='') parser.add_argument('--add-word-emb', default=True, type=int, help='') parser.add_argument('--pretrained-all', default=True, type=int, help='') parser.add_argument('--concat-his', default=False, type=int, help='') # model parameters parser.add_argument('--model', '-m', default='', type=str, help='Attention model to be output') # multimodal encoder parameters parser.add_argument('--enc-psize', '-p', nargs='+', type=int, help='Number of projection layer units') parser.add_argument('--enc-hsize', '-u', nargs='+', type=int, help='Number of hidden units') parser.add_argument('--att-size', '-a', default=100, type=int, help='Number of attention layer units') parser.add_argument('--mout-size', default=100, type=int, help='Number of output layer units') parser.add_argument('--mm-att', default='baseline', type=str, help="") parser.add_argument('--mm-fusioning', default='baseline', type=str, help="") parser.add_argument('--mm-att-hops', default=1, type=int, help='') parser.add_argument('--caption-mm-att', action='store_true', help='') # input (question/caption) encoder parameters parser.add_argument('--embed-size', default=200, type=int, help='Word embedding size') parser.add_argument('--in-enc-layers', default=2, type=int, help='Number of input encoder layers') parser.add_argument('--in-enc-hsize', default=200, type=int, help='Number of input encoder hidden layer units') parser.add_argument('--q-att', default=None, type=str, help='') parser.add_argument('--c-att', default=None, type=str, help='') parser.add_argument('--rnn-type', default='lstm', type=str, help='') parser.add_argument('--caption-states-att', default=False, type=bool, help='') # history (QA pairs) encoder parameters parser.add_argument('--hist-enc-layers', nargs='+', type=int, help='Number of history encoder layers') parser.add_argument('--hist-enc-hsize', default=200, type=int, help='History embedding size') parser.add_argument('--hist-out-size', default=200, type=int, help='History embedding size') parser.add_argument('--ft-fusioning', default='baseline', type=str, help='Fusioning fetures between images and text') parser.add_argument('--caption-mm-fusion-out-size', default=-1, type=int, help='') # response (answer) decoder parameters parser.add_argument('--dec-layers', default=2, type=int, help='Number of decoder layers') parser.add_argument('--dec-psize', '-P', default=200, type=int, help='Number of decoder projection layer units') parser.add_argument('--dec-hsize', '-d', default=200, type=int, help='Number of decoder hidden layer units') parser.add_argument('--classifier', default='baseline', type=str, help='') # Training conditions parser.add_argument('--optimizer', '-o', default='AdaDelta', type=str, choices=['SGD', 'Adam', 'AdaDelta', 'RMSprop'], help="optimizer") parser.add_argument('--rand-seed', '-s', default=1, type=int, help="seed for generating random numbers") parser.add_argument('--batch-size', '-b', default=20, type=int, help='Batch size in training') parser.add_argument('--num-epochs', '-e', default=15, type=int, help='Number of epochs') parser.add_argument('--max-length', default=20, type=int, help='Maximum length for controling batch size') parser.add_argument('--n-batches', default=-1, type=int, help='Number of batches in training') parser.add_argument('--weight-decay', default=0, type=float, help='') parser.add_argument('--lr-scheduler', action='store_true', help='') parser.add_argument('--lr', default=-1, type=float, help='') # others parser.add_argument('--verbose', '-v', default=0, type=int, help='verbose level') args = parser.parse_args() args.pretrained_elmo = bool(args.pretrained_elmo) args.finetune_elmo = bool(args.finetune_elmo) args.pretrained_bert = bool(args.pretrained_bert) args.finetune_bert = bool(args.finetune_bert) args.add_word_emb = bool(args.add_word_emb) args.pretrained_all = bool(args.pretrained_all) random.seed(args.rand_seed) np.random.seed(args.rand_seed) if args.dictmap != '': dictmap = json.load(open(args.dictmap, 'r')) else: dictmap = None if args.verbose >= 1: logging.basicConfig(level=logging.DEBUG, format='%(asctime)s (%(module)s:%(lineno)d) %(levelname)s: %(message)s') else: logging.basicConfig(level=logging.INFO, format='%(asctime)s %(levelname)s: %(message)s') for arg in vars(args): print("{}={}".format(arg, getattr(args, arg))) # get vocabulary if args.pretrained_bert: if 'bert' in args.bert_model: bert_tokenizer = BertTokenizer.from_pretrained(args.bert_model) elif 'openai-gpt' in args.bert_model: bert_tokenizer = OpenAIGPTTokenizer.from_pretrained(args.bert_model) elif 'gpt2' in args.bert_model: bert_tokenizer = GPT2Tokenizer.from_pretrained(args.bert_model) elif 'transfo-xl' in args.bert_model: bert_tokenizer = TransfoXLTokenizer.from_pretrained(args.bert_model) else: bert_tokenizer = None logging.info('Extracting words from ' + args.train_set) vocab = dh.get_vocabulary(args.train_set, include_caption=args.include_caption, tokenizer=bert_tokenizer) if args.pretrained_word_emb is not None and 'none' not in args.pretrained_word_emb: pretrained_word_emb = dh.get_word_emb(vocab, args.pretrained_word_emb) else: pretrained_word_emb = None # load data logging.info('Loading training data from ' + args.train_set) train_data = dh.load(args.fea_type, args.train_path, args.train_set, vocabfile=args.vocabfile, include_caption=args.include_caption, separate_caption=args.separate_caption, vocab=vocab, dictmap=dictmap, pretrained_elmo=args.pretrained_elmo, pretrained_bert=args.pretrained_bert, bert_model=args.bert_model, tokenizer=bert_tokenizer, pretrained_all=args.pretrained_all, concat_his=args.concat_his) logging.info('Loading validation data from ' + args.valid_set) valid_data = dh.load(args.fea_type, args.valid_path, args.valid_set, vocabfile=args.vocabfile, include_caption=args.include_caption, separate_caption=args.separate_caption, vocab=vocab, dictmap=dictmap, pretrained_elmo=args.pretrained_elmo, pretrained_bert=args.pretrained_bert, bert_model=args.bert_model, tokenizer=bert_tokenizer, pretrained_all=args.pretrained_all, concat_his=args.concat_his) feature_dims = dh.feature_shape(train_data) logging.info("Detected feature dims: {}".format(feature_dims)); # Prepare RNN model and load data caption_state_size = -1 if args.pretrained_weights: pretrained_weights = pickle.load(open(args.pretrained_weights, 'rb')) pretrained_conf = ('/').join(args.pretrained_weights.split('/')[:-1]) + '/avsd_model.conf' pretrained_vocab, _ = pickle.load(open(pretrained_conf, 'rb')) pretrained_weights = dh.align_vocab(pretrained_vocab, vocab, pretrained_weights) else: pretrained_weights = None if args.q_att: in_size_decoder = args.mout_size + args.hist_out_size + args.in_enc_hsize*2 state_size = args.in_enc_hsize*2 else: in_size_decoder = args.mout_size + args.hist_out_size + args.in_enc_hsize state_size = args.in_enc_hsize if args.separate_caption: if args.c_att == 'conv_sum': caption_state_size = args.in_enc_hsize*2 if not args.caption_states_att: in_size_decoder += caption_state_size else: caption_state_size = args.in_enc_hsize if not args.caption_states_att: in_size_decoder += caption_state_size if args.exclude_video: mm_encoder = None in_size_decoder -= args.mout_size else: if args.caption_mm_att: mm_state_size = caption_state_size else: mm_state_size = state_size mm_encoder = MMEncoder(feature_dims, args.mout_size, enc_psize=args.enc_psize, enc_hsize=args.enc_hsize, att_size=args.att_size, state_size=mm_state_size, attention=args.mm_att, fusioning=args.mm_fusioning, att_hops=args.mm_att_hops) if args.ft_fusioning == 'caption_mm_nonlinear_multiply': in_size_decoder = in_size_decoder - args.mout_size - caption_state_size + args.caption_mm_fusion_out_size weights_init=pretrained_weights['history_encoder'] if pretrained_weights is not None else None hlstm_encoder = HLSTMEncoder(args.hist_enc_layers[0], args.hist_enc_layers[1], len(vocab), args.hist_out_size, args.embed_size, args.hist_enc_hsize, rnn_type=args.rnn_type, embedding_init=pretrained_word_emb, weights_init=weights_init, elmo_init=args.pretrained_elmo, elmo_num_outputs=args.elmo_num_outputs, finetune_elmo=args.finetune_elmo, bert_init=args.pretrained_bert, bert_model=args.bert_model, finetune_bert=args.finetune_bert, add_word_emb=args.add_word_emb, pretrained_all=args.pretrained_all, concat_his=args.concat_his) weights_init=pretrained_weights['input_encoder'] if pretrained_weights is not None else None input_encoder = LSTMEncoder(args.in_enc_layers, len(vocab), args.in_enc_hsize, args.embed_size, attention=args.q_att, rnn_type=args.rnn_type, embedding_init=pretrained_word_emb, weights_init=weights_init, elmo_init=args.pretrained_elmo, elmo_num_outputs=args.elmo_num_outputs, finetune_elmo=args.finetune_elmo, bert_init=args.pretrained_bert, bert_model=args.bert_model, finetune_bert=args.finetune_bert, add_word_emb=args.add_word_emb) weights_init=pretrained_weights['response_decoder'] if pretrained_weights is not None else None hlstm_decoder = HLSTMDecoder(args.dec_layers, len(vocab), len(vocab), args.embed_size, in_size_decoder, args.dec_hsize, args.dec_psize, independent=True, rnn_type=args.rnn_type, classifier=args.classifier, states_att=args.caption_states_att, state_size=caption_state_size, embedding_init=pretrained_word_emb, weights_init=weights_init, elmo_init=args.pretrained_elmo, elmo_num_outputs=args.elmo_num_outputs, finetune_elmo=args.finetune_elmo, bert_init=args.pretrained_bert, bert_model=args.bert_model, finetune_bert=args.finetune_bert, add_word_emb=args.add_word_emb, pretrained_all=args.pretrained_all) if args.separate_caption: weights_init=pretrained_weights['caption_encoder'] if pretrained_weights is not None else None caption_encoder = LSTMEncoder(args.in_enc_layers, len(vocab), args.in_enc_hsize, args.embed_size, attention=args.c_att, rnn_type=args.rnn_type, q_size=state_size, weights_init=weights_init) else: caption_encoder = None model = MMSeq2SeqModel(mm_encoder, hlstm_encoder, input_encoder, hlstm_decoder, fusioning=args.ft_fusioning, caption_encoder=caption_encoder, caption_states_att = args.caption_states_att, caption_mm_att = args.caption_mm_att, c_in_size=caption_state_size, mm_in_size=args.mout_size, out_size=args.caption_mm_fusion_out_size) # report data summary logging.info('#vocab = %d' % len(vocab)) # make batchset for training logging.info('Making mini batches for training data') train_indices, train_samples = dh.make_batch_indices(train_data, args.batch_size, max_length=args.max_length, separate_caption=args.separate_caption) logging.info('#train sample = %d' % train_samples) logging.info('#train batch = %d' % len(train_indices)) # make batchset for validation logging.info('Making mini batches for validation data') valid_indices, valid_samples = dh.make_batch_indices(valid_data, args.batch_size, max_length=args.max_length, separate_caption=args.separate_caption) logging.info('#validation sample = %d' % valid_samples) logging.info('#validation batch = %d' % len(valid_indices)) # copy model to gpu device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") model.to(device) parallel = False path = args.model + '.conf' with open(path, 'wb') as f: pickle.dump((vocab, args), f, -1) path2 = args.model + '_params.txt' with open(path2, "w") as f: for arg in vars(args): f.write("{}={}\n".format(arg, getattr(args, arg))) # start training logging.info('----------------') logging.info('Start training') logging.info('----------------') # Setup optimizer if args.optimizer == 'SGD': optimizer = torch.optim.SGD(model.parameters(), weight_decay=args.weight_decay) elif args.optimizer == 'Adam': optimizer = torch.optim.Adam(model.parameters(), weight_decay=args.weight_decay) elif args.optimizer == 'AdaDelta': optimizer = torch.optim.Adadelta(model.parameters(), weight_decay=args.weight_decay) elif args.optimizer == 'RMSprop': optimizer = torch.optim.RMSprop(model.parameters(), weight_decay=args.weight_decay) if args.lr > 0: for g in optim.param_groups: g['lr'] = args.lr if args.lr_scheduler: scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=1, gamma=0.5) # initialize status parameters modelext = '.pth.tar' cur_loss = 0. cur_num_words = 0 epoch = 0 start_at = time.time() cur_at = start_at min_valid_ppl = 1.0e+10 n = 0 report_interval = int(100/args.batch_size) bestmodel_num = 0 random.shuffle(train_indices) trace_log_path = args.model+'_trace.csv' with open(trace_log_path, "w") as f: f.write('epoch,split,perplexity\n') train_log_path = args.model+'_train.csv' with open(train_log_path, "w") as f: f.write('epoch,step,perplexity\n') print("Saving training results to {}".format(train_log_path)) print("Saving val results to {}".format(trace_log_path)) # do training iterations for i in six.moves.range(args.num_epochs): if args.lr_scheduler: scheduler.step() logging.info('-------------------------Epoch %d : %s-----------------------' % (i+1, args.optimizer)) train_loss = 0. train_num_words = 0 # fetch the first batch batch = [dh.make_batch(train_data, train_indices[0], separate_caption=args.separate_caption, pretrained_elmo=args.pretrained_elmo, pretrained_bert=args.pretrained_bert, bert_tokenizer=bert_tokenizer, pretrained_all=args.pretrained_all, bert_model=args.bert_model, concat_his=args.concat_his)] # train iterations if args.n_batches > 0: n_batches = args.n_batches else: n_batches = len(train_indices) it = tqdm(six.moves.range(n_batches), desc="epoch {}/{}".format(i, args.num_epochs), ncols=0) for j in it: b = batch.pop() # fetch the next batch in parallel if j < len(train_indices)-1: prefetch = threading.Thread(target=fetch_batch, args=([dh, train_data, train_indices[j+1], args.separate_caption, batch])) prefetch.start() # propagate for training x = [torch.from_numpy(x) for x in b[0]] if args.concat_his: h = [torch.from_numpy(h_i) for h_i in b[1]] else: h = [[torch.from_numpy(h) for h in hb] for hb in b[1]] q = [torch.from_numpy(q) for q in b[2]] ai = [torch.from_numpy(ai) for ai in b[3]] ao = [torch.from_numpy(ao) for ao in b[4]] if args.separate_caption: c = [torch.from_numpy(c) for c in b[5]] else: c = None if args.pretrained_elmo or args.pretrained_bert: if args.pretrained_all: context_q, context_h, context_ai = b[-3:] else: context_q = b[-1] context_h = None context_ai = None else: context_q = None context_h = None context_ai = None if args.exclude_video: x = None if parallel: _, _, loss = model.module.loss(x, h, q, ai, ao, c, context_q, context_h, context_ai) else: _, _, loss = model.loss(x, h, q, ai, ao, c, context_q, context_h, context_ai) num_words = sum([len(s) for s in ao]) batch_loss = loss.cpu().data.numpy() train_loss += batch_loss * num_words train_num_words += num_words cur_loss += batch_loss * num_words cur_num_words += num_words if (n + 1) % report_interval == 0: now = time.time() throuput = report_interval / (now - cur_at) perp = math.exp(cur_loss / cur_num_words) it.set_postfix(train_perplexity='{:.3f}'.format(perp)) with open(train_log_path, "a") as f: f.write("{},{},{:e}\n".format(i+1,n+1,perp)) cur_at = now cur_loss = 0. cur_num_words = 0 n += 1 # Run truncated BPTT optimizer.zero_grad() loss.backward() optimizer.step() # wait prefetch completion prefetch.join() train_ppl = math.exp(train_loss/train_num_words) logging.info("epoch: %d train perplexity: %f" % (i+1, train_ppl)) # validation step logging.info('-------validation--------') now = time.time() valid_ppl, valid_time = evaluate(model, valid_data, valid_indices, parallel) logging.info('validation perplexity: %.4f' % (valid_ppl)) with open(trace_log_path,"a") as f: f.write("{},train,{:e}\n".format(i+1,train_ppl)) f.write("{},val,{:e}\n".format(i+1,valid_ppl)) # update the model via comparing with the lowest perplexity modelfile = args.model + '_' + str(i + 1) + modelext logging.info('writing model params to ' + modelfile) torch.save(model, modelfile) if min_valid_ppl > valid_ppl: bestmodel_num = i+1 logging.info('validation perplexity reduced %.4f -> %.4f' % (min_valid_ppl, valid_ppl)) min_valid_ppl = valid_ppl logging.info('a symbolic link is made as ' + args.model + '_best' + modelext) if os.path.exists(args.model + '_best' + modelext): os.remove(args.model + '_best' + modelext) os.symlink(os.path.basename(args.model + '_' + str(bestmodel_num) + modelext), args.model + '_best' + modelext) cur_at += time.time() - now # skip time of evaluation and file I/O logging.info('----------------') # make a symlink to the best model logging.info('the best model is epoch %d.' % bestmodel_num)
[ "pytorch_pretrained_bert.BertTokenizer.from_pretrained", "data_handler.get_vocabulary", "model.multimodal_encoder.MMEncoder", "data_handler.make_batch_indices", "torch.from_numpy", "data_handler.align_vocab", "pytorch_pretrained_bert.TransfoXLTokenizer.from_pretrained", "torch.cuda.is_available", "p...
[((1051, 1062), 'time.time', 'time.time', ([], {}), '()\n', (1060, 1062), False, 'import time\n'), ((3748, 3773), 'argparse.ArgumentParser', 'argparse.ArgumentParser', ([], {}), '()\n', (3771, 3773), False, 'import argparse\n'), ((10962, 10989), 'random.seed', 'random.seed', (['args.rand_seed'], {}), '(args.rand_seed)\n', (10973, 10989), False, 'import random\n'), ((10994, 11024), 'numpy.random.seed', 'np.random.seed', (['args.rand_seed'], {}), '(args.rand_seed)\n', (11008, 11024), True, 'import numpy as np\n'), ((12089, 12144), 'logging.info', 'logging.info', (["('Extracting words from ' + args.train_set)"], {}), "('Extracting words from ' + args.train_set)\n", (12101, 12144), False, 'import logging\n'), ((12157, 12258), 'data_handler.get_vocabulary', 'dh.get_vocabulary', (['args.train_set'], {'include_caption': 'args.include_caption', 'tokenizer': 'bert_tokenizer'}), '(args.train_set, include_caption=args.include_caption,\n tokenizer=bert_tokenizer)\n', (12174, 12258), True, 'import data_handler as dh\n'), ((12489, 12549), 'logging.info', 'logging.info', (["('Loading training data from ' + args.train_set)"], {}), "('Loading training data from ' + args.train_set)\n", (12501, 12549), False, 'import logging\n'), ((12567, 12975), 'data_handler.load', 'dh.load', (['args.fea_type', 'args.train_path', 'args.train_set'], {'vocabfile': 'args.vocabfile', 'include_caption': 'args.include_caption', 'separate_caption': 'args.separate_caption', 'vocab': 'vocab', 'dictmap': 'dictmap', 'pretrained_elmo': 'args.pretrained_elmo', 'pretrained_bert': 'args.pretrained_bert', 'bert_model': 'args.bert_model', 'tokenizer': 'bert_tokenizer', 'pretrained_all': 'args.pretrained_all', 'concat_his': 'args.concat_his'}), '(args.fea_type, args.train_path, args.train_set, vocabfile=args.\n vocabfile, include_caption=args.include_caption, separate_caption=args.\n separate_caption, vocab=vocab, dictmap=dictmap, pretrained_elmo=args.\n pretrained_elmo, pretrained_bert=args.pretrained_bert, bert_model=args.\n bert_model, tokenizer=bert_tokenizer, pretrained_all=args.\n pretrained_all, concat_his=args.concat_his)\n', (12574, 12975), True, 'import data_handler as dh\n'), ((13109, 13171), 'logging.info', 'logging.info', (["('Loading validation data from ' + args.valid_set)"], {}), "('Loading validation data from ' + args.valid_set)\n", (13121, 13171), False, 'import logging\n'), ((13189, 13597), 'data_handler.load', 'dh.load', (['args.fea_type', 'args.valid_path', 'args.valid_set'], {'vocabfile': 'args.vocabfile', 'include_caption': 'args.include_caption', 'separate_caption': 'args.separate_caption', 'vocab': 'vocab', 'dictmap': 'dictmap', 'pretrained_elmo': 'args.pretrained_elmo', 'pretrained_bert': 'args.pretrained_bert', 'bert_model': 'args.bert_model', 'tokenizer': 'bert_tokenizer', 'pretrained_all': 'args.pretrained_all', 'concat_his': 'args.concat_his'}), '(args.fea_type, args.valid_path, args.valid_set, vocabfile=args.\n vocabfile, include_caption=args.include_caption, separate_caption=args.\n separate_caption, vocab=vocab, dictmap=dictmap, pretrained_elmo=args.\n pretrained_elmo, pretrained_bert=args.pretrained_bert, bert_model=args.\n bert_model, tokenizer=bert_tokenizer, pretrained_all=args.\n pretrained_all, concat_his=args.concat_his)\n', (13196, 13597), True, 'import data_handler as dh\n'), ((13747, 13775), 'data_handler.feature_shape', 'dh.feature_shape', (['train_data'], {}), '(train_data)\n', (13763, 13775), True, 'import data_handler as dh\n'), ((18470, 18799), 'model.mmseq2seq_model.MMSeq2SeqModel', 'MMSeq2SeqModel', (['mm_encoder', 'hlstm_encoder', 'input_encoder', 'hlstm_decoder'], {'fusioning': 'args.ft_fusioning', 'caption_encoder': 'caption_encoder', 'caption_states_att': 'args.caption_states_att', 'caption_mm_att': 'args.caption_mm_att', 'c_in_size': 'caption_state_size', 'mm_in_size': 'args.mout_size', 'out_size': 'args.caption_mm_fusion_out_size'}), '(mm_encoder, hlstm_encoder, input_encoder, hlstm_decoder,\n fusioning=args.ft_fusioning, caption_encoder=caption_encoder,\n caption_states_att=args.caption_states_att, caption_mm_att=args.\n caption_mm_att, c_in_size=caption_state_size, mm_in_size=args.mout_size,\n out_size=args.caption_mm_fusion_out_size)\n', (18484, 18799), False, 'from model.mmseq2seq_model import MMSeq2SeqModel\n'), ((18905, 18958), 'logging.info', 'logging.info', (['"""Making mini batches for training data"""'], {}), "('Making mini batches for training data')\n", (18917, 18958), False, 'import logging\n'), ((18994, 19117), 'data_handler.make_batch_indices', 'dh.make_batch_indices', (['train_data', 'args.batch_size'], {'max_length': 'args.max_length', 'separate_caption': 'args.separate_caption'}), '(train_data, args.batch_size, max_length=args.\n max_length, separate_caption=args.separate_caption)\n', (19015, 19117), True, 'import data_handler as dh\n'), ((19174, 19224), 'logging.info', 'logging.info', (["('#train sample = %d' % train_samples)"], {}), "('#train sample = %d' % train_samples)\n", (19186, 19224), False, 'import logging\n'), ((19323, 19378), 'logging.info', 'logging.info', (['"""Making mini batches for validation data"""'], {}), "('Making mini batches for validation data')\n", (19335, 19378), False, 'import logging\n'), ((19414, 19537), 'data_handler.make_batch_indices', 'dh.make_batch_indices', (['valid_data', 'args.batch_size'], {'max_length': 'args.max_length', 'separate_caption': 'args.separate_caption'}), '(valid_data, args.batch_size, max_length=args.\n max_length, separate_caption=args.separate_caption)\n', (19435, 19537), True, 'import data_handler as dh\n'), ((19590, 19645), 'logging.info', 'logging.info', (["('#validation sample = %d' % valid_samples)"], {}), "('#validation sample = %d' % valid_samples)\n", (19602, 19645), False, 'import logging\n'), ((20152, 20184), 'logging.info', 'logging.info', (['"""----------------"""'], {}), "('----------------')\n", (20164, 20184), False, 'import logging\n'), ((20189, 20219), 'logging.info', 'logging.info', (['"""Start training"""'], {}), "('Start training')\n", (20201, 20219), False, 'import logging\n'), ((20224, 20256), 'logging.info', 'logging.info', (['"""----------------"""'], {}), "('----------------')\n", (20236, 20256), False, 'import logging\n'), ((21120, 21131), 'time.time', 'time.time', ([], {}), '()\n', (21129, 21131), False, 'import time\n'), ((21265, 21294), 'random.shuffle', 'random.shuffle', (['train_indices'], {}), '(train_indices)\n', (21279, 21294), False, 'import random\n'), ((21729, 21761), 'six.moves.range', 'six.moves.range', (['args.num_epochs'], {}), '(args.num_epochs)\n', (21744, 21761), False, 'import six\n'), ((26713, 26772), 'logging.info', 'logging.info', (["('the best model is epoch %d.' % bestmodel_num)"], {}), "('the best model is epoch %d.' % bestmodel_num)\n", (26725, 26772), False, 'import logging\n'), ((698, 972), 'data_handler.make_batch', 'dh.make_batch', (['data', 'index'], {'separate_caption': 'separate_caption', 'pretrained_elmo': 'args.pretrained_elmo', 'pretrained_bert': 'args.pretrained_bert', 'bert_tokenizer': 'bert_tokenizer', 'pretrained_all': 'args.pretrained_all', 'bert_model': 'args.bert_model', 'concat_his': 'args.concat_his'}), '(data, index, separate_caption=separate_caption,\n pretrained_elmo=args.pretrained_elmo, pretrained_bert=args.\n pretrained_bert, bert_tokenizer=bert_tokenizer, pretrained_all=args.\n pretrained_all, bert_model=args.bert_model, concat_his=args.concat_his)\n', (711, 972), True, 'import data_handler as dh\n'), ((1131, 1146), 'torch.no_grad', 'torch.no_grad', ([], {}), '()\n', (1144, 1146), False, 'import torch\n'), ((3584, 3595), 'time.time', 'time.time', ([], {}), '()\n', (3593, 3595), False, 'import time\n'), ((3620, 3656), 'math.exp', 'math.exp', (['(eval_loss / eval_num_words)'], {}), '(eval_loss / eval_num_words)\n', (3628, 3656), False, 'import math\n'), ((11174, 11292), 'logging.basicConfig', 'logging.basicConfig', ([], {'level': 'logging.DEBUG', 'format': '"""%(asctime)s (%(module)s:%(lineno)d) %(levelname)s: %(message)s"""'}), "(level=logging.DEBUG, format=\n '%(asctime)s (%(module)s:%(lineno)d) %(levelname)s: %(message)s')\n", (11193, 11292), False, 'import logging\n'), ((11319, 11412), 'logging.basicConfig', 'logging.basicConfig', ([], {'level': 'logging.INFO', 'format': '"""%(asctime)s %(levelname)s: %(message)s"""'}), "(level=logging.INFO, format=\n '%(asctime)s %(levelname)s: %(message)s')\n", (11338, 11412), False, 'import logging\n'), ((12374, 12422), 'data_handler.get_word_emb', 'dh.get_word_emb', (['vocab', 'args.pretrained_word_emb'], {}), '(vocab, args.pretrained_word_emb)\n', (12389, 12422), True, 'import data_handler as dh\n'), ((14221, 14280), 'data_handler.align_vocab', 'dh.align_vocab', (['pretrained_vocab', 'vocab', 'pretrained_weights'], {}), '(pretrained_vocab, vocab, pretrained_weights)\n', (14235, 14280), True, 'import data_handler as dh\n'), ((15241, 15475), 'model.multimodal_encoder.MMEncoder', 'MMEncoder', (['feature_dims', 'args.mout_size'], {'enc_psize': 'args.enc_psize', 'enc_hsize': 'args.enc_hsize', 'att_size': 'args.att_size', 'state_size': 'mm_state_size', 'attention': 'args.mm_att', 'fusioning': 'args.mm_fusioning', 'att_hops': 'args.mm_att_hops'}), '(feature_dims, args.mout_size, enc_psize=args.enc_psize, enc_hsize\n =args.enc_hsize, att_size=args.att_size, state_size=mm_state_size,\n attention=args.mm_att, fusioning=args.mm_fusioning, att_hops=args.\n mm_att_hops)\n', (15250, 15475), False, 'from model.multimodal_encoder import MMEncoder\n'), ((19924, 19957), 'pickle.dump', 'pickle.dump', (['(vocab, args)', 'f', '(-1)'], {}), '((vocab, args), f, -1)\n', (19935, 19957), False, 'import pickle\n'), ((20922, 20988), 'torch.optim.lr_scheduler.StepLR', 'torch.optim.lr_scheduler.StepLR', (['optimizer'], {'step_size': '(1)', 'gamma': '(0.5)'}), '(optimizer, step_size=1, gamma=0.5)\n', (20953, 20988), False, 'import torch\n'), ((21830, 21943), 'logging.info', 'logging.info', (["('-------------------------Epoch %d : %s-----------------------' % (i + 1,\n args.optimizer))"], {}), "(\n '-------------------------Epoch %d : %s-----------------------' % (i + \n 1, args.optimizer))\n", (21842, 21943), False, 'import logging\n'), ((25236, 25274), 'math.exp', 'math.exp', (['(train_loss / train_num_words)'], {}), '(train_loss / train_num_words)\n', (25244, 25274), False, 'import math\n'), ((25281, 25349), 'logging.info', 'logging.info', (["('epoch: %d train perplexity: %f' % (i + 1, train_ppl))"], {}), "('epoch: %d train perplexity: %f' % (i + 1, train_ppl))\n", (25293, 25349), False, 'import logging\n'), ((25384, 25425), 'logging.info', 'logging.info', (['"""-------validation--------"""'], {}), "('-------validation--------')\n", (25396, 25425), False, 'import logging\n'), ((25440, 25451), 'time.time', 'time.time', ([], {}), '()\n', (25449, 25451), False, 'import time\n'), ((25545, 25600), 'logging.info', 'logging.info', (["('validation perplexity: %.4f' % valid_ppl)"], {}), "('validation perplexity: %.4f' % valid_ppl)\n", (25557, 25600), False, 'import logging\n'), ((25915, 25967), 'logging.info', 'logging.info', (["('writing model params to ' + modelfile)"], {}), "('writing model params to ' + modelfile)\n", (25927, 25967), False, 'import logging\n'), ((25976, 26004), 'torch.save', 'torch.save', (['model', 'modelfile'], {}), '(model, modelfile)\n', (25986, 26004), False, 'import torch\n'), ((26636, 26668), 'logging.info', 'logging.info', (['"""----------------"""'], {}), "('----------------')\n", (26648, 26668), False, 'import logging\n'), ((1197, 1481), 'data_handler.make_batch', 'dh.make_batch', (['data', 'indices[0]'], {'separate_caption': 'args.separate_caption', 'pretrained_elmo': 'args.pretrained_elmo', 'pretrained_bert': 'args.pretrained_bert', 'bert_tokenizer': 'bert_tokenizer', 'pretrained_all': 'args.pretrained_all', 'bert_model': 'args.bert_model', 'concat_his': 'args.concat_his'}), '(data, indices[0], separate_caption=args.separate_caption,\n pretrained_elmo=args.pretrained_elmo, pretrained_bert=args.\n pretrained_bert, bert_tokenizer=bert_tokenizer, pretrained_all=args.\n pretrained_all, bert_model=args.bert_model, concat_his=args.concat_his)\n', (1210, 1481), True, 'import data_handler as dh\n'), ((11626, 11672), 'pytorch_pretrained_bert.BertTokenizer.from_pretrained', 'BertTokenizer.from_pretrained', (['args.bert_model'], {}), '(args.bert_model)\n', (11655, 11672), False, 'from pytorch_pretrained_bert import BertTokenizer, OpenAIGPTTokenizer, GPT2Tokenizer, TransfoXLTokenizer\n'), ((19772, 19797), 'torch.cuda.is_available', 'torch.cuda.is_available', ([], {}), '()\n', (19795, 19797), False, 'import torch\n'), ((22033, 22335), 'data_handler.make_batch', 'dh.make_batch', (['train_data', 'train_indices[0]'], {'separate_caption': 'args.separate_caption', 'pretrained_elmo': 'args.pretrained_elmo', 'pretrained_bert': 'args.pretrained_bert', 'bert_tokenizer': 'bert_tokenizer', 'pretrained_all': 'args.pretrained_all', 'bert_model': 'args.bert_model', 'concat_his': 'args.concat_his'}), '(train_data, train_indices[0], separate_caption=args.\n separate_caption, pretrained_elmo=args.pretrained_elmo, pretrained_bert\n =args.pretrained_bert, bert_tokenizer=bert_tokenizer, pretrained_all=\n args.pretrained_all, bert_model=args.bert_model, concat_his=args.concat_his\n )\n', (22046, 22335), True, 'import data_handler as dh\n'), ((22523, 22549), 'six.moves.range', 'six.moves.range', (['n_batches'], {}), '(n_batches)\n', (22538, 22549), False, 'import six\n'), ((26088, 26179), 'logging.info', 'logging.info', (["('validation perplexity reduced %.4f -> %.4f' % (min_valid_ppl, valid_ppl))"], {}), "('validation perplexity reduced %.4f -> %.4f' % (min_valid_ppl,\n valid_ppl))\n", (26100, 26179), False, 'import logging\n'), ((26226, 26303), 'logging.info', 'logging.info', (["('a symbolic link is made as ' + args.model + '_best' + modelext)"], {}), "('a symbolic link is made as ' + args.model + '_best' + modelext)\n", (26238, 26303), False, 'import logging\n'), ((26319, 26366), 'os.path.exists', 'os.path.exists', (["(args.model + '_best' + modelext)"], {}), "(args.model + '_best' + modelext)\n", (26333, 26366), False, 'import os\n'), ((26570, 26581), 'time.time', 'time.time', ([], {}), '()\n', (26579, 26581), False, 'import time\n'), ((1915, 2019), 'threading.Thread', 'threading.Thread', ([], {'target': 'fetch_batch', 'args': '[dh, data, indices[j + 1], args.separate_caption, batch]'}), '(target=fetch_batch, args=[dh, data, indices[j + 1], args.\n separate_caption, batch])\n', (1931, 2019), False, 'import threading\n'), ((2135, 2154), 'torch.from_numpy', 'torch.from_numpy', (['x'], {}), '(x)\n', (2151, 2154), False, 'import torch\n'), ((2368, 2387), 'torch.from_numpy', 'torch.from_numpy', (['q'], {}), '(q)\n', (2384, 2387), False, 'import torch\n'), ((2421, 2441), 'torch.from_numpy', 'torch.from_numpy', (['ai'], {}), '(ai)\n', (2437, 2441), False, 'import torch\n'), ((2476, 2496), 'torch.from_numpy', 'torch.from_numpy', (['ao'], {}), '(ao)\n', (2492, 2496), False, 'import torch\n'), ((11748, 11799), 'pytorch_pretrained_bert.OpenAIGPTTokenizer.from_pretrained', 'OpenAIGPTTokenizer.from_pretrained', (['args.bert_model'], {}), '(args.bert_model)\n', (11782, 11799), False, 'from pytorch_pretrained_bert import BertTokenizer, OpenAIGPTTokenizer, GPT2Tokenizer, TransfoXLTokenizer\n'), ((22771, 22886), 'threading.Thread', 'threading.Thread', ([], {'target': 'fetch_batch', 'args': '[dh, train_data, train_indices[j + 1], args.separate_caption, batch]'}), '(target=fetch_batch, args=[dh, train_data, train_indices[j +\n 1], args.separate_caption, batch])\n', (22787, 22886), False, 'import threading\n'), ((23016, 23035), 'torch.from_numpy', 'torch.from_numpy', (['x'], {}), '(x)\n', (23032, 23035), False, 'import torch\n'), ((23249, 23268), 'torch.from_numpy', 'torch.from_numpy', (['q'], {}), '(q)\n', (23265, 23268), False, 'import torch\n'), ((23302, 23322), 'torch.from_numpy', 'torch.from_numpy', (['ai'], {}), '(ai)\n', (23318, 23322), False, 'import torch\n'), ((23357, 23377), 'torch.from_numpy', 'torch.from_numpy', (['ao'], {}), '(ao)\n', (23373, 23377), False, 'import torch\n'), ((24593, 24604), 'time.time', 'time.time', ([], {}), '()\n', (24602, 24604), False, 'import time\n'), ((24688, 24722), 'math.exp', 'math.exp', (['(cur_loss / cur_num_words)'], {}), '(cur_loss / cur_num_words)\n', (24696, 24722), False, 'import math\n'), ((26384, 26426), 'os.remove', 'os.remove', (["(args.model + '_best' + modelext)"], {}), "(args.model + '_best' + modelext)\n", (26393, 26426), False, 'import os\n'), ((2223, 2244), 'torch.from_numpy', 'torch.from_numpy', (['h_i'], {}), '(h_i)\n', (2239, 2244), False, 'import torch\n'), ((2572, 2591), 'torch.from_numpy', 'torch.from_numpy', (['c'], {}), '(c)\n', (2588, 2591), False, 'import torch\n'), ((11869, 11915), 'pytorch_pretrained_bert.GPT2Tokenizer.from_pretrained', 'GPT2Tokenizer.from_pretrained', (['args.bert_model'], {}), '(args.bert_model)\n', (11898, 11915), False, 'from pytorch_pretrained_bert import BertTokenizer, OpenAIGPTTokenizer, GPT2Tokenizer, TransfoXLTokenizer\n'), ((23104, 23125), 'torch.from_numpy', 'torch.from_numpy', (['h_i'], {}), '(h_i)\n', (23120, 23125), False, 'import torch\n'), ((23453, 23472), 'torch.from_numpy', 'torch.from_numpy', (['c'], {}), '(c)\n', (23469, 23472), False, 'import torch\n'), ((2302, 2321), 'torch.from_numpy', 'torch.from_numpy', (['h'], {}), '(h)\n', (2318, 2321), False, 'import torch\n'), ((11991, 12042), 'pytorch_pretrained_bert.TransfoXLTokenizer.from_pretrained', 'TransfoXLTokenizer.from_pretrained', (['args.bert_model'], {}), '(args.bert_model)\n', (12025, 12042), False, 'from pytorch_pretrained_bert import BertTokenizer, OpenAIGPTTokenizer, GPT2Tokenizer, TransfoXLTokenizer\n'), ((23183, 23202), 'torch.from_numpy', 'torch.from_numpy', (['h'], {}), '(h)\n', (23199, 23202), False, 'import torch\n')]
""" Contributions from: DSEverything - Mean Mix - Math, Geo, Harmonic (LB 0.493) https://www.kaggle.com/dongxu027/mean-mix-math-geo-harmonic-lb-0-493 JdPaletto - Surprised Yet? - Part2 - (LB: 0.503) https://www.kaggle.com/jdpaletto/surprised-yet-part2-lb-0-503 hklee - weighted mean comparisons, LB 0.497, 1ST https://www.kaggle.com/zeemeen/weighted-mean-comparisons-lb-0-497-1st Also all comments for changes, encouragement, and forked scripts rock Keep the Surprise Going """ import glob, re import numpy as np import pandas as pd from sklearn import * from datetime import datetime from xgboost import XGBRegressor data = { 'tra': pd.read_csv('../input/air_visit_data.csv'), 'as': pd.read_csv('../input/air_store_info.csv'), 'hs': pd.read_csv('../input/hpg_store_info.csv'), 'ar': pd.read_csv('../input/air_reserve.csv'), 'hr': pd.read_csv('../input/hpg_reserve.csv'), 'id': pd.read_csv('../input/store_id_relation.csv'), 'tes': pd.read_csv('../input/sample_submission.csv'), 'hol': pd.read_csv('../input/date_info.csv').rename(columns={'calendar_date':'visit_date'}) } data['hr'] = pd.merge(data['hr'], data['id'], how='inner', on=['hpg_store_id']) for df in ['ar','hr']: data[df]['visit_datetime'] = pd.to_datetime(data[df]['visit_datetime']) data[df]['visit_datetime'] = data[df]['visit_datetime'].dt.date data[df]['reserve_datetime'] = pd.to_datetime(data[df]['reserve_datetime']) data[df]['reserve_datetime'] = data[df]['reserve_datetime'].dt.date data[df]['reserve_datetime_diff'] = data[df].apply(lambda r: (r['visit_datetime'] - r['reserve_datetime']).days, axis=1) tmp1 = data[df].groupby(['air_store_id','visit_datetime'], as_index=False)[['reserve_datetime_diff', 'reserve_visitors']].sum().rename(columns={'visit_datetime':'visit_date', 'reserve_datetime_diff': 'rs1', 'reserve_visitors':'rv1'}) tmp2 = data[df].groupby(['air_store_id','visit_datetime'], as_index=False)[['reserve_datetime_diff', 'reserve_visitors']].mean().rename(columns={'visit_datetime':'visit_date', 'reserve_datetime_diff': 'rs2', 'reserve_visitors':'rv2'}) data[df] = pd.merge(tmp1, tmp2, how='inner', on=['air_store_id','visit_date']) data['tra']['visit_date'] = pd.to_datetime(data['tra']['visit_date']) data['tra']['dow'] = data['tra']['visit_date'].dt.dayofweek data['tra']['year'] = data['tra']['visit_date'].dt.year data['tra']['month'] = data['tra']['visit_date'].dt.month data['tra']['visit_date'] = data['tra']['visit_date'].dt.date data['tes']['visit_date'] = data['tes']['id'].map(lambda x: str(x).split('_')[2]) data['tes']['air_store_id'] = data['tes']['id'].map(lambda x: '_'.join(x.split('_')[:2])) data['tes']['visit_date'] = pd.to_datetime(data['tes']['visit_date']) data['tes']['dow'] = data['tes']['visit_date'].dt.dayofweek data['tes']['year'] = data['tes']['visit_date'].dt.year data['tes']['month'] = data['tes']['visit_date'].dt.month data['tes']['visit_date'] = data['tes']['visit_date'].dt.date unique_stores = data['tes']['air_store_id'].unique() stores = pd.concat([pd.DataFrame({'air_store_id': unique_stores, 'dow': [i]*len(unique_stores)}) for i in range(7)], axis=0, ignore_index=True).reset_index(drop=True) #sure it can be compressed... tmp = data['tra'].groupby(['air_store_id','dow'], as_index=False)['visitors'].min().rename(columns={'visitors':'min_visitors'}) stores = pd.merge(stores, tmp, how='left', on=['air_store_id','dow']) tmp = data['tra'].groupby(['air_store_id','dow'], as_index=False)['visitors'].mean().rename(columns={'visitors':'mean_visitors'}) stores = pd.merge(stores, tmp, how='left', on=['air_store_id','dow']) tmp = data['tra'].groupby(['air_store_id','dow'], as_index=False)['visitors'].median().rename(columns={'visitors':'median_visitors'}) stores = pd.merge(stores, tmp, how='left', on=['air_store_id','dow']) tmp = data['tra'].groupby(['air_store_id','dow'], as_index=False)['visitors'].max().rename(columns={'visitors':'max_visitors'}) stores = pd.merge(stores, tmp, how='left', on=['air_store_id','dow']) tmp = data['tra'].groupby(['air_store_id','dow'], as_index=False)['visitors'].count().rename(columns={'visitors':'count_observations'}) stores = pd.merge(stores, tmp, how='left', on=['air_store_id','dow']) stores = pd.merge(stores, data['as'], how='left', on=['air_store_id']) # NEW FEATURES FROM <NAME>nia stores['air_genre_name'] = stores['air_genre_name'].map(lambda x: str(str(x).replace('/',' '))) stores['air_area_name'] = stores['air_area_name'].map(lambda x: str(str(x).replace('-',' '))) lbl = preprocessing.LabelEncoder() for i in range(10): stores['air_genre_name'+str(i)] = lbl.fit_transform(stores['air_genre_name'].map(lambda x: str(str(x).split(' ')[i]) if len(str(x).split(' '))>i else '')) stores['air_area_name'+str(i)] = lbl.fit_transform(stores['air_area_name'].map(lambda x: str(str(x).split(' ')[i]) if len(str(x).split(' '))>i else '')) stores['air_genre_name'] = lbl.fit_transform(stores['air_genre_name']) stores['air_area_name'] = lbl.fit_transform(stores['air_area_name']) data['hol']['visit_date'] = pd.to_datetime(data['hol']['visit_date']) data['hol']['day_of_week'] = lbl.fit_transform(data['hol']['day_of_week']) data['hol']['visit_date'] = data['hol']['visit_date'].dt.date train = pd.merge(data['tra'], data['hol'], how='left', on=['visit_date']) test = pd.merge(data['tes'], data['hol'], how='left', on=['visit_date']) train = pd.merge(train, stores, how='left', on=['air_store_id','dow']) test = pd.merge(test, stores, how='left', on=['air_store_id','dow']) for df in ['ar','hr']: train = pd.merge(train, data[df], how='left', on=['air_store_id','visit_date']) test = pd.merge(test, data[df], how='left', on=['air_store_id','visit_date']) train['id'] = train.apply(lambda r: '_'.join([str(r['air_store_id']), str(r['visit_date'])]), axis=1) train['total_reserv_sum'] = train['rv1_x'] + train['rv1_y'] train['total_reserv_mean'] = (train['rv2_x'] + train['rv2_y']) / 2 train['total_reserv_dt_diff_mean'] = (train['rs2_x'] + train['rs2_y']) / 2 test['total_reserv_sum'] = test['rv1_x'] + test['rv1_y'] test['total_reserv_mean'] = (test['rv2_x'] + test['rv2_y']) / 2 test['total_reserv_dt_diff_mean'] = (test['rs2_x'] + test['rs2_y']) / 2 # NEW FEATURES FROM JMBULL train['date_int'] = train['visit_date'].apply(lambda x: x.strftime('%Y%m%d')).astype(int) test['date_int'] = test['visit_date'].apply(lambda x: x.strftime('%Y%m%d')).astype(int) train['var_max_lat'] = train['latitude'].max() - train['latitude'] train['var_max_long'] = train['longitude'].max() - train['longitude'] test['var_max_lat'] = test['latitude'].max() - test['latitude'] test['var_max_long'] = test['longitude'].max() - test['longitude'] # NEW FEATURES FROM Georgii Vyshnia train['lon_plus_lat'] = train['longitude'] + train['latitude'] test['lon_plus_lat'] = test['longitude'] + test['latitude'] lbl = preprocessing.LabelEncoder() train['air_store_id2'] = lbl.fit_transform(train['air_store_id']) test['air_store_id2'] = lbl.transform(test['air_store_id']) col = [c for c in train if c not in ['id', 'air_store_id', 'visit_date','visitors']] train = train.fillna(-1) test = test.fillna(-1) def RMSLE(y, pred): return metrics.mean_squared_error(y, pred)**0.5 model1 = ensemble.GradientBoostingRegressor(learning_rate=0.2, random_state=3, n_estimators=200, subsample=0.8, max_depth =10) model2 = neighbors.KNeighborsRegressor(n_jobs=-1, n_neighbors=4) model3 = XGBRegressor(learning_rate=0.2, random_state=3, n_estimators=280, subsample=0.8, colsample_bytree=0.8, max_depth =12) model1.fit(train[col], np.log1p(train['visitors'].values)) model2.fit(train[col], np.log1p(train['visitors'].values)) model3.fit(train[col], np.log1p(train['visitors'].values)) preds1 = model1.predict(train[col]) preds2 = model2.predict(train[col]) preds3 = model3.predict(train[col]) print('RMSE GradientBoostingRegressor: ', RMSLE(np.log1p(train['visitors'].values), preds1)) print('RMSE KNeighborsRegressor: ', RMSLE(np.log1p(train['visitors'].values), preds2)) print('RMSE XGBRegressor: ', RMSLE(np.log1p(train['visitors'].values), preds3)) preds1 = model1.predict(test[col]) preds2 = model2.predict(test[col]) preds3 = model3.predict(test[col]) test['visitors'] = 0.3*preds1+0.3*preds2+0.4*preds3 test['visitors'] = np.expm1(test['visitors']).clip(lower=0.) sub1 = test[['id','visitors']].copy() del train; del data; # from hklee # https://www.kaggle.com/zeemeen/weighted-mean-comparisons-lb-0-497-1st/code dfs = { re.search('/([^/\.]*)\.csv', fn).group(1): pd.read_csv(fn)for fn in glob.glob('../input/*.csv')} for k, v in dfs.items(): locals()[k] = v wkend_holidays = date_info.apply( (lambda x:(x.day_of_week=='Sunday' or x.day_of_week=='Saturday') and x.holiday_flg==1), axis=1) date_info.loc[wkend_holidays, 'holiday_flg'] = 0 date_info['weight'] = ((date_info.index + 1) / len(date_info)) ** 5 visit_data = air_visit_data.merge(date_info, left_on='visit_date', right_on='calendar_date', how='left') visit_data.drop('calendar_date', axis=1, inplace=True) visit_data['visitors'] = visit_data.visitors.map(pd.np.log1p) wmean = lambda x:( (x.weight * x.visitors).sum() / x.weight.sum() ) visitors = visit_data.groupby(['air_store_id', 'day_of_week', 'holiday_flg']).apply(wmean).reset_index() visitors.rename(columns={0:'visitors'}, inplace=True) # cumbersome, should be better ways. sample_submission['air_store_id'] = sample_submission.id.map(lambda x: '_'.join(x.split('_')[:-1])) sample_submission['calendar_date'] = sample_submission.id.map(lambda x: x.split('_')[2]) sample_submission.drop('visitors', axis=1, inplace=True) sample_submission = sample_submission.merge(date_info, on='calendar_date', how='left') sample_submission = sample_submission.merge(visitors, on=[ 'air_store_id', 'day_of_week', 'holiday_flg'], how='left') missings = sample_submission.visitors.isnull() sample_submission.loc[missings, 'visitors'] = sample_submission[missings].merge( visitors[visitors.holiday_flg==0], on=('air_store_id', 'day_of_week'), how='left')['visitors_y'].values missings = sample_submission.visitors.isnull() sample_submission.loc[missings, 'visitors'] = sample_submission[missings].merge( visitors[['air_store_id', 'visitors']].groupby('air_store_id').mean().reset_index(), on='air_store_id', how='left')['visitors_y'].values sample_submission['visitors'] = sample_submission.visitors.map(pd.np.expm1) sub2 = sample_submission[['id', 'visitors']].copy() sub_merge = pd.merge(sub1, sub2, on='id', how='inner') sub_merge['visitors'] = 0.7*sub_merge['visitors_x'] + 0.3*sub_merge['visitors_y']* 1.1 sub_merge[['id', 'visitors']].to_csv('submission.csv', index=False)
[ "pandas.read_csv", "pandas.merge", "numpy.expm1", "xgboost.XGBRegressor", "glob.glob", "numpy.log1p", "pandas.to_datetime", "re.search" ]
[((1128, 1194), 'pandas.merge', 'pd.merge', (["data['hr']", "data['id']"], {'how': '"""inner"""', 'on': "['hpg_store_id']"}), "(data['hr'], data['id'], how='inner', on=['hpg_store_id'])\n", (1136, 1194), True, 'import pandas as pd\n'), ((2229, 2270), 'pandas.to_datetime', 'pd.to_datetime', (["data['tra']['visit_date']"], {}), "(data['tra']['visit_date'])\n", (2243, 2270), True, 'import pandas as pd\n'), ((2708, 2749), 'pandas.to_datetime', 'pd.to_datetime', (["data['tes']['visit_date']"], {}), "(data['tes']['visit_date'])\n", (2722, 2749), True, 'import pandas as pd\n'), ((3375, 3436), 'pandas.merge', 'pd.merge', (['stores', 'tmp'], {'how': '"""left"""', 'on': "['air_store_id', 'dow']"}), "(stores, tmp, how='left', on=['air_store_id', 'dow'])\n", (3383, 3436), True, 'import pandas as pd\n'), ((3576, 3637), 'pandas.merge', 'pd.merge', (['stores', 'tmp'], {'how': '"""left"""', 'on': "['air_store_id', 'dow']"}), "(stores, tmp, how='left', on=['air_store_id', 'dow'])\n", (3584, 3637), True, 'import pandas as pd\n'), ((3780, 3841), 'pandas.merge', 'pd.merge', (['stores', 'tmp'], {'how': '"""left"""', 'on': "['air_store_id', 'dow']"}), "(stores, tmp, how='left', on=['air_store_id', 'dow'])\n", (3788, 3841), True, 'import pandas as pd\n'), ((3978, 4039), 'pandas.merge', 'pd.merge', (['stores', 'tmp'], {'how': '"""left"""', 'on': "['air_store_id', 'dow']"}), "(stores, tmp, how='left', on=['air_store_id', 'dow'])\n", (3986, 4039), True, 'import pandas as pd\n'), ((4184, 4245), 'pandas.merge', 'pd.merge', (['stores', 'tmp'], {'how': '"""left"""', 'on': "['air_store_id', 'dow']"}), "(stores, tmp, how='left', on=['air_store_id', 'dow'])\n", (4192, 4245), True, 'import pandas as pd\n'), ((4256, 4317), 'pandas.merge', 'pd.merge', (['stores', "data['as']"], {'how': '"""left"""', 'on': "['air_store_id']"}), "(stores, data['as'], how='left', on=['air_store_id'])\n", (4264, 4317), True, 'import pandas as pd\n'), ((5079, 5120), 'pandas.to_datetime', 'pd.to_datetime', (["data['hol']['visit_date']"], {}), "(data['hol']['visit_date'])\n", (5093, 5120), True, 'import pandas as pd\n'), ((5266, 5331), 'pandas.merge', 'pd.merge', (["data['tra']", "data['hol']"], {'how': '"""left"""', 'on': "['visit_date']"}), "(data['tra'], data['hol'], how='left', on=['visit_date'])\n", (5274, 5331), True, 'import pandas as pd\n'), ((5340, 5405), 'pandas.merge', 'pd.merge', (["data['tes']", "data['hol']"], {'how': '"""left"""', 'on': "['visit_date']"}), "(data['tes'], data['hol'], how='left', on=['visit_date'])\n", (5348, 5405), True, 'import pandas as pd\n'), ((5416, 5479), 'pandas.merge', 'pd.merge', (['train', 'stores'], {'how': '"""left"""', 'on': "['air_store_id', 'dow']"}), "(train, stores, how='left', on=['air_store_id', 'dow'])\n", (5424, 5479), True, 'import pandas as pd\n'), ((5487, 5549), 'pandas.merge', 'pd.merge', (['test', 'stores'], {'how': '"""left"""', 'on': "['air_store_id', 'dow']"}), "(test, stores, how='left', on=['air_store_id', 'dow'])\n", (5495, 5549), True, 'import pandas as pd\n'), ((7473, 7594), 'xgboost.XGBRegressor', 'XGBRegressor', ([], {'learning_rate': '(0.2)', 'random_state': '(3)', 'n_estimators': '(280)', 'subsample': '(0.8)', 'colsample_bytree': '(0.8)', 'max_depth': '(12)'}), '(learning_rate=0.2, random_state=3, n_estimators=280, subsample\n =0.8, colsample_bytree=0.8, max_depth=12)\n', (7485, 7594), False, 'from xgboost import XGBRegressor\n'), ((10538, 10580), 'pandas.merge', 'pd.merge', (['sub1', 'sub2'], {'on': '"""id"""', 'how': '"""inner"""'}), "(sub1, sub2, on='id', how='inner')\n", (10546, 10580), True, 'import pandas as pd\n'), ((643, 685), 'pandas.read_csv', 'pd.read_csv', (['"""../input/air_visit_data.csv"""'], {}), "('../input/air_visit_data.csv')\n", (654, 685), True, 'import pandas as pd\n'), ((697, 739), 'pandas.read_csv', 'pd.read_csv', (['"""../input/air_store_info.csv"""'], {}), "('../input/air_store_info.csv')\n", (708, 739), True, 'import pandas as pd\n'), ((751, 793), 'pandas.read_csv', 'pd.read_csv', (['"""../input/hpg_store_info.csv"""'], {}), "('../input/hpg_store_info.csv')\n", (762, 793), True, 'import pandas as pd\n'), ((805, 844), 'pandas.read_csv', 'pd.read_csv', (['"""../input/air_reserve.csv"""'], {}), "('../input/air_reserve.csv')\n", (816, 844), True, 'import pandas as pd\n'), ((856, 895), 'pandas.read_csv', 'pd.read_csv', (['"""../input/hpg_reserve.csv"""'], {}), "('../input/hpg_reserve.csv')\n", (867, 895), True, 'import pandas as pd\n'), ((907, 952), 'pandas.read_csv', 'pd.read_csv', (['"""../input/store_id_relation.csv"""'], {}), "('../input/store_id_relation.csv')\n", (918, 952), True, 'import pandas as pd\n'), ((965, 1010), 'pandas.read_csv', 'pd.read_csv', (['"""../input/sample_submission.csv"""'], {}), "('../input/sample_submission.csv')\n", (976, 1010), True, 'import pandas as pd\n'), ((1252, 1294), 'pandas.to_datetime', 'pd.to_datetime', (["data[df]['visit_datetime']"], {}), "(data[df]['visit_datetime'])\n", (1266, 1294), True, 'import pandas as pd\n'), ((1398, 1442), 'pandas.to_datetime', 'pd.to_datetime', (["data[df]['reserve_datetime']"], {}), "(data[df]['reserve_datetime'])\n", (1412, 1442), True, 'import pandas as pd\n'), ((2132, 2200), 'pandas.merge', 'pd.merge', (['tmp1', 'tmp2'], {'how': '"""inner"""', 'on': "['air_store_id', 'visit_date']"}), "(tmp1, tmp2, how='inner', on=['air_store_id', 'visit_date'])\n", (2140, 2200), True, 'import pandas as pd\n'), ((5585, 5657), 'pandas.merge', 'pd.merge', (['train', 'data[df]'], {'how': '"""left"""', 'on': "['air_store_id', 'visit_date']"}), "(train, data[df], how='left', on=['air_store_id', 'visit_date'])\n", (5593, 5657), True, 'import pandas as pd\n'), ((5669, 5740), 'pandas.merge', 'pd.merge', (['test', 'data[df]'], {'how': '"""left"""', 'on': "['air_store_id', 'visit_date']"}), "(test, data[df], how='left', on=['air_store_id', 'visit_date'])\n", (5677, 5740), True, 'import pandas as pd\n'), ((7638, 7672), 'numpy.log1p', 'np.log1p', (["train['visitors'].values"], {}), "(train['visitors'].values)\n", (7646, 7672), True, 'import numpy as np\n'), ((7697, 7731), 'numpy.log1p', 'np.log1p', (["train['visitors'].values"], {}), "(train['visitors'].values)\n", (7705, 7731), True, 'import numpy as np\n'), ((7756, 7790), 'numpy.log1p', 'np.log1p', (["train['visitors'].values"], {}), "(train['visitors'].values)\n", (7764, 7790), True, 'import numpy as np\n'), ((8586, 8601), 'pandas.read_csv', 'pd.read_csv', (['fn'], {}), '(fn)\n', (8597, 8601), True, 'import pandas as pd\n'), ((7950, 7984), 'numpy.log1p', 'np.log1p', (["train['visitors'].values"], {}), "(train['visitors'].values)\n", (7958, 7984), True, 'import numpy as np\n'), ((8037, 8071), 'numpy.log1p', 'np.log1p', (["train['visitors'].values"], {}), "(train['visitors'].values)\n", (8045, 8071), True, 'import numpy as np\n'), ((8117, 8151), 'numpy.log1p', 'np.log1p', (["train['visitors'].values"], {}), "(train['visitors'].values)\n", (8125, 8151), True, 'import numpy as np\n'), ((8339, 8365), 'numpy.expm1', 'np.expm1', (["test['visitors']"], {}), "(test['visitors'])\n", (8347, 8365), True, 'import numpy as np\n'), ((8611, 8638), 'glob.glob', 'glob.glob', (['"""../input/*.csv"""'], {}), "('../input/*.csv')\n", (8620, 8638), False, 'import glob, re\n'), ((1023, 1060), 'pandas.read_csv', 'pd.read_csv', (['"""../input/date_info.csv"""'], {}), "('../input/date_info.csv')\n", (1034, 1060), True, 'import pandas as pd\n'), ((8539, 8573), 're.search', 're.search', (['"""/([^/\\\\.]*)\\\\.csv"""', 'fn'], {}), "('/([^/\\\\.]*)\\\\.csv', fn)\n", (8548, 8573), False, 'import glob, re\n')]
import sys from setuptools import setup, find_packages, Extension import numpy as np if '--use-cython' in sys.argv: USE_CYTHON = True sys.argv.remove('--use-cython') else: USE_CYTHON = False ext = '.pyx' if USE_CYTHON else '.c' # cppext = '' if USE_CYTHON else 'pp' extensions = [ Extension( "deepgraph._triu_indices", ["deepgraph/_triu_indices" + ext], include_dirs=[np.get_include()], # language='c++', ), Extension( "deepgraph._find_selected_indices", ["deepgraph/_find_selected_indices" + ext], include_dirs=[np.get_include()] ) ] if USE_CYTHON: from Cython.Build import cythonize extensions = cythonize( extensions, compiler_directives={'language_level': sys.version_info[0]} ) setup( name="DeepGraph", version='0.2.3', packages=find_packages(), author="<NAME>", author_email="<EMAIL>", url='https://github.com/deepgraph/deepgraph/', download_url='https://github.com/deepgraph/deepgraph/tarball/v0.2.3', description=("Analyze Data with Pandas-based Networks."), long_description=open('README.rst').read(), install_requires=['numpy>=1.6', 'pandas>=0.17.0'], license="BSD", classifiers=[ 'License :: OSI Approved :: BSD License', 'Operating System :: OS Independent', 'Programming Language :: Python :: 2', 'Programming Language :: Python :: 2.7', 'Programming Language :: Python :: 3', 'Programming Language :: Python :: 3.4', 'Programming Language :: Python :: 3.5', 'Programming Language :: Python :: 3.6', 'Programming Language :: Python :: 3.7', 'Programming Language :: Python :: 3.8', 'Programming Language :: Cython', 'Topic :: Software Development :: Libraries :: Python Modules', 'Topic :: Scientific/Engineering :: Information Analysis', 'Topic :: Scientific/Engineering :: Mathematics', 'Topic :: Scientific/Engineering :: Physics'], ext_modules=extensions, package_data={'deepgraph': ['../tests/*.py', '../LICENSE.txt', './*.pyx', './*.c', './*.cpp', ]}, )
[ "Cython.Build.cythonize", "setuptools.find_packages", "sys.argv.remove", "numpy.get_include" ]
[((143, 174), 'sys.argv.remove', 'sys.argv.remove', (['"""--use-cython"""'], {}), "('--use-cython')\n", (158, 174), False, 'import sys\n'), ((694, 781), 'Cython.Build.cythonize', 'cythonize', (['extensions'], {'compiler_directives': "{'language_level': sys.version_info[0]}"}), "(extensions, compiler_directives={'language_level': sys.\n version_info[0]})\n", (703, 781), False, 'from Cython.Build import cythonize\n'), ((863, 878), 'setuptools.find_packages', 'find_packages', ([], {}), '()\n', (876, 878), False, 'from setuptools import setup, find_packages, Extension\n'), ((411, 427), 'numpy.get_include', 'np.get_include', ([], {}), '()\n', (425, 427), True, 'import numpy as np\n'), ((596, 612), 'numpy.get_include', 'np.get_include', ([], {}), '()\n', (610, 612), True, 'import numpy as np\n')]
import numpy as np import pandas as pd def mtr(val, brackets, rates): """Calculates the marginal tax rate applied to a value depending on a tax schedule. :param val: Value to assess tax on, e.g. wealth or income (list or Series). :param brackets: Left side of each bracket (list or Series). :param rates: Rate corresponding to each bracket. :returns: Series of the size of val representing the marginal tax rate. """ df_tax = pd.DataFrame({"brackets": brackets, "rates": rates}) df_tax["base_tax"] = ( df_tax.brackets.sub(df_tax.brackets.shift(fill_value=0)) .mul(df_tax.rates.shift(fill_value=0)) .cumsum() ) rows = df_tax.brackets.searchsorted(val, side="right") - 1 income_bracket_df = df_tax.loc[rows].reset_index(drop=True) return income_bracket_df.rates def tax_from_mtrs( val, brackets, rates, avoidance_rate=0, avoidance_elasticity=0, avoidance_elasticity_flat=0, ): """Calculates tax liability based on a marginal tax rate schedule. :param val: Value to assess tax on, e.g. wealth or income (list or Series). :param brackets: Left side of each bracket (list or Series). :param rates: Rate corresponding to each bracket. :param avoidance_rate: Constant avoidance/evasion rate in percentage terms. Defaults to zero. :param avoidance_elasticity: Avoidance/evasion elasticity. Response of log taxable value with respect to tax rate. Defaults to zero. Should be positive. :param avoidance_elasticity_flat: Response of taxable value with respect to tax rate. Use avoidance_elasticity in most cases. Defaults to zero. Should be positive. :returns: Series of tax liabilities with the same size as val. """ assert ( avoidance_rate == 0 or avoidance_elasticity == 0 or avoidance_elasticity_flat == 0 ), "Cannot supply multiple avoidance parameters." assert ( avoidance_elasticity >= 0 ), "Provide nonnegative avoidance_elasticity." df_tax = pd.DataFrame({"brackets": brackets, "rates": rates}) df_tax["base_tax"] = ( df_tax.brackets.sub(df_tax.brackets.shift(fill_value=0)) .mul(df_tax.rates.shift(fill_value=0)) .cumsum() ) if avoidance_rate == 0: # Only need MTRs if elasticity is supplied. mtrs = mtr(val, brackets, rates) if avoidance_elasticity > 0: avoidance_rate = 1 - np.exp(-avoidance_elasticity * mtrs) if avoidance_elasticity_flat > 0: avoidance_rate = avoidance_elasticity_flat * mtrs taxable = pd.Series(val) * (1 - avoidance_rate) rows = df_tax.brackets.searchsorted(taxable, side="right") - 1 income_bracket_df = df_tax.loc[rows].reset_index(drop=True) return ( pd.Series(taxable) .sub(income_bracket_df.brackets) .mul(income_bracket_df.rates) .add(income_bracket_df.base_tax) )
[ "pandas.DataFrame", "numpy.exp", "pandas.Series" ]
[((462, 514), 'pandas.DataFrame', 'pd.DataFrame', (["{'brackets': brackets, 'rates': rates}"], {}), "({'brackets': brackets, 'rates': rates})\n", (474, 514), True, 'import pandas as pd\n'), ((2255, 2307), 'pandas.DataFrame', 'pd.DataFrame', (["{'brackets': brackets, 'rates': rates}"], {}), "({'brackets': brackets, 'rates': rates})\n", (2267, 2307), True, 'import pandas as pd\n'), ((2794, 2808), 'pandas.Series', 'pd.Series', (['val'], {}), '(val)\n', (2803, 2808), True, 'import pandas as pd\n'), ((2647, 2683), 'numpy.exp', 'np.exp', (['(-avoidance_elasticity * mtrs)'], {}), '(-avoidance_elasticity * mtrs)\n', (2653, 2683), True, 'import numpy as np\n'), ((2984, 3002), 'pandas.Series', 'pd.Series', (['taxable'], {}), '(taxable)\n', (2993, 3002), True, 'import pandas as pd\n')]
import numpy as np class Agent(object): def __init__(self, k, policy, init_exploration, prior=0, gamma=None): self.policy = policy self.k = k self.prior = prior self.gamma = gamma self._value_estimates = prior * np.ones(self.k) # Estimated Mean reward self.action_attempts = np.zeros(self.k) self.t = 0 self.last_action = None self.init_exploration = init_exploration def reset(self): """ Resets the agent's memory to an initial state. """ self._value_estimates[:] = self.prior * np.ones(self.k) self.action_attempts[:] = np.zeros(self.k) self.last_action = None self.t = 0 def choose(self): if self.t < self.init_exploration: action = np.random.randint(self.k) else: action = self.policy.choose(self) self.last_action = action return action def observe(self, reward): # Updating value_estimates ! (calculating mean rewards) self.action_attempts[self.last_action] += 1 if self.gamma is None: g = 1 / self.action_attempts[self.last_action] else: g = self.gamma q = self._value_estimates[self.last_action] self._value_estimates[self.last_action] += g * (reward - q) self.t += 1 @property def value_estimates(self): return self._value_estimates class ContextualAgent(Agent): """ ( linUCB disjoint model) """ def __init__(self, k, d, policy, init_exploration, prior=0, gamma=None): super().__init__(k, policy, init_exploration, prior, gamma) self.d = d self.memory = {action: {'A': np.identity(self.d), 'b': np.zeros((self.d, 1))} for action in range(self.k)} self.states = np.array([]) self.reset() def reset(self): self._value_estimates[:] = self.prior * np.ones(self.k) self.action_attempts[:] = 0 self.last_action = None self.t = 0 self.memory = {action: {'A': np.identity(self.d), 'b': np.zeros((self.d, 1))} for action in range(self.k)} self.states = np.array([]) # FIXME def get_state(self, bandit): self.states = bandit.states for action, memory in self.memory.items(): A = memory['A'] b = memory['b'] A_inv = np.linalg.inv(A) theta_hat = np.dot(A_inv, b) x_t = self.states[action] self._value_estimates[action] = np.dot(x_t.T, theta_hat) def observe(self, reward): self.action_attempts[self.last_action] += 1 self.memory[self.last_action]['A'] += np.outer(self.states[self.last_action], self.states[self.last_action]) self.memory[self.last_action]['b'] += reward * self.states[self.last_action].reshape((self.d, 1)) self.t += 1
[ "numpy.identity", "numpy.ones", "numpy.array", "numpy.zeros", "numpy.outer", "numpy.random.randint", "numpy.linalg.inv", "numpy.dot" ]
[((330, 346), 'numpy.zeros', 'np.zeros', (['self.k'], {}), '(self.k)\n', (338, 346), True, 'import numpy as np\n'), ((646, 662), 'numpy.zeros', 'np.zeros', (['self.k'], {}), '(self.k)\n', (654, 662), True, 'import numpy as np\n'), ((1852, 1864), 'numpy.array', 'np.array', (['[]'], {}), '([])\n', (1860, 1864), True, 'import numpy as np\n'), ((2228, 2240), 'numpy.array', 'np.array', (['[]'], {}), '([])\n', (2236, 2240), True, 'import numpy as np\n'), ((2745, 2815), 'numpy.outer', 'np.outer', (['self.states[self.last_action]', 'self.states[self.last_action]'], {}), '(self.states[self.last_action], self.states[self.last_action])\n', (2753, 2815), True, 'import numpy as np\n'), ((258, 273), 'numpy.ones', 'np.ones', (['self.k'], {}), '(self.k)\n', (265, 273), True, 'import numpy as np\n'), ((596, 611), 'numpy.ones', 'np.ones', (['self.k'], {}), '(self.k)\n', (603, 611), True, 'import numpy as np\n'), ((801, 826), 'numpy.random.randint', 'np.random.randint', (['self.k'], {}), '(self.k)\n', (818, 826), True, 'import numpy as np\n'), ((1956, 1971), 'numpy.ones', 'np.ones', (['self.k'], {}), '(self.k)\n', (1963, 1971), True, 'import numpy as np\n'), ((2450, 2466), 'numpy.linalg.inv', 'np.linalg.inv', (['A'], {}), '(A)\n', (2463, 2466), True, 'import numpy as np\n'), ((2491, 2507), 'numpy.dot', 'np.dot', (['A_inv', 'b'], {}), '(A_inv, b)\n', (2497, 2507), True, 'import numpy as np\n'), ((2590, 2614), 'numpy.dot', 'np.dot', (['x_t.T', 'theta_hat'], {}), '(x_t.T, theta_hat)\n', (2596, 2614), True, 'import numpy as np\n'), ((1720, 1739), 'numpy.identity', 'np.identity', (['self.d'], {}), '(self.d)\n', (1731, 1739), True, 'import numpy as np\n'), ((1778, 1799), 'numpy.zeros', 'np.zeros', (['(self.d, 1)'], {}), '((self.d, 1))\n', (1786, 1799), True, 'import numpy as np\n'), ((2096, 2115), 'numpy.identity', 'np.identity', (['self.d'], {}), '(self.d)\n', (2107, 2115), True, 'import numpy as np\n'), ((2154, 2175), 'numpy.zeros', 'np.zeros', (['(self.d, 1)'], {}), '((self.d, 1))\n', (2162, 2175), True, 'import numpy as np\n')]
import pandas as pd import numpy as np from trading_gym.envs.portfolio_gym.portfolio_gym import PortfolioTradingGym np.random.seed(64) def create_mock_data(order_book_ids, start_date="2019-01-01", end_date="2022-01-02", number_feature=3): trading_dates = pd.date_range(start=start_date, end=end_date, freq="D") number = len(trading_dates) * len(order_book_ids) multi_index = pd.MultiIndex.from_product([order_book_ids, trading_dates], names=["order_book_id", "datetime"]) mock_data = pd.DataFrame(np.random.randn(number, number_feature + 1), index=multi_index, columns=["feature1", "feature2", "feature3", "returns"]) mock_data["returns"] = mock_data["returns"] / 100 # 当期收益率 mock_data["returns"] = round(mock_data["returns"], 4) return mock_data def test_single_array_R(): order_book_ids = ["000001.XSHE"] mock_data = create_mock_data(order_book_ids=order_book_ids, start_date="2019-01-01", end_date="2019-01-14") sequence_window = 1 env = PortfolioTradingGym(data_df=mock_data, sequence_window=sequence_window, add_cash=True) state = env.reset() h_t_list = [] orderlist = [0.5, 0.8, 0.4, 1.0, 0.0, -0.5, -0.7, -0.4, -1.0, 0.3] for i in range(len(orderlist)): next_state, reward, done, info = env.step([orderlist[i], 0]) h_t_list.append(info["h_t"]) ''' 000001.XSHE 2019-01-01 0.0219 2019-01-02 -0.0103 2019-01-03 0.0175 2019-01-04 -0.0017 2019-01-05 -0.0039 2019-01-06 0.0059 2019-01-07 -0.0049 2019-01-08 -0.0003 2019-01-09 -0.0136 2019-01-10 0.0068 2019-01-11 0.0077 2019-01-12 0.0136 2019-01-13 -0.0022 2019-01-14 -0.0012 ''' expected_h_t = ([494850, 500050], [809848.6, 198999.898], [402853.3822, 605369.6309], [1004290.9, 0], [0, 1004391.359], [-499734.9212, 1506737.699], [-704690.4898, 1712075.95], [-397473.9834, 1410480.594], [-1019895.146, 2026216.001], [304220.9, 704495.1425]) np.testing.assert_almost_equal(h_t_list, expected_h_t, decimal=1) def test_single_number_R(): order_book_ids = ["000001.XSHE"] mock_data = create_mock_data(order_book_ids=order_book_ids, start_date="2019-01-01", end_date="2019-01-14") sequence_window = 1 env = PortfolioTradingGym(data_df=mock_data, sequence_window=sequence_window, add_cash=True) state = env.reset() h_t_list = [] orderlist = [0.5, 0.8, 0.4, 1.0, 0.0, -0.5, -0.7, -0.4, -1.0, 0.3] for i in range(len(orderlist)): next_state, reward, done, info = env.step(orderlist[i]) h_t_list.append(info["h_t"]) expected_h_t = ([494850, 500050], [809848.6, 198999.898], [402853.3822, 605369.6309], [1004290.9, 0], [0, 1004391.359], [-499734.9212, 1506737.699], [-704690.4898, 1712075.95], [-397473.9834, 1410480.594], [-1019895.146, 2026216.001], [304220.9, 704495.1425]) np.testing.assert_almost_equal(h_t_list, expected_h_t, decimal=1) def test_single_cashfalse_R(): order_book_ids = ["000001.XSHE"] mock_data = create_mock_data(order_book_ids=order_book_ids, start_date="2019-01-01", end_date="2019-01-14") sequence_window = 1 env = PortfolioTradingGym(data_df=mock_data, sequence_window=sequence_window, add_cash=False) state = env.reset() h_t_list = [] orderlist = [0.5, 0.8, 0.4, 1.0, 0.0, -0.5, -0.7, -0.4, -1.0, 0.3] for i in range(len(orderlist)): next_state, reward, done, info = env.step(orderlist[i]) h_t_list.append(info["h_t"]) expected_h_t = ([494850], [809848.6], [402853.3822],[1004290.9], [0], [-499734.9212], [-704690.4898], [-397473.9834], [-1019895.146], [304220.9]) np.testing.assert_almost_equal(h_t_list, expected_h_t, decimal=1) if __name__ == "__main__": #test_single_array_R() test_single_number_R() #test_single_cashfalse_R() ''' Results here are same as those in test_single_stock.py. '''
[ "pandas.MultiIndex.from_product", "trading_gym.envs.portfolio_gym.portfolio_gym.PortfolioTradingGym", "numpy.testing.assert_almost_equal", "numpy.random.seed", "numpy.random.randn", "pandas.date_range" ]
[((117, 135), 'numpy.random.seed', 'np.random.seed', (['(64)'], {}), '(64)\n', (131, 135), True, 'import numpy as np\n'), ((262, 317), 'pandas.date_range', 'pd.date_range', ([], {'start': 'start_date', 'end': 'end_date', 'freq': '"""D"""'}), "(start=start_date, end=end_date, freq='D')\n", (275, 317), True, 'import pandas as pd\n'), ((390, 491), 'pandas.MultiIndex.from_product', 'pd.MultiIndex.from_product', (['[order_book_ids, trading_dates]'], {'names': "['order_book_id', 'datetime']"}), "([order_book_ids, trading_dates], names=[\n 'order_book_id', 'datetime'])\n", (416, 491), True, 'import pandas as pd\n'), ((1020, 1110), 'trading_gym.envs.portfolio_gym.portfolio_gym.PortfolioTradingGym', 'PortfolioTradingGym', ([], {'data_df': 'mock_data', 'sequence_window': 'sequence_window', 'add_cash': '(True)'}), '(data_df=mock_data, sequence_window=sequence_window,\n add_cash=True)\n', (1039, 1110), False, 'from trading_gym.envs.portfolio_gym.portfolio_gym import PortfolioTradingGym\n'), ((2194, 2259), 'numpy.testing.assert_almost_equal', 'np.testing.assert_almost_equal', (['h_t_list', 'expected_h_t'], {'decimal': '(1)'}), '(h_t_list, expected_h_t, decimal=1)\n', (2224, 2259), True, 'import numpy as np\n'), ((2473, 2563), 'trading_gym.envs.portfolio_gym.portfolio_gym.PortfolioTradingGym', 'PortfolioTradingGym', ([], {'data_df': 'mock_data', 'sequence_window': 'sequence_window', 'add_cash': '(True)'}), '(data_df=mock_data, sequence_window=sequence_window,\n add_cash=True)\n', (2492, 2563), False, 'from trading_gym.envs.portfolio_gym.portfolio_gym import PortfolioTradingGym\n'), ((3118, 3183), 'numpy.testing.assert_almost_equal', 'np.testing.assert_almost_equal', (['h_t_list', 'expected_h_t'], {'decimal': '(1)'}), '(h_t_list, expected_h_t, decimal=1)\n', (3148, 3183), True, 'import numpy as np\n'), ((3400, 3491), 'trading_gym.envs.portfolio_gym.portfolio_gym.PortfolioTradingGym', 'PortfolioTradingGym', ([], {'data_df': 'mock_data', 'sequence_window': 'sequence_window', 'add_cash': '(False)'}), '(data_df=mock_data, sequence_window=sequence_window,\n add_cash=False)\n', (3419, 3491), False, 'from trading_gym.envs.portfolio_gym.portfolio_gym import PortfolioTradingGym\n'), ((3912, 3977), 'numpy.testing.assert_almost_equal', 'np.testing.assert_almost_equal', (['h_t_list', 'expected_h_t'], {'decimal': '(1)'}), '(h_t_list, expected_h_t, decimal=1)\n', (3942, 3977), True, 'import numpy as np\n'), ((516, 559), 'numpy.random.randn', 'np.random.randn', (['number', '(number_feature + 1)'], {}), '(number, number_feature + 1)\n', (531, 559), True, 'import numpy as np\n')]
import numpy as np from Common import DIRECTION, TURN, BOARD_OBJECT, ALL_TURNS, ALL_DIRECTIONS, DIRECTION_UNIT_VECTORS, DIRECTION_MARKERS, Log from GameStatistics import GameStatistics from Network import Network class SnakeGame: @staticmethod def play(board_size, predictor: Network, max_moves, print_sensory = False): game = SnakeGame(board_size, max_moves=max_moves) while True: inputs = game.get_sensory_inputs() if(print_sensory): Log(str(inputs)) index = predictor.predict(inputs) turn = ALL_TURNS[index] if(game.move(turn)): break return game @staticmethod def __init_free_positions(D: dict, board_size): D.clear() m = board_size x, y = np.meshgrid(range(0, m), range(0, m)) coordinates = np.concatenate((x.reshape(-1, 1), y.reshape(-1, 1)), axis=1) for point in coordinates: D[tuple(point)] = 0 def __init__(self, board_size, max_moves): self.__max_moves = max_moves # defines maximum number of moves consecutively without eating apple self.__no_apple_counter = 0 # number of consecutive moves without eating apple self.__history = [] # list of game states for each move self.__statistics = GameStatistics() # statistics object that holds count for left/right/forward moves and snake size self.__body = [] # list of snakes body part positions [x,y]. first is tail, last is head self.__apple = None # position of apple [x,y] self.__free_positions = dict() # dictionary of free positions. keys are positions (x,y) self.__board_size = board_size # side length of board self.__proximities = [1/i for i in range(1,board_size+1)] # list of proximities denoting how much close to an object (apple, wall or body) self.__init_free_positions(self.__free_positions, board_size) self.__head_direction = np.random.choice(ALL_DIRECTIONS) head = np.random.randint(0, board_size, 2) self.__put_head(head) apple = self.__get_random_free_position() self.__put_apple(apple) self.__save_state() def __turn_head(self, turn: TURN): self.__statistics.turn(turn) current_direction_value = self.__head_direction.value self.__head_direction = ALL_DIRECTIONS[(current_direction_value + turn.value) % 4] def __get_random_free_position(self): frees = list(self.__free_positions.keys()) n_free = len(frees) if n_free == 0: return None index = int(np.random.randint(0, n_free)) free = frees[index] # todo: faster way to access random item in dictionary return np.array(free) def __is_free(self, position): return tuple(position) in self.__free_positions def __set_free(self, position): """ Adds given position to list of empty positions. """ self.__free_positions[tuple(position)] = 0 def __set_occupied(self, position): """ Removes given position from list of empty positions. """ if position is None: return del self.__free_positions[tuple(position)] def __put_apple(self, position): self.__apple = position self.__set_occupied(position) def __put_head(self, position): self.__body.append(position) self.__set_occupied(position) Log("head at: "+str(position) + " Looking: " + DIRECTION_MARKERS[self.__head_direction], level=0) def __move_head(self, position, is_apple_eaten): self.__body.append(position) # add new head to body parts if is_apple_eaten == False: self.__no_apple_counter += 1 # increase no apple count tail = self.__body.pop(0) # remove old tail self.__set_free(tail) self.__set_occupied(position) else: self.__no_apple_counter = 0 # reset no apple counter self.__statistics.eat() # update statistics new_apple = self.__get_random_free_position() self.__put_apple(new_apple) # put new apple on board Log("head at: "+str(position) + " Looking: " + DIRECTION_MARKERS[self.__head_direction], level=0) def __is_outside(self, position): m = self.__board_size x, y = position return x < 0 or x >= m or y < 0 or y >= m def __is_perfect_game(self): return len(self.__body) == (self.__board_size**2) def __save_state(self): if self.__apple is None: snapshot = self.__body[:], None, self.__head_direction else: snapshot = self.__body[:], self.__apple[:], self.__head_direction self.__history.append(snapshot) def __scan_direction(self, position, direction: DIRECTION): """ Scans obstacles in given direction starting from position. Returns [apple_proximity, obstacle_proximity] """ i = 0 unit_step = DIRECTION_UNIT_VECTORS[direction] next = position while True: next = next + unit_step proximity = self.__proximities[i] if(self.__is_outside(next)): # wall return [0, proximity] elif(np.array_equal(next, self.__apple)): # apple return [proximity, 0] elif(self.__is_free(next) == False): # body return [0, proximity] i += 1 ### PUBLIC METHODS def get_sensory_inputs(self): """ Scans 3 directions relative to head and gets proximity measures in order: Forward Apple, Forward Obstacle, Left Apple, Left Obstacle, Right Apple, Right Obstacle Returns [f_apple, f_obstacle, l_apple, l_obstacle, r_apple, r_obstacle] """ head = self.__body[-1] d_forward = self.__head_direction d_left = ALL_DIRECTIONS[(d_forward.value + TURN.LEFT.value) % 4] d_right = ALL_DIRECTIONS[(d_forward.value + TURN.RIGHT.value) % 4] s_forward = self.__scan_direction(head, d_forward) s_left= self.__scan_direction(head, d_left) s_right = self.__scan_direction(head, d_right) return s_forward + s_left + s_right def get_statistics(self): """ Returns move counts for left,right,forward and snake size """ return self.__statistics.get() def get_fitness(self): l,r,f,size = self.get_statistics() bias_coef = 1 if(l == 0 or r == 0): bias_coef = 0.1 #total_moves = l+r+f #delta = abs(l-r) score = (bias_coef * size) #- delta*(0.1) return score def get_board_size(self): return self.__board_size def get_history(self): return self.__history def move(self, turn: TURN): Log("Command: "+str(turn), level=0) is_finished = False # change snake heads direction self.__turn_head(turn) # calculate heads next position head = self.__body[-1] next = head + DIRECTION_UNIT_VECTORS[self.__head_direction] # check if new head position is inside borders if self.__is_outside(next): is_finished = True # check if new head overlaps with apple elif np.array_equal(next, self.__apple): self.__move_head(next, is_apple_eaten=True) if(self.__is_perfect_game()): is_finished = True # check if new head overlaps with body elif self.__is_free(next) == False: is_finished = True # head is moving to empty else: self.__move_head(next, is_apple_eaten=False) if(self.__no_apple_counter >= self.__max_moves): is_finished = True self.__save_state() return is_finished
[ "numpy.random.choice", "GameStatistics.GameStatistics", "numpy.array", "numpy.random.randint", "numpy.array_equal" ]
[((1382, 1398), 'GameStatistics.GameStatistics', 'GameStatistics', ([], {}), '()\n', (1396, 1398), False, 'from GameStatistics import GameStatistics\n'), ((2110, 2142), 'numpy.random.choice', 'np.random.choice', (['ALL_DIRECTIONS'], {}), '(ALL_DIRECTIONS)\n', (2126, 2142), True, 'import numpy as np\n'), ((2159, 2194), 'numpy.random.randint', 'np.random.randint', (['(0)', 'board_size', '(2)'], {}), '(0, board_size, 2)\n', (2176, 2194), True, 'import numpy as np\n'), ((2901, 2915), 'numpy.array', 'np.array', (['free'], {}), '(free)\n', (2909, 2915), True, 'import numpy as np\n'), ((2765, 2793), 'numpy.random.randint', 'np.random.randint', (['(0)', 'n_free'], {}), '(0, n_free)\n', (2782, 2793), True, 'import numpy as np\n'), ((7598, 7632), 'numpy.array_equal', 'np.array_equal', (['next', 'self.__apple'], {}), '(next, self.__apple)\n', (7612, 7632), True, 'import numpy as np\n'), ((5519, 5553), 'numpy.array_equal', 'np.array_equal', (['next', 'self.__apple'], {}), '(next, self.__apple)\n', (5533, 5553), True, 'import numpy as np\n')]
import os import uuid from datetime import datetime import numpy as np import pandas as pd import pytest import pytz from google.protobuf.duration_pb2 import Duration from pandas.testing import assert_frame_equal from feast.client import Client from feast.data_source import BigQuerySource, FileSource, KafkaSource from feast.entity import Entity from feast.feature import Feature from feast.feature_table import FeatureTable from feast.value_type import ValueType from feast.wait import wait_retry_backoff DIR_PATH = os.path.dirname(os.path.realpath(__file__)) PROJECT_NAME = "basic_" + uuid.uuid4().hex.upper()[0:6] SUFFIX = str(int(datetime.now().timestamp())) @pytest.fixture(scope="module") def client(pytestconfig): core_url = pytestconfig.getoption("core_url") serving_url = pytestconfig.getoption("serving_url") client = Client(core_url=core_url, serving_url=serving_url,) client.set_project(PROJECT_NAME) return client @pytest.fixture def bq_table_id(): return f"kf-feast:feaste2e.table{SUFFIX}" @pytest.fixture def customer_entity(): return Entity( name="customer_id", description="Customer entity for rides", value_type=ValueType.STRING, labels={"team": "customer_service", "common_key": "common_val"}, ) @pytest.fixture def driver_entity(): return Entity( name="driver_id", description="Driver entity for car rides", value_type=ValueType.STRING, labels={"team": "matchmaking", "common_key": "common_val"}, ) @pytest.fixture def basic_featuretable(): batch_source = FileSource( field_mapping={ "dev_entity": "dev_entity_field", "dev_feature_float": "dev_feature_float_field", "dev_feature_string": "dev_feature_string_field", }, file_format="PARQUET", file_url="gs://example/feast/*", timestamp_column="datetime_col", date_partition_column="datetime", ) stream_source = KafkaSource( field_mapping={ "dev_entity": "dev_entity_field", "dev_feature_float": "dev_feature_float_field", "dev_feature_string": "dev_feature_string_field", }, bootstrap_servers="localhost:9094", class_path="random/path/to/class", topic="test_topic", timestamp_column="datetime_col", ) return FeatureTable( name="basic_featuretable", entities=["driver_id", "customer_id"], features=[ Feature(name="dev_feature_float", dtype=ValueType.FLOAT), Feature(name="dev_feature_string", dtype=ValueType.STRING), ], max_age=Duration(seconds=3600), batch_source=batch_source, stream_source=stream_source, labels={"key1": "val1", "key2": "val2"}, ) @pytest.fixture def bq_dataset(): N_ROWS = 100 time_offset = datetime.utcnow().replace(tzinfo=pytz.utc) return pd.DataFrame( { "datetime": [time_offset] * N_ROWS, "dev_feature_float": [np.float(row) for row in range(N_ROWS)], "dev_feature_string": ["feat_" + str(row) for row in range(N_ROWS)], } ) @pytest.fixture def bq_featuretable(bq_table_id): batch_source = BigQuerySource(table_ref=bq_table_id, timestamp_column="datetime",) return FeatureTable( name="basic_featuretable", entities=["driver_id", "customer_id"], features=[ Feature(name="dev_feature_float", dtype=ValueType.FLOAT), Feature(name="dev_feature_string", dtype=ValueType.STRING), ], max_age=Duration(seconds=3600), batch_source=batch_source, ) @pytest.fixture def alltypes_entity(): return Entity( name="alltypes_id", description="Driver entity for car rides", value_type=ValueType.STRING, labels={"cat": "alltypes"}, ) @pytest.fixture def alltypes_featuretable(): batch_source = FileSource( file_format="parquet", file_url="file://feast/*", timestamp_column="ts_col", date_partition_column="date_partition_col", ) return FeatureTable( name="alltypes", entities=["alltypes_id"], features=[ Feature(name="float_feature", dtype=ValueType.FLOAT), Feature(name="int64_feature", dtype=ValueType.INT64), Feature(name="int32_feature", dtype=ValueType.INT32), Feature(name="string_feature", dtype=ValueType.STRING), Feature(name="bytes_feature", dtype=ValueType.BYTES), Feature(name="bool_feature", dtype=ValueType.BOOL), Feature(name="double_feature", dtype=ValueType.DOUBLE), Feature(name="double_list_feature", dtype=ValueType.DOUBLE_LIST), Feature(name="float_list_feature", dtype=ValueType.FLOAT_LIST), Feature(name="int64_list_feature", dtype=ValueType.INT64_LIST), Feature(name="int32_list_feature", dtype=ValueType.INT32_LIST), Feature(name="string_list_feature", dtype=ValueType.STRING_LIST), Feature(name="bytes_list_feature", dtype=ValueType.BYTES_LIST), Feature(name="bool_list_feature", dtype=ValueType.BOOL_LIST), ], max_age=Duration(seconds=3600), batch_source=batch_source, labels={"cat": "alltypes"}, ) def test_get_list_basic( client: Client, customer_entity: Entity, driver_entity: Entity, basic_featuretable: FeatureTable, ): # ApplyEntity client.apply_entity(customer_entity) client.apply_entity(driver_entity) # GetEntity Check assert client.get_entity(name="customer_id") == customer_entity assert client.get_entity(name="driver_id") == driver_entity # ListEntities Check common_filtering_labels = {"common_key": "common_val"} matchmaking_filtering_labels = {"team": "matchmaking"} actual_common_entities = client.list_entities(labels=common_filtering_labels) actual_matchmaking_entities = client.list_entities( labels=matchmaking_filtering_labels ) assert len(actual_common_entities) == 2 assert len(actual_matchmaking_entities) == 1 # ApplyFeatureTable client.apply_feature_table(basic_featuretable) # GetFeatureTable Check actual_get_feature_table = client.get_feature_table(name="basic_featuretable") assert actual_get_feature_table == basic_featuretable # ListFeatureTables Check actual_list_feature_table = [ ft for ft in client.list_feature_tables() if ft.name == "basic_featuretable" ][0] assert actual_list_feature_table == basic_featuretable def test_get_list_alltypes( client: Client, alltypes_entity: Entity, alltypes_featuretable: FeatureTable ): # ApplyEntity client.apply_entity(alltypes_entity) # GetEntity Check assert client.get_entity(name="alltypes_id") == alltypes_entity # ListEntities Check alltypes_filtering_labels = {"cat": "alltypes"} actual_alltypes_entities = client.list_entities(labels=alltypes_filtering_labels) assert len(actual_alltypes_entities) == 1 # ApplyFeatureTable client.apply_feature_table(alltypes_featuretable) # GetFeatureTable Check actual_get_feature_table = client.get_feature_table(name="alltypes") assert actual_get_feature_table == alltypes_featuretable # ListFeatureTables Check actual_list_feature_table = [ ft for ft in client.list_feature_tables() if ft.name == "alltypes" ][0] assert actual_list_feature_table == alltypes_featuretable @pytest.mark.bq def test_ingest( client: Client, customer_entity: Entity, driver_entity: Entity, bq_featuretable: FeatureTable, bq_dataset: pd.DataFrame, bq_table_id: str, ): gcp_project, _ = bq_table_id.split(":") bq_table_id = bq_table_id.replace(":", ".") # ApplyEntity client.apply_entity(customer_entity) client.apply_entity(driver_entity) # ApplyFeatureTable client.apply_feature_table(bq_featuretable) client.ingest(bq_featuretable, bq_dataset, timeout=120) from google.api_core.exceptions import NotFound from google.cloud import bigquery bq_client = bigquery.Client(project=gcp_project) # Poll BQ for table until the table has been created def try_get_table(): table_exist = False table_resp = None try: table_resp = bq_client.get_table(bq_table_id) if table_resp and table_resp.table_id == bq_table_id.split(".")[-1]: table_exist = True except NotFound: pass return table_resp, table_exist wait_retry_backoff( retry_fn=try_get_table, timeout_secs=30, timeout_msg="Timed out trying to get bigquery table", ) query_string = f"SELECT * FROM `{bq_table_id}`" job = bq_client.query(query_string) query_df = job.to_dataframe() assert_frame_equal(query_df, bq_dataset) bq_client.delete_table(bq_table_id, not_found_ok=True)
[ "google.protobuf.duration_pb2.Duration", "numpy.float", "datetime.datetime.utcnow", "feast.feature.Feature", "feast.data_source.BigQuerySource", "feast.client.Client", "feast.entity.Entity", "feast.data_source.FileSource", "feast.wait.wait_retry_backoff", "os.path.realpath", "feast.data_source.K...
[((670, 700), 'pytest.fixture', 'pytest.fixture', ([], {'scope': '"""module"""'}), "(scope='module')\n", (684, 700), False, 'import pytest\n'), ((537, 563), 'os.path.realpath', 'os.path.realpath', (['__file__'], {}), '(__file__)\n', (553, 563), False, 'import os\n'), ((847, 897), 'feast.client.Client', 'Client', ([], {'core_url': 'core_url', 'serving_url': 'serving_url'}), '(core_url=core_url, serving_url=serving_url)\n', (853, 897), False, 'from feast.client import Client\n'), ((1091, 1260), 'feast.entity.Entity', 'Entity', ([], {'name': '"""customer_id"""', 'description': '"""Customer entity for rides"""', 'value_type': 'ValueType.STRING', 'labels': "{'team': 'customer_service', 'common_key': 'common_val'}"}), "(name='customer_id', description='Customer entity for rides',\n value_type=ValueType.STRING, labels={'team': 'customer_service',\n 'common_key': 'common_val'})\n", (1097, 1260), False, 'from feast.entity import Entity\n'), ((1342, 1506), 'feast.entity.Entity', 'Entity', ([], {'name': '"""driver_id"""', 'description': '"""Driver entity for car rides"""', 'value_type': 'ValueType.STRING', 'labels': "{'team': 'matchmaking', 'common_key': 'common_val'}"}), "(name='driver_id', description='Driver entity for car rides',\n value_type=ValueType.STRING, labels={'team': 'matchmaking',\n 'common_key': 'common_val'})\n", (1348, 1506), False, 'from feast.entity import Entity\n'), ((1601, 1899), 'feast.data_source.FileSource', 'FileSource', ([], {'field_mapping': "{'dev_entity': 'dev_entity_field', 'dev_feature_float':\n 'dev_feature_float_field', 'dev_feature_string': 'dev_feature_string_field'\n }", 'file_format': '"""PARQUET"""', 'file_url': '"""gs://example/feast/*"""', 'timestamp_column': '"""datetime_col"""', 'date_partition_column': '"""datetime"""'}), "(field_mapping={'dev_entity': 'dev_entity_field',\n 'dev_feature_float': 'dev_feature_float_field', 'dev_feature_string':\n 'dev_feature_string_field'}, file_format='PARQUET', file_url=\n 'gs://example/feast/*', timestamp_column='datetime_col',\n date_partition_column='datetime')\n", (1611, 1899), False, 'from feast.data_source import BigQuerySource, FileSource, KafkaSource\n'), ((1997, 2297), 'feast.data_source.KafkaSource', 'KafkaSource', ([], {'field_mapping': "{'dev_entity': 'dev_entity_field', 'dev_feature_float':\n 'dev_feature_float_field', 'dev_feature_string': 'dev_feature_string_field'\n }", 'bootstrap_servers': '"""localhost:9094"""', 'class_path': '"""random/path/to/class"""', 'topic': '"""test_topic"""', 'timestamp_column': '"""datetime_col"""'}), "(field_mapping={'dev_entity': 'dev_entity_field',\n 'dev_feature_float': 'dev_feature_float_field', 'dev_feature_string':\n 'dev_feature_string_field'}, bootstrap_servers='localhost:9094',\n class_path='random/path/to/class', topic='test_topic', timestamp_column\n ='datetime_col')\n", (2008, 2297), False, 'from feast.data_source import BigQuerySource, FileSource, KafkaSource\n'), ((3261, 3327), 'feast.data_source.BigQuerySource', 'BigQuerySource', ([], {'table_ref': 'bq_table_id', 'timestamp_column': '"""datetime"""'}), "(table_ref=bq_table_id, timestamp_column='datetime')\n", (3275, 3327), False, 'from feast.data_source import BigQuerySource, FileSource, KafkaSource\n'), ((3741, 3871), 'feast.entity.Entity', 'Entity', ([], {'name': '"""alltypes_id"""', 'description': '"""Driver entity for car rides"""', 'value_type': 'ValueType.STRING', 'labels': "{'cat': 'alltypes'}"}), "(name='alltypes_id', description='Driver entity for car rides',\n value_type=ValueType.STRING, labels={'cat': 'alltypes'})\n", (3747, 3871), False, 'from feast.entity import Entity\n'), ((3973, 4108), 'feast.data_source.FileSource', 'FileSource', ([], {'file_format': '"""parquet"""', 'file_url': '"""file://feast/*"""', 'timestamp_column': '"""ts_col"""', 'date_partition_column': '"""date_partition_col"""'}), "(file_format='parquet', file_url='file://feast/*',\n timestamp_column='ts_col', date_partition_column='date_partition_col')\n", (3983, 4108), False, 'from feast.data_source import BigQuerySource, FileSource, KafkaSource\n'), ((8221, 8257), 'google.cloud.bigquery.Client', 'bigquery.Client', ([], {'project': 'gcp_project'}), '(project=gcp_project)\n', (8236, 8257), False, 'from google.cloud import bigquery\n'), ((8670, 8788), 'feast.wait.wait_retry_backoff', 'wait_retry_backoff', ([], {'retry_fn': 'try_get_table', 'timeout_secs': '(30)', 'timeout_msg': '"""Timed out trying to get bigquery table"""'}), "(retry_fn=try_get_table, timeout_secs=30, timeout_msg=\n 'Timed out trying to get bigquery table')\n", (8688, 8788), False, 'from feast.wait import wait_retry_backoff\n'), ((8948, 8988), 'pandas.testing.assert_frame_equal', 'assert_frame_equal', (['query_df', 'bq_dataset'], {}), '(query_df, bq_dataset)\n', (8966, 8988), False, 'from pandas.testing import assert_frame_equal\n'), ((2670, 2692), 'google.protobuf.duration_pb2.Duration', 'Duration', ([], {'seconds': '(3600)'}), '(seconds=3600)\n', (2678, 2692), False, 'from google.protobuf.duration_pb2 import Duration\n'), ((2892, 2909), 'datetime.datetime.utcnow', 'datetime.utcnow', ([], {}), '()\n', (2907, 2909), False, 'from datetime import datetime\n'), ((3624, 3646), 'google.protobuf.duration_pb2.Duration', 'Duration', ([], {'seconds': '(3600)'}), '(seconds=3600)\n', (3632, 3646), False, 'from google.protobuf.duration_pb2 import Duration\n'), ((5272, 5294), 'google.protobuf.duration_pb2.Duration', 'Duration', ([], {'seconds': '(3600)'}), '(seconds=3600)\n', (5280, 5294), False, 'from google.protobuf.duration_pb2 import Duration\n'), ((638, 652), 'datetime.datetime.now', 'datetime.now', ([], {}), '()\n', (650, 652), False, 'from datetime import datetime\n'), ((2513, 2569), 'feast.feature.Feature', 'Feature', ([], {'name': '"""dev_feature_float"""', 'dtype': 'ValueType.FLOAT'}), "(name='dev_feature_float', dtype=ValueType.FLOAT)\n", (2520, 2569), False, 'from feast.feature import Feature\n'), ((2583, 2641), 'feast.feature.Feature', 'Feature', ([], {'name': '"""dev_feature_string"""', 'dtype': 'ValueType.STRING'}), "(name='dev_feature_string', dtype=ValueType.STRING)\n", (2590, 2641), False, 'from feast.feature import Feature\n'), ((3052, 3065), 'numpy.float', 'np.float', (['row'], {}), '(row)\n', (3060, 3065), True, 'import numpy as np\n'), ((3467, 3523), 'feast.feature.Feature', 'Feature', ([], {'name': '"""dev_feature_float"""', 'dtype': 'ValueType.FLOAT'}), "(name='dev_feature_float', dtype=ValueType.FLOAT)\n", (3474, 3523), False, 'from feast.feature import Feature\n'), ((3537, 3595), 'feast.feature.Feature', 'Feature', ([], {'name': '"""dev_feature_string"""', 'dtype': 'ValueType.STRING'}), "(name='dev_feature_string', dtype=ValueType.STRING)\n", (3544, 3595), False, 'from feast.feature import Feature\n'), ((4259, 4311), 'feast.feature.Feature', 'Feature', ([], {'name': '"""float_feature"""', 'dtype': 'ValueType.FLOAT'}), "(name='float_feature', dtype=ValueType.FLOAT)\n", (4266, 4311), False, 'from feast.feature import Feature\n'), ((4325, 4377), 'feast.feature.Feature', 'Feature', ([], {'name': '"""int64_feature"""', 'dtype': 'ValueType.INT64'}), "(name='int64_feature', dtype=ValueType.INT64)\n", (4332, 4377), False, 'from feast.feature import Feature\n'), ((4391, 4443), 'feast.feature.Feature', 'Feature', ([], {'name': '"""int32_feature"""', 'dtype': 'ValueType.INT32'}), "(name='int32_feature', dtype=ValueType.INT32)\n", (4398, 4443), False, 'from feast.feature import Feature\n'), ((4457, 4511), 'feast.feature.Feature', 'Feature', ([], {'name': '"""string_feature"""', 'dtype': 'ValueType.STRING'}), "(name='string_feature', dtype=ValueType.STRING)\n", (4464, 4511), False, 'from feast.feature import Feature\n'), ((4525, 4577), 'feast.feature.Feature', 'Feature', ([], {'name': '"""bytes_feature"""', 'dtype': 'ValueType.BYTES'}), "(name='bytes_feature', dtype=ValueType.BYTES)\n", (4532, 4577), False, 'from feast.feature import Feature\n'), ((4591, 4641), 'feast.feature.Feature', 'Feature', ([], {'name': '"""bool_feature"""', 'dtype': 'ValueType.BOOL'}), "(name='bool_feature', dtype=ValueType.BOOL)\n", (4598, 4641), False, 'from feast.feature import Feature\n'), ((4655, 4709), 'feast.feature.Feature', 'Feature', ([], {'name': '"""double_feature"""', 'dtype': 'ValueType.DOUBLE'}), "(name='double_feature', dtype=ValueType.DOUBLE)\n", (4662, 4709), False, 'from feast.feature import Feature\n'), ((4723, 4787), 'feast.feature.Feature', 'Feature', ([], {'name': '"""double_list_feature"""', 'dtype': 'ValueType.DOUBLE_LIST'}), "(name='double_list_feature', dtype=ValueType.DOUBLE_LIST)\n", (4730, 4787), False, 'from feast.feature import Feature\n'), ((4801, 4863), 'feast.feature.Feature', 'Feature', ([], {'name': '"""float_list_feature"""', 'dtype': 'ValueType.FLOAT_LIST'}), "(name='float_list_feature', dtype=ValueType.FLOAT_LIST)\n", (4808, 4863), False, 'from feast.feature import Feature\n'), ((4877, 4939), 'feast.feature.Feature', 'Feature', ([], {'name': '"""int64_list_feature"""', 'dtype': 'ValueType.INT64_LIST'}), "(name='int64_list_feature', dtype=ValueType.INT64_LIST)\n", (4884, 4939), False, 'from feast.feature import Feature\n'), ((4953, 5015), 'feast.feature.Feature', 'Feature', ([], {'name': '"""int32_list_feature"""', 'dtype': 'ValueType.INT32_LIST'}), "(name='int32_list_feature', dtype=ValueType.INT32_LIST)\n", (4960, 5015), False, 'from feast.feature import Feature\n'), ((5029, 5093), 'feast.feature.Feature', 'Feature', ([], {'name': '"""string_list_feature"""', 'dtype': 'ValueType.STRING_LIST'}), "(name='string_list_feature', dtype=ValueType.STRING_LIST)\n", (5036, 5093), False, 'from feast.feature import Feature\n'), ((5107, 5169), 'feast.feature.Feature', 'Feature', ([], {'name': '"""bytes_list_feature"""', 'dtype': 'ValueType.BYTES_LIST'}), "(name='bytes_list_feature', dtype=ValueType.BYTES_LIST)\n", (5114, 5169), False, 'from feast.feature import Feature\n'), ((5183, 5243), 'feast.feature.Feature', 'Feature', ([], {'name': '"""bool_list_feature"""', 'dtype': 'ValueType.BOOL_LIST'}), "(name='bool_list_feature', dtype=ValueType.BOOL_LIST)\n", (5190, 5243), False, 'from feast.feature import Feature\n'), ((591, 603), 'uuid.uuid4', 'uuid.uuid4', ([], {}), '()\n', (601, 603), False, 'import uuid\n')]
import numpy as np from bunch import Bunch from analysis import bunch_coord_array from analysis import BunchStats from analysis import DanilovEnvelopeBunch from orbit.teapot import DriftTEAPOT class AnalysisNode(DriftTEAPOT): def __init__(self, name, skip=0): DriftTEAPOT.__init__(self, name) self.setLength(0.0) self.position = None self.data = [] self.turn = 0 self.turns_stored = [] self.skip = skip self.active = True def set_position(self, position): self.position = position def get_data(self, turn='all'): if turn == 'all': return self.data return self.data[turn] def clear_data(self): self.data = [] def should_store(self): return self.turn % (self.skip + 1) == 0 def set_active(self, active): self.active = active class BunchMonitorNode(AnalysisNode): def __init__(self, name='bunch_monitor', mm_mrad=False, transverse_only=True, skip=0): AnalysisNode.__init__(self, name, skip) self.mm_mrad = mm_mrad self.transverse_only = transverse_only def track(self, params_dict): if self.should_store() and self.active: bunch = params_dict['bunch'] X = bunch_coord_array(bunch, self.mm_mrad, self.transverse_only) self.data.append(X) self.turns_stored.append(self.turn) self.turn += 1 class BunchStatsNode(AnalysisNode): def __init__(self, name='bunch_stats', mm_mrad=False, skip=0): AnalysisNode.__init__(self, name, skip) self.mm_mrad = mm_mrad def track(self, params_dict): if self.should_store() and self.active: bunch = params_dict['bunch'] X = bunch_coord_array(bunch, self.mm_mrad, transverse_only=True) Sigma = np.cov(X.T) bunch_stats = BunchStats(Sigma) self.data.append(bunch_stats) self.turns_stored.append(self.turn) self.turn += 1 class DanilovEnvelopeBunchMonitorNode(AnalysisNode): def __init__(self, name='envelope_monitor', mm_mrad=False, skip=0): AnalysisNode.__init__(self, name, skip) self.mm_mrad = mm_mrad def track(self, params_dict): if self.should_store() and self.active: bunch = params_dict['bunch'] X = bunch_coord_array(bunch, self.mm_mrad, transverse_only=True) env_bunch = DanilovEnvelopeBunch(X) self.data.append(env_bunch) self.turns_stored.append(self.turn) self.turn += 1
[ "orbit.teapot.DriftTEAPOT.__init__", "analysis.bunch_coord_array", "analysis.DanilovEnvelopeBunch", "analysis.BunchStats", "numpy.cov" ]
[((280, 312), 'orbit.teapot.DriftTEAPOT.__init__', 'DriftTEAPOT.__init__', (['self', 'name'], {}), '(self, name)\n', (300, 312), False, 'from orbit.teapot import DriftTEAPOT\n'), ((1358, 1418), 'analysis.bunch_coord_array', 'bunch_coord_array', (['bunch', 'self.mm_mrad', 'self.transverse_only'], {}), '(bunch, self.mm_mrad, self.transverse_only)\n', (1375, 1418), False, 'from analysis import bunch_coord_array\n'), ((1871, 1931), 'analysis.bunch_coord_array', 'bunch_coord_array', (['bunch', 'self.mm_mrad'], {'transverse_only': '(True)'}), '(bunch, self.mm_mrad, transverse_only=True)\n', (1888, 1931), False, 'from analysis import bunch_coord_array\n'), ((1952, 1963), 'numpy.cov', 'np.cov', (['X.T'], {}), '(X.T)\n', (1958, 1963), True, 'import numpy as np\n'), ((1990, 2007), 'analysis.BunchStats', 'BunchStats', (['Sigma'], {}), '(Sigma)\n', (2000, 2007), False, 'from analysis import BunchStats\n'), ((2492, 2552), 'analysis.bunch_coord_array', 'bunch_coord_array', (['bunch', 'self.mm_mrad'], {'transverse_only': '(True)'}), '(bunch, self.mm_mrad, transverse_only=True)\n', (2509, 2552), False, 'from analysis import bunch_coord_array\n'), ((2577, 2600), 'analysis.DanilovEnvelopeBunch', 'DanilovEnvelopeBunch', (['X'], {}), '(X)\n', (2597, 2600), False, 'from analysis import DanilovEnvelopeBunch\n')]
import numpy as np import matplotlib.pyplot as plt GENERATE_POINTS_DIST = './data/generatePoints_distance.txt' GENERATE_POINTS = './data/generatePoints.txt' r = np.random.RandomState(24) o = r.randn(400, 2) o[:, 0] += 2 o[:, 1] += 6 u = r.randn(400, 2) u[:, 0] += 4 u[:, 1] -= 0.5 v = r.randn(400, 2) v[:, 0] += 7 v[:, 1] -= 0.5 p = r.randn(400, 2) q = r.randn(400, 2) + 3 # q[:, 0] += 3 # q[:, 1] += 9 s = r.randn(400, 2) + 6 t = np.concatenate((o, p, q, s, u, v), axis=0) with open(GENERATE_POINTS, 'w', encoding='utf-8') as f: for pos in range(len(t)): cor = t[pos] f.write(str(pos) + ' ' + str(cor[0]) + ' ' + str(cor[1]) + '\n') d = lambda x, y: np.sqrt(np.power((x[0] - y[0]), 2) + np.power((x[1] - y[1]), 2)) with open(GENERATE_POINTS_DIST, 'w', encoding='utf-8') as f: for i in range(len(t)): for j in range(i + 1, len(t)): distance = d(t[i], t[j]) f.write(str(i) + ' ' + str(j) + ' ' + str(distance) + '\n') # Without labels x, y = t[:, 0], t[:, 1] plt.plot(x, y, 'ok', markersize=1, alpha=0.5) # plt.axis([-3, 10, -3, 9]) plt.xlabel('x') plt.ylabel('y') plt.title('Generated Points Plot') plt.savefig('./images/generatedPoints.png') plt.close() color = {0: 'c', 1: 'r', 2: 'g', 3: 'b', 4: 'm', 5: 'y'} cluster = [o, p, q, s, u, v] for i in range(len(cluster)): cur = cluster[i] x, y = cur[:, 0], cur[:, 1] plt.scatter(x, y, s=1, c=color[i], alpha=0.7, label=i + 1) plt.legend() plt.xlabel('x') plt.ylabel('y') plt.title('Generated Points with Lable') plt.savefig('./images/generatedColoredPoints.png') plt.show()
[ "matplotlib.pyplot.savefig", "matplotlib.pyplot.ylabel", "numpy.power", "matplotlib.pyplot.legend", "matplotlib.pyplot.xlabel", "matplotlib.pyplot.plot", "matplotlib.pyplot.close", "numpy.concatenate", "matplotlib.pyplot.scatter", "matplotlib.pyplot.title", "numpy.random.RandomState", "matplot...
[((163, 188), 'numpy.random.RandomState', 'np.random.RandomState', (['(24)'], {}), '(24)\n', (184, 188), True, 'import numpy as np\n'), ((434, 476), 'numpy.concatenate', 'np.concatenate', (['(o, p, q, s, u, v)'], {'axis': '(0)'}), '((o, p, q, s, u, v), axis=0)\n', (448, 476), True, 'import numpy as np\n'), ((1021, 1066), 'matplotlib.pyplot.plot', 'plt.plot', (['x', 'y', '"""ok"""'], {'markersize': '(1)', 'alpha': '(0.5)'}), "(x, y, 'ok', markersize=1, alpha=0.5)\n", (1029, 1066), True, 'import matplotlib.pyplot as plt\n'), ((1095, 1110), 'matplotlib.pyplot.xlabel', 'plt.xlabel', (['"""x"""'], {}), "('x')\n", (1105, 1110), True, 'import matplotlib.pyplot as plt\n'), ((1111, 1126), 'matplotlib.pyplot.ylabel', 'plt.ylabel', (['"""y"""'], {}), "('y')\n", (1121, 1126), True, 'import matplotlib.pyplot as plt\n'), ((1127, 1161), 'matplotlib.pyplot.title', 'plt.title', (['"""Generated Points Plot"""'], {}), "('Generated Points Plot')\n", (1136, 1161), True, 'import matplotlib.pyplot as plt\n'), ((1162, 1205), 'matplotlib.pyplot.savefig', 'plt.savefig', (['"""./images/generatedPoints.png"""'], {}), "('./images/generatedPoints.png')\n", (1173, 1205), True, 'import matplotlib.pyplot as plt\n'), ((1206, 1217), 'matplotlib.pyplot.close', 'plt.close', ([], {}), '()\n', (1215, 1217), True, 'import matplotlib.pyplot as plt\n'), ((1451, 1463), 'matplotlib.pyplot.legend', 'plt.legend', ([], {}), '()\n', (1461, 1463), True, 'import matplotlib.pyplot as plt\n'), ((1464, 1479), 'matplotlib.pyplot.xlabel', 'plt.xlabel', (['"""x"""'], {}), "('x')\n", (1474, 1479), True, 'import matplotlib.pyplot as plt\n'), ((1480, 1495), 'matplotlib.pyplot.ylabel', 'plt.ylabel', (['"""y"""'], {}), "('y')\n", (1490, 1495), True, 'import matplotlib.pyplot as plt\n'), ((1496, 1536), 'matplotlib.pyplot.title', 'plt.title', (['"""Generated Points with Lable"""'], {}), "('Generated Points with Lable')\n", (1505, 1536), True, 'import matplotlib.pyplot as plt\n'), ((1537, 1587), 'matplotlib.pyplot.savefig', 'plt.savefig', (['"""./images/generatedColoredPoints.png"""'], {}), "('./images/generatedColoredPoints.png')\n", (1548, 1587), True, 'import matplotlib.pyplot as plt\n'), ((1588, 1598), 'matplotlib.pyplot.show', 'plt.show', ([], {}), '()\n', (1596, 1598), True, 'import matplotlib.pyplot as plt\n'), ((1392, 1450), 'matplotlib.pyplot.scatter', 'plt.scatter', (['x', 'y'], {'s': '(1)', 'c': 'color[i]', 'alpha': '(0.7)', 'label': '(i + 1)'}), '(x, y, s=1, c=color[i], alpha=0.7, label=i + 1)\n', (1403, 1450), True, 'import matplotlib.pyplot as plt\n'), ((684, 708), 'numpy.power', 'np.power', (['(x[0] - y[0])', '(2)'], {}), '(x[0] - y[0], 2)\n', (692, 708), True, 'import numpy as np\n'), ((713, 737), 'numpy.power', 'np.power', (['(x[1] - y[1])', '(2)'], {}), '(x[1] - y[1], 2)\n', (721, 737), True, 'import numpy as np\n')]
# Copyright (c) Microsoft Corporation. All rights reserved. # Licensed under the MIT license. import torch from PIL import Image, ImageDraw import torch.utils.data as data import numpy as np import random import sys; sys.path.append('../') from utils.augmentations import preprocess class WIDERDetection(data.Dataset): """docstring for WIDERDetection""" def __init__(self, list_file, mode='train', mono_mode=False): super(WIDERDetection, self).__init__() self.mode = mode self.mono_mode = mono_mode self.fnames = [] self.boxes = [] self.labels = [] with open(list_file) as f: lines = f.readlines() for line in lines: line = line.strip().split() num_faces = int(line[1]) box = [] label = [] for i in range(num_faces): x = float(line[2 + 5 * i]) y = float(line[3 + 5 * i]) w = float(line[4 + 5 * i]) h = float(line[5 + 5 * i]) c = int(line[6 + 5 * i]) if w <= 0 or h <= 0: continue box.append([x, y, x + w, y + h]) label.append(c) if len(box) > 0: self.fnames.append(line[0]) self.boxes.append(box) self.labels.append(label) self.num_samples = len(self.boxes) def __len__(self): return self.num_samples def __getitem__(self, index): img, target, h, w = self.pull_item(index) return img, target def pull_item(self, index): while True: image_path = self.fnames[index] img = Image.open(image_path) if img.mode == 'L': img = img.convert('RGB') im_width, im_height = img.size boxes = self.annotransform( np.array(self.boxes[index]), im_width, im_height) label = np.array(self.labels[index]) bbox_labels = np.hstack((label[:, np.newaxis], boxes)).tolist() img, sample_labels = preprocess( img, bbox_labels, self.mode, image_path) sample_labels = np.array(sample_labels) if len(sample_labels) > 0: target = np.hstack( (sample_labels[:, 1:], sample_labels[:, 0][:, np.newaxis])) assert (target[:, 2] > target[:, 0]).any() assert (target[:, 3] > target[:, 1]).any() break else: index = random.randrange(0, self.num_samples) if self.mono_mode==True: im = 0.299 * img[0] + 0.587 * img[1] + 0.114 * img[2] return torch.from_numpy(np.expand_dims(im,axis=0)), target, im_height, im_width return torch.from_numpy(img), target, im_height, im_width def annotransform(self, boxes, im_width, im_height): boxes[:, 0] /= im_width boxes[:, 1] /= im_height boxes[:, 2] /= im_width boxes[:, 3] /= im_height return boxes def detection_collate(batch): """Custom collate fn for dealing with batches of images that have a different number of associated object annotations (bounding boxes). Arguments: batch: (tuple) A tuple of tensor images and lists of annotations Return: A tuple containing: 1) (tensor) batch of images stacked on their 0 dim 2) (list of tensors) annotations for a given image are stacked on 0 dim """ targets = [] imgs = [] for sample in batch: imgs.append(sample[0]) targets.append(torch.FloatTensor(sample[1])) return torch.stack(imgs, 0), targets
[ "PIL.Image.open", "numpy.hstack", "random.randrange", "torch.stack", "torch.from_numpy", "numpy.array", "utils.augmentations.preprocess", "numpy.expand_dims", "sys.path.append", "torch.FloatTensor" ]
[((218, 240), 'sys.path.append', 'sys.path.append', (['"""../"""'], {}), "('../')\n", (233, 240), False, 'import sys\n'), ((3733, 3753), 'torch.stack', 'torch.stack', (['imgs', '(0)'], {}), '(imgs, 0)\n', (3744, 3753), False, 'import torch\n'), ((1708, 1730), 'PIL.Image.open', 'Image.open', (['image_path'], {}), '(image_path)\n', (1718, 1730), False, 'from PIL import Image, ImageDraw\n'), ((1974, 2002), 'numpy.array', 'np.array', (['self.labels[index]'], {}), '(self.labels[index])\n', (1982, 2002), True, 'import numpy as np\n'), ((2112, 2163), 'utils.augmentations.preprocess', 'preprocess', (['img', 'bbox_labels', 'self.mode', 'image_path'], {}), '(img, bbox_labels, self.mode, image_path)\n', (2122, 2163), False, 'from utils.augmentations import preprocess\n'), ((2209, 2232), 'numpy.array', 'np.array', (['sample_labels'], {}), '(sample_labels)\n', (2217, 2232), True, 'import numpy as np\n'), ((2819, 2840), 'torch.from_numpy', 'torch.from_numpy', (['img'], {}), '(img)\n', (2835, 2840), False, 'import torch\n'), ((3692, 3720), 'torch.FloatTensor', 'torch.FloatTensor', (['sample[1]'], {}), '(sample[1])\n', (3709, 3720), False, 'import torch\n'), ((1904, 1931), 'numpy.array', 'np.array', (['self.boxes[index]'], {}), '(self.boxes[index])\n', (1912, 1931), True, 'import numpy as np\n'), ((2297, 2366), 'numpy.hstack', 'np.hstack', (['(sample_labels[:, 1:], sample_labels[:, 0][:, np.newaxis])'], {}), '((sample_labels[:, 1:], sample_labels[:, 0][:, np.newaxis]))\n', (2306, 2366), True, 'import numpy as np\n'), ((2572, 2609), 'random.randrange', 'random.randrange', (['(0)', 'self.num_samples'], {}), '(0, self.num_samples)\n', (2588, 2609), False, 'import random\n'), ((2029, 2069), 'numpy.hstack', 'np.hstack', (['(label[:, np.newaxis], boxes)'], {}), '((label[:, np.newaxis], boxes))\n', (2038, 2069), True, 'import numpy as np\n'), ((2747, 2773), 'numpy.expand_dims', 'np.expand_dims', (['im'], {'axis': '(0)'}), '(im, axis=0)\n', (2761, 2773), True, 'import numpy as np\n')]
from inference.gp_tools import GpOptimiser import matplotlib.pyplot as plt import matplotlib as mpl from numpy import sin, cos, linspace, array, meshgrid mpl.rcParams['axes.autolimit_mode'] = 'round_numbers' mpl.rcParams['axes.xmargin'] = 0 mpl.rcParams['axes.ymargin'] = 0 def example_plot_1d(): mu, sig = GP(x_gp) fig, (ax1, ax2, ax3) = plt.subplots(3, 1, gridspec_kw={'height_ratios': [1, 3, 1]}, figsize = (10,8)) ax1.plot(evaluations, max_values, marker = 'o', ls = 'solid', c = 'orange', label = 'highest observed value', zorder = 5) ax1.plot([2,12], [max(y_func), max(y_func)], ls = 'dashed', label = 'actual max', c = 'black') ax1.set_xlabel('function evaluations') ax1.set_xlim([2,12]) ax1.set_ylim([max(y)-0.3, max(y_func)+0.3]) ax1.xaxis.set_label_position('top') ax1.yaxis.set_label_position('right') ax1.xaxis.tick_top() ax1.set_yticks([]) ax1.legend(loc=4) ax2.plot(GP.x, GP.y, 'o', c = 'red', label = 'observations', zorder = 5) ax2.plot(x_gp, y_func, lw = 1.5, c = 'red', ls = 'dashed', label = 'actual function') ax2.plot(x_gp, mu, lw = 2, c = 'blue', label = 'GP prediction') ax2.fill_between(x_gp, (mu-2*sig), y2=(mu+2*sig), color = 'blue', alpha = 0.15, label = '95% confidence interval') ax2.set_ylim([-1.5,4]) ax2.set_ylabel('y') ax2.set_xticks([]) aq = array([abs(GP.acquisition(array([k]))) for k in x_gp]).squeeze() proposal = x_gp[aq.argmax()] ax3.fill_between(x_gp, 0.9*aq/aq.max(), color = 'green', alpha = 0.15) ax3.plot(x_gp, 0.9*aq/aq.max(), color = 'green', label = 'acquisition function') ax3.plot([proposal]*2, [0.,1.], c = 'green', ls = 'dashed', label = 'acquisition maximum') ax2.plot([proposal]*2, [-1.5,search_function(proposal)], c = 'green', ls = 'dashed') ax2.plot(proposal, search_function(proposal), 'o', c = 'green', label = 'proposed observation') ax3.set_ylim([0,1]) ax3.set_yticks([]) ax3.set_xlabel('x') ax3.legend(loc=1) ax2.legend(loc=2) plt.tight_layout() plt.subplots_adjust(hspace=0) plt.show() def example_plot_2d(): fig, (ax1, ax2) = plt.subplots(2, 1, gridspec_kw={'height_ratios': [1, 3]}, figsize=(10, 8)) plt.subplots_adjust(hspace=0) ax1.plot(evaluations, max_values, marker='o', ls='solid', c='orange', label='optimum value', zorder=5) ax1.plot([5, 30], [z_func.max(), z_func.max()], ls='dashed', label='actual max', c='black') ax1.set_xlabel('function evaluations') ax1.set_xlim([5, 30]) ax1.set_ylim([max(y) - 0.3, z_func.max() + 0.3]) ax1.xaxis.set_label_position('top') ax1.yaxis.set_label_position('right') ax1.xaxis.tick_top() ax1.set_yticks([]) ax1.legend(loc=4) ax2.contour(*mesh, z_func, 40) ax2.plot([i[0] for i in GP.x], [i[1] for i in GP.x], 'D', c='red', markeredgecolor='black') plt.show() """ GpOptimiser extends the functionality of GpRegressor to perform 'Bayesian optimisation'. Bayesian optimisation is suited to problems for which a single evaluation of the function being explored is expensive, such that the total number of function evaluations must be made as small as possible. """ # define the function whose maximum we will search for def search_function(x): return sin(0.5*x) + 3 / (1 + (x-1)**2) # define bounds for the optimisation bounds = [(-8,8)] # create some initialisation data x = array([-8,8]) y = search_function(x) # create an instance of GpOptimiser GP = GpOptimiser(x,y,bounds=bounds) # here we evaluate the search function for plotting purposes M = 500 x_gp = linspace(*bounds[0],M) y_func = search_function(x_gp) max_values = [max(GP.y)] evaluations = [len(GP.y)] for i in range(11): # plot the current state of the optimisation example_plot_1d() # request the proposed evaluation new_x = GP.propose_evaluation() # evaluate the new point new_y = search_function(new_x) # update the gaussian process with the new information GP.add_evaluation(new_x, new_y) # track the optimum value for plotting max_values.append(max(GP.y)) evaluations.append(len(GP.y)) """ 2D example """ from mpl_toolkits.mplot3d import Axes3D # define a new 2D search function def search_function(v): x,y = v z = ((x-1)/2)**2 + ((y+3)/1.5)**2 return sin(0.5*x) + cos(0.4*y) + 5/(1 + z) # set bounds bounds = [(-8,8), (-8,8)] # evaluate function for plotting N = 80 x = linspace(*bounds[0], N) y = linspace(*bounds[1], N) mesh = meshgrid(x, y) z_func = search_function(mesh) # create some initialisation data # we've picked a point at each corner and one in the middle x = [(-8,-8), (8,-8), (-8,8), (8,8), (0,0)] y = [search_function(k) for k in x] # initiate the optimiser GP = GpOptimiser(x,y,bounds=bounds) max_values = [max(GP.y)] evaluations = [len(GP.y)] for i in range(25): new_x = GP.propose_evaluation() new_y = search_function(new_x) GP.add_evaluation(new_x, new_y) # track the optimum value for plotting max_values.append(max(GP.y)) evaluations.append(len(GP.y)) # plot the results example_plot_2d()
[ "inference.gp_tools.GpOptimiser", "numpy.array", "numpy.linspace", "numpy.cos", "matplotlib.pyplot.tight_layout", "numpy.sin", "numpy.meshgrid", "matplotlib.pyplot.subplots", "matplotlib.pyplot.subplots_adjust", "matplotlib.pyplot.show" ]
[((3398, 3412), 'numpy.array', 'array', (['[-8, 8]'], {}), '([-8, 8])\n', (3403, 3412), False, 'from numpy import sin, cos, linspace, array, meshgrid\n'), ((3477, 3509), 'inference.gp_tools.GpOptimiser', 'GpOptimiser', (['x', 'y'], {'bounds': 'bounds'}), '(x, y, bounds=bounds)\n', (3488, 3509), False, 'from inference.gp_tools import GpOptimiser\n'), ((3586, 3609), 'numpy.linspace', 'linspace', (['*bounds[0]', 'M'], {}), '(*bounds[0], M)\n', (3594, 3609), False, 'from numpy import sin, cos, linspace, array, meshgrid\n'), ((4441, 4464), 'numpy.linspace', 'linspace', (['*bounds[0]', 'N'], {}), '(*bounds[0], N)\n', (4449, 4464), False, 'from numpy import sin, cos, linspace, array, meshgrid\n'), ((4469, 4492), 'numpy.linspace', 'linspace', (['*bounds[1]', 'N'], {}), '(*bounds[1], N)\n', (4477, 4492), False, 'from numpy import sin, cos, linspace, array, meshgrid\n'), ((4500, 4514), 'numpy.meshgrid', 'meshgrid', (['x', 'y'], {}), '(x, y)\n', (4508, 4514), False, 'from numpy import sin, cos, linspace, array, meshgrid\n'), ((4754, 4786), 'inference.gp_tools.GpOptimiser', 'GpOptimiser', (['x', 'y'], {'bounds': 'bounds'}), '(x, y, bounds=bounds)\n', (4765, 4786), False, 'from inference.gp_tools import GpOptimiser\n'), ((350, 427), 'matplotlib.pyplot.subplots', 'plt.subplots', (['(3)', '(1)'], {'gridspec_kw': "{'height_ratios': [1, 3, 1]}", 'figsize': '(10, 8)'}), "(3, 1, gridspec_kw={'height_ratios': [1, 3, 1]}, figsize=(10, 8))\n", (362, 427), True, 'import matplotlib.pyplot as plt\n'), ((2024, 2042), 'matplotlib.pyplot.tight_layout', 'plt.tight_layout', ([], {}), '()\n', (2040, 2042), True, 'import matplotlib.pyplot as plt\n'), ((2047, 2076), 'matplotlib.pyplot.subplots_adjust', 'plt.subplots_adjust', ([], {'hspace': '(0)'}), '(hspace=0)\n', (2066, 2076), True, 'import matplotlib.pyplot as plt\n'), ((2081, 2091), 'matplotlib.pyplot.show', 'plt.show', ([], {}), '()\n', (2089, 2091), True, 'import matplotlib.pyplot as plt\n'), ((2138, 2212), 'matplotlib.pyplot.subplots', 'plt.subplots', (['(2)', '(1)'], {'gridspec_kw': "{'height_ratios': [1, 3]}", 'figsize': '(10, 8)'}), "(2, 1, gridspec_kw={'height_ratios': [1, 3]}, figsize=(10, 8))\n", (2150, 2212), True, 'import matplotlib.pyplot as plt\n'), ((2217, 2246), 'matplotlib.pyplot.subplots_adjust', 'plt.subplots_adjust', ([], {'hspace': '(0)'}), '(hspace=0)\n', (2236, 2246), True, 'import matplotlib.pyplot as plt\n'), ((2861, 2871), 'matplotlib.pyplot.show', 'plt.show', ([], {}), '()\n', (2869, 2871), True, 'import matplotlib.pyplot as plt\n'), ((3271, 3283), 'numpy.sin', 'sin', (['(0.5 * x)'], {}), '(0.5 * x)\n', (3274, 3283), False, 'from numpy import sin, cos, linspace, array, meshgrid\n'), ((4320, 4332), 'numpy.sin', 'sin', (['(0.5 * x)'], {}), '(0.5 * x)\n', (4323, 4332), False, 'from numpy import sin, cos, linspace, array, meshgrid\n'), ((4333, 4345), 'numpy.cos', 'cos', (['(0.4 * y)'], {}), '(0.4 * y)\n', (4336, 4345), False, 'from numpy import sin, cos, linspace, array, meshgrid\n'), ((1388, 1398), 'numpy.array', 'array', (['[k]'], {}), '([k])\n', (1393, 1398), False, 'from numpy import sin, cos, linspace, array, meshgrid\n')]
import numpy as np def alignChannels(red, green, blue): """Given 3 images corresponding to different channels of a color image, compute the best aligned result with minimum abberations Args: red, green, blue - each is a HxW matrix corresponding to an HxW image Returns: rgb_output - HxWx3 color image output, aligned as desired""" bestGreenAlignment = None minSSD = 1e9 for horizontalShift in range(-30, 31 , 1): #shiftedGreen = np.roll(green, horizontalShift, axis=0) for virticalShift in range(-30, 31 , 1): shiftedGreen = np.roll(np.roll(green, horizontalShift, axis=0), virticalShift, axis=1) shiftedSSD = np.sqrt(np.sum(np.square(shiftedGreen-red))) if shiftedSSD < minSSD: minSSD = shiftedSSD bestGreenAlignment = shiftedGreen bestBlueAlignment = None minSSD = 1e9 for horizontalShift in range(-30, 31 , 1): for virticalShift in range(-30, 31 , 1): shiftedBlue = np.roll(np.roll(blue,horizontalShift, axis=0), virticalShift, axis=1) shiftedSSD = np.sqrt(np.sum(np.square(shiftedBlue-red))) if shiftedSSD < minSSD: minSSD = shiftedSSD bestBlueAlignment = shiftedBlue alignedImage = np.zeros([blue.shape[0], blue.shape[1], 3]) alignedImage[:,:,0] = red alignedImage[:,:,1] = bestGreenAlignment alignedImage[:,:,2] = bestBlueAlignment return alignedImage
[ "numpy.zeros", "numpy.roll", "numpy.square" ]
[((1191, 1234), 'numpy.zeros', 'np.zeros', (['[blue.shape[0], blue.shape[1], 3]'], {}), '([blue.shape[0], blue.shape[1], 3])\n', (1199, 1234), True, 'import numpy as np\n'), ((558, 597), 'numpy.roll', 'np.roll', (['green', 'horizontalShift'], {'axis': '(0)'}), '(green, horizontalShift, axis=0)\n', (565, 597), True, 'import numpy as np\n'), ((949, 987), 'numpy.roll', 'np.roll', (['blue', 'horizontalShift'], {'axis': '(0)'}), '(blue, horizontalShift, axis=0)\n', (956, 987), True, 'import numpy as np\n'), ((664, 693), 'numpy.square', 'np.square', (['(shiftedGreen - red)'], {}), '(shiftedGreen - red)\n', (673, 693), True, 'import numpy as np\n'), ((1053, 1081), 'numpy.square', 'np.square', (['(shiftedBlue - red)'], {}), '(shiftedBlue - red)\n', (1062, 1081), True, 'import numpy as np\n')]
""" This code using models on cluster, which needs large computing power. Now we deployed an Object Detection model on AWS Lambda, and accompied with Amazon API gateway, see obj_detector.py """ from models.yolo.models import * from models.yolo.utils import * import cv2 import os, sys, time, datetime, random import torch import numpy as np from torch.utils.data import DataLoader from torchvision import datasets, transforms from torch.autograd import Variable import matplotlib.pyplot as plt import matplotlib.patches as patches from PIL import Image from models.yolo.sort import * class YoloDetectron(): def __init__(self): # Basical setting related to the YOLO mdoel self.config_path = '/home/ubuntu/workspace/CarsMemory/models/yolo/config/yolov3.cfg' self.weights_path = '/home/ubuntu/workspace/CarsMemory/models/yolo/config/yolov3.weights' self.class_path = '/home/ubuntu/workspace/CarsMemory/models/yolo/config/coco.names' self.img_size = 416 self.conf_thres = 0.8 self.nms_thres = 0.4 # Load model and weights self.model = Darknet(self.config_path, img_size=self.img_size) self.model.load_weights(self.weights_path) self.model.eval() self.classes = utils.load_classes(self.class_path) self.Tensor = torch.FloatTensor # About image color self.cmap = plt.get_cmap('tab20b') self.colors = [self.cmap(i)[:3] for i in np.linspace(0, 1, 20)] self.mot_tracker = Sort() def _detect_image(self, img): # scale and pad image ratio = min(self.img_size/img.size[0], self.img_size/img.size[1]) imw = round(img.size[0] * ratio) imh = round(img.size[1] * ratio) img_transforms = transforms.Compose([transforms.Resize((imh, imw)), transforms.Pad((max(int((imh-imw)/2),0), max(int((imw-imh)/2),0), max(int((imh-imw)/2),0), max(int((imw-imh)/2),0)), (128, 128, 128)), transforms.ToTensor(), ]) # convert image to Tensor image_tensor = img_transforms(img).float() image_tensor = image_tensor.unsqueeze_(0) input_img = Variable(image_tensor.type(self.Tensor)) # run inference on the model and get detections with torch.no_grad(): detections = self.model(input_img) detections = utils.non_max_suppression(detections, 80) return detections[0] def processing(self, frame): frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) pilimg = Image.fromarray(frame) detections = self._detect_image(pilimg) img = np.array(pilimg) pad_x = max(img.shape[0] - img.shape[1], 0) * (self.img_size / max(img.shape)) pad_y = max(img.shape[1] - img.shape[0], 0) * (self.img_size / max(img.shape)) unpad_h = self.img_size - pad_y unpad_w = self.img_size - pad_x if detections is not None: tracked_objects = self.mot_tracker.update(detections.cpu()) # unique_labels = detections[:, -1].cpu().unique() # n_cls_preds = len(unique_labels) for x1, y1, x2, y2, obj_id, cls_pred in tracked_objects: box_h = int(((y2 - y1) / unpad_h) * img.shape[0]) box_w = int(((x2 - x1) / unpad_w) * img.shape[1]) y1 = int(((y1 - pad_y // 2) / unpad_h) * img.shape[0]) x1 = int(((x1 - pad_x // 2) / unpad_w) * img.shape[1]) color = self.colors[int(obj_id) % len(self.colors)] color = [i * 255 for i in color] cls = self.classes[int(cls_pred)] cv2.rectangle(frame, (x1, y1), (x1+box_w, y1+box_h), color, 4) cv2.rectangle(frame, (x1, y1-35), (x1+len(cls)*19+60, y1), color, -1) cv2.putText(frame, cls + "-" + str(int(obj_id)), (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 1, (255,255,255), 3) return frame
[ "cv2.rectangle", "PIL.Image.fromarray", "numpy.array", "numpy.linspace", "cv2.cvtColor", "torchvision.transforms.Resize", "torch.no_grad", "torchvision.transforms.ToTensor", "matplotlib.pyplot.get_cmap" ]
[((1390, 1412), 'matplotlib.pyplot.get_cmap', 'plt.get_cmap', (['"""tab20b"""'], {}), "('tab20b')\n", (1402, 1412), True, 'import matplotlib.pyplot as plt\n'), ((2519, 2557), 'cv2.cvtColor', 'cv2.cvtColor', (['frame', 'cv2.COLOR_BGR2RGB'], {}), '(frame, cv2.COLOR_BGR2RGB)\n', (2531, 2557), False, 'import cv2\n'), ((2575, 2597), 'PIL.Image.fromarray', 'Image.fromarray', (['frame'], {}), '(frame)\n', (2590, 2597), False, 'from PIL import Image\n'), ((2661, 2677), 'numpy.array', 'np.array', (['pilimg'], {}), '(pilimg)\n', (2669, 2677), True, 'import numpy as np\n'), ((2309, 2324), 'torch.no_grad', 'torch.no_grad', ([], {}), '()\n', (2322, 2324), False, 'import torch\n'), ((1462, 1483), 'numpy.linspace', 'np.linspace', (['(0)', '(1)', '(20)'], {}), '(0, 1, 20)\n', (1473, 1483), True, 'import numpy as np\n'), ((1785, 1814), 'torchvision.transforms.Resize', 'transforms.Resize', (['(imh, imw)'], {}), '((imh, imw))\n', (1802, 1814), False, 'from torchvision import datasets, transforms\n'), ((2005, 2026), 'torchvision.transforms.ToTensor', 'transforms.ToTensor', ([], {}), '()\n', (2024, 2026), False, 'from torchvision import datasets, transforms\n'), ((3677, 3743), 'cv2.rectangle', 'cv2.rectangle', (['frame', '(x1, y1)', '(x1 + box_w, y1 + box_h)', 'color', '(4)'], {}), '(frame, (x1, y1), (x1 + box_w, y1 + box_h), color, 4)\n', (3690, 3743), False, 'import cv2\n')]
#!/usr/bin/env python """ Video Animation Functions ========================= This module contains functions for displaying sequences of 2D data as animations. - animate Animate a 2D video sequence. - animate2 Animate two 2D video sequences simultaneously. - animate_compare Animate two 2D video sequences and their difference. - frame_compare Display frames from two sequences and their difference. """ # Copyright (c) 2009-2015, <NAME> # All rights reserved. # Distributed under the terms of the BSD license: # http://www.opensource.org/licenses/bsd-license __all__ = ['animate', 'animate2', 'animate_compare', 'frame_compare'] import time import numpy as np import pylab as p # Fix for animation problems with Qt4Agg backend: if p.get_backend() == 'Qt4Agg': from PyQt4.QtGui import QApplication animate_fix = QApplication.processEvents else: def animate_fix(): pass def animate(data, step=1, delay=0): """ Animate sequence of frames. Animate a sequence of `Ny x Nx` bitmap frames stored in a `M x Ny x Nx` data array. Parameters ---------- data : numpy.ndarray Sequence of `M` 2D bitmaps stored as an array with shape `(M, Ny, Nx)`. step : int Skip `step` frames between each displayed frames. delay : float Wait `delay` seconds between each frame refresh. """ # Get maximum value in data to scale the luminance appropriately: mx = np.max(np.abs(data)) img = p.imshow(data[0, :, :], vmin=-mx, vmax=mx) for k in np.arange(0, data.shape[0], step): time.sleep(delay) img.set_data(data[k, :, :]) p.draw() animate_fix() def animate2(data_1, data_2, step=1, delay=0): """ Animate two sequence of frames simultaneously. Animate two sequences of `Ny x Nx` bitmap frames stored in two `M x Ny x Nx` data arrays. Parameters ---------- data_1 : numpy.ndarray Sequence of `M` 2D bitmaps stored as an array with shape `(M, Ny, Nx)`. data_2 : numpy.ndarray Sequence of `M` 2D bitmaps stored as an array with shape `(M, Ny, Nx)`. step : int Skip `step` frames between each displayed frames. delay : float Wait `delay` seconds between each frame refresh. """ if data_1.shape != data_2.shape: raise ValueError('cannot animate two video sequences with ' 'different dimensions') # Get maximum value in data to scale the luminance appropriately: mx_1 = np.max(np.abs(data_1)) mx_2 = np.max(np.abs(data_2)) p.subplot(121) img_1 = p.imshow(data_1[0, :, :], vmin=-mx_1, vmax=mx_1) p.subplot(122) img_2 = p.imshow(data_2[0, :, :], vmin=-mx_2, vmax=mx_2) for k in np.arange(0, data_1.shape[0], step): time.sleep(delay) img_1.set_data(data_1[k, :, :]) img_2.set_data(data_2[k, :, :]) p.draw() animate_fix() def animate_compare(data_1, data_2, step=1, delay=0): """ Animate two sequence of frames and their difference simultaneously. Animate two sequences of `Ny x Nx` bitmap frames stored in two `M x Ny x Nx` data arrays simultaneously with their difference. Parameters ---------- data_1 : numpy.ndarray Sequence of `M` 2D bitmaps stored as an array with shape `(M, Ny, Nx)`. data_2 : numpy.ndarray Sequence of `M` 2D bitmaps stored as an array with shape `(M, Ny, Nx)`. step : int Skip `step` frames between each displayed frames. delay : float Wait `delay` seconds between each frame refresh. """ if data_1.shape != data_2.shape: raise ValueError('cannot animate two video sequences with ' 'different dimensions') # Get maximum value in data to scale the luminance appropriately: mx_1 = np.max(np.abs(data_1)) mx_2 = np.max(np.abs(data_2)) mx_err = max(mx_1, mx_2) p.subplot(131) img_1 = p.imshow(data_1[0, :, :], vmin=-mx_1, vmax=mx_1) p.subplot(132) img_2 = p.imshow(data_2[0, :, :], vmin=-mx_2, vmax=mx_2) p.subplot(133) img_err = p.imshow(data_1[0, :, :]-data_2[0, :, :], vmin=-mx_err, vmax=mx_err) for k in np.arange(0, data_1.shape[0], step): time.sleep(delay) img_1.set_data(data_1[k, :, :]) img_2.set_data(data_2[k, :, :]) img_err.set_data(data_1[k, :, :]-data_2[k, :, :]) p.draw() animate_fix() def frame_compare(data_1, data_2, i=0): """ Compare corresponding frames in two video sequences. Simultaneously display two corresponding frames from two video sequences of identical length. Parameters ---------- data_1 : numpy.ndarray Sequence of `M` 2D bitmaps stored as an array with shape `(M, Ny, Nx)`. data_2 : numpy.ndarray Sequence of `M` 2D bitmaps stored as an array with shape `(M, Ny, Nx)`. i : int Index of frame to display. """ if data_1.shape != data_2.shape: raise ValueError('cannot compare frames from two video sequences with ' 'different dimensions') mx_1 = np.max(np.abs(data_1)) mx_2 = np.max(np.abs(data_2)) p.subplot(121) p.imshow(data_1[i, :, :], vmin=-mx_1, vmax=mx_1) p.subplot(122) p.imshow(data_2[i, :, :], vmin=-mx_2, vmax=mx_2) p.draw()
[ "numpy.abs", "pylab.draw", "pylab.subplot", "time.sleep", "pylab.get_backend", "numpy.arange", "pylab.imshow" ]
[((763, 778), 'pylab.get_backend', 'p.get_backend', ([], {}), '()\n', (776, 778), True, 'import pylab as p\n'), ((1501, 1543), 'pylab.imshow', 'p.imshow', (['data[0, :, :]'], {'vmin': '(-mx)', 'vmax': 'mx'}), '(data[0, :, :], vmin=-mx, vmax=mx)\n', (1509, 1543), True, 'import pylab as p\n'), ((1557, 1590), 'numpy.arange', 'np.arange', (['(0)', 'data.shape[0]', 'step'], {}), '(0, data.shape[0], step)\n', (1566, 1590), True, 'import numpy as np\n'), ((2614, 2628), 'pylab.subplot', 'p.subplot', (['(121)'], {}), '(121)\n', (2623, 2628), True, 'import pylab as p\n'), ((2641, 2689), 'pylab.imshow', 'p.imshow', (['data_1[0, :, :]'], {'vmin': '(-mx_1)', 'vmax': 'mx_1'}), '(data_1[0, :, :], vmin=-mx_1, vmax=mx_1)\n', (2649, 2689), True, 'import pylab as p\n'), ((2694, 2708), 'pylab.subplot', 'p.subplot', (['(122)'], {}), '(122)\n', (2703, 2708), True, 'import pylab as p\n'), ((2721, 2769), 'pylab.imshow', 'p.imshow', (['data_2[0, :, :]'], {'vmin': '(-mx_2)', 'vmax': 'mx_2'}), '(data_2[0, :, :], vmin=-mx_2, vmax=mx_2)\n', (2729, 2769), True, 'import pylab as p\n'), ((2783, 2818), 'numpy.arange', 'np.arange', (['(0)', 'data_1.shape[0]', 'step'], {}), '(0, data_1.shape[0], step)\n', (2792, 2818), True, 'import numpy as np\n'), ((3980, 3994), 'pylab.subplot', 'p.subplot', (['(131)'], {}), '(131)\n', (3989, 3994), True, 'import pylab as p\n'), ((4007, 4055), 'pylab.imshow', 'p.imshow', (['data_1[0, :, :]'], {'vmin': '(-mx_1)', 'vmax': 'mx_1'}), '(data_1[0, :, :], vmin=-mx_1, vmax=mx_1)\n', (4015, 4055), True, 'import pylab as p\n'), ((4060, 4074), 'pylab.subplot', 'p.subplot', (['(132)'], {}), '(132)\n', (4069, 4074), True, 'import pylab as p\n'), ((4087, 4135), 'pylab.imshow', 'p.imshow', (['data_2[0, :, :]'], {'vmin': '(-mx_2)', 'vmax': 'mx_2'}), '(data_2[0, :, :], vmin=-mx_2, vmax=mx_2)\n', (4095, 4135), True, 'import pylab as p\n'), ((4140, 4154), 'pylab.subplot', 'p.subplot', (['(133)'], {}), '(133)\n', (4149, 4154), True, 'import pylab as p\n'), ((4169, 4239), 'pylab.imshow', 'p.imshow', (['(data_1[0, :, :] - data_2[0, :, :])'], {'vmin': '(-mx_err)', 'vmax': 'mx_err'}), '(data_1[0, :, :] - data_2[0, :, :], vmin=-mx_err, vmax=mx_err)\n', (4177, 4239), True, 'import pylab as p\n'), ((4251, 4286), 'numpy.arange', 'np.arange', (['(0)', 'data_1.shape[0]', 'step'], {}), '(0, data_1.shape[0], step)\n', (4260, 4286), True, 'import numpy as np\n'), ((5256, 5270), 'pylab.subplot', 'p.subplot', (['(121)'], {}), '(121)\n', (5265, 5270), True, 'import pylab as p\n'), ((5275, 5323), 'pylab.imshow', 'p.imshow', (['data_1[i, :, :]'], {'vmin': '(-mx_1)', 'vmax': 'mx_1'}), '(data_1[i, :, :], vmin=-mx_1, vmax=mx_1)\n', (5283, 5323), True, 'import pylab as p\n'), ((5328, 5342), 'pylab.subplot', 'p.subplot', (['(122)'], {}), '(122)\n', (5337, 5342), True, 'import pylab as p\n'), ((5347, 5395), 'pylab.imshow', 'p.imshow', (['data_2[i, :, :]'], {'vmin': '(-mx_2)', 'vmax': 'mx_2'}), '(data_2[i, :, :], vmin=-mx_2, vmax=mx_2)\n', (5355, 5395), True, 'import pylab as p\n'), ((5400, 5408), 'pylab.draw', 'p.draw', ([], {}), '()\n', (5406, 5408), True, 'import pylab as p\n'), ((1477, 1489), 'numpy.abs', 'np.abs', (['data'], {}), '(data)\n', (1483, 1489), True, 'import numpy as np\n'), ((1600, 1617), 'time.sleep', 'time.sleep', (['delay'], {}), '(delay)\n', (1610, 1617), False, 'import time\n'), ((1662, 1670), 'pylab.draw', 'p.draw', ([], {}), '()\n', (1668, 1670), True, 'import pylab as p\n'), ((2560, 2574), 'numpy.abs', 'np.abs', (['data_1'], {}), '(data_1)\n', (2566, 2574), True, 'import numpy as np\n'), ((2594, 2608), 'numpy.abs', 'np.abs', (['data_2'], {}), '(data_2)\n', (2600, 2608), True, 'import numpy as np\n'), ((2828, 2845), 'time.sleep', 'time.sleep', (['delay'], {}), '(delay)\n', (2838, 2845), False, 'import time\n'), ((2934, 2942), 'pylab.draw', 'p.draw', ([], {}), '()\n', (2940, 2942), True, 'import pylab as p\n'), ((3897, 3911), 'numpy.abs', 'np.abs', (['data_1'], {}), '(data_1)\n', (3903, 3911), True, 'import numpy as np\n'), ((3931, 3945), 'numpy.abs', 'np.abs', (['data_2'], {}), '(data_2)\n', (3937, 3945), True, 'import numpy as np\n'), ((4296, 4313), 'time.sleep', 'time.sleep', (['delay'], {}), '(delay)\n', (4306, 4313), False, 'import time\n'), ((4460, 4468), 'pylab.draw', 'p.draw', ([], {}), '()\n', (4466, 4468), True, 'import pylab as p\n'), ((5202, 5216), 'numpy.abs', 'np.abs', (['data_1'], {}), '(data_1)\n', (5208, 5216), True, 'import numpy as np\n'), ((5236, 5250), 'numpy.abs', 'np.abs', (['data_2'], {}), '(data_2)\n', (5242, 5250), True, 'import numpy as np\n')]
import numpy as np import os import sys utils_path = os.path.join(os.path.abspath(os.getenv('PROCESSING_DIR')),'utils') if utils_path not in sys.path: sys.path.append(utils_path) import util_files import util_cloud import urllib import ee from google.cloud import storage import logging import gzip import shutil import rasterio # set up logging # get the top-level logger object logger = logging.getLogger() for handler in logger.handlers: logger.removeHandler(handler) logger.setLevel(logging.INFO) # make it print to the console. console = logging.StreamHandler() logger.addHandler(console) logging.basicConfig(format='%(asctime)s - %(name)s - %(levelname)s - %(message)s') # name of asset on GEE where you want to upload data # this should be an asset name that is not currently in use dataset_name = 'req_019_facebook_total_precipitation' #check logger.info('Executing script for dataset: ' + dataset_name) # first, set the directory that you are working in with the path variable path = os.path.abspath(os.path.join(os.getenv('BLOG_DIR'),dataset_name)) # move to this directory os.chdir(path) # create a new sub-directory within your specified dir called 'data' # within this directory, create files to store raw and processed data data_dir = 'data' if not os.path.exists(data_dir): os.mkdir(data_dir) ''' Download data and save to your data directory ''' logger.info('Downloading raw data') # list the urls used to download the data from the source website url_list = ['https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_1891_1900_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_1901_1910_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_1911_1920_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_1921_1930_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_1931_1940_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_1941_1950_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_1951_1960_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_1961_1970_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_1971_1980_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_1981_1990_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_1991_2000_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_2001_2010_025.nc.gz', 'https://opendata.dwd.de/climate_environment/GPCC/full_data_monthly_v2020/025/full_data_monthly_v2020_2011_2019_025.nc.gz'] # download the data from the source raw_data_file = [os.path.join(data_dir,os.path.basename(url)) for url in url_list] for url, file in zip(url_list, raw_data_file): urllib.request.urlretrieve(url, file) # unzip source data raw_data_file_unzipped = [file[:-3] for file in raw_data_file] for file,file_unzipped in zip(raw_data_file,raw_data_file_unzipped): with gzip.open(file, 'rb') as f_in: with open(file_unzipped, 'wb') as f_out: shutil.copyfileobj(f_in, f_out) # convert the netcdf files to tif files for raw_file in raw_data_file_unzipped: util_files.convert_netcdf(raw_file, ['precip']) processed_data_file = [os.path.join(raw_file[:-3]+'_precip.tif') for raw_file in raw_data_file_unzipped] processed_data_annual=[os.path.join(data_dir,'full_data_annual_v2020_'+str(year)+'_025_precip.tif') for year in range(1891,2020)] n_layers=[int(rasterio.open(file).meta['count']/12) for file in processed_data_file] # calculate annual total precipitation for id, file in enumerate(processed_data_file,start=0): with rasterio.open(file) as src0: # update metedata for annual aggregation meta = src0.meta meta.update(count = 1) meta.update(nodata = meta['nodata']*12) # sum and export annual total precipitation as tif file for i in range(int(src0.meta['count']/12)): with rasterio.open(processed_data_annual[sum(n_layers[:id])+i], 'w', **meta) as dst: dst.write_band(1,np.sum(src0.read(range((i*12+1), (i*12+13))), axis = 0)) ''' Upload processed data to Google Earth Engine ''' logger.info('Uploading processed data to Google Cloud Storage.') # set up Google Cloud Storage project and bucket objects gcsClient = storage.Client(os.environ.get("CLOUDSDK_CORE_PROJECT")) gcsBucket = gcsClient.bucket(os.environ.get("GEE_STAGING_BUCKET")) # upload files to Google Cloud Storage gcs_uris= util_cloud.gcs_upload(processed_data_annual, dataset_name, gcs_bucket=gcsBucket) logger.info('Uploading processed data to Google Earth Engine.') # initialize ee and eeUtil modules for uploading to Google Earth Engine auth = ee.ServiceAccountCredentials(os.getenv('GEE_SERVICE_ACCOUNT'), os.getenv('GOOGLE_APPLICATION_CREDENTIALS')) ee.Initialize(auth) # set pyramiding policy for GEE upload pyramiding_policy = 'MEAN' #check # create an image collection where we will put the processed data files in GEE image_collection = f'projects/resource-watch-gee/Facebook/PrecipitationAnalysis/GPCC_annual' ee.data.createAsset({'type': 'ImageCollection'}, image_collection) # set image collection's privacy to public acl = {"all_users_can_read": True} ee.data.setAssetAcl(image_collection, acl) print('Privacy set to public.') # list the bands in each image band_ids = ['b1'] task_id = [] # upload processed data files to GEE for uri in gcs_uris: # generate an asset name for the current file by using the filename (minus the file type extension) asset_name = f'projects/resource-watch-gee/Facebook/PrecipitationAnalysis/GPCC_annual/{os.path.basename(uri)[:-4]}' # create the band manifest for this asset mf_bands = [{'id': band_id, 'tileset_band_index': band_ids.index(band_id), 'tileset_id': os.path.basename(uri)[:-4], 'pyramidingPolicy': pyramiding_policy} for band_id in band_ids] # create complete manifest for asset upload manifest = util_cloud.gee_manifest_complete(asset_name, uri, mf_bands) # upload the file from GCS to GEE task = util_cloud.gee_ingest(manifest) print(asset_name + ' uploaded to GEE') task_id.append(task) # remove files from Google Cloud Storage util_cloud.gcs_remove(gcs_uris, gcs_bucket=gcsBucket) logger.info('Files deleted from Google Cloud Storage.') ''' Data processing on Google Earth Engine ''' # Initialize earth engine try: ee.Initialize() except Exception as e: ee.Authenticate() ee.Initialize() # load image collection GPCC_annual = ee.ImageCollection("projects/resource-watch-gee/Facebook/PrecipitationAnalysis/GPCC_annual") # save projection (crs, crsTransform, scale) projection = GPCC_annual.first().projection().getInfo() projection_gee = GPCC_annual.first().projection() crs = projection.get('crs') crsTransform = projection.get('transform') scale = GPCC_annual.first().projection().nominalScale().getInfo() print('crs: ', crs) print('crsTransform: ', crsTransform) print('scale: ', scale) # convert ImageCollection to Image band_names = ee.List([str(i) for i in np.arange(1891,2020)]) GPCC_annual_img = GPCC_annual.toBands() GPCC_annual_img = GPCC_annual_img.rename(band_names) # define countries to calculate statistics for country_list = ["USA","GBR","FRA","DEU","CAN", "SWE", "BRA", "MEX", "BEL", "IRL", "NLD", "NGA", "SAU", "ZAF", "ESP", "IND", "IDN", "TWN"] # load country and state shapefiles countries = ee.FeatureCollection("projects/resource-watch-gee/gadm36_0_simplified") states = ee.FeatureCollection("projects/resource-watch-gee/gadm36_1_simplified") for country_ios in country_list: # national-level analysis # load country data, filter to desired ISO Codes country = countries.filterMetadata('GID_0', 'equals', ee.String(country_ios)) # export to Google Drive output_name='GPCC_annual_country_'+country_ios output_folder='Facebook' export_results_task = ee.batch.Export.table.toDrive( collection = GPCC_annual_img.reduceRegions(country, ee.Reducer.mean(), scale = scale, tileScale = 16).select(ee.List(['GID_0','NAME_0']).cat(band_names), retainGeometry = False), description = output_name, fileNamePrefix = output_name, folder = output_folder) # start the task export_results_task.start() # state-level analysis # load state data, filter to desired ISO Codes state = states.filterMetadata('GID_0', 'equals', ee.String(country_ios)) # export to Google Drive output_name='GPCC_annual_state_'+country_ios output_folder='Facebook' export_results_task = ee.batch.Export.table.toDrive( collection = GPCC_annual_img.reduceRegions(state, ee.Reducer.mean(), scale = scale, tileScale = 16).select(ee.List(['CC_1','ENGTYPE_1','GID_0','GID_1','HASC_1','NAME_0','NAME_1','NL_NAME_1','TYPE_1','VARNAME_1']).cat(band_names), retainGeometry = False), description = output_name, fileNamePrefix = output_name, folder = output_folder) # start the task export_results_task.start()
[ "logging.getLogger", "logging.StreamHandler", "gzip.open", "util_cloud.gcs_remove", "ee.ImageCollection", "ee.String", "sys.path.append", "numpy.arange", "os.path.exists", "ee.Authenticate", "ee.Reducer.mean", "urllib.request.urlretrieve", "util_cloud.gee_ingest", "os.mkdir", "ee.data.se...
[((395, 414), 'logging.getLogger', 'logging.getLogger', ([], {}), '()\n', (412, 414), False, 'import logging\n'), ((549, 572), 'logging.StreamHandler', 'logging.StreamHandler', ([], {}), '()\n', (570, 572), False, 'import logging\n'), ((600, 687), 'logging.basicConfig', 'logging.basicConfig', ([], {'format': '"""%(asctime)s - %(name)s - %(levelname)s - %(message)s"""'}), "(format=\n '%(asctime)s - %(name)s - %(levelname)s - %(message)s')\n", (619, 687), False, 'import logging\n'), ((1092, 1106), 'os.chdir', 'os.chdir', (['path'], {}), '(path)\n', (1100, 1106), False, 'import os\n'), ((5148, 5233), 'util_cloud.gcs_upload', 'util_cloud.gcs_upload', (['processed_data_annual', 'dataset_name'], {'gcs_bucket': 'gcsBucket'}), '(processed_data_annual, dataset_name, gcs_bucket=gcsBucket\n )\n', (5169, 5233), False, 'import util_cloud\n'), ((5481, 5500), 'ee.Initialize', 'ee.Initialize', (['auth'], {}), '(auth)\n', (5494, 5500), False, 'import ee\n'), ((5748, 5814), 'ee.data.createAsset', 'ee.data.createAsset', (["{'type': 'ImageCollection'}", 'image_collection'], {}), "({'type': 'ImageCollection'}, image_collection)\n", (5767, 5814), False, 'import ee\n'), ((5894, 5936), 'ee.data.setAssetAcl', 'ee.data.setAssetAcl', (['image_collection', 'acl'], {}), '(image_collection, acl)\n', (5913, 5936), False, 'import ee\n'), ((6873, 6926), 'util_cloud.gcs_remove', 'util_cloud.gcs_remove', (['gcs_uris'], {'gcs_bucket': 'gcsBucket'}), '(gcs_uris, gcs_bucket=gcsBucket)\n', (6894, 6926), False, 'import util_cloud\n'), ((7191, 7288), 'ee.ImageCollection', 'ee.ImageCollection', (['"""projects/resource-watch-gee/Facebook/PrecipitationAnalysis/GPCC_annual"""'], {}), "(\n 'projects/resource-watch-gee/Facebook/PrecipitationAnalysis/GPCC_annual')\n", (7209, 7288), False, 'import ee\n'), ((8079, 8150), 'ee.FeatureCollection', 'ee.FeatureCollection', (['"""projects/resource-watch-gee/gadm36_0_simplified"""'], {}), "('projects/resource-watch-gee/gadm36_0_simplified')\n", (8099, 8150), False, 'import ee\n'), ((8160, 8231), 'ee.FeatureCollection', 'ee.FeatureCollection', (['"""projects/resource-watch-gee/gadm36_1_simplified"""'], {}), "('projects/resource-watch-gee/gadm36_1_simplified')\n", (8180, 8231), False, 'import ee\n'), ((155, 182), 'sys.path.append', 'sys.path.append', (['utils_path'], {}), '(utils_path)\n', (170, 182), False, 'import sys\n'), ((1271, 1295), 'os.path.exists', 'os.path.exists', (['data_dir'], {}), '(data_dir)\n', (1285, 1295), False, 'import os\n'), ((1301, 1319), 'os.mkdir', 'os.mkdir', (['data_dir'], {}), '(data_dir)\n', (1309, 1319), False, 'import os\n'), ((3418, 3455), 'urllib.request.urlretrieve', 'urllib.request.urlretrieve', (['url', 'file'], {}), '(url, file)\n', (3444, 3455), False, 'import urllib\n'), ((3827, 3874), 'util_files.convert_netcdf', 'util_files.convert_netcdf', (['raw_file', "['precip']"], {}), "(raw_file, ['precip'])\n", (3852, 3874), False, 'import util_files\n'), ((3898, 3941), 'os.path.join', 'os.path.join', (["(raw_file[:-3] + '_precip.tif')"], {}), "(raw_file[:-3] + '_precip.tif')\n", (3910, 3941), False, 'import os\n'), ((4990, 5029), 'os.environ.get', 'os.environ.get', (['"""CLOUDSDK_CORE_PROJECT"""'], {}), "('CLOUDSDK_CORE_PROJECT')\n", (5004, 5029), False, 'import os\n'), ((5060, 5096), 'os.environ.get', 'os.environ.get', (['"""GEE_STAGING_BUCKET"""'], {}), "('GEE_STAGING_BUCKET')\n", (5074, 5096), False, 'import os\n'), ((5402, 5434), 'os.getenv', 'os.getenv', (['"""GEE_SERVICE_ACCOUNT"""'], {}), "('GEE_SERVICE_ACCOUNT')\n", (5411, 5434), False, 'import os\n'), ((5436, 5479), 'os.getenv', 'os.getenv', (['"""GOOGLE_APPLICATION_CREDENTIALS"""'], {}), "('GOOGLE_APPLICATION_CREDENTIALS')\n", (5445, 5479), False, 'import os\n'), ((6622, 6681), 'util_cloud.gee_manifest_complete', 'util_cloud.gee_manifest_complete', (['asset_name', 'uri', 'mf_bands'], {}), '(asset_name, uri, mf_bands)\n', (6654, 6681), False, 'import util_cloud\n'), ((6731, 6762), 'util_cloud.gee_ingest', 'util_cloud.gee_ingest', (['manifest'], {}), '(manifest)\n', (6752, 6762), False, 'import util_cloud\n'), ((7067, 7082), 'ee.Initialize', 'ee.Initialize', ([], {}), '()\n', (7080, 7082), False, 'import ee\n'), ((82, 109), 'os.getenv', 'os.getenv', (['"""PROCESSING_DIR"""'], {}), "('PROCESSING_DIR')\n", (91, 109), False, 'import os\n'), ((1030, 1051), 'os.getenv', 'os.getenv', (['"""BLOG_DIR"""'], {}), "('BLOG_DIR')\n", (1039, 1051), False, 'import os\n'), ((3322, 3343), 'os.path.basename', 'os.path.basename', (['url'], {}), '(url)\n', (3338, 3343), False, 'import os\n'), ((3618, 3639), 'gzip.open', 'gzip.open', (['file', '"""rb"""'], {}), "(file, 'rb')\n", (3627, 3639), False, 'import gzip\n'), ((4301, 4320), 'rasterio.open', 'rasterio.open', (['file'], {}), '(file)\n', (4314, 4320), False, 'import rasterio\n'), ((7110, 7127), 'ee.Authenticate', 'ee.Authenticate', ([], {}), '()\n', (7125, 7127), False, 'import ee\n'), ((7132, 7147), 'ee.Initialize', 'ee.Initialize', ([], {}), '()\n', (7145, 7147), False, 'import ee\n'), ((8407, 8429), 'ee.String', 'ee.String', (['country_ios'], {}), '(country_ios)\n', (8416, 8429), False, 'import ee\n'), ((9078, 9100), 'ee.String', 'ee.String', (['country_ios'], {}), '(country_ios)\n', (9087, 9100), False, 'import ee\n'), ((3710, 3741), 'shutil.copyfileobj', 'shutil.copyfileobj', (['f_in', 'f_out'], {}), '(f_in, f_out)\n', (3728, 3741), False, 'import shutil\n'), ((7729, 7750), 'numpy.arange', 'np.arange', (['(1891)', '(2020)'], {}), '(1891, 2020)\n', (7738, 7750), True, 'import numpy as np\n'), ((6286, 6307), 'os.path.basename', 'os.path.basename', (['uri'], {}), '(uri)\n', (6302, 6307), False, 'import os\n'), ((6454, 6475), 'os.path.basename', 'os.path.basename', (['uri'], {}), '(uri)\n', (6470, 6475), False, 'import os\n'), ((4125, 4144), 'rasterio.open', 'rasterio.open', (['file'], {}), '(file)\n', (4138, 4144), False, 'import rasterio\n'), ((8658, 8675), 'ee.Reducer.mean', 'ee.Reducer.mean', ([], {}), '()\n', (8673, 8675), False, 'import ee\n'), ((8715, 8743), 'ee.List', 'ee.List', (["['GID_0', 'NAME_0']"], {}), "(['GID_0', 'NAME_0'])\n", (8722, 8743), False, 'import ee\n'), ((9325, 9342), 'ee.Reducer.mean', 'ee.Reducer.mean', ([], {}), '()\n', (9340, 9342), False, 'import ee\n'), ((9382, 9500), 'ee.List', 'ee.List', (["['CC_1', 'ENGTYPE_1', 'GID_0', 'GID_1', 'HASC_1', 'NAME_0', 'NAME_1',\n 'NL_NAME_1', 'TYPE_1', 'VARNAME_1']"], {}), "(['CC_1', 'ENGTYPE_1', 'GID_0', 'GID_1', 'HASC_1', 'NAME_0',\n 'NAME_1', 'NL_NAME_1', 'TYPE_1', 'VARNAME_1'])\n", (9389, 9500), False, 'import ee\n')]
''' Some info on various layers, so you know what to expect depending on which layer you choose: layer 1: wavy layer 2: lines layer 3: boxes layer 4: circles? layer 6: dogs, bears, cute animals. layer 7: faces, buildings layer 8: fish begin to appear, frogs/reptilian eyes. layer 10: Monkies, lizards, snakes, duck Choosing various parameters like num_iterations, rescale, and num_repeats really varies on which layer you're doing. We could probably come up with some sort of formula. The deeper the layer is, the more iterations and repeats you will want. Layer 3: 20 iterations, 0.5 rescale, and 8 repeats is decent start Layer 10: 40 iterations and 25 repeats is good. ''' from deepdreamer import model, load_image, recursive_optimize import numpy as np import PIL.Image import os layer_tensor = model.layer_tensors[2] # Number from 1 to 10 for diffrent layers as mentioned above os.system('cls') for file_name in os.listdir('input'): if file_name.endswith(".jpg"): print('Starting ', file_name) img_result = load_image(filename='{}'.format('input/{}'.format(file_name))) img_result = recursive_optimize(layer_tensor=layer_tensor, image=img_result, # how clear is the dream vs original image num_iterations=20, step_size=1.0, rescale_factor=0.5, # How many "passes" over the data. More passes, the more granular the gradients will be. num_repeats=8, blend=0.2) img_result = np.clip(img_result, 0.0, 255.0) img_result = img_result.astype(np.uint8) result = PIL.Image.fromarray(img_result, mode='RGB') result.save('output/{}'.format(file_name)) print('Finished processing ', file_name) print('Done')
[ "deepdreamer.recursive_optimize", "os.system", "os.listdir", "numpy.clip" ]
[((924, 940), 'os.system', 'os.system', (['"""cls"""'], {}), "('cls')\n", (933, 940), False, 'import os\n'), ((959, 978), 'os.listdir', 'os.listdir', (['"""input"""'], {}), "('input')\n", (969, 978), False, 'import os\n'), ((1164, 1315), 'deepdreamer.recursive_optimize', 'recursive_optimize', ([], {'layer_tensor': 'layer_tensor', 'image': 'img_result', 'num_iterations': '(20)', 'step_size': '(1.0)', 'rescale_factor': '(0.5)', 'num_repeats': '(8)', 'blend': '(0.2)'}), '(layer_tensor=layer_tensor, image=img_result,\n num_iterations=20, step_size=1.0, rescale_factor=0.5, num_repeats=8,\n blend=0.2)\n', (1182, 1315), False, 'from deepdreamer import model, load_image, recursive_optimize\n'), ((1572, 1603), 'numpy.clip', 'np.clip', (['img_result', '(0.0)', '(255.0)'], {}), '(img_result, 0.0, 255.0)\n', (1579, 1603), True, 'import numpy as np\n')]
import numpy class Fickett: ''' calculate Fickett TESTCODE for full sequence NAR 10(17) 5303-531 modified from source code of CPAT 1.2.1 downloaded from https://sourceforge.net/projects/rna-cpat/files/?source=navbar ''' def __init__(self): '''new compiled Fickett look-up table''' self.position_parameter = [1.9,1.8,1.7,1.6,1.5,1.4,1.3,1.2,1.1,0.0] self.content_parameter = [0.33,0.31,0.29,0.27,0.25,0.23,0.21,0.19,0.17,0] self.position_probability = { "A":[0.51,0.55,0.57,0.52,0.48,0.58,0.57,0.54,0.50,0.36], "C":[0.29,0.44,0.55,0.49,0.52,0.60,0.60,0.56,0.51,0.38], "G":[0.62,0.67,0.74,0.65,0.61,0.62,0.52,0.41,0.31,0.17], "T":[0.51,0.60,0.69,0.64,0.62,0.67,0.58,0.48,0.39,0.24], } self.position_weight = {"A":0.062,"C":0.093,"G":0.205,"T":0.154} self.content_probability = { "A":[0.40,0.55,0.58,0.58,0.52,0.48,0.45,0.45,0.38,0.19], "C":[0.50,0.63,0.59,0.50,0.46,0.45,0.47,0.56,0.59,0.33], "G":[0.21,0.40,0.47,0.50,0.52,0.56,0.57,0.52,0.44,0.23], "T":[0.30,0.49,0.56,0.53,0.48,0.48,0.52,0.57,0.60,0.51] } self.content_weight = {"A":0.084,"C":0.076,"G":0.081,"T":0.055} def look_up_position_probability(self,value, base): ''' look up positional probability by base and value ''' if float(value) < 0: return None for idx,val in enumerate (self.position_parameter): if (float(value) >= val): return float(self.position_probability[base][idx]) * float(self.position_weight[base]) def look_up_content_probability(self,value, base): ''' look up content probability by base and value ''' if float(value) < 0: return None for idx,val in enumerate (self.content_parameter): if (float(value) >= val): return float(self.content_probability[base][idx]) * float(self.content_weight[base]) def fickett_value(self,dna): ''' calculate Fickett value from full RNA transcript sequence ''' if len(dna) < 2: return 0 fickett_score=0 dna=dna total_base = len(dna) A_content = float(dna.count("A"))/total_base C_content = float(dna.count("C"))/total_base G_content = float(dna.count("G"))/total_base T_content = float(dna.count("T"))/total_base phase_0 = dna[::3] phase_1 = dna[1::3] phase_2 = dna[2::3] phase_0_A = phase_0.count("A") phase_1_A = phase_1.count("A") phase_2_A = phase_2.count("A") phase_0_C = phase_0.count("C") phase_1_C = phase_1.count("C") phase_2_C = phase_2.count("C") phase_0_G = phase_0.count("G") phase_1_G = phase_1.count("G") phase_2_G = phase_2.count("G") phase_0_T = phase_0.count("T") phase_1_T = phase_1.count("T") phase_2_T = phase_2.count("T") A_content = float(phase_0_A + phase_1_A + phase_2_A)/total_base C_content = float(phase_0_C + phase_1_C + phase_2_C)/total_base G_content = float(phase_0_G + phase_1_G + phase_2_G)/total_base T_content = float(phase_0_T + phase_1_T + phase_2_T)/total_base A_position= numpy.max([phase_0_A,phase_1_A,phase_2_A])/(numpy.min([phase_0_A,phase_1_A,phase_2_A]) +1.0) C_position= numpy.max([phase_0_C,phase_1_C,phase_2_C])/(numpy.min([phase_0_C,phase_1_C,phase_2_C]) +1.0) G_position= numpy.max([phase_0_G,phase_1_G,phase_2_G])/(numpy.min([phase_0_G,phase_1_G,phase_2_G]) +1.0) T_position= numpy.max([phase_0_T,phase_1_T,phase_2_T])/(numpy.min([phase_0_T,phase_1_T,phase_2_T]) +1.0) fickett_score += self.look_up_content_probability(A_content,"A") fickett_score += self.look_up_content_probability(C_content,"C") fickett_score += self.look_up_content_probability(G_content,"G") fickett_score += self.look_up_content_probability(T_content,"T") fickett_score += self.look_up_position_probability(A_position,"A") fickett_score += self.look_up_position_probability(C_position,"C") fickett_score += self.look_up_position_probability(G_position,"G") fickett_score += self.look_up_position_probability(T_position,"T") return fickett_score
[ "numpy.min", "numpy.max" ]
[((3374, 3418), 'numpy.max', 'numpy.max', (['[phase_0_A, phase_1_A, phase_2_A]'], {}), '([phase_0_A, phase_1_A, phase_2_A])\n', (3383, 3418), False, 'import numpy\n'), ((3487, 3531), 'numpy.max', 'numpy.max', (['[phase_0_C, phase_1_C, phase_2_C]'], {}), '([phase_0_C, phase_1_C, phase_2_C])\n', (3496, 3531), False, 'import numpy\n'), ((3600, 3644), 'numpy.max', 'numpy.max', (['[phase_0_G, phase_1_G, phase_2_G]'], {}), '([phase_0_G, phase_1_G, phase_2_G])\n', (3609, 3644), False, 'import numpy\n'), ((3713, 3757), 'numpy.max', 'numpy.max', (['[phase_0_T, phase_1_T, phase_2_T]'], {}), '([phase_0_T, phase_1_T, phase_2_T])\n', (3722, 3757), False, 'import numpy\n'), ((3418, 3462), 'numpy.min', 'numpy.min', (['[phase_0_A, phase_1_A, phase_2_A]'], {}), '([phase_0_A, phase_1_A, phase_2_A])\n', (3427, 3462), False, 'import numpy\n'), ((3531, 3575), 'numpy.min', 'numpy.min', (['[phase_0_C, phase_1_C, phase_2_C]'], {}), '([phase_0_C, phase_1_C, phase_2_C])\n', (3540, 3575), False, 'import numpy\n'), ((3644, 3688), 'numpy.min', 'numpy.min', (['[phase_0_G, phase_1_G, phase_2_G]'], {}), '([phase_0_G, phase_1_G, phase_2_G])\n', (3653, 3688), False, 'import numpy\n'), ((3757, 3801), 'numpy.min', 'numpy.min', (['[phase_0_T, phase_1_T, phase_2_T]'], {}), '([phase_0_T, phase_1_T, phase_2_T])\n', (3766, 3801), False, 'import numpy\n')]
""" logreg.py This module contains functions to run and analyse logistic regressions to predict stimulus information from ROI activity for data generated by the Allen Institute OpenScope experiments for the Credit Assignment Project. Authors: <NAME> Date: October, 2018 Note: this code uses python 3.7. """ import os import copy import logging import warnings from matplotlib import pyplot as plt import numpy as np import pandas as pd import torch from analysis import quint_analys from util import data_util, file_util, gen_util, logger_util, logreg_util, \ math_util, plot_util from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util from plot_fcts import logreg_plots from util import gen_util logger = logging.getLogger(__name__) TAB = " " #### ALWAYS SET TO FALSE - CHANGE ONLY FOR TESTING PURPOSES TEST_BRICKS_VARIATIONS = False ############################################# def get_comps(stimtype="gabors", q1v4=False, regvsurp=False): """ get_comps() Returns comparisons that fit the criteria. Optional args: - stimtype (str) : stimtype default: "gabors" - q1v4 (bool) : if True, analysis is trained on first and tested on last quintiles default: False - regvsurp (bool): if True, analysis is trained on regular and tested on regular sequences default: False Returns: - comps (list): list of comparisons that fit the criteria """ if stimtype == "gabors": if regvsurp: raise ValueError("regvsurp can only be used with bricks.") comps = ["surp", "AvB", "AvC", "BvC", "DvU", "Aori", "Bori", "Cori", "Dori", "Uori", "DoriU", "DoriA", "BCDoriA", "BCDoriU", "ABCoriD", "ABCoriU"] elif stimtype == "bricks": comps = ["surp", "dir_all", "dir_surp", "dir_reg", "half_right", "half_left", "half_diff"] if regvsurp: comps = gen_util.remove_if( comps, ["surp", "dir_surp", "dir_all", "half_right", "half_left", "half_diff"]) if q1v4: comps = gen_util.remove_if( comps, ["half_left", "half_right", "half_diff"]) else: gen_util.accepted_values_error( "stimtype", stimtype, ["gabors", "bricks"]) return comps ############################################# def get_class_pars(comp="surp", stimtype="gabors"): """ get_class_pars() Returns name of the class determining variable, and the surprise values to use for the classes. Optional args: - comp (str) : type of comparison default: "surp" - stimtype (str) : stimulus type default: "gabors" Returns: - class_var (str) : variable separating classes (e.g., "surps", "gab_ori", "bri_dir") - surps (str or list): surprise values (for each class, if list) """ if stimtype == "gabors": if comp == "surp": class_var = "surps" surps = [0, 1] elif comp == "DvU": class_var = "surps" surps = [0, 1] elif "ori" in comp: class_var = "gab_ori" gab_letts = [lett.upper() for lett in comp.split("ori") if len(lett) > 0] surps = [] for lett in gab_letts: if ("D" in lett) ^ ("U" in lett): # exclusive or surp_val = 1 if "U" in lett else 0 surps.append(surp_val) else: surps.append("any") if len(gab_letts) == 1: surps = surps[0] elif "dir" in comp: raise ValueError("dir comparison not valid for gabors.") else: class_var = "gabfr" surps = "any" elif stimtype == "bricks": class_var = "bri_dir" if comp == "dir_all": surps = "any" elif comp == "dir_reg": surps = 0 elif comp == "dir_surp": surps = 1 elif comp == "surp": surps = [0, 1] class_var = "surps" elif comp in ["half_right", "half_left", "half_diff"]: surps = "any" class_var = comp else: raise ValueError("Only surp, dir_all, dir_reg, dir_surp, " "samehalf, diffhalf comparisons supported for Bricks.") return class_var, surps ############################################# def get_stimpar(comp="surp", stimtype="gabors", bri_dir="both", bri_size=128, gabfr=0, gabk=16, gab_ori="all", bri_pre=0.0): """ get_stimpar() Returns a stimulus parameter named tuple based on the stimulus parameters passed and comparison type. Optional args: - comp (str) : type of comparison default: "surp" - stimtype (str) : stimulus type default: "gabors" - bri_dir (str or list) : brick direction default: "both" - bri_size (int or list): brick direction default: 128 - gabfr (int or list) : gabor frame of reference (may be a list depending on "comp") default: 0 - gabk (int or list) : gabor kappa default: 16 - gab_ori (str) : gabor orientations ("all" or "shared"), for comp values like DoriU, DoriA, etc. default: "all" - bri_pre (int) : pre parameter for Bricks default: 0.0 Returns: - stimpar (StimPar) : named tuple containing stimulus parameters """ if stimtype == "bricks" and "half" in bri_dir or "dir" in bri_dir: logger.info("Ignoring brick dir setting.") if not (len(comp.replace("ori", "").upper()) > 1): gab_ori = "all" [bri_dir, bri_size, gabfr, gabk, gab_ori] = sess_gen_util.get_params( stimtype, bri_dir, bri_size, gabfr, gabk, gab_ori) if stimtype == "gabors": # DO NOT ALLOW OVERLAPPING if comp == "surp": stimpar = sess_ntuple_util.init_stimpar( stimtype, bri_dir, bri_size, gabfr, gabk, gab_ori, 0, 1.5) elif comp == "DvU": gabfr = sess_str_util.gabfr_nbrs(comp[0]) stimpar = sess_ntuple_util.init_stimpar( stimtype, bri_dir, bri_size, gabfr, gabk, gab_ori, 0, 0.45) elif "ori" in comp: gab_letts = [lett.upper() for lett in comp.split("ori") if len(lett) > 0] act_gabfr = [[sess_str_util.gabfr_nbrs(lett) for lett in letts] for letts in gab_letts] if len(act_gabfr) == 1: pre, post = 0, 0.45 if comp in ["Dori", "Uori"]: pre, post = 0, 0.6 act_gabfr = act_gabfr[0] if act_gabfr != gabfr: logger.info( f"Setting gabfr to {act_gabfr} instead of {gabfr}.") else: pre, post = -0.15, 0.45 gab_ori = sess_gen_util.gab_oris_shared_U(gab_letts, gab_ori) stimpar = sess_ntuple_util.init_stimpar( stimtype, bri_dir, bri_size, act_gabfr, gabk, gab_ori, pre, post) elif "dir" in comp or "half" in comp: raise ValueError("dir/half comparison not valid for gabors.") else: gabfrs = sess_str_util.gabfr_nbrs([comp[0], comp[2]]) stimpar = sess_ntuple_util.init_stimpar( stimtype, bri_dir, bri_size, gabfrs, gabk, gab_ori, 0, 0.45) elif stimtype == "bricks": # DO NOT ALLOW OVERLAPPING if "right" in comp: bri_dir = "right" elif "left" in comp: bri_dir = "left" stimpar = sess_ntuple_util.init_stimpar( stimtype, bri_dir, bri_size, gabfr, gabk, gab_ori, bri_pre, 1.0) # for brick logreg test analyses if TEST_BRICKS_VARIATIONS: logger.warning("Setting bricks pre/post to 2 for testing purposes.") stimpar = sess_ntuple_util.init_stimpar( stimtype, bri_dir, bri_size, gabfr, gabk, gab_ori, 2, 2) return stimpar ############################################# def get_rundir(run_val, uniqueid=None, alg="sklearn"): """ get_rundir(run_val) Returns the name of the specific subdirectory in which an analysis is saved, based on a run number and unique ID. Required args: - run_val (int): run number ("pytorch" alg) or number of run ("sklearn" alg) Optional args: - uniqueid (str or int): unique ID for analysis default: None - alg (str) : algorithm used to run logistic regression ("sklearn" or "pytorch") default: "sklearn" Returns: - rundir (str): name of subdirectory to save analysis in """ if uniqueid is None: if alg == "sklearn": rundir = f"{run_val}_runs" elif alg == "pytorch": rundir = f"run_{run_val}" else: gen_util.accepted_values_error("alg", alg, ["sklearn", "pytorch"]) else: rundir = f"{uniqueid}_{run_val}" return rundir ############################################# def get_compdir_dict(rundir, no_lists=False): """ get_compdir_dict(rundir) Returns a dictionary with analysis parameters based on the full analysis path. Required args: - rundir (str): path of subdirectory in which analysis is saved, structured as ".../m_s_plane_stim_fluor_scaled_comp_shuffled/ uniqueid_run" Optional args: - no_lists (bool): if True, list parameters are replaced with a string, e.g. "both" False Returns: - compdir_dict (dict): parameter dictionary - bri_dir (str or list) : Bricks direction parameter ("right", "left", ["right", "left"] or "none") - bri_size (int or list): Bricks size parameter (128, 256, [128, 256] or "none") - comp (str) : comparison parameter ("surp", "AvB", "AvC", "BvC" or "DvU", None) - fluor (str) : fluorescence parameter ("raw" or "dff") - gabk (int or list) : Gabor kappa parameter (4, 16, [4, 16] or "none") - plane (str) : plane ("soma" or "dend") - mouse_n (int) : mouse number - sess_n (int) : session number - scale (bool) : scaling parameter - run_n (int) : run number - shuffle (bool) : shuffle parameter - stimtype (str) : stimulus type ("gabors" or "bricks") - uniqueid (str) : unique ID (datetime, 6 digit number or None) """ parts = rundir.split(os.sep) param_str = parts[-2] run_str = parts[-1] compdir_dict = sess_gen_util.get_params_from_str(param_str, no_lists) if "run" in run_str: compdir_dict["uniqueid"] = None compdir_dict["run_n"] = int(run_str.split("_")[1]) else: compdir_dict["uniqueid"] = "_".join( [str(sub) for sub in run_str.split("_")[:-1]]) compdir_dict["run_n"] = int(run_str.split("_")[-1]) return compdir_dict ############################################# def get_df_name(task="analyse", stimtype="gabors", comp="surp", ctrl=False, alg="sklearn"): """ get_df_name() Returns a dictionary with analysis parameters based on the full analysis path. Optional args: - task (str) : type of task for which to get the dataframe default: "analyse" - stimtype (str): type of stimulus default: "gabors" - comp (str) : type of comparison default: "surp" - ctrl (bool) : if True, control comparisons are analysed default: False - alg (str) : algorithm used to run logistic regression ("sklearn" or "pytorch") default: "sklearn" Returns: - df_name (str): name of the dataframe """ alg_str = "" if alg == "pytorch": alg_str = "_pt" elif alg != "sklearn": gen_util.accepted_values_error("alg", alg, ["pytorch", "sklearn"]) ctrl_str = sess_str_util.ctrl_par_str(ctrl) sub_str = f"{stimtype[0:3]}_{comp}{ctrl_str}{alg_str}" if task == "collate": df_name = f"{sub_str}_all_scores_df.csv" elif task == "analyse": df_name = f"{sub_str}_score_stats_df.csv" return df_name ############################################# def info_dict(analyspar=None, sesspar=None, stimpar=None, extrapar=None, comp="surp", alg="sklearn", n_rois=None, epoch_n=None): """ info_dict() Returns an info dictionary from the parameters. Includes epoch number if it is passed. Returns an ordered list of keys instead if any of the dictionaries or namedtuples are None. Required args: - analyspar (AnalysPar): named tuple containing analysis parameters default: None - sesspar (SessPar) : named tuple containing session parameters default: None - stimpar (StimPar) : named tuple containing stimulus parameters default: None - extrapar (dict) : dictionary with extra parameters default: None ["run_n"] (int) : run number ["shuffle"] (bool): whether data is shuffled ["uniqueid"] (str): uniqueid Optional args: - comp (str) : comparison type default: "surp" - alg (str) : algorithm used to run logistic regression ("sklearn" or "pytorch") default: "sklearn" - n_rois (int) : number of ROIs default: None - epoch_n (int): epoch number default: None Returns: if all namedtuples and dictionaries are passed: - info (dict): analysis dictionary else if any are None: - info (list): list of dictionary keys """ if not any(par is None for par in [analyspar, sesspar, stimpar, extrapar]): if stimpar.stimtype == "bricks": bri_dir = gen_util.list_if_not(stimpar.bri_dir) if len(bri_dir) == 2: bri_dir = "both" else: bri_dir = bri_dir[0] else: bri_dir = stimpar.bri_dir info = {"mouse_n" : sesspar.mouse_n, "sess_n" : sesspar.sess_n, "plane" : sesspar.plane, "line" : sesspar.line, "fluor" : analyspar.fluor, "scale" : analyspar.scale, "shuffle" : extrapar["shuffle"], "stimtype": stimpar.stimtype, "bri_dir" : bri_dir, "comp" : comp, "uniqueid": extrapar["uniqueid"], "runtype" : sesspar.runtype, "n_rois" : n_rois } if alg == "pytorch": info["run_n"] = extrapar["run_n"] if epoch_n is not None: info["epoch_n"] = epoch_n # if no args are passed, just returns keys else: info = ["mouse_n", "sess_n", "plane", "line", "fluor", "scale", "shuffle", "stimtype", "bri_dir", "comp", "uniqueid", "run_n", "runtype", "n_rois", "epoch_n"] return info ############################################# def save_hyperpar(analyspar, logregpar, sesspar, stimpar, extrapar): """ save_hyperpar(analyspar, logregpar, sesspar, stimpar, extrapar) Saves the hyperparameters for an analysis. Required args: - analyspar (AnalysPar): named tuple containing analysis parameters - logregpar (LogRegPar): named tuple containing logistic regression parameters - sesspar (SessPar) : named tuple containing session parameters - stimpar (StimPar) : named tuple containing stimulus parameters - extrapar (dict) : dictionary with extra parameters ["dirname"] (str): directory in which to save hyperparameters Returns: - hyperpars (dict): hyperparameter dictionary with inputs as keys and named tuples converted to dictionaries """ hyperpars = {"analyspar": analyspar._asdict(), "logregpar": logregpar._asdict(), "sesspar" : sesspar._asdict(), "stimpar" : stimpar._asdict(), "extrapar" : extrapar } file_util.saveinfo(hyperpars, "hyperparameters.json", extrapar["dirname"]) return hyperpars ############################################# def get_classes(comp="surp", gab_ori="shared"): """ get_classes() Returns names for classes based on the comparison type. Optional args: - comp (str) : type of comparison default: "surp" - gab_ori (str or list): Gabor orientations default: "all" Returns: - classes (list): list of class names """ if gab_ori == "all": gab_ori = [0, 45, 90, 135] if comp == "surp": classes = ["Regular", "Surprise"] elif comp in ["AvB", "AvC", "BvC", "DvU"]: classes = [f"Gabor {fr}" for fr in [comp[0], comp[2]]] elif "ori" in comp: deg_vals = gab_ori stripped = comp.replace("ori", "") if stripped == "U": deg_vals = [val + 90 for val in deg_vals] elif len(stripped) == 2: deg_vals = gab_ori[0] deg = u"\u00B0" classes = [f"{val}{deg}" for val in deg_vals] elif "dir" in comp: classes = [sess_str_util.dir_par_str( direc, str_type="print").replace( "bricks (", "").replace(", ", " (").capitalize() for direc in ["right", "left"]] elif "half" in comp: classes = ["First half", "Second half"] else: gen_util.accepted_values_error("comp", comp, ["surp", "AvB", "AvC", "BvC", "DvU", "dir...", "...ori..."]) return classes ############################################# def get_data(stim, analyspar, stimpar, quintpar, qu_i=0, surp=[0, 1], n=1, remconsec_surps=False, get_2nd=False): """ get_data(sess, quintpar, stimpar) Returns ROI data based on specified criteria. Required args: - stim (Stim) : stimulus object - analyspar (AnalysPar): named tuple containing analysis parameters - stimpar (StimPar) : named tuple containing stimulus parameters - quintpar (QuintPar) : named tuple containing quintile parameters Optional args: - qu_i (int) : quintile index default: 0 - surp (list) : surprise values default: [0, 1] - n (int) : factor by which to multiply number of surprise values default: 1 - remconsec_surps (bool): whether consecutive segments are removed for surprise segments default: False - get_2nd (bool) : if True, every second segment is retained default: False Returns: - roi_data (3D array): ROI data, as sequences x frames x ROIs - surp_n (int) : Number of surprise sequences """ # data for single quintile # first number of surprises, then segs for t, surp_use in enumerate([1, surp]): remconsec = (remconsec_surps and surp_use == 1) segs = quint_analys.quint_segs( stim, stimpar, quintpar.n_quints, qu_i, surp_use, remconsec=remconsec)[0][0] # get alternating for consecutive segments if get_2nd and not remconsec: segs = gen_util.get_alternating_consec(segs, first=False) if t == 0: surp_n = len(segs) * n twop_fr = stim.get_twop_fr_by_seg(segs, first=True)["first_twop_fr"] # do not scale (scaling factors cannot be based on test data) roi_data = gen_util.reshape_df_data( stim.get_roi_data(twop_fr, stimpar.pre, stimpar.post, analyspar.fluor, remnans=True, scale=False), squeeze_cols=True) # for brick logreg test analyses if TEST_BRICKS_VARIATIONS: if remconsec_surps: # Normalize to first half mid = roi_data.shape[-1] // 2 div = np.median(roi_data[:, :, : mid], axis=-1) roi_data = roi_data - np.expand_dims(div, -1) # # Mean only if TEST_BRICKS_VARIATIONS == "mean": logger.warning("Using mean across ROIs, for testing purposes.") # 1 x seqs x frames roi_data = np.expand_dims(np.nanmean(roi_data, axis=0), axis=0) # Mean and std elif TEST_BRICKS_VARIATIONS == "mean_std": logger.warning("Using mean and standard deviation across ROIs, " "for testing purposes.") roi_data = np.stack([np.nanmean(roi_data, axis=0), np.nanstd(roi_data, axis=0)], axis=0) # transpose to seqs x frames x ROIs roi_data = np.transpose(roi_data, [1, 2, 0]) return roi_data, surp_n ############################################# def get_sess_data(sess, analyspar, stimpar, quintpar, class_var="surps", surps=[0, 1], regvsurp=False, split_oris=False): """ get_sess_data(sess, analyspar, stimpar, quintpar) Logs session information and returns ROI trace segments, target classes and class information and number of surprise segments in the dataset. Required args: - sess (Session) : session - analyspar (AnalysPar): named tuple containing analysis parameters - stimpar (StimPar) : named tuple containing stimulus parameters - quintpar (QuintPar) : named tuple containing quintile parameters Optional args: - class_var (str) : class determining variable ("surps" or stimpar attribute) default: "surps" - surps (list, str, int) : surprise value(s) (list if class_var is "surps", otherwise 0, 1 or "any") - regvsurp (bool) : if True, the first dataset will include regular sequences and the second will include surprise sequences default: False - split_oris (bool or list): List of Gabor frames for each split, or False if splitting orientation comparison is not applicable. default: False Returns: - roi_seqs (list) : list of 3D arrays of selected ROI trace seqs (1 or 2 if an additional test set is included), each structured as sequences x frames x ROIs - seq_classes (list): list of 2D arrays of sequence classes (1 or 2 if an additional test set is included), each structured as class values x 1 - n_surps (list) : list of lists of number of surprise sequences (doubled if "half" comparison), structured as datasets x class """ stim = sess.get_stim(stimpar.stimtype) split_oris = split_oris is not False # set to boolean if (regvsurp + (len(quintpar.qu_idx) > 1) + ("half" in class_var) + split_oris) > 1: raise ValueError("Cannot combine any of the following: separating " "quintiles, regvsurp, half comparisons, multiple Gabor frame " "orientation comparisons.") elif len(quintpar.qu_idx) > 2: raise ValueError("Max of 2 quintiles expected.") elif split_oris and len(stimpar.gabfr) > 2: raise ValueError("Max of 2 Gabor frame sets expected for orientation " "classification.") # check for stimulus pre/post problems pre_post_err = False get_2nd, remconsec_surps = False, False if stimpar.pre > 0: if stimpar.stimtype == "bricks": if class_var == "surps": remconsec_surps = True elif stimpar.pre == 1: get_2nd = True else: pre_post_err = True else: pre_post_err = True if stimpar.post > 1.0: if not stimpar.stimtype == "gabors" and stimpar.post <= 1.5: pre_post_err = True if pre_post_err: raise NotImplementedError("Not implemented to prevent sequence overlap " f"for {stimpar.stimtype}: {stimpar.pre} pre/{stimpar.post} post " f"for {class_var} classification") n = 1 if class_var == "surps": n_cl = len(surps) elif "half" in class_var: n_cl = 2 # DOUBLE surp ns to compensate for shorter blocks, if using control n = 2 if "diff" in class_var: quintpar = sess_ntuple_util.init_quintpar( 4, [[1, 2]], [None], [None]) if len(np.unique(stim.direcs)) != 2: raise ValueError( "Segments do not fit these criteria (missing directions).") else: quintpar = sess_ntuple_util.init_quintpar( 2, [[0, 1]], [None], [None]) else: n_cl = len(stimpar._asdict()[class_var]) # modify surps, qu_idx, gabfr to cycle through datasets if len(quintpar.qu_idx) == 2: surps = [surps, surps] gabfr_idxs = ["ignore", "ignore"] if regvsurp: raise ValueError( "Cannot set regvsurp to True if more than 1 quintile.") if "part" in class_var: raise ValueError("Cannot do half comparisons with quintiles.") elif regvsurp: surps = [surps, 1-surps] gabfr_idxs = ["ignore", "ignore"] quintpar = sess_ntuple_util.init_quintpar( 1, [0, 0], [None, None], [None, None]) elif split_oris: surps = surps gabfr_idxs = [0, 1] quintpar = sess_ntuple_util.init_quintpar( 1, [0, 0], [None, None], [None, None]) else: surps = [surps] gabfr_idxs = ["ignore"] gabfr_idxs = [0, 1] if split_oris else ["ignore", "ignore"] # cycle through classes roi_seqs = [[] for _ in range(len(quintpar.qu_idx))] seq_classes = [[] for _ in range(len(quintpar.qu_idx))] surp_ns = [[] for _ in range(len(quintpar.qu_idx))] # cycle through data groups (quint or regvsurp or gabfr for oris) for d, (qu_i, subsurps, gabfr_idx) in enumerate( zip(quintpar.qu_idx, surps, gabfr_idxs)): for cl in range(n_cl): use_qu_i = [qu_i] surp = subsurps stimpar_sp = stimpar if class_var == "surps": surp = subsurps[cl] elif "half" in class_var: use_qu_i = [qu_i[cl]] else: keys = class_var vals = stimpar._asdict()[class_var][cl] if split_oris: keys = [keys, "gabfr", "gab_ori"] vals = [vals, stimpar.gabfr[gabfr_idx], stimpar.gab_ori[gabfr_idx][cl]] # modify stimpar stimpar_sp = sess_ntuple_util.get_modif_ntuple( stimpar, keys, vals) roi_data, surp_n = get_data( stim, analyspar, stimpar_sp, quintpar, qu_i=use_qu_i, surp=surp, remconsec_surps=remconsec_surps, n=n, get_2nd=get_2nd) roi_seqs[d].append(roi_data) seq_classes[d].append(np.full(len(roi_data), cl)) surp_ns[d].append(surp_n) # concatenate data split by class along trial seqs axis roi_seqs[d] = np.concatenate(roi_seqs[d], axis=0) seq_classes[d] = np.concatenate(seq_classes[d], axis=0) # get logistic variance across datasets log_var = np.log(np.var(np.concatenate(roi_seqs, axis=0))) n_fr, nrois = roi_seqs[0].shape[1:] # in training set if stimpar.stimtype == "gabors": surp_use = surps[0] if surp_use == [0, 1] and not isinstance(stimpar.gabfr, list): surp_use = "any" if split_oris: gabfr_lett = [sess_str_util.gabfr_letters( gabfr, surp=surp_use) for gabfr in stimpar.gabfr] gabfr_lett = " -> ".join([str(lett) for lett in gabfr_lett]) else: gabfr_lett = sess_str_util.gabfr_letters( stimpar.gabfr, surp=surp_use) stim_info = f"\nGab fr: {gabfr_lett}\nGab K: {stimpar.gabk}" elif stimpar.stimtype == "bricks": stim_info = (f"\nBri dir: {stimpar.bri_dir}\n" f"Bri size: {stimpar.bri_size}") logger.info(f"Runtype: {sess.runtype}\nMouse: {sess.mouse_n}\n" f"Sess: {sess.sess_n}\nPlane: {sess.plane}\nLine: {sess.line}\n" f"Fluor: {analyspar.fluor}\nROIs: {nrois}{stim_info}\n" f"Frames per seg: {n_fr}\nLogvar: {log_var:.2f}", extra={"spacing": "\n"}) return roi_seqs, seq_classes, surp_ns ############################################# def sample_seqs(roi_seqs, seq_classes, n_surp): """ sample_seqs(roi_seqs, seq_classes, n_surp) Samples sequences to correspond to the ratio of surprise to regular sequences. Required args: - roi_seqs (3D array) : array of all ROI trace sequences, structured as: sequences x frames x ROIs - seq_classes (2D array): array of all sequence classes (0, 1), structured as class values x 1 - n_surp (int) : number of surprise sequences Returns: - roi_seqs (3D array) : array of selected ROI trace sequences, structured as sequences x frames x ROIs - seq_classes (2D array): array of sequence classes, structured as class values x 1 """ if np.unique(seq_classes).tolist() != [0, 1]: raise ValueError("Function expects classes 0 and 1 only.") class0_all = np.where(seq_classes == 0)[0] class1_all = np.where(seq_classes == 1)[0] n_reg = (len(class0_all) + len(class1_all))//2 - n_surp class0_idx = np.random.choice(class0_all, n_reg, replace=False) class1_idx = np.random.choice(class1_all, n_surp, replace=False) roi_seqs = np.concatenate( [roi_seqs[class0_idx], roi_seqs[class1_idx]], axis=0) seq_classes = np.concatenate( [seq_classes[class0_idx], seq_classes[class1_idx]], axis=0) return roi_seqs, seq_classes ############################################# def save_tr_stats(plot_data, plot_targ, data_names, analyspar, stimpar, n_rois, alg="sklearn",mod=None, dirname="."): """ save_tr_stats(plot_data, plot_targ, data_names, stimpar, n_rois) Extracts, saves and returns trace statistics in a json in the specified directory. Required args: - plot_data (list) : list of 3D arrays of selected ROI trace seqs to be plotted, each structured as sequences x frames x ROIs - plot_targ (list) : list of 2D arrays of sequence classes to be plotted, each structured as class values x 1 - data_names (list) : names for each plot_data array - analyspar (AnalysPar): named tuple containing analysis parameters - stimpar (SessPar) : named tuple containing stimulus parameters - n_rois (int) : number of ROIs in data Optional args: - alg (str) : algorithm used to run logistic regression ("sklearn" or "pytorch") default: "sklearn" - mod (sklearn pipeline): sklearn pipeline (model). Required if alg is "sklearn" default: None - dirname (str) : directory in which to save the traces default: "." Returns: - tr_stats (dict) : dictionary of trace stats data ["n_rois"] (int) : number of ROIs ["train_ns"] (list) : number of segments per class ["train_class_stats"] (3D array) : training statistics, structured as class x stats (me, err) x frames ["xran"] (array-like) : x values for frames optionally, if an additional named set (e.g., "test_Q4") is passed: ["set_ns"] (list) : number of segments per class ["set_class_stats"] (3D array): trace statistics, structured as class x stats (me, err) x frames """ if len(data_names) != len(plot_data): raise ValueError("Not as many 'plot_data' items as 'data_names'.") tr_stats = {"n_rois": n_rois} classes = np.unique(plot_targ[0]) for data, targ, name in zip(plot_data, plot_targ, data_names): # get stats if data is None: continue if alg == "sklearn": # scales, flattens and optionally shuffles data data = logreg_util.get_transf_data_sk(mod, data, False, name=="train") elif alg == "pytorch": data = data.numpy() targ = targ.numpy() else: gen_util.accepted_values_error("alg", alg, ["sklearn", "pytorch"]) xran, class_stats, ns = logreg_util.get_stats( data, targ, stimpar.pre, stimpar.post, classes, analyspar.stats, analyspar.error) tr_stats["xran"] = xran.tolist() tr_stats[f"{name}_class_stats"] = class_stats.tolist() tr_stats[f"{name}_ns"] = ns file_util.saveinfo(tr_stats, "tr_stats.json", dirname) return tr_stats ############################################# @logreg_util.catch_set_problem def init_logreg_model_pt(roi_seqs, seq_classes, logregpar, extrapar, scale=True, device="cpu", thresh_cl=2): """ init_logreg_model_pt(roi_seqs, seq_classes, logregpar, extrapar) Initializes and returns the pytorch logreg model and dataloaders. Required args: - roi_seqs (list) : list of 3D arrays of selected ROI trace seqs (1 or 2 if an additional test set is included), each structured as sequences x frames x ROIs - seq_classes (list) : list of 2D arrays of sequence classes (0 or 1) (1 or 2 if an additional test set is included), each structured as class values x 1 - logregpar (LogRegPar) : named tuple containing logistic regression parameters - extrapar (dict) : dictionary with extra parameters ["shuffle"] (bool): if True, data is shuffled Optional args: - scale (bool) : whether data is scaled by ROI default: True - device (str) : device to use default: "cpu" - thresh_cl (int) : size threshold for classes in each non empty set beneath which the indices are reselected (only if targets are passed). Not checked if thresh_cl is 0. default: 2 Returns: - model (torch.nn.Module) : Neural network module with optimizer and loss function as attributes - dls (list of torch DataLoaders): list of torch DataLoaders for each set. If a set is empty, the corresponding dls value is None. - extrapar (dict) : dictionary with extra parameters ["cl_wei"] (list) : list of weights for each class ["loss_name"] (str) : name of the loss function used ["sc_facts"] (list) : min/max value(s) with which to scale ["shuffle"] (bool) : if True, data is shuffled ["shuff_reidx"] (list) : list of indices with which targets were shuffled """ sc_dim = "last" if scale else "none" if np.unique(seq_classes[0]).tolist() != [0, 1]: raise NotImplementedError("This Pytorch logreg function is " "implemented only for classes 0 and 1.") if len(roi_seqs) > 2: raise ValueError("Must pass no more than 2 sets of data, but " f"found {len(roi_seqs)}.") dl_info = data_util.create_dls( roi_seqs[0], seq_classes[0], train_p=logregpar.train_p, sc_dim=sc_dim, sc_type="stand_rob", extrem="perc", shuffle=extrapar["shuffle"], batchsize=logregpar.batchsize, thresh_cl=thresh_cl) dls = dl_info[0] extrapar = copy.deepcopy(extrapar) if scale: extrapar["sc_facts"] = dl_info[-1] if len(roi_seqs) == 2: class_vals = seq_classes[1] if extrapar["shuffle"]: np.random.shuffle(class_vals) if scale: roi_seqs[1] = data_util.scale_datasets( torch.Tensor(roi_seqs[1]), sc_facts=extrapar["sc_facts"])[0] dls.append(data_util.init_dl( roi_seqs[1], class_vals, logregpar.batchsize)) if extrapar["shuffle"]: extrapar["shuff_reidx"] = dl_info[1] # from train targets extrapar["cl_wei"] = logreg_util.class_weights(dls[0].dataset.targets) n_fr, n_rois = roi_seqs[0].shape[1:] model = logreg_util.LogReg(n_rois, n_fr).to(device) model.opt = torch.optim.Adam( model.parameters(), lr=logregpar.lr, weight_decay=logregpar.wd) model.loss_fn = logreg_util.weighted_BCE(extrapar["cl_wei"]) extrapar["loss_name"] = model.loss_fn.name return model, dls, extrapar ############################################# def save_scores(info, scores, key_order=None, dirname=""): """ save_scores(args, scores, saved_eps) Saves run information and scores per epoch as a dataframe. Required args: - info (dict) : dictionary of parameters (see info_dict()) - scores (pd DataFrame): dataframe of recorded scores Optional args: - key_order (list): ordered list of keys default: None - dirname (str) : directory in which to save scores default: "" Returns: - summ_df (pd DataFrame): dataframe with scores and recorded parameters """ summ_df = copy.deepcopy(scores) if key_order is None: key_order = info.keys() for key in reversed(key_order): if key in info.keys(): summ_df.insert(0, key, info[key]) file_util.saveinfo(summ_df, "scores_df.csv", dirname) return summ_df ############################################# def setup_run(quintpar, extrapar, techpar, sess_data, comp="surp", gab_ori="all"): """ setup_run(quintpar, extrapar, techpar, sess_data) Sets up run(s) by setting seed, getting classes and number of ROIs. Required args: - quintpar (QuintPar) : named tuple containing quintile parameters - extrapar (dict) : dictionary with extra parameters ["seed"] (int) : seed to use ["shuffle"] (bool): if analysis is on shuffled data ["uniqueid"] (str): unique ID for the analysis - techpar (dict) : dictionary with technical parameters ["alg"] (str) : algorithm used ("sklearn" or "pytorch") ["compdir"] (str) : specific output comparison directory ["device"] (str) : device to use (e.g., "cpu" or "cuda") ["fontdir"] (str) : directory in which additional fonts are located ["output"] (str) : main output directiory ["plt_bkend"] (str): plt backend to use (e.g., "agg" or None) ["reseed"] (bool) : if True, run is reseeded - sess_data (list): list of session data: - roi_seqs (list) : list of 3D arrays of selected ROI trace seqs (1 or 2 if an additional test set is included), each structured as sequences x frames x ROIs - seq_classes (list): list of 2D arrays of sequence classes (1 or 2 if an additional test set is included), each structured as class values x 1 - n_surps (list) : list of lists of number of surprise sequences (doubled if "half" comparison), structured as datasets x class Optional args: - comp (str) : comparison default: "surp" - gab_ori (str or list): Gabor orientations, for comp values like DoriU, DoriA, etc. default: "shared" Returns: - extrapar (dict) : dictionary with extra parameters ["classes"] (list): list of class names ["n_rois"] (int) : number of ROIs ["seed"] (int) : seed to use ["shuffle"] (bool): if analysis is on shuffled data ["uniqueid"] (str): unique ID for the analysis - roi_seqs (list) : list of 3D arrays of selected ROI trace seqs (1 or 2 if an additional test set is included), each structured as sequences x frames x ROIs - seq_classes (list): list of 2D arrays of sequence classes (1 or 2 if an additional test set is included), each structured as class values x 1 - n_surps (list) : list of lists of number of surprise sequences (doubled if "half" comparison), structured as datasets x class """ extrapar = copy.deepcopy(extrapar) if techpar["reseed"]: # reset seed extrapar["seed"] = None extrapar["seed"] = gen_util.seed_all( extrapar["seed"], techpar["device"], seed_torch=(techpar["alg"] == "pytorch")) # seed torch, if needed extrapar["classes"] = get_classes(comp, gab_ori) [roi_seqs, seq_classes, n_surps] = copy.deepcopy(sess_data) extrapar["n_rois"] = roi_seqs[0].shape[-1] return extrapar, roi_seqs, seq_classes, n_surps ############################################# def all_runs_sk(n_runs, analyspar, logregpar, quintpar, sesspar, stimpar, extrapar, techpar, sess_data): """ all_runs_sk(n_runs, analyspar, logregpar, quintpar, sesspar, stimpar, extrapar, techpar, sess_data) Does all runs of a logistic regression on the specified comparison and session data using sklearn. Records hyperparameters, all models, tr_stats dictionary. Plots scores and training statistics. Required args: - n_runs (int) : number of runs to do - analyspar (AnalysPar): named tuple containing analysis parameters - logregpar (LogRegPar): named tuple containing logistic regression parameters - quintpar (QuintPar) : named tuple containing quintile parameters - sesspar (SessPar) : named tuple containing session parameters - stimpar (StimPar) : named tuple containing stimulus parameters - extrapar (dict) : dictionary with extra parameters ["seed"] (int) : seed to use ["shuffle"] (bool): if analysis is on shuffled data ["uniqueid"] (str): unique ID for the analysis - techpar (dict) : dictionary with technical parameters ["compdir"] (str) : specific output comparison directory ["device"] (str) : device to use (e.g., "cpu" or "cuda") ["fontdir"] (str) : directory in which additional fonts are located ["output"] (str) : main output directiory ["plt_bkend"] (str): plt backend to use (e.g., "agg" or None) ["reseed"] (bool) : if True, run is reseeded - sess_data (list): list of session data: - roi_seqs (list) : list of 3D arrays of selected ROI trace seqs (1 or 2 if an additional test set is included), each structured as sequences x frames x ROIs - seq_classes (list): list of 2D arrays of sequence classes (1 or 2 if an additional test set is included), each structured as class values x 1 - n_surps (list) : list of lists of number of surprise sequences (doubled if "half" comparison), structured as datasets x class """ logregpar = sess_ntuple_util.get_modif_ntuple( logregpar, ["batchsize", "lr", "wd"], [None, None, None]) if techpar["device"] == "cuda": warnings.warn("sklearn method not implemented with GPU.") [extrapar, roi_seqs, seq_classes, n_surps] = setup_run( quintpar, extrapar, techpar, sess_data, logregpar.comp, gab_ori=stimpar.gab_ori) main_data = [roi_seqs[0], seq_classes[0]] samples = [False for _ in n_surps] if logregpar.ctrl: # get n_surp for last class samples = [n_surp[-1] for n_surp in n_surps] keep_all = [val in logregpar.comp for val in ["ori", "dir_reg", "half"]] if sum(keep_all) > 0: # get n_surp for all classes samples = n_surps plot_util.manage_mpl(techpar["plt_bkend"], fontdir=techpar["fontdir"]) extrapar = copy.deepcopy(extrapar) extrapar["n_runs"] = n_runs rundir = get_rundir(extrapar["n_runs"], extrapar["uniqueid"], logregpar.alg) extrapar["dirname"] = file_util.createdir([techpar["output"], techpar["compdir"], rundir]) scale_str = sess_str_util.scale_par_str(analyspar.scale, "print") shuff_str = sess_str_util.shuff_par_str(extrapar["shuffle"], "labels") logger.info(f"Runs ({n_runs}): {scale_str}{shuff_str}", extra={"spacing": "\n"}) split_test = False returns = logreg_util.run_logreg_cv_sk( main_data[0], main_data[1], logregpar._asdict(), extrapar, analyspar.scale, samples[0], split_test, extrapar["seed"], techpar["parallel"], catch_set_prob=True) if returns is None: return else: mod_cvs, cv, extrapar = returns hyperpars = save_hyperpar(analyspar, logregpar, sesspar, stimpar, extrapar) # Additional data for scoring set_names = ["train", "test"] extra_data, extra_cv = None, None extra_name = sess_str_util.ext_test_str( logregpar.q1v4, logregpar.regvsurp, logregpar.comp) if cv._split_test: set_names.append("test_out") if len(roi_seqs) == 2: extra_data = [roi_seqs[1], seq_classes[1]] extra_cv = logreg_util.StratifiedShuffleSplitMod( n_splits=cv.n_splits, train_p=0.5, sample=samples[1], bal=logregpar.bal) # since train_p cannot be 1.0 if extra_name is None: raise ValueError("Extra test dataset not labelled.") set_names.append(extra_name) mod_cvs = logreg_util.test_logreg_cv_sk( mod_cvs, cv, extrapar["scoring"], main_data=main_data, extra_data=extra_data, extra_name=extra_name, extra_cv=extra_cv, catch_set_prob=True) if mod_cvs is None: return # Get and save best model best_mod_idx = np.argmax(mod_cvs[f"{set_names[-1]}_neg_log_loss"]) best_mod = mod_cvs["estimator"][best_mod_idx] # Save data used for best model plot_data, plot_targ, data_names = [], [], [] for name, subcv, data in zip( ["train", extra_name], [cv, extra_cv], [main_data, extra_data]): if data is None: continue if name == "train": idx = subcv._set_idx[best_mod_idx][0] else: idx = [i for sub in subcv._set_idx[best_mod_idx] for i in sub] plot_data.append(data[0][idx]) plot_targ.append(data[1][idx]) data_names.append(name) tr_stats = save_tr_stats( plot_data, plot_targ, data_names, analyspar, stimpar, extrapar["n_rois"], logregpar.alg, best_mod, dirname=extrapar["dirname"]) # Get scores dataframe scores = logreg_util.create_score_df_sk( mod_cvs, best_mod_idx, set_names, extrapar["scoring"]) info = info_dict( analyspar, sesspar, stimpar, extrapar, logregpar.comp, logregpar.alg, n_rois=extrapar["n_rois"]) full_scores = save_scores( info, scores, key_order=info_dict(), dirname=extrapar["dirname"]) # Save best model logreg_plots.plot_traces_scores( hyperpars, tr_stats, full_scores, plot_wei=best_mod_idx) plt.close("all") ############################################# def single_run_pt(run_n, analyspar, logregpar, quintpar, sesspar, stimpar, extrapar, techpar, sess_data): """ single_run_pt(run_n, analyspar, logregpar, quintpar, sesspar, stimpar, extrapar, techpar, sess_data) Does a single run of a logistic regression using PyTorch on the specified comparison and session data. Records hyperparameters, best model, last model, tr_stats dictionary. Plots scores and training statistics. Required args: - run_n (int) : run number - analyspar (AnalysPar): named tuple containing analysis parameters - logregpar (LogRegPar): named tuple containing logistic regression parameters - quintpar (QuintPar) : named tuple containing quintile parameters - sesspar (SessPar) : named tuple containing session parameters - stimpar (StimPar) : named tuple containing stimulus parameters - extrapar (dict) : dictionary with extra parameters ["seed"] (int) : seed to use ["shuffle"] (bool): if analysis is on shuffled data ["uniqueid"] (str): unique ID for the analysis - techpar (dict) : dictionary with technical parameters ["compdir"] (str) : specific output comparison directory ["device"] (str) : device to use (e.g., "cpu" or "cuda") ["fontdir"] (str) : directory in which additional fonts are located ["output"] (str) : main output directiory ["plt_bkend"] (str): plt backend to use (e.g., "agg" or None) ["reseed"] (bool) : if True, run is reseeded - sess_data (list): list of session data: - roi_seqs (list) : list of 3D arrays of selected ROI trace seqs (1 or 2 if an additional test set is included), each structured as sequences x frames x ROIs - seq_classes (list): list of 2D arrays of sequence classes (1 or 2 if an additional test set is included), each structured as class values x 1 - n_surps (list) : number of surprise sequences, listed by quintile, (doubled if "half" comparison) """ extrapar = copy.deepcopy(extrapar) extrapar["seed"] *= run_n + 1 # ensure different seed for each run [extrapar, roi_seqs, seq_classes, n_surps] = setup_run( quintpar, extrapar, techpar, sess_data, logregpar.comp, gab_ori=stimpar.gab_ori) extrapar["run_n"] = run_n scale_str = sess_str_util.scale_par_str(analyspar.scale, "print") shuff_str = sess_str_util.shuff_par_str(extrapar["shuffle"], "labels") run_n = extrapar["run_n"] logger.info(f"Run: {run_n}{scale_str}{shuff_str}", extra={"spacing": "\n"}) for i in range(len(roi_seqs)): if logregpar.ctrl: # select a random subsample roi_seqs[i], seq_classes[i] = sample_seqs( roi_seqs[i], seq_classes[i], n_surps[i][-1]) if logregpar.bal: # balance classes roi_seqs[i], seq_classes[i] = data_util.bal_classes( roi_seqs[i], seq_classes[i]) plot_util.manage_mpl(techpar["plt_bkend"], fontdir=techpar["fontdir"]) thresh_cl = 2 if sesspar.runtype == "pilot": thresh_cl = 1 rundir = get_rundir(extrapar["run_n"], extrapar["uniqueid"], logregpar.alg) extrapar["dirname"] = file_util.createdir( [techpar["output"], techpar["compdir"], rundir]) returns = init_logreg_model_pt( roi_seqs, seq_classes, logregpar, extrapar, analyspar.scale, techpar["device"], thresh_cl=thresh_cl, catch_set_prob=True) if returns is None: return else: mod, dls, extrapar = returns hyperpars = save_hyperpar(analyspar, logregpar, sesspar, stimpar, extrapar) data_names = ["train"] extra_name = sess_str_util.ext_test_str( logregpar.q1v4, logregpar.regvsurp, logregpar.comp) idx = [0] if len(roi_seqs) == 2: idx.append(-1) if extra_name == "": raise ValueError("Extra test dataset not labelled.") data_names.append(extra_name) plot_data = [dls[i].dataset.data for i in idx] plot_targ = [dls[i].dataset.targets for i in idx] tr_stats = save_tr_stats( plot_data, plot_targ, data_names, analyspar, stimpar, extrapar["n_rois"], alg=logregpar.alg, dirname=extrapar["dirname"]) info = info_dict( analyspar, sesspar, stimpar, extrapar, logregpar.comp, logregpar.alg, n_rois=tr_stats["n_rois"]) scores = logreg_util.fit_model_pt( info, logregpar.n_epochs, mod, dls, techpar["device"], extrapar["dirname"], ep_freq=techpar["ep_freq"], test_dl2_name=extra_name) logger.info("Run {}: training done.\n".format(extrapar["run_n"])) # save scores in dataframe full_scores = save_scores( info, scores, key_order=info_dict(), dirname=extrapar["dirname"]) # plot traces and scores (only for first 5 runs) # no plotting for the rest to reduce number of files generated if run_n <= 5: logreg_plots.plot_traces_scores( hyperpars, tr_stats, full_scores, plot_wei=True) plt.close("all") ############################################# def run_regr(sess, analyspar, stimpar, logregpar, quintpar, extrapar, techpar): """ Does runs of a logistic regressions on the specified comparison on a session. Required args: - sess (Session) : Session on which to run logistic regression - analyspar (AnalysPar): named tuple containing analysis parameters - stimpar (StimPar) : named tuple containing stimulus parameters - logregpar (LogRegPar): named tuple containing logistic regression analysis parameters - quintpar (QuintPar) : named tuple containing quintile analysis parameters - extrapar (dict) : dictionary containing additional analysis parameters ["uniqueid"] (str or int): unique ID for analysis ["seed"] (int) : seed to seed random processes with - techpar (dict) : dictionary containing technical analysi parameters ["device"] (str) : device name (i.e., "cuda" or "cpu") ["ep_freq"] (int) : frequency at which to log loss to console ["fontdir"] (str) : directory in which additional fonts are located ["n_reg"] (int) : number of regular runs ["n_shuff"] (int) : number of shuffled runs ["output"] (str) : general directory in which to save output ["parallel"] (bool): if True, runs are done in parallel ["plt_bkend"] (str): pyplot backend to use ["reseed"] (bool) : if True, each run is reseeded """ techpar = copy.deepcopy(techpar) sesspar = sess_ntuple_util.init_sesspar( sess.sess_n, False, sess.plane,sess.line, runtype=sess.runtype, mouse_n=sess.mouse_n) class_var, surps = get_class_pars(logregpar.comp, stimpar.stimtype) split_oris = sess_str_util.get_split_oris(logregpar.comp) try: sess_data = get_sess_data( sess, analyspar, stimpar, quintpar, class_var, surps, regvsurp=logregpar.regvsurp, split_oris=split_oris) except ValueError as err: catch_phr = ["fit these criteria", "No frames"] catch = sum(phr in str(err) for phr in catch_phr) if catch: warnings.warn(str(err)) return else: raise err for n_runs, shuffle in zip( [techpar["n_reg"], techpar["n_shuff"]], [False, True]): if n_runs == 0: continue extrapar = copy.deepcopy(extrapar) extrapar["shuffle"] = shuffle techpar = copy.deepcopy(techpar) techpar["compdir"] = sess_gen_util.get_analysdir( sesspar.mouse_n, sesspar.sess_n, sesspar.plane, analyspar.fluor, analyspar.scale, stimpar.stimtype, stimpar.bri_dir, stimpar.bri_size, stimpar.gabk, logregpar.comp, logregpar.ctrl, extrapar["shuffle"]) if logregpar.alg == "pytorch": techpar["compdir"] = "{}_pt".format(techpar["compdir"]) if n_runs == 0: continue # optionally runs in parallel args_list = [analyspar, logregpar, quintpar, sesspar, stimpar, extrapar, techpar, sess_data] gen_util.parallel_wrap( single_run_pt, range(n_runs), args_list, parallel=techpar["parallel"]) elif logregpar.alg == "sklearn": all_runs_sk(n_runs, analyspar, logregpar, quintpar, sesspar, stimpar, extrapar, techpar, sess_data) else: gen_util.accepted_values_error("logregpar.alg", logregpar.alg, ["pytorch", "sklearn"]) ############################################# def collate_scores(direc, all_labels, alg="sklearn"): """ collate_scores(direc, all_labels) Collects the analysis information and scores from the last epoch recorded for a run and returns in dataframe. Required args: - direc (str) : path to the specific comparison run folder - all_labels (list): ordered list of columns to save to dataframe Optional args: - alg (str): algorithm used to run logistic regression ("sklearn" or "pytorch") default: "sklearn" Return: - scores (pd DataFrame): Dataframe containing run analysis information and scores from the last epoch recorded. """ logger.info(direc) scores = pd.DataFrame(columns=all_labels) ep_info, hyperpars = logreg_util.get_scores(direc, alg) if ep_info is None: comp_dict = get_compdir_dict(direc, no_lists=True) comp_dict["scale"] = hyperpars["analyspar"]["scale"] comp_dict["runtype"] = hyperpars["sesspar"]["runtype"] comp_dict["line"] = hyperpars["sesspar"]["line"] comp_dict["n_rois"] = hyperpars["extrapar"]["n_rois"] for col in all_labels: # ensures correct order if col in comp_dict.keys(): scores.loc[0, col] = comp_dict[col] else: for col in all_labels: if col in ep_info.columns: if alg == "pytorch": scores.loc[0, col] = ep_info[col].item() elif alg == "sklearn": scores[col] = ep_info[col] return scores ############################################# def remove_overlap_comp_dir(gen_dirs, stimtype="gabors", comp="surp"): """ remove_overlap_comp_dir(gen_dirs) Returns list of directories with those corresponding to comparisons with overlapping names removed. Required args: - gen_dirs (list): list of directories Optional args: - stimtype (str) : stimulus type default: "gabors" - comp (str) : type of comparison default: "surp" """ all_comps = get_comps(stimtype) for other_comp in all_comps: if comp != other_comp and comp in other_comp: gen_dirs = [gen_dir for gen_dir in gen_dirs if other_comp not in gen_dir] return gen_dirs ############################################# def adjust_duplicate_runs(all_scores): """ adjust_duplicate_runs(all_scores) Adjust the run numbers for duplicate runs so that they are sequential and not repeating. Required args: - all_scores (pd DataFrame): dataframe compiling all the scores for logistic regression runs Returns: - all_scores (pd DataFrame): dataframe compiling all the scores for logistic regression runs, with run numbers adjusted if needed to be sequential and non repeating. """ grouping_keys = list(filter( lambda x: x not in ["uniqueid", "run_n", "epoch_n"], info_dict())) # runs are expected to occur sequentially without gaps in the dataframe # for any grouping. for _, group in all_scores.groupby(grouping_keys): breaks = group.loc[group["run_n"].diff() < 0].index for break_pt in breaks: current_max = group.loc[0 : break_pt, "run_n"].max() group.loc[break_pt :, "run_n"] += current_max + 1 if len(breaks): all_scores.loc[group.index, "run_n"] = group["run_n"] return all_scores ############################################# def run_collate(output, stimtype="gabors", comp="surp", ctrl=False, alg="sklearn", parallel=False): """ run_collate(output) Collects the analysis information and scores from the last epochs recorded for all runs for a comparison type, and saves to a dataframe. Overwrites any existing dataframe of collated data. Required args: - output (str): general directory in which summary dataframe is saved Optional args: - stimtype (str) : stimulus type default: "gabors" - comp (str) : type of comparison default: "surp" - ctrl (bool) : if True, control comparisons are analysed default: False - alg (str) : algorithm used to run logistic regression ("sklearn" or "pytorch") default: "sklearn" - parallel (bool): if True, run information is collected in parallel default: False Returns: - all_scores (pd DataFrame): dataframe compiling all the scores for logistic regression runs in the output folder that correspond to the stimulus type and comparison type criteria """ if not os.path.isdir(output): logger.info(f"{output} does not exist.") return ext_test = sess_str_util.ext_test_str( ("q1v4" in output), ("rvs" in output), comp) if ext_test == "": ext_test = None ctrl_str = sess_str_util.ctrl_par_str(ctrl) gen_dirs = file_util.getfiles( output, "subdirs", [stimtype[0:3], comp, ctrl_str]) if alg == "sklearn": gen_dirs = [gen_dir for gen_dir in gen_dirs if "_pt" not in gen_dir] elif alg == "pytorch": gen_dirs = [gen_dir for gen_dir in gen_dirs if "_pt" in gen_dir] else: gen_util.accepted_values_error("alg", alg, ["sklearn", "pytorch"]) gen_dirs = remove_overlap_comp_dir(gen_dirs, stimtype, comp) if not ctrl: gen_dirs = [gen_dir for gen_dir in gen_dirs if "ctrl" not in gen_dir] if len(gen_dirs) == 0: logger.info("No runs found.") return run_dirs = [run_dir for gen_dir in gen_dirs for run_dir in file_util.getfiles(gen_dir, "subdirs")] # identify, remove and flag empty directories empty_dirs = [run_dir for run_dir in run_dirs if len(os.listdir(run_dir)) == 0] if len(empty_dirs) != 0: logger.info("EMPTY DIRECTORIES:") for empty_dir in empty_dirs: run_dirs.remove(empty_dir) logger.info(f"{empty_dir}", extra={"spacing": TAB}) all_labels = info_dict() + \ logreg_util.get_sc_labs(True, ext_test_name=ext_test) + ["saved"] scores_list = gen_util.parallel_wrap( collate_scores, run_dirs, [all_labels, alg], parallel=parallel) # check for repeated run numbers if len(scores_list) != 0: all_scores = pd.concat(scores_list) all_scores = all_scores[all_labels] # reorder else: all_scores = pd.DataFrame(columns=all_labels) # check for and adjust duplicate run numbers all_scores = all_scores.reset_index(drop=True) all_scores = adjust_duplicate_runs(all_scores) # sort df by mouse, session, plane, line, fluor, scale, shuffle, stimtype, # comp, uniqueid, run_n, runtype sorter = info_dict()[0:13] all_scores = all_scores.sort_values(by=sorter).reset_index(drop=True) savename = get_df_name("collate", stimtype, comp, ctrl, alg) file_util.saveinfo(all_scores, savename, output, overwrite=True) return all_scores ############################################# def calc_stats(scores_summ, curr_lines, curr_idx, CI=0.95, ext_test=None, stats="mean", shuffle=False): """ calc_stats(scores_summ, curr_lines, curr_idx) Calculates statistics on scores from runs with specific analysis criteria and records them in the summary scores dataframe. Required args: - scores_summ (pd DataFrame): DataFrame containing scores summary - curr_lines (pd DataFrame) : DataFrame lines corresponding to specific analysis criteria - curr_idx (int) : Current row in the scores summary DataFrame Optional args: - CI (num) : Confidence interval around which to collect percentile values default: 0.95 - extra_test (str): Name of extra test set, if any (None if none) default: None - stats (str) : stats to take, i.e., "mean" or "median" default: "mean" - shuffle (bool) : If True, data is for shuffled, and will be averaged across runs before taking stats default: False Returns: - scores_summ (pd DataFrame): Updated DataFrame containing scores, as well as epoch_n, runs_total, runs_nan summaries """ scores_summ = copy.deepcopy(scores_summ) # score labels to perform statistics on sc_labs = ["epoch_n"] + logreg_util.get_sc_labs( True, ext_test_name=ext_test) # avoids accidental nuisance dropping by pandas curr_lines["epoch_n"] = curr_lines["epoch_n"].astype(float) if shuffle: # group runs and take mean or median across scores_summ.loc[curr_idx, "mouse_n"] = -1 keep_lines = \ [col for col in curr_lines.columns if col in sc_labs] + ["run_n"] grped_lines = curr_lines[keep_lines].groupby("run_n", as_index=False) if stats == "mean": curr_lines = grped_lines.mean() # automatically skips NaNs elif stats == "median": curr_lines = grped_lines.median() # automatically skips NaNs else: gen_util.accepted_values_error("stats", stats, ["mean", "median"]) # calculate n_runs (without nans and with) scores_summ.loc[curr_idx, "runs_total"] = len(curr_lines) scores_summ.loc[curr_idx, "runs_nan"] = curr_lines["epoch_n"].isna().sum() # percentiles to record ps, p_names = math_util.get_percentiles(CI) for sc_lab in sc_labs: if sc_lab in curr_lines.keys(): cols = [] vals = [] data = curr_lines[sc_lab].astype(float) for stat in ["mean", "median"]: cols.extend([stat]) vals.extend( [math_util.mean_med(data, stats=stat, nanpol="omit")]) for error in ["std", "sem"]: cols.extend([error]) vals.extend([math_util.error_stat( data, stats="mean", error=error, nanpol="omit")]) # get 25th and 75th quartiles cols.extend(["q25", "q75"]) vals.extend(math_util.error_stat( data, stats="median", error="std", nanpol="omit")) # get other percentiles (for CI) cols.extend(p_names) vals.extend(math_util.error_stat( data, stats="median", error="std", nanpol="omit", qu=ps)) # get MAD cols.extend(["mad"]) vals.extend([math_util.error_stat( data, stats="median", error="sem", nanpol="omit")]) # plug in values cols = [f"{sc_lab}_{name}" for name in cols] gen_util.set_df_vals(scores_summ, curr_idx, cols, vals) return scores_summ ############################################# def run_analysis(output, stimtype="gabors", comp="surp", ctrl=False, CI=0.95, alg="sklearn", parallel=False, all_scores_df=None): """ run_analysis(output) Calculates statistics on scores from runs for each specific analysis criteria and saves them in the summary scores dataframe. Overwrites any existing dataframe of analysed data. Required args: - output (str): general directory in which summary dataframe is saved. Optional args: - stimtype (str) : stimulus type default: "gabors" - comp (str) : type of comparison default: "surp" - ctrl (bool) : if True, control comparisons are analysed default: False - CI (num) : CI for shuffled data default: 0.95 - alg (str) : algorithm used to run logistic regression ("sklearn" or "pytorch") default: "sklearn" - parallel (bool) : if True, run information is collected in parallel default: False - all_scores_df (pd df): already collated scores dataframe default: None Returns: - scores_summ (pd DataFrame): dataframe with analysed scores """ if all_scores_df is None: all_scores_df = run_collate(output, stimtype, comp, ctrl, alg, parallel) stats = "mean" # across runs for shuffle CIs if all_scores_df is None: return scores_summ = pd.DataFrame() ext_test = sess_str_util.ext_test_str( ("q1v4" in output), ("rvs" in output), comp) if ext_test == "": ext_test = None # common labels comm_labs = gen_util.remove_if(info_dict(), ["uniqueid", "run_n", "epoch_n"]) # get all unique comb of labels for acr_shuff in [False, True]: if not acr_shuff: df_unique = all_scores_df[comm_labs].drop_duplicates() else: df_unique = all_scores_df[gen_util.remove_if(comm_labs, ["mouse_n", "n_rois"])].drop_duplicates() for _, df_row in df_unique.iterrows(): if acr_shuff and not df_row["shuffle"]: # second pass, only shuffle continue vals = [df_row[x] for x in comm_labs] curr_lines = gen_util.get_df_vals(all_scores_df, comm_labs, vals) # assign values to current line in summary df curr_idx = len(scores_summ) gen_util.set_df_vals(scores_summ, curr_idx, comm_labs, vals) # calculate stats scores_summ = calc_stats(scores_summ, curr_lines, curr_idx, CI, ext_test, stats=stats, shuffle=acr_shuff) savename = get_df_name("analyse", stimtype, comp, ctrl, alg) file_util.saveinfo(scores_summ, savename, output, overwrite=True) return scores_summ ############################################# def run_plot(output, stimtype="gabors", comp="surp", ctrl=False, bri_dir="both", fluor="dff", scale=True, CI=0.95, alg="sklearn", plt_bkend=None, fontdir=None, modif=False): """ run_plot(output) Plots summary data for a specific comparison, for each datatype in a separate figure and saves figures. Required args: - output (str): general directory in which summary dataframe is saved Optional args: - stimtype (str) : stimulus type default: "gabors" - comp (str) : type of comparison default: "surp" - ctrl (bool) : if True, control comparisons are analysed default: False - bri_dir (str) : brick direction default: "both" - fluor (str) : fluorescence trace type default: "dff" - scale (bool) : whether data is scaled by ROI default: True - CI (num) : CI for shuffled data default: 0.95 - alg (str) : algorithm used to run logistic regression ("sklearn" or "pytorch") default: "sklearn" - plt_bkend (str): pyplot backend to use default: None - fontdir (str) : directory in which additional fonts are located default: None - modif (bool) : if True, plots are made in a modified (simplified way) default: False """ if comp in ["half_right", "half_left"]: bri_dir = comp.replace("half_", "") savename = get_df_name("analyse", stimtype, comp, ctrl, alg) logreg_plots.plot_summ(output, savename, stimtype, comp, ctrl, bri_dir, fluor, scale, CI, plt_bkend, fontdir, modif)
[ "logging.getLogger", "sess_util.sess_ntuple_util.init_quintpar", "util.gen_util.parallel_wrap", "util.file_util.createdir", "plot_fcts.logreg_plots.plot_traces_scores", "util.logreg_util.fit_model_pt", "util.logreg_util.get_transf_data_sk", "util.gen_util.list_if_not", "util.gen_util.get_df_vals", ...
[((736, 763), 'logging.getLogger', 'logging.getLogger', (['__name__'], {}), '(__name__)\n', (753, 763), False, 'import logging\n'), ((6334, 6409), 'sess_util.sess_gen_util.get_params', 'sess_gen_util.get_params', (['stimtype', 'bri_dir', 'bri_size', 'gabfr', 'gabk', 'gab_ori'], {}), '(stimtype, bri_dir, bri_size, gabfr, gabk, gab_ori)\n', (6358, 6409), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((11794, 11848), 'sess_util.sess_gen_util.get_params_from_str', 'sess_gen_util.get_params_from_str', (['param_str', 'no_lists'], {}), '(param_str, no_lists)\n', (11827, 11848), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((13297, 13329), 'sess_util.sess_str_util.ctrl_par_str', 'sess_str_util.ctrl_par_str', (['ctrl'], {}), '(ctrl)\n', (13323, 13329), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((17797, 17871), 'util.file_util.saveinfo', 'file_util.saveinfo', (['hyperpars', '"""hyperparameters.json"""', "extrapar['dirname']"], {}), "(hyperpars, 'hyperparameters.json', extrapar['dirname'])\n", (17815, 17871), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((22648, 22681), 'numpy.transpose', 'np.transpose', (['roi_data', '[1, 2, 0]'], {}), '(roi_data, [1, 2, 0])\n', (22660, 22681), True, 'import numpy as np\n'), ((32023, 32073), 'numpy.random.choice', 'np.random.choice', (['class0_all', 'n_reg'], {'replace': '(False)'}), '(class0_all, n_reg, replace=False)\n', (32039, 32073), True, 'import numpy as np\n'), ((32091, 32142), 'numpy.random.choice', 'np.random.choice', (['class1_all', 'n_surp'], {'replace': '(False)'}), '(class1_all, n_surp, replace=False)\n', (32107, 32142), True, 'import numpy as np\n'), ((32163, 32231), 'numpy.concatenate', 'np.concatenate', (['[roi_seqs[class0_idx], roi_seqs[class1_idx]]'], {'axis': '(0)'}), '([roi_seqs[class0_idx], roi_seqs[class1_idx]], axis=0)\n', (32177, 32231), True, 'import numpy as np\n'), ((32260, 32334), 'numpy.concatenate', 'np.concatenate', (['[seq_classes[class0_idx], seq_classes[class1_idx]]'], {'axis': '(0)'}), '([seq_classes[class0_idx], seq_classes[class1_idx]], axis=0)\n', (32274, 32334), True, 'import numpy as np\n'), ((35005, 35028), 'numpy.unique', 'np.unique', (['plot_targ[0]'], {}), '(plot_targ[0])\n', (35014, 35028), True, 'import numpy as np\n'), ((35839, 35893), 'util.file_util.saveinfo', 'file_util.saveinfo', (['tr_stats', '"""tr_stats.json"""', 'dirname'], {}), "(tr_stats, 'tr_stats.json', dirname)\n", (35857, 35893), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((38920, 39137), 'util.data_util.create_dls', 'data_util.create_dls', (['roi_seqs[0]', 'seq_classes[0]'], {'train_p': 'logregpar.train_p', 'sc_dim': 'sc_dim', 'sc_type': '"""stand_rob"""', 'extrem': '"""perc"""', 'shuffle': "extrapar['shuffle']", 'batchsize': 'logregpar.batchsize', 'thresh_cl': 'thresh_cl'}), "(roi_seqs[0], seq_classes[0], train_p=logregpar.train_p,\n sc_dim=sc_dim, sc_type='stand_rob', extrem='perc', shuffle=extrapar[\n 'shuffle'], batchsize=logregpar.batchsize, thresh_cl=thresh_cl)\n", (38940, 39137), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((39193, 39216), 'copy.deepcopy', 'copy.deepcopy', (['extrapar'], {}), '(extrapar)\n', (39206, 39216), False, 'import copy\n'), ((39786, 39835), 'util.logreg_util.class_weights', 'logreg_util.class_weights', (['dls[0].dataset.targets'], {}), '(dls[0].dataset.targets)\n', (39811, 39835), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((40074, 40118), 'util.logreg_util.weighted_BCE', 'logreg_util.weighted_BCE', (["extrapar['cl_wei']"], {}), "(extrapar['cl_wei'])\n", (40098, 40118), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((40923, 40944), 'copy.deepcopy', 'copy.deepcopy', (['scores'], {}), '(scores)\n', (40936, 40944), False, 'import copy\n'), ((41123, 41176), 'util.file_util.saveinfo', 'file_util.saveinfo', (['summ_df', '"""scores_df.csv"""', 'dirname'], {}), "(summ_df, 'scores_df.csv', dirname)\n", (41141, 41176), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((44507, 44530), 'copy.deepcopy', 'copy.deepcopy', (['extrapar'], {}), '(extrapar)\n', (44520, 44530), False, 'import copy\n'), ((44634, 44733), 'util.gen_util.seed_all', 'gen_util.seed_all', (["extrapar['seed']", "techpar['device']"], {'seed_torch': "(techpar['alg'] == 'pytorch')"}), "(extrapar['seed'], techpar['device'], seed_torch=techpar[\n 'alg'] == 'pytorch')\n", (44651, 44733), False, 'from util import gen_util\n'), ((44870, 44894), 'copy.deepcopy', 'copy.deepcopy', (['sess_data'], {}), '(sess_data)\n', (44883, 44894), False, 'import copy\n'), ((47587, 47683), 'sess_util.sess_ntuple_util.get_modif_ntuple', 'sess_ntuple_util.get_modif_ntuple', (['logregpar', "['batchsize', 'lr', 'wd']", '[None, None, None]'], {}), "(logregpar, ['batchsize', 'lr', 'wd'], [\n None, None, None])\n", (47620, 47683), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((48324, 48394), 'util.plot_util.manage_mpl', 'plot_util.manage_mpl', (["techpar['plt_bkend']"], {'fontdir': "techpar['fontdir']"}), "(techpar['plt_bkend'], fontdir=techpar['fontdir'])\n", (48344, 48394), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((48411, 48434), 'copy.deepcopy', 'copy.deepcopy', (['extrapar'], {}), '(extrapar)\n', (48424, 48434), False, 'import copy\n'), ((48574, 48642), 'util.file_util.createdir', 'file_util.createdir', (["[techpar['output'], techpar['compdir'], rundir]"], {}), "([techpar['output'], techpar['compdir'], rundir])\n", (48593, 48642), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((48673, 48726), 'sess_util.sess_str_util.scale_par_str', 'sess_str_util.scale_par_str', (['analyspar.scale', '"""print"""'], {}), "(analyspar.scale, 'print')\n", (48700, 48726), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((48743, 48801), 'sess_util.sess_str_util.shuff_par_str', 'sess_str_util.shuff_par_str', (["extrapar['shuffle']", '"""labels"""'], {}), "(extrapar['shuffle'], 'labels')\n", (48770, 48801), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((49447, 49525), 'sess_util.sess_str_util.ext_test_str', 'sess_str_util.ext_test_str', (['logregpar.q1v4', 'logregpar.regvsurp', 'logregpar.comp'], {}), '(logregpar.q1v4, logregpar.regvsurp, logregpar.comp)\n', (49473, 49525), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((50007, 50187), 'util.logreg_util.test_logreg_cv_sk', 'logreg_util.test_logreg_cv_sk', (['mod_cvs', 'cv', "extrapar['scoring']"], {'main_data': 'main_data', 'extra_data': 'extra_data', 'extra_name': 'extra_name', 'extra_cv': 'extra_cv', 'catch_set_prob': '(True)'}), "(mod_cvs, cv, extrapar['scoring'], main_data=\n main_data, extra_data=extra_data, extra_name=extra_name, extra_cv=\n extra_cv, catch_set_prob=True)\n", (50036, 50187), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((50294, 50345), 'numpy.argmax', 'np.argmax', (["mod_cvs[f'{set_names[-1]}_neg_log_loss']"], {}), "(mod_cvs[f'{set_names[-1]}_neg_log_loss'])\n", (50303, 50345), True, 'import numpy as np\n'), ((51145, 51235), 'util.logreg_util.create_score_df_sk', 'logreg_util.create_score_df_sk', (['mod_cvs', 'best_mod_idx', 'set_names', "extrapar['scoring']"], {}), "(mod_cvs, best_mod_idx, set_names, extrapar[\n 'scoring'])\n", (51175, 51235), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((51508, 51601), 'plot_fcts.logreg_plots.plot_traces_scores', 'logreg_plots.plot_traces_scores', (['hyperpars', 'tr_stats', 'full_scores'], {'plot_wei': 'best_mod_idx'}), '(hyperpars, tr_stats, full_scores, plot_wei=\n best_mod_idx)\n', (51539, 51601), False, 'from plot_fcts import logreg_plots\n'), ((51611, 51627), 'matplotlib.pyplot.close', 'plt.close', (['"""all"""'], {}), "('all')\n", (51620, 51627), True, 'from matplotlib import pyplot as plt\n'), ((54169, 54192), 'copy.deepcopy', 'copy.deepcopy', (['extrapar'], {}), '(extrapar)\n', (54182, 54192), False, 'import copy\n'), ((54470, 54523), 'sess_util.sess_str_util.scale_par_str', 'sess_str_util.scale_par_str', (['analyspar.scale', '"""print"""'], {}), "(analyspar.scale, 'print')\n", (54497, 54523), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((54540, 54598), 'sess_util.sess_str_util.shuff_par_str', 'sess_str_util.shuff_par_str', (["extrapar['shuffle']", '"""labels"""'], {}), "(extrapar['shuffle'], 'labels')\n", (54567, 54598), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((55086, 55156), 'util.plot_util.manage_mpl', 'plot_util.manage_mpl', (["techpar['plt_bkend']"], {'fontdir': "techpar['fontdir']"}), "(techpar['plt_bkend'], fontdir=techpar['fontdir'])\n", (55106, 55156), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((55344, 55412), 'util.file_util.createdir', 'file_util.createdir', (["[techpar['output'], techpar['compdir'], rundir]"], {}), "([techpar['output'], techpar['compdir'], rundir])\n", (55363, 55412), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((55810, 55888), 'sess_util.sess_str_util.ext_test_str', 'sess_str_util.ext_test_str', (['logregpar.q1v4', 'logregpar.regvsurp', 'logregpar.comp'], {}), '(logregpar.q1v4, logregpar.regvsurp, logregpar.comp)\n', (55836, 55888), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((56519, 56682), 'util.logreg_util.fit_model_pt', 'logreg_util.fit_model_pt', (['info', 'logregpar.n_epochs', 'mod', 'dls', "techpar['device']", "extrapar['dirname']"], {'ep_freq': "techpar['ep_freq']", 'test_dl2_name': 'extra_name'}), "(info, logregpar.n_epochs, mod, dls, techpar[\n 'device'], extrapar['dirname'], ep_freq=techpar['ep_freq'],\n test_dl2_name=extra_name)\n", (56543, 56682), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((57156, 57172), 'matplotlib.pyplot.close', 'plt.close', (['"""all"""'], {}), "('all')\n", (57165, 57172), True, 'from matplotlib import pyplot as plt\n'), ((58924, 58946), 'copy.deepcopy', 'copy.deepcopy', (['techpar'], {}), '(techpar)\n', (58937, 58946), False, 'import copy\n'), ((58961, 59081), 'sess_util.sess_ntuple_util.init_sesspar', 'sess_ntuple_util.init_sesspar', (['sess.sess_n', '(False)', 'sess.plane', 'sess.line'], {'runtype': 'sess.runtype', 'mouse_n': 'sess.mouse_n'}), '(sess.sess_n, False, sess.plane, sess.line,\n runtype=sess.runtype, mouse_n=sess.mouse_n)\n', (58990, 59081), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((59185, 59229), 'sess_util.sess_str_util.get_split_oris', 'sess_str_util.get_split_oris', (['logregpar.comp'], {}), '(logregpar.comp)\n', (59213, 59229), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((61831, 61863), 'pandas.DataFrame', 'pd.DataFrame', ([], {'columns': 'all_labels'}), '(columns=all_labels)\n', (61843, 61863), True, 'import pandas as pd\n'), ((61890, 61924), 'util.logreg_util.get_scores', 'logreg_util.get_scores', (['direc', 'alg'], {}), '(direc, alg)\n', (61912, 61924), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((66317, 66384), 'sess_util.sess_str_util.ext_test_str', 'sess_str_util.ext_test_str', (["('q1v4' in output)", "('rvs' in output)", 'comp'], {}), "('q1v4' in output, 'rvs' in output, comp)\n", (66343, 66384), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((66461, 66493), 'sess_util.sess_str_util.ctrl_par_str', 'sess_str_util.ctrl_par_str', (['ctrl'], {}), '(ctrl)\n', (66487, 66493), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((66509, 66579), 'util.file_util.getfiles', 'file_util.getfiles', (['output', '"""subdirs"""', '[stimtype[0:3], comp, ctrl_str]'], {}), "(output, 'subdirs', [stimtype[0:3], comp, ctrl_str])\n", (66527, 66579), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((67757, 67847), 'util.gen_util.parallel_wrap', 'gen_util.parallel_wrap', (['collate_scores', 'run_dirs', '[all_labels, alg]'], {'parallel': 'parallel'}), '(collate_scores, run_dirs, [all_labels, alg],\n parallel=parallel)\n', (67779, 67847), False, 'from util import gen_util\n'), ((68531, 68595), 'util.file_util.saveinfo', 'file_util.saveinfo', (['all_scores', 'savename', 'output'], {'overwrite': '(True)'}), '(all_scores, savename, output, overwrite=True)\n', (68549, 68595), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((70177, 70203), 'copy.deepcopy', 'copy.deepcopy', (['scores_summ'], {}), '(scores_summ)\n', (70190, 70203), False, 'import copy\n'), ((71281, 71310), 'util.math_util.get_percentiles', 'math_util.get_percentiles', (['CI'], {}), '(CI)\n', (71306, 71310), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((74414, 74428), 'pandas.DataFrame', 'pd.DataFrame', ([], {}), '()\n', (74426, 74428), True, 'import pandas as pd\n'), ((74445, 74512), 'sess_util.sess_str_util.ext_test_str', 'sess_str_util.ext_test_str', (["('q1v4' in output)", "('rvs' in output)", 'comp'], {}), "('q1v4' in output, 'rvs' in output, comp)\n", (74471, 74512), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((75698, 75763), 'util.file_util.saveinfo', 'file_util.saveinfo', (['scores_summ', 'savename', 'output'], {'overwrite': '(True)'}), '(scores_summ, savename, output, overwrite=True)\n', (75716, 75763), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((77658, 77778), 'plot_fcts.logreg_plots.plot_summ', 'logreg_plots.plot_summ', (['output', 'savename', 'stimtype', 'comp', 'ctrl', 'bri_dir', 'fluor', 'scale', 'CI', 'plt_bkend', 'fontdir', 'modif'], {}), '(output, savename, stimtype, comp, ctrl, bri_dir,\n fluor, scale, CI, plt_bkend, fontdir, modif)\n', (77680, 77778), False, 'from plot_fcts import logreg_plots\n'), ((29507, 29542), 'numpy.concatenate', 'np.concatenate', (['roi_seqs[d]'], {'axis': '(0)'}), '(roi_seqs[d], axis=0)\n', (29521, 29542), True, 'import numpy as np\n'), ((29568, 29606), 'numpy.concatenate', 'np.concatenate', (['seq_classes[d]'], {'axis': '(0)'}), '(seq_classes[d], axis=0)\n', (29582, 29606), True, 'import numpy as np\n'), ((31868, 31894), 'numpy.where', 'np.where', (['(seq_classes == 0)'], {}), '(seq_classes == 0)\n', (31876, 31894), True, 'import numpy as np\n'), ((31915, 31941), 'numpy.where', 'np.where', (['(seq_classes == 1)'], {}), '(seq_classes == 1)\n', (31923, 31941), True, 'import numpy as np\n'), ((35564, 35671), 'util.logreg_util.get_stats', 'logreg_util.get_stats', (['data', 'targ', 'stimpar.pre', 'stimpar.post', 'classes', 'analyspar.stats', 'analyspar.error'], {}), '(data, targ, stimpar.pre, stimpar.post, classes,\n analyspar.stats, analyspar.error)\n', (35585, 35671), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((47732, 47789), 'warnings.warn', 'warnings.warn', (['"""sklearn method not implemented with GPU."""'], {}), "('sklearn method not implemented with GPU.')\n", (47745, 47789), False, 'import warnings\n'), ((49692, 49806), 'util.logreg_util.StratifiedShuffleSplitMod', 'logreg_util.StratifiedShuffleSplitMod', ([], {'n_splits': 'cv.n_splits', 'train_p': '(0.5)', 'sample': 'samples[1]', 'bal': 'logregpar.bal'}), '(n_splits=cv.n_splits, train_p=0.5,\n sample=samples[1], bal=logregpar.bal)\n', (49729, 49806), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((57057, 57142), 'plot_fcts.logreg_plots.plot_traces_scores', 'logreg_plots.plot_traces_scores', (['hyperpars', 'tr_stats', 'full_scores'], {'plot_wei': '(True)'}), '(hyperpars, tr_stats, full_scores, plot_wei=True\n )\n', (57088, 57142), False, 'from plot_fcts import logreg_plots\n'), ((59832, 59855), 'copy.deepcopy', 'copy.deepcopy', (['extrapar'], {}), '(extrapar)\n', (59845, 59855), False, 'import copy\n'), ((59912, 59934), 'copy.deepcopy', 'copy.deepcopy', (['techpar'], {}), '(techpar)\n', (59925, 59934), False, 'import copy\n'), ((59964, 60205), 'sess_util.sess_gen_util.get_analysdir', 'sess_gen_util.get_analysdir', (['sesspar.mouse_n', 'sesspar.sess_n', 'sesspar.plane', 'analyspar.fluor', 'analyspar.scale', 'stimpar.stimtype', 'stimpar.bri_dir', 'stimpar.bri_size', 'stimpar.gabk', 'logregpar.comp', 'logregpar.ctrl', "extrapar['shuffle']"], {}), "(sesspar.mouse_n, sesspar.sess_n, sesspar.plane,\n analyspar.fluor, analyspar.scale, stimpar.stimtype, stimpar.bri_dir,\n stimpar.bri_size, stimpar.gabk, logregpar.comp, logregpar.ctrl,\n extrapar['shuffle'])\n", (59991, 60205), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((66214, 66235), 'os.path.isdir', 'os.path.isdir', (['output'], {}), '(output)\n', (66227, 66235), False, 'import os\n'), ((67941, 67963), 'pandas.concat', 'pd.concat', (['scores_list'], {}), '(scores_list)\n', (67950, 67963), True, 'import pandas as pd\n'), ((68049, 68081), 'pandas.DataFrame', 'pd.DataFrame', ([], {'columns': 'all_labels'}), '(columns=all_labels)\n', (68061, 68081), True, 'import pandas as pd\n'), ((70277, 70330), 'util.logreg_util.get_sc_labs', 'logreg_util.get_sc_labs', (['(True)'], {'ext_test_name': 'ext_test'}), '(True, ext_test_name=ext_test)\n', (70300, 70330), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((2341, 2415), 'util.gen_util.accepted_values_error', 'gen_util.accepted_values_error', (['"""stimtype"""', 'stimtype', "['gabors', 'bricks']"], {}), "('stimtype', stimtype, ['gabors', 'bricks'])\n", (2371, 2415), False, 'from util import gen_util\n'), ((6537, 6629), 'sess_util.sess_ntuple_util.init_stimpar', 'sess_ntuple_util.init_stimpar', (['stimtype', 'bri_dir', 'bri_size', 'gabfr', 'gabk', 'gab_ori', '(0)', '(1.5)'], {}), '(stimtype, bri_dir, bri_size, gabfr, gabk,\n gab_ori, 0, 1.5)\n', (6566, 6629), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((8277, 8375), 'sess_util.sess_ntuple_util.init_stimpar', 'sess_ntuple_util.init_stimpar', (['stimtype', 'bri_dir', 'bri_size', 'gabfr', 'gabk', 'gab_ori', 'bri_pre', '(1.0)'], {}), '(stimtype, bri_dir, bri_size, gabfr, gabk,\n gab_ori, bri_pre, 1.0)\n', (8306, 8375), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((13214, 13280), 'util.gen_util.accepted_values_error', 'gen_util.accepted_values_error', (['"""alg"""', 'alg', "['pytorch', 'sklearn']"], {}), "('alg', alg, ['pytorch', 'sklearn'])\n", (13244, 13280), False, 'from util import gen_util\n'), ((15403, 15440), 'util.gen_util.list_if_not', 'gen_util.list_if_not', (['stimpar.bri_dir'], {}), '(stimpar.bri_dir)\n', (15423, 15440), False, 'from util import gen_util\n'), ((21240, 21290), 'util.gen_util.get_alternating_consec', 'gen_util.get_alternating_consec', (['segs'], {'first': '(False)'}), '(segs, first=False)\n', (21271, 21290), False, 'from util import gen_util\n'), ((21865, 21905), 'numpy.median', 'np.median', (['roi_data[:, :, :mid]'], {'axis': '(-1)'}), '(roi_data[:, :, :mid], axis=-1)\n', (21874, 21905), True, 'import numpy as np\n'), ((27570, 27639), 'sess_util.sess_ntuple_util.init_quintpar', 'sess_ntuple_util.init_quintpar', (['(1)', '[0, 0]', '[None, None]', '[None, None]'], {}), '(1, [0, 0], [None, None], [None, None])\n', (27600, 27639), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((29688, 29720), 'numpy.concatenate', 'np.concatenate', (['roi_seqs'], {'axis': '(0)'}), '(roi_seqs, axis=0)\n', (29702, 29720), True, 'import numpy as np\n'), ((30203, 30260), 'sess_util.sess_str_util.gabfr_letters', 'sess_str_util.gabfr_letters', (['stimpar.gabfr'], {'surp': 'surp_use'}), '(stimpar.gabfr, surp=surp_use)\n', (30230, 30260), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((35263, 35328), 'util.logreg_util.get_transf_data_sk', 'logreg_util.get_transf_data_sk', (['mod', 'data', '(False)', "(name == 'train')"], {}), "(mod, data, False, name == 'train')\n", (35293, 35328), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((39386, 39415), 'numpy.random.shuffle', 'np.random.shuffle', (['class_vals'], {}), '(class_vals)\n', (39403, 39415), True, 'import numpy as np\n'), ((39582, 39645), 'util.data_util.init_dl', 'data_util.init_dl', (['roi_seqs[1]', 'class_vals', 'logregpar.batchsize'], {}), '(roi_seqs[1], class_vals, logregpar.batchsize)\n', (39599, 39645), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((39900, 39932), 'util.logreg_util.LogReg', 'logreg_util.LogReg', (['n_rois', 'n_fr'], {}), '(n_rois, n_fr)\n', (39918, 39932), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((55013, 55063), 'util.data_util.bal_classes', 'data_util.bal_classes', (['roi_seqs[i]', 'seq_classes[i]'], {}), '(roi_seqs[i], seq_classes[i])\n', (55034, 55063), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((66835, 66901), 'util.gen_util.accepted_values_error', 'gen_util.accepted_values_error', (['"""alg"""', 'alg', "['sklearn', 'pytorch']"], {}), "('alg', alg, ['sklearn', 'pytorch'])\n", (66865, 66901), False, 'from util import gen_util\n'), ((67229, 67267), 'util.file_util.getfiles', 'file_util.getfiles', (['gen_dir', '"""subdirs"""'], {}), "(gen_dir, 'subdirs')\n", (67247, 67267), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((67672, 67725), 'util.logreg_util.get_sc_labs', 'logreg_util.get_sc_labs', (['(True)'], {'ext_test_name': 'ext_test'}), '(True, ext_test_name=ext_test)\n', (67695, 67725), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((72577, 72632), 'util.gen_util.set_df_vals', 'gen_util.set_df_vals', (['scores_summ', 'curr_idx', 'cols', 'vals'], {}), '(scores_summ, curr_idx, cols, vals)\n', (72597, 72632), False, 'from util import gen_util\n'), ((75235, 75287), 'util.gen_util.get_df_vals', 'gen_util.get_df_vals', (['all_scores_df', 'comm_labs', 'vals'], {}), '(all_scores_df, comm_labs, vals)\n', (75255, 75287), False, 'from util import gen_util\n'), ((75398, 75458), 'util.gen_util.set_df_vals', 'gen_util.set_df_vals', (['scores_summ', 'curr_idx', 'comm_labs', 'vals'], {}), '(scores_summ, curr_idx, comm_labs, vals)\n', (75418, 75458), False, 'from util import gen_util\n'), ((2068, 2170), 'util.gen_util.remove_if', 'gen_util.remove_if', (['comps', "['surp', 'dir_surp', 'dir_all', 'half_right', 'half_left', 'half_diff']"], {}), "(comps, ['surp', 'dir_surp', 'dir_all', 'half_right',\n 'half_left', 'half_diff'])\n", (2086, 2170), False, 'from util import gen_util\n'), ((2238, 2305), 'util.gen_util.remove_if', 'gen_util.remove_if', (['comps', "['half_left', 'half_right', 'half_diff']"], {}), "(comps, ['half_left', 'half_right', 'half_diff'])\n", (2256, 2305), False, 'from util import gen_util\n'), ((6694, 6727), 'sess_util.sess_str_util.gabfr_nbrs', 'sess_str_util.gabfr_nbrs', (['comp[0]'], {}), '(comp[0])\n', (6718, 6727), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((6750, 6843), 'sess_util.sess_ntuple_util.init_stimpar', 'sess_ntuple_util.init_stimpar', (['stimtype', 'bri_dir', 'bri_size', 'gabfr', 'gabk', 'gab_ori', '(0)', '(0.45)'], {}), '(stimtype, bri_dir, bri_size, gabfr, gabk,\n gab_ori, 0, 0.45)\n', (6779, 6843), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((8565, 8655), 'sess_util.sess_ntuple_util.init_stimpar', 'sess_ntuple_util.init_stimpar', (['stimtype', 'bri_dir', 'bri_size', 'gabfr', 'gabk', 'gab_ori', '(2)', '(2)'], {}), '(stimtype, bri_dir, bri_size, gabfr, gabk,\n gab_ori, 2, 2)\n', (8594, 8655), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((9672, 9738), 'util.gen_util.accepted_values_error', 'gen_util.accepted_values_error', (['"""alg"""', 'alg', "['sklearn', 'pytorch']"], {}), "('alg', alg, ['sklearn', 'pytorch'])\n", (9702, 9738), False, 'from util import gen_util\n'), ((21004, 21102), 'analysis.quint_analys.quint_segs', 'quint_analys.quint_segs', (['stim', 'stimpar', 'quintpar.n_quints', 'qu_i', 'surp_use'], {'remconsec': 'remconsec'}), '(stim, stimpar, quintpar.n_quints, qu_i, surp_use,\n remconsec=remconsec)\n', (21027, 21102), False, 'from analysis import quint_analys\n'), ((21941, 21964), 'numpy.expand_dims', 'np.expand_dims', (['div', '(-1)'], {}), '(div, -1)\n', (21955, 21964), True, 'import numpy as np\n'), ((26646, 26705), 'sess_util.sess_ntuple_util.init_quintpar', 'sess_ntuple_util.init_quintpar', (['(4)', '[[1, 2]]', '[None]', '[None]'], {}), '(4, [[1, 2]], [None], [None])\n', (26676, 26705), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((26923, 26982), 'sess_util.sess_ntuple_util.init_quintpar', 'sess_ntuple_util.init_quintpar', (['(2)', '[[0, 1]]', '[None]', '[None]'], {}), '(2, [[0, 1]], [None], [None])\n', (26953, 26982), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((27743, 27812), 'sess_util.sess_ntuple_util.init_quintpar', 'sess_ntuple_util.init_quintpar', (['(1)', '[0, 0]', '[None, None]', '[None, None]'], {}), '(1, [0, 0], [None, None], [None, None])\n', (27773, 27812), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((29996, 30045), 'sess_util.sess_str_util.gabfr_letters', 'sess_str_util.gabfr_letters', (['gabfr'], {'surp': 'surp_use'}), '(gabfr, surp=surp_use)\n', (30023, 30045), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((31740, 31762), 'numpy.unique', 'np.unique', (['seq_classes'], {}), '(seq_classes)\n', (31749, 31762), True, 'import numpy as np\n'), ((35465, 35531), 'util.gen_util.accepted_values_error', 'gen_util.accepted_values_error', (['"""alg"""', 'alg', "['sklearn', 'pytorch']"], {}), "('alg', alg, ['sklearn', 'pytorch'])\n", (35495, 35531), False, 'from util import gen_util\n'), ((38600, 38625), 'numpy.unique', 'np.unique', (['seq_classes[0]'], {}), '(seq_classes[0])\n', (38609, 38625), True, 'import numpy as np\n'), ((60906, 60996), 'util.gen_util.accepted_values_error', 'gen_util.accepted_values_error', (['"""logregpar.alg"""', 'logregpar.alg', "['pytorch', 'sklearn']"], {}), "('logregpar.alg', logregpar.alg, ['pytorch',\n 'sklearn'])\n", (60936, 60996), False, 'from util import gen_util\n'), ((70977, 71043), 'util.gen_util.accepted_values_error', 'gen_util.accepted_values_error', (['"""stats"""', 'stats', "['mean', 'median']"], {}), "('stats', stats, ['mean', 'median'])\n", (71007, 71043), False, 'from util import gen_util\n'), ((71964, 72034), 'util.math_util.error_stat', 'math_util.error_stat', (['data'], {'stats': '"""median"""', 'error': '"""std"""', 'nanpol': '"""omit"""'}), "(data, stats='median', error='std', nanpol='omit')\n", (71984, 72034), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((72199, 72276), 'util.math_util.error_stat', 'math_util.error_stat', (['data'], {'stats': '"""median"""', 'error': '"""std"""', 'nanpol': '"""omit"""', 'qu': 'ps'}), "(data, stats='median', error='std', nanpol='omit', qu=ps)\n", (72219, 72276), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((7608, 7707), 'sess_util.sess_ntuple_util.init_stimpar', 'sess_ntuple_util.init_stimpar', (['stimtype', 'bri_dir', 'bri_size', 'act_gabfr', 'gabk', 'gab_ori', 'pre', 'post'], {}), '(stimtype, bri_dir, bri_size, act_gabfr, gabk,\n gab_ori, pre, post)\n', (7637, 7707), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((22211, 22239), 'numpy.nanmean', 'np.nanmean', (['roi_data'], {'axis': '(0)'}), '(roi_data, axis=0)\n', (22221, 22239), True, 'import numpy as np\n'), ((28989, 29043), 'sess_util.sess_ntuple_util.get_modif_ntuple', 'sess_ntuple_util.get_modif_ntuple', (['stimpar', 'keys', 'vals'], {}), '(stimpar, keys, vals)\n', (29022, 29043), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((39502, 39527), 'torch.Tensor', 'torch.Tensor', (['roi_seqs[1]'], {}), '(roi_seqs[1])\n', (39514, 39527), False, 'import torch\n'), ((67391, 67410), 'os.listdir', 'os.listdir', (['run_dir'], {}), '(run_dir)\n', (67401, 67410), False, 'import os\n'), ((72388, 72458), 'util.math_util.error_stat', 'math_util.error_stat', (['data'], {'stats': '"""median"""', 'error': '"""sem"""', 'nanpol': '"""omit"""'}), "(data, stats='median', error='sem', nanpol='omit')\n", (72408, 72458), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((7534, 7585), 'sess_util.sess_gen_util.gab_oris_shared_U', 'sess_gen_util.gab_oris_shared_U', (['gab_letts', 'gab_ori'], {}), '(gab_letts, gab_ori)\n', (7565, 7585), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((7902, 7946), 'sess_util.sess_str_util.gabfr_nbrs', 'sess_str_util.gabfr_nbrs', (['[comp[0], comp[2]]'], {}), '([comp[0], comp[2]])\n', (7926, 7946), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((7969, 8063), 'sess_util.sess_ntuple_util.init_stimpar', 'sess_ntuple_util.init_stimpar', (['stimtype', 'bri_dir', 'bri_size', 'gabfrs', 'gabk', 'gab_ori', '(0)', '(0.45)'], {}), '(stimtype, bri_dir, bri_size, gabfrs, gabk,\n gab_ori, 0, 0.45)\n', (7998, 8063), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((19248, 19357), 'util.gen_util.accepted_values_error', 'gen_util.accepted_values_error', (['"""comp"""', 'comp', "['surp', 'AvB', 'AvC', 'BvC', 'DvU', 'dir...', '...ori...']"], {}), "('comp', comp, ['surp', 'AvB', 'AvC', 'BvC',\n 'DvU', 'dir...', '...ori...'])\n", (19278, 19357), False, 'from util import gen_util\n'), ((26742, 26764), 'numpy.unique', 'np.unique', (['stim.direcs'], {}), '(stim.direcs)\n', (26751, 26764), True, 'import numpy as np\n'), ((71605, 71656), 'util.math_util.mean_med', 'math_util.mean_med', (['data'], {'stats': 'stat', 'nanpol': '"""omit"""'}), "(data, stats=stat, nanpol='omit')\n", (71623, 71656), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((71766, 71834), 'util.math_util.error_stat', 'math_util.error_stat', (['data'], {'stats': '"""mean"""', 'error': 'error', 'nanpol': '"""omit"""'}), "(data, stats='mean', error=error, nanpol='omit')\n", (71786, 71834), False, 'from util import data_util, file_util, gen_util, logger_util, logreg_util, math_util, plot_util\n'), ((74903, 74955), 'util.gen_util.remove_if', 'gen_util.remove_if', (['comm_labs', "['mouse_n', 'n_rois']"], {}), "(comm_labs, ['mouse_n', 'n_rois'])\n", (74921, 74955), False, 'from util import gen_util\n'), ((7013, 7043), 'sess_util.sess_str_util.gabfr_nbrs', 'sess_str_util.gabfr_nbrs', (['lett'], {}), '(lett)\n', (7037, 7043), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n'), ((22495, 22523), 'numpy.nanmean', 'np.nanmean', (['roi_data'], {'axis': '(0)'}), '(roi_data, axis=0)\n', (22505, 22523), True, 'import numpy as np\n'), ((22545, 22572), 'numpy.nanstd', 'np.nanstd', (['roi_data'], {'axis': '(0)'}), '(roi_data, axis=0)\n', (22554, 22572), True, 'import numpy as np\n'), ((18974, 19024), 'sess_util.sess_str_util.dir_par_str', 'sess_str_util.dir_par_str', (['direc'], {'str_type': '"""print"""'}), "(direc, str_type='print')\n", (18999, 19024), False, 'from sess_util import sess_gen_util, sess_ntuple_util, sess_str_util\n')]
#!/usr/bin/env python2 # -*- coding: utf-8 -*- """ Created on March 10 2020 @author: <NAME> Translated from C++ code by <NAME> --- Camera models for pixel-to-vector transformation """ import os import numpy as np import xmltodict import math class OcamCalibCameraModel: ''' OcamCalib camera model for fisheye cameras ''' def __init__(self,calib_file): param_dict = self.read_opencv_storage_from_file(calib_file) self.set_params(param_dict) return None def set_params(self,opencv_storage_dict): #Etract parameters from opencv storage dictionary object d = opencv_storage_dict self.fx = np.zeros(5) self.fx[0] = float(d['ss4']) self.fx[1] = float(d['ss3']) self.fx[2] = float(d['ss2']) self.fx[3] = float(d['ss1']) self.fx[4] = float(d['ss0']) self.M = np.zeros((2,2)) self.M[0,0] = float(d['c']) self.M[0,1] = float(d['d']) self.M[1,0] = float(d['e']) self.M[1,1] = 1.0 self.xc = float(d['xc']) self.yc = float(d['yc']) self.width = int(d['width']) self.height = int(d['height']) self.image_circle_FOV = float(d['imageCircleFOV']) #Derived parameters self.dfdx = np.zeros(4) self.dfdx[0] = 4 * self.fx[0] self.dfdx[1] = 3 * self.fx[1] self.dfdx[2] = 2 * self.fx[2] self.dfdx[3] = self.fx[3] self.inv_M = np.linalg.inv(self.M) self.initial_x = self.width/4 #starting point for polynomial solver def read_opencv_storage_from_file(self,calib_file): print(calib_file) with open(calib_file) as fd: dict_ = xmltodict.parse(fd.read()) model_dict = dict_['opencv_storage']['cam_model'] return model_dict def vector_to_pixel(self,point): ''' Go from vector (in camera coordinates) to pixel (image coordinates) input: point - x,y,z in camera frame ''' forward = np.array([0,0,1]) r = vec3_normalise(point) alpha = np.arccos(np.dot(r,forward)) R = self.alpha_to_R(alpha) if R <0 : x = -1.0 y = -1.0 return False #Scale to get ideal fisheye pixel coordinates: mag = np.sqrt(r[0]**2 + r[1]**2) if (mag != 0): mag = R / mag # NOTE: model (x,y) is (height,width) so we swap px = r[1] * mag py = r[0] * mag # Account for non ideal fisheye effects (shear and translation): y = self.M[0,0]*px + self.M[0,1]*py + self.xc x = self.M[1,0]*px + self.M[1,1]*py + self.yc x = np.round(x).astype(int) y = np.round(y).astype(int) return x, y, R**2 def pixel_to_vector(self,x,y): #NOTE: model (x,y) is (height,width) so we swap dx = y - self.xc dy = x - self.yc; px = self.inv_M[0,0]*dx +self.inv_M[0,1]*dy; py = self.inv_M[1,0]*dx + self.inv_M[1,1]*dy; R2 = px*px + py*py; direction = np.array([0,0,0]) direction[0] = py; direction[1] = px; direction[2] = -eval_poly4(self.fx, np.sqrt(R2)); return direction def alpha_to_R(self,alpha): ''' Solves polynomial to go from alpha (angle between rays) to R (distance from center point on sensor) ''' #Newton-Raphson search for the solution newFx3 = self.fx[3] - np.tan(alpha - np.pi/2); fx = np.array([self.fx[0], self.fx[1], self.fx[2], newFx3, self.fx[4]]) dfdx = np.array([self.dfdx[0], self.dfdx[1], self.dfdx[2], newFx3]) x = self.initial_x; while True: px = x x -= eval_poly4(fx,x) / eval_poly3(dfdx,x) if (abs(x - px) > 1e-3): break R = x return R class RectilinearCameraModel: ''' Rectilinear camera model(compatible with Realsense) ''' def __init__(self,calib_file): ''' Load xml file with intrinsic calibration parameters ''' param_dict = self.read_opencv_storage_from_file(calib_file) self.set_params(param_dict) #Derived parameters self.focalLengthPixels = (self.height * 0.5) / math.tan(self.verticalFOV * 0.5) R = self.focalLengthPixels * math.tan(self.imageCircleFOV * 0.5) if (self.imageCircleFOV <= 0): R = self.width + self.height; # allows everything self.imageCircleR2 = R * R def set_params(self,opencv_storage_dict): ''' Extract parameters from opencv storage dictionary object ''' d = opencv_storage_dict self.xc = float(d['centreX']) self.yc = float(d['centreY']) self.imageCircleFOV = float(d['imageCircleFOV']) self.verticalFOV = float(d['verticalFOV']) self.width = int(d['width']) self.height = int(d['height']) def read_opencv_storage_from_file(self,calib_file): with open(calib_file) as fd: dict_ = xmltodict.parse(fd.read()) model_dict = dict_['opencv_storage']['cam_model'] return model_dict def vector_to_pixel(self, point): ''' Go from vector (in camera coordinates) to pixel (image coordinates) input: point (list) - x,y,z in camera frame ''' s = self.focalLengthPixels / point[2] dx = point[0] * s dy = point[1] * s x = dx + self.xc y = dy + self.yc x = np.round(x).astype(int) y = np.round(y).astype(int) R_squared = dx**2 + dy**2 return x, y, R_squared def pixel_to_vector(self, x, y): ''' Go from pixel in image coordinates to vector in camera coordinates ''' #NOTE: model (x,y) is (height,width) so we swap dx = x - self.xc dy = y - self.yc direction = np.array([0,0,0]) direction[0] = dx direction[1] = dy direction[2] = self.focalLengthPixels return direction # Utility functions def vec3_normalise(point): norm = np.linalg.norm(point) if norm == 0: return point return point / norm #Evaluation of polynomials #cubic def eval_poly3(poly,x): return ((poly[0]*x + poly[1])*x + poly[2])*x + poly[3] #quartic def eval_poly4(poly, x): return (((poly[0]*x + poly[1])*x + poly[2])*x + poly[3])*x + poly[4]
[ "numpy.sqrt", "numpy.tan", "math.tan", "numpy.array", "numpy.zeros", "numpy.linalg.inv", "numpy.dot", "numpy.linalg.norm", "numpy.round" ]
[((6173, 6194), 'numpy.linalg.norm', 'np.linalg.norm', (['point'], {}), '(point)\n', (6187, 6194), True, 'import numpy as np\n'), ((663, 674), 'numpy.zeros', 'np.zeros', (['(5)'], {}), '(5)\n', (671, 674), True, 'import numpy as np\n'), ((877, 893), 'numpy.zeros', 'np.zeros', (['(2, 2)'], {}), '((2, 2))\n', (885, 893), True, 'import numpy as np\n'), ((1285, 1296), 'numpy.zeros', 'np.zeros', (['(4)'], {}), '(4)\n', (1293, 1296), True, 'import numpy as np\n'), ((1466, 1487), 'numpy.linalg.inv', 'np.linalg.inv', (['self.M'], {}), '(self.M)\n', (1479, 1487), True, 'import numpy as np\n'), ((2045, 2064), 'numpy.array', 'np.array', (['[0, 0, 1]'], {}), '([0, 0, 1])\n', (2053, 2064), True, 'import numpy as np\n'), ((2331, 2361), 'numpy.sqrt', 'np.sqrt', (['(r[0] ** 2 + r[1] ** 2)'], {}), '(r[0] ** 2 + r[1] ** 2)\n', (2338, 2361), True, 'import numpy as np\n'), ((3103, 3122), 'numpy.array', 'np.array', (['[0, 0, 0]'], {}), '([0, 0, 0])\n', (3111, 3122), True, 'import numpy as np\n'), ((3545, 3611), 'numpy.array', 'np.array', (['[self.fx[0], self.fx[1], self.fx[2], newFx3, self.fx[4]]'], {}), '([self.fx[0], self.fx[1], self.fx[2], newFx3, self.fx[4]])\n', (3553, 3611), True, 'import numpy as np\n'), ((3627, 3687), 'numpy.array', 'np.array', (['[self.dfdx[0], self.dfdx[1], self.dfdx[2], newFx3]'], {}), '([self.dfdx[0], self.dfdx[1], self.dfdx[2], newFx3])\n', (3635, 3687), True, 'import numpy as np\n'), ((5960, 5979), 'numpy.array', 'np.array', (['[0, 0, 0]'], {}), '([0, 0, 0])\n', (5968, 5979), True, 'import numpy as np\n'), ((2123, 2141), 'numpy.dot', 'np.dot', (['r', 'forward'], {}), '(r, forward)\n', (2129, 2141), True, 'import numpy as np\n'), ((3507, 3532), 'numpy.tan', 'np.tan', (['(alpha - np.pi / 2)'], {}), '(alpha - np.pi / 2)\n', (3513, 3532), True, 'import numpy as np\n'), ((4319, 4351), 'math.tan', 'math.tan', (['(self.verticalFOV * 0.5)'], {}), '(self.verticalFOV * 0.5)\n', (4327, 4351), False, 'import math\n'), ((4389, 4424), 'math.tan', 'math.tan', (['(self.imageCircleFOV * 0.5)'], {}), '(self.imageCircleFOV * 0.5)\n', (4397, 4424), False, 'import math\n'), ((2714, 2725), 'numpy.round', 'np.round', (['x'], {}), '(x)\n', (2722, 2725), True, 'import numpy as np\n'), ((2750, 2761), 'numpy.round', 'np.round', (['y'], {}), '(y)\n', (2758, 2761), True, 'import numpy as np\n'), ((3219, 3230), 'numpy.sqrt', 'np.sqrt', (['R2'], {}), '(R2)\n', (3226, 3230), True, 'import numpy as np\n'), ((5569, 5580), 'numpy.round', 'np.round', (['x'], {}), '(x)\n', (5577, 5580), True, 'import numpy as np\n'), ((5605, 5616), 'numpy.round', 'np.round', (['y'], {}), '(y)\n', (5613, 5616), True, 'import numpy as np\n')]
import numpy as np from sklearn.datasets import load_iris from sklearn.metrics import accuracy_score from deepforest import CascadeForestClassifier from sklearn.model_selection import train_test_split X, y = load_iris(return_X_y=True) X = np.array(X) y = np.array(y) X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2) model = CascadeForestClassifier() model.fit(X_train, y_train) y_pred = model.predict(X_test) acc = accuracy_score(y_test, y_pred) * 100 print("Accuracy: {:.3f} %".format(acc))
[ "sklearn.datasets.load_iris", "sklearn.model_selection.train_test_split", "deepforest.CascadeForestClassifier", "numpy.array", "sklearn.metrics.accuracy_score" ]
[((209, 235), 'sklearn.datasets.load_iris', 'load_iris', ([], {'return_X_y': '(True)'}), '(return_X_y=True)\n', (218, 235), False, 'from sklearn.datasets import load_iris\n'), ((241, 252), 'numpy.array', 'np.array', (['X'], {}), '(X)\n', (249, 252), True, 'import numpy as np\n'), ((257, 268), 'numpy.array', 'np.array', (['y'], {}), '(y)\n', (265, 268), True, 'import numpy as np\n'), ((305, 342), 'sklearn.model_selection.train_test_split', 'train_test_split', (['X', 'y'], {'test_size': '(0.2)'}), '(X, y, test_size=0.2)\n', (321, 342), False, 'from sklearn.model_selection import train_test_split\n'), ((352, 377), 'deepforest.CascadeForestClassifier', 'CascadeForestClassifier', ([], {}), '()\n', (375, 377), False, 'from deepforest import CascadeForestClassifier\n'), ((447, 477), 'sklearn.metrics.accuracy_score', 'accuracy_score', (['y_test', 'y_pred'], {}), '(y_test, y_pred)\n', (461, 477), False, 'from sklearn.metrics import accuracy_score\n')]
import numpy as np import cv2 from utils import FacialRecoLogin # Time before locking/unlocking (in seconds) time_before_lock = 3 if __name__ == '__main__': # Facial Reco object reco_login = FacialRecoLogin(time_before_lock=time_before_lock) # Grasp webcam video_capture = cv2.VideoCapture(0) while True: # Grab a single frame of video ret, frame = video_capture.read() # Resize frame of video to 1/4 size for faster face recognition processing small_frame = cv2.resize(frame, (0, 0), fx=0.25, fy=0.25) # Convert the image from BGR color (which OpenCV uses) to RGB color (which face_recognition uses) rgb_small_frame = small_frame[:, :, ::-1] # List of names and face locations face_names, face_locations = reco_login.frame_match(rgb_small_frame) # Looping on the face detected on the frame for (top, right, bottom, left), name in zip(face_locations, face_names): # Scale back up face locations since the frame we detected in was scaled to 1/4 size top *= 4 right *= 4 bottom *= 4 left *= 4 # If user recognized if name in reco_login.known_face_names: cv2.rectangle(frame, (left, top), (right, bottom), (0, 255, 0), 2) # center1 = int((left+right)/2) # center2 = int((top+bottom)/2) # cv2.circle(frame, (center1, center2), int( # right - (left+right)/2), (0, 255, 0), 7) font = cv2.FONT_HERSHEY_DUPLEX cv2.putText(frame, name, (left, bottom + 30), font, 1.0, (0, 255, 0), 1) cv2.putText(frame, "Authorized... " + str(np.round(reco_login.time_recognized, 1)), (left, bottom + 60), font, 1.0, (0, 255, 0), 1) # If unknown else: cv2.rectangle(frame, (left, top), (right, bottom), (0, 0, 255), 2) font = cv2.FONT_HERSHEY_DUPLEX cv2.putText(frame, name, (left + 60, bottom - 40), font, 1.0, (0, 0, 255), 1) cv2.putText(frame, "No access... " + str(np.round(reco_login.time_unknown, 1)), (left + 60, bottom + 6), font, 1.0, (0, 0, 255), 1) # Check for lock/unlock timing if reco_login.time_recognized <= 0: reco_login.unlock(video_capture, speaker=True) elif reco_login.time_unknown <= 0: reco_login.lock(video_capture, speaker=True) # Show webcam output cv2.imshow('Video', frame) if (cv2.waitKey(1) & 0xFF == ord('q')): break
[ "cv2.rectangle", "utils.FacialRecoLogin", "cv2.imshow", "cv2.putText", "cv2.VideoCapture", "cv2.resize", "cv2.waitKey", "numpy.round" ]
[((203, 253), 'utils.FacialRecoLogin', 'FacialRecoLogin', ([], {'time_before_lock': 'time_before_lock'}), '(time_before_lock=time_before_lock)\n', (218, 253), False, 'from utils import FacialRecoLogin\n'), ((294, 313), 'cv2.VideoCapture', 'cv2.VideoCapture', (['(0)'], {}), '(0)\n', (310, 313), False, 'import cv2\n'), ((518, 561), 'cv2.resize', 'cv2.resize', (['frame', '(0, 0)'], {'fx': '(0.25)', 'fy': '(0.25)'}), '(frame, (0, 0), fx=0.25, fy=0.25)\n', (528, 561), False, 'import cv2\n'), ((2699, 2725), 'cv2.imshow', 'cv2.imshow', (['"""Video"""', 'frame'], {}), "('Video', frame)\n", (2709, 2725), False, 'import cv2\n'), ((1263, 1329), 'cv2.rectangle', 'cv2.rectangle', (['frame', '(left, top)', '(right, bottom)', '(0, 255, 0)', '(2)'], {}), '(frame, (left, top), (right, bottom), (0, 255, 0), 2)\n', (1276, 1329), False, 'import cv2\n'), ((1641, 1713), 'cv2.putText', 'cv2.putText', (['frame', 'name', '(left, bottom + 30)', 'font', '(1.0)', '(0, 255, 0)', '(1)'], {}), '(frame, name, (left, bottom + 30), font, 1.0, (0, 255, 0), 1)\n', (1652, 1713), False, 'import cv2\n'), ((1977, 2043), 'cv2.rectangle', 'cv2.rectangle', (['frame', '(left, top)', '(right, bottom)', '(0, 0, 255)', '(2)'], {}), '(frame, (left, top), (right, bottom), (0, 0, 255), 2)\n', (1990, 2043), False, 'import cv2\n'), ((2135, 2212), 'cv2.putText', 'cv2.putText', (['frame', 'name', '(left + 60, bottom - 40)', 'font', '(1.0)', '(0, 0, 255)', '(1)'], {}), '(frame, name, (left + 60, bottom - 40), font, 1.0, (0, 0, 255), 1)\n', (2146, 2212), False, 'import cv2\n'), ((2738, 2752), 'cv2.waitKey', 'cv2.waitKey', (['(1)'], {}), '(1)\n', (2749, 2752), False, 'import cv2\n'), ((1800, 1839), 'numpy.round', 'np.round', (['reco_login.time_recognized', '(1)'], {}), '(reco_login.time_recognized, 1)\n', (1808, 1839), True, 'import numpy as np\n'), ((2298, 2334), 'numpy.round', 'np.round', (['reco_login.time_unknown', '(1)'], {}), '(reco_login.time_unknown, 1)\n', (2306, 2334), True, 'import numpy as np\n')]
import numpy as np import torch.utils.data as tud from sklearn.model_selection import train_test_split from noge.constants import REAL_DATASETS, PLACES, DATA_DIR from xlog.utils import load_pickle class GraphDataset(tud.Dataset): def __init__(self, graphs, mazes=None): self.graphs = graphs self.mazes = mazes self.num_nodes_per_graph = np.array([G.number_of_nodes() for G in graphs], dtype=int) self.num_edges_per_graph = np.array([G.number_of_edges() for G in graphs], dtype=int) n_graphs = len(graphs) self._pairs = [(g, s) for g in range(n_graphs) for s in graphs[g].nodes] graph_idx, sources = zip(*self._pairs) self.samples_graph_idx = np.array(graph_idx) self._samples_sources = np.array(sources) def __len__(self): return len(self._pairs) def __getitem__(self, item): graph_index, source = self._pairs[item] graph = self.graphs[graph_index] sample = dict(graph=graph, source=source) if self.mazes is not None: sample.update(maze=self.mazes[graph_index]) return sample @property def max_nodes(self): return max(self.num_nodes_per_graph) @property def max_edges(self): return max(self.num_edges_per_graph) @property def num_graphs(self): return len(self.graphs) class SubsetSampler(tud.Sampler): def __init__(self, dataset, seed, num_samples=50): assert num_samples <= len(dataset) self.dataset = dataset self.seed = seed self.rng = np.random.RandomState(seed=seed) # for evaluation only choose pairs once (to be consistent across epochs) n_graphs = len(dataset.graphs) if n_graphs >= num_samples: # sample one source node per graph num_nodes_per_graph = self.dataset.num_nodes_per_graph indices = [] offset = 0 for num_nodes in num_nodes_per_graph: # num_nodes = num_nodes_per_graph[g] index = self.rng.randint(num_nodes) indices.append(offset + index) offset += num_nodes if len(indices) == num_samples: break self._indices = indices else: # the number of graphs is less than the required num_samples n_total = len(dataset) if n_total <= num_samples: # if the total number of samples is less than or equal to required, use all samples self._indices = list(range(n_total)) else: # if the total number of samples is larger than required, sub-sample self._indices = self.rng.choice(n_total, size=num_samples, replace=False).tolist() self._indices.sort() def __iter__(self): return iter(self._indices) def __len__(self): return len(self._indices) def get_test_loader(dataset, seed, num_samples): sampler = SubsetSampler(dataset, seed=seed, num_samples=num_samples) # set batch_size = None to get each sample without a batch dimension # set collate_fn = identity to not trigger auto_collate which converts to torch types loader = tud.DataLoader(dataset=dataset, batch_size=None, collate_fn=lambda x: x, sampler=sampler) return loader class BalancedInfiniteRandomSampler: """ Sample each graph with equal probability (in the limit) """ def __init__(self, dataset, seed, cycle_size=100_000, replace=True): self.dataset = dataset self.seed = seed self.rng = np.random.RandomState(seed=seed) # each node's weight should be proportional to 1 over the graph size of the node inverse_graph_sizes = 1. / dataset.num_nodes_per_graph self.p = inverse_graph_sizes[dataset.samples_graph_idx] self.p = self.p / self.p.sum() # self.weights = torch.as_tensor(self.p, dtype=torch.double) self.cycle_size = cycle_size self.replacement = replace def __iter__(self): while True: # sample once every `cycle_size` (rng.choice is slow) indices = self.rng.choice(len(self.dataset), self.cycle_size, self.replacement, p=self.p).tolist() # items = torch.multinomial(self.weights, self.cycle_size, self.replacement).tolist() for index in indices: yield index def get_train_generator(dataset, seed): sampler = BalancedInfiniteRandomSampler(dataset, seed) sampler_iter = iter(sampler) while True: index = next(sampler_iter) sample = dataset[index] yield sample def _get_real_graph(dataset): proc_dir = DATA_DIR / 'osm' / 'processed' # get place and name place = PLACES[dataset] name = place['city'].replace(' ', '') path_train = proc_dir / f"{name}_train.pkl" path_test = proc_dir / f"{name}_test.pkl" train_graph = load_pickle(path_train) test_graph = load_pickle(path_test) train_set = GraphDataset([train_graph]) test_set = GraphDataset([test_graph]) return train_set, test_set def get_datasets(dataset, seed, test_size, val_size=0): if dataset in REAL_DATASETS: return _get_real_graph(dataset) path_graphs = DATA_DIR / f"{dataset}.pkl" graphs = load_pickle(path_graphs) graphs_train, graphs_test = train_test_split(graphs, test_size=test_size, random_state=seed) graphs_val = None if val_size > 0: graphs_train, graphs_val = train_test_split(graphs_train, test_size=val_size, random_state=seed) mazes_train = None mazes_test = None mazes_val = None if dataset in ('maze', 'hypermaze'): graphs_train, mazes_train = zip(*graphs_train) graphs_test, mazes_test = zip(*graphs_test) if graphs_val is not None: graphs_val, mazes_val = zip(*graphs_val) train_set = GraphDataset(graphs_train, mazes_train) test_set = GraphDataset(graphs_test, mazes_test) if graphs_val is not None: val_set = GraphDataset(graphs_val, mazes_val) return train_set, val_set, test_set return train_set, test_set
[ "sklearn.model_selection.train_test_split", "numpy.array", "torch.utils.data.DataLoader", "xlog.utils.load_pickle", "numpy.random.RandomState" ]
[((3265, 3358), 'torch.utils.data.DataLoader', 'tud.DataLoader', ([], {'dataset': 'dataset', 'batch_size': 'None', 'collate_fn': '(lambda x: x)', 'sampler': 'sampler'}), '(dataset=dataset, batch_size=None, collate_fn=lambda x: x,\n sampler=sampler)\n', (3279, 3358), True, 'import torch.utils.data as tud\n'), ((4997, 5020), 'xlog.utils.load_pickle', 'load_pickle', (['path_train'], {}), '(path_train)\n', (5008, 5020), False, 'from xlog.utils import load_pickle\n'), ((5038, 5060), 'xlog.utils.load_pickle', 'load_pickle', (['path_test'], {}), '(path_test)\n', (5049, 5060), False, 'from xlog.utils import load_pickle\n'), ((5371, 5395), 'xlog.utils.load_pickle', 'load_pickle', (['path_graphs'], {}), '(path_graphs)\n', (5382, 5395), False, 'from xlog.utils import load_pickle\n'), ((5428, 5492), 'sklearn.model_selection.train_test_split', 'train_test_split', (['graphs'], {'test_size': 'test_size', 'random_state': 'seed'}), '(graphs, test_size=test_size, random_state=seed)\n', (5444, 5492), False, 'from sklearn.model_selection import train_test_split\n'), ((717, 736), 'numpy.array', 'np.array', (['graph_idx'], {}), '(graph_idx)\n', (725, 736), True, 'import numpy as np\n'), ((769, 786), 'numpy.array', 'np.array', (['sources'], {}), '(sources)\n', (777, 786), True, 'import numpy as np\n'), ((1587, 1619), 'numpy.random.RandomState', 'np.random.RandomState', ([], {'seed': 'seed'}), '(seed=seed)\n', (1608, 1619), True, 'import numpy as np\n'), ((3657, 3689), 'numpy.random.RandomState', 'np.random.RandomState', ([], {'seed': 'seed'}), '(seed=seed)\n', (3678, 3689), True, 'import numpy as np\n'), ((5572, 5641), 'sklearn.model_selection.train_test_split', 'train_test_split', (['graphs_train'], {'test_size': 'val_size', 'random_state': 'seed'}), '(graphs_train, test_size=val_size, random_state=seed)\n', (5588, 5641), False, 'from sklearn.model_selection import train_test_split\n')]
# based on: https://stanford.edu/~shervine/blog/keras-how-to-generate-data-on-the-fly.html import os, sys import numpy as np import keras from . import helper as hp class OneHotEncode(): def __init__(self, max_len_model, n_chars, indices_token, token_indices, pad_char, start_char, end_char): 'Initialization' self.max_len_model = max_len_model self.n_chars = n_chars self.pad_char = pad_char self.start_char = start_char self.end_char = end_char self.indices_token = indices_token self.token_indices = token_indices def one_hot_encode(self, token_list, n_chars): output = np.zeros((token_list.shape[0], n_chars)) for j, token in enumerate(token_list): output[j, token] = 1 return output def smi_to_int(self, smi): """ this will turn a list of smiles in string format and turn them into a np array of int, with padding """ token_list = hp.smi_tokenizer(smi) token_list = [self.start_char] + token_list + [self.end_char] padding = [self.pad_char]*(self.max_len_model - len(token_list)) token_list.extend(padding) int_list = [self.token_indices[x] for x in token_list] return np.asarray(int_list) def int_to_smile(self, array): """ From an array of int, return a list of molecules in string smile format Note: remove the padding char """ all_smi = [] for seq in array: new_mol = [self.indices_token[int(x)] for x in seq] all_smi.append(''.join(new_mol).replace(self.pad_char, '')) return all_smi def clean_smile(self, smi): """ remove return line symbols """ smi = smi.replace('\n', '') return smi def smile_to_onehot(self, path_data): f = open(path_data) lines = f.readlines() n_data = len(lines) X = np.empty((n_data, self.max_len_model, self.n_chars), dtype=int) for i,smi in enumerate(lines): # remove return line symbols smi = self.clean_smile(smi) # tokenize int_smi = self.smi_to_int(smi) # one hot encode X[i] = self.one_hot_encode(int_smi, self.n_chars) return X def generator_smile_to_onehot(self, smi): smi = self.clean_smile(smi) int_smi = self.smi_to_int(smi) one_hot = self.one_hot_encode(int_smi, self.n_chars) return one_hot class DataGenerator(keras.utils.Sequence): 'Generates data for Keras' def __init__(self, list_IDs, batch_size, max_len_model, path_data, n_chars, indices_token, token_indices, pad_char, start_char, end_char, shuffle=True): 'Initialization' self.max_len_model = max_len_model self.batch_size = batch_size self.list_IDs = list_IDs self.shuffle = shuffle self.path_data = path_data self.n_chars = n_chars self.OneHotEncoder = OneHotEncode(max_len_model, n_chars, indices_token, token_indices, pad_char, start_char, end_char) self.on_epoch_end() f=open(self.path_data) self.lines=f.readlines() self.indices_token = indices_token self.token_indices = token_indices def __len__(self): 'Denotes the number of batches per epoch' return int(np.floor(len(self.list_IDs) / self.batch_size)) def __getitem__(self, index): 'Generate one batch of data' # Generate indexes of the batch indexes = self.indexes[index*self.batch_size:(index+1)*self.batch_size] # Find list of IDs list_IDs_temp = [self.list_IDs[k] for k in indexes] # Generate data X, y = self.__data_generation(list_IDs_temp) return X, y def on_epoch_end(self): 'Updates indexes after each epoch' self.indexes = np.arange(len(self.list_IDs)) if self.shuffle == True: np.random.shuffle(self.indexes) def __data_generation(self, list_IDs_temp): 'Generates batch of data containing batch_size samples' switch = 1 y = np.empty((self.batch_size, self.max_len_model-switch, self.n_chars), dtype=int) X = np.empty((self.batch_size, self.max_len_model-switch, self.n_chars), dtype=int) # Generate data for i, ID in enumerate(list_IDs_temp): smi = self.lines[ID] one_hot_smi = self.OneHotEncoder.generator_smile_to_onehot(smi) X[i] = one_hot_smi[:-1] y[i] = one_hot_smi[1:] return X, y
[ "numpy.zeros", "numpy.asarray", "numpy.empty", "numpy.random.shuffle" ]
[((667, 707), 'numpy.zeros', 'np.zeros', (['(token_list.shape[0], n_chars)'], {}), '((token_list.shape[0], n_chars))\n', (675, 707), True, 'import numpy as np\n'), ((1285, 1305), 'numpy.asarray', 'np.asarray', (['int_list'], {}), '(int_list)\n', (1295, 1305), True, 'import numpy as np\n'), ((1998, 2061), 'numpy.empty', 'np.empty', (['(n_data, self.max_len_model, self.n_chars)'], {'dtype': 'int'}), '((n_data, self.max_len_model, self.n_chars), dtype=int)\n', (2006, 2061), True, 'import numpy as np\n'), ((4403, 4488), 'numpy.empty', 'np.empty', (['(self.batch_size, self.max_len_model - switch, self.n_chars)'], {'dtype': 'int'}), '((self.batch_size, self.max_len_model - switch, self.n_chars),\n dtype=int)\n', (4411, 4488), True, 'import numpy as np\n'), ((4495, 4580), 'numpy.empty', 'np.empty', (['(self.batch_size, self.max_len_model - switch, self.n_chars)'], {'dtype': 'int'}), '((self.batch_size, self.max_len_model - switch, self.n_chars),\n dtype=int)\n', (4503, 4580), True, 'import numpy as np\n'), ((4200, 4231), 'numpy.random.shuffle', 'np.random.shuffle', (['self.indexes'], {}), '(self.indexes)\n', (4217, 4231), True, 'import numpy as np\n')]
import sys import numpy as np vert = np.loadtxt(sys.argv[1]) tri = np.loadtxt(sys.argv[2]) with open(sys.argv[3], 'w') as f: f.write('OFF\n') f.write('{} {} {}\n'.format(int(vert.shape[0]), int(tri.shape[0]), 0)) with open(sys.argv[3], 'ab') as f: np.savetxt(f, vert, fmt='%.6f') np.savetxt(f, np.hstack([np.ones((tri.shape[0],1))*3, tri]), fmt='%d')
[ "numpy.loadtxt", "numpy.ones", "numpy.savetxt" ]
[((38, 61), 'numpy.loadtxt', 'np.loadtxt', (['sys.argv[1]'], {}), '(sys.argv[1])\n', (48, 61), True, 'import numpy as np\n'), ((68, 91), 'numpy.loadtxt', 'np.loadtxt', (['sys.argv[2]'], {}), '(sys.argv[2])\n', (78, 91), True, 'import numpy as np\n'), ((265, 296), 'numpy.savetxt', 'np.savetxt', (['f', 'vert'], {'fmt': '"""%.6f"""'}), "(f, vert, fmt='%.6f')\n", (275, 296), True, 'import numpy as np\n'), ((326, 352), 'numpy.ones', 'np.ones', (['(tri.shape[0], 1)'], {}), '((tri.shape[0], 1))\n', (333, 352), True, 'import numpy as np\n')]
import pandas as pd from scipy.stats import ttest_ind import numpy as np from statsmodels.stats.multitest import multipletests as multit from warnings import warn def count_reps(inseries): inseries = inseries.tolist() counts = {k:0 for k in list(set(inseries))} out = [878 for i in range(len(inseries))] for ind, ite in enumerate(inseries): out[ind] = counts[ite] counts[ite] += 1 return out from scipy.stats import mstats_basic def interpret(ld, condition_column, strain_column, values_column, control_condition, out_prefix, circularity=None, set_missing_na=False): ''' Interpret experimental data report produced by pyphe-analyse. ''' ###Check if essential columns exist print('Checking input table') print('Checking if axis_column exists') if condition_column not in list(ld): raise NameError('Axis_column not found in table.') print('....OK') print('Checking if grouping_column exists') if strain_column not in list(ld): raise NameError('grouping_column not found in table.') print('....OK') print('Checking if values_column exists') if values_column not in list(ld): raise NameError('values_column not found in table.') print('....OK') print('Checking if control exists in axis_column') if control_condition not in ld[condition_column].unique(): raise NameError('control not found in axis_column.') print('....OK') if circularity: print('Circularity filter is set. Checking if Colony_circularity column exists') if 'Colony_circularity' not in list(ld): raise NameError('Input data has no column named Colony_circularity. Cannot apply circularity filter.') ###Report some simple numbers print('Data report loaded successfully') initial_conditions = ld[condition_column].unique() print('Number of unique elements in axis column: %i'%len(initial_conditions)) initial_strains = ld[strain_column].unique() print('Number of unique elements in grouping column: %i'%len(initial_strains)) print('Number of plates: %i'%len(ld['Plate'].unique())) print('Number of non-NA data points: %i'%len(ld.loc[~pd.isnull(ld[values_column])].index)) ###Simple QC filters n_datapoints = (~ld[values_column].isnull()).sum() if circularity: ld.loc[ld['Colony_circularity']<circularity, values_column] = np.nan nn_datapoints = (~ld[values_column].isnull()).sum() print('Removed %i entries with circularity < %f'%(n_datapoints-nn_datapoints, circularity)) n_datapoints = nn_datapoints if set_missing_na: ld.loc[ld[values_column]==0, values_column] = np.nan nn_datapoints = (~ld[values_column].isnull()).sum() print('Removed %i entries with fitness 0'%(n_datapoints-nn_datapoints)) n_datapoints = nn_datapoints ###Group by replicates ld_stats = ld.copy() #drop any NA ld_stats = ld_stats.loc[~ld_stats[values_column].isnull()] #Recompute number of axis and grouping elements conditions = ld_stats[condition_column].unique() print('Number of unique elements in axis column after filtering: %i'%len(conditions)) strains = ld_stats[strain_column].unique() print('Number of unique elements in grouping column: %i'%len(strains)) ld_stats['condition---strain'] = ld_stats[condition_column] + '---' + ld_stats[strain_column] ld_stats['rep'] = count_reps(ld_stats['condition---strain']) #Pivot this into wide format ld_stats_piv = ld_stats.pivot_table(index=strain_column, columns=[condition_column,'rep'], values=values_column) #assert that there are no duplicates, i.e. that count_reps() worked as expected assert (ld_stats.pivot_table(index=strain_column, columns=[condition_column,'rep'], values=values_column, aggfunc=len).unstack().dropna()==1.0).all() #Save this table: ld_stats_piv.to_csv(out_prefix+'_reps.csv') ###Compute summary stats mean_fitness = ld_stats_piv.mean(axis=1, level=0) median_fitness = ld_stats_piv.median(axis=1, level=0) fitness_stdev = ld_stats_piv.std(axis=1, level=0) obs_count = ld_stats_piv.count(axis=1, level=0) #Compute effect sizes median_effect_size = median_fitness.div(median_fitness[control_condition], axis=0) mean_effect_size = mean_fitness.div(mean_fitness[control_condition], axis=0) ###run Welch's t-test print('Running t-tests') p_Welch = {} b = ld_stats_piv.xs(control_condition,axis=1, level=0).values b = np.ma.masked_invalid(b) for co in conditions: a = ld_stats_piv.xs(co, axis=1, level=0).values a = np.ma.masked_invalid(a) pvals_temp = mstats_basic.ttest_ind(a, b, axis=1, equal_var=False)[1].filled(np.nan) p_Welch[co] = pd.Series(pvals_temp, index=ld_stats_piv.index) p_Welch = pd.concat(p_Welch, axis=1) #multiple testing correction by BH p_Welch_BH = p_Welch.copy() for c in p_Welch_BH: if p_Welch_BH[c].isnull().all(): warn('No p-values obtained for %s (probably not enaough replicates)'%c) else: p_Welch_BH.loc[~p_Welch_BH[c].isnull(), c] = multit(p_Welch_BH.loc[~p_Welch_BH[c].isnull(), c], method='fdr_bh')[1] #aggregate data in table and save #And join together in one big data frame combined_data = pd.concat({'mean_fitness' : mean_fitness, 'mean_fitness_log2' : mean_fitness.applymap(np.log2), 'median_fitness' : median_fitness, 'median_fitness_log2' : median_fitness.applymap(np.log2), 'mean_effect_size' : mean_effect_size, 'mean_effect_size_log2' : mean_effect_size.applymap(np.log2), 'median_effect_size' : median_effect_size, 'median_effect_size_log2' : median_effect_size.applymap(np.log2), 'observation_count' : obs_count, 'stdev_fitness' : fitness_stdev, 'p_Welch' : p_Welch, 'p_Welch_BH' : p_Welch_BH, 'p_Welch_BH_-log10' : -p_Welch_BH.applymap(np.log10)}, axis=1) combined_data = combined_data.swaplevel(axis=1).sort_index(axis=1) combined_data.to_csv(out_prefix+'_summaryStats.csv') print('Interpretation completed and results saved.') return combined_data
[ "pandas.Series", "pandas.isnull", "scipy.stats.mstats_basic.ttest_ind", "numpy.ma.masked_invalid", "warnings.warn", "pandas.concat" ]
[((4556, 4579), 'numpy.ma.masked_invalid', 'np.ma.masked_invalid', (['b'], {}), '(b)\n', (4576, 4579), True, 'import numpy as np\n'), ((4876, 4902), 'pandas.concat', 'pd.concat', (['p_Welch'], {'axis': '(1)'}), '(p_Welch, axis=1)\n', (4885, 4902), True, 'import pandas as pd\n'), ((4675, 4698), 'numpy.ma.masked_invalid', 'np.ma.masked_invalid', (['a'], {}), '(a)\n', (4695, 4698), True, 'import numpy as np\n'), ((4814, 4861), 'pandas.Series', 'pd.Series', (['pvals_temp'], {'index': 'ld_stats_piv.index'}), '(pvals_temp, index=ld_stats_piv.index)\n', (4823, 4861), True, 'import pandas as pd\n'), ((5052, 5125), 'warnings.warn', 'warn', (["('No p-values obtained for %s (probably not enaough replicates)' % c)"], {}), "('No p-values obtained for %s (probably not enaough replicates)' % c)\n", (5056, 5125), False, 'from warnings import warn\n'), ((4720, 4773), 'scipy.stats.mstats_basic.ttest_ind', 'mstats_basic.ttest_ind', (['a', 'b'], {'axis': '(1)', 'equal_var': '(False)'}), '(a, b, axis=1, equal_var=False)\n', (4742, 4773), False, 'from scipy.stats import mstats_basic\n'), ((2218, 2246), 'pandas.isnull', 'pd.isnull', (['ld[values_column]'], {}), '(ld[values_column])\n', (2227, 2246), True, 'import pandas as pd\n')]
#==============================================================# # This script classifies an instance of a connection # #==============================================================# import sys import math import pickle import numpy as np import pandas as pd from sklearn import svm from sklearn import preprocessing from sklearn.covariance import EllipticEnvelope from sklearn.preprocessing import StandardScaler from sklearn.model_selection import train_test_split from sklearn.metrics import classification_report, confusion_matrix, accuracy_score #================================# # Load the model # #================================# model = pickle.load(open("./elliptic_envelope.mlmodel", 'rb')) n_features = 7 #===============================# # Input connection string # #===============================# #conn = sys.argv[1] #feature_values = conn.split(",") #conn = "1198897,9,51,13,22,0,10043209,7,7,16,14,32,850,459,17,8,59,5,3,1,3,1,2,0,2,1,1,1,1,4,4,2,1,1,1,1,1" #conn = "0,0,0,0,0,0,0,4,0,0,1043,1373,0,0,0,0,0,0,0,0,0,0,1,1,0,0,18,18,0,0,0,0,1" #conn = "0,0,0,0,0,0,0,2,0,0,1746,14789,0,0,0,0,0,0,0,0,0,0,1,1,0,0,18,18,0,0,0,0,1" #conn = "6,39,0,0,0,0,0,3,0,0,4,31,0,0,22,0,1,1,0,0,0,0,1,1,0,0,1,1,0,0,1,0,1" #conn = "0,0,799,2291,0,0,1006,16,-1" #conn = "0,1007,1235,0,18,18,1" conn = "0,1234,4649,0,3,3,1" #if(sys.argv[1] == 0): if(len(sys.argv) == 1): feature_values = conn.split(",") if(len(feature_values) == n_features): feature_values = [int(i) for i in feature_values] df = pd.DataFrame(np.array(feature_values).reshape(1,n_features)) feature_values = df.values output = model.predict(feature_values) print(output) else: leg_count = 0 att_count = 0 file_to_classify = sys.argv[1] with open(file_to_classify,'r') as fp: for line in fp: line = line.strip() if("class" in line): continue conn = line feature_values = conn.split(",") if(len(feature_values) == n_features): feature_values = [int(i) for i in feature_values] df = pd.DataFrame(np.array(feature_values).reshape(1,n_features)) feature_values = df.values output = model.predict(feature_values) if(output[0] == 1): leg_count = leg_count + 1 else: att_count = att_count + 1 print("Classified connection " + line + " as " + str(output[0])) print("Leg Count = " + str(leg_count)) print("Att Count = " + str(att_count))
[ "numpy.array" ]
[((1571, 1595), 'numpy.array', 'np.array', (['feature_values'], {}), '(feature_values)\n', (1579, 1595), True, 'import numpy as np\n'), ((2178, 2202), 'numpy.array', 'np.array', (['feature_values'], {}), '(feature_values)\n', (2186, 2202), True, 'import numpy as np\n')]
""" TONAS Loader .. admonition:: Dataset Info :class: dropdown This dataset contains a music collection of 72 sung excerpts representative of three a cappella singing styles (Deblas, and two variants of Martinete). It has been developed within the COFLA research project context. The distribution is as follows: 1. 16 Deblas 2. 36 Martinete 1 3. 20 Martinete 2 This collection was built in the context of a study on similarity and style classification of flamenco a cappella singing styles (Tonas) by the flamenco expert Dr. <NAME>, Universidad de Sevilla. We refer to (Mora et al. 2010) for a comprehensive description of the considered styles and their musical characteristics. All 72 excerpts are monophonic, their average duration is 30 seconds and there is enough variability for a proper evaluation of our methods, including a variety of singers, recording conditions, presence of percussion, clapping, background voices and noise. We also provide manual melodic transcriptions, generated by the COFLA team and <NAME>. The annotations are represented by specifying the value (in this case, Notes and F0) at the related timestamps. TONAS' note and F0 annotations also have "Energy" information, which refers to the average energy value through all the frames in which a note or a F0 value is comprised. Using this dataset: TONAS dataset can be obtained upon request. Please refer to this link: https://zenodo.org/record/1290722 to request access and follow the indications of the .download() method for a proper storing and organization of the TONAS dataset. Citing this dataset: When TONAS is used for academic research, we would highly appreciate if scientific publications of works partly based on the TONAS dataset quote the following publication: - Music material: <NAME>., <NAME>., <NAME>., <NAME>., <NAME>. (2010). Melodic Characterization and Similarity in A Cappella Flamenco Cantes. 11th International Society for Music Information Retrieval Conference (ISMIR 2010). - Transcriptions: <NAME>., <NAME>. (in Press). Towards Computer-Assisted Flamenco Transcription: An Experimental Comparison of Automatic Transcription Algorithms As Applied to A Cappella Singing. Computer Music Journal. """ import csv import logging import os from typing import TextIO, Tuple, Optional from deprecated.sphinx import deprecated import librosa import numpy as np from smart_open import open from mirdata import annotations, jams_utils, core, io BIBTEX = """ Music material: @inproceedings{tonas_music, author = {<NAME> <NAME> <NAME> and <NAME> and <NAME>}, year = {2010}, month = {01}, pages = {351-356}, title = {Characterization and Similarity in A Cappella Flamenco Cantes.} } Transcriptions: @inproceedings{tonas_annotations, author = {E. {Gómez} and J. {Bonada}}, journal = {Computer Music Journal}, title = {Towards Computer-Assisted Flamenco Transcription: An Experimental Comparison of Automatic Transcription Algorithms as Applied to A Cappella Singing}, year = {2013}, volume = {37}, number = {2}, pages = {73-90}, doi = {10.1162/COMJ_a_00180}} """ INDEXES = { "default": "1.0", "test": "1.0", "1.0": core.Index(filename="tonas_index_1.0.json"), } DOWNLOAD_INFO = """ PLEASE READ CAREFULLY ALL THE INFORMATION SO YOU DON'T MISS ANY STEP: Unfortunately, the TONAS dataset is not available to be shared openly. However, you can request access to the dataset in the following link, providing a brief explanation of what your are going to use the dataset for: ==> https://zenodo.org/record/1290722 Then, unzip the dataset, change the dataset name to: "tonas" (with lowercase), and locate it to {}. If you unzip it into a different path, please remember to set the right data_home when initializing the dataset. """ LICENSE_INFO = """ The TONAS dataset is offered free of charge for internal non-commercial use only. You can not redistribute it nor modify it. Dataset by COFLA team. Copyright © 2012 COFLA project, Universidad de Sevilla. Distribution rights granted to Music Technology Group, Universitat Pompeu Fabra. All Rights Reserved. """ class Track(core.Track): """TONAS track class Args: track_id (str): track id of the track data_home (str): Local path where the dataset is stored. If `None`, looks for the data in the default directory, `~/mir_datasets/TONAS` Attributes: f0_path (str): local path where f0 melody annotation file is stored notes_path(str): local path where notation annotation file is stored audio_path(str): local path where audio file is stored track_id (str): track id singer (str): performing singer (cantaor) title (str): title of the track song tuning_frequency (float): tuning frequency of the symbolic notation Cached Properties: f0_automatic (F0Data): automatically extracted f0 f0_corrected (F0Data): manually corrected f0 annotations notes (NoteData): annotated notes """ def __init__( self, track_id, data_home, dataset_name, index, metadata, ): super().__init__( track_id, data_home, dataset_name, index, metadata, ) self.f0_path = self.get_path("f0") self.notes_path = self.get_path("notes") self.audio_path = self.get_path("audio") @property def style(self): return self._track_metadata.get("style") @property def singer(self): return self._track_metadata.get("singer") @property def title(self): return self._track_metadata.get("title") @property def tuning_frequency(self): return _load_tuning_frequency(self.notes_path) @property def audio(self) -> Tuple[np.ndarray, float]: """The track's audio Returns: * np.ndarray - audio signal * float - sample rate """ return load_audio(self.audio_path) @core.cached_property def f0_automatic(self) -> Optional[annotations.F0Data]: return load_f0(self.f0_path, False) @core.cached_property def f0_corrected(self) -> Optional[annotations.F0Data]: return load_f0(self.f0_path, True) @property def f0(self): logging.warning( "Track.f0 is deprecated as of 0.3.4 and will be removed in a future version. Use" " Track.f0_automatic or Track.f0_corrected" ) return self.f0_corrected @core.cached_property def notes(self) -> Optional[annotations.NoteData]: return load_notes(self.notes_path) def to_jams(self): """Get the track's data in jams format Returns: jams.JAMS: the track's data in jams format """ return jams_utils.jams_converter( audio_path=self.audio_path, f0_data=[(self.f0, "pitch_contour")], note_data=[(self.notes, "note_hz")], metadata=self._track_metadata, ) def load_audio(fhandle: str) -> Tuple[np.ndarray, float]: """Load a TONAS audio file. Args: fhandle (str): path to an audio file Returns: * np.ndarray - the mono audio signal * float - The sample rate of the audio file """ return librosa.load(fhandle, sr=44100, mono=True) # no decorator because of https://github.com/mir-dataset-loaders/mirdata/issues/503 def load_f0(fpath: str, corrected: bool) -> Optional[annotations.F0Data]: """Load TONAS f0 annotations Args: fpath (str): path pointing to f0 annotation file corrected (bool): if True, loads manually corrected frequency values otherwise, loads automatically extracted frequency values Returns: F0Data: predominant f0 melody """ times = [] freqs = [] freqs_corr = [] energies = [] with open(fpath, "r") as fhandle: reader = np.genfromtxt(fhandle) for line in reader: times.append(float(line[0])) freqs.append(float(line[2])) freqs_corr.append(float(line[3])) energies.append(float(line[1])) freq_array = np.array(freqs_corr if corrected else freqs, dtype="float") energy_array = np.array(energies, dtype="float") voicing_array = (freq_array > 0).astype("float") return annotations.F0Data( np.array(times, dtype="float"), "s", freq_array, "hz", voicing_array, "binary", energy_array, "energy", ) @io.coerce_to_string_io def load_notes(fhandle: TextIO) -> Optional[annotations.NoteData]: """Load TONAS note data from the annotation files Args: fhandle (str or file-like): path or file-like object pointing to a notes annotation file Returns: NoteData: note annotations """ intervals = [] pitches = [] energy = [] reader = csv.reader(fhandle, delimiter=",") tuning = next(reader)[0] for line in reader: intervals.append([line[0], float(line[0]) + float(line[1])]) # Convert midi value to frequency note_hz = _midi_to_hz(float(line[2]), float(tuning)) pitches.append(note_hz) energy.append(float(line[3])) note_data = annotations.NoteData( np.array(intervals, dtype="float"), "s", np.array(pitches, dtype="float"), "hz", np.array(energy, dtype="float"), "energy", ) return note_data @io.coerce_to_string_io def _load_tuning_frequency(fhandle: TextIO) -> float: """Load tuning frequency of the track with re Args: fhandle (str or file-like): path or file-like object pointing to a notes annotation file Returns: tuning_frequency (float): returns new tuning frequency considering the deviation """ # Compute tuning frequency cents_deviation = float(next(csv.reader(fhandle, delimiter=","))[0]) tuning_frequency = 440 * ( 2 ** (cents_deviation / 1200) ) # Frequency of A (common value is 440Hz) return tuning_frequency def _midi_to_hz(midi_note, tuning_deviation): """Function to convert MIDI to Hz with certain tuning freq Args: midi_note (float): note represented in midi value tuning_deviation (float): deviation in cents with respect to 440Hz Returns: (float): note in Hz considering the new tuning frequency """ tuning_frequency = 440 * ( 2 ** (tuning_deviation / 1200) ) # Frequency of A (common value is 440Hz) return (tuning_frequency / 32) * (2 ** ((midi_note - 9) / 12)) @core.docstring_inherit(core.Dataset) class Dataset(core.Dataset): """ The TONAS dataset """ def __init__(self, data_home=None, version="default"): super().__init__( data_home, version, name="tonas", track_class=Track, bibtex=BIBTEX, indexes=INDEXES, download_info=DOWNLOAD_INFO, license_info=LICENSE_INFO, ) @core.cached_property def _metadata(self): metadata_path = os.path.join(self.data_home, "TONAS-Metadata.txt") metadata = {} try: with open(metadata_path, "r", errors="ignore") as f: reader = csv.reader( (x.replace("\0", "") for x in f), delimiter="\t" ) # Fix wrong byte for line in reader: if line: # Do not consider empty lines index = line[0].replace(".wav", "") metadata[index] = { "style": line[1], "title": line[2], "singer": line[3], } except FileNotFoundError: raise FileNotFoundError("Metadata not found. Did you run .download()?") return metadata @deprecated( reason="Use mirdata.datasets.tonas.load_audio", version="0.3.4", ) def load_audio(self, *args, **kwargs): return load_audio(*args, **kwargs) @deprecated( reason="Use mirdata.datasets.tonas.load_f0", version="0.3.4", ) def load_f0(self, *args, **kwargs): return load_f0(*args, **kwargs) @deprecated( reason="Use mirdata.datasets.tonas.load_notes", version="0.3.4", ) def load_notes(self, *args, **kwargs): return load_notes(*args, **kwargs)
[ "numpy.genfromtxt", "logging.warning", "mirdata.core.Index", "smart_open.open", "os.path.join", "mirdata.core.docstring_inherit", "numpy.array", "deprecated.sphinx.deprecated", "mirdata.jams_utils.jams_converter", "csv.reader", "librosa.load" ]
[((10911, 10947), 'mirdata.core.docstring_inherit', 'core.docstring_inherit', (['core.Dataset'], {}), '(core.Dataset)\n', (10933, 10947), False, 'from mirdata import annotations, jams_utils, core, io\n'), ((3345, 3388), 'mirdata.core.Index', 'core.Index', ([], {'filename': '"""tonas_index_1.0.json"""'}), "(filename='tonas_index_1.0.json')\n", (3355, 3388), False, 'from mirdata import annotations, jams_utils, core, io\n'), ((7583, 7625), 'librosa.load', 'librosa.load', (['fhandle'], {'sr': '(44100)', 'mono': '(True)'}), '(fhandle, sr=44100, mono=True)\n', (7595, 7625), False, 'import librosa\n'), ((8459, 8518), 'numpy.array', 'np.array', (['(freqs_corr if corrected else freqs)'], {'dtype': '"""float"""'}), "(freqs_corr if corrected else freqs, dtype='float')\n", (8467, 8518), True, 'import numpy as np\n'), ((8538, 8571), 'numpy.array', 'np.array', (['energies'], {'dtype': '"""float"""'}), "(energies, dtype='float')\n", (8546, 8571), True, 'import numpy as np\n'), ((9210, 9244), 'csv.reader', 'csv.reader', (['fhandle'], {'delimiter': '""","""'}), "(fhandle, delimiter=',')\n", (9220, 9244), False, 'import csv\n'), ((12232, 12307), 'deprecated.sphinx.deprecated', 'deprecated', ([], {'reason': '"""Use mirdata.datasets.tonas.load_audio"""', 'version': '"""0.3.4"""'}), "(reason='Use mirdata.datasets.tonas.load_audio', version='0.3.4')\n", (12242, 12307), False, 'from deprecated.sphinx import deprecated\n'), ((12423, 12495), 'deprecated.sphinx.deprecated', 'deprecated', ([], {'reason': '"""Use mirdata.datasets.tonas.load_f0"""', 'version': '"""0.3.4"""'}), "(reason='Use mirdata.datasets.tonas.load_f0', version='0.3.4')\n", (12433, 12495), False, 'from deprecated.sphinx import deprecated\n'), ((12605, 12680), 'deprecated.sphinx.deprecated', 'deprecated', ([], {'reason': '"""Use mirdata.datasets.tonas.load_notes"""', 'version': '"""0.3.4"""'}), "(reason='Use mirdata.datasets.tonas.load_notes', version='0.3.4')\n", (12615, 12680), False, 'from deprecated.sphinx import deprecated\n'), ((6578, 6727), 'logging.warning', 'logging.warning', (['"""Track.f0 is deprecated as of 0.3.4 and will be removed in a future version. Use Track.f0_automatic or Track.f0_corrected"""'], {}), "(\n 'Track.f0 is deprecated as of 0.3.4 and will be removed in a future version. Use Track.f0_automatic or Track.f0_corrected'\n )\n", (6593, 6727), False, 'import logging\n'), ((7085, 7253), 'mirdata.jams_utils.jams_converter', 'jams_utils.jams_converter', ([], {'audio_path': 'self.audio_path', 'f0_data': "[(self.f0, 'pitch_contour')]", 'note_data': "[(self.notes, 'note_hz')]", 'metadata': 'self._track_metadata'}), "(audio_path=self.audio_path, f0_data=[(self.f0,\n 'pitch_contour')], note_data=[(self.notes, 'note_hz')], metadata=self.\n _track_metadata)\n", (7110, 7253), False, 'from mirdata import annotations, jams_utils, core, io\n'), ((8172, 8188), 'smart_open.open', 'open', (['fpath', '"""r"""'], {}), "(fpath, 'r')\n", (8176, 8188), False, 'from smart_open import open\n'), ((8218, 8240), 'numpy.genfromtxt', 'np.genfromtxt', (['fhandle'], {}), '(fhandle)\n', (8231, 8240), True, 'import numpy as np\n'), ((8665, 8695), 'numpy.array', 'np.array', (['times'], {'dtype': '"""float"""'}), "(times, dtype='float')\n", (8673, 8695), True, 'import numpy as np\n'), ((9587, 9621), 'numpy.array', 'np.array', (['intervals'], {'dtype': '"""float"""'}), "(intervals, dtype='float')\n", (9595, 9621), True, 'import numpy as np\n'), ((9644, 9676), 'numpy.array', 'np.array', (['pitches'], {'dtype': '"""float"""'}), "(pitches, dtype='float')\n", (9652, 9676), True, 'import numpy as np\n'), ((9700, 9731), 'numpy.array', 'np.array', (['energy'], {'dtype': '"""float"""'}), "(energy, dtype='float')\n", (9708, 9731), True, 'import numpy as np\n'), ((11424, 11474), 'os.path.join', 'os.path.join', (['self.data_home', '"""TONAS-Metadata.txt"""'], {}), "(self.data_home, 'TONAS-Metadata.txt')\n", (11436, 11474), False, 'import os\n'), ((10194, 10228), 'csv.reader', 'csv.reader', (['fhandle'], {'delimiter': '""","""'}), "(fhandle, delimiter=',')\n", (10204, 10228), False, 'import csv\n'), ((11528, 11569), 'smart_open.open', 'open', (['metadata_path', '"""r"""'], {'errors': '"""ignore"""'}), "(metadata_path, 'r', errors='ignore')\n", (11532, 11569), False, 'from smart_open import open\n')]
import os import shutil from functools import partial import neptune import numpy as np import pandas as pd from attrdict import AttrDict from steppy.adapter import Adapter, E from steppy.base import IdentityOperation, Step from common_blocks import augmentation as aug from common_blocks import metrics from common_blocks import models from common_blocks import pipelines from common_blocks import postprocessing from common_blocks.utils import io, misc CTX = neptune.Context() LOGGER = misc.init_logger() # ______ ______ .__ __. _______ __ _______ _______. # / | / __ \ | \ | | | ____|| | / _____| / | # | ,----'| | | | | \| | | |__ | | | | __ | (----` # | | | | | | | . ` | | __| | | | | |_ | \ \ # | `----.| `--' | | |\ | | | | | | |__| | .----) | # \______| \______/ |__| \__| |__| |__| \______| |_______/ # EXPERIMENT_DIR = '/output/experiment' CLONE_EXPERIMENT_DIR_FROM = '' # When running eval in the cloud specify this as for example /input/SHIP-14/output/experiment OVERWRITE_EXPERIMENT_DIR = False DEV_MODE = False USE_TTA = True INFERENCE_WITH_SHIP_NO_SHIP = True if OVERWRITE_EXPERIMENT_DIR and os.path.isdir(EXPERIMENT_DIR): shutil.rmtree(EXPERIMENT_DIR) if CLONE_EXPERIMENT_DIR_FROM != '': if os.path.exists(EXPERIMENT_DIR): shutil.rmtree(EXPERIMENT_DIR) shutil.copytree(CLONE_EXPERIMENT_DIR_FROM, EXPERIMENT_DIR) if CTX.params.__class__.__name__ == 'OfflineContextParams': PARAMS = misc.read_yaml().parameters else: PARAMS = CTX.params MEAN = [0.485, 0.456, 0.406] STD = [0.229, 0.224, 0.225] CHUNK_SIZE_SHIP_NO_SHIP = 2500 CHUNK_SIZE_SEGMENTATION = 2500 SEED = 1234 ID_COLUMN = 'id' ID_BIG_IMAGE = 'BigImageId' IS_NOT_EMPTY_COLUMN = 'is_not_empty' X_COLUMN = 'file_path_image' Y_COLUMN = 'file_path_mask' CONFIG = AttrDict({ 'execution': {'experiment_dir': EXPERIMENT_DIR, 'num_workers': PARAMS.num_workers, 'num_threads': PARAMS.num_threads }, 'general': {'img_H-W': (PARAMS.image_h, PARAMS.image_w), 'loader_mode': PARAMS.loader_mode, 'num_classes': 2, 'original_size': (768, 768), }, 'meta_reader': { 'segmentation_network': {'x_columns': [X_COLUMN], 'y_columns': [Y_COLUMN, IS_NOT_EMPTY_COLUMN], }, 'ship_no_ship_network': {'x_columns': [X_COLUMN], 'y_columns': [IS_NOT_EMPTY_COLUMN], }, }, 'loaders': {'resize': {'dataset_params': {'h': PARAMS.image_h, 'w': PARAMS.image_w, 'sns_h': PARAMS.sns_image_h, 'sns_w': PARAMS.sns_image_w, 'image_source': PARAMS.image_source, 'target_format': PARAMS.target_format, 'empty_fraction': PARAMS.training_sampler_empty_fraction, 'sample_size': PARAMS.training_sampler_size, 'sns_empty_fraction': PARAMS.sns_training_sampler_empty_fracion, 'MEAN': MEAN, 'STD': STD }, 'loader_params': {'training': {'batch_size': PARAMS.batch_size_train, 'shuffle': False, 'num_workers': PARAMS.num_workers, 'pin_memory': PARAMS.pin_memory }, 'inference': {'batch_size': PARAMS.batch_size_inference, 'shuffle': False, 'num_workers': PARAMS.num_workers, 'pin_memory': PARAMS.pin_memory }, }, 'augmentation_params': {'image_augment_train': aug.intensity_seq, 'image_augment_with_target_train': aug.resize_seq( resize_target_size=PARAMS.resize_target_size), 'image_augment_inference': aug.resize_to_fit_net( resize_target_size=PARAMS.resize_target_size), 'image_augment_with_target_inference': aug.resize_to_fit_net( resize_target_size=PARAMS.resize_target_size) }, }, 'resize_tta': {'dataset_params': {'h': PARAMS.image_h, 'w': PARAMS.image_w, 'image_source': PARAMS.image_source, 'target_format': PARAMS.target_format, 'empty_fraction': PARAMS.training_sampler_empty_fraction, 'sample_size': PARAMS.training_sampler_size, 'MEAN': MEAN, 'STD': STD }, 'loader_params': {'training': {'batch_size': PARAMS.batch_size_train, 'shuffle': False, 'num_workers': PARAMS.num_workers, 'pin_memory': PARAMS.pin_memory }, 'inference': {'batch_size': PARAMS.batch_size_inference, 'shuffle': False, 'num_workers': PARAMS.num_workers, 'pin_memory': PARAMS.pin_memory }, }, 'augmentation_params': { 'image_augment_inference': aug.resize_to_fit_net( resize_target_size=PARAMS.resize_target_size), 'image_augment_with_target_inference': aug.resize_to_fit_net( resize_target_size=PARAMS.resize_target_size), 'tta_transform': aug.test_time_augmentation_transform }, }, }, 'model': { 'segmentation_network': { 'architecture_config': {'model_params': {'in_channels': PARAMS.image_channels, 'out_channels': PARAMS.network_output_channels, 'architecture': PARAMS.architecture, 'encoder': PARAMS.encoder, 'activation': PARAMS.network_activation, }, 'optimizer_params': {'lr': PARAMS.lr, }, 'regularizer_params': {'regularize': True, 'weight_decay_conv2d': PARAMS.l2_reg_conv, }, 'weights_init': {'function': 'xavier', }, }, 'training_config': {'epochs': PARAMS.epochs_nr, 'shuffle': True, 'batch_size': PARAMS.batch_size_train, 'fine_tuning': PARAMS.fine_tuning, }, 'callbacks_config': {'model_checkpoint': { 'filepath': os.path.join(EXPERIMENT_DIR, 'checkpoints', 'segmentation_network', 'best.torch'), 'epoch_every': 1, 'metric_name': PARAMS.validation_metric_name, 'minimize': PARAMS.minimize_validation_metric}, "one_cycle_scheduler": { "enabled": PARAMS.use_one_cycle, "number_of_batches_per_full_cycle": PARAMS.one_cycle_number_of_batches_per_full_cycle, "max_lr": PARAMS.one_cycle_max_lr, "momentum_range": (0.95, 0.8), "prcnt_annihilate": 10, "div": 10 }, 'exponential_lr_scheduler': {'gamma': PARAMS.gamma, 'epoch_every': 1}, 'reduce_lr_on_plateau_scheduler': {'metric_name': PARAMS.validation_metric_name, 'minimize': PARAMS.minimize_validation_metric, 'reduce_factor': PARAMS.reduce_factor, 'reduce_patience': PARAMS.reduce_patience, 'min_lr': PARAMS.min_lr}, 'training_monitor': {'batch_every': 1, 'epoch_every': 1}, 'experiment_timing': {'batch_every': 10, 'epoch_every': 1}, 'validation_monitor': {'epoch_every': 1, 'data_dir': PARAMS.train_images_dir, 'loader_mode': PARAMS.loader_mode}, 'neptune_monitor': {'model_name': 'network', 'image_nr': 16, 'image_resize': 1.0, 'image_every': 1}, 'early_stopping': {'patience': PARAMS.patience, 'metric_name': PARAMS.validation_metric_name, 'minimize': PARAMS.minimize_validation_metric}, } }, 'ship_no_ship_network': { 'architecture_config': {'model_params': {'architecture': PARAMS.sns_architecture, 'activation': 'sigmoid'}, 'optimizer_params': {'lr': PARAMS.sns_lr, }, 'regularizer_params': {'regularize': True, 'weight_decay_conv2d': PARAMS.sns_l2_reg_conv, }, 'weights_init': {'function': 'xavier', }, }, 'training_config': {'epochs': PARAMS.sns_epochs_nr, 'shuffle': True, 'batch_size': PARAMS.sns_batch_size_train, 'fine_tuning': PARAMS.fine_tuning, }, 'callbacks_config': {'model_checkpoint': { 'filepath': os.path.join(EXPERIMENT_DIR, 'checkpoints', 'ship_no_ship_network', 'best.torch'), 'epoch_every': 1, 'metric_name': PARAMS.sns_validation_metric_name, 'minimize': PARAMS.sns_minimize_validation_metric }, "one_cycle_scheduler": { "enabled": PARAMS.sns_use_one_cycle, "number_of_batches_per_full_cycle": PARAMS.sns_one_cycle_number_of_batches_per_full_cycle, "max_lr": PARAMS.sns_one_cycle_max_lr, "momentum_range": (0.95, 0.7), "prcnt_annihilate": 10, "div": 10 }, 'exponential_lr_scheduler': {'gamma': PARAMS.gamma, 'epoch_every': 1}, 'reduce_lr_on_plateau_scheduler': {'metric_name': PARAMS.sns_validation_metric_name, 'minimize': PARAMS.sns_minimize_validation_metric, 'reduce_factor': PARAMS.reduce_factor, 'reduce_patience': PARAMS.reduce_patience, 'min_lr': PARAMS.min_lr}, 'training_monitor': {'batch_every': 10, 'epoch_every': 1}, 'experiment_timing': {'batch_every': 10, 'epoch_every': 1}, 'validation_monitor': {'epoch_every': 1, 'data_dir': PARAMS.train_images_dir, 'loader_mode': PARAMS.loader_mode}, 'neptune_monitor': {'model_name': 'network', 'image_nr': 16, 'image_resize': 1.0, 'image_every': 1}, 'early_stopping': {'patience': PARAMS.patience, 'metric_name': PARAMS.validation_metric_name, 'minimize': PARAMS.minimize_validation_metric}, } }, }, 'tta_generator': {'flip_ud': True, 'flip_lr': True, 'rotation': True, 'color_shift_runs': False}, 'tta_aggregator': {'tta_inverse_transform': aug.test_time_augmentation_inverse_transform, 'method': PARAMS.tta_aggregation_method, 'nthreads': PARAMS.num_threads }, 'thresholder': {'threshold_masks': PARAMS.threshold_masks, 'threshold_ship_no_ship': PARAMS.sns_threshold, }, }) # .______ __ .______ _______ __ __ .__ __. _______ _______. # | _ \ | | | _ \ | ____|| | | | | \ | | | ____| / | # | |_) | | | | |_) | | |__ | | | | | \| | | |__ | (----` # | ___/ | | | ___/ | __| | | | | | . ` | | __| \ \ # | | | | | | | |____ | `----.| | | |\ | | |____.----) | # | _| |__| | _| |_______||_______||__| |__| \__| |_______|_______/ # def ship_no_ship_pipeline(config, suffix='_ship_no_ship', train_mode=True): if train_mode: preprocessing = pipelines.preprocessing_binary_train(config, model_name='ship_no_ship_network', suffix=suffix) else: preprocessing = pipelines.preprocessing_binary_inference(config, model_name='ship_no_ship_network', suffix=suffix) preprocessing.set_parameters_upstream({'is_fittable': False}) sns_network = misc.FineTuneStep(name='ship_no_ship_network', transformer=models.BinaryModel(**config.model['ship_no_ship_network']), input_steps=[preprocessing], ) sns_network.set_mode_train() sns_network.set_parameters_upstream({'experiment_directory': config.execution.experiment_dir, }) sns_network.force_fitting = False sns_network.fine_tuning = config.model.segmentation_network.training_config.fine_tuning if train_mode: return sns_network else: class_prediction = Step(name='class_prediction', transformer=misc.make_apply_transformer( partial(postprocessing.get_class, threshold=config.thresholder.threshold_ship_no_ship), output_name='classes', apply_on=['predictions']), input_steps=[sns_network], adapter=Adapter({'predictions': E(sns_network.name, 'ship_no_ship_prediction'), }), is_fittable=False ) return class_prediction def train_segmentation_pipeline(config): preprocessing = pipelines.preprocessing_train(config, model_name='segmentation_network') segmentation_network = misc.FineTuneStep(name='segmentation_network', transformer=models.SegmentationModel( **config.model['segmentation_network']), input_data=['callback_input'], input_steps=[preprocessing], adapter=Adapter({'datagen': E(preprocessing.name, 'datagen'), 'validation_datagen': E(preprocessing.name, 'validation_datagen'), 'meta_valid': E('callback_input', 'meta_valid'), })) segmentation_network.set_mode_train() segmentation_network.set_parameters_upstream({'experiment_directory': config.execution.experiment_dir, }) segmentation_network.force_fitting = False segmentation_network.fine_tuning = config.model.segmentation_network.training_config.fine_tuning return segmentation_network def inference_segmentation_pipeline(config): if config.general.loader_mode == 'resize_and_pad': size_adjustment_function = partial(postprocessing.crop_image, target_size=config.general.original_size) elif config.general.loader_mode == 'resize' or config.general.loader_mode == 'stacking': size_adjustment_function = partial(postprocessing.resize_image, target_size=config.general.original_size) else: raise NotImplementedError if USE_TTA: preprocessing, tta_generator = pipelines.preprocessing_inference_tta(config, model_name='segmentation_network') segmentation_network = Step(name='segmentation_network', transformer=models.SegmentationModel(**config.model['segmentation_network']), input_steps=[preprocessing]) tta_aggregator = pipelines.aggregator('tta_aggregator', segmentation_network, tta_generator=tta_generator, config=config.tta_aggregator) prediction_renamed = Step(name='prediction_renamed', transformer=IdentityOperation(), input_steps=[tta_aggregator], adapter=Adapter({'mask_prediction': E(tta_aggregator.name, 'aggregated_prediction') })) mask_resize = Step(name='mask_resize', transformer=misc.make_apply_transformer(size_adjustment_function, output_name='resized_images', apply_on=['images'], n_threads=config.execution.num_threads, ), input_steps=[prediction_renamed], adapter=Adapter({'images': E(prediction_renamed.name, 'mask_prediction'), })) else: preprocessing = pipelines.preprocessing_inference(config, model_name='segmentation_network') segmentation_network = misc.FineTuneStep(name='segmentation_network', transformer=models.SegmentationModel( **config.model['segmentation_network']), input_steps=[preprocessing], ) mask_resize = Step(name='mask_resize', transformer=misc.make_apply_transformer(size_adjustment_function, output_name='resized_images', apply_on=['images'], n_threads=config.execution.num_threads, ), input_steps=[segmentation_network], adapter=Adapter({'images': E(segmentation_network.name, 'mask_prediction'), }), ) binarizer = Step(name='binarizer', transformer=misc.make_apply_transformer( partial(postprocessing.binarize, threshold=config.thresholder.threshold_masks), output_name='binarized_images', apply_on=['images'], n_threads=config.execution.num_threads ), input_steps=[mask_resize], adapter=Adapter({'images': E(mask_resize.name, 'resized_images'), })) labeler = Step(name='labeler', transformer=misc.make_apply_transformer(postprocessing.label, output_name='labeled_images', apply_on=['images'], n_threads=config.execution.num_threads, ), input_steps=[binarizer], adapter=Adapter({'images': E(binarizer.name, 'binarized_images'), })) mask_postprocessing = Step(name='mask_postprocessing', transformer=misc.make_apply_transformer(postprocessing.mask_postprocessing, output_name='labeled_images', apply_on=['images'], n_threads=config.execution.num_threads, ), input_steps=[labeler], adapter=Adapter({'images': E(labeler.name, 'labeled_images'), })) mask_postprocessing.set_mode_inference() mask_postprocessing.set_parameters_upstream({'experiment_directory': config.execution.experiment_dir, 'is_fittable': False }) segmentation_network.is_fittable = True return mask_postprocessing # __________ ___ _______ ______ __ __ .___________. __ ______ .__ __. # | ____\ \ / / | ____| / || | | | | || | / __ \ | \ | | # | |__ \ V / | |__ | ,----'| | | | `---| |----`| | | | | | | \| | # | __| > < | __| | | | | | | | | | | | | | | | . ` | # | |____ / . \ | |____ | `----.| `--' | | | | | | `--' | | |\ | # |_______/__/ \__\ |_______| \______| \______/ |__| |__| \______/ |__| \__| # def train_ship_no_ship(): meta = pd.read_csv(PARAMS.metadata_filepath) meta_train = meta[meta['is_train'] == 1] meta_train = add_big_image_id(meta_train) meta_train_split, meta_valid_split = misc.train_test_split_with_empty_fraction_with_groups(meta_train, groups=meta_train[ ID_BIG_IMAGE], empty_fraction=PARAMS.evaluation_empty_fraction, test_size=PARAMS.evaluation_size, shuffle=True, random_state=SEED) meta_train_split = meta_train_split.sample(frac=1, random_state=SEED) meta_valid_split = meta_valid_split.sample(frac=1, random_state=SEED) if DEV_MODE: meta_train_split = meta_train_split.sample(PARAMS.dev_mode_size, random_state=SEED) meta_valid_split = meta_valid_split.sample(int(PARAMS.dev_mode_size / 2), random_state=SEED) data = {'input': {'meta': meta_train_split }, 'callback_input': {'meta_valid': meta_valid_split } } sns_pipe = ship_no_ship_pipeline(config=CONFIG, train_mode=True) sns_pipe.fit_transform(data) def train(): meta = pd.read_csv(PARAMS.metadata_filepath) meta_train = meta[meta['is_train'] == 1] meta_train = add_big_image_id(meta_train) meta_train_split, meta_valid_split = misc.train_test_split_with_empty_fraction_with_groups(meta_train, groups=meta_train[ ID_BIG_IMAGE], empty_fraction=PARAMS.evaluation_empty_fraction, test_size=PARAMS.evaluation_size, shuffle=True, random_state=SEED) meta_valid_split = meta_valid_split[meta_valid_split[IS_NOT_EMPTY_COLUMN] == 1].sample( PARAMS.in_train_evaluation_size, random_state=SEED) if DEV_MODE: meta_train_split = meta_train_split.sample(PARAMS.dev_mode_size, random_state=SEED) meta_valid_split = meta_valid_split.sample(int(PARAMS.dev_mode_size / 2), random_state=SEED) data = {'input': {'meta': meta_train_split }, 'callback_input': {'meta_valid': meta_valid_split } } pipeline = train_segmentation_pipeline(config=CONFIG) pipeline.fit_transform(data) def evaluate(): meta = pd.read_csv(PARAMS.metadata_filepath) meta_train = meta[meta['is_train'] == 1] meta_train = add_big_image_id(meta_train) _, meta_valid_split = misc.train_test_split_with_empty_fraction_with_groups(meta_train, groups=meta_train[ID_BIG_IMAGE], empty_fraction=PARAMS.evaluation_empty_fraction, test_size=PARAMS.evaluation_size, shuffle=True, random_state=SEED) if DEV_MODE: _, meta_valid_split = misc.train_test_split_with_empty_fraction_with_groups(meta_valid_split, groups=meta_valid_split[ ID_BIG_IMAGE], empty_fraction=PARAMS.evaluation_empty_fraction, test_size=PARAMS.evaluation_size, shuffle=True, random_state=SEED) segm_pipe = inference_segmentation_pipeline(config=CONFIG) valid_ids = meta_valid_split[ID_COLUMN] + '.jpg' if INFERENCE_WITH_SHIP_NO_SHIP: sns_pipe = ship_no_ship_pipeline(config=CONFIG, train_mode=False) ids_ship, ids_no_ship = predict_ship_no_ship(meta_valid_split, sns_pipe, CHUNK_SIZE_SHIP_NO_SHIP) meta_valid_ship = meta_valid_split[valid_ids.isin(ids_ship)] prediction_ship = generate_submission(meta_valid_ship, segm_pipe, CHUNK_SIZE_SEGMENTATION) prediction = misc.combine_two_stage_predictions(ids_no_ship, prediction_ship, valid_ids) else: prediction = generate_submission(meta_valid_split, segm_pipe, CHUNK_SIZE_SEGMENTATION) gt = io.read_gt_subset(PARAMS.annotation_file, valid_ids) f2_per_image, image_ids = metrics.f_beta_metric(gt, prediction, beta=2, apply_mean=False) f2 = np.mean(f2_per_image) LOGGER.info('f2 {}'.format(f2)) CTX.channel_send('f2', 0, f2) LOGGER.info('preparing results') results = misc.prepare_results(gt, prediction, meta_valid_split, f2_per_image, image_ids) results_filepath = os.path.join(EXPERIMENT_DIR, 'validation_results.csv') results.to_csv(results_filepath, index=None) def predict(): meta = pd.read_csv(PARAMS.metadata_filepath) meta_test = meta[meta['is_train'] == 0] if DEV_MODE: meta_test = meta_test.sample(PARAMS.dev_mode_size, random_state=SEED) segm_pipe = inference_segmentation_pipeline(config=CONFIG) test_ids = meta_test[ID_COLUMN] + '.jpg' if INFERENCE_WITH_SHIP_NO_SHIP: sns_pipe = ship_no_ship_pipeline(config=CONFIG, train_mode=False) ids_ship, ids_no_ship = predict_ship_no_ship(meta_test, sns_pipe, CHUNK_SIZE_SHIP_NO_SHIP) meta_test_ship = meta_test[test_ids.isin(ids_ship)] prediction_ship = generate_submission(meta_test_ship, segm_pipe, CHUNK_SIZE_SEGMENTATION) submission = misc.combine_two_stage_predictions(ids_no_ship, prediction_ship, test_ids) else: submission = generate_submission(meta_test, segm_pipe, CHUNK_SIZE_SEGMENTATION) submission_filepath = os.path.join(EXPERIMENT_DIR, 'submission.csv') submission.to_csv(submission_filepath, index=None, encoding='utf-8') LOGGER.info('submission saved to {}'.format(submission_filepath)) LOGGER.info('submission head \n\n{}'.format(submission.head())) # __ __ .___________. __ __ _______. # | | | | | || | | | / | # | | | | `---| |----`| | | | | (----` # | | | | | | | | | | \ \ # | `--' | | | | | | `----.----) | # \______/ |__| |__| |_______|_______/ # def add_big_image_id(meta): big_image_ids = pd.read_csv('big-images-ids_v2.csv') meta['ImageId'] = meta[ID_COLUMN] + '.jpg' meta_joined = pd.merge(meta, big_image_ids, on='ImageId') return meta_joined def generate_submission(meta_data, pipeline, chunk_size): if chunk_size is not None: return _generate_submission_in_chunks(meta_data, pipeline, chunk_size) else: return _generate_submission(meta_data, pipeline) def _generate_submission(meta_data, pipeline): prediction = _generate_prediction(meta_data, pipeline) submission = misc.create_submission(meta_data[ID_COLUMN] + '.jpg', prediction) return submission def _generate_submission_in_chunks(meta_data, pipeline, chunk_size): submissions = [] for meta_chunk in misc.generate_data_frame_chunks(meta_data, chunk_size): prediction_chunk = _generate_prediction(meta_chunk, pipeline) submission_chunk = misc.create_submission(meta_chunk[ID_COLUMN] + '.jpg', prediction_chunk) submissions.append(submission_chunk) submission = pd.concat(submissions) return submission def _generate_prediction(meta_data, pipeline): data = {'input': {'meta': meta_data, }, 'callback_input': {'meta_valid': None } } output = pipeline.transform(data) y_pred = output['labeled_images'] return y_pred def predict_ship_no_ship(meta_data, pipeline, chunk_size): if chunk_size is not None: return _predict_ship_no_ship_in_chunks(meta_data, pipeline, chunk_size) else: return _predict_ship_no_ship(meta_data, pipeline) def _predict_ship_no_ship(meta_data, pipeline): prediction = _generate_prediction_ship_no_ship(meta_data, pipeline) ids_ship, ids_no_ship = misc.get_ship_no_ship_ids(meta_data[ID_COLUMN] + '.jpg', prediction) return ids_ship, ids_no_ship def _predict_ship_no_ship_in_chunks(meta_data, pipeline, chunk_size): ids_ship, ids_no_ship = [], [] for meta_chunk in misc.generate_data_frame_chunks(meta_data, chunk_size): prediction_chunk = _generate_prediction_ship_no_ship(meta_chunk, pipeline) ids_ship_chunk, ids_no_ship_chunk = misc.get_ship_no_ship_ids(meta_chunk[ID_COLUMN] + '.jpg', prediction_chunk) ids_ship.extend(ids_ship_chunk) ids_no_ship.extend(ids_no_ship_chunk) return ids_ship, ids_no_ship def _generate_prediction_ship_no_ship(meta_data, pipeline): data = {'input': {'meta': meta_data, }, 'callback_input': {'meta_valid': None } } output = pipeline.transform(data) y_pred = output['classes'] return y_pred # .___ ___. ___ __ .__ __. # | \/ | / \ | | | \ | | # | \ / | / ^ \ | | | \| | # | |\/| | / /_\ \ | | | . ` | # | | | | / _____ \ | | | |\ | # |__| |__| /__/ \__\ |__| |__| \__| # if __name__ == '__main__': train_ship_no_ship() train() evaluate() predict()
[ "pandas.read_csv", "steppy.adapter.E", "common_blocks.utils.misc.generate_data_frame_chunks", "common_blocks.augmentation.resize_to_fit_net", "common_blocks.metrics.f_beta_metric", "os.path.exists", "numpy.mean", "common_blocks.utils.misc.read_yaml", "common_blocks.models.BinaryModel", "common_blo...
[((464, 481), 'neptune.Context', 'neptune.Context', ([], {}), '()\n', (479, 481), False, 'import neptune\n'), ((491, 509), 'common_blocks.utils.misc.init_logger', 'misc.init_logger', ([], {}), '()\n', (507, 509), False, 'from common_blocks.utils import io, misc\n'), ((1227, 1256), 'os.path.isdir', 'os.path.isdir', (['EXPERIMENT_DIR'], {}), '(EXPERIMENT_DIR)\n', (1240, 1256), False, 'import os\n'), ((1262, 1291), 'shutil.rmtree', 'shutil.rmtree', (['EXPERIMENT_DIR'], {}), '(EXPERIMENT_DIR)\n', (1275, 1291), False, 'import shutil\n'), ((1335, 1365), 'os.path.exists', 'os.path.exists', (['EXPERIMENT_DIR'], {}), '(EXPERIMENT_DIR)\n', (1349, 1365), False, 'import os\n'), ((1409, 1467), 'shutil.copytree', 'shutil.copytree', (['CLONE_EXPERIMENT_DIR_FROM', 'EXPERIMENT_DIR'], {}), '(CLONE_EXPERIMENT_DIR_FROM, EXPERIMENT_DIR)\n', (1424, 1467), False, 'import shutil\n'), ((17421, 17493), 'common_blocks.pipelines.preprocessing_train', 'pipelines.preprocessing_train', (['config'], {'model_name': '"""segmentation_network"""'}), "(config, model_name='segmentation_network')\n", (17450, 17493), False, 'from common_blocks import pipelines\n'), ((25046, 25083), 'pandas.read_csv', 'pd.read_csv', (['PARAMS.metadata_filepath'], {}), '(PARAMS.metadata_filepath)\n', (25057, 25083), True, 'import pandas as pd\n'), ((25217, 25446), 'common_blocks.utils.misc.train_test_split_with_empty_fraction_with_groups', 'misc.train_test_split_with_empty_fraction_with_groups', (['meta_train'], {'groups': 'meta_train[ID_BIG_IMAGE]', 'empty_fraction': 'PARAMS.evaluation_empty_fraction', 'test_size': 'PARAMS.evaluation_size', 'shuffle': '(True)', 'random_state': 'SEED'}), '(meta_train, groups=\n meta_train[ID_BIG_IMAGE], empty_fraction=PARAMS.\n evaluation_empty_fraction, test_size=PARAMS.evaluation_size, shuffle=\n True, random_state=SEED)\n', (25270, 25446), False, 'from common_blocks.utils import io, misc\n'), ((26678, 26715), 'pandas.read_csv', 'pd.read_csv', (['PARAMS.metadata_filepath'], {}), '(PARAMS.metadata_filepath)\n', (26689, 26715), True, 'import pandas as pd\n'), ((26849, 27078), 'common_blocks.utils.misc.train_test_split_with_empty_fraction_with_groups', 'misc.train_test_split_with_empty_fraction_with_groups', (['meta_train'], {'groups': 'meta_train[ID_BIG_IMAGE]', 'empty_fraction': 'PARAMS.evaluation_empty_fraction', 'test_size': 'PARAMS.evaluation_size', 'shuffle': '(True)', 'random_state': 'SEED'}), '(meta_train, groups=\n meta_train[ID_BIG_IMAGE], empty_fraction=PARAMS.\n evaluation_empty_fraction, test_size=PARAMS.evaluation_size, shuffle=\n True, random_state=SEED)\n', (26902, 27078), False, 'from common_blocks.utils import io, misc\n'), ((28306, 28343), 'pandas.read_csv', 'pd.read_csv', (['PARAMS.metadata_filepath'], {}), '(PARAMS.metadata_filepath)\n', (28317, 28343), True, 'import pandas as pd\n'), ((28462, 28691), 'common_blocks.utils.misc.train_test_split_with_empty_fraction_with_groups', 'misc.train_test_split_with_empty_fraction_with_groups', (['meta_train'], {'groups': 'meta_train[ID_BIG_IMAGE]', 'empty_fraction': 'PARAMS.evaluation_empty_fraction', 'test_size': 'PARAMS.evaluation_size', 'shuffle': '(True)', 'random_state': 'SEED'}), '(meta_train, groups=\n meta_train[ID_BIG_IMAGE], empty_fraction=PARAMS.\n evaluation_empty_fraction, test_size=PARAMS.evaluation_size, shuffle=\n True, random_state=SEED)\n', (28515, 28691), False, 'from common_blocks.utils import io, misc\n'), ((30411, 30463), 'common_blocks.utils.io.read_gt_subset', 'io.read_gt_subset', (['PARAMS.annotation_file', 'valid_ids'], {}), '(PARAMS.annotation_file, valid_ids)\n', (30428, 30463), False, 'from common_blocks.utils import io, misc\n'), ((30494, 30557), 'common_blocks.metrics.f_beta_metric', 'metrics.f_beta_metric', (['gt', 'prediction'], {'beta': '(2)', 'apply_mean': '(False)'}), '(gt, prediction, beta=2, apply_mean=False)\n', (30515, 30557), False, 'from common_blocks import metrics\n'), ((30567, 30588), 'numpy.mean', 'np.mean', (['f2_per_image'], {}), '(f2_per_image)\n', (30574, 30588), True, 'import numpy as np\n'), ((30711, 30790), 'common_blocks.utils.misc.prepare_results', 'misc.prepare_results', (['gt', 'prediction', 'meta_valid_split', 'f2_per_image', 'image_ids'], {}), '(gt, prediction, meta_valid_split, f2_per_image, image_ids)\n', (30731, 30790), False, 'from common_blocks.utils import io, misc\n'), ((30814, 30868), 'os.path.join', 'os.path.join', (['EXPERIMENT_DIR', '"""validation_results.csv"""'], {}), "(EXPERIMENT_DIR, 'validation_results.csv')\n", (30826, 30868), False, 'import os\n'), ((30946, 30983), 'pandas.read_csv', 'pd.read_csv', (['PARAMS.metadata_filepath'], {}), '(PARAMS.metadata_filepath)\n', (30957, 30983), True, 'import pandas as pd\n'), ((31822, 31868), 'os.path.join', 'os.path.join', (['EXPERIMENT_DIR', '"""submission.csv"""'], {}), "(EXPERIMENT_DIR, 'submission.csv')\n", (31834, 31868), False, 'import os\n'), ((32447, 32483), 'pandas.read_csv', 'pd.read_csv', (['"""big-images-ids_v2.csv"""'], {}), "('big-images-ids_v2.csv')\n", (32458, 32483), True, 'import pandas as pd\n'), ((32549, 32592), 'pandas.merge', 'pd.merge', (['meta', 'big_image_ids'], {'on': '"""ImageId"""'}), "(meta, big_image_ids, on='ImageId')\n", (32557, 32592), True, 'import pandas as pd\n'), ((32978, 33043), 'common_blocks.utils.misc.create_submission', 'misc.create_submission', (["(meta_data[ID_COLUMN] + '.jpg')", 'prediction'], {}), "(meta_data[ID_COLUMN] + '.jpg', prediction)\n", (33000, 33043), False, 'from common_blocks.utils import io, misc\n'), ((33180, 33234), 'common_blocks.utils.misc.generate_data_frame_chunks', 'misc.generate_data_frame_chunks', (['meta_data', 'chunk_size'], {}), '(meta_data, chunk_size)\n', (33211, 33234), False, 'from common_blocks.utils import io, misc\n'), ((33468, 33490), 'pandas.concat', 'pd.concat', (['submissions'], {}), '(submissions)\n', (33477, 33490), True, 'import pandas as pd\n'), ((34209, 34277), 'common_blocks.utils.misc.get_ship_no_ship_ids', 'misc.get_ship_no_ship_ids', (["(meta_data[ID_COLUMN] + '.jpg')", 'prediction'], {}), "(meta_data[ID_COLUMN] + '.jpg', prediction)\n", (34234, 34277), False, 'from common_blocks.utils import io, misc\n'), ((34440, 34494), 'common_blocks.utils.misc.generate_data_frame_chunks', 'misc.generate_data_frame_chunks', (['meta_data', 'chunk_size'], {}), '(meta_data, chunk_size)\n', (34471, 34494), False, 'from common_blocks.utils import io, misc\n'), ((1375, 1404), 'shutil.rmtree', 'shutil.rmtree', (['EXPERIMENT_DIR'], {}), '(EXPERIMENT_DIR)\n', (1388, 1404), False, 'import shutil\n'), ((1542, 1558), 'common_blocks.utils.misc.read_yaml', 'misc.read_yaml', ([], {}), '()\n', (1556, 1558), False, 'from common_blocks.utils import io, misc\n'), ((15599, 15698), 'common_blocks.pipelines.preprocessing_binary_train', 'pipelines.preprocessing_binary_train', (['config'], {'model_name': '"""ship_no_ship_network"""', 'suffix': 'suffix'}), "(config, model_name=\n 'ship_no_ship_network', suffix=suffix)\n", (15635, 15698), False, 'from common_blocks import pipelines\n'), ((15728, 15831), 'common_blocks.pipelines.preprocessing_binary_inference', 'pipelines.preprocessing_binary_inference', (['config'], {'model_name': '"""ship_no_ship_network"""', 'suffix': 'suffix'}), "(config, model_name=\n 'ship_no_ship_network', suffix=suffix)\n", (15768, 15831), False, 'from common_blocks import pipelines\n'), ((18911, 18987), 'functools.partial', 'partial', (['postprocessing.crop_image'], {'target_size': 'config.general.original_size'}), '(postprocessing.crop_image, target_size=config.general.original_size)\n', (18918, 18987), False, 'from functools import partial\n'), ((19295, 19380), 'common_blocks.pipelines.preprocessing_inference_tta', 'pipelines.preprocessing_inference_tta', (['config'], {'model_name': '"""segmentation_network"""'}), "(config, model_name='segmentation_network'\n )\n", (19332, 19380), False, 'from common_blocks import pipelines\n'), ((19647, 19771), 'common_blocks.pipelines.aggregator', 'pipelines.aggregator', (['"""tta_aggregator"""', 'segmentation_network'], {'tta_generator': 'tta_generator', 'config': 'config.tta_aggregator'}), "('tta_aggregator', segmentation_network, tta_generator=\n tta_generator, config=config.tta_aggregator)\n", (19667, 19771), False, 'from common_blocks import pipelines\n'), ((20974, 21050), 'common_blocks.pipelines.preprocessing_inference', 'pipelines.preprocessing_inference', (['config'], {'model_name': '"""segmentation_network"""'}), "(config, model_name='segmentation_network')\n", (21007, 21050), False, 'from common_blocks import pipelines\n'), ((29045, 29285), 'common_blocks.utils.misc.train_test_split_with_empty_fraction_with_groups', 'misc.train_test_split_with_empty_fraction_with_groups', (['meta_valid_split'], {'groups': 'meta_valid_split[ID_BIG_IMAGE]', 'empty_fraction': 'PARAMS.evaluation_empty_fraction', 'test_size': 'PARAMS.evaluation_size', 'shuffle': '(True)', 'random_state': 'SEED'}), '(meta_valid_split,\n groups=meta_valid_split[ID_BIG_IMAGE], empty_fraction=PARAMS.\n evaluation_empty_fraction, test_size=PARAMS.evaluation_size, shuffle=\n True, random_state=SEED)\n', (29098, 29285), False, 'from common_blocks.utils import io, misc\n'), ((30220, 30295), 'common_blocks.utils.misc.combine_two_stage_predictions', 'misc.combine_two_stage_predictions', (['ids_no_ship', 'prediction_ship', 'valid_ids'], {}), '(ids_no_ship, prediction_ship, valid_ids)\n', (30254, 30295), False, 'from common_blocks.utils import io, misc\n'), ((31622, 31696), 'common_blocks.utils.misc.combine_two_stage_predictions', 'misc.combine_two_stage_predictions', (['ids_no_ship', 'prediction_ship', 'test_ids'], {}), '(ids_no_ship, prediction_ship, test_ids)\n', (31656, 31696), False, 'from common_blocks.utils import io, misc\n'), ((33333, 33405), 'common_blocks.utils.misc.create_submission', 'misc.create_submission', (["(meta_chunk[ID_COLUMN] + '.jpg')", 'prediction_chunk'], {}), "(meta_chunk[ID_COLUMN] + '.jpg', prediction_chunk)\n", (33355, 33405), False, 'from common_blocks.utils import io, misc\n'), ((34623, 34698), 'common_blocks.utils.misc.get_ship_no_ship_ids', 'misc.get_ship_no_ship_ids', (["(meta_chunk[ID_COLUMN] + '.jpg')", 'prediction_chunk'], {}), "(meta_chunk[ID_COLUMN] + '.jpg', prediction_chunk)\n", (34648, 34698), False, 'from common_blocks.utils import io, misc\n'), ((16072, 16130), 'common_blocks.models.BinaryModel', 'models.BinaryModel', ([], {}), "(**config.model['ship_no_ship_network'])\n", (16090, 16130), False, 'from common_blocks import models\n'), ((17626, 17690), 'common_blocks.models.SegmentationModel', 'models.SegmentationModel', ([], {}), "(**config.model['segmentation_network'])\n", (17650, 17690), False, 'from common_blocks import models\n'), ((19116, 19194), 'functools.partial', 'partial', (['postprocessing.resize_image'], {'target_size': 'config.general.original_size'}), '(postprocessing.resize_image, target_size=config.general.original_size)\n', (19123, 19194), False, 'from functools import partial\n'), ((22829, 22979), 'common_blocks.utils.misc.make_apply_transformer', 'misc.make_apply_transformer', (['postprocessing.label'], {'output_name': '"""labeled_images"""', 'apply_on': "['images']", 'n_threads': 'config.execution.num_threads'}), "(postprocessing.label, output_name=\n 'labeled_images', apply_on=['images'], n_threads=config.execution.\n num_threads)\n", (22856, 22979), False, 'from common_blocks.utils import io, misc\n'), ((23480, 23644), 'common_blocks.utils.misc.make_apply_transformer', 'misc.make_apply_transformer', (['postprocessing.mask_postprocessing'], {'output_name': '"""labeled_images"""', 'apply_on': "['images']", 'n_threads': 'config.execution.num_threads'}), "(postprocessing.mask_postprocessing, output_name\n ='labeled_images', apply_on=['images'], n_threads=config.execution.\n num_threads)\n", (23507, 23644), False, 'from common_blocks.utils import io, misc\n'), ((19490, 19554), 'common_blocks.models.SegmentationModel', 'models.SegmentationModel', ([], {}), "(**config.model['segmentation_network'])\n", (19514, 19554), False, 'from common_blocks import models\n'), ((19968, 19987), 'steppy.base.IdentityOperation', 'IdentityOperation', ([], {}), '()\n', (19985, 19987), False, 'from steppy.base import IdentityOperation, Step\n'), ((20313, 20467), 'common_blocks.utils.misc.make_apply_transformer', 'misc.make_apply_transformer', (['size_adjustment_function'], {'output_name': '"""resized_images"""', 'apply_on': "['images']", 'n_threads': 'config.execution.num_threads'}), "(size_adjustment_function, output_name=\n 'resized_images', apply_on=['images'], n_threads=config.execution.\n num_threads)\n", (20340, 20467), False, 'from common_blocks.utils import io, misc\n'), ((21191, 21255), 'common_blocks.models.SegmentationModel', 'models.SegmentationModel', ([], {}), "(**config.model['segmentation_network'])\n", (21215, 21255), False, 'from common_blocks import models\n'), ((21528, 21682), 'common_blocks.utils.misc.make_apply_transformer', 'misc.make_apply_transformer', (['size_adjustment_function'], {'output_name': '"""resized_images"""', 'apply_on': "['images']", 'n_threads': 'config.execution.num_threads'}), "(size_adjustment_function, output_name=\n 'resized_images', apply_on=['images'], n_threads=config.execution.\n num_threads)\n", (21555, 21682), False, 'from common_blocks.utils import io, misc\n'), ((22314, 22392), 'functools.partial', 'partial', (['postprocessing.binarize'], {'threshold': 'config.thresholder.threshold_masks'}), '(postprocessing.binarize, threshold=config.thresholder.threshold_masks)\n', (22321, 22392), False, 'from functools import partial\n'), ((4640, 4700), 'common_blocks.augmentation.resize_seq', 'aug.resize_seq', ([], {'resize_target_size': 'PARAMS.resize_target_size'}), '(resize_target_size=PARAMS.resize_target_size)\n', (4654, 4700), True, 'from common_blocks import augmentation as aug\n'), ((4836, 4903), 'common_blocks.augmentation.resize_to_fit_net', 'aug.resize_to_fit_net', ([], {'resize_target_size': 'PARAMS.resize_target_size'}), '(resize_target_size=PARAMS.resize_target_size)\n', (4857, 4903), True, 'from common_blocks import augmentation as aug\n'), ((5051, 5118), 'common_blocks.augmentation.resize_to_fit_net', 'aug.resize_to_fit_net', ([], {'resize_target_size': 'PARAMS.resize_target_size'}), '(resize_target_size=PARAMS.resize_target_size)\n', (5072, 5118), True, 'from common_blocks import augmentation as aug\n'), ((7012, 7079), 'common_blocks.augmentation.resize_to_fit_net', 'aug.resize_to_fit_net', ([], {'resize_target_size': 'PARAMS.resize_target_size'}), '(resize_target_size=PARAMS.resize_target_size)\n', (7033, 7079), True, 'from common_blocks import augmentation as aug\n'), ((7195, 7262), 'common_blocks.augmentation.resize_to_fit_net', 'aug.resize_to_fit_net', ([], {'resize_target_size': 'PARAMS.resize_target_size'}), '(resize_target_size=PARAMS.resize_target_size)\n', (7216, 7262), True, 'from common_blocks import augmentation as aug\n'), ((16764, 16855), 'functools.partial', 'partial', (['postprocessing.get_class'], {'threshold': 'config.thresholder.threshold_ship_no_ship'}), '(postprocessing.get_class, threshold=config.thresholder.\n threshold_ship_no_ship)\n', (16771, 16855), False, 'from functools import partial\n'), ((17965, 17997), 'steppy.adapter.E', 'E', (['preprocessing.name', '"""datagen"""'], {}), "(preprocessing.name, 'datagen')\n", (17966, 17997), False, 'from steppy.adapter import Adapter, E\n'), ((18083, 18126), 'steppy.adapter.E', 'E', (['preprocessing.name', '"""validation_datagen"""'], {}), "(preprocessing.name, 'validation_datagen')\n", (18084, 18126), False, 'from steppy.adapter import Adapter, E\n'), ((18290, 18323), 'steppy.adapter.E', 'E', (['"""callback_input"""', '"""meta_valid"""'], {}), "('callback_input', 'meta_valid')\n", (18291, 18323), False, 'from steppy.adapter import Adapter, E\n'), ((22681, 22718), 'steppy.adapter.E', 'E', (['mask_resize.name', '"""resized_images"""'], {}), "(mask_resize.name, 'resized_images')\n", (22682, 22718), False, 'from steppy.adapter import Adapter, E\n'), ((23299, 23336), 'steppy.adapter.E', 'E', (['binarizer.name', '"""binarized_images"""'], {}), "(binarizer.name, 'binarized_images')\n", (23300, 23336), False, 'from steppy.adapter import Adapter, E\n'), ((24034, 24067), 'steppy.adapter.E', 'E', (['labeler.name', '"""labeled_images"""'], {}), "(labeler.name, 'labeled_images')\n", (24035, 24067), False, 'from steppy.adapter import Adapter, E\n'), ((8953, 9038), 'os.path.join', 'os.path.join', (['EXPERIMENT_DIR', '"""checkpoints"""', '"""segmentation_network"""', '"""best.torch"""'], {}), "(EXPERIMENT_DIR, 'checkpoints', 'segmentation_network',\n 'best.torch')\n", (8965, 9038), False, 'import os\n'), ((12232, 12317), 'os.path.join', 'os.path.join', (['EXPERIMENT_DIR', '"""checkpoints"""', '"""ship_no_ship_network"""', '"""best.torch"""'], {}), "(EXPERIMENT_DIR, 'checkpoints', 'ship_no_ship_network',\n 'best.torch')\n", (12244, 12317), False, 'import os\n'), ((17141, 17187), 'steppy.adapter.E', 'E', (['sns_network.name', '"""ship_no_ship_prediction"""'], {}), "(sns_network.name, 'ship_no_ship_prediction')\n", (17142, 17187), False, 'from steppy.adapter import Adapter, E\n'), ((20123, 20170), 'steppy.adapter.E', 'E', (['tta_aggregator.name', '"""aggregated_prediction"""'], {}), "(tta_aggregator.name, 'aggregated_prediction')\n", (20124, 20170), False, 'from steppy.adapter import Adapter, E\n'), ((20844, 20889), 'steppy.adapter.E', 'E', (['prediction_renamed.name', '"""mask_prediction"""'], {}), "(prediction_renamed.name, 'mask_prediction')\n", (20845, 20889), False, 'from steppy.adapter import Adapter, E\n'), ((22061, 22108), 'steppy.adapter.E', 'E', (['segmentation_network.name', '"""mask_prediction"""'], {}), "(segmentation_network.name, 'mask_prediction')\n", (22062, 22108), False, 'from steppy.adapter import Adapter, E\n')]
# @Author: <NAME> <narsi> # @Date: 2020-01-18T20:22:17-06:00 # @Email: <EMAIL> # @Last modified by: narsi # @Last modified time: 2020-01-18T20:24:43-06:00 import numpy as np import json from PIL import Image from .msssim import MultiScaleSSIM from tqdm import tqdm def evaluate2(submission_images, target_images, settings={}): """ Calculates metrics for the given images. """ if settings is None: settings = {} metrics = settings.get('metrics', ['PSNR', 'MSSSIM']) num_dims = 0 sqerror_values = [] msssim_values = [] for img1, img0 in tqdm(zip(submission_images, target_images)): image0 = np.asarray(Image.open(img0).convert('RGB'), dtype=np.float32) image1 = np.asarray(Image.open(img1).convert('RGB'), dtype=np.float32) num_dims += image0.size if 'PSNR' in metrics: sqerror_values.append(mse(image1, image0)) if 'MSSSIM' in metrics: msssim_values.append(msssim(image0, image1) * image0.size) results = {} if 'PSNR' in metrics: results['PSNR'] = mse2psnr(np.sum(sqerror_values) / num_dims) if 'MSSSIM' in metrics: results['MSSSIM'] = np.sum(msssim_values) / num_dims return results def evaluate(submission_images, target_images, settings={}): """ Calculates metrics for the given images. """ if settings is None: settings = {} metrics = settings.get('metrics', ['PSNR', 'MSSSIM']) num_dims = 0 sqerror_values = [] msssim_values = [] for name in target_images: image0 = np.asarray(Image.open(target_images[name]).convert('RGB'), dtype=np.float32) image1 = np.asarray(Image.open(submission_images[name]).convert('RGB'), dtype=np.float32) num_dims += image0.size if 'PSNR' in metrics: sqerror_values.append(mse(image1, image0)) if 'MSSSIM' in metrics: msssim_values.append(msssim(image0, image1) * image0.size) results = {} if 'PSNR' in metrics: results['PSNR'] = mse2psnr(np.sum(sqerror_values) / num_dims) if 'MSSSIM' in metrics: results['MSSSIM'] = np.sum(msssim_values) / num_dims return results def mse(image0, image1): return np.sum(np.square(image1 - image0)) def mse2psnr(mse): return 20. * np.log10(255.) - 10. * np.log10(mse) def msssim(image0, image1): return MultiScaleSSIM(image0[None], image1[None])
[ "numpy.sum", "numpy.log10", "PIL.Image.open", "numpy.square" ]
[((2032, 2058), 'numpy.square', 'np.square', (['(image1 - image0)'], {}), '(image1 - image0)\n', (2041, 2058), True, 'import numpy as np\n'), ((1082, 1103), 'numpy.sum', 'np.sum', (['msssim_values'], {}), '(msssim_values)\n', (1088, 1103), True, 'import numpy as np\n'), ((1940, 1961), 'numpy.sum', 'np.sum', (['msssim_values'], {}), '(msssim_values)\n', (1946, 1961), True, 'import numpy as np\n'), ((2095, 2110), 'numpy.log10', 'np.log10', (['(255.0)'], {}), '(255.0)\n', (2103, 2110), True, 'import numpy as np\n'), ((2118, 2131), 'numpy.log10', 'np.log10', (['mse'], {}), '(mse)\n', (2126, 2131), True, 'import numpy as np\n'), ((1000, 1022), 'numpy.sum', 'np.sum', (['sqerror_values'], {}), '(sqerror_values)\n', (1006, 1022), True, 'import numpy as np\n'), ((1858, 1880), 'numpy.sum', 'np.sum', (['sqerror_values'], {}), '(sqerror_values)\n', (1864, 1880), True, 'import numpy as np\n'), ((622, 638), 'PIL.Image.open', 'Image.open', (['img0'], {}), '(img0)\n', (632, 638), False, 'from PIL import Image\n'), ((695, 711), 'PIL.Image.open', 'Image.open', (['img1'], {}), '(img1)\n', (705, 711), False, 'from PIL import Image\n'), ((1446, 1477), 'PIL.Image.open', 'Image.open', (['target_images[name]'], {}), '(target_images[name])\n', (1456, 1477), False, 'from PIL import Image\n'), ((1534, 1569), 'PIL.Image.open', 'Image.open', (['submission_images[name]'], {}), '(submission_images[name])\n', (1544, 1569), False, 'from PIL import Image\n')]
# # This file is part of the chi repository # (https://github.com/DavAug/chi/) which is released under the # BSD 3-clause license. See accompanying LICENSE.md for copyright notice and # full license details. # import unittest import numpy as np from chi import plots from chi.library import DataLibrary class TestResidualPlot(unittest.TestCase): """ Tests the chi.plots.ResidualPlot class. """ @classmethod def setUpClass(cls): # Create test datasets cls.measurements = DataLibrary().lung_cancer_control_group() cls.data = cls.measurements.rename( columns={'Measurement': 'Sample'}) # Create test figure cls.fig = plots.ResidualPlot(cls.measurements) def test_bad_instantiation(self): # Create data of wrong type measurements = np.ones(shape=(10, 4)) with self.assertRaisesRegex(TypeError, 'Measurements has to be'): plots.ResidualPlot(measurements) # Wrong ID key with self.assertRaisesRegex(ValueError, 'Measurements does not have'): plots.ResidualPlot(self.measurements, id_key='Wrong') # Wrong time key with self.assertRaisesRegex(ValueError, 'Measurements does not have'): plots.ResidualPlot(self.measurements, time_key='Wrong') # Wrong biomarker key with self.assertRaisesRegex(ValueError, 'Measurements does not have'): plots.ResidualPlot(self.measurements, biom_key='Wrong') # Wrong measurement key with self.assertRaisesRegex(ValueError, 'Measurements does not have'): plots.ResidualPlot(self.measurements, meas_key='Wrong') def test_add_data_wrong_data_type(self): # Create data of wrong type data = np.ones(shape=(10, 4)) with self.assertRaisesRegex(TypeError, 'Data has to be'): self.fig.add_data(data) def test_add_data_wrong_biomarker(self): # Biomarker does not exist in prediction dataframe biomarker = 'Does not exist' with self.assertRaisesRegex(ValueError, 'The biomarker could not be'): self.fig.add_data(self.data, biomarker) # Biomarker does not exist in measurement dataframe data = self.data.copy() data['Biomarker'] = 'Does not exist' biomarker = 'Does not exist' with self.assertRaisesRegex(ValueError, 'The biomarker <Does not'): self.fig.add_data(data, biomarker) def test_add_data_wrong_individual(self): individual = 'does not exist' self.assertRaisesRegex( ValueError, 'The ID <does not exist> does not exist.', self.fig.add_data, self.data, individual=individual) def test_add_data_wrong_time_key(self): # Rename time key data = self.data.rename(columns={'Time': 'SOME NON-STANDARD KEY'}) self.assertRaisesRegex( ValueError, 'Data does not have the key <Time>.', self.fig.add_data, data) def test_add_data_wrong_biom_key(self): # Rename biomarker key data = self.data.rename(columns={'Biomarker': 'SOME NON-STANDARD KEY'}) self.assertRaisesRegex( ValueError, 'Data does not have the key <Biomarker>.', self.fig.add_data, data) def test_add_data_wrong_sample_key(self): # Rename measurement key data = self.data.rename( columns={'Sample': 'SOME NON-STANDARD KEY'}) self.assertRaisesRegex( ValueError, 'Data does not have the key <Sample>.', self.fig.add_data, data) def test_add_data_time_key_mapping(self): # Rename time key data = self.data.rename(columns={'Time': 'SOME NON-STANDARD KEY'}) # Test that it works with correct mapping self.fig.add_data( data, time_key='SOME NON-STANDARD KEY') # Test that it fails with wrong mapping with self.assertRaisesRegex( ValueError, 'Data does not have the key <SOME WRONG KEY>.'): self.fig.add_data( data, time_key='SOME WRONG KEY') def test_add_data_biom_key_mapping(self): # Rename biomarker key data = self.data.rename(columns={'Biomarker': 'SOME NON-STANDARD KEY'}) # Test that it works with correct mapping self.fig.add_data( data, biom_key='SOME NON-STANDARD KEY') # Test that it fails with wrong mapping with self.assertRaisesRegex( ValueError, 'Data does not have the key <SOME WRONG KEY>.'): self.fig.add_data( data, biom_key='SOME WRONG KEY') def test_add_data_sample_key_mapping(self): # Rename measurement key data = self.data.rename( columns={'Sample': 'SOME NON-STANDARD KEY'}) # Test that it works with correct mapping self.fig.add_data( data, sample_key='SOME NON-STANDARD KEY') # Test that it fails with wrong mapping with self.assertRaisesRegex( ValueError, 'Data does not have the key <SOME WRONG KEY>.'): self.fig.add_data( data, sample_key='SOME WRONG KEY') def test_add_data_show_relative(self): self.fig.add_data(self.data, show_relative=True) def test_data_wrong_time_points(self): # Not all measured time points can be found in the prediction # dataframe data = self.data.copy() data['Time'] = 1 with self.assertRaisesRegex( ValueError, 'The prediction dataframe is not'): self.fig.add_data(data) def test_add_data_individual(self): # Select an individual self.fig.add_data(self.data, individual=40) if __name__ == '__main__': unittest.main()
[ "unittest.main", "chi.plots.ResidualPlot", "chi.library.DataLibrary", "numpy.ones" ]
[((5779, 5794), 'unittest.main', 'unittest.main', ([], {}), '()\n', (5792, 5794), False, 'import unittest\n'), ((693, 729), 'chi.plots.ResidualPlot', 'plots.ResidualPlot', (['cls.measurements'], {}), '(cls.measurements)\n', (711, 729), False, 'from chi import plots\n'), ((828, 850), 'numpy.ones', 'np.ones', ([], {'shape': '(10, 4)'}), '(shape=(10, 4))\n', (835, 850), True, 'import numpy as np\n'), ((1767, 1789), 'numpy.ones', 'np.ones', ([], {'shape': '(10, 4)'}), '(shape=(10, 4))\n', (1774, 1789), True, 'import numpy as np\n'), ((937, 969), 'chi.plots.ResidualPlot', 'plots.ResidualPlot', (['measurements'], {}), '(measurements)\n', (955, 969), False, 'from chi import plots\n'), ((1085, 1138), 'chi.plots.ResidualPlot', 'plots.ResidualPlot', (['self.measurements'], {'id_key': '"""Wrong"""'}), "(self.measurements, id_key='Wrong')\n", (1103, 1138), False, 'from chi import plots\n'), ((1256, 1311), 'chi.plots.ResidualPlot', 'plots.ResidualPlot', (['self.measurements'], {'time_key': '"""Wrong"""'}), "(self.measurements, time_key='Wrong')\n", (1274, 1311), False, 'from chi import plots\n'), ((1434, 1489), 'chi.plots.ResidualPlot', 'plots.ResidualPlot', (['self.measurements'], {'biom_key': '"""Wrong"""'}), "(self.measurements, biom_key='Wrong')\n", (1452, 1489), False, 'from chi import plots\n'), ((1614, 1669), 'chi.plots.ResidualPlot', 'plots.ResidualPlot', (['self.measurements'], {'meas_key': '"""Wrong"""'}), "(self.measurements, meas_key='Wrong')\n", (1632, 1669), False, 'from chi import plots\n'), ((512, 525), 'chi.library.DataLibrary', 'DataLibrary', ([], {}), '()\n', (523, 525), False, 'from chi.library import DataLibrary\n')]
""" This data implement model's metric :paper author: hxq :code author: hxq :code convert: shy """ import math from sklearn import metrics import numpy as np import bottleneck as bn def ndcg_binary_at_k_batch(x_pred, heldout_batch, k=100): """ normalized discounted cumulative gain@k for binary relevance ASSUMPTIONS: all the 0's in heldout_data indicate 0 relevance """ batch_users = x_pred.shape[0] idx_topk_part = bn.argpartition(-x_pred, k, axis=1) # shape和x_pred一样 topk_part = x_pred[np.arange(batch_users)[:, np.newaxis], idx_topk_part[:, :k]] # 所有用户的topk item同时挑出来 idx_part = np.argsort(-topk_part, axis=1) # X_pred[np.arange(batch_users)[:, np.newaxis], idx_topk] is the sorted # topk predicted score idx_topk = idx_topk_part[np.arange(batch_users)[:, np.newaxis], idx_part] # build the discount template tset_pre = 1. / np.log2(np.arange(2, k + 2)) dcg = (heldout_batch[np.arange(batch_users)[:, np.newaxis], idx_topk].toarray() * tset_pre).sum(axis=1) idcg = np.array([(tset_pre[:min(n, k)]).sum() for n in heldout_batch.getnnz(axis=1)]) ndcg = dcg / idcg ndcg[np.isnan(ndcg)] = 0 return ndcg def precision_recall_at_k_batch(x_pred, heldout_batch, k=100, observe_fair=False, attr_indicator_list=None): """[summary] Args: x_pred ([type]): [description] heldout_batch ([type]): [description] k (int, optional): [description]. Defaults to 100. observe_fair (bool, optional): [description]. Defaults to False. attr_indicator_list ([type], optional): [description]. Defaults to None. Returns: [type]: [description] """ print(observe_fair, attr_indicator_list) batch_users = x_pred.shape[0] idx = bn.argpartition(-x_pred, k, axis=1) x_pred_binary = np.zeros_like(x_pred, dtype=bool) x_pred_binary[np.arange(batch_users)[:, np.newaxis], idx[:, :k]] = True x_true_binary = (heldout_batch > 0).toarray() # ranked_list = x_pred_binary tmp = (np.logical_and(x_true_binary, x_pred_binary).sum(axis=1)).astype( np.float32) recall = tmp / x_true_binary.sum(axis=1) precision = tmp / k precision[np.isnan(precision)] = 0 recall[np.isnan(recall)] = 0 f1_recall = 2 * recall * precision / (precision + recall) f1_recall[np.isnan(f1_recall)] = 0 return precision, recall, f1_recall def update_threshold(x_pred, id_onehots_ph, threshold_ph, k=100): """[summary] Args: x_pred ([type]): [description] id_onehots_ph ([type]): [description] threshold_ph ([type]): [description] k (int, optional): [description]. Defaults to 100. Returns: [type]: [description] """ batch_users = x_pred.shape[0] idx = bn.argpartition(-x_pred, k, axis=1) #epsion = 1e-10 #threshold_ph_batch = x_pred[:, idx[:, k]]-epsion #print('shape(threshold_ph_batch)', threshold_ph_batch.shape) threshold_ph[np.nonzero(id_onehots_ph)[1]] = \ x_pred[np.arange(batch_users), idx[:, k]].reshape(-1, 1) #threshold_ph = np.dot(threshold_ph.T, id_onehots_ph.toarray()) return threshold_ph def average_precision(ranked_list, ground_truth): """Compute the average precision (AP) of a list of ranked items """ hits = 0 sum_precs = 0 for index in range(len(ranked_list)): if ranked_list[index] in ground_truth: hits += 1 sum_precs += hits / (index + 1.0) if hits > 0: return sum_precs / len(ground_truth) return 0 def hit(gt_items, pred_items): # HR为所有用户的hits/所有用户的grounf truth总个数 """[summary] Args: gt_items ([type]): [description] pred_items ([type]): [description] Returns: [type]: [description] """ count = 0 for item in pred_items: if item in gt_items: count += 1 return count def auc(label, prob): # prob 为预测为正的概率 """[summary] Args: label ([type]): [description] prob ([type]): [description] Returns: [type]: [description] """ precision, recall, thresholds = metrics.precision_recall_curve(label, prob) print(thresholds) area = metrics.auc(recall, precision) return area # sklearn # precision, recall, _thresholds = metrics.precision_recall_curve(label, prob) # area = metrics.auc(recall, precision) # return area # area = metrics.roc_auc_score(label, prob) # return area def hit_precision_recall_ndcg_k(train_set_batch, test_set_batch, pred_scores_batch, max_train_count, k=20, ranked_tag=False, vad_set_batch=None): """[summary] Args: train_set_batch ([type]): [description] test_set_batch ([type]): [description] pred_scores_batch ([type]): [description] max_train_count ([type]): [description] k (int, optional): [description]. Defaults to 20. ranked_tag (bool, optional): [description]. Defaults to False. vad_set_batch ([type], optional): [description]. Defaults to None. Returns: [type]: [description] """ recall_k, precision_k, ndcg_k, hits_list = [], [], [], [] if not ranked_tag: batch_users = pred_scores_batch.shape[0] idx_topk_part = bn.argpartition(-pred_scores_batch, k+max_train_count, axis=1) topk_part = pred_scores_batch[np.arange(batch_users)[:, np.newaxis], idx_topk_part[:, :(k+max_train_count)]] idx_part = np.argsort(-topk_part, axis=1) top_items = idx_topk_part[np.arange(batch_users)[:, np.newaxis], idx_part] else: top_items = pred_scores_batch if vad_set_batch is None: for train_set, test_set, ranked in zip(train_set_batch, test_set_batch, top_items): n_k = k if len(test_set) > k else len(test_set) # n_k = min(k, len(test_k)) n_idcg, n_dcg = 0, 0 for pos in range(n_k): n_idcg += 1.0 / math.log(pos + 2, 2) tops_sub_train = [] n_top_items = 0 for val in ranked: if val not in train_set: tops_sub_train.append(val) n_top_items += 1 if n_top_items >= k: # 控制topK个item是从用户没交互过的商品中选的 break hits_set = [(idx, item_id) for idx, item_id in enumerate(tops_sub_train) if item_id in test_set] cnt_hits = len(hits_set) for idx in range(cnt_hits): n_dcg += 1.0 / math.log(hits_set[idx][0] + 2, 2) precision_k.append(float(cnt_hits / k)) recall_k.append(float(cnt_hits / len(test_set))) ndcg_k.append(float(n_dcg / n_idcg)) hits_list.append(cnt_hits) else: hits_list, precision_k, recall_k, ndcg_k = \ calc_second(train_set_batch, test_set_batch, top_items, vad_set_batch, k) return hits_list, precision_k, recall_k, ndcg_k def calc_second(train_set_batch, test_set_batch, top_items, vad_set_batch, k): """[summary] Args: train_set_batch ([type]): [description] test_set_batch ([type]): [description] top_items ([type]): [description] vad_set_batch ([type]): [description] k ([type]): [description] Returns: [type]: [description] """ recall_k, precision_k, ndcg_k, hits_list = [], [], [], [] for train_set, test_set, ranked, vad_set in \ zip(train_set_batch, test_set_batch, top_items, vad_set_batch): n_k = k if len(test_set) > k else len(test_set) # n_k = min(k, len(test_k)) n_idcg, n_dcg = 0, 0 for pos in range(n_k): n_idcg += 1.0 / math.log(pos + 2, 2) tops_sub_train = [] n_top_items = 0 for val in ranked: if val not in train_set and val not in vad_set: tops_sub_train.append(val) n_top_items += 1 if n_top_items >= k: # 控制topK个item是从用户没交互过的商品中选的 break hits_set = [(idx, item_id) for idx, item_id in \ enumerate(tops_sub_train) if item_id in test_set] cnt_hits = len(hits_set) for idx in range(cnt_hits): n_dcg += 1.0 /math.log(hits_set[idx][0] + 2, 2) precision_k.append(float(cnt_hits / k)) recall_k.append(float(cnt_hits / len(test_set))) ndcg_k.append(float(n_dcg / n_idcg)) hits_list.append(cnt_hits) return hits_list, precision_k, recall_k, ndcg_k
[ "numpy.logical_and", "numpy.arange", "sklearn.metrics.auc", "sklearn.metrics.precision_recall_curve", "math.log", "numpy.argsort", "numpy.isnan", "numpy.nonzero", "numpy.zeros_like", "bottleneck.argpartition" ]
[((444, 479), 'bottleneck.argpartition', 'bn.argpartition', (['(-x_pred)', 'k'], {'axis': '(1)'}), '(-x_pred, k, axis=1)\n', (459, 479), True, 'import bottleneck as bn\n'), ((641, 671), 'numpy.argsort', 'np.argsort', (['(-topk_part)'], {'axis': '(1)'}), '(-topk_part, axis=1)\n', (651, 671), True, 'import numpy as np\n'), ((1859, 1894), 'bottleneck.argpartition', 'bn.argpartition', (['(-x_pred)', 'k'], {'axis': '(1)'}), '(-x_pred, k, axis=1)\n', (1874, 1894), True, 'import bottleneck as bn\n'), ((1915, 1948), 'numpy.zeros_like', 'np.zeros_like', (['x_pred'], {'dtype': 'bool'}), '(x_pred, dtype=bool)\n', (1928, 1948), True, 'import numpy as np\n'), ((2872, 2907), 'bottleneck.argpartition', 'bn.argpartition', (['(-x_pred)', 'k'], {'axis': '(1)'}), '(-x_pred, k, axis=1)\n', (2887, 2907), True, 'import bottleneck as bn\n'), ((4219, 4262), 'sklearn.metrics.precision_recall_curve', 'metrics.precision_recall_curve', (['label', 'prob'], {}), '(label, prob)\n', (4249, 4262), False, 'from sklearn import metrics\n'), ((4296, 4326), 'sklearn.metrics.auc', 'metrics.auc', (['recall', 'precision'], {}), '(recall, precision)\n', (4307, 4326), False, 'from sklearn import metrics\n'), ((1212, 1226), 'numpy.isnan', 'np.isnan', (['ndcg'], {}), '(ndcg)\n', (1220, 1226), True, 'import numpy as np\n'), ((2292, 2311), 'numpy.isnan', 'np.isnan', (['precision'], {}), '(precision)\n', (2300, 2311), True, 'import numpy as np\n'), ((2328, 2344), 'numpy.isnan', 'np.isnan', (['recall'], {}), '(recall)\n', (2336, 2344), True, 'import numpy as np\n'), ((2426, 2445), 'numpy.isnan', 'np.isnan', (['f1_recall'], {}), '(f1_recall)\n', (2434, 2445), True, 'import numpy as np\n'), ((5393, 5457), 'bottleneck.argpartition', 'bn.argpartition', (['(-pred_scores_batch)', '(k + max_train_count)'], {'axis': '(1)'}), '(-pred_scores_batch, k + max_train_count, axis=1)\n', (5408, 5457), True, 'import bottleneck as bn\n'), ((5630, 5660), 'numpy.argsort', 'np.argsort', (['(-topk_part)'], {'axis': '(1)'}), '(-topk_part, axis=1)\n', (5640, 5660), True, 'import numpy as np\n'), ((915, 934), 'numpy.arange', 'np.arange', (['(2)', '(k + 2)'], {}), '(2, k + 2)\n', (924, 934), True, 'import numpy as np\n'), ((3065, 3090), 'numpy.nonzero', 'np.nonzero', (['id_onehots_ph'], {}), '(id_onehots_ph)\n', (3075, 3090), True, 'import numpy as np\n'), ((520, 542), 'numpy.arange', 'np.arange', (['batch_users'], {}), '(batch_users)\n', (529, 542), True, 'import numpy as np\n'), ((804, 826), 'numpy.arange', 'np.arange', (['batch_users'], {}), '(batch_users)\n', (813, 826), True, 'import numpy as np\n'), ((1967, 1989), 'numpy.arange', 'np.arange', (['batch_users'], {}), '(batch_users)\n', (1976, 1989), True, 'import numpy as np\n'), ((7822, 7842), 'math.log', 'math.log', (['(pos + 2)', '(2)'], {}), '(pos + 2, 2)\n', (7830, 7842), False, 'import math\n'), ((8363, 8396), 'math.log', 'math.log', (['(hits_set[idx][0] + 2)', '(2)'], {}), '(hits_set[idx][0] + 2, 2)\n', (8371, 8396), False, 'import math\n'), ((2123, 2167), 'numpy.logical_and', 'np.logical_and', (['x_true_binary', 'x_pred_binary'], {}), '(x_true_binary, x_pred_binary)\n', (2137, 2167), True, 'import numpy as np\n'), ((3110, 3132), 'numpy.arange', 'np.arange', (['batch_users'], {}), '(batch_users)\n', (3119, 3132), True, 'import numpy as np\n'), ((5494, 5516), 'numpy.arange', 'np.arange', (['batch_users'], {}), '(batch_users)\n', (5503, 5516), True, 'import numpy as np\n'), ((5695, 5717), 'numpy.arange', 'np.arange', (['batch_users'], {}), '(batch_users)\n', (5704, 5717), True, 'import numpy as np\n'), ((6105, 6125), 'math.log', 'math.log', (['(pos + 2)', '(2)'], {}), '(pos + 2, 2)\n', (6113, 6125), False, 'import math\n'), ((6650, 6683), 'math.log', 'math.log', (['(hits_set[idx][0] + 2)', '(2)'], {}), '(hits_set[idx][0] + 2, 2)\n', (6658, 6683), False, 'import math\n'), ((962, 984), 'numpy.arange', 'np.arange', (['batch_users'], {}), '(batch_users)\n', (971, 984), True, 'import numpy as np\n')]
""" Function for working with patches from tensors. See the :doc:`tutorials/patches` tutorial for more details. """ from typing import Iterable import numpy as np from .shape_ops import crop_to_box from .axes import broadcast_to_axes, fill_by_indices, AxesLike from .box import make_box_, Box from dpipe.itertools import zip_equal, peek from .shape_utils import shape_after_full_convolution from .utils import build_slices __all__ = 'get_boxes', 'divide', 'combine' def get_boxes(shape: AxesLike, box_size: AxesLike, stride: AxesLike, axes: AxesLike = None, valid: bool = True) -> Iterable[Box]: """ Yield boxes appropriate for a tensor of shape ``shape`` in a convolution-like fashion. Parameters ---------- shape the input tensor's shape. box_size axes axes along which the slices will be taken. stride the stride (step-size) of the slice. valid whether boxes of size smaller than ``box_size`` should be left out. References ---------- See the :doc:`tutorials/patches` tutorial for more details. """ final_shape = shape_after_full_convolution(shape, box_size, axes, stride, valid=valid) box_size, stride = np.broadcast_arrays(box_size, stride) full_box = fill_by_indices(shape, box_size, axes) full_stride = fill_by_indices(np.ones_like(shape), stride, axes) for start in np.ndindex(*final_shape): start = np.asarray(start) * full_stride yield make_box_([start, np.minimum(start + full_box, shape)]) def divide(x: np.ndarray, patch_size: AxesLike, stride: AxesLike, axes: AxesLike = None, valid: bool = False) -> Iterable[np.ndarray]: """ A convolution-like approach to generating patches from a tensor. Parameters ---------- x patch_size axes dimensions along which the slices will be taken. stride the stride (step-size) of the slice. valid whether patches of size smaller than ``patch_size`` should be left out. References ---------- See the :doc:`tutorials/patches` tutorial for more details. """ for box in get_boxes(x.shape, patch_size, stride, axes, valid=valid): yield crop_to_box(x, box) def combine(patches: Iterable[np.ndarray], output_shape: AxesLike, stride: AxesLike, axes: AxesLike = None, valid: bool = False) -> np.ndarray: """ Build a tensor of shape ``output_shape`` from ``patches`` obtained in a convolution-like approach with corresponding parameters. The overlapping parts are averaged. References ---------- See the :doc:`tutorials/patches` tutorial for more details. """ axes, stride = broadcast_to_axes(axes, stride) patch, patches = peek(patches) patch_size = np.array(patch.shape)[list(axes)] if len(np.atleast_1d(output_shape)) != patch.ndim: output_shape = fill_by_indices(patch.shape, output_shape, axes) dtype = patch.dtype if not np.issubdtype(dtype, np.floating): dtype = float result = np.zeros(output_shape, dtype) counts = np.zeros(output_shape, int) for box, patch in zip_equal(get_boxes(output_shape, patch_size, stride, axes, valid), patches): slc = build_slices(*box) result[slc] += patch counts[slc] += 1 np.true_divide(result, counts, out=result, where=counts > 0) return result
[ "numpy.ones_like", "numpy.minimum", "numpy.asarray", "numpy.ndindex", "numpy.array", "numpy.zeros", "numpy.issubdtype", "numpy.true_divide", "dpipe.itertools.peek", "numpy.broadcast_arrays", "numpy.atleast_1d" ]
[((1221, 1258), 'numpy.broadcast_arrays', 'np.broadcast_arrays', (['box_size', 'stride'], {}), '(box_size, stride)\n', (1240, 1258), True, 'import numpy as np\n'), ((1401, 1425), 'numpy.ndindex', 'np.ndindex', (['*final_shape'], {}), '(*final_shape)\n', (1411, 1425), True, 'import numpy as np\n'), ((2759, 2772), 'dpipe.itertools.peek', 'peek', (['patches'], {}), '(patches)\n', (2763, 2772), False, 'from dpipe.itertools import zip_equal, peek\n'), ((3058, 3087), 'numpy.zeros', 'np.zeros', (['output_shape', 'dtype'], {}), '(output_shape, dtype)\n', (3066, 3087), True, 'import numpy as np\n'), ((3101, 3128), 'numpy.zeros', 'np.zeros', (['output_shape', 'int'], {}), '(output_shape, int)\n', (3109, 3128), True, 'import numpy as np\n'), ((3321, 3381), 'numpy.true_divide', 'np.true_divide', (['result', 'counts'], {'out': 'result', 'where': '(counts > 0)'}), '(result, counts, out=result, where=counts > 0)\n', (3335, 3381), True, 'import numpy as np\n'), ((1348, 1367), 'numpy.ones_like', 'np.ones_like', (['shape'], {}), '(shape)\n', (1360, 1367), True, 'import numpy as np\n'), ((2790, 2811), 'numpy.array', 'np.array', (['patch.shape'], {}), '(patch.shape)\n', (2798, 2811), True, 'import numpy as np\n'), ((2987, 3020), 'numpy.issubdtype', 'np.issubdtype', (['dtype', 'np.floating'], {}), '(dtype, np.floating)\n', (3000, 3020), True, 'import numpy as np\n'), ((1443, 1460), 'numpy.asarray', 'np.asarray', (['start'], {}), '(start)\n', (1453, 1460), True, 'import numpy as np\n'), ((2835, 2862), 'numpy.atleast_1d', 'np.atleast_1d', (['output_shape'], {}), '(output_shape)\n', (2848, 2862), True, 'import numpy as np\n'), ((1507, 1542), 'numpy.minimum', 'np.minimum', (['(start + full_box)', 'shape'], {}), '(start + full_box, shape)\n', (1517, 1542), True, 'import numpy as np\n')]
import glob from pylab import * import brewer2mpl import numpy as np import sys import math import gzip import matplotlib.gridspec as gridspec from collections import defaultdict from matplotlib import pyplot as plt # brewer2mpl.get_map args: set name set type number of colors bmap = brewer2mpl.get_map('Set2', 'qualitative', 7) colors = bmap.mpl_colors params = { 'axes.labelsize': 8, 'legend.fontsize': 10, 'xtick.labelsize': 10, 'ytick.labelsize': 10, 'text.usetex': False, 'figure.figsize': [6, 8] } rcParams.update(params) def customize_axis(ax): ax.spines['top'].set_visible(False) ax.spines['right'].set_visible(False) ax.spines['left'].set_visible(False) ax.get_xaxis().tick_bottom() ax.tick_params(axis='y', length=0) #ax.get_yaxis().tick_left() # offset the spines for spine in ax.spines.values(): spine.set_position(('outward', 5)) # put the grid behind ax.set_axisbelow(True) ax.grid(axis='y', color="0.9", linestyle='--', linewidth=1) fig = figure(frameon=False) # no frame #plt.box(False) #plt.ticklabel_format(axis='both', style='sci', scilimits=(-2,2)) ax1 = fig.add_subplot(311) k = 0 for i in sys.argv[1:]: data = np.loadtxt(i) ax1.plot(data[:,0], data[:, 1], '-', linewidth=2, color=colors[k], label=i) k += 1 ax1.set_title('Coverage') customize_axis(ax1) ax2 = fig.add_subplot(312) k = 0 for i in sys.argv[1:]: data = np.loadtxt(i) ax2.plot(data[:,0], data[:, 3], '-', linewidth=2, color=colors[k], label=i) k += 1 ax2.set_title('Mean fitness') customize_axis(ax2) ax3 = fig.add_subplot(313) ax3.grid(axis='y', color="0.9", linestyle='--', linewidth=1) k = 0 for i in sys.argv[1:]: data = np.loadtxt(i) ax3.plot(data[:,0], data[:, 2], '-', linewidth=2, color=colors[k], label=i) k += 1 ax3.set_title('Max fitness') customize_axis(ax3) legend = ax1.legend(loc=4)#bbox_to_anchor=(0., 1.02, 1., .102), loc=3, ncol=(3)) frame = legend.get_frame() frame.set_facecolor('0.9') frame.set_edgecolor('1.0') fig.tight_layout() fig.savefig('progress.pdf') fig.savefig('progress.svg')
[ "numpy.loadtxt", "brewer2mpl.get_map" ]
[((290, 334), 'brewer2mpl.get_map', 'brewer2mpl.get_map', (['"""Set2"""', '"""qualitative"""', '(7)'], {}), "('Set2', 'qualitative', 7)\n", (308, 334), False, 'import brewer2mpl\n'), ((1225, 1238), 'numpy.loadtxt', 'np.loadtxt', (['i'], {}), '(i)\n', (1235, 1238), True, 'import numpy as np\n'), ((1444, 1457), 'numpy.loadtxt', 'np.loadtxt', (['i'], {}), '(i)\n', (1454, 1457), True, 'import numpy as np\n'), ((1730, 1743), 'numpy.loadtxt', 'np.loadtxt', (['i'], {}), '(i)\n', (1740, 1743), True, 'import numpy as np\n')]
import numpy as np import argparse import config import os import datetime import sys import tensorflow.keras as keras from tensorflow.keras.layers import Input, Conv2D, Flatten, Dense, Conv2DTranspose, Lambda, Reshape, Layer from tensorflow.keras.models import Model from tensorflow.keras.optimizers import Adam from tensorflow.keras import backend as K import tensorflow as tf DIR_NAME = './data/rollout/' SCREEN_SIZE_X = 64 SCREEN_SIZE_Y = 64 batch_size = 10 IM_DIM = 64 DEPTH = 32 LATENT_DEPTH = 512 K_SIZE = 5 def sampling(args): mean, logsigma = args epsilon = keras.backend.random_normal(shape=keras.backend.shape(mean)) return mean + tf.exp(logsigma / 2) * epsilon def encoder(): input_E = keras.layers.Input(shape=(IM_DIM, IM_DIM, 3)) X = keras.layers.Conv2D(filters=DEPTH*2, kernel_size=K_SIZE, strides=2, padding='same')(input_E) X = keras.layers.BatchNormalization()(X) X = keras.layers.LeakyReLU(alpha=0.2)(X) X = keras.layers.Conv2D(filters=DEPTH*4, kernel_size=K_SIZE, strides=2, padding='same')(X) X = keras.layers.BatchNormalization()(X) X = keras.layers.LeakyReLU(alpha=0.2)(X) X = keras.layers.Conv2D(filters=DEPTH*8, kernel_size=K_SIZE, strides=2, padding='same')(X) X = keras.layers.BatchNormalization()(X) X = keras.layers.LeakyReLU(alpha=0.2)(X) X = keras.layers.Flatten()(X) X = keras.layers.Dense(LATENT_DEPTH)(X) X = keras.layers.BatchNormalization()(X) X = keras.layers.LeakyReLU(alpha=0.2)(X) mean = keras.layers.Dense(LATENT_DEPTH,activation="tanh")(X) logsigma = keras.layers.Dense(LATENT_DEPTH,activation="tanh")(X) latent = keras.layers.Lambda(sampling, output_shape=(LATENT_DEPTH,))([mean, logsigma]) kl_loss = 1 + logsigma - keras.backend.square(mean) - keras.backend.exp(logsigma) kl_loss = keras.backend.mean(kl_loss, axis=-1) kl_loss *= -0.5 return keras.models.Model(input_E, [latent,kl_loss]) def generator(): input_G = keras.layers.Input(shape=(LATENT_DEPTH,)) X = keras.layers.Dense(8*8*DEPTH*8)(input_G) X = keras.layers.BatchNormalization()(X) X = keras.layers.LeakyReLU(alpha=0.2)(X) X = keras.layers.Reshape((8, 8, DEPTH * 8))(X) X = keras.layers.Conv2DTranspose(filters=DEPTH*8, kernel_size=K_SIZE, strides=2, padding='same')(X) X = keras.layers.BatchNormalization()(X) X = keras.layers.LeakyReLU(alpha=0.2)(X) X = keras.layers.Conv2DTranspose(filters=DEPTH*4, kernel_size=K_SIZE, strides=2, padding='same')(X) X = keras.layers.BatchNormalization()(X) X = keras.layers.LeakyReLU(alpha=0.2)(X) X = keras.layers.Conv2DTranspose(filters=DEPTH, kernel_size=K_SIZE, strides=2, padding='same')(X) X = keras.layers.BatchNormalization()(X) X = keras.layers.LeakyReLU(alpha=0.2)(X) X = keras.layers.Conv2D(filters=3, kernel_size=K_SIZE, padding='same')(X) X = keras.layers.Activation('sigmoid')(X) return keras.models.Model(input_G, X) def discriminator(): input_D = keras.layers.Input(shape=(IM_DIM, IM_DIM, 3)) X = keras.layers.Conv2D(filters=DEPTH, kernel_size=K_SIZE, strides=2, padding='same')(input_D) X = keras.layers.LeakyReLU(alpha=0.2)(X) X = keras.layers.Conv2D(filters=DEPTH*4, kernel_size=K_SIZE, strides=2, padding='same')(input_D) X = keras.layers.LeakyReLU(alpha=0.2)(X) X = keras.layers.BatchNormalization()(X) X = keras.layers.Conv2D(filters=DEPTH*8, kernel_size=K_SIZE, strides=2, padding='same')(X) X = keras.layers.BatchNormalization()(X) X = keras.layers.LeakyReLU(alpha=0.2)(X) X = keras.layers.Conv2D(filters=DEPTH*8, kernel_size=K_SIZE, padding='same')(X) inner_output = keras.layers.Flatten()(X) X = keras.layers.BatchNormalization()(X) X = keras.layers.LeakyReLU(alpha=0.2)(X) X = keras.layers.Flatten()(X) X = keras.layers.Dense(DEPTH*8)(X) X = keras.layers.BatchNormalization()(X) X = keras.layers.LeakyReLU(alpha=0.2)(X) output = keras.layers.Dense(1)(X) return keras.models.Model(input_D, [output, inner_output]) def import_data(N, M): filelist = os.listdir(DIR_NAME) filelist = [x for x in filelist if x != '.DS_Store'] filelist.sort() length_filelist = len(filelist) if length_filelist > N: filelist = filelist[:N] if length_filelist < N: N = length_filelist data = np.zeros((M*N, SCREEN_SIZE_X, SCREEN_SIZE_Y, 3), dtype=np.float32) idx = 0 file_count = 0 for file in filelist: try: new_data = np.load(DIR_NAME + file)['obs'] data[idx:(idx + M), :, :, :] = new_data idx = idx + M file_count += 1 if file_count%50==0: print('Imported {} / {} ::: Current data size = {} observations'.format(file_count, N, idx)) except Exception as e: print(e) print('Skipped {}...'.format(file)) print('Imported {} / {} ::: Current data size = {} observations'.format(file_count, N, idx)) return data, N E = encoder() G = generator() D = discriminator() lr=0.001 #lr=0.0001 E_opt = keras.optimizers.Adam(lr=lr) G_opt = keras.optimizers.Adam(lr=lr) D_opt = keras.optimizers.Adam(lr=lr) inner_loss_coef = 1 normal_coef = 0.1 kl_coef = 0.5 @tf.function def train_step_vaegan(x): lattent_r = tf.random.normal((100, LATENT_DEPTH)) with tf.GradientTape(persistent=True) as tape : lattent,kl_loss = E(x) fake = G(lattent) dis_fake,dis_inner_fake = D(fake) dis_fake_r,_ = D(G(lattent_r)) dis_true,dis_inner_true = D(x) vae_inner = dis_inner_fake-dis_inner_true vae_inner = vae_inner*vae_inner mean,var = tf.nn.moments(E(x)[0], axes=0) var_to_one = var - 1 normal_loss = tf.reduce_mean(mean*mean) + tf.reduce_mean(var_to_one*var_to_one) kl_loss = tf.reduce_mean(kl_loss) vae_diff_loss = tf.reduce_mean(vae_inner) f_dis_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(tf.zeros_like(dis_fake), dis_fake)) r_dis_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(tf.zeros_like(dis_fake_r), dis_fake_r)) t_dis_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(tf.ones_like(dis_true), dis_true)) gan_loss = (0.5*t_dis_loss + 0.25*f_dis_loss + 0.25*r_dis_loss) vae_loss = tf.reduce_mean(tf.abs(x-fake)) E_loss = vae_diff_loss + kl_coef*kl_loss + normal_coef*normal_loss G_loss = inner_loss_coef*vae_diff_loss - gan_loss D_loss = gan_loss E_grad = tape.gradient(E_loss,E.trainable_variables) G_grad = tape.gradient(G_loss,G.trainable_variables) D_grad = tape.gradient(D_loss,D.trainable_variables) del tape E_opt.apply_gradients(zip(E_grad, E.trainable_variables)) G_opt.apply_gradients(zip(G_grad, G.trainable_variables)) D_opt.apply_gradients(zip(D_grad, D.trainable_variables)) return [gan_loss, vae_loss, f_dis_loss, r_dis_loss, t_dis_loss, vae_diff_loss, E_loss, D_loss, kl_loss, normal_loss] def main(args): new_model = args.new_model N = int(args.N) M = int(args.time_steps) epochs = int(args.epochs) try: data, N = import_data(N, M) except: print('NO DATA FOUND') raise if not new_model: try: D.load_weights("./saved-models/D_training_.h5") E.load_weights("./saved-models/E_training_.h5") G.load_weights("./saved-models/G_training_.h5") except: print("Either set --new_model or ensure ./vae/weights.h5 exists") raise print('DATA SHAPE = {}'.format(data.shape)) step = 0 max_step = 100 log_freq = 1 metrics_names = ["gan_loss", "vae_loss", "fake_dis_loss", "r_dis_loss", "t_dis_loss", "vae_inner_loss", "E_loss", "D_loss", "kl_loss", "normal_loss"] metrics = [] for m in metrics_names : metrics.append(tf.keras.metrics.Mean('m', dtype=tf.float32)) def save_model(): D.save('saved-models/D_training_' + '.h5') G.save('saved-models/G_training_' + '.h5') E.save('saved-models/E_training_' + '.h5') def print_metrics(): s = "" for name,metric in zip(metrics_names,metrics) : s+= " " + name + " " + str(np.around(metric.result().numpy(), 3)) print(f"\rStep : " + str(step) + " " + s, end="", flush=True) for metric in metrics : metric.reset_states() for i in range(2000,5001,100): step+=1 if not i % log_freq : print_metrics() results = train_step_vaegan(data[i-100:i]) for metric,result in zip(metrics, results) : metric(result) save_model() if __name__ == "__main__": parser = argparse.ArgumentParser(description=('Train VAE')) parser.add_argument('--N',default = 10000, help='number of episodes to use to train') parser.add_argument('--new_model', action='store_true', help='start a new model from scratch?') parser.add_argument('--time_steps', type=int, default=300, help='how many timesteps at start of episode?') parser.add_argument('--epochs', default = 10, help='number of epochs to train for') args = parser.parse_args() main(args)
[ "tensorflow.keras.layers.BatchNormalization", "tensorflow.GradientTape", "tensorflow.keras.layers.Dense", "tensorflow.ones_like", "tensorflow.reduce_mean", "tensorflow.keras.backend.shape", "tensorflow.keras.layers.Input", "tensorflow.random.normal", "os.listdir", "tensorflow.keras.backend.mean", ...
[((5084, 5112), 'tensorflow.keras.optimizers.Adam', 'keras.optimizers.Adam', ([], {'lr': 'lr'}), '(lr=lr)\n', (5105, 5112), True, 'import tensorflow.keras as keras\n'), ((5121, 5149), 'tensorflow.keras.optimizers.Adam', 'keras.optimizers.Adam', ([], {'lr': 'lr'}), '(lr=lr)\n', (5142, 5149), True, 'import tensorflow.keras as keras\n'), ((5158, 5186), 'tensorflow.keras.optimizers.Adam', 'keras.optimizers.Adam', ([], {'lr': 'lr'}), '(lr=lr)\n', (5179, 5186), True, 'import tensorflow.keras as keras\n'), ((720, 765), 'tensorflow.keras.layers.Input', 'keras.layers.Input', ([], {'shape': '(IM_DIM, IM_DIM, 3)'}), '(shape=(IM_DIM, IM_DIM, 3))\n', (738, 765), True, 'import tensorflow.keras as keras\n'), ((1846, 1882), 'tensorflow.keras.backend.mean', 'keras.backend.mean', (['kl_loss'], {'axis': '(-1)'}), '(kl_loss, axis=-1)\n', (1864, 1882), True, 'import tensorflow.keras as keras\n'), ((1919, 1965), 'tensorflow.keras.models.Model', 'keras.models.Model', (['input_E', '[latent, kl_loss]'], {}), '(input_E, [latent, kl_loss])\n', (1937, 1965), True, 'import tensorflow.keras as keras\n'), ((1997, 2038), 'tensorflow.keras.layers.Input', 'keras.layers.Input', ([], {'shape': '(LATENT_DEPTH,)'}), '(shape=(LATENT_DEPTH,))\n', (2015, 2038), True, 'import tensorflow.keras as keras\n'), ((2962, 2992), 'tensorflow.keras.models.Model', 'keras.models.Model', (['input_G', 'X'], {}), '(input_G, X)\n', (2980, 2992), True, 'import tensorflow.keras as keras\n'), ((3029, 3074), 'tensorflow.keras.layers.Input', 'keras.layers.Input', ([], {'shape': '(IM_DIM, IM_DIM, 3)'}), '(shape=(IM_DIM, IM_DIM, 3))\n', (3047, 3074), True, 'import tensorflow.keras as keras\n'), ((4057, 4108), 'tensorflow.keras.models.Model', 'keras.models.Model', (['input_D', '[output, inner_output]'], {}), '(input_D, [output, inner_output])\n', (4075, 4108), True, 'import tensorflow.keras as keras\n'), ((4146, 4166), 'os.listdir', 'os.listdir', (['DIR_NAME'], {}), '(DIR_NAME)\n', (4156, 4166), False, 'import os\n'), ((4391, 4459), 'numpy.zeros', 'np.zeros', (['(M * N, SCREEN_SIZE_X, SCREEN_SIZE_Y, 3)'], {'dtype': 'np.float32'}), '((M * N, SCREEN_SIZE_X, SCREEN_SIZE_Y, 3), dtype=np.float32)\n', (4399, 4459), True, 'import numpy as np\n'), ((5297, 5334), 'tensorflow.random.normal', 'tf.random.normal', (['(100, LATENT_DEPTH)'], {}), '((100, LATENT_DEPTH))\n', (5313, 5334), True, 'import tensorflow as tf\n'), ((8775, 8823), 'argparse.ArgumentParser', 'argparse.ArgumentParser', ([], {'description': '"""Train VAE"""'}), "(description='Train VAE')\n", (8798, 8823), False, 'import argparse\n'), ((779, 868), 'tensorflow.keras.layers.Conv2D', 'keras.layers.Conv2D', ([], {'filters': '(DEPTH * 2)', 'kernel_size': 'K_SIZE', 'strides': '(2)', 'padding': '"""same"""'}), "(filters=DEPTH * 2, kernel_size=K_SIZE, strides=2,\n padding='same')\n", (798, 868), True, 'import tensorflow.keras as keras\n'), ((880, 913), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (911, 913), True, 'import tensorflow.keras as keras\n'), ((925, 958), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (947, 958), True, 'import tensorflow.keras as keras\n'), ((971, 1060), 'tensorflow.keras.layers.Conv2D', 'keras.layers.Conv2D', ([], {'filters': '(DEPTH * 4)', 'kernel_size': 'K_SIZE', 'strides': '(2)', 'padding': '"""same"""'}), "(filters=DEPTH * 4, kernel_size=K_SIZE, strides=2,\n padding='same')\n", (990, 1060), True, 'import tensorflow.keras as keras\n'), ((1066, 1099), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (1097, 1099), True, 'import tensorflow.keras as keras\n'), ((1111, 1144), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (1133, 1144), True, 'import tensorflow.keras as keras\n'), ((1157, 1246), 'tensorflow.keras.layers.Conv2D', 'keras.layers.Conv2D', ([], {'filters': '(DEPTH * 8)', 'kernel_size': 'K_SIZE', 'strides': '(2)', 'padding': '"""same"""'}), "(filters=DEPTH * 8, kernel_size=K_SIZE, strides=2,\n padding='same')\n", (1176, 1246), True, 'import tensorflow.keras as keras\n'), ((1252, 1285), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (1283, 1285), True, 'import tensorflow.keras as keras\n'), ((1297, 1330), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (1319, 1330), True, 'import tensorflow.keras as keras\n'), ((1347, 1369), 'tensorflow.keras.layers.Flatten', 'keras.layers.Flatten', ([], {}), '()\n', (1367, 1369), True, 'import tensorflow.keras as keras\n'), ((1381, 1413), 'tensorflow.keras.layers.Dense', 'keras.layers.Dense', (['LATENT_DEPTH'], {}), '(LATENT_DEPTH)\n', (1399, 1413), True, 'import tensorflow.keras as keras\n'), ((1429, 1462), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (1460, 1462), True, 'import tensorflow.keras as keras\n'), ((1474, 1507), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (1496, 1507), True, 'import tensorflow.keras as keras\n'), ((1527, 1578), 'tensorflow.keras.layers.Dense', 'keras.layers.Dense', (['LATENT_DEPTH'], {'activation': '"""tanh"""'}), "(LATENT_DEPTH, activation='tanh')\n", (1545, 1578), True, 'import tensorflow.keras as keras\n'), ((1596, 1647), 'tensorflow.keras.layers.Dense', 'keras.layers.Dense', (['LATENT_DEPTH'], {'activation': '"""tanh"""'}), "(LATENT_DEPTH, activation='tanh')\n", (1614, 1647), True, 'import tensorflow.keras as keras\n'), ((1663, 1722), 'tensorflow.keras.layers.Lambda', 'keras.layers.Lambda', (['sampling'], {'output_shape': '(LATENT_DEPTH,)'}), '(sampling, output_shape=(LATENT_DEPTH,))\n', (1682, 1722), True, 'import tensorflow.keras as keras\n'), ((1804, 1831), 'tensorflow.keras.backend.exp', 'keras.backend.exp', (['logsigma'], {}), '(logsigma)\n', (1821, 1831), True, 'import tensorflow.keras as keras\n'), ((2048, 2085), 'tensorflow.keras.layers.Dense', 'keras.layers.Dense', (['(8 * 8 * DEPTH * 8)'], {}), '(8 * 8 * DEPTH * 8)\n', (2066, 2085), True, 'import tensorflow.keras as keras\n'), ((2097, 2130), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (2128, 2130), True, 'import tensorflow.keras as keras\n'), ((2142, 2175), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (2164, 2175), True, 'import tensorflow.keras as keras\n'), ((2187, 2226), 'tensorflow.keras.layers.Reshape', 'keras.layers.Reshape', (['(8, 8, DEPTH * 8)'], {}), '((8, 8, DEPTH * 8))\n', (2207, 2226), True, 'import tensorflow.keras as keras\n'), ((2243, 2342), 'tensorflow.keras.layers.Conv2DTranspose', 'keras.layers.Conv2DTranspose', ([], {'filters': '(DEPTH * 8)', 'kernel_size': 'K_SIZE', 'strides': '(2)', 'padding': '"""same"""'}), "(filters=DEPTH * 8, kernel_size=K_SIZE, strides\n =2, padding='same')\n", (2271, 2342), True, 'import tensorflow.keras as keras\n'), ((2347, 2380), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (2378, 2380), True, 'import tensorflow.keras as keras\n'), ((2392, 2425), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (2414, 2425), True, 'import tensorflow.keras as keras\n'), ((2438, 2537), 'tensorflow.keras.layers.Conv2DTranspose', 'keras.layers.Conv2DTranspose', ([], {'filters': '(DEPTH * 4)', 'kernel_size': 'K_SIZE', 'strides': '(2)', 'padding': '"""same"""'}), "(filters=DEPTH * 4, kernel_size=K_SIZE, strides\n =2, padding='same')\n", (2466, 2537), True, 'import tensorflow.keras as keras\n'), ((2542, 2575), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (2573, 2575), True, 'import tensorflow.keras as keras\n'), ((2587, 2620), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (2609, 2620), True, 'import tensorflow.keras as keras\n'), ((2637, 2731), 'tensorflow.keras.layers.Conv2DTranspose', 'keras.layers.Conv2DTranspose', ([], {'filters': 'DEPTH', 'kernel_size': 'K_SIZE', 'strides': '(2)', 'padding': '"""same"""'}), "(filters=DEPTH, kernel_size=K_SIZE, strides=2,\n padding='same')\n", (2665, 2731), True, 'import tensorflow.keras as keras\n'), ((2739, 2772), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (2770, 2772), True, 'import tensorflow.keras as keras\n'), ((2784, 2817), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (2806, 2817), True, 'import tensorflow.keras as keras\n'), ((2834, 2900), 'tensorflow.keras.layers.Conv2D', 'keras.layers.Conv2D', ([], {'filters': '(3)', 'kernel_size': 'K_SIZE', 'padding': '"""same"""'}), "(filters=3, kernel_size=K_SIZE, padding='same')\n", (2853, 2900), True, 'import tensorflow.keras as keras\n'), ((2912, 2946), 'tensorflow.keras.layers.Activation', 'keras.layers.Activation', (['"""sigmoid"""'], {}), "('sigmoid')\n", (2935, 2946), True, 'import tensorflow.keras as keras\n'), ((3088, 3174), 'tensorflow.keras.layers.Conv2D', 'keras.layers.Conv2D', ([], {'filters': 'DEPTH', 'kernel_size': 'K_SIZE', 'strides': '(2)', 'padding': '"""same"""'}), "(filters=DEPTH, kernel_size=K_SIZE, strides=2, padding=\n 'same')\n", (3107, 3174), True, 'import tensorflow.keras as keras\n'), ((3187, 3220), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (3209, 3220), True, 'import tensorflow.keras as keras\n'), ((3237, 3326), 'tensorflow.keras.layers.Conv2D', 'keras.layers.Conv2D', ([], {'filters': '(DEPTH * 4)', 'kernel_size': 'K_SIZE', 'strides': '(2)', 'padding': '"""same"""'}), "(filters=DEPTH * 4, kernel_size=K_SIZE, strides=2,\n padding='same')\n", (3256, 3326), True, 'import tensorflow.keras as keras\n'), ((3338, 3371), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (3360, 3371), True, 'import tensorflow.keras as keras\n'), ((3383, 3416), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (3414, 3416), True, 'import tensorflow.keras as keras\n'), ((3429, 3518), 'tensorflow.keras.layers.Conv2D', 'keras.layers.Conv2D', ([], {'filters': '(DEPTH * 8)', 'kernel_size': 'K_SIZE', 'strides': '(2)', 'padding': '"""same"""'}), "(filters=DEPTH * 8, kernel_size=K_SIZE, strides=2,\n padding='same')\n", (3448, 3518), True, 'import tensorflow.keras as keras\n'), ((3524, 3557), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (3555, 3557), True, 'import tensorflow.keras as keras\n'), ((3569, 3602), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (3591, 3602), True, 'import tensorflow.keras as keras\n'), ((3615, 3689), 'tensorflow.keras.layers.Conv2D', 'keras.layers.Conv2D', ([], {'filters': '(DEPTH * 8)', 'kernel_size': 'K_SIZE', 'padding': '"""same"""'}), "(filters=DEPTH * 8, kernel_size=K_SIZE, padding='same')\n", (3634, 3689), True, 'import tensorflow.keras as keras\n'), ((3710, 3732), 'tensorflow.keras.layers.Flatten', 'keras.layers.Flatten', ([], {}), '()\n', (3730, 3732), True, 'import tensorflow.keras as keras\n'), ((3744, 3777), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (3775, 3777), True, 'import tensorflow.keras as keras\n'), ((3789, 3822), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (3811, 3822), True, 'import tensorflow.keras as keras\n'), ((3839, 3861), 'tensorflow.keras.layers.Flatten', 'keras.layers.Flatten', ([], {}), '()\n', (3859, 3861), True, 'import tensorflow.keras as keras\n'), ((3873, 3902), 'tensorflow.keras.layers.Dense', 'keras.layers.Dense', (['(DEPTH * 8)'], {}), '(DEPTH * 8)\n', (3891, 3902), True, 'import tensorflow.keras as keras\n'), ((3912, 3945), 'tensorflow.keras.layers.BatchNormalization', 'keras.layers.BatchNormalization', ([], {}), '()\n', (3943, 3945), True, 'import tensorflow.keras as keras\n'), ((3957, 3990), 'tensorflow.keras.layers.LeakyReLU', 'keras.layers.LeakyReLU', ([], {'alpha': '(0.2)'}), '(alpha=0.2)\n', (3979, 3990), True, 'import tensorflow.keras as keras\n'), ((4012, 4033), 'tensorflow.keras.layers.Dense', 'keras.layers.Dense', (['(1)'], {}), '(1)\n', (4030, 4033), True, 'import tensorflow.keras as keras\n'), ((5344, 5376), 'tensorflow.GradientTape', 'tf.GradientTape', ([], {'persistent': '(True)'}), '(persistent=True)\n', (5359, 5376), True, 'import tensorflow as tf\n'), ((5867, 5890), 'tensorflow.reduce_mean', 'tf.reduce_mean', (['kl_loss'], {}), '(kl_loss)\n', (5881, 5890), True, 'import tensorflow as tf\n'), ((5915, 5940), 'tensorflow.reduce_mean', 'tf.reduce_mean', (['vae_inner'], {}), '(vae_inner)\n', (5929, 5940), True, 'import tensorflow as tf\n'), ((614, 639), 'tensorflow.keras.backend.shape', 'keras.backend.shape', (['mean'], {}), '(mean)\n', (633, 639), True, 'import tensorflow.keras as keras\n'), ((659, 679), 'tensorflow.exp', 'tf.exp', (['(logsigma / 2)'], {}), '(logsigma / 2)\n', (665, 679), True, 'import tensorflow as tf\n'), ((1775, 1801), 'tensorflow.keras.backend.square', 'keras.backend.square', (['mean'], {}), '(mean)\n', (1795, 1801), True, 'import tensorflow.keras as keras\n'), ((5774, 5801), 'tensorflow.reduce_mean', 'tf.reduce_mean', (['(mean * mean)'], {}), '(mean * mean)\n', (5788, 5801), True, 'import tensorflow as tf\n'), ((5802, 5841), 'tensorflow.reduce_mean', 'tf.reduce_mean', (['(var_to_one * var_to_one)'], {}), '(var_to_one * var_to_one)\n', (5816, 5841), True, 'import tensorflow as tf\n'), ((6386, 6402), 'tensorflow.abs', 'tf.abs', (['(x - fake)'], {}), '(x - fake)\n', (6392, 6402), True, 'import tensorflow as tf\n'), ((7943, 7987), 'tensorflow.keras.metrics.Mean', 'tf.keras.metrics.Mean', (['"""m"""'], {'dtype': 'tf.float32'}), "('m', dtype=tf.float32)\n", (7964, 7987), True, 'import tensorflow as tf\n'), ((4541, 4565), 'numpy.load', 'np.load', (['(DIR_NAME + file)'], {}), '(DIR_NAME + file)\n', (4548, 4565), True, 'import numpy as np\n'), ((6017, 6040), 'tensorflow.zeros_like', 'tf.zeros_like', (['dis_fake'], {}), '(dis_fake)\n', (6030, 6040), True, 'import tensorflow as tf\n'), ((6129, 6154), 'tensorflow.zeros_like', 'tf.zeros_like', (['dis_fake_r'], {}), '(dis_fake_r)\n', (6142, 6154), True, 'import tensorflow as tf\n'), ((6245, 6267), 'tensorflow.ones_like', 'tf.ones_like', (['dis_true'], {}), '(dis_true)\n', (6257, 6267), True, 'import tensorflow as tf\n')]
from torch.utils.data import DataLoader, Dataset from abc import abstractmethod from torch import Tensor import json import numpy as np from sqlalchemy import create_engine from torchvision.io import read_image import pandas as pd import configparser import os from gensim.models import Word2Vec config = configparser.ConfigParser() config.read('config.ini') COVER_LOC = config['IMAGES']['ImageLocation'] dbconf = config["DATABASE"] uname = dbconf['UserName'] pword = dbconf['Password'] addrs = dbconf['Address'] dname = dbconf['Database'] connstring = f'mysql+pymysql://{uname}:{pword}@{addrs}/{dname}?charset=utf8mb4' ENGINE = create_engine(connstring) class BandcampDatasetBase(Dataset): def __init__(self, engine=ENGINE, loc=COVER_LOC, transform=None): self.df = pd.read_sql('SELECT * FROM albums', engine) self.img_dir = loc self.transform = transform self.img_lines = self.df['id'] + '.jpg' def __len__(self): return len(self.img_lines) def get_img(self, item): img_path = os.path.join(self.img_dir, self.img_lines[item]) image = read_image(img_path).moveaxis([0, 1, 2], [-1, -3, -2]) if self.transform: image = self.transform(image) return image @abstractmethod def __getitem__(self, item): raise NotImplementedError class BandcampTagDataset(BandcampDatasetBase): def __init__(self, **kwargs): super(BandcampTagDataset, self).__init__(**kwargs) self.tag_jsons = self.df['tags'] self.w2v_model = Word2Vec.load('./models/tags.model') def __getitem__(self, item): image = self.get_img(item) tags = self.tag_jsons.iloc[item] tag_list = json.loads(tags) tag_list = np.concatenate([s.split(' ') for s in tag_list]) cbow = np.mean(self.w2v_model.wv[tag_list], axis=0) return image, cbow
[ "numpy.mean", "json.loads", "configparser.ConfigParser", "sqlalchemy.create_engine", "gensim.models.Word2Vec.load", "os.path.join", "torchvision.io.read_image", "pandas.read_sql" ]
[((307, 334), 'configparser.ConfigParser', 'configparser.ConfigParser', ([], {}), '()\n', (332, 334), False, 'import configparser\n'), ((631, 656), 'sqlalchemy.create_engine', 'create_engine', (['connstring'], {}), '(connstring)\n', (644, 656), False, 'from sqlalchemy import create_engine\n'), ((783, 826), 'pandas.read_sql', 'pd.read_sql', (['"""SELECT * FROM albums"""', 'engine'], {}), "('SELECT * FROM albums', engine)\n", (794, 826), True, 'import pandas as pd\n'), ((1045, 1093), 'os.path.join', 'os.path.join', (['self.img_dir', 'self.img_lines[item]'], {}), '(self.img_dir, self.img_lines[item])\n', (1057, 1093), False, 'import os\n'), ((1551, 1587), 'gensim.models.Word2Vec.load', 'Word2Vec.load', (['"""./models/tags.model"""'], {}), "('./models/tags.model')\n", (1564, 1587), False, 'from gensim.models import Word2Vec\n'), ((1717, 1733), 'json.loads', 'json.loads', (['tags'], {}), '(tags)\n', (1727, 1733), False, 'import json\n'), ((1817, 1861), 'numpy.mean', 'np.mean', (['self.w2v_model.wv[tag_list]'], {'axis': '(0)'}), '(self.w2v_model.wv[tag_list], axis=0)\n', (1824, 1861), True, 'import numpy as np\n'), ((1110, 1130), 'torchvision.io.read_image', 'read_image', (['img_path'], {}), '(img_path)\n', (1120, 1130), False, 'from torchvision.io import read_image\n')]
import argparse import skimage import skimage.io import skimage.transform from PIL import Image from skimage import io from math import log10 import sys import shutil import os import torch import torch.nn as nn import torch.nn.parallel import torch.backends.cudnn as cudnn import torch.optim as optim from torch.autograd import Variable from torch.utils.data import DataLoader from time import time from struct import unpack import matplotlib.pyplot as plt import re import numpy as np import pdb from path import Path from utils.preprocess import scale_disp, default_transform from networks.FADNet import FADNet from networks.DispNetC import DispNetC from networks.GANet_deep import GANet from networks.stackhourglass import PSMNet from networks.aanet import AANet parser = argparse.ArgumentParser(description='FADNet') parser.add_argument('--crop_height', type=int, required=True, help="crop height") parser.add_argument('--crop_width', type=int, required=True, help="crop width") parser.add_argument('--sceneflow', type=int, default=0, help='sceneflow dataset? Default=False') parser.add_argument('--kitti2012', type=int, default=0, help='kitti 2012? Default=False') parser.add_argument('--kitti2015', type=int, default=0, help='kitti 2015? Default=False') parser.add_argument('--middlebury', type=int, default=0, help='Middlebury? Default=False') parser.add_argument('--eth3d', type=int, default=0, help='ETH3D? Default=False') parser.add_argument('--datapath', default='/media/jiaren/ImageNet/data_scene_flow_2015/testing/', help='data path') parser.add_argument('--list', default='lists/middeval_test.list', help='list of stereo images') parser.add_argument('--loadmodel', default=None, help='loading model') parser.add_argument('--savepath', default='results/', help='path to save the results.') parser.add_argument('--model', default='fadnet', help='select model') parser.add_argument('--maxdisp', type=int, default=192, help='maxium disparity') parser.add_argument('--no-cuda', action='store_true', default=False, help='enables CUDA training') parser.add_argument('--devices', type=str, help='indicates CUDA devices, e.g. 0,1,2', default='0') parser.add_argument('--seed', type=int, default=1, metavar='S', help='random seed (default: 1)') opt = parser.parse_args() print(opt) torch.backends.cudnn.benchmark = True opt.cuda = not opt.no_cuda and torch.cuda.is_available() if not os.path.exists(opt.savepath): os.makedirs(opt.savepath) torch.manual_seed(opt.seed) if opt.cuda: torch.cuda.manual_seed(opt.seed) if opt.model == 'psmnet': model = PSMNet(opt.maxdisp) elif opt.model == 'ganet': model = GANet(opt.maxdisp) elif opt.model == 'aanet': model = AANet(opt.maxdisp) elif opt.model == 'fadnet': model = FADNet(maxdisp=opt.maxdisp) elif opt.model == 'dispnetc': model = DispNetC(resBlock=False, maxdisp=opt.maxdisp) elif opt.model == 'crl': model = FADNet(resBlock=False, maxdisp=opt.maxdisp) else: print('no model') sys.exit(-1) model = nn.DataParallel(model, device_ids=[0]) model.cuda() if opt.loadmodel is not None: state_dict = torch.load(opt.loadmodel) model.load_state_dict(state_dict['state_dict']) print('Number of model parameters: {}'.format(sum([p.data.nelement() for p in model.parameters()]))) def readPFM(file): with open(file, "rb") as f: # Line 1: PF=>RGB (3 channels), Pf=>Greyscale (1 channel) type = f.readline().decode('latin-1') if "PF" in type: channels = 3 elif "Pf" in type: channels = 1 else: sys.exit(1) # Line 2: width height line = f.readline().decode('latin-1') width, height = re.findall('\d+', line) width = int(width) height = int(height) # Line 3: +ve number means big endian, negative means little endian line = f.readline().decode('latin-1') BigEndian = True if "-" in line: BigEndian = False # Slurp all binary data samples = width * height * channels; buffer = f.read(samples * 4) # Unpack floats with appropriate endianness if BigEndian: fmt = ">" else: fmt = "<" fmt = fmt + str(samples) + "f" img = unpack(fmt, buffer) img = np.reshape(img, (height, width)) img = np.flipud(img) return img, height, width def save_pfm(filename, image, scale=1): ''' Save a Numpy array to a PFM file. ''' color = None file = open(filename, "w") if image.dtype.name != 'float32': raise Exception('Image dtype must be float32.') if len(image.shape) == 3 and image.shape[2] == 3: # color image color = True elif len(image.shape) == 2 or len(image.shape) == 3 and image.shape[2] == 1: # greyscale color = False else: raise Exception('Image must have H x W x 3, H x W x 1 or H x W dimensions.') file.write('PF\n' if color else 'Pf\n') file.write('%d %d\n' % (image.shape[1], image.shape[0])) endian = image.dtype.byteorder if endian == '<' or endian == '=' and sys.byteorder == 'little': scale = -scale file.write('%f\n' % scale) image.tofile(file) def test_transform(temp_data, crop_height, crop_width): _, h, w=np.shape(temp_data) if h <= crop_height and w <= crop_width: # padding zero temp = temp_data temp_data = np.zeros([6, crop_height, crop_width], 'float32') temp_data[:, crop_height - h: crop_height, crop_width - w: crop_width] = temp else: start_x = int((w - crop_width) / 2) start_y = int((h - crop_height) / 2) temp_data = temp_data[:, start_y: start_y + crop_height, start_x: start_x + crop_width] left = np.ones([1, 3,crop_height,crop_width],'float32') left[0, :, :, :] = temp_data[0: 3, :, :] right = np.ones([1, 3, crop_height, crop_width], 'float32') right[0, :, :, :] = temp_data[3: 6, :, :] return torch.from_numpy(left).float(), torch.from_numpy(right).float(), h, w def load_data(leftname, rightname): left = Image.open(leftname) right = Image.open(rightname) size = np.shape(left) height = size[0] width = size[1] temp_data = np.zeros([6, height, width], 'float32') left = np.asarray(left) right = np.asarray(right) r = left[:, :, 0] g = left[:, :, 1] b = left[:, :, 2] temp_data[0, :, :] = (r - np.mean(r[:])) / np.std(r[:]) temp_data[1, :, :] = (g - np.mean(g[:])) / np.std(g[:]) temp_data[2, :, :] = (b - np.mean(b[:])) / np.std(b[:]) r = right[:, :, 0] g = right[:, :, 1] b = right[:, :, 2] #r,g,b,_ = right.split() temp_data[3, :, :] = (r - np.mean(r[:])) / np.std(r[:]) temp_data[4, :, :] = (g - np.mean(g[:])) / np.std(g[:]) temp_data[5, :, :] = (b - np.mean(b[:])) / np.std(b[:]) return temp_data def load_data_imn(leftname, rightname): #left = Image.open(leftname) #right = Image.open(rightname) #h, w, c = np.shape(left) left = io.imread(leftname) right = io.imread(rightname) h, w, _ = left.shape normalize = 'instnorm' if normalize == 'imagenet': rgb_transform = default_transform() img_left = rgb_transform(left) img_right = rgb_transform(right) else: img_left = np.zeros([3, h, w], 'float32') img_right = np.zeros([3, h, w], 'float32') for c in range(3): img_left[c, :, :] = (left[:, :, c] - np.mean(left[:, :, c])) / np.std(left[:, :, c]) img_right[c, :, :] = (right[:, :, c] - np.mean(right[:, :, c])) / np.std(right[:, :, c]) print(h, w) bottom_pad = opt.crop_height-h right_pad = opt.crop_width-w img_left = np.lib.pad(img_left,((0,0),(0,bottom_pad),(0,right_pad)),mode='constant',constant_values=0) img_right = np.lib.pad(img_right,((0,0),(0,bottom_pad),(0,right_pad)),mode='constant',constant_values=0) return torch.from_numpy(img_left).float(), torch.from_numpy(img_right).float(), h, w def test_md(leftname, rightname, savename): print(savename) epe = 0 input1, input2, height, width = load_data_imn(leftname, rightname) input1 = Variable(input1, requires_grad = False) input2 = Variable(input2, requires_grad = False) model.eval() if opt.cuda: input1 = input1.cuda() input2 = input2.cuda() input_var = torch.cat((input1, input2), 0) input_var = input_var.unsqueeze(0) torch.cuda.synchronize() start_time = time() with torch.no_grad(): prediction = model(input_var)[1] prediction = prediction.squeeze(0) torch.cuda.synchronize() end_time = time() print("Processing time: {:.4f}".format(end_time - start_time)) temp = prediction.cpu() temp = temp.detach().numpy() temp = temp[0, :height, :width] # print epe thres = 400 if 'trainingQ' in leftname: thres = 192 if 'training' in leftname: gt_disp, _, _ = readPFM(leftname.replace('im0.png', 'disp0GT.pfm')) gt_disp[np.isinf(gt_disp)] = 0 mask = (gt_disp > 0) & (gt_disp < thres) epe = np.mean(np.abs(gt_disp[mask] - temp[mask])) print(savename, epe, np.min(gt_disp), np.max(gt_disp)) savepfm_path = savename.replace('.png','') temp = np.flipud(temp) disppath = Path(savepfm_path) disppath.makedirs_p() save_pfm(savepfm_path+'/disp0FADNet++.pfm', temp, scale=1) ##########write time txt######## fp = open(savepfm_path+'/timeFADNet++.txt', 'w') runtime = "%.4f" % (end_time - start_time) fp.write(runtime) fp.close() return epe def test_eth3d(leftname, rightname, savename): print(savename) epe = 0 input1, input2, height, width = load_data_imn(leftname, rightname) input1 = Variable(input1, requires_grad = False) input2 = Variable(input2, requires_grad = False) model.eval() if opt.cuda: input1 = input1.cuda() input2 = input2.cuda() input_var = torch.cat((input1, input2), 0) input_var = input_var.unsqueeze(0) torch.cuda.synchronize() start_time = time() with torch.no_grad(): prediction = model(input_var)[1] prediction = prediction.squeeze(0) torch.cuda.synchronize() end_time = time() print("Processing time: {:.4f}".format(end_time - start_time)) temp = prediction.cpu() temp = temp.detach().numpy() temp = temp[0, :height, :width] # print epe if 'training' in leftname: gt_disp, _, _ = readPFM(leftname.replace('im0.png', 'disp0GT.pfm').replace('training/', 'training_gt/')) gt_disp[np.isinf(gt_disp)] = 0 mask = (gt_disp > 0) & (gt_disp < 192) epe = np.mean(np.abs(gt_disp[mask] - temp[mask])) print(savename, epe, np.min(gt_disp), np.max(gt_disp)) temp = np.flipud(temp) disppath = Path('/'.join(savename.split('/')[:-1])) disppath.makedirs_p() save_pfm(savename, temp, scale=1) ##########write time txt######## fp = open(savename.replace("pfm", "txt"), 'w') runtime = "runtime %.4f" % (end_time - start_time) fp.write(runtime) fp.close() return epe def test_kitti(leftname, rightname, savename): print(savename) epe = 0 input1, input2, height, width = load_data_imn(leftname, rightname) input1 = Variable(input1, requires_grad = False) input2 = Variable(input2, requires_grad = False) model.eval() if opt.cuda: input1 = input1.cuda() input2 = input2.cuda() input_var = torch.cat((input1, input2), 0) input_var = input_var.unsqueeze(0) with torch.no_grad(): prediction = model(input_var)[1] prediction = prediction.squeeze(0) temp = prediction.cpu() temp = temp.detach().numpy() temp = temp[0, :height, :width] # print epe if 'training' in leftname: gt_disp = Image.open(leftname.replace('colored_0', 'disp_occ').replace('image_2', 'disp_occ_0')) gt_disp = np.ascontiguousarray(gt_disp,dtype=np.float32)/256 mask = (gt_disp > 0) & (gt_disp < 192) epe = np.mean(np.abs(gt_disp[mask] - temp[mask])) print(savename, epe, np.min(gt_disp), np.max(gt_disp)) skimage.io.imsave(savename, (temp * 256).astype('uint16')) return epe def test(leftname, rightname, savename, gt_disp): input1, input2, height, width = load_data_imn(leftname, rightname) input1 = Variable(input1, requires_grad = False) input2 = Variable(input2, requires_grad = False) model.eval() start_time = time() if opt.cuda: input1 = input1.cuda() input2 = input2.cuda() input_var = torch.cat((input1, input2), 0) input_var = input_var.unsqueeze(0) with torch.no_grad(): predictions = model(input_var) if len(predictions) == 5: # aanet prediction = predictions[-1] elif len(predictions) > 2: # ganet, psmnet prediction = predictions[0] elif len(predictions) == 2: # two-stage nets prediction = predictions[1] else: prediction = predictions if prediction.dim() == 4: prediction = prediction.squeeze(0) end_time = time() print("Processing time: {:.4f}".format(end_time - start_time)) temp = prediction.cpu() temp = temp.detach().numpy() temp = temp[0, :height, :width] plot_disparity(savename, temp, np.max(gt_disp)+5, cmap='rainbow') mask = (gt_disp > 0) & (gt_disp < 192) epe = np.mean(np.abs(gt_disp[mask] - temp[mask])) err_map = np.abs(temp - gt_disp) err_name = savename.replace('disp', 'err') plot_disparity(err_name, err_map, 30, cmap='turbo') savename_pfm = savename.replace('png','pfm') temp = np.flipud(temp) return epe def plot_disparity(savename, data, max_disp, cmap='turbo'): plt.imsave(savename, data, vmin=0, vmax=max_disp, cmap=cmap) if __name__ == "__main__": file_path = opt.datapath file_list = opt.list f = open(file_list, 'r') filelist = f.readlines() error = 0 for index in range(len(filelist)): current_file = filelist[index].split() if opt.kitti2015 or opt.kitti2012: leftname = os.path.join(file_path, current_file[0]) rightname = os.path.join(file_path, current_file[1]) savename = os.path.join(opt.savepath, current_file[0].split("/")[-1]) error += test_kitti(leftname, rightname, savename) if opt.sceneflow: leftname = file_path + current_file[0] rightname = file_path + current_file[1] leftgtname = file_path + current_file[2] disp_left_gt, height, width = readPFM(leftgtname) savenamegt = opt.savepath + "gt_" + "_".join(current_file[2].split("/")[-4:]).replace('pfm', 'png').replace('left', 'disp') plot_disparity(savenamegt, disp_left_gt, np.max(disp_left_gt)+5, cmap='rainbow') savename = opt.savepath + "%s_" % opt.model + "_".join(current_file[2].split("/")[-4:]).replace('pfm', 'png').replace('left', 'disp') epe = test(leftname, rightname, savename, disp_left_gt) error += epe print(leftname, rightname, savename, epe) if opt.middlebury: leftname = file_path + current_file[0] rightname = file_path + current_file[1] temppath = opt.savepath.replace(opt.savepath.split("/")[-2], opt.savepath.split("/")[-2]+"/images") #img_path = Path(temppath) #img_path.makedirs_p() savename = opt.savepath + "/".join(leftname.split("/")[-4:-1]) + ".png" error += test_md(leftname, rightname, savename) if opt.eth3d: leftname = file_path + current_file[0] rightname = file_path + current_file[1] savename = opt.savepath + "low_res_two_view/" + leftname.split("/")[-2] + ".pfm" error += test_eth3d(leftname, rightname, savename) if index > 200: break print("EPE:", error / len(filelist))
[ "torch.from_numpy", "numpy.ascontiguousarray", "torch.cuda.synchronize", "numpy.lib.pad", "path.Path", "torch.cuda.is_available", "sys.exit", "networks.DispNetC.DispNetC", "networks.GANet_deep.GANet", "networks.aanet.AANet", "os.path.exists", "numpy.mean", "numpy.reshape", "argparse.Argume...
[((780, 825), 'argparse.ArgumentParser', 'argparse.ArgumentParser', ([], {'description': '"""FADNet"""'}), "(description='FADNet')\n", (803, 825), False, 'import argparse\n'), ((2617, 2644), 'torch.manual_seed', 'torch.manual_seed', (['opt.seed'], {}), '(opt.seed)\n', (2634, 2644), False, 'import torch\n'), ((3161, 3199), 'torch.nn.DataParallel', 'nn.DataParallel', (['model'], {'device_ids': '[0]'}), '(model, device_ids=[0])\n', (3176, 3199), True, 'import torch.nn as nn\n'), ((2522, 2547), 'torch.cuda.is_available', 'torch.cuda.is_available', ([], {}), '()\n', (2545, 2547), False, 'import torch\n'), ((2556, 2584), 'os.path.exists', 'os.path.exists', (['opt.savepath'], {}), '(opt.savepath)\n', (2570, 2584), False, 'import os\n'), ((2590, 2615), 'os.makedirs', 'os.makedirs', (['opt.savepath'], {}), '(opt.savepath)\n', (2601, 2615), False, 'import os\n'), ((2662, 2694), 'torch.cuda.manual_seed', 'torch.cuda.manual_seed', (['opt.seed'], {}), '(opt.seed)\n', (2684, 2694), False, 'import torch\n'), ((2734, 2753), 'networks.stackhourglass.PSMNet', 'PSMNet', (['opt.maxdisp'], {}), '(opt.maxdisp)\n', (2740, 2753), False, 'from networks.stackhourglass import PSMNet\n'), ((3261, 3286), 'torch.load', 'torch.load', (['opt.loadmodel'], {}), '(opt.loadmodel)\n', (3271, 3286), False, 'import torch\n'), ((5457, 5476), 'numpy.shape', 'np.shape', (['temp_data'], {}), '(temp_data)\n', (5465, 5476), True, 'import numpy as np\n'), ((5939, 5990), 'numpy.ones', 'np.ones', (['[1, 3, crop_height, crop_width]', '"""float32"""'], {}), "([1, 3, crop_height, crop_width], 'float32')\n", (5946, 5990), True, 'import numpy as np\n'), ((6045, 6096), 'numpy.ones', 'np.ones', (['[1, 3, crop_height, crop_width]', '"""float32"""'], {}), "([1, 3, crop_height, crop_width], 'float32')\n", (6052, 6096), True, 'import numpy as np\n'), ((6272, 6292), 'PIL.Image.open', 'Image.open', (['leftname'], {}), '(leftname)\n', (6282, 6292), False, 'from PIL import Image\n'), ((6305, 6326), 'PIL.Image.open', 'Image.open', (['rightname'], {}), '(rightname)\n', (6315, 6326), False, 'from PIL import Image\n'), ((6338, 6352), 'numpy.shape', 'np.shape', (['left'], {}), '(left)\n', (6346, 6352), True, 'import numpy as np\n'), ((6410, 6449), 'numpy.zeros', 'np.zeros', (['[6, height, width]', '"""float32"""'], {}), "([6, height, width], 'float32')\n", (6418, 6449), True, 'import numpy as np\n'), ((6461, 6477), 'numpy.asarray', 'np.asarray', (['left'], {}), '(left)\n', (6471, 6477), True, 'import numpy as np\n'), ((6490, 6507), 'numpy.asarray', 'np.asarray', (['right'], {}), '(right)\n', (6500, 6507), True, 'import numpy as np\n'), ((7204, 7223), 'skimage.io.imread', 'io.imread', (['leftname'], {}), '(leftname)\n', (7213, 7223), False, 'from skimage import io\n'), ((7236, 7256), 'skimage.io.imread', 'io.imread', (['rightname'], {}), '(rightname)\n', (7245, 7256), False, 'from skimage import io\n'), ((7902, 8006), 'numpy.lib.pad', 'np.lib.pad', (['img_left', '((0, 0), (0, bottom_pad), (0, right_pad))'], {'mode': '"""constant"""', 'constant_values': '(0)'}), "(img_left, ((0, 0), (0, bottom_pad), (0, right_pad)), mode=\n 'constant', constant_values=0)\n", (7912, 8006), True, 'import numpy as np\n'), ((8010, 8115), 'numpy.lib.pad', 'np.lib.pad', (['img_right', '((0, 0), (0, bottom_pad), (0, right_pad))'], {'mode': '"""constant"""', 'constant_values': '(0)'}), "(img_right, ((0, 0), (0, bottom_pad), (0, right_pad)), mode=\n 'constant', constant_values=0)\n", (8020, 8115), True, 'import numpy as np\n'), ((8355, 8392), 'torch.autograd.Variable', 'Variable', (['input1'], {'requires_grad': '(False)'}), '(input1, requires_grad=False)\n', (8363, 8392), False, 'from torch.autograd import Variable\n'), ((8408, 8445), 'torch.autograd.Variable', 'Variable', (['input2'], {'requires_grad': '(False)'}), '(input2, requires_grad=False)\n', (8416, 8445), False, 'from torch.autograd import Variable\n'), ((8643, 8667), 'torch.cuda.synchronize', 'torch.cuda.synchronize', ([], {}), '()\n', (8665, 8667), False, 'import torch\n'), ((8685, 8691), 'time.time', 'time', ([], {}), '()\n', (8689, 8691), False, 'from time import time\n'), ((8806, 8830), 'torch.cuda.synchronize', 'torch.cuda.synchronize', ([], {}), '()\n', (8828, 8830), False, 'import torch\n'), ((8846, 8852), 'time.time', 'time', ([], {}), '()\n', (8850, 8852), False, 'from time import time\n'), ((9484, 9499), 'numpy.flipud', 'np.flipud', (['temp'], {}), '(temp)\n', (9493, 9499), True, 'import numpy as np\n'), ((9516, 9534), 'path.Path', 'Path', (['savepfm_path'], {}), '(savepfm_path)\n', (9520, 9534), False, 'from path import Path\n'), ((9985, 10022), 'torch.autograd.Variable', 'Variable', (['input1'], {'requires_grad': '(False)'}), '(input1, requires_grad=False)\n', (9993, 10022), False, 'from torch.autograd import Variable\n'), ((10038, 10075), 'torch.autograd.Variable', 'Variable', (['input2'], {'requires_grad': '(False)'}), '(input2, requires_grad=False)\n', (10046, 10075), False, 'from torch.autograd import Variable\n'), ((10273, 10297), 'torch.cuda.synchronize', 'torch.cuda.synchronize', ([], {}), '()\n', (10295, 10297), False, 'import torch\n'), ((10315, 10321), 'time.time', 'time', ([], {}), '()\n', (10319, 10321), False, 'from time import time\n'), ((10436, 10460), 'torch.cuda.synchronize', 'torch.cuda.synchronize', ([], {}), '()\n', (10458, 10460), False, 'import torch\n'), ((10476, 10482), 'time.time', 'time', ([], {}), '()\n', (10480, 10482), False, 'from time import time\n'), ((11033, 11048), 'numpy.flipud', 'np.flipud', (['temp'], {}), '(temp)\n', (11042, 11048), True, 'import numpy as np\n'), ((11536, 11573), 'torch.autograd.Variable', 'Variable', (['input1'], {'requires_grad': '(False)'}), '(input1, requires_grad=False)\n', (11544, 11573), False, 'from torch.autograd import Variable\n'), ((11589, 11626), 'torch.autograd.Variable', 'Variable', (['input2'], {'requires_grad': '(False)'}), '(input2, requires_grad=False)\n', (11597, 11626), False, 'from torch.autograd import Variable\n'), ((12654, 12691), 'torch.autograd.Variable', 'Variable', (['input1'], {'requires_grad': '(False)'}), '(input1, requires_grad=False)\n', (12662, 12691), False, 'from torch.autograd import Variable\n'), ((12707, 12744), 'torch.autograd.Variable', 'Variable', (['input2'], {'requires_grad': '(False)'}), '(input2, requires_grad=False)\n', (12715, 12744), False, 'from torch.autograd import Variable\n'), ((12782, 12788), 'time.time', 'time', ([], {}), '()\n', (12786, 12788), False, 'from time import time\n'), ((13443, 13449), 'time.time', 'time', ([], {}), '()\n', (13447, 13449), False, 'from time import time\n'), ((13799, 13821), 'numpy.abs', 'np.abs', (['(temp - gt_disp)'], {}), '(temp - gt_disp)\n', (13805, 13821), True, 'import numpy as np\n'), ((13987, 14002), 'numpy.flipud', 'np.flipud', (['temp'], {}), '(temp)\n', (13996, 14002), True, 'import numpy as np\n'), ((14084, 14144), 'matplotlib.pyplot.imsave', 'plt.imsave', (['savename', 'data'], {'vmin': '(0)', 'vmax': 'max_disp', 'cmap': 'cmap'}), '(savename, data, vmin=0, vmax=max_disp, cmap=cmap)\n', (14094, 14144), True, 'import matplotlib.pyplot as plt\n'), ((2793, 2811), 'networks.GANet_deep.GANet', 'GANet', (['opt.maxdisp'], {}), '(opt.maxdisp)\n', (2798, 2811), False, 'from networks.GANet_deep import GANet\n'), ((3848, 3872), 're.findall', 're.findall', (['"""\\\\d+"""', 'line'], {}), "('\\\\d+', line)\n", (3858, 3872), False, 'import re\n'), ((4433, 4452), 'struct.unpack', 'unpack', (['fmt', 'buffer'], {}), '(fmt, buffer)\n', (4439, 4452), False, 'from struct import unpack\n'), ((4467, 4499), 'numpy.reshape', 'np.reshape', (['img', '(height, width)'], {}), '(img, (height, width))\n', (4477, 4499), True, 'import numpy as np\n'), ((4514, 4528), 'numpy.flipud', 'np.flipud', (['img'], {}), '(img)\n', (4523, 4528), True, 'import numpy as np\n'), ((5593, 5642), 'numpy.zeros', 'np.zeros', (['[6, crop_height, crop_width]', '"""float32"""'], {}), "([6, crop_height, crop_width], 'float32')\n", (5601, 5642), True, 'import numpy as np\n'), ((6621, 6633), 'numpy.std', 'np.std', (['r[:]'], {}), '(r[:])\n', (6627, 6633), True, 'import numpy as np\n'), ((6681, 6693), 'numpy.std', 'np.std', (['g[:]'], {}), '(g[:])\n', (6687, 6693), True, 'import numpy as np\n'), ((6741, 6753), 'numpy.std', 'np.std', (['b[:]'], {}), '(b[:])\n', (6747, 6753), True, 'import numpy as np\n'), ((6900, 6912), 'numpy.std', 'np.std', (['r[:]'], {}), '(r[:])\n', (6906, 6912), True, 'import numpy as np\n'), ((6960, 6972), 'numpy.std', 'np.std', (['g[:]'], {}), '(g[:])\n', (6966, 6972), True, 'import numpy as np\n'), ((7020, 7032), 'numpy.std', 'np.std', (['b[:]'], {}), '(b[:])\n', (7026, 7032), True, 'import numpy as np\n'), ((7366, 7385), 'utils.preprocess.default_transform', 'default_transform', ([], {}), '()\n', (7383, 7385), False, 'from utils.preprocess import scale_disp, default_transform\n'), ((7495, 7525), 'numpy.zeros', 'np.zeros', (['[3, h, w]', '"""float32"""'], {}), "([3, h, w], 'float32')\n", (7503, 7525), True, 'import numpy as np\n'), ((7546, 7576), 'numpy.zeros', 'np.zeros', (['[3, h, w]', '"""float32"""'], {}), "([3, h, w], 'float32')\n", (7554, 7576), True, 'import numpy as np\n'), ((8565, 8595), 'torch.cat', 'torch.cat', (['(input1, input2)', '(0)'], {}), '((input1, input2), 0)\n', (8574, 8595), False, 'import torch\n'), ((8701, 8716), 'torch.no_grad', 'torch.no_grad', ([], {}), '()\n', (8714, 8716), False, 'import torch\n'), ((10195, 10225), 'torch.cat', 'torch.cat', (['(input1, input2)', '(0)'], {}), '((input1, input2), 0)\n', (10204, 10225), False, 'import torch\n'), ((10331, 10346), 'torch.no_grad', 'torch.no_grad', ([], {}), '()\n', (10344, 10346), False, 'import torch\n'), ((11746, 11776), 'torch.cat', 'torch.cat', (['(input1, input2)', '(0)'], {}), '((input1, input2), 0)\n', (11755, 11776), False, 'import torch\n'), ((11829, 11844), 'torch.no_grad', 'torch.no_grad', ([], {}), '()\n', (11842, 11844), False, 'import torch\n'), ((12889, 12919), 'torch.cat', 'torch.cat', (['(input1, input2)', '(0)'], {}), '((input1, input2), 0)\n', (12898, 12919), False, 'import torch\n'), ((12972, 12987), 'torch.no_grad', 'torch.no_grad', ([], {}), '()\n', (12985, 12987), False, 'import torch\n'), ((13748, 13782), 'numpy.abs', 'np.abs', (['(gt_disp[mask] - temp[mask])'], {}), '(gt_disp[mask] - temp[mask])\n', (13754, 13782), True, 'import numpy as np\n'), ((2851, 2869), 'networks.aanet.AANet', 'AANet', (['opt.maxdisp'], {}), '(opt.maxdisp)\n', (2856, 2869), False, 'from networks.aanet import AANet\n'), ((6604, 6617), 'numpy.mean', 'np.mean', (['r[:]'], {}), '(r[:])\n', (6611, 6617), True, 'import numpy as np\n'), ((6664, 6677), 'numpy.mean', 'np.mean', (['g[:]'], {}), '(g[:])\n', (6671, 6677), True, 'import numpy as np\n'), ((6724, 6737), 'numpy.mean', 'np.mean', (['b[:]'], {}), '(b[:])\n', (6731, 6737), True, 'import numpy as np\n'), ((6883, 6896), 'numpy.mean', 'np.mean', (['r[:]'], {}), '(r[:])\n', (6890, 6896), True, 'import numpy as np\n'), ((6943, 6956), 'numpy.mean', 'np.mean', (['g[:]'], {}), '(g[:])\n', (6950, 6956), True, 'import numpy as np\n'), ((7003, 7016), 'numpy.mean', 'np.mean', (['b[:]'], {}), '(b[:])\n', (7010, 7016), True, 'import numpy as np\n'), ((9230, 9247), 'numpy.isinf', 'np.isinf', (['gt_disp'], {}), '(gt_disp)\n', (9238, 9247), True, 'import numpy as np\n'), ((9324, 9358), 'numpy.abs', 'np.abs', (['(gt_disp[mask] - temp[mask])'], {}), '(gt_disp[mask] - temp[mask])\n', (9330, 9358), True, 'import numpy as np\n'), ((9390, 9405), 'numpy.min', 'np.min', (['gt_disp'], {}), '(gt_disp)\n', (9396, 9405), True, 'import numpy as np\n'), ((9407, 9422), 'numpy.max', 'np.max', (['gt_disp'], {}), '(gt_disp)\n', (9413, 9422), True, 'import numpy as np\n'), ((10829, 10846), 'numpy.isinf', 'np.isinf', (['gt_disp'], {}), '(gt_disp)\n', (10837, 10846), True, 'import numpy as np\n'), ((10921, 10955), 'numpy.abs', 'np.abs', (['(gt_disp[mask] - temp[mask])'], {}), '(gt_disp[mask] - temp[mask])\n', (10927, 10955), True, 'import numpy as np\n'), ((10987, 11002), 'numpy.min', 'np.min', (['gt_disp'], {}), '(gt_disp)\n', (10993, 11002), True, 'import numpy as np\n'), ((11004, 11019), 'numpy.max', 'np.max', (['gt_disp'], {}), '(gt_disp)\n', (11010, 11019), True, 'import numpy as np\n'), ((12215, 12262), 'numpy.ascontiguousarray', 'np.ascontiguousarray', (['gt_disp'], {'dtype': 'np.float32'}), '(gt_disp, dtype=np.float32)\n', (12235, 12262), True, 'import numpy as np\n'), ((12335, 12369), 'numpy.abs', 'np.abs', (['(gt_disp[mask] - temp[mask])'], {}), '(gt_disp[mask] - temp[mask])\n', (12341, 12369), True, 'import numpy as np\n'), ((12401, 12416), 'numpy.min', 'np.min', (['gt_disp'], {}), '(gt_disp)\n', (12407, 12416), True, 'import numpy as np\n'), ((12418, 12433), 'numpy.max', 'np.max', (['gt_disp'], {}), '(gt_disp)\n', (12424, 12433), True, 'import numpy as np\n'), ((13651, 13666), 'numpy.max', 'np.max', (['gt_disp'], {}), '(gt_disp)\n', (13657, 13666), True, 'import numpy as np\n'), ((14456, 14496), 'os.path.join', 'os.path.join', (['file_path', 'current_file[0]'], {}), '(file_path, current_file[0])\n', (14468, 14496), False, 'import os\n'), ((14521, 14561), 'os.path.join', 'os.path.join', (['file_path', 'current_file[1]'], {}), '(file_path, current_file[1])\n', (14533, 14561), False, 'import os\n'), ((2910, 2937), 'networks.FADNet.FADNet', 'FADNet', ([], {'maxdisp': 'opt.maxdisp'}), '(maxdisp=opt.maxdisp)\n', (2916, 2937), False, 'from networks.FADNet import FADNet\n'), ((3735, 3746), 'sys.exit', 'sys.exit', (['(1)'], {}), '(1)\n', (3743, 3746), False, 'import sys\n'), ((6154, 6176), 'torch.from_numpy', 'torch.from_numpy', (['left'], {}), '(left)\n', (6170, 6176), False, 'import torch\n'), ((6186, 6209), 'torch.from_numpy', 'torch.from_numpy', (['right'], {}), '(right)\n', (6202, 6209), False, 'import torch\n'), ((7679, 7700), 'numpy.std', 'np.std', (['left[:, :, c]'], {}), '(left[:, :, c])\n', (7685, 7700), True, 'import numpy as np\n'), ((7779, 7801), 'numpy.std', 'np.std', (['right[:, :, c]'], {}), '(right[:, :, c])\n', (7785, 7801), True, 'import numpy as np\n'), ((8114, 8140), 'torch.from_numpy', 'torch.from_numpy', (['img_left'], {}), '(img_left)\n', (8130, 8140), False, 'import torch\n'), ((8150, 8177), 'torch.from_numpy', 'torch.from_numpy', (['img_right'], {}), '(img_right)\n', (8166, 8177), False, 'import torch\n'), ((2980, 3025), 'networks.DispNetC.DispNetC', 'DispNetC', ([], {'resBlock': '(False)', 'maxdisp': 'opt.maxdisp'}), '(resBlock=False, maxdisp=opt.maxdisp)\n', (2988, 3025), False, 'from networks.DispNetC import DispNetC\n'), ((7653, 7675), 'numpy.mean', 'np.mean', (['left[:, :, c]'], {}), '(left[:, :, c])\n', (7660, 7675), True, 'import numpy as np\n'), ((7752, 7775), 'numpy.mean', 'np.mean', (['right[:, :, c]'], {}), '(right[:, :, c])\n', (7759, 7775), True, 'import numpy as np\n'), ((15142, 15162), 'numpy.max', 'np.max', (['disp_left_gt'], {}), '(disp_left_gt)\n', (15148, 15162), True, 'import numpy as np\n'), ((3063, 3106), 'networks.FADNet.FADNet', 'FADNet', ([], {'resBlock': '(False)', 'maxdisp': 'opt.maxdisp'}), '(resBlock=False, maxdisp=opt.maxdisp)\n', (3069, 3106), False, 'from networks.FADNet import FADNet\n'), ((3139, 3151), 'sys.exit', 'sys.exit', (['(-1)'], {}), '(-1)\n', (3147, 3151), False, 'import sys\n')]
import numpy as np def init_spin_state_2d(nsize=16): """Initialize spin state""" return 2*np.random.randint(2, size=(nsize, nsize)) - 1 def mcmh_algorithm(state, beta=1): """Apply Monte Carlo Metropolis-Hastings algorithm""" # Get input dimensions height, width = state.shape energy = 0 for i in range(height): for j in range(width): # Periodic neighbors up, down, left, right = ( (i - 1) % height, (i + 1) & height, (j - 1) % width, (j + 1) & width ) # Spin interaction energies e_spin_init = J*( state[i, j]*state[up, j] + state[i, j]*state[down, j] + state[i, j]*state[i, left] + state[i, j]*state[i, right] ) e_spin_flip = J*( -state[i, j]*state[up, j] - state[i, j]*state[down, j] - state[i, j]*state[i, left] - state[i, j]*state[i, right] ) delta_e = e_spin_flip - e_spin_init energy += e_spin_flip # Metropolis updates if delta: state[i, j] = -state[i, j] elif np.random.rand() < np.exp(-beta*delta_e) : state[i, j] = -state[i, j] else: pass return state, energy/nsize**2., np.sum(state)/nsize**2. def run_simulation(num_iter, beta=16): """Run Ising model simulation""" # Randomly initialize spin state state = init_spin_state(nsize=10) for step in num_iter: # Update state state, energy, mag = mcmh_algorithm(state, beta) # Update values if step < relaxation: # Relaxation continue elif step == relaxation: # init values avg_energy = energy/num_iter avg_mag = mag/num_iter sqd_energy = energy**2./num_iter sqd_mag = mag**2./num_iter else: avg_energy += energy/num_iter avg_mag += mag/num_iter sqd_energy += energy**2./num_iter sqd_mag += mag**2./num_iter specific_heat = beta*(sqd_energy - avg_energy**2.) magnetic_susceptibility = beta**2.*(sqd_mag - avg_mag**2.)
[ "numpy.exp", "numpy.sum", "numpy.random.randint", "numpy.random.rand" ]
[((100, 141), 'numpy.random.randint', 'np.random.randint', (['(2)'], {'size': '(nsize, nsize)'}), '(2, size=(nsize, nsize))\n', (117, 141), True, 'import numpy as np\n'), ((1408, 1421), 'numpy.sum', 'np.sum', (['state'], {}), '(state)\n', (1414, 1421), True, 'import numpy as np\n'), ((1246, 1262), 'numpy.random.rand', 'np.random.rand', ([], {}), '()\n', (1260, 1262), True, 'import numpy as np\n'), ((1265, 1288), 'numpy.exp', 'np.exp', (['(-beta * delta_e)'], {}), '(-beta * delta_e)\n', (1271, 1288), True, 'import numpy as np\n')]
"""Utility method to draw the eye contact""" import cv2 as cv import numpy as np def draw_contact(image_in, face, gaze_vector): image_out = image_in if (is_contact(gaze_vector)): #Draw green cv.rectangle( image_out, tuple(np.round(face[:2]).astype(np.int32)), tuple(np.round(np.add(face[:2], face[2:])).astype(np.int32)), color=(0, 255, 0), thickness=1, lineType=cv.LINE_AA, ) else: #Draw red cv.rectangle( image_out, tuple(np.round(face[:2]).astype(np.int32)), tuple(np.round(np.add(face[:2], face[2:])).astype(np.int32)), color=(0, 0, 255), thickness=1, lineType=cv.LINE_AA, ) return image_out def is_contact(gaze_vector): #The eye contact is defined as looking within a 3° range around the camera #This seems to be a good compromise between accurracy and realism in live testing max_angle = np.radians(3) phi = np.absolute(gaze_vector[1]) theta = np.absolute(gaze_vector[0]) return phi < max_angle and theta < max_angle
[ "numpy.radians", "numpy.absolute", "numpy.add", "numpy.round" ]
[((1033, 1046), 'numpy.radians', 'np.radians', (['(3)'], {}), '(3)\n', (1043, 1046), True, 'import numpy as np\n'), ((1060, 1087), 'numpy.absolute', 'np.absolute', (['gaze_vector[1]'], {}), '(gaze_vector[1])\n', (1071, 1087), True, 'import numpy as np\n'), ((1101, 1128), 'numpy.absolute', 'np.absolute', (['gaze_vector[0]'], {}), '(gaze_vector[0])\n', (1112, 1128), True, 'import numpy as np\n'), ((267, 285), 'numpy.round', 'np.round', (['face[:2]'], {}), '(face[:2])\n', (275, 285), True, 'import numpy as np\n'), ((563, 581), 'numpy.round', 'np.round', (['face[:2]'], {}), '(face[:2])\n', (571, 581), True, 'import numpy as np\n'), ((341, 367), 'numpy.add', 'np.add', (['face[:2]', 'face[2:]'], {}), '(face[:2], face[2:])\n', (347, 367), True, 'import numpy as np\n'), ((637, 663), 'numpy.add', 'np.add', (['face[:2]', 'face[2:]'], {}), '(face[:2], face[2:])\n', (643, 663), True, 'import numpy as np\n')]
# Copyright 2018 Google LLC # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # https://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Minimal data reader for GQN TFRecord datasets.""" import collections import os import tensorflow as tf import numpy as np from PIL import Image DatasetInfo = collections.namedtuple( 'DatasetInfo', ['basepath', 'train_size', 'test_size', 'frame_size', 'sequence_size'] ) Context = collections.namedtuple('Context', ['frames', 'cameras']) Query = collections.namedtuple('Query', ['context', 'query_camera']) TaskData = collections.namedtuple('TaskData', ['query', 'target']) _DATASETS = dict( jaco=DatasetInfo( basepath='jaco', train_size=3600, test_size=400, frame_size=64, sequence_size=11), mazes=DatasetInfo( basepath='mazes', train_size=1080, test_size=120, frame_size=84, sequence_size=300), rooms_free_camera_with_object_rotations=DatasetInfo( basepath='rooms_free_camera_with_object_rotations', train_size=2034, test_size=226, frame_size=128, sequence_size=10), rooms_ring_camera=DatasetInfo( basepath='rooms_ring_camera', train_size=2160, test_size=240, frame_size=64, sequence_size=10), rooms_free_camera_no_object_rotations=DatasetInfo( basepath='rooms_free_camera_no_object_rotations', train_size=2160, test_size=240, frame_size=64, sequence_size=10), shepard_metzler_5_parts=DatasetInfo( basepath='shepard_metzler_5_parts', train_size=900, test_size=100, frame_size=64, sequence_size=15), shepard_metzler_7_parts=DatasetInfo( basepath='shepard_metzler_7_parts', train_size=900, test_size=100, frame_size=64, sequence_size=15) ) _NUM_CHANNELS = 3 _NUM_RAW_CAMERA_PARAMS = 5 def _get_dataset_files(dateset_info, mode, root): """Generates lists of files for a given dataset version.""" basepath = dateset_info.basepath base = os.path.join(root, basepath, mode) if mode == 'train': num_files = dateset_info.train_size else: num_files = dateset_info.test_size length = len(str(num_files)) template = '{:0%d}-of-{:0%d}.tfrecord' % (length, length) return [os.path.join(base, template.format(i + 1, num_files)) for i in range(num_files)] def _convert_frame_data(jpeg_data): decoded_frames = tf.image.decode_jpeg(jpeg_data) return tf.image.convert_image_dtype(decoded_frames, dtype=tf.float32) def _get_randomized_indices(dataset_info, example_size): """Generates randomized indices into a sequence of a specific length.""" indices = tf.range(0, dataset_info.sequence_size) indices = tf.random_shuffle(indices) indices = tf.slice(indices, begin=[0], size=[example_size]) return indices def _preprocess_frames(example, indices, dataset_info, example_size, custom_frame_size): """Instantiates the ops used to preprocess the frames data.""" frames = tf.concat(example['frames'], axis=0) frames = tf.gather(frames, indices, axis=0) frames = tf.map_fn( _convert_frame_data, tf.reshape(frames, [-1]), dtype=tf.float32, back_prop=False) dataset_image_dimensions = tuple( [dataset_info.frame_size] * 2 + [_NUM_CHANNELS]) # tf.Print(tf.shape(frames), [tf.shape(frames)], "Shape: ") frames = tf.reshape( frames, (-1, example_size) + dataset_image_dimensions) # tf.Print(tf.shape(frames), [tf.shape(frames)], "Shape: ") if (custom_frame_size and custom_frame_size != dataset_info.frame_size): frames = tf.reshape(frames, (-1,) + dataset_image_dimensions) new_frame_dimensions = (custom_frame_size,) * 2 + (_NUM_CHANNELS,) frames = tf.image.resize_bilinear( frames, new_frame_dimensions[:2], align_corners=True) frames = tf.reshape( frames, (-1, example_size) + new_frame_dimensions) return frames def _preprocess_cameras(example, indices, dataset_info): """Instantiates the ops used to preprocess the cameras data.""" raw_pose_params = example['cameras'] raw_pose_params = tf.reshape( raw_pose_params, [-1, dataset_info.sequence_size, _NUM_RAW_CAMERA_PARAMS]) raw_pose_params = tf.gather(raw_pose_params, indices, axis=1) pos = raw_pose_params[:, :, 0:3] yaw = raw_pose_params[:, :, 3:4] pitch = raw_pose_params[:, :, 4:5] cameras = tf.concat( [pos, yaw, pitch], axis=2) return cameras def _parse_function(example_proto, dataset_info, example_size, custom_frame_size): feature_map = { 'frames': tf.FixedLenFeature( shape=dataset_info.sequence_size, dtype=tf.string), 'cameras': tf.FixedLenFeature( shape=[dataset_info.sequence_size * _NUM_RAW_CAMERA_PARAMS], dtype=tf.float32) } example = tf.parse_single_example(example_proto, feature_map) indices = _get_randomized_indices(dataset_info, example_size) frames = _preprocess_frames(example, indices, dataset_info, example_size, custom_frame_size) cameras = _preprocess_cameras(example, indices, dataset_info) return frames, cameras def make_dataset(dataset, root, context_size=5, mode='train', custom_frame_size=None, load_all=False): dataset_info = _DATASETS[dataset] file_names = _get_dataset_files(dataset_info, mode, root) dataset = tf.data.TFRecordDataset(file_names) if load_all: context_size = dataset_info.sequence_size - 1 def parse_func(example_proto): return _parse_function(example_proto, dataset_info=dataset_info, example_size=context_size + 1, custom_frame_size=custom_frame_size) dataset = dataset.map(parse_func) dataset = dataset.repeat(1) return dataset class DatasetWriter: def __init__(self, dataset, mode, root): """ Writes images to files, and camera info csv """ self.dataset_info = _DATASETS[dataset] self.mode = mode self.root = root self.counter = 0 # csv_header = "scene,view,x,y,z,yaw,pitch" path = os.path.join(self.root, self.dataset_info.basepath, self.mode) os.makedirs(path, exist_ok=True) self.meta_file = os.path.join(path, "info.meta") # self.csv = os.path.join(path, "info.csv") # with open(self.csv, 'w+') as f: # f.write(csv_header) def save_multiple(self, records): img_dir = os.path.join(self.root, self.dataset_info.basepath, self.mode) frames = [] cameras = [] for rec_frames, rec_cameras in records: rec_frames = np.squeeze(rec_frames) rec_cameras = np.squeeze(rec_cameras) frames.append(rec_frames) cameras.append(rec_cameras) frames = np.array(frames) cameras = np.array(cameras) # np.savez(os.path.join(img_dir, "record-{}.npy".format(self.counter + 1)), frames=frames, cameras=cameras) np.savez_compressed(os.path.join(img_dir, "record-{}.npz".format(self.counter + 1)), frames=frames, cameras=cameras) self.counter += 1 if self.mode == 'train': num_files = 2e6 * 9 / 10 else: num_files = 2e6 * 1 / 10 if self.counter % 1000 == 0: print( "{}% complete".format(self.counter * 100 / num_files)) def save_record(self, record): # image1, context1, image2, context2, ..., imageN, contextN # context1 = x1, y1, z1, yaw1, pitch1 frames, cameras = record img_dir = os.path.join(self.root, self.dataset_info.basepath, self.mode) try: frames = np.squeeze(frames) cameras = np.squeeze(cameras) except ValueError: pass if self.mode == 'train': num_files = 2e6 * 9 / 10 else: num_files = 2e6 * 1 / 10 rows = [] if self.counter % 1000 == 0: print("{}% complete".format(self.counter * 100 / num_files)) self.counter += 1
[ "tensorflow.data.TFRecordDataset", "tensorflow.slice", "collections.namedtuple", "tensorflow.image.convert_image_dtype", "os.makedirs", "tensorflow.parse_single_example", "tensorflow.image.resize_bilinear", "os.path.join", "tensorflow.FixedLenFeature", "tensorflow.random_shuffle", "tensorflow.ra...
[((739, 852), 'collections.namedtuple', 'collections.namedtuple', (['"""DatasetInfo"""', "['basepath', 'train_size', 'test_size', 'frame_size', 'sequence_size']"], {}), "('DatasetInfo', ['basepath', 'train_size',\n 'test_size', 'frame_size', 'sequence_size'])\n", (761, 852), False, 'import collections\n'), ((869, 925), 'collections.namedtuple', 'collections.namedtuple', (['"""Context"""', "['frames', 'cameras']"], {}), "('Context', ['frames', 'cameras'])\n", (891, 925), False, 'import collections\n'), ((934, 994), 'collections.namedtuple', 'collections.namedtuple', (['"""Query"""', "['context', 'query_camera']"], {}), "('Query', ['context', 'query_camera'])\n", (956, 994), False, 'import collections\n'), ((1006, 1061), 'collections.namedtuple', 'collections.namedtuple', (['"""TaskData"""', "['query', 'target']"], {}), "('TaskData', ['query', 'target'])\n", (1028, 1061), False, 'import collections\n'), ((2547, 2581), 'os.path.join', 'os.path.join', (['root', 'basepath', 'mode'], {}), '(root, basepath, mode)\n', (2559, 2581), False, 'import os\n'), ((2941, 2972), 'tensorflow.image.decode_jpeg', 'tf.image.decode_jpeg', (['jpeg_data'], {}), '(jpeg_data)\n', (2961, 2972), True, 'import tensorflow as tf\n'), ((2982, 3044), 'tensorflow.image.convert_image_dtype', 'tf.image.convert_image_dtype', (['decoded_frames'], {'dtype': 'tf.float32'}), '(decoded_frames, dtype=tf.float32)\n', (3010, 3044), True, 'import tensorflow as tf\n'), ((3194, 3233), 'tensorflow.range', 'tf.range', (['(0)', 'dataset_info.sequence_size'], {}), '(0, dataset_info.sequence_size)\n', (3202, 3233), True, 'import tensorflow as tf\n'), ((3248, 3274), 'tensorflow.random_shuffle', 'tf.random_shuffle', (['indices'], {}), '(indices)\n', (3265, 3274), True, 'import tensorflow as tf\n'), ((3289, 3338), 'tensorflow.slice', 'tf.slice', (['indices'], {'begin': '[0]', 'size': '[example_size]'}), '(indices, begin=[0], size=[example_size])\n', (3297, 3338), True, 'import tensorflow as tf\n'), ((3528, 3564), 'tensorflow.concat', 'tf.concat', (["example['frames']"], {'axis': '(0)'}), "(example['frames'], axis=0)\n", (3537, 3564), True, 'import tensorflow as tf\n'), ((3578, 3612), 'tensorflow.gather', 'tf.gather', (['frames', 'indices'], {'axis': '(0)'}), '(frames, indices, axis=0)\n', (3587, 3612), True, 'import tensorflow as tf\n'), ((3907, 3972), 'tensorflow.reshape', 'tf.reshape', (['frames', '((-1, example_size) + dataset_image_dimensions)'], {}), '(frames, (-1, example_size) + dataset_image_dimensions)\n', (3917, 3972), True, 'import tensorflow as tf\n'), ((4673, 4762), 'tensorflow.reshape', 'tf.reshape', (['raw_pose_params', '[-1, dataset_info.sequence_size, _NUM_RAW_CAMERA_PARAMS]'], {}), '(raw_pose_params, [-1, dataset_info.sequence_size,\n _NUM_RAW_CAMERA_PARAMS])\n', (4683, 4762), True, 'import tensorflow as tf\n'), ((4798, 4841), 'tensorflow.gather', 'tf.gather', (['raw_pose_params', 'indices'], {'axis': '(1)'}), '(raw_pose_params, indices, axis=1)\n', (4807, 4841), True, 'import tensorflow as tf\n'), ((4969, 5005), 'tensorflow.concat', 'tf.concat', (['[pos, yaw, pitch]'], {'axis': '(2)'}), '([pos, yaw, pitch], axis=2)\n', (4978, 5005), True, 'import tensorflow as tf\n'), ((5402, 5453), 'tensorflow.parse_single_example', 'tf.parse_single_example', (['example_proto', 'feature_map'], {}), '(example_proto, feature_map)\n', (5425, 5453), True, 'import tensorflow as tf\n'), ((5928, 5963), 'tensorflow.data.TFRecordDataset', 'tf.data.TFRecordDataset', (['file_names'], {}), '(file_names)\n', (5951, 5963), True, 'import tensorflow as tf\n'), ((3666, 3690), 'tensorflow.reshape', 'tf.reshape', (['frames', '[-1]'], {}), '(frames, [-1])\n', (3676, 3690), True, 'import tensorflow as tf\n'), ((4147, 4199), 'tensorflow.reshape', 'tf.reshape', (['frames', '((-1,) + dataset_image_dimensions)'], {}), '(frames, (-1,) + dataset_image_dimensions)\n', (4157, 4199), True, 'import tensorflow as tf\n'), ((4288, 4366), 'tensorflow.image.resize_bilinear', 'tf.image.resize_bilinear', (['frames', 'new_frame_dimensions[:2]'], {'align_corners': '(True)'}), '(frames, new_frame_dimensions[:2], align_corners=True)\n', (4312, 4366), True, 'import tensorflow as tf\n'), ((4393, 4454), 'tensorflow.reshape', 'tf.reshape', (['frames', '((-1, example_size) + new_frame_dimensions)'], {}), '(frames, (-1, example_size) + new_frame_dimensions)\n', (4403, 4454), True, 'import tensorflow as tf\n'), ((5156, 5225), 'tensorflow.FixedLenFeature', 'tf.FixedLenFeature', ([], {'shape': 'dataset_info.sequence_size', 'dtype': 'tf.string'}), '(shape=dataset_info.sequence_size, dtype=tf.string)\n', (5174, 5225), True, 'import tensorflow as tf\n'), ((5259, 5360), 'tensorflow.FixedLenFeature', 'tf.FixedLenFeature', ([], {'shape': '[dataset_info.sequence_size * _NUM_RAW_CAMERA_PARAMS]', 'dtype': 'tf.float32'}), '(shape=[dataset_info.sequence_size *\n _NUM_RAW_CAMERA_PARAMS], dtype=tf.float32)\n', (5277, 5360), True, 'import tensorflow as tf\n'), ((6633, 6695), 'os.path.join', 'os.path.join', (['self.root', 'self.dataset_info.basepath', 'self.mode'], {}), '(self.root, self.dataset_info.basepath, self.mode)\n', (6645, 6695), False, 'import os\n'), ((6704, 6736), 'os.makedirs', 'os.makedirs', (['path'], {'exist_ok': '(True)'}), '(path, exist_ok=True)\n', (6715, 6736), False, 'import os\n'), ((6762, 6793), 'os.path.join', 'os.path.join', (['path', '"""info.meta"""'], {}), "(path, 'info.meta')\n", (6774, 6793), False, 'import os\n'), ((6992, 7054), 'os.path.join', 'os.path.join', (['self.root', 'self.dataset_info.basepath', 'self.mode'], {}), '(self.root, self.dataset_info.basepath, self.mode)\n', (7004, 7054), False, 'import os\n'), ((7337, 7353), 'numpy.array', 'np.array', (['frames'], {}), '(frames)\n', (7345, 7353), True, 'import numpy as np\n'), ((7372, 7389), 'numpy.array', 'np.array', (['cameras'], {}), '(cameras)\n', (7380, 7389), True, 'import numpy as np\n'), ((8095, 8157), 'os.path.join', 'os.path.join', (['self.root', 'self.dataset_info.basepath', 'self.mode'], {}), '(self.root, self.dataset_info.basepath, self.mode)\n', (8107, 8157), False, 'import os\n'), ((7169, 7191), 'numpy.squeeze', 'np.squeeze', (['rec_frames'], {}), '(rec_frames)\n', (7179, 7191), True, 'import numpy as np\n'), ((7218, 7241), 'numpy.squeeze', 'np.squeeze', (['rec_cameras'], {}), '(rec_cameras)\n', (7228, 7241), True, 'import numpy as np\n'), ((8192, 8210), 'numpy.squeeze', 'np.squeeze', (['frames'], {}), '(frames)\n', (8202, 8210), True, 'import numpy as np\n'), ((8233, 8252), 'numpy.squeeze', 'np.squeeze', (['cameras'], {}), '(cameras)\n', (8243, 8252), True, 'import numpy as np\n')]
import joblib import pytest from pydefect.analyzer.grids import Grids from pymatgen.core import Lattice, Structure import numpy as np from pymatgen.io.vasp import Chgcar from vise.tests.helpers.assertion import assert_dataclass_almost_equal @pytest.fixture def grids(): return Grids(lattice=Lattice.cubic(10), dim=(1, 1, 5), distance_data=np.array([[[0.0, 2.0, 4.0, 4.0, 2.0]]])) @pytest.fixture def chgcar(): struc = Structure(lattice=Lattice.cubic(10), species=["H"], coords=[[0]*3]) data = {"total": np.array([[[0.0, 2.0, 4.0, 6.0, 8.0]]]), "diff": np.array([[[0.0, 1.0, 2.0, 3.0, 4.0]]])} return Chgcar(struc, data=data) def test_grids_joblib_roundtrip(tmpdir, grids): tmpdir.chdir() print(tmpdir) with open("tmp.joblib", mode="wb") as f: joblib.dump(grids, f, compress=3) with open("tmp.joblib", mode="rb") as f: actual = joblib.load(f) assert_dataclass_almost_equal(actual, grids) def test_grids_np_save_load_roundtrip(tmpdir, grids): tmpdir.chdir() print(tmpdir) grids.dump() actual = grids.from_file() assert_dataclass_almost_equal(actual, grids) def test_grids_from_chgcar(grids, chgcar): actual = Grids.from_chgcar(chgcar) assert_dataclass_almost_equal(actual, grids) def test_shift_distance_data(grids): actual = grids.shifted_distance_data(center=[0, 0, 1]) expected = np.array([[[2.0, 0.0, 2.0, 4.0, 4.0]]]) np.testing.assert_array_almost_equal(actual, expected) def test_shift_distance_data2(): grids = Grids(lattice=Lattice.cubic(10), dim=(2, 2, 2), distance_data=np.array([[[0.0, 5.0], [5.0, 7.07]], [[5.0, 7.07], [7.07, 8.66]]])) actual = grids.shifted_distance_data(center=[1, 1, 1]) expected = np.array([[[8.66, 7.07], [7.07, 5.0]], [[7.07, 5.0], [5.0, 0.0]]]) np.testing.assert_array_almost_equal(actual, expected) def test_spherical_dist(grids): # distance_data=np.array([[[0.0, 2.0, 4.0, 4.0, 2.0]]])) actual = grids.spherical_dist(data=np.array([[[0.0, 0.0, 0.0, 0.0, 1.0]]]), center=[0, 0, 1], distance_bins=np.array([0.0, 2.5, 5.0])) # Divide by 2 since there are 2 points at 4.0 distance. volume=1000. expected = [0.0, 1.0 / 2 / 1000] assert actual == expected """ TODO - Check how to revolve numpy array. - Add distances_data - Add _calc_histogram(chgcar, distances_data, center) method DONE - Add defect_center_idxs - Add defect_center_coords property """
[ "numpy.testing.assert_array_almost_equal", "vise.tests.helpers.assertion.assert_dataclass_almost_equal", "numpy.array", "pymatgen.core.Lattice.cubic", "joblib.load", "pydefect.analyzer.grids.Grids.from_chgcar", "joblib.dump", "pymatgen.io.vasp.Chgcar" ]
[((666, 690), 'pymatgen.io.vasp.Chgcar', 'Chgcar', (['struc'], {'data': 'data'}), '(struc, data=data)\n', (672, 690), False, 'from pymatgen.io.vasp import Chgcar\n'), ((946, 990), 'vise.tests.helpers.assertion.assert_dataclass_almost_equal', 'assert_dataclass_almost_equal', (['actual', 'grids'], {}), '(actual, grids)\n', (975, 990), False, 'from vise.tests.helpers.assertion import assert_dataclass_almost_equal\n'), ((1136, 1180), 'vise.tests.helpers.assertion.assert_dataclass_almost_equal', 'assert_dataclass_almost_equal', (['actual', 'grids'], {}), '(actual, grids)\n', (1165, 1180), False, 'from vise.tests.helpers.assertion import assert_dataclass_almost_equal\n'), ((1239, 1264), 'pydefect.analyzer.grids.Grids.from_chgcar', 'Grids.from_chgcar', (['chgcar'], {}), '(chgcar)\n', (1256, 1264), False, 'from pydefect.analyzer.grids import Grids\n'), ((1269, 1313), 'vise.tests.helpers.assertion.assert_dataclass_almost_equal', 'assert_dataclass_almost_equal', (['actual', 'grids'], {}), '(actual, grids)\n', (1298, 1313), False, 'from vise.tests.helpers.assertion import assert_dataclass_almost_equal\n'), ((1427, 1466), 'numpy.array', 'np.array', (['[[[2.0, 0.0, 2.0, 4.0, 4.0]]]'], {}), '([[[2.0, 0.0, 2.0, 4.0, 4.0]]])\n', (1435, 1466), True, 'import numpy as np\n'), ((1471, 1525), 'numpy.testing.assert_array_almost_equal', 'np.testing.assert_array_almost_equal', (['actual', 'expected'], {}), '(actual, expected)\n', (1507, 1525), True, 'import numpy as np\n'), ((1855, 1921), 'numpy.array', 'np.array', (['[[[8.66, 7.07], [7.07, 5.0]], [[7.07, 5.0], [5.0, 0.0]]]'], {}), '([[[8.66, 7.07], [7.07, 5.0]], [[7.07, 5.0], [5.0, 0.0]]])\n', (1863, 1921), True, 'import numpy as np\n'), ((1951, 2005), 'numpy.testing.assert_array_almost_equal', 'np.testing.assert_array_almost_equal', (['actual', 'expected'], {}), '(actual, expected)\n', (1987, 2005), True, 'import numpy as np\n'), ((553, 592), 'numpy.array', 'np.array', (['[[[0.0, 2.0, 4.0, 6.0, 8.0]]]'], {}), '([[[0.0, 2.0, 4.0, 6.0, 8.0]]])\n', (561, 592), True, 'import numpy as np\n'), ((614, 653), 'numpy.array', 'np.array', (['[[[0.0, 1.0, 2.0, 3.0, 4.0]]]'], {}), '([[[0.0, 1.0, 2.0, 3.0, 4.0]]])\n', (622, 653), True, 'import numpy as np\n'), ((831, 864), 'joblib.dump', 'joblib.dump', (['grids', 'f'], {'compress': '(3)'}), '(grids, f, compress=3)\n', (842, 864), False, 'import joblib\n'), ((927, 941), 'joblib.load', 'joblib.load', (['f'], {}), '(f)\n', (938, 941), False, 'import joblib\n'), ((297, 314), 'pymatgen.core.Lattice.cubic', 'Lattice.cubic', (['(10)'], {}), '(10)\n', (310, 314), False, 'from pymatgen.core import Lattice, Structure\n'), ((379, 418), 'numpy.array', 'np.array', (['[[[0.0, 2.0, 4.0, 4.0, 2.0]]]'], {}), '([[[0.0, 2.0, 4.0, 4.0, 2.0]]])\n', (387, 418), True, 'import numpy as np\n'), ((482, 499), 'pymatgen.core.Lattice.cubic', 'Lattice.cubic', (['(10)'], {}), '(10)\n', (495, 499), False, 'from pymatgen.core import Lattice, Structure\n'), ((1587, 1604), 'pymatgen.core.Lattice.cubic', 'Lattice.cubic', (['(10)'], {}), '(10)\n', (1600, 1604), False, 'from pymatgen.core import Lattice, Structure\n'), ((1671, 1737), 'numpy.array', 'np.array', (['[[[0.0, 5.0], [5.0, 7.07]], [[5.0, 7.07], [7.07, 8.66]]]'], {}), '([[[0.0, 5.0], [5.0, 7.07]], [[5.0, 7.07], [7.07, 8.66]]])\n', (1679, 1737), True, 'import numpy as np\n'), ((2140, 2179), 'numpy.array', 'np.array', (['[[[0.0, 0.0, 0.0, 0.0, 1.0]]]'], {}), '([[[0.0, 0.0, 0.0, 0.0, 1.0]]])\n', (2148, 2179), True, 'import numpy as np\n'), ((2281, 2306), 'numpy.array', 'np.array', (['[0.0, 2.5, 5.0]'], {}), '([0.0, 2.5, 5.0])\n', (2289, 2306), True, 'import numpy as np\n')]
import torch import random import numpy as np import torch.nn.functional as F from .min_norm_solvers import MinNormSolver, gradient_normalizers from torch.autograd import Variable class backprop_scheduler(object): def __init__(self, model, mode=None): self.model = model self.mode = mode self.num_worker = len(self.model.regression_workers) + len(self.model.classification_workers) self.Q = torch.zeros(self.num_worker).detach() self.last_loss = torch.zeros(self.num_worker).detach() self.pi = torch.ones(self.num_worker).detach() def __call__(self, preds, label, cls_optim, regr_optim, frontend_optim, device, h=None, dropout_rate=None, delta=None, temperture=None, alpha=None, batch=None): if self.mode == "base": return self._base_scheduler(preds, label, cls_optim, regr_optim, frontend_optim, device) elif self.mode == "adversarial": return self._adversarial(preds, label, cls_optim, regr_optim, frontend_optim, device) elif self.mode == "select_one": return self._select_one(preds, label, cls_optim, regr_optim, frontend_optim, device) elif self.mode == "select_half": return self._select_half(preds, label, cls_optim, regr_optim, frontend_optim, device) elif self.mode == "dropout": return self._drop_out(preds, label, cls_optim, regr_optim, frontend_optim, device=device, dropout_rate=dropout_rate) elif self.mode == "hyper_volume": return self._hyper_volume(preds, label, cls_optim, regr_optim, frontend_optim, device=device, delta=delta) elif self.mode == "softmax": return self._softmax(preds, label, cls_optim, regr_optim, frontend_optim, temperture=temperture, device=device) elif self.mode == "adaptive": return self._online_adaptive(preds, label, cls_optim, regr_optim, frontend_optim, temperture=temperture, alpha=alpha, device=device) elif self.mode == "MGD": return self._MGDA(preds, label, cls_optim, regr_optim, frontend_optim, batch=batch, device=device) else: raise NotImplementedError def _base_scheduler(self, preds, label, cls_optim, regr_optim, frontend_optim, device): frontend_optim.zero_grad() tot_loss = 0 losses = {} for worker in self.model.classification_workers: cls_optim[worker.name].zero_grad() loss = worker.loss_weight * worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss tot_loss += loss for worker in self.model.regression_workers: regr_optim[worker.name].zero_grad() loss = worker.loss_weight * worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss tot_loss += loss for worker in self.model.regularizer_workers: loss = worker.loss_weight * worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss tot_loss += loss tot_loss.backward() for _, optim in cls_optim.items(): optim.step() for _, optim in regr_optim.items(): optim.step() frontend_optim.step() losses["total"] = tot_loss return losses, 1 def _select_one(self, preds, label, cls_optim, regr_optim, frontend_optim, device): self.count += 1 loss_lst = [] num_worker = len(self.model.regression_workers) + len(self.model.classification_workers) frontend_optim.zero_grad() losses = {} selected = self.count % num_worker # select one if selected > 3: worker = self.model.classification_workers[selected - 4] loss = worker.loss(preds[worker.name], label[worker.name]) else: worker = self.model.classification_workers[selected] loss = worker.loss(preds[worker.name], label[worker.name]) tot_loss = loss tot_loss.backward() for _, optim in cls_optim.items(): optim.step() for _, optim in regr_optim.items(): optim.step() frontend_optim.step() losses["total"] = tot_loss return losses, 1 def _select_half(self, preds, label, cls_optim, regr_optim, frontend_optim, device): num_worker = len(self.model.regression_workers) + len(self.model.classification_workers) loss_tmp = torch.zeros(num_worker).to(device) idx = 0 frontend_optim.zero_grad() losses = {} for worker in self.model.classification_workers: cls_optim[worker.name].zero_grad() loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss loss_tmp[idx] = loss idx += 1 for worker in self.model.regression_workers: regr_optim[worker.name].zero_grad() loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss loss_tmp[idx] = loss idx += 1 # generate mask mask = np.random.randint(2, size=num_worker) while np.sum(mask) > 4 or np.sum(mask) < 3: mask = np.random.randint(2, size=num_worker) mask = torch.from_numpy(mask).type(torch.FloatTensor).to(device) #sum up losses tot_loss = torch.sum(mask * loss_tmp, dim=0) tot_loss.backward() for _, optim in cls_optim.items(): optim.step() for _, optim in regr_optim.items(): optim.step() frontend_optim.step() losses["total"] = tot_loss return losses, 1 def _drop_out(self, preds, label, cls_optim, regr_optim, frontend_optim, dropout_rate, device): loss_tmp = torch.zeros(7, requires_grad=True).to(device) idx = 0 assert dropout_rate is not None re_mask = np.random.binomial(1, dropout_rate, size=len(self.model.regression_workers)) cls_mask = np.random.binomial(1, dropout_rate, size=len(self.model.classification_workers)) frontend_optim.zero_grad() losses = {} for i, worker in enumerate(self.model.classification_workers): cls_optim[worker.name].zero_grad() if cls_mask[i] == 1: loss = worker.loss(preds[worker.name], label[worker.name]) else: loss = 0 losses[worker.name] = loss loss_tmp[idx] = loss idx += 1 for worker in self.model.regression_workers: regr_optim[worker.name].zero_grad() if re_mask[i] == 1: loss = worker.loss(preds[worker.name], label[worker.name]) else: loss = 0 losses[worker.name] = loss loss_tmp[idx] = loss idx += 1 #sum up losses tot_loss = torch.sum(loss_tmp, dim=0) tot_loss.backward() for _, optim in cls_optim.items(): optim.step() for _, optim in regr_optim.items(): optim.step() frontend_optim.step() losses["total"] = tot_loss return losses, 1 def _hyper_volume(self, preds, label, cls_optim, regr_optim, frontend_optim, delta ,device): assert delta > 1 num_worker = len(self.model.regression_workers) + len(self.model.classification_workers) loss_tmp = torch.zeros(num_worker).to(device) idx = 0 frontend_optim.zero_grad() losses = {} for worker in self.model.classification_workers: cls_optim[worker.name].zero_grad() loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss loss_tmp[idx] = loss idx += 1 for worker in self.model.regression_workers: regr_optim[worker.name].zero_grad() loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss loss_tmp[idx] = loss idx += 1 #sum up losses eta = delta * torch.max(loss_tmp.detach()).item() hyper_votolume = torch.sum(loss_tmp) alpha = 1 / (eta - loss_tmp + 1e-6) hyper_votolume.backward() for _, optim in cls_optim.items(): optim.step() for _, optim in regr_optim.items(): optim.step() frontend_optim.step() losses["total"] = hyper_votolume return losses, alpha def _softmax(self, preds, label, cls_optim, regr_optim, frontend_optim, temperture, device): assert temperture > 0 num_worker = len(self.model.regression_workers) + len(self.model.classification_workers) loss_tmp = [] idx = 0 frontend_optim.zero_grad() losses = {} for worker in self.model.classification_workers: cls_optim[worker.name].zero_grad() loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss loss_tmp.append(loss.item() * temperture) # idx += 1 for worker in self.model.regression_workers: regr_optim[worker.name].zero_grad() loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss loss_tmp.append(loss.item() * temperture) # idx += 1 alpha = self._stable_softmax(loss_tmp) tot_loss = 0 for worker in self.model.classification_workers: # tot_loss += alpha[idx] * losses[worker.name] tot_loss += losses[worker.name] idx += 1 for worker in self.model.regression_workers: # tot_loss += alpha[idx] * losses[worker.name] tot_loss += losses[worker.name] idx += 1 # tot_loss = torch.sum(alpha.detach() * loss_vec) tot_loss.backward() for _, optim in cls_optim.items(): optim.step() for _, optim in regr_optim.items(): optim.step() frontend_optim.step() losses["total"] = tot_loss return losses, alpha def _online_adaptive(self, preds, label, cls_optim, regr_optim, frontend_optim, temperture, alpha, device): assert temperture > 0 and alpha > 0 # device = preds['chunk'].device num_worker = len(self.model.regression_workers) + len(self.model.classification_workers) loss_tmp = torch.zeros(num_worker).to(device) idx = 0 frontend_optim.zero_grad() losses = {} for worker in self.model.classification_workers: cls_optim[worker.name].zero_grad() loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss loss_tmp[idx] = loss idx += 1 for worker in self.model.regression_workers: regr_optim[worker.name].zero_grad() loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss loss_tmp[idx] = loss idx += 1 R_t = self.last_loss.to(device) - loss_tmp with torch.no_grad(): Q_t = alpha * R_t.detach() + (1 - alpha) * self.Q.to(device) self.pi = F.softmax(temperture * Q_t, dim=0) tot_loss = torch.sum(loss_tmp) tot_loss.backward() self.last_loss = loss_tmp.detach() self.Q = Q_t.detach() for _, optim in cls_optim.items(): optim.step() for _, optim in regr_optim.items(): optim.step() frontend_optim.step() losses["total"] = tot_loss return losses, self.pi def _MGDA(self, preds, label, cls_optim, regr_optim, frontend_optim, batch, device): frontend_optim.zero_grad() losses = {} grads = {} for worker in self.model.classification_workers: self.model.zero_grad() h, chunk, preds, labels = self.model.forward(batch, 1, device) # print(worker.name) loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss grads[worker.name] = self._get_gen_grads(loss) for worker in self.model.regression_workers: self.model.zero_grad() h, chunk, preds, labels = self.model.forward(batch, 1, device) # print(worker.name) loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss grads[worker.name] = self._get_gen_grads(loss) sol, min_norm = MinNormSolver.find_min_norm_element([grads[worker].unsqueeze(0) for worker, _ in grads.items()]) alpha = sol tot_loss = 0 # idx = 0 self.model.zero_grad() h, chunk, preds, labels = self.model.forward(batch, 1, device) for worker in self.model.classification_workers: loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss tot_loss += loss # tot_loss += sol[idx] * loss for worker in self.model.regression_workers: loss = worker.loss(preds[worker.name], label[worker.name]) losses[worker.name] = loss tot_loss += loss # tot_loss += sol[idx] * loss tot_loss.backward() for _, optim in cls_optim.items(): optim.step() for _, optim in regr_optim.items(): optim.step() frontend_optim.step() losses["total"] = tot_loss return losses, alpha def _get_gen_grads(self, loss_): # grads = torch.autograd.grad(outputs=loss_, inputs=self.model.frontend.parameters()) self.model.frontend.zero_grad() loss_.backward() # grads = self.model.frontend.grad() for params in self.model.frontend.parameters(): try: grads_ = torch.cat([grads_, params.grad.view(-1)], 0) except: grads_ = params.grad.view(-1) return grads_ / grads_.norm() def _stable_softmax(self, x): z = np.asarray(x, np.float) - np.max(x) numerator = np.exp(z) denominator = np.sum(numerator) softmax = numerator / denominator return softmax
[ "numpy.asarray", "torch.from_numpy", "numpy.max", "numpy.exp", "numpy.sum", "numpy.random.randint", "torch.zeros", "torch.sum", "torch.no_grad", "torch.nn.functional.softmax", "torch.ones" ]
[((5181, 5218), 'numpy.random.randint', 'np.random.randint', (['(2)'], {'size': 'num_worker'}), '(2, size=num_worker)\n', (5198, 5218), True, 'import numpy as np\n'), ((5444, 5477), 'torch.sum', 'torch.sum', (['(mask * loss_tmp)'], {'dim': '(0)'}), '(mask * loss_tmp, dim=0)\n', (5453, 5477), False, 'import torch\n'), ((6967, 6993), 'torch.sum', 'torch.sum', (['loss_tmp'], {'dim': '(0)'}), '(loss_tmp, dim=0)\n', (6976, 6993), False, 'import torch\n'), ((8243, 8262), 'torch.sum', 'torch.sum', (['loss_tmp'], {}), '(loss_tmp)\n', (8252, 8262), False, 'import torch\n'), ((11427, 11446), 'torch.sum', 'torch.sum', (['loss_tmp'], {}), '(loss_tmp)\n', (11436, 11446), False, 'import torch\n'), ((14309, 14318), 'numpy.exp', 'np.exp', (['z'], {}), '(z)\n', (14315, 14318), True, 'import numpy as np\n'), ((14341, 14358), 'numpy.sum', 'np.sum', (['numerator'], {}), '(numerator)\n', (14347, 14358), True, 'import numpy as np\n'), ((5290, 5327), 'numpy.random.randint', 'np.random.randint', (['(2)'], {'size': 'num_worker'}), '(2, size=num_worker)\n', (5307, 5327), True, 'import numpy as np\n'), ((11258, 11273), 'torch.no_grad', 'torch.no_grad', ([], {}), '()\n', (11271, 11273), False, 'import torch\n'), ((11371, 11405), 'torch.nn.functional.softmax', 'F.softmax', (['(temperture * Q_t)'], {'dim': '(0)'}), '(temperture * Q_t, dim=0)\n', (11380, 11405), True, 'import torch.nn.functional as F\n'), ((14253, 14276), 'numpy.asarray', 'np.asarray', (['x', 'np.float'], {}), '(x, np.float)\n', (14263, 14276), True, 'import numpy as np\n'), ((14279, 14288), 'numpy.max', 'np.max', (['x'], {}), '(x)\n', (14285, 14288), True, 'import numpy as np\n'), ((430, 458), 'torch.zeros', 'torch.zeros', (['self.num_worker'], {}), '(self.num_worker)\n', (441, 458), False, 'import torch\n'), ((493, 521), 'torch.zeros', 'torch.zeros', (['self.num_worker'], {}), '(self.num_worker)\n', (504, 521), False, 'import torch\n'), ((549, 576), 'torch.ones', 'torch.ones', (['self.num_worker'], {}), '(self.num_worker)\n', (559, 576), False, 'import torch\n'), ((4500, 4523), 'torch.zeros', 'torch.zeros', (['num_worker'], {}), '(num_worker)\n', (4511, 4523), False, 'import torch\n'), ((5233, 5245), 'numpy.sum', 'np.sum', (['mask'], {}), '(mask)\n', (5239, 5245), True, 'import numpy as np\n'), ((5253, 5265), 'numpy.sum', 'np.sum', (['mask'], {}), '(mask)\n', (5259, 5265), True, 'import numpy as np\n'), ((5858, 5892), 'torch.zeros', 'torch.zeros', (['(7)'], {'requires_grad': '(True)'}), '(7, requires_grad=True)\n', (5869, 5892), False, 'import torch\n'), ((7494, 7517), 'torch.zeros', 'torch.zeros', (['num_worker'], {}), '(num_worker)\n', (7505, 7517), False, 'import torch\n'), ((10551, 10574), 'torch.zeros', 'torch.zeros', (['num_worker'], {}), '(num_worker)\n', (10562, 10574), False, 'import torch\n'), ((5343, 5365), 'torch.from_numpy', 'torch.from_numpy', (['mask'], {}), '(mask)\n', (5359, 5365), False, 'import torch\n')]
import numpy as np class DualTask: def __init__(self, A: list, b: list, c: list, d_lo: list, d_hi: list): self.A: np.array = np.array(A) self.b: list = b self.c: list = c self.d_lo: np.array = np.array(d_lo) self.d_hi: np.array = np.array(d_hi) self.m, self.n = self.A.shape def remove(self, row, col): self.A = np.delete(self.A, row, axis=0) self.A = np.delete(self.A, col, axis=1) del self.b[row] del self.c[col] self.d_lo = np.delete(self.d_lo, col) self.d_hi = np.delete(self.d_hi, col) self.n -= 1 self.m -= 1 def extend(self, a, b, J): J_not_base = [j for j in range(self.n) if j not in J] restriction = np.zeros(self.n) restriction[J_not_base] = -a[J_not_base] self.n += 1 self.m += 1 self.A = np.vstack((self.A, restriction)) self.A = np.hstack((self.A, np.zeros((self.m, 1)))) self.A[-1, -1] = 1 self.b.append(-b) self.c.append(0) self.d_lo = np.append(self.d_lo, 0) self.d_hi = np.append(self.d_hi, 1e9)
[ "numpy.delete", "numpy.append", "numpy.array", "numpy.zeros", "numpy.vstack" ]
[((139, 150), 'numpy.array', 'np.array', (['A'], {}), '(A)\n', (147, 150), True, 'import numpy as np\n'), ((231, 245), 'numpy.array', 'np.array', (['d_lo'], {}), '(d_lo)\n', (239, 245), True, 'import numpy as np\n'), ((276, 290), 'numpy.array', 'np.array', (['d_hi'], {}), '(d_hi)\n', (284, 290), True, 'import numpy as np\n'), ((383, 413), 'numpy.delete', 'np.delete', (['self.A', 'row'], {'axis': '(0)'}), '(self.A, row, axis=0)\n', (392, 413), True, 'import numpy as np\n'), ((431, 461), 'numpy.delete', 'np.delete', (['self.A', 'col'], {'axis': '(1)'}), '(self.A, col, axis=1)\n', (440, 461), True, 'import numpy as np\n'), ((532, 557), 'numpy.delete', 'np.delete', (['self.d_lo', 'col'], {}), '(self.d_lo, col)\n', (541, 557), True, 'import numpy as np\n'), ((578, 603), 'numpy.delete', 'np.delete', (['self.d_hi', 'col'], {}), '(self.d_hi, col)\n', (587, 603), True, 'import numpy as np\n'), ((762, 778), 'numpy.zeros', 'np.zeros', (['self.n'], {}), '(self.n)\n', (770, 778), True, 'import numpy as np\n'), ((895, 927), 'numpy.vstack', 'np.vstack', (['(self.A, restriction)'], {}), '((self.A, restriction))\n', (904, 927), True, 'import numpy as np\n'), ((1088, 1111), 'numpy.append', 'np.append', (['self.d_lo', '(0)'], {}), '(self.d_lo, 0)\n', (1097, 1111), True, 'import numpy as np\n'), ((1132, 1166), 'numpy.append', 'np.append', (['self.d_hi', '(1000000000.0)'], {}), '(self.d_hi, 1000000000.0)\n', (1141, 1166), True, 'import numpy as np\n'), ((964, 985), 'numpy.zeros', 'np.zeros', (['(self.m, 1)'], {}), '((self.m, 1))\n', (972, 985), True, 'import numpy as np\n')]
import logging import random from abc import abstractmethod from typing import List, Callable, Tuple import matplotlib.pyplot as plt import numpy as np from matplotlib import colors from matplotlib.colors import LinearSegmentedColormap logger = logging.getLogger(__name__) class SAModel: """ Use case specific SA model that needs to be inherited by user of SARouteOptimizer. Contains methods used in SARouteOptimizer to - calculate cost for route - mutate route - calculate temperature - calculate probability from cost change and temperature """ @abstractmethod def cost(self, route: List[int]) -> float: """ Calculate cost for route. """ raise NotImplementedError @staticmethod def schedule(t: int, max_temperature: float = 1.0, decay_constant: float = 0.005) -> float: """ Calculate current temperature from iteration round t. """ return max_temperature * np.exp(-decay_constant * t) @staticmethod def probability(delta_cost: float, temperature: float, k: float = 1) -> float: """ Calculate acceptance probability for mutated route, based on cost change (vs. current solution) and temperature. """ if delta_cost < 0: return 1 else: return np.exp(-delta_cost / (k * temperature)) def mutate(self, input_route: List[int], mutation_probability: float = 0.2) -> List[int]: """ Mutate (modify) given route. This mutated route will be solution candidate that will be accepted or not, based on calculated probability. """ route = input_route.copy() for k in range(len(route)): if random.random() < mutation_probability: self._swap(route) # Make sure that at least one change is made to input route if route == input_route: self._swap(route) return route def _swap(self, route): i, j = random.sample([k + 1 for k in range(len(route) - 2)], 2) value_i = route[i] value_j = route[j] route[i] = value_j route[j] = value_i class SARouteOptimizer: """ Simulated annealing route optimizer. With give model and termination criteria, finds optimal route that minimizes cost function defined by model. """ def __init__(self, model: SAModel, max_iter: int = 10000, max_iter_without_improvement: int = 2000, min_temperature: float = 1e-12, cost_threshold: float = -np.inf, ): self.model = model self.max_iter = max_iter self.max_iter_without_improvement = max_iter_without_improvement self.min_temperature = min_temperature self.cost_threshold = cost_threshold self.temperatures = [] self.costs = [] self.probabilities = [] self.delta_costs = [] self.is_accepted = [] def run(self, init_route: List[int]) -> Tuple[List[int], float]: """ Find optimal route. :param init_route: Init guess for route. :return: optimal route, route cost """ current_route = init_route.copy() best_route = current_route.copy() current_cost = self.model.cost(current_route) best_cost = self.model.cost(best_route) probability, delta_cost = 1, 0 is_accepted = True no_improvement_counter = 0 for t in range(self.max_iter): no_improvement_counter += 1 temperature = self.model.schedule(t) if temperature < self.min_temperature: logger.info("Minimum temperature reached. Return solution") return best_route, best_cost self.temperatures.append(temperature) self.costs.append(current_cost) self.probabilities.append(probability) self.delta_costs.append(delta_cost) self.is_accepted.append(is_accepted) mutated_route = self.model.mutate(current_route.copy()) mutated_route_cost = self.model.cost(mutated_route) logger.debug(f"Mutated route: {mutated_route}; cost {mutated_route_cost}") delta_cost = mutated_route_cost - current_cost is_accepted = False probability = self.model.probability(delta_cost, temperature) if probability >= random.uniform(0.0, 1.0): is_accepted = True current_route = mutated_route.copy() current_cost = mutated_route_cost if current_cost < best_cost: best_route = current_route.copy() best_cost = current_cost logger.info(f"Found better solution; round {t}; cost {best_cost}") no_improvement_counter = 0 if best_cost < self.cost_threshold: logger.info("Cost reached required threshold value. Return solution.") return best_route, best_cost logger.debug(f"Round {t}: temperature {temperature}; cost {current_cost}") if no_improvement_counter > self.max_iter_without_improvement: logger.info("Max iteration number without improvement reached. Return solution.") return best_route, best_cost logger.info("Max iteration number reached. Return solution.") return best_route, best_cost def plot_solution(self): plt.figure(1) ax1 = plt.subplot(1, 2, 1) color = 'tab:blue' ax1.plot(self.costs, color=color) ax1.set_ylabel('Cost', color=color) color = 'tab:red' ax2 = ax1.twinx() ax2.plot(self.temperatures, color=color) ax2.set_ylabel('Temperature', color=color) ax1.set_xlabel("Iteration") plt.title("Cost & temperature") plt.subplot(1, 2, 2) sc = plt.scatter(self.temperatures, self.delta_costs, c=np.array(self.probabilities) + 0.001, norm=colors.LogNorm(), edgecolors="k", cmap=LinearSegmentedColormap.from_list("MyCmapName", ["b", "r"])) plt.colorbar(sc) plt.gca().invert_xaxis() plt.plot(np.array(self.temperatures)[self.is_accepted], np.array(self.delta_costs)[self.is_accepted], "kx", markersize=3) plt.xlabel("Temperature") plt.ylabel("Cost change") plt.title("Probability") plt.tight_layout() plt.show()
[ "logging.getLogger", "random.uniform", "matplotlib.pyplot.ylabel", "matplotlib.pyplot.gca", "matplotlib.pyplot.xlabel", "matplotlib.pyplot.colorbar", "matplotlib.colors.LinearSegmentedColormap.from_list", "numpy.exp", "numpy.array", "matplotlib.pyplot.figure", "matplotlib.pyplot.tight_layout", ...
[((247, 274), 'logging.getLogger', 'logging.getLogger', (['__name__'], {}), '(__name__)\n', (264, 274), False, 'import logging\n'), ((5581, 5594), 'matplotlib.pyplot.figure', 'plt.figure', (['(1)'], {}), '(1)\n', (5591, 5594), True, 'import matplotlib.pyplot as plt\n'), ((5610, 5630), 'matplotlib.pyplot.subplot', 'plt.subplot', (['(1)', '(2)', '(1)'], {}), '(1, 2, 1)\n', (5621, 5630), True, 'import matplotlib.pyplot as plt\n'), ((5943, 5974), 'matplotlib.pyplot.title', 'plt.title', (['"""Cost & temperature"""'], {}), "('Cost & temperature')\n", (5952, 5974), True, 'import matplotlib.pyplot as plt\n'), ((5984, 6004), 'matplotlib.pyplot.subplot', 'plt.subplot', (['(1)', '(2)', '(2)'], {}), '(1, 2, 2)\n', (5995, 6004), True, 'import matplotlib.pyplot as plt\n'), ((6345, 6361), 'matplotlib.pyplot.colorbar', 'plt.colorbar', (['sc'], {}), '(sc)\n', (6357, 6361), True, 'import matplotlib.pyplot as plt\n'), ((6585, 6610), 'matplotlib.pyplot.xlabel', 'plt.xlabel', (['"""Temperature"""'], {}), "('Temperature')\n", (6595, 6610), True, 'import matplotlib.pyplot as plt\n'), ((6619, 6644), 'matplotlib.pyplot.ylabel', 'plt.ylabel', (['"""Cost change"""'], {}), "('Cost change')\n", (6629, 6644), True, 'import matplotlib.pyplot as plt\n'), ((6653, 6677), 'matplotlib.pyplot.title', 'plt.title', (['"""Probability"""'], {}), "('Probability')\n", (6662, 6677), True, 'import matplotlib.pyplot as plt\n'), ((6687, 6705), 'matplotlib.pyplot.tight_layout', 'plt.tight_layout', ([], {}), '()\n', (6703, 6705), True, 'import matplotlib.pyplot as plt\n'), ((6714, 6724), 'matplotlib.pyplot.show', 'plt.show', ([], {}), '()\n', (6722, 6724), True, 'import matplotlib.pyplot as plt\n'), ((978, 1005), 'numpy.exp', 'np.exp', (['(-decay_constant * t)'], {}), '(-decay_constant * t)\n', (984, 1005), True, 'import numpy as np\n'), ((1334, 1373), 'numpy.exp', 'np.exp', (['(-delta_cost / (k * temperature))'], {}), '(-delta_cost / (k * temperature))\n', (1340, 1373), True, 'import numpy as np\n'), ((1733, 1748), 'random.random', 'random.random', ([], {}), '()\n', (1746, 1748), False, 'import random\n'), ((4482, 4506), 'random.uniform', 'random.uniform', (['(0.0)', '(1.0)'], {}), '(0.0, 1.0)\n', (4496, 4506), False, 'import random\n'), ((6187, 6203), 'matplotlib.colors.LogNorm', 'colors.LogNorm', ([], {}), '()\n', (6201, 6203), False, 'from matplotlib import colors\n'), ((6276, 6335), 'matplotlib.colors.LinearSegmentedColormap.from_list', 'LinearSegmentedColormap.from_list', (['"""MyCmapName"""', "['b', 'r']"], {}), "('MyCmapName', ['b', 'r'])\n", (6309, 6335), False, 'from matplotlib.colors import LinearSegmentedColormap\n'), ((6370, 6379), 'matplotlib.pyplot.gca', 'plt.gca', ([], {}), '()\n', (6377, 6379), True, 'import matplotlib.pyplot as plt\n'), ((6412, 6439), 'numpy.array', 'np.array', (['self.temperatures'], {}), '(self.temperatures)\n', (6420, 6439), True, 'import numpy as np\n'), ((6476, 6502), 'numpy.array', 'np.array', (['self.delta_costs'], {}), '(self.delta_costs)\n', (6484, 6502), True, 'import numpy as np\n'), ((6119, 6147), 'numpy.array', 'np.array', (['self.probabilities'], {}), '(self.probabilities)\n', (6127, 6147), True, 'import numpy as np\n')]
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. from __future__ import division import numpy as np from numba import jit from paddle import fluid import paddle.fluid.layers as F import paddle.fluid.dygraph as dg def masked_mean(inputs, mask): """ Args: inputs (Variable): shape(B, T, C), dtype float32, the input. mask (Variable): shape(B, T), dtype float32, a mask. Returns: loss (Variable): shape(1, ), dtype float32, masked mean. """ channels = inputs.shape[-1] masked_inputs = F.elementwise_mul(inputs, mask, axis=0) loss = F.reduce_sum(masked_inputs) / (channels * F.reduce_sum(mask)) return loss @jit(nopython=True) def guided_attention(N, max_N, T, max_T, g): """Generate an diagonal attention guide. Args: N (int): valid length of encoder. max_N (int): max length of encoder. T (int): valid length of decoder. max_T (int): max length of decoder. g (float): sigma to adjust the degree of diagonal guide. Returns: np.ndarray: shape(max_N, max_T), dtype float32, the diagonal guide. """ W = np.zeros((max_N, max_T), dtype=np.float32) for n in range(N): for t in range(T): W[n, t] = 1 - np.exp(-(n / N - t / T)**2 / (2 * g * g)) return W def guided_attentions(encoder_lengths, decoder_lengths, max_decoder_len, g=0.2): """Generate a diagonal attention guide for a batch. Args: encoder_lengths (np.ndarray): shape(B, ), dtype: int64, encoder valid lengths. decoder_lengths (np.ndarray): shape(B, ), dtype: int64, decoder valid lengths. max_decoder_len (int): max length of decoder. g (float, optional): sigma to adjust the degree of diagonal guide.. Defaults to 0.2. Returns: np.ndarray: shape(B, max_T, max_N), dtype float32, the diagonal guide. (max_N: max encoder length, max_T: max decoder length.) """ B = len(encoder_lengths) max_input_len = encoder_lengths.max() W = np.zeros((B, max_decoder_len, max_input_len), dtype=np.float32) for b in range(B): W[b] = guided_attention(encoder_lengths[b], max_input_len, decoder_lengths[b], max_decoder_len, g).T return W class TTSLoss(object): def __init__(self, masked_weight=0.0, priority_bin=None, priority_weight=0.0, binary_divergence_weight=0.0, guided_attention_sigma=0.2, downsample_factor=4, r=1): """Compute loss for Deep Voice 3 model. Args: masked_weight (float, optional): the weight of masked loss. Defaults to 0.0. priority_bin ([type], optional): frequency bands for linear spectrogram loss to be prioritized. Defaults to None. priority_weight (float, optional): weight for the prioritized frequency bands. Defaults to 0.0. binary_divergence_weight (float, optional): weight for binary cross entropy (used for spectrogram loss). Defaults to 0.0. guided_attention_sigma (float, optional): `sigma` for attention guide. Defaults to 0.2. downsample_factor (int, optional): the downsample factor for mel spectrogram. Defaults to 4. r (int, optional): frames per decoder step. Defaults to 1. """ self.masked_weight = masked_weight self.priority_bin = priority_bin # only used for lin-spec loss self.priority_weight = priority_weight # only used for lin-spec loss self.binary_divergence_weight = binary_divergence_weight self.guided_attention_sigma = guided_attention_sigma self.time_shift = r self.r = r self.downsample_factor = downsample_factor def l1_loss(self, prediction, target, mask, priority_bin=None): """L1 loss for spectrogram. Args: prediction (Variable): shape(B, T, C), dtype float32, predicted spectrogram. target (Variable): shape(B, T, C), dtype float32, target spectrogram. mask (Variable): shape(B, T), mask. priority_bin (int, optional): frequency bands for linear spectrogram loss to be prioritized. Defaults to None. Returns: Variable: shape(1,), dtype float32, l1 loss(with mask and possibly priority bin applied.) """ abs_diff = F.abs(prediction - target) # basic mask-weighted l1 loss w = self.masked_weight if w > 0 and mask is not None: base_l1_loss = w * masked_mean(abs_diff, mask) \ + (1 - w) * F.reduce_mean(abs_diff) else: base_l1_loss = F.reduce_mean(abs_diff) if self.priority_weight > 0 and priority_bin is not None: # mask-weighted priority channels' l1-loss priority_abs_diff = abs_diff[:, :, :priority_bin] if w > 0 and mask is not None: priority_loss = w * masked_mean(priority_abs_diff, mask) \ + (1 - w) * F.reduce_mean(priority_abs_diff) else: priority_loss = F.reduce_mean(priority_abs_diff) # priority weighted sum p = self.priority_weight loss = p * priority_loss + (1 - p) * base_l1_loss else: loss = base_l1_loss return loss def binary_divergence(self, prediction, target, mask): """Binary cross entropy loss for spectrogram. All the values in the spectrogram are treated as logits in a logistic regression. Args: prediction (Variable): shape(B, T, C), dtype float32, predicted spectrogram. target (Variable): shape(B, T, C), dtype float32, target spectrogram. mask (Variable): shape(B, T), mask. Returns: Variable: shape(1,), dtype float32, binary cross entropy loss. """ flattened_prediction = F.reshape(prediction, [-1, 1]) flattened_target = F.reshape(target, [-1, 1]) flattened_loss = F.log_loss( flattened_prediction, flattened_target, epsilon=1e-8) bin_div = fluid.layers.reshape(flattened_loss, prediction.shape) w = self.masked_weight if w > 0 and mask is not None: loss = w * masked_mean(bin_div, mask) \ + (1 - w) * F.reduce_mean(bin_div) else: loss = F.reduce_mean(bin_div) return loss @staticmethod def done_loss(done_hat, done): """Compute done loss Args: done_hat (Variable): shape(B, T), dtype float32, predicted done probability(the probability that the final frame has been generated.) done (Variable): shape(B, T), dtype float32, ground truth done probability(the probability that the final frame has been generated.) Returns: Variable: shape(1, ), dtype float32, done loss. """ flat_done_hat = F.reshape(done_hat, [-1, 1]) flat_done = F.reshape(done, [-1, 1]) loss = F.log_loss(flat_done_hat, flat_done, epsilon=1e-8) loss = F.reduce_mean(loss) return loss def attention_loss(self, predicted_attention, input_lengths, target_lengths): """ Given valid encoder_lengths and decoder_lengths, compute a diagonal guide, and compute loss from the predicted attention and the guide. Args: predicted_attention (Variable): shape(*, B, T_dec, T_enc), dtype float32, the alignment tensor, where B means batch size, T_dec means number of time steps of the decoder, T_enc means the number of time steps of the encoder, * means other possible dimensions. input_lengths (numpy.ndarray): shape(B,), dtype:int64, valid lengths (time steps) of encoder outputs. target_lengths (numpy.ndarray): shape(batch_size,), dtype:int64, valid lengths (time steps) of decoder outputs. Returns: loss (Variable): shape(1, ), dtype float32, attention loss. """ n_attention, batch_size, max_target_len, max_input_len = ( predicted_attention.shape) soft_mask = guided_attentions(input_lengths, target_lengths, max_target_len, self.guided_attention_sigma) soft_mask_ = dg.to_variable(soft_mask) loss = fluid.layers.reduce_mean(predicted_attention * soft_mask_) return loss def __call__(self, outputs, inputs): """Total loss Args: outpus is a tuple of (mel_hyp, lin_hyp, attn_hyp, done_hyp). mel_hyp (Variable): shape(B, T, C_mel), dtype float32, predicted mel spectrogram. lin_hyp (Variable): shape(B, T, C_lin), dtype float32, predicted linear spectrogram. done_hyp (Variable): shape(B, T), dtype float32, predicted done probability. attn_hyp (Variable): shape(N, B, T_dec, T_enc), dtype float32, predicted attention. inputs is a tuple of (mel_ref, lin_ref, done_ref, input_lengths, n_frames) mel_ref (Variable): shape(B, T, C_mel), dtype float32, ground truth mel spectrogram. lin_ref (Variable): shape(B, T, C_lin), dtype float32, ground truth linear spectrogram. done_ref (Variable): shape(B, T), dtype float32, ground truth done flag. input_lengths (Variable): shape(B, ), dtype: int, encoder valid lengths. n_frames (Variable): shape(B, ), dtype: int, decoder valid lengths. Returns: Dict(str, Variable): details of loss. """ total_loss = 0. mel_hyp, lin_hyp, attn_hyp, done_hyp = outputs mel_ref, lin_ref, done_ref, input_lengths, n_frames = inputs # n_frames # mel_lengths # decoder_lengths max_frames = lin_hyp.shape[1] max_mel_steps = max_frames // self.downsample_factor # max_decoder_steps = max_mel_steps // self.r # decoder_mask = F.sequence_mask(n_frames // self.downsample_factor // # self.r, # max_decoder_steps, # dtype="float32") mel_mask = F.sequence_mask( n_frames // self.downsample_factor, max_mel_steps, dtype="float32") lin_mask = F.sequence_mask(n_frames, max_frames, dtype="float32") lin_hyp = lin_hyp[:, :-self.time_shift, :] lin_ref = lin_ref[:, self.time_shift:, :] lin_mask = lin_mask[:, self.time_shift:] lin_l1_loss = self.l1_loss( lin_hyp, lin_ref, lin_mask, priority_bin=self.priority_bin) lin_bce_loss = self.binary_divergence(lin_hyp, lin_ref, lin_mask) lin_loss = self.binary_divergence_weight * lin_bce_loss \ + (1 - self.binary_divergence_weight) * lin_l1_loss total_loss += lin_loss mel_hyp = mel_hyp[:, :-self.time_shift, :] mel_ref = mel_ref[:, self.time_shift:, :] mel_mask = mel_mask[:, self.time_shift:] mel_l1_loss = self.l1_loss(mel_hyp, mel_ref, mel_mask) mel_bce_loss = self.binary_divergence(mel_hyp, mel_ref, mel_mask) # print("=====>", mel_l1_loss.numpy()[0], mel_bce_loss.numpy()[0]) mel_loss = self.binary_divergence_weight * mel_bce_loss \ + (1 - self.binary_divergence_weight) * mel_l1_loss total_loss += mel_loss attn_loss = self.attention_loss(attn_hyp, input_lengths.numpy(), n_frames.numpy() // (self.downsample_factor * self.r)) total_loss += attn_loss done_loss = self.done_loss(done_hyp, done_ref) total_loss += done_loss losses = { "loss": total_loss, "mel/mel_loss": mel_loss, "mel/l1_loss": mel_l1_loss, "mel/bce_loss": mel_bce_loss, "lin/lin_loss": lin_loss, "lin/l1_loss": lin_l1_loss, "lin/bce_loss": lin_bce_loss, "done": done_loss, "attn": attn_loss, } return losses
[ "paddle.fluid.layers.reduce_mean", "paddle.fluid.dygraph.to_variable", "paddle.fluid.layers.reduce_sum", "numpy.exp", "numpy.zeros", "numba.jit", "paddle.fluid.layers.abs", "paddle.fluid.layers.sequence_mask", "paddle.fluid.layers.elementwise_mul", "paddle.fluid.layers.log_loss", "paddle.fluid.l...
[((1228, 1246), 'numba.jit', 'jit', ([], {'nopython': '(True)'}), '(nopython=True)\n', (1231, 1246), False, 'from numba import jit\n'), ((1096, 1135), 'paddle.fluid.layers.elementwise_mul', 'F.elementwise_mul', (['inputs', 'mask'], {'axis': '(0)'}), '(inputs, mask, axis=0)\n', (1113, 1135), True, 'import paddle.fluid.layers as F\n'), ((1695, 1737), 'numpy.zeros', 'np.zeros', (['(max_N, max_T)'], {'dtype': 'np.float32'}), '((max_N, max_T), dtype=np.float32)\n', (1703, 1737), True, 'import numpy as np\n'), ((2598, 2661), 'numpy.zeros', 'np.zeros', (['(B, max_decoder_len, max_input_len)'], {'dtype': 'np.float32'}), '((B, max_decoder_len, max_input_len), dtype=np.float32)\n', (2606, 2661), True, 'import numpy as np\n'), ((1147, 1174), 'paddle.fluid.layers.reduce_sum', 'F.reduce_sum', (['masked_inputs'], {}), '(masked_inputs)\n', (1159, 1174), True, 'import paddle.fluid.layers as F\n'), ((4989, 5015), 'paddle.fluid.layers.abs', 'F.abs', (['(prediction - target)'], {}), '(prediction - target)\n', (4994, 5015), True, 'import paddle.fluid.layers as F\n'), ((6540, 6570), 'paddle.fluid.layers.reshape', 'F.reshape', (['prediction', '[-1, 1]'], {}), '(prediction, [-1, 1])\n', (6549, 6570), True, 'import paddle.fluid.layers as F\n'), ((6598, 6624), 'paddle.fluid.layers.reshape', 'F.reshape', (['target', '[-1, 1]'], {}), '(target, [-1, 1])\n', (6607, 6624), True, 'import paddle.fluid.layers as F\n'), ((6650, 6715), 'paddle.fluid.layers.log_loss', 'F.log_loss', (['flattened_prediction', 'flattened_target'], {'epsilon': '(1e-08)'}), '(flattened_prediction, flattened_target, epsilon=1e-08)\n', (6660, 6715), True, 'import paddle.fluid.layers as F\n'), ((6746, 6800), 'paddle.fluid.layers.reshape', 'fluid.layers.reshape', (['flattened_loss', 'prediction.shape'], {}), '(flattened_loss, prediction.shape)\n', (6766, 6800), False, 'from paddle import fluid\n'), ((7555, 7583), 'paddle.fluid.layers.reshape', 'F.reshape', (['done_hat', '[-1, 1]'], {}), '(done_hat, [-1, 1])\n', (7564, 7583), True, 'import paddle.fluid.layers as F\n'), ((7604, 7628), 'paddle.fluid.layers.reshape', 'F.reshape', (['done', '[-1, 1]'], {}), '(done, [-1, 1])\n', (7613, 7628), True, 'import paddle.fluid.layers as F\n'), ((7644, 7695), 'paddle.fluid.layers.log_loss', 'F.log_loss', (['flat_done_hat', 'flat_done'], {'epsilon': '(1e-08)'}), '(flat_done_hat, flat_done, epsilon=1e-08)\n', (7654, 7695), True, 'import paddle.fluid.layers as F\n'), ((7710, 7729), 'paddle.fluid.layers.reduce_mean', 'F.reduce_mean', (['loss'], {}), '(loss)\n', (7723, 7729), True, 'import paddle.fluid.layers as F\n'), ((8971, 8996), 'paddle.fluid.dygraph.to_variable', 'dg.to_variable', (['soft_mask'], {}), '(soft_mask)\n', (8985, 8996), True, 'import paddle.fluid.dygraph as dg\n'), ((9012, 9070), 'paddle.fluid.layers.reduce_mean', 'fluid.layers.reduce_mean', (['(predicted_attention * soft_mask_)'], {}), '(predicted_attention * soft_mask_)\n', (9036, 9070), False, 'from paddle import fluid\n'), ((10853, 10941), 'paddle.fluid.layers.sequence_mask', 'F.sequence_mask', (['(n_frames // self.downsample_factor)', 'max_mel_steps'], {'dtype': '"""float32"""'}), "(n_frames // self.downsample_factor, max_mel_steps, dtype=\n 'float32')\n", (10868, 10941), True, 'import paddle.fluid.layers as F\n'), ((10969, 11023), 'paddle.fluid.layers.sequence_mask', 'F.sequence_mask', (['n_frames', 'max_frames'], {'dtype': '"""float32"""'}), "(n_frames, max_frames, dtype='float32')\n", (10984, 11023), True, 'import paddle.fluid.layers as F\n'), ((1189, 1207), 'paddle.fluid.layers.reduce_sum', 'F.reduce_sum', (['mask'], {}), '(mask)\n', (1201, 1207), True, 'import paddle.fluid.layers as F\n'), ((5288, 5311), 'paddle.fluid.layers.reduce_mean', 'F.reduce_mean', (['abs_diff'], {}), '(abs_diff)\n', (5301, 5311), True, 'import paddle.fluid.layers as F\n'), ((7009, 7031), 'paddle.fluid.layers.reduce_mean', 'F.reduce_mean', (['bin_div'], {}), '(bin_div)\n', (7022, 7031), True, 'import paddle.fluid.layers as F\n'), ((1814, 1857), 'numpy.exp', 'np.exp', (['(-(n / N - t / T) ** 2 / (2 * g * g))'], {}), '(-(n / N - t / T) ** 2 / (2 * g * g))\n', (1820, 1857), True, 'import numpy as np\n'), ((5739, 5771), 'paddle.fluid.layers.reduce_mean', 'F.reduce_mean', (['priority_abs_diff'], {}), '(priority_abs_diff)\n', (5752, 5771), True, 'import paddle.fluid.layers as F\n'), ((5223, 5246), 'paddle.fluid.layers.reduce_mean', 'F.reduce_mean', (['abs_diff'], {}), '(abs_diff)\n', (5236, 5246), True, 'import paddle.fluid.layers as F\n'), ((6953, 6975), 'paddle.fluid.layers.reduce_mean', 'F.reduce_mean', (['bin_div'], {}), '(bin_div)\n', (6966, 6975), True, 'import paddle.fluid.layers as F\n'), ((5656, 5688), 'paddle.fluid.layers.reduce_mean', 'F.reduce_mean', (['priority_abs_diff'], {}), '(priority_abs_diff)\n', (5669, 5688), True, 'import paddle.fluid.layers as F\n')]
r"""@package motsfinder.metric.discrete.discretize Helpers to create discrete versions of (\eg analytical) metrics. These can be used to compare results obtained with analytically implemented metrics with those of the discrete metric classes. @b Examples ``` # Use a simple Brill-Lindquist metric as example here. m1 = 0.2; m2 = 0.8; d = 0.6 gBL = BrillLindquistMetric(d=d, m1=m1, m2=m2) # This creates the discrete version of it. res = 128; radius = 2 g = DiscretizedMetric( patch=DiscretizedMetric.construct_patch(res=res, radius=radius), metric=gBL, curv=gBL.get_curv(), # not needed here as gBL is time symmetric ) # To demonstrate that this metric can be used as usual, we find the four # MOTSs in it. h = InitHelper( metric=g, out_base="some/output/folder_res%s" % res, suffix="discrete", ) curves = h.find_four_MOTSs(m1=m1, m2=m2, d=d, plot=True) ``` """ import numpy as np from ...numutils import nan_mat, raise_all_warnings from .patch import GridPatch, DataPatch, BBox from .metric import DiscreteMetric from .tensors import DiscreteScalarField, DiscreteVectorField from .tensors import DiscreteSym2TensorField __all__ = [ "DiscretizedMetric", ] class _ScalarField(DiscreteScalarField): r"""Axisymmetric discrete scalar field created from a scalar function. This samples a (e.g. analytical) function on a grid to create a discrete version of it. The grid is defined by the configuration of the metric this field is associated with. """ def __init__(self, metric, func): r"""Create a scalar field from the given function. The `metric` defines the discretization (i.e. resolution, domain). """ super(_ScalarField, self).__init__(metric) self.__func = func def _load_data(self): f = self.__func patch = self.metric.patch mat = self.metric.empty_mat() with raise_all_warnings(): for (i, j, k), pt in patch.grid(full_output=True): mat[i, j, k] = _eval(f, pt, np.nan) return [DataPatch.from_patch(patch, mat, 'even')] class _VectorField(DiscreteVectorField): r"""Axisymmetric discrete vector field created from a vector-valued function. This samples a (e.g. analytical) function on a grid to create a discrete version of it. The grid is defined by the configuration of the metric this field is associated with. """ def __init__(self, metric, func): r"""Create a vector field from the given function. The `metric` defines the discretization (i.e. resolution, domain). `func` should be a callable returning three floats: the x-, y-, and z-components of the vector. """ super(_VectorField, self).__init__(metric) self.__func = func def _load_data(self): f = self.__func patch = self.metric.patch x_mat = self.metric.empty_mat() y_mat = self.metric.empty_mat() z_mat = self.metric.empty_mat() nan = nan_mat(3) with raise_all_warnings(): for (i, j, k), pt in patch.grid(full_output=True): x, y, z = _eval(f, pt, nan) x_mat[i, j, k] = x y_mat[i, j, k] = y z_mat[i, j, k] = z data = [(x_mat, 'odd'), (y_mat, 'odd'), (z_mat, 'even')] return [DataPatch.from_patch(patch, mat, sym) for mat, sym in data] class _Sym2TensorField(DiscreteSym2TensorField): r"""Axisymmetric discrete tensor field created from a matrix-valued function. This samples a (e.g. analytical) function on a grid to create a discrete version of it. The grid is defined by the configuration of the metric this field is associated with. """ def __init__(self, metric, func): r"""Create a tensor field from the given function. The `metric` defines the discretization (i.e. resolution, domain). `func` should be a callable returning a symmetric 3x3 matrix. """ super(_Sym2TensorField, self).__init__(metric) self.__func = func def _load_data(self): f = self.__func patch = self.metric.patch xx_mat = self.metric.empty_mat() xy_mat = self.metric.empty_mat() xz_mat = self.metric.empty_mat() yy_mat = self.metric.empty_mat() yz_mat = self.metric.empty_mat() zz_mat = self.metric.empty_mat() nan = nan_mat((3, 3)) with raise_all_warnings(): for (i, j, k), pt in patch.grid(full_output=True): mat = _eval(f, pt, nan) xx_mat[i, j, k] = mat[0, 0] xy_mat[i, j, k] = mat[0, 1] xz_mat[i, j, k] = mat[0, 2] yy_mat[i, j, k] = mat[1, 1] yz_mat[i, j, k] = mat[1, 2] zz_mat[i, j, k] = mat[2, 2] data = [ (xx_mat, 'even'), (xy_mat, 'even'), (xz_mat, 'odd'), (yy_mat, 'even'), (yz_mat, 'odd'), (zz_mat, 'even'), ] return [DataPatch.from_patch(patch, mat, sym) for mat, sym in data] class DiscretizedMetric(DiscreteMetric): r"""Full discrete slice geometry generated from (analytical) functions. This takes a 3-metric (..base._ThreeMetric) and optionally callables to evaluate e.g. the extrinsic curvature and builds matrices for all the different components. This allows discretization of analytical metrics to e.g. compare results at different discrete resolutions. """ def __init__(self, patch, metric, curv=None, lapse=None, shift=None, dtlapse=None, dtshift=None): r"""Construct a discrete metric on a given patch. @param patch Definition of the discretization (i.e. domain, resolution). Use the convenience class method construct_patch() to easily generate such a patch. @param metric 3-metric tensor field. Should be axisymmetric. A valid example is ..analytical.simple.BrillLindquistMetric. @param curv Callable returning symmetric 3x3 matrices representing the extrinsic curvature of the 3-slice embedded in spacetime. @param lapse,dtlapse Lapse function (scalar) and its time derivative (also scalar), respectively. Both are callables returning scalar values. @param shift,dtshift Shift vector field and its time derivative, respectively. Both callables should return 3 floats for the x-, y-, z-components of the field at the specified point. """ super(DiscretizedMetric, self).__init__() self._patch = patch self._metric = _Sym2TensorField(self, metric) self._curv = _Sym2TensorField(self, curv) if curv else None self._lapse = _ScalarField(self, lapse) if lapse else None self._shift = _VectorField(self, shift) if shift else None self._dtlapse = _ScalarField(self, dtlapse) if dtlapse else None self._dtshift = _VectorField(self, dtshift) if dtshift else None @classmethod def construct_patch(cls, res, radius, origin=(0., 0., 0.)): r"""Class method to create a patch definition from a given resolution. This creates a patch to be used for constructing a DiscretizedMetric. The domain is specified using an origin and "radius" (the domain is, of course, rectangular). @param res Resolution. There will be `res` grid points per unit per axis. @param radius Coordinate distance from `origin` to include in the domain. @param origin Origin of the patch around which the radius defines the full domain. Default is ``0,0,0``. """ origin = np.array(origin) # make a copy to not mutate input xmin = max(0., origin[0] - radius) xmax = origin[0] + radius origin[0] = xmin origin[2] -= radius deltas = 1./res * np.identity(3) box = BBox( lower=[0, 0, 0], upper=[int(round(res*(xmax-xmin)))+1, 1, int(round(2*res*radius))+1], ) return GridPatch(origin=origin, deltas=deltas, box=box) @property def patch(self): r"""Patch property used during construction.""" return self._patch def empty_mat(self): r"""Empty (zero) matrix of the correct shape for the full domain.""" return np.zeros(shape=self.patch.shape) def all_field_objects(self): return [ self._metric, self._curv, self._lapse, self._shift, self._dtlapse, self._dtshift, ] def _get_metric(self): return self._metric def get_curv(self): return self._curv def get_lapse(self): return self._lapse def get_dtlapse(self): return self._dtlapse def get_shift(self): return self._shift def get_dtshift(self): return self._dtshift def _eval(func, arg, default): r"""Evaluate a function, returning a given default in case of error. If evaluating the function succeeds, the produced value is returned. If, however, a `FloatingPointError` is raised, the given default value is returned instead. @param func Callable to evaluate. @param arg Argument to call `func` with. @param default Value to return in case of `FloatingPointError`. """ try: return func(arg) except (FloatingPointError, ZeroDivisionError): return default
[ "numpy.identity", "numpy.array", "numpy.zeros" ]
[((7924, 7940), 'numpy.array', 'np.array', (['origin'], {}), '(origin)\n', (7932, 7940), True, 'import numpy as np\n'), ((8626, 8658), 'numpy.zeros', 'np.zeros', ([], {'shape': 'self.patch.shape'}), '(shape=self.patch.shape)\n', (8634, 8658), True, 'import numpy as np\n'), ((8131, 8145), 'numpy.identity', 'np.identity', (['(3)'], {}), '(3)\n', (8142, 8145), True, 'import numpy as np\n')]
import numpy as np import matplotlib.pyplot as pp omega = .1 omega_n = 5 def analytic(t): return (np.cos(omega*t) - np.cos(omega_n*t) ) / (omega_n * omega_n - omega * omega) num=[1000,10000,100000] for type in ('euler', 'euler_symplectic', 'rk4'): for value in num: s = type + '.' + str(value) + '.out' t,x,xprime = np.loadtxt(s,unpack=True) labelstring = 'Nsteps = ' + str(value) #if type != 'euler_symplectic': # pp.plot(x,xprime,label=labelstring) if value == 10000: pp.plot(x,xprime,label=labelstring,lw=2.5) elif value == 1000: pp.plot(x,xprime,label=labelstring,lw=2) else: pp.plot(x,xprime,label=labelstring) # pp.plot(np.linspace(0.,100.,1000),analytic(np.linspace(0.,100.,1000)),label="True") pp.xlabel("position") pp.ylabel("velocity") pp.xlim(-.1,.1) pp.ylim(-.3,.3) if type == 'euler': pp.xlim(-1,1) pp.ylim(-1,1) pp.legend(loc='best') pp.title(type) s = 'pdf/' + type + '_phase_plot.pdf' pp.savefig(s) pp.close() #Now plot for a given nsteps all the types for value in num: for type in ('euler', 'euler_symplectic', 'rk4'): s = type + '.' + str(value) + '.out' t,x,xprime = np.loadtxt(s,unpack=True) if type == 'euler_symplectic': pp.plot(x,xprime,label=type,lw=4) else: pp.plot(x,xprime,label=type) #pp.plot(np.linspace(0.,100.,100000),analytic(np.linspace(0.,100.,100000)),label="True") pp.xlabel("position") pp.ylabel("velocity") pp.xlim(-.2,.15) pp.ylim(-1,1) pp.legend(loc='best') titlestring = 'Nsteps = ' + str(value) pp.title(titlestring) s = 'pdf/' + str(value) + '_phase_plot.pdf' pp.savefig(s) pp.close()
[ "matplotlib.pyplot.title", "matplotlib.pyplot.savefig", "matplotlib.pyplot.ylabel", "matplotlib.pyplot.xlabel", "matplotlib.pyplot.plot", "matplotlib.pyplot.close", "numpy.loadtxt", "numpy.cos", "matplotlib.pyplot.ylim", "matplotlib.pyplot.xlim", "matplotlib.pyplot.legend" ]
[((838, 859), 'matplotlib.pyplot.xlabel', 'pp.xlabel', (['"""position"""'], {}), "('position')\n", (847, 859), True, 'import matplotlib.pyplot as pp\n'), ((864, 885), 'matplotlib.pyplot.ylabel', 'pp.ylabel', (['"""velocity"""'], {}), "('velocity')\n", (873, 885), True, 'import matplotlib.pyplot as pp\n'), ((890, 908), 'matplotlib.pyplot.xlim', 'pp.xlim', (['(-0.1)', '(0.1)'], {}), '(-0.1, 0.1)\n', (897, 908), True, 'import matplotlib.pyplot as pp\n'), ((910, 928), 'matplotlib.pyplot.ylim', 'pp.ylim', (['(-0.3)', '(0.3)'], {}), '(-0.3, 0.3)\n', (917, 928), True, 'import matplotlib.pyplot as pp\n'), ((998, 1019), 'matplotlib.pyplot.legend', 'pp.legend', ([], {'loc': '"""best"""'}), "(loc='best')\n", (1007, 1019), True, 'import matplotlib.pyplot as pp\n'), ((1024, 1038), 'matplotlib.pyplot.title', 'pp.title', (['type'], {}), '(type)\n', (1032, 1038), True, 'import matplotlib.pyplot as pp\n'), ((1085, 1098), 'matplotlib.pyplot.savefig', 'pp.savefig', (['s'], {}), '(s)\n', (1095, 1098), True, 'import matplotlib.pyplot as pp\n'), ((1103, 1113), 'matplotlib.pyplot.close', 'pp.close', ([], {}), '()\n', (1111, 1113), True, 'import matplotlib.pyplot as pp\n'), ((1569, 1590), 'matplotlib.pyplot.xlabel', 'pp.xlabel', (['"""position"""'], {}), "('position')\n", (1578, 1590), True, 'import matplotlib.pyplot as pp\n'), ((1595, 1616), 'matplotlib.pyplot.ylabel', 'pp.ylabel', (['"""velocity"""'], {}), "('velocity')\n", (1604, 1616), True, 'import matplotlib.pyplot as pp\n'), ((1621, 1640), 'matplotlib.pyplot.xlim', 'pp.xlim', (['(-0.2)', '(0.15)'], {}), '(-0.2, 0.15)\n', (1628, 1640), True, 'import matplotlib.pyplot as pp\n'), ((1642, 1656), 'matplotlib.pyplot.ylim', 'pp.ylim', (['(-1)', '(1)'], {}), '(-1, 1)\n', (1649, 1656), True, 'import matplotlib.pyplot as pp\n'), ((1660, 1681), 'matplotlib.pyplot.legend', 'pp.legend', ([], {'loc': '"""best"""'}), "(loc='best')\n", (1669, 1681), True, 'import matplotlib.pyplot as pp\n'), ((1729, 1750), 'matplotlib.pyplot.title', 'pp.title', (['titlestring'], {}), '(titlestring)\n', (1737, 1750), True, 'import matplotlib.pyplot as pp\n'), ((1803, 1816), 'matplotlib.pyplot.savefig', 'pp.savefig', (['s'], {}), '(s)\n', (1813, 1816), True, 'import matplotlib.pyplot as pp\n'), ((1821, 1831), 'matplotlib.pyplot.close', 'pp.close', ([], {}), '()\n', (1829, 1831), True, 'import matplotlib.pyplot as pp\n'), ((343, 369), 'numpy.loadtxt', 'np.loadtxt', (['s'], {'unpack': '(True)'}), '(s, unpack=True)\n', (353, 369), True, 'import numpy as np\n'), ((958, 972), 'matplotlib.pyplot.xlim', 'pp.xlim', (['(-1)', '(1)'], {}), '(-1, 1)\n', (965, 972), True, 'import matplotlib.pyplot as pp\n'), ((980, 994), 'matplotlib.pyplot.ylim', 'pp.ylim', (['(-1)', '(1)'], {}), '(-1, 1)\n', (987, 994), True, 'import matplotlib.pyplot as pp\n'), ((1305, 1331), 'numpy.loadtxt', 'np.loadtxt', (['s'], {'unpack': '(True)'}), '(s, unpack=True)\n', (1315, 1331), True, 'import numpy as np\n'), ((103, 120), 'numpy.cos', 'np.cos', (['(omega * t)'], {}), '(omega * t)\n', (109, 120), True, 'import numpy as np\n'), ((121, 140), 'numpy.cos', 'np.cos', (['(omega_n * t)'], {}), '(omega_n * t)\n', (127, 140), True, 'import numpy as np\n'), ((544, 589), 'matplotlib.pyplot.plot', 'pp.plot', (['x', 'xprime'], {'label': 'labelstring', 'lw': '(2.5)'}), '(x, xprime, label=labelstring, lw=2.5)\n', (551, 589), True, 'import matplotlib.pyplot as pp\n'), ((1382, 1418), 'matplotlib.pyplot.plot', 'pp.plot', (['x', 'xprime'], {'label': 'type', 'lw': '(4)'}), '(x, xprime, label=type, lw=4)\n', (1389, 1418), True, 'import matplotlib.pyplot as pp\n'), ((1442, 1472), 'matplotlib.pyplot.plot', 'pp.plot', (['x', 'xprime'], {'label': 'type'}), '(x, xprime, label=type)\n', (1449, 1472), True, 'import matplotlib.pyplot as pp\n'), ((627, 670), 'matplotlib.pyplot.plot', 'pp.plot', (['x', 'xprime'], {'label': 'labelstring', 'lw': '(2)'}), '(x, xprime, label=labelstring, lw=2)\n', (634, 670), True, 'import matplotlib.pyplot as pp\n'), ((694, 731), 'matplotlib.pyplot.plot', 'pp.plot', (['x', 'xprime'], {'label': 'labelstring'}), '(x, xprime, label=labelstring)\n', (701, 731), True, 'import matplotlib.pyplot as pp\n')]
import numpy as np import properscoring as ps from scipy.stats import norm from sklearn.preprocessing import MinMaxScaler import matplotlib.pyplot as plt EPS = 1e-6 def rmse(predictions, targets): """ Root Mean Squared Error Args: predictions (np.ndarray): Point Predictions of the model targets (np.ndarray): Point Targets of the model Returns: float: RMSE """ return np.sqrt(((predictions - targets) ** 2).mean()) def mape(predictions, targets): """ Mean Absolute Percentage Error Args: predictions (np.ndarray): Predictions of the model targets (np.ndarray): Targets of the model Returns: float: MAPE """ return np.mean(np.abs((predictions - targets) / targets)) * 100 # target ground truth # mean - def crps(mean, std, targets): """ Quantile-based CRPS Args: mean (np.ndarray): Mean of the distribution (N) std (np.ndarray): Standard Deviation of the distribution (N) targets (np.ndarray): Targets of the model (N) Returns: float: CRPS """ return ps.crps_gaussian(targets, mean, std).mean() # -1 def crps_samples(samples, targets): """ Quantile-based CRPS Args: samples (np.ndarray): Samples of the distribution (N, samples) targets (np.ndarray): Targets of the model (N) Returns: float: CRPS """ return ps.crps_ensemble(targets, samples).mean() def log_score( mean, std, targets, window = 0.1 ): """ Log Score Args: mean (np.ndarray): Mean of the distribution (N) std (np.ndarray): Standard Deviation of the distribution (N) targets (np.ndarray): Targets of the model (N) Returns: float: Log Score """ # rescale, changed code scale = MinMaxScaler() targets = scale.fit_transform(targets) t1 = norm.cdf(targets - window / 2.0, mean, std) t2 = norm.cdf(targets + window / 2.0, mean, std) a = np.log(np.clip(t2 - t1, EPS, 1.0)).mean() return np.clip(a, -10, 10) # put in slack def interval_score( mean, std, targets, window = 1.0 ): """ Interval Score Args: mean (np.ndarray): Mean of the distribution (N) std (np.ndarray): Standard Deviation of the distribution (N) targets (np.ndarray): Targets of the model (N) Returns: float: Interval Score """ # rescale, changed code scale = MinMaxScaler() targets = scale.fit_transform(targets) rd_val = np.round(targets, decimals=1) low_val = np.clip(rd_val - window / 2, a_min=0.0, a_max=None) high_val = np.clip(rd_val + window / 2, a_min=None, a_max=13) t1 = norm.cdf(low_val, loc=mean, scale=std) t2 = norm.cdf(high_val, loc=mean, scale=std) return np.log(np.clip(t2 - t1, a_min=EPS, a_max=1.0)).mean() def conf_interval( mean, var, conf ): """ Confintance Interval for given confidence level Args: mean (np.ndarray): Mean of the distribution (N) var (np.ndarray): Variance of the distribution (N) conf (float): Confidence level Returns: tuple: (low, high) interval """ out_prob = 1.0 - conf high = norm.ppf(1.0 - (out_prob / 2), loc=mean, scale=var**0.5) low = norm.ppf((1.0 - conf) / 2, loc=mean, scale=var**0.5) return low, high def pres_recall( mean, var, target, conf ): """ Fraction of GT points within the confidence interval Args: mean (np.ndarray): Mean of the distribution (N) var (np.ndarray): Variance of the distribution (N) target (np.ndarray): Target of the model (N) conf (float): Confidence level Returns: np.ndarray: Fraction of GT points within the confidence interval """ low, high = conf_interval(mean, var, conf) truth = ((target > low) & (target < high)).astype("float32") return truth.mean(-1) # Plot def get_pr(pred, var, target, color="blue", label="FluFNP"): """ Plot confidence and return Confidence score and AUC Args: pred (np.ndarray): Predictions of the model (N) var (np.ndarray): Variance of the distribution (N) target (np.ndarray): Target of the model (N) color (str): Color of the line label (str): Label of the model Returns: tuple: (Confidence score, AUC, fraction values) """ x = np.arange(0.05, 1.0, 0.01).reshape((95, 1)) y = np.array([pres_recall(pred, var, target, c) for c in x]) # plt.plot(list(x) + [1.0], list(y) + [1.0], label=label, color=color) conf_score = np.abs(y - x).sum() * 0.01 auc = y.sum() * 0.01 return auc, conf_score, list(y) + [1.0]
[ "numpy.clip", "numpy.abs", "numpy.arange", "scipy.stats.norm.ppf", "properscoring.crps_ensemble", "properscoring.crps_gaussian", "scipy.stats.norm.cdf", "sklearn.preprocessing.MinMaxScaler", "numpy.round" ]
[((1820, 1834), 'sklearn.preprocessing.MinMaxScaler', 'MinMaxScaler', ([], {}), '()\n', (1832, 1834), False, 'from sklearn.preprocessing import MinMaxScaler\n'), ((1893, 1936), 'scipy.stats.norm.cdf', 'norm.cdf', (['(targets - window / 2.0)', 'mean', 'std'], {}), '(targets - window / 2.0, mean, std)\n', (1901, 1936), False, 'from scipy.stats import norm\n'), ((1946, 1989), 'scipy.stats.norm.cdf', 'norm.cdf', (['(targets + window / 2.0)', 'mean', 'std'], {}), '(targets + window / 2.0, mean, std)\n', (1954, 1989), False, 'from scipy.stats import norm\n'), ((2052, 2071), 'numpy.clip', 'np.clip', (['a', '(-10)', '(10)'], {}), '(a, -10, 10)\n', (2059, 2071), True, 'import numpy as np\n'), ((2457, 2471), 'sklearn.preprocessing.MinMaxScaler', 'MinMaxScaler', ([], {}), '()\n', (2469, 2471), False, 'from sklearn.preprocessing import MinMaxScaler\n'), ((2530, 2559), 'numpy.round', 'np.round', (['targets'], {'decimals': '(1)'}), '(targets, decimals=1)\n', (2538, 2559), True, 'import numpy as np\n'), ((2574, 2625), 'numpy.clip', 'np.clip', (['(rd_val - window / 2)'], {'a_min': '(0.0)', 'a_max': 'None'}), '(rd_val - window / 2, a_min=0.0, a_max=None)\n', (2581, 2625), True, 'import numpy as np\n'), ((2641, 2691), 'numpy.clip', 'np.clip', (['(rd_val + window / 2)'], {'a_min': 'None', 'a_max': '(13)'}), '(rd_val + window / 2, a_min=None, a_max=13)\n', (2648, 2691), True, 'import numpy as np\n'), ((2701, 2739), 'scipy.stats.norm.cdf', 'norm.cdf', (['low_val'], {'loc': 'mean', 'scale': 'std'}), '(low_val, loc=mean, scale=std)\n', (2709, 2739), False, 'from scipy.stats import norm\n'), ((2749, 2788), 'scipy.stats.norm.cdf', 'norm.cdf', (['high_val'], {'loc': 'mean', 'scale': 'std'}), '(high_val, loc=mean, scale=std)\n', (2757, 2788), False, 'from scipy.stats import norm\n'), ((3215, 3271), 'scipy.stats.norm.ppf', 'norm.ppf', (['(1.0 - out_prob / 2)'], {'loc': 'mean', 'scale': '(var ** 0.5)'}), '(1.0 - out_prob / 2, loc=mean, scale=var ** 0.5)\n', (3223, 3271), False, 'from scipy.stats import norm\n'), ((3282, 3336), 'scipy.stats.norm.ppf', 'norm.ppf', (['((1.0 - conf) / 2)'], {'loc': 'mean', 'scale': '(var ** 0.5)'}), '((1.0 - conf) / 2, loc=mean, scale=var ** 0.5)\n', (3290, 3336), False, 'from scipy.stats import norm\n'), ((729, 770), 'numpy.abs', 'np.abs', (['((predictions - targets) / targets)'], {}), '((predictions - targets) / targets)\n', (735, 770), True, 'import numpy as np\n'), ((1116, 1152), 'properscoring.crps_gaussian', 'ps.crps_gaussian', (['targets', 'mean', 'std'], {}), '(targets, mean, std)\n', (1132, 1152), True, 'import properscoring as ps\n'), ((1422, 1456), 'properscoring.crps_ensemble', 'ps.crps_ensemble', (['targets', 'samples'], {}), '(targets, samples)\n', (1438, 1456), True, 'import properscoring as ps\n'), ((4399, 4425), 'numpy.arange', 'np.arange', (['(0.05)', '(1.0)', '(0.01)'], {}), '(0.05, 1.0, 0.01)\n', (4408, 4425), True, 'import numpy as np\n'), ((2006, 2032), 'numpy.clip', 'np.clip', (['(t2 - t1)', 'EPS', '(1.0)'], {}), '(t2 - t1, EPS, 1.0)\n', (2013, 2032), True, 'import numpy as np\n'), ((2807, 2845), 'numpy.clip', 'np.clip', (['(t2 - t1)'], {'a_min': 'EPS', 'a_max': '(1.0)'}), '(t2 - t1, a_min=EPS, a_max=1.0)\n', (2814, 2845), True, 'import numpy as np\n'), ((4600, 4613), 'numpy.abs', 'np.abs', (['(y - x)'], {}), '(y - x)\n', (4606, 4613), True, 'import numpy as np\n')]
""" Train and test the key identification CNN """ import database import generator import eda from glob import glob import os import numpy as np import keras from keras.models import Model from keras.layers import Input, Conv2D, MaxPooling2D, add from keras.layers import Dropout, Flatten, Dense from keras.callbacks import ModelCheckpoint from keras.optimizers import Adam from sklearn.metrics import f1_score, accuracy_score, recall_score, precision_score def build_model(): """Builds the key identification CNN model""" # o = 0.0388 inputs = Input(shape=(15,155,1)) nb_filters = 32 # 15 x 155 x 1 x = Conv2D(filters=nb_filters, kernel_size=(3,3), padding='same', activation='relu')(inputs) x = Conv2D(filters=2*nb_filters, kernel_size=(3,3), padding='same', activation='relu')(x) x = Dropout(0.2)(x) pool = MaxPooling2D(pool_size=(1,2), padding='same')(x) # 15 x 78 x 32 x = Conv2D(filters=nb_filters, kernel_size=(3,3), padding='same', activation='relu')(pool) x = Conv2D(filters=2*nb_filters, kernel_size=(3,3), padding='same', activation='relu')(x) x = Dropout(0.2)(x) x = add([pool,x]) pool = MaxPooling2D(pool_size=(1,2), padding='same')(x) # 15 x 39 x 32 x = Conv2D(filters=nb_filters, kernel_size=(3,3), padding='same', activation='relu')(pool) x = Conv2D(filters=2*nb_filters, kernel_size=(3,3), padding='same', activation='relu')(x) x = Dropout(0.2)(x) x = add([pool,x]) pool = MaxPooling2D(pool_size=(1,2), padding='same')(x) # 15 x 20 x 32 x = Conv2D(filters=nb_filters, kernel_size=(3,3), padding='same', activation='relu')(pool) x = Conv2D(filters=2*nb_filters, kernel_size=(3,3), padding='same', activation='relu')(x) x = Dropout(0.2)(x) x = add([pool,x]) pool = MaxPooling2D(pool_size=(2,2), padding='same')(x) # 8 x 10 x 32 x = Conv2D(filters=nb_filters, kernel_size=(3,3), padding='same', activation='relu')(pool) x = Conv2D(filters=2*nb_filters, kernel_size=(3,3), padding='same', activation='relu')(x) x = Dropout(0.2)(x) x = add([pool,x]) #pool = MaxPooling2D(pool_size=(2,2), padding='same')(x) pool = x # 8 x 10 x 64 x = Conv2D(filters=2*nb_filters, kernel_size=(3,3), padding='same', activation='relu')(pool) x = add([pool,x]) x = Conv2D(filters=4*nb_filters, kernel_size=(3,3), padding='same', activation='relu')(x) x = Dropout(0.2)(x) pool = MaxPooling2D(pool_size=(2,2), padding='same')(x) # 4 x 5 x 128 x = Conv2D(filters=4*nb_filters, kernel_size=(2,2), padding='same', activation='relu')(pool) x = add([pool,x]) x = Conv2D(filters=8*nb_filters, kernel_size=(2,2), padding='same', activation='relu')(x) x = Dropout(0.2)(x) pool = MaxPooling2D(pool_size=(2,2), padding='same')(x) # 2 x 3 x 256 x = Dropout(0.2)(pool) x = Flatten()(x) ## x = Dense(512, activation='relu')(x) ## x = Dropout(0.2)(x) ## x = Dense(256, activation='relu')(x) ## x = Dropout(0.2)(x) predictions = Dense(88, activation='sigmoid')(x) model = Model(inputs=inputs, outputs=predictions) model.compile(optimizer=Adam(), loss='binary_crossentropy', metrics=['accuracy']) return model def train_model(model, db, output=None, epochs=300): """ Train a key identification model on a database :param model: model to train :param db: database to train on :param output: output filename of the best trained model :param epochs: number of iterations :return: train history """ if output==None: dbname = db.name() output = dbname[:dbname.rindex('.')] + '_note.hdf5' #model.summary() checkpointer = ModelCheckpoint(filepath=output, verbose=1, save_best_only=True) train_group = ['train','cqt'] xgroup = db.get_subgroup(['train','cqt']) ygroup = db.get_subgroup(['train','note_labels']) step = min(500, db.get_total_points(train_group)) frac_val = 0.2 frac = 1.-frac_val nb_steps, tmp = generator.get_nb_steps(xgroup, step, frac, dict_arrays=True) nb_steps_val, tmp = generator.get_nb_steps(xgroup, step, frac_val, shift='end', dict_arrays=True) print('step: ',step) print('nb_steps: ',nb_steps, nb_steps_val) hist = model.fit_generator(generator.generator( (xgroup, ygroup), nb=step, frac=frac, dict_arrays=True), steps_per_epoch= 100,#max(4,nb_steps), max_queue_size=1, validation_data= generator.generator( (xgroup, ygroup), nb=step, frac=frac_val, shift='end', dict_arrays=True), validation_steps= nb_steps_val, epochs=epochs, callbacks=[checkpointer], verbose=2 ) return hist.history def compute_scores(model, set_name, subset, log, categories=False): """ Do predictions on a dataset and compute scores :param model: model to test :param set_name: string in the form of set_1 (must be in data folder) :param subset: string in the form of 'test', 'train' (must be in the set_name folder) :param log: opened log file to write results into :param categories: set to True to compute scores per category :return: None """ filenames = glob('data/{}/{}/*.mid'.format(set_name, subset)) basenames = [] for f in filenames: basenames.append(os.path.splitext(os.path.basename(f))[0]) db = database.DatabaseReader('data_{}.hdf5'.format(set_name)) print("compute scores for : {} / {}".format(set_name, subset)) dgroup = db.get_subgroup([subset, 'cqt']) lgroup = db.get_subgroup([subset, 'note_labels']) res = {} true_positive = np.zeros(88) true_negative = np.zeros(88) false_positive = np.zeros(88) false_negative = np.zeros(88) nb_class = np.zeros(88) nb_total = 0 tp_total = 0 tn_total = 0 fp_total = 0 fn_total = 0 for name in dgroup: test_preds = 1. * (model.predict(dgroup[name]) >= 0.5) y_test = lgroup[name] x = test_preds y = y_test[:] z = y + 2*x # foreach note (88 arrays): true_positive += np.sum(z==3, axis=0) true_negative += np.sum(z==0, axis=0) false_positive += np.sum(z==2, axis=0) false_negative += np.sum(z==1, axis=0) nb_class += np.sum(y, axis=0) nb_total += x.shape[0] tp_total += np.sum(true_positive) tn_total += np.sum(true_negative) fp_total += np.sum(false_positive) fn_total += np.sum(false_negative) score = (accuracy_score(y_test, test_preds), precision_score(y_test, test_preds, average='weighted'), recall_score(y_test, test_preds, average='weighted'), f1_score(y_test, test_preds, average='weighted')) for f, b in zip(filenames, basenames): if b in name: break stats = eda.analyze(f) stats_simp = [] for s in stats: stats_simp.append(s[0][s[1].argmax()]) res[name] = (score, stats_simp) print(f'--- {name} ---') print("accuracy {:.4f}".format(score[0])) print("precision {:.4f}".format(score[1])) print("recall {:.4f}".format(score[2])) print("F1-score {:.4f}".format(score[3])) log.write(name) for r in score: log.write('\t{:.4f}'.format(r)) for s in stats_simp: log.write('\t{}'.format(s)) log.write('\n') log.flush() log.write(f'TOTAL_{set_name}_{subset}') # score per class accu = (true_positive+true_negative)/nb_total prec = np.divide(true_positive, (true_positive+false_positive)) reca = np.divide(true_positive, (true_positive+false_negative)) f1 = np.divide(2*prec*reca, (prec+reca)) # macro score (unused) macro_f1 = np.nanmean(f1) # micro score (unused) micro_prec = tp_total/(tp_total+fp_total) micro_reca = tp_total/(tp_total+fn_total) micro_f1 = 2*micro_prec*micro_reca/(micro_reca+micro_prec) # weighted score w = nb_class/np.sum(nb_class) w_accu = np.nansum(accu*w) w_prec = np.nansum(prec*w) w_reca = np.nansum(reca*w) w_f1 = np.nansum(f1*w) for r in (w_accu, w_prec, w_reca, w_f1): log.write('\t{:.4f}'.format(r)) for s in stats_simp: log.write('\t-') log.write('\n') # category score if categories: for i in range(len(accu)): log.write(f'carac_note\t{i}\t{accu[i]}\t{prec[i]}\t{reca[i]}\t{f1[i]}\n') log.flush() def train_sets(sets, epochs): """ Train several models on datasets (each model is trained on one dataset) :param sets: list of datasets :param epochs: number of iterations :return: None """ for s in sets: print(f'-------- Training on {s} --------') db = database.DatabaseReader(f'data_{s}.hdf5') model = build_model() #model.summary() hist = train_model(model, db, output=f'best_note_{s}.hdf5', epochs=epochs) with open(f'best_note_{s}_training.log','w') as log: log.write('epoch\t') for k in hist: log.write(k+'\t') nb = len(hist[k]) log.write('\n') for i in range(nb): log.write(f'{i+1}\t') for k in hist: log.write(f'{hist[k][i]}\t') log.write('\n') def test_sets(sets, doTrain=False, categories=False): """ Test models on datasets. Each model is tested on all datasets. :param sets: list of datasets :param doTrain: set to True to also test on training set :param categories: set to True to also compute scores per category :return: None (writes log files) """ for s in sets: with open(f'best_note_{s}_predictions.log', "w") as log: print(f'-------- Predictions for model trained on {s} --------') log.write('song\tacc\tprec\trecall\tF1\tstroke\tnote\tvolume\tnb\ttempo\n') model = keras.models.load_model(f'best_note_{s}.hdf5') for s2 in sets: print(f' -> computing predictions on {s2}') if doTrain: compute_scores(model, s2, 'train', log, categories) compute_scores(model, s2, 'test', log, categories) if __name__ == '__main__': sets = ['set_1', 'set_2', 'set_3', 'set_4' ] # train_sets(sets, epochs=50) test_sets(sets, doTrain=True, categories=False) # test_sets(sets, doTrain=False, categories=True)
[ "keras.layers.Conv2D", "eda.analyze", "sklearn.metrics.precision_score", "sklearn.metrics.recall_score", "numpy.nanmean", "keras.layers.Dense", "numpy.divide", "generator.generator", "database.DatabaseReader", "keras.models.Model", "generator.get_nb_steps", "keras.optimizers.Adam", "keras.la...
[((563, 588), 'keras.layers.Input', 'Input', ([], {'shape': '(15, 155, 1)'}), '(shape=(15, 155, 1))\n', (568, 588), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1146, 1160), 'keras.layers.add', 'add', (['[pool, x]'], {}), '([pool, x])\n', (1149, 1160), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1462, 1476), 'keras.layers.add', 'add', (['[pool, x]'], {}), '([pool, x])\n', (1465, 1476), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1778, 1792), 'keras.layers.add', 'add', (['[pool, x]'], {}), '([pool, x])\n', (1781, 1792), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((2093, 2107), 'keras.layers.add', 'add', (['[pool, x]'], {}), '([pool, x])\n', (2096, 2107), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((2306, 2320), 'keras.layers.add', 'add', (['[pool, x]'], {}), '([pool, x])\n', (2309, 2320), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((2623, 2637), 'keras.layers.add', 'add', (['[pool, x]'], {}), '([pool, x])\n', (2626, 2637), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((3095, 3136), 'keras.models.Model', 'Model', ([], {'inputs': 'inputs', 'outputs': 'predictions'}), '(inputs=inputs, outputs=predictions)\n', (3100, 3136), False, 'from keras.models import Model\n'), ((3759, 3823), 'keras.callbacks.ModelCheckpoint', 'ModelCheckpoint', ([], {'filepath': 'output', 'verbose': '(1)', 'save_best_only': '(True)'}), '(filepath=output, verbose=1, save_best_only=True)\n', (3774, 3823), False, 'from keras.callbacks import ModelCheckpoint\n'), ((4118, 4178), 'generator.get_nb_steps', 'generator.get_nb_steps', (['xgroup', 'step', 'frac'], {'dict_arrays': '(True)'}), '(xgroup, step, frac, dict_arrays=True)\n', (4140, 4178), False, 'import generator\n'), ((4203, 4280), 'generator.get_nb_steps', 'generator.get_nb_steps', (['xgroup', 'step', 'frac_val'], {'shift': '"""end"""', 'dict_arrays': '(True)'}), "(xgroup, step, frac_val, shift='end', dict_arrays=True)\n", (4225, 4280), False, 'import generator\n'), ((6254, 6266), 'numpy.zeros', 'np.zeros', (['(88)'], {}), '(88)\n', (6262, 6266), True, 'import numpy as np\n'), ((6287, 6299), 'numpy.zeros', 'np.zeros', (['(88)'], {}), '(88)\n', (6295, 6299), True, 'import numpy as np\n'), ((6321, 6333), 'numpy.zeros', 'np.zeros', (['(88)'], {}), '(88)\n', (6329, 6333), True, 'import numpy as np\n'), ((6355, 6367), 'numpy.zeros', 'np.zeros', (['(88)'], {}), '(88)\n', (6363, 6367), True, 'import numpy as np\n'), ((6383, 6395), 'numpy.zeros', 'np.zeros', (['(88)'], {}), '(88)\n', (6391, 6395), True, 'import numpy as np\n'), ((8263, 8319), 'numpy.divide', 'np.divide', (['true_positive', '(true_positive + false_positive)'], {}), '(true_positive, true_positive + false_positive)\n', (8272, 8319), True, 'import numpy as np\n'), ((8331, 8387), 'numpy.divide', 'np.divide', (['true_positive', '(true_positive + false_negative)'], {}), '(true_positive, true_positive + false_negative)\n', (8340, 8387), True, 'import numpy as np\n'), ((8397, 8436), 'numpy.divide', 'np.divide', (['(2 * prec * reca)', '(prec + reca)'], {}), '(2 * prec * reca, prec + reca)\n', (8406, 8436), True, 'import numpy as np\n'), ((8476, 8490), 'numpy.nanmean', 'np.nanmean', (['f1'], {}), '(f1)\n', (8486, 8490), True, 'import numpy as np\n'), ((8743, 8762), 'numpy.nansum', 'np.nansum', (['(accu * w)'], {}), '(accu * w)\n', (8752, 8762), True, 'import numpy as np\n'), ((8774, 8793), 'numpy.nansum', 'np.nansum', (['(prec * w)'], {}), '(prec * w)\n', (8783, 8793), True, 'import numpy as np\n'), ((8805, 8824), 'numpy.nansum', 'np.nansum', (['(reca * w)'], {}), '(reca * w)\n', (8814, 8824), True, 'import numpy as np\n'), ((8834, 8851), 'numpy.nansum', 'np.nansum', (['(f1 * w)'], {}), '(f1 * w)\n', (8843, 8851), True, 'import numpy as np\n'), ((637, 723), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': 'nb_filters', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=nb_filters, kernel_size=(3, 3), padding='same', activation=\n 'relu')\n", (643, 723), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((734, 823), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': '(2 * nb_filters)', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=2 * nb_filters, kernel_size=(3, 3), padding='same',\n activation='relu')\n", (740, 823), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((828, 840), 'keras.layers.Dropout', 'Dropout', (['(0.2)'], {}), '(0.2)\n', (835, 840), False, 'from keras.layers import Dropout, Flatten, Dense\n'), ((855, 901), 'keras.layers.MaxPooling2D', 'MaxPooling2D', ([], {'pool_size': '(1, 2)', 'padding': '"""same"""'}), "(pool_size=(1, 2), padding='same')\n", (867, 901), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((933, 1019), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': 'nb_filters', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=nb_filters, kernel_size=(3, 3), padding='same', activation=\n 'relu')\n", (939, 1019), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1028, 1117), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': '(2 * nb_filters)', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=2 * nb_filters, kernel_size=(3, 3), padding='same',\n activation='relu')\n", (1034, 1117), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1122, 1134), 'keras.layers.Dropout', 'Dropout', (['(0.2)'], {}), '(0.2)\n', (1129, 1134), False, 'from keras.layers import Dropout, Flatten, Dense\n'), ((1171, 1217), 'keras.layers.MaxPooling2D', 'MaxPooling2D', ([], {'pool_size': '(1, 2)', 'padding': '"""same"""'}), "(pool_size=(1, 2), padding='same')\n", (1183, 1217), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1249, 1335), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': 'nb_filters', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=nb_filters, kernel_size=(3, 3), padding='same', activation=\n 'relu')\n", (1255, 1335), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1344, 1433), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': '(2 * nb_filters)', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=2 * nb_filters, kernel_size=(3, 3), padding='same',\n activation='relu')\n", (1350, 1433), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1438, 1450), 'keras.layers.Dropout', 'Dropout', (['(0.2)'], {}), '(0.2)\n', (1445, 1450), False, 'from keras.layers import Dropout, Flatten, Dense\n'), ((1487, 1533), 'keras.layers.MaxPooling2D', 'MaxPooling2D', ([], {'pool_size': '(1, 2)', 'padding': '"""same"""'}), "(pool_size=(1, 2), padding='same')\n", (1499, 1533), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1565, 1651), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': 'nb_filters', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=nb_filters, kernel_size=(3, 3), padding='same', activation=\n 'relu')\n", (1571, 1651), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1660, 1749), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': '(2 * nb_filters)', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=2 * nb_filters, kernel_size=(3, 3), padding='same',\n activation='relu')\n", (1666, 1749), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1754, 1766), 'keras.layers.Dropout', 'Dropout', (['(0.2)'], {}), '(0.2)\n', (1761, 1766), False, 'from keras.layers import Dropout, Flatten, Dense\n'), ((1803, 1849), 'keras.layers.MaxPooling2D', 'MaxPooling2D', ([], {'pool_size': '(2, 2)', 'padding': '"""same"""'}), "(pool_size=(2, 2), padding='same')\n", (1815, 1849), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1880, 1966), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': 'nb_filters', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=nb_filters, kernel_size=(3, 3), padding='same', activation=\n 'relu')\n", (1886, 1966), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((1975, 2064), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': '(2 * nb_filters)', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=2 * nb_filters, kernel_size=(3, 3), padding='same',\n activation='relu')\n", (1981, 2064), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((2069, 2081), 'keras.layers.Dropout', 'Dropout', (['(0.2)'], {}), '(0.2)\n', (2076, 2081), False, 'from keras.layers import Dropout, Flatten, Dense\n'), ((2209, 2298), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': '(2 * nb_filters)', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=2 * nb_filters, kernel_size=(3, 3), padding='same',\n activation='relu')\n", (2215, 2298), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((2328, 2417), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': '(4 * nb_filters)', 'kernel_size': '(3, 3)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=4 * nb_filters, kernel_size=(3, 3), padding='same',\n activation='relu')\n", (2334, 2417), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((2422, 2434), 'keras.layers.Dropout', 'Dropout', (['(0.2)'], {}), '(0.2)\n', (2429, 2434), False, 'from keras.layers import Dropout, Flatten, Dense\n'), ((2449, 2495), 'keras.layers.MaxPooling2D', 'MaxPooling2D', ([], {'pool_size': '(2, 2)', 'padding': '"""same"""'}), "(pool_size=(2, 2), padding='same')\n", (2461, 2495), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((2526, 2615), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': '(4 * nb_filters)', 'kernel_size': '(2, 2)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=4 * nb_filters, kernel_size=(2, 2), padding='same',\n activation='relu')\n", (2532, 2615), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((2645, 2734), 'keras.layers.Conv2D', 'Conv2D', ([], {'filters': '(8 * nb_filters)', 'kernel_size': '(2, 2)', 'padding': '"""same"""', 'activation': '"""relu"""'}), "(filters=8 * nb_filters, kernel_size=(2, 2), padding='same',\n activation='relu')\n", (2651, 2734), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((2739, 2751), 'keras.layers.Dropout', 'Dropout', (['(0.2)'], {}), '(0.2)\n', (2746, 2751), False, 'from keras.layers import Dropout, Flatten, Dense\n'), ((2766, 2812), 'keras.layers.MaxPooling2D', 'MaxPooling2D', ([], {'pool_size': '(2, 2)', 'padding': '"""same"""'}), "(pool_size=(2, 2), padding='same')\n", (2778, 2812), False, 'from keras.layers import Input, Conv2D, MaxPooling2D, add\n'), ((2851, 2863), 'keras.layers.Dropout', 'Dropout', (['(0.2)'], {}), '(0.2)\n', (2858, 2863), False, 'from keras.layers import Dropout, Flatten, Dense\n'), ((2878, 2887), 'keras.layers.Flatten', 'Flatten', ([], {}), '()\n', (2885, 2887), False, 'from keras.layers import Dropout, Flatten, Dense\n'), ((3047, 3078), 'keras.layers.Dense', 'Dense', (['(88)'], {'activation': '"""sigmoid"""'}), "(88, activation='sigmoid')\n", (3052, 3078), False, 'from keras.layers import Dropout, Flatten, Dense\n'), ((4390, 4465), 'generator.generator', 'generator.generator', (['(xgroup, ygroup)'], {'nb': 'step', 'frac': 'frac', 'dict_arrays': '(True)'}), '((xgroup, ygroup), nb=step, frac=frac, dict_arrays=True)\n', (4409, 4465), False, 'import generator\n'), ((6737, 6759), 'numpy.sum', 'np.sum', (['(z == 3)'], {'axis': '(0)'}), '(z == 3, axis=0)\n', (6743, 6759), True, 'import numpy as np\n'), ((6783, 6805), 'numpy.sum', 'np.sum', (['(z == 0)'], {'axis': '(0)'}), '(z == 0, axis=0)\n', (6789, 6805), True, 'import numpy as np\n'), ((6830, 6852), 'numpy.sum', 'np.sum', (['(z == 2)'], {'axis': '(0)'}), '(z == 2, axis=0)\n', (6836, 6852), True, 'import numpy as np\n'), ((6877, 6899), 'numpy.sum', 'np.sum', (['(z == 1)'], {'axis': '(0)'}), '(z == 1, axis=0)\n', (6883, 6899), True, 'import numpy as np\n'), ((6918, 6935), 'numpy.sum', 'np.sum', (['y'], {'axis': '(0)'}), '(y, axis=0)\n', (6924, 6935), True, 'import numpy as np\n'), ((6997, 7018), 'numpy.sum', 'np.sum', (['true_positive'], {}), '(true_positive)\n', (7003, 7018), True, 'import numpy as np\n'), ((7039, 7060), 'numpy.sum', 'np.sum', (['true_negative'], {}), '(true_negative)\n', (7045, 7060), True, 'import numpy as np\n'), ((7081, 7103), 'numpy.sum', 'np.sum', (['false_positive'], {}), '(false_positive)\n', (7087, 7103), True, 'import numpy as np\n'), ((7124, 7146), 'numpy.sum', 'np.sum', (['false_negative'], {}), '(false_negative)\n', (7130, 7146), True, 'import numpy as np\n'), ((7526, 7540), 'eda.analyze', 'eda.analyze', (['f'], {}), '(f)\n', (7537, 7540), False, 'import eda\n'), ((8713, 8729), 'numpy.sum', 'np.sum', (['nb_class'], {}), '(nb_class)\n', (8719, 8729), True, 'import numpy as np\n'), ((9488, 9529), 'database.DatabaseReader', 'database.DatabaseReader', (['f"""data_{s}.hdf5"""'], {}), "(f'data_{s}.hdf5')\n", (9511, 9529), False, 'import database\n'), ((3166, 3172), 'keras.optimizers.Adam', 'Adam', ([], {}), '()\n', (3170, 3172), False, 'from keras.optimizers import Adam\n'), ((4735, 4831), 'generator.generator', 'generator.generator', (['(xgroup, ygroup)'], {'nb': 'step', 'frac': 'frac_val', 'shift': '"""end"""', 'dict_arrays': '(True)'}), "((xgroup, ygroup), nb=step, frac=frac_val, shift='end',\n dict_arrays=True)\n", (4754, 4831), False, 'import generator\n'), ((7165, 7199), 'sklearn.metrics.accuracy_score', 'accuracy_score', (['y_test', 'test_preds'], {}), '(y_test, test_preds)\n', (7179, 7199), False, 'from sklearn.metrics import f1_score, accuracy_score, recall_score, precision_score\n'), ((7218, 7273), 'sklearn.metrics.precision_score', 'precision_score', (['y_test', 'test_preds'], {'average': '"""weighted"""'}), "(y_test, test_preds, average='weighted')\n", (7233, 7273), False, 'from sklearn.metrics import f1_score, accuracy_score, recall_score, precision_score\n'), ((7292, 7344), 'sklearn.metrics.recall_score', 'recall_score', (['y_test', 'test_preds'], {'average': '"""weighted"""'}), "(y_test, test_preds, average='weighted')\n", (7304, 7344), False, 'from sklearn.metrics import f1_score, accuracy_score, recall_score, precision_score\n'), ((7363, 7411), 'sklearn.metrics.f1_score', 'f1_score', (['y_test', 'test_preds'], {'average': '"""weighted"""'}), "(y_test, test_preds, average='weighted')\n", (7371, 7411), False, 'from sklearn.metrics import f1_score, accuracy_score, recall_score, precision_score\n'), ((10693, 10739), 'keras.models.load_model', 'keras.models.load_model', (['f"""best_note_{s}.hdf5"""'], {}), "(f'best_note_{s}.hdf5')\n", (10716, 10739), False, 'import keras\n'), ((5959, 5978), 'os.path.basename', 'os.path.basename', (['f'], {}), '(f)\n', (5975, 5978), False, 'import os\n')]
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. from __future__ import print_function import argparse import logging import os import time import math import reader import google.protobuf.text_format as text_format import numpy as np import six import paddle import paddle.fluid as fluid import paddle.fluid.proto.framework_pb2 as framework_pb2 import paddle.fluid.core as core from multiprocessing import cpu_count # disable gpu training for this example os.environ["CUDA_VISIBLE_DEVICES"] = "" logging.basicConfig(format='%(asctime)s - %(levelname)s - %(message)s') logger = logging.getLogger("fluid") logger.setLevel(logging.INFO) dense_feature_dim = 13 def parse_args(): parser = argparse.ArgumentParser(description="PaddlePaddle CTR example") parser.add_argument( '--train_data_path', type=str, default='./data/raw/train.txt', help="The path of training dataset") parser.add_argument( '--batch_size', type=int, default=1000, help="The size of mini-batch (default:1000)") parser.add_argument( '--embedding_size', type=int, default=10, help="The size for embedding layer (default:10)") parser.add_argument( '--sparse_feature_dim', type=int, default=1000001, help='sparse feature hashing space for index processing') parser.add_argument( '--model_output_dir', type=str, default='models', help='The path for model to store (default: models)') return parser.parse_args() def ctr_dnn_model(embedding_size, sparse_feature_dim, use_py_reader=True): def embedding_layer(input): emb = fluid.layers.embedding( input=input, is_sparse=True, # you need to patch https://github.com/PaddlePaddle/Paddle/pull/14190 # if you want to set is_distributed to True is_distributed=False, size=[sparse_feature_dim, embedding_size], param_attr=fluid.ParamAttr( name="SparseFeatFactors", initializer=fluid.initializer.Uniform())) seq = fluid.layers.sequence_pool(input=emb, pool_type='average') return emb, seq dense_input = fluid.layers.data( name="dense_input", shape=[dense_feature_dim], dtype='float32') sparse_input_ids = [ fluid.layers.data( name="C" + str(i), shape=[1], lod_level=1, dtype='int64') for i in range(1, 27) ] label = fluid.layers.data(name='label', shape=[1], dtype='int64') words = [dense_input] + sparse_input_ids + [label] sparse_embed_and_seq = list(map(embedding_layer, words[1:-1])) emb_list = [x[0] for x in sparse_embed_and_seq] sparse_embed_seq = [x[1] for x in sparse_embed_and_seq] concated = fluid.layers.concat(sparse_embed_seq + words[0:1], axis=1) train_feed_vars = words inference_feed_vars = emb_list + words[0:1] fc1 = fluid.layers.fc(input=concated, size=400, act='relu', param_attr=fluid.ParamAttr( initializer=fluid.initializer.Normal( scale=1 / math.sqrt(concated.shape[1])))) fc2 = fluid.layers.fc(input=fc1, size=400, act='relu', param_attr=fluid.ParamAttr( initializer=fluid.initializer.Normal( scale=1 / math.sqrt(fc1.shape[1])))) fc3 = fluid.layers.fc(input=fc2, size=400, act='relu', param_attr=fluid.ParamAttr( initializer=fluid.initializer.Normal( scale=1 / math.sqrt(fc2.shape[1])))) predict = fluid.layers.fc(input=fc3, size=2, act='softmax', param_attr=fluid.ParamAttr( initializer=fluid.initializer.Normal( scale=1 / math.sqrt(fc3.shape[1])))) cost = fluid.layers.cross_entropy(input=predict, label=words[-1]) avg_cost = fluid.layers.reduce_sum(cost) accuracy = fluid.layers.accuracy(input=predict, label=words[-1]) auc_var, batch_auc_var, auc_states = \ fluid.layers.auc(input=predict, label=words[-1], num_thresholds=2 ** 12, slide_steps=20) fetch_vars = [predict] return avg_cost, auc_var, batch_auc_var, train_feed_vars, inference_feed_vars, fetch_vars def train_loop(args, train_program, feed_vars, loss, auc_var, batch_auc_var, trainer_num, trainer_id): dataset = reader.CriteoDataset(args.sparse_feature_dim) train_reader = paddle.batch( paddle.reader.shuffle( dataset.train([args.train_data_path], trainer_num, trainer_id), buf_size=args.batch_size * 100), batch_size=args.batch_size) feed_var_names = [var.name for var in feed_vars] place = fluid.CPUPlace() exe = fluid.Executor(place) exe.run(fluid.default_startup_program()) total_time = 0 pass_id = 0 batch_id = 0 feeder = fluid.DataFeeder(feed_var_names, place) for data in train_reader(): loss_val, auc_val, batch_auc_val = exe.run( fluid.default_main_program(), feed=feeder.feed(data), fetch_list=[loss.name, auc_var.name, batch_auc_var.name]) break loss_val = np.mean(loss_val) auc_val = np.mean(auc_val) batch_auc_val = np.mean(batch_auc_val) logger.info("TRAIN --> pass: {} batch: {} loss: {} auc: {}, batch_auc: {}" .format(pass_id, batch_id, loss_val / args.batch_size, auc_val, batch_auc_val)) def save_program(): args = parse_args() if not os.path.isdir(args.model_output_dir): os.mkdir(args.model_output_dir) loss, auc_var, batch_auc_var, train_feed_vars, inference_feed_vars, fetch_vars = ctr_dnn_model( args.embedding_size, args.sparse_feature_dim, use_py_reader=False) optimizer = fluid.optimizer.Adam(learning_rate=1e-4) optimizer.minimize(loss) main_program = fluid.default_main_program() place = fluid.CPUPlace() exe = fluid.Executor(place) train_loop(args, main_program, train_feed_vars, loss, auc_var, batch_auc_var, 1, 0) model_dir = args.model_output_dir + "/inference_only" feed_var_names = [var.name for var in inference_feed_vars] fluid.io.save_inference_model(model_dir, feed_var_names, fetch_vars, exe, fluid.default_main_program()) def prune_program(): args = parse_args() model_dir = args.model_output_dir + "/inference_only" model_file = model_dir + "/__model__" with open(model_file, "rb") as f: protostr = f.read() f.close() proto = framework_pb2.ProgramDesc.FromString(six.binary_type(protostr)) block = proto.blocks[0] kept_ops = [op for op in block.ops if op.type != "lookup_table"] del block.ops[:] block.ops.extend(kept_ops) kept_vars = [var for var in block.vars if var.name != "SparseFeatFactors"] del block.vars[:] block.vars.extend(kept_vars) with open(model_file, "wb") as f: f.write(proto.SerializePartialToString()) f.close() with open(model_file + ".prototxt.pruned", "w") as f: f.write(text_format.MessageToString(proto)) f.close() def remove_embedding_param_file(): args = parse_args() model_dir = args.model_output_dir + "/inference_only" embedding_file = model_dir + "/SparseFeatFactors" os.remove(embedding_file) if __name__ == '__main__': save_program() prune_program() remove_embedding_param_file()
[ "logging.getLogger", "paddle.fluid.DataFeeder", "paddle.fluid.layers.auc", "paddle.fluid.layers.data", "paddle.fluid.layers.cross_entropy", "math.sqrt", "paddle.fluid.Executor", "paddle.fluid.optimizer.Adam", "os.remove", "numpy.mean", "argparse.ArgumentParser", "paddle.fluid.default_startup_p...
[((1065, 1136), 'logging.basicConfig', 'logging.basicConfig', ([], {'format': '"""%(asctime)s - %(levelname)s - %(message)s"""'}), "(format='%(asctime)s - %(levelname)s - %(message)s')\n", (1084, 1136), False, 'import logging\n'), ((1146, 1172), 'logging.getLogger', 'logging.getLogger', (['"""fluid"""'], {}), "('fluid')\n", (1163, 1172), False, 'import logging\n'), ((1260, 1323), 'argparse.ArgumentParser', 'argparse.ArgumentParser', ([], {'description': '"""PaddlePaddle CTR example"""'}), "(description='PaddlePaddle CTR example')\n", (1283, 1323), False, 'import argparse\n'), ((2814, 2900), 'paddle.fluid.layers.data', 'fluid.layers.data', ([], {'name': '"""dense_input"""', 'shape': '[dense_feature_dim]', 'dtype': '"""float32"""'}), "(name='dense_input', shape=[dense_feature_dim], dtype=\n 'float32')\n", (2831, 2900), True, 'import paddle.fluid as fluid\n'), ((3077, 3134), 'paddle.fluid.layers.data', 'fluid.layers.data', ([], {'name': '"""label"""', 'shape': '[1]', 'dtype': '"""int64"""'}), "(name='label', shape=[1], dtype='int64')\n", (3094, 3134), True, 'import paddle.fluid as fluid\n'), ((3387, 3445), 'paddle.fluid.layers.concat', 'fluid.layers.concat', (['(sparse_embed_seq + words[0:1])'], {'axis': '(1)'}), '(sparse_embed_seq + words[0:1], axis=1)\n', (3406, 3445), True, 'import paddle.fluid as fluid\n'), ((4787, 4845), 'paddle.fluid.layers.cross_entropy', 'fluid.layers.cross_entropy', ([], {'input': 'predict', 'label': 'words[-1]'}), '(input=predict, label=words[-1])\n', (4813, 4845), True, 'import paddle.fluid as fluid\n'), ((4861, 4890), 'paddle.fluid.layers.reduce_sum', 'fluid.layers.reduce_sum', (['cost'], {}), '(cost)\n', (4884, 4890), True, 'import paddle.fluid as fluid\n'), ((4906, 4959), 'paddle.fluid.layers.accuracy', 'fluid.layers.accuracy', ([], {'input': 'predict', 'label': 'words[-1]'}), '(input=predict, label=words[-1])\n', (4927, 4959), True, 'import paddle.fluid as fluid\n'), ((5011, 5103), 'paddle.fluid.layers.auc', 'fluid.layers.auc', ([], {'input': 'predict', 'label': 'words[-1]', 'num_thresholds': '(2 ** 12)', 'slide_steps': '(20)'}), '(input=predict, label=words[-1], num_thresholds=2 ** 12,\n slide_steps=20)\n', (5027, 5103), True, 'import paddle.fluid as fluid\n'), ((5356, 5401), 'reader.CriteoDataset', 'reader.CriteoDataset', (['args.sparse_feature_dim'], {}), '(args.sparse_feature_dim)\n', (5376, 5401), False, 'import reader\n'), ((5689, 5705), 'paddle.fluid.CPUPlace', 'fluid.CPUPlace', ([], {}), '()\n', (5703, 5705), True, 'import paddle.fluid as fluid\n'), ((5716, 5737), 'paddle.fluid.Executor', 'fluid.Executor', (['place'], {}), '(place)\n', (5730, 5737), True, 'import paddle.fluid as fluid\n'), ((5851, 5890), 'paddle.fluid.DataFeeder', 'fluid.DataFeeder', (['feed_var_names', 'place'], {}), '(feed_var_names, place)\n', (5867, 5890), True, 'import paddle.fluid as fluid\n'), ((6153, 6170), 'numpy.mean', 'np.mean', (['loss_val'], {}), '(loss_val)\n', (6160, 6170), True, 'import numpy as np\n'), ((6185, 6201), 'numpy.mean', 'np.mean', (['auc_val'], {}), '(auc_val)\n', (6192, 6201), True, 'import numpy as np\n'), ((6222, 6244), 'numpy.mean', 'np.mean', (['batch_auc_val'], {}), '(batch_auc_val)\n', (6229, 6244), True, 'import numpy as np\n'), ((6774, 6816), 'paddle.fluid.optimizer.Adam', 'fluid.optimizer.Adam', ([], {'learning_rate': '(0.0001)'}), '(learning_rate=0.0001)\n', (6794, 6816), True, 'import paddle.fluid as fluid\n'), ((6863, 6891), 'paddle.fluid.default_main_program', 'fluid.default_main_program', ([], {}), '()\n', (6889, 6891), True, 'import paddle.fluid as fluid\n'), ((6905, 6921), 'paddle.fluid.CPUPlace', 'fluid.CPUPlace', ([], {}), '()\n', (6919, 6921), True, 'import paddle.fluid as fluid\n'), ((6932, 6953), 'paddle.fluid.Executor', 'fluid.Executor', (['place'], {}), '(place)\n', (6946, 6953), True, 'import paddle.fluid as fluid\n'), ((8312, 8337), 'os.remove', 'os.remove', (['embedding_file'], {}), '(embedding_file)\n', (8321, 8337), False, 'import os\n'), ((2712, 2770), 'paddle.fluid.layers.sequence_pool', 'fluid.layers.sequence_pool', ([], {'input': 'emb', 'pool_type': '"""average"""'}), "(input=emb, pool_type='average')\n", (2738, 2770), True, 'import paddle.fluid as fluid\n'), ((5751, 5782), 'paddle.fluid.default_startup_program', 'fluid.default_startup_program', ([], {}), '()\n', (5780, 5782), True, 'import paddle.fluid as fluid\n'), ((6503, 6539), 'os.path.isdir', 'os.path.isdir', (['args.model_output_dir'], {}), '(args.model_output_dir)\n', (6516, 6539), False, 'import os\n'), ((6549, 6580), 'os.mkdir', 'os.mkdir', (['args.model_output_dir'], {}), '(args.model_output_dir)\n', (6557, 6580), False, 'import os\n'), ((7291, 7319), 'paddle.fluid.default_main_program', 'fluid.default_main_program', ([], {}), '()\n', (7317, 7319), True, 'import paddle.fluid as fluid\n'), ((7597, 7622), 'six.binary_type', 'six.binary_type', (['protostr'], {}), '(protostr)\n', (7612, 7622), False, 'import six\n'), ((5987, 6015), 'paddle.fluid.default_main_program', 'fluid.default_main_program', ([], {}), '()\n', (6013, 6015), True, 'import paddle.fluid as fluid\n'), ((8085, 8119), 'google.protobuf.text_format.MessageToString', 'text_format.MessageToString', (['proto'], {}), '(proto)\n', (8112, 8119), True, 'import google.protobuf.text_format as text_format\n'), ((2668, 2695), 'paddle.fluid.initializer.Uniform', 'fluid.initializer.Uniform', ([], {}), '()\n', (2693, 2695), True, 'import paddle.fluid as fluid\n'), ((3806, 3834), 'math.sqrt', 'math.sqrt', (['concated.shape[1]'], {}), '(concated.shape[1])\n', (3815, 3834), False, 'import math\n'), ((4115, 4138), 'math.sqrt', 'math.sqrt', (['fc1.shape[1]'], {}), '(fc1.shape[1])\n', (4124, 4138), False, 'import math\n'), ((4419, 4442), 'math.sqrt', 'math.sqrt', (['fc2.shape[1]'], {}), '(fc2.shape[1])\n', (4428, 4442), False, 'import math\n'), ((4748, 4771), 'math.sqrt', 'math.sqrt', (['fc3.shape[1]'], {}), '(fc3.shape[1])\n', (4757, 4771), False, 'import math\n')]
# -*- coding: utf-8 -*- # test_calculateTF.py # This module provides the tests for the calculateTF() function. # Copyright 2014 <NAME> & <NAME> # This file is part of python-deltasigma. # # python-deltasigma is a 1:1 Python replacement of Richard Schreier's # MATLAB delta sigma toolbox (aka "delsigma"), upon which it is heavily based. # The delta sigma toolbox is (c) 2009, <NAME>. # # python-deltasigma 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 # LICENSE file for the licensing terms. import unittest import numpy as np import deltasigma as ds from deltasigma._utils import cplxpair class TestCalculateTF(unittest.TestCase): """Test function for calculateTF()""" def setUp(self): ABCD = [[1.0, 0.0, 0.0, 0.044408783846879, -0.044408783846879], [0.999036450096481, 0.997109907515262, -0.005777399147297, 0.0, 0.499759089304780], [0.499759089304780, 0.999036450096481, 0.997109907515262, 0.0, -0.260002096136488], [0.0, 0.0, 1.0, 0.0, 0.0]] ABCD = np.array(ABCD) ntf, stf = ds.calculateTF(ABCD) ntf_zeros, ntf_poles, _ = ntf stf_zeros, stf_poles, _ = stf mntf_poles = np.array((1.498975311463384, 1.102565142679772, 0.132677264750882)) mntf_zeros = np.array((0.997109907515262 + 0.075972576202904j, 0.997109907515262 - 0.075972576202904j, 1.000000000000000 + 0.000000000000000j)) mstf_zeros = np.array((-0.999999999999996,)) mstf_poles = np.array((1.498975311463384, 1.102565142679772, 0.132677264750882)) # for some reason, sometimes the zeros are in different order. self.ntf_zeros, self.mntf_zeros = (cplxpair(ntf_zeros), cplxpair(mntf_zeros)) self.stf_zeros, self.mstf_zeros = (cplxpair(stf_zeros), cplxpair(mstf_zeros)) self.ntf_poles, self.mntf_poles = (cplxpair(ntf_poles), cplxpair(mntf_poles)) self.stf_poles, self.mstf_poles = (cplxpair(stf_poles), cplxpair(mstf_poles)) def test_calculateTF1(self): """Test function for calculateTF() 1/4""" # ntf zeros self.assertTrue(np.allclose(self.ntf_zeros, self.mntf_zeros, rtol=1e-5, atol=1e-8)) # ntf poles self.assertTrue(np.allclose(self.ntf_poles, self.mntf_poles, rtol=1e-5, atol=1e-8)) # stf zeros self.assertTrue(np.allclose(self.stf_zeros, self.mstf_zeros, rtol=1e-5, atol=1e-8)) # stf poles self.assertTrue(np.allclose(self.stf_poles, self.mstf_poles, rtol=1e-5, atol=1e-8)) def test_calculateTF2(self): """Test function for calculateTF() 2/4""" # test an easy TF ABCD = np.array([[1., 1., -1.], [1., 0., 0.]]) k = 1. ntf, stf = ds.calculateTF(ABCD, k) ntf_zeros, ntf_poles, ntf_gain = ntf stf_zeros, stf_poles, stf_gain = stf self.assertTrue(np.allclose(stf_poles, [0.], rtol=1e-5, atol=1e-8)) self.assertTrue(not len(stf_zeros)) self.assertTrue(np.allclose(stf_gain, 1., rtol=1e-5, atol=1e-8)) self.assertTrue(np.allclose(ntf_poles, [0.], rtol=1e-5, atol=1e-8)) self.assertTrue(np.allclose(ntf_zeros, [1.], rtol=1e-5, atol=1e-8)) self.assertTrue(np.allclose(ntf_gain, 1., rtol=1e-5, atol=1e-8)) def test_calculateTF3(self): """Test function for calculateTF() 3/4""" # test for the default k value ABCD = np.array([[1., 1., -1.], [1., 0., 0.]]) ntf, stf = ds.calculateTF(ABCD) ntf_zeros, ntf_poles, ntf_gain = ntf stf_zeros, stf_poles, stf_gain = stf self.assertTrue(np.allclose(stf_poles, [0.], rtol=1e-5, atol=1e-8)) self.assertTrue(not len(stf_zeros)) self.assertTrue(np.allclose(stf_gain, 1., rtol=1e-5, atol=1e-8)) self.assertTrue(np.allclose(ntf_poles, [0.], rtol=1e-5, atol=1e-8)) self.assertTrue(np.allclose(ntf_zeros, [1.], rtol=1e-5, atol=1e-8)) self.assertTrue(np.allclose(ntf_gain, 1., rtol=1e-5, atol=1e-8)) def test_calculateTF4(self): """Test function for calculateTF() 4/4""" # Easy test for a 2-quantizers system # MASH 1-0 cascade ABCD = [[1, 1, -1, 0], [1, 0, 0, 0], [1, 0, -1, 0]] ABCD = np.array(ABCD, dtype=np.float_) k = [1., 1.] # here we get back arrays of transfer functions ntfs, stfs = ds.calculateTF(ABCD, k=k) # stfs self.assertTrue(np.allclose(stfs[0][1], [0.], rtol=1e-5, atol=1e-8)) self.assertTrue(not len(stfs[0][0])) self.assertTrue(np.allclose(stfs[0][2], 1., rtol=1e-5, atol=1e-8)) self.assertTrue(np.allclose(stfs[1][1], [0.], rtol=1e-5, atol=1e-8)) self.assertTrue(not len(stfs[1][0])) self.assertTrue(np.allclose(stfs[1][2], 1., rtol=1e-5, atol=1e-8)) # e1 to V1 self.assertTrue(np.allclose(ntfs[0, 0][1], [0.], rtol=1e-5, atol=1e-8)) self.assertTrue(np.allclose(ntfs[0, 0][0], [1.], rtol=1e-5, atol=1e-8)) self.assertTrue(np.allclose(ntfs[0, 0][2], 1., rtol=1e-5, atol=1e-8)) # e1 to V2 self.assertTrue(np.allclose(ntfs[1, 0][1], [0.], rtol=1e-5, atol=1e-8)) self.assertTrue(not len(ntfs[1, 0][0])) self.assertTrue(np.allclose(ntfs[1, 0][2], -1., rtol=1e-5, atol=1e-8)) # e2 to V2 self.assertTrue(not len(ntfs[1, 1][0])) self.assertTrue(not len(ntfs[1, 1][1])) self.assertTrue(np.allclose(ntfs[1, 1][2], 1., rtol=1e-5, atol=1e-8)) # e2 to V1 self.assertTrue(np.allclose(ntfs[0, 1][2], 0., rtol=1e-5, atol=1e-8))
[ "numpy.array", "numpy.allclose", "deltasigma.calculateTF", "deltasigma._utils.cplxpair" ]
[((1198, 1212), 'numpy.array', 'np.array', (['ABCD'], {}), '(ABCD)\n', (1206, 1212), True, 'import numpy as np\n'), ((1232, 1252), 'deltasigma.calculateTF', 'ds.calculateTF', (['ABCD'], {}), '(ABCD)\n', (1246, 1252), True, 'import deltasigma as ds\n'), ((1350, 1417), 'numpy.array', 'np.array', (['(1.498975311463384, 1.102565142679772, 0.132677264750882)'], {}), '((1.498975311463384, 1.102565142679772, 0.132677264750882))\n', (1358, 1417), True, 'import numpy as np\n'), ((1470, 1577), 'numpy.array', 'np.array', (['(0.997109907515262 + 0.075972576202904j, 0.997109907515262 - \n 0.075972576202904j, 1.0 + 0.0j)'], {}), '((0.997109907515262 + 0.075972576202904j, 0.997109907515262 - \n 0.075972576202904j, 1.0 + 0.0j))\n', (1478, 1577), True, 'import numpy as np\n'), ((1684, 1715), 'numpy.array', 'np.array', (['(-0.999999999999996,)'], {}), '((-0.999999999999996,))\n', (1692, 1715), True, 'import numpy as np\n'), ((1737, 1804), 'numpy.array', 'np.array', (['(1.498975311463384, 1.102565142679772, 0.132677264750882)'], {}), '((1.498975311463384, 1.102565142679772, 0.132677264750882))\n', (1745, 1804), True, 'import numpy as np\n'), ((3177, 3222), 'numpy.array', 'np.array', (['[[1.0, 1.0, -1.0], [1.0, 0.0, 0.0]]'], {}), '([[1.0, 1.0, -1.0], [1.0, 0.0, 0.0]])\n', (3185, 3222), True, 'import numpy as np\n'), ((3275, 3298), 'deltasigma.calculateTF', 'ds.calculateTF', (['ABCD', 'k'], {}), '(ABCD, k)\n', (3289, 3298), True, 'import deltasigma as ds\n'), ((3946, 3991), 'numpy.array', 'np.array', (['[[1.0, 1.0, -1.0], [1.0, 0.0, 0.0]]'], {}), '([[1.0, 1.0, -1.0], [1.0, 0.0, 0.0]])\n', (3954, 3991), True, 'import numpy as np\n'), ((4029, 4049), 'deltasigma.calculateTF', 'ds.calculateTF', (['ABCD'], {}), '(ABCD)\n', (4043, 4049), True, 'import deltasigma as ds\n'), ((4822, 4853), 'numpy.array', 'np.array', (['ABCD'], {'dtype': 'np.float_'}), '(ABCD, dtype=np.float_)\n', (4830, 4853), True, 'import numpy as np\n'), ((4952, 4977), 'deltasigma.calculateTF', 'ds.calculateTF', (['ABCD'], {'k': 'k'}), '(ABCD, k=k)\n', (4966, 4977), True, 'import deltasigma as ds\n'), ((1951, 1970), 'deltasigma._utils.cplxpair', 'cplxpair', (['ntf_zeros'], {}), '(ntf_zeros)\n', (1959, 1970), False, 'from deltasigma._utils import cplxpair\n'), ((2015, 2035), 'deltasigma._utils.cplxpair', 'cplxpair', (['mntf_zeros'], {}), '(mntf_zeros)\n', (2023, 2035), False, 'from deltasigma._utils import cplxpair\n'), ((2080, 2099), 'deltasigma._utils.cplxpair', 'cplxpair', (['stf_zeros'], {}), '(stf_zeros)\n', (2088, 2099), False, 'from deltasigma._utils import cplxpair\n'), ((2144, 2164), 'deltasigma._utils.cplxpair', 'cplxpair', (['mstf_zeros'], {}), '(mstf_zeros)\n', (2152, 2164), False, 'from deltasigma._utils import cplxpair\n'), ((2209, 2228), 'deltasigma._utils.cplxpair', 'cplxpair', (['ntf_poles'], {}), '(ntf_poles)\n', (2217, 2228), False, 'from deltasigma._utils import cplxpair\n'), ((2273, 2293), 'deltasigma._utils.cplxpair', 'cplxpair', (['mntf_poles'], {}), '(mntf_poles)\n', (2281, 2293), False, 'from deltasigma._utils import cplxpair\n'), ((2338, 2357), 'deltasigma._utils.cplxpair', 'cplxpair', (['stf_poles'], {}), '(stf_poles)\n', (2346, 2357), False, 'from deltasigma._utils import cplxpair\n'), ((2402, 2422), 'deltasigma._utils.cplxpair', 'cplxpair', (['mstf_poles'], {}), '(mstf_poles)\n', (2410, 2422), False, 'from deltasigma._utils import cplxpair\n'), ((2552, 2620), 'numpy.allclose', 'np.allclose', (['self.ntf_zeros', 'self.mntf_zeros'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(self.ntf_zeros, self.mntf_zeros, rtol=1e-05, atol=1e-08)\n', (2563, 2620), True, 'import numpy as np\n'), ((2688, 2756), 'numpy.allclose', 'np.allclose', (['self.ntf_poles', 'self.mntf_poles'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(self.ntf_poles, self.mntf_poles, rtol=1e-05, atol=1e-08)\n', (2699, 2756), True, 'import numpy as np\n'), ((2824, 2892), 'numpy.allclose', 'np.allclose', (['self.stf_zeros', 'self.mstf_zeros'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(self.stf_zeros, self.mstf_zeros, rtol=1e-05, atol=1e-08)\n', (2835, 2892), True, 'import numpy as np\n'), ((2960, 3028), 'numpy.allclose', 'np.allclose', (['self.stf_poles', 'self.mstf_poles'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(self.stf_poles, self.mstf_poles, rtol=1e-05, atol=1e-08)\n', (2971, 3028), True, 'import numpy as np\n'), ((3414, 3467), 'numpy.allclose', 'np.allclose', (['stf_poles', '[0.0]'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(stf_poles, [0.0], rtol=1e-05, atol=1e-08)\n', (3425, 3467), True, 'import numpy as np\n'), ((3534, 3584), 'numpy.allclose', 'np.allclose', (['stf_gain', '(1.0)'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(stf_gain, 1.0, rtol=1e-05, atol=1e-08)\n', (3545, 3584), True, 'import numpy as np\n'), ((3607, 3660), 'numpy.allclose', 'np.allclose', (['ntf_poles', '[0.0]'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntf_poles, [0.0], rtol=1e-05, atol=1e-08)\n', (3618, 3660), True, 'import numpy as np\n'), ((3683, 3736), 'numpy.allclose', 'np.allclose', (['ntf_zeros', '[1.0]'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntf_zeros, [1.0], rtol=1e-05, atol=1e-08)\n', (3694, 3736), True, 'import numpy as np\n'), ((3759, 3809), 'numpy.allclose', 'np.allclose', (['ntf_gain', '(1.0)'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntf_gain, 1.0, rtol=1e-05, atol=1e-08)\n', (3770, 3809), True, 'import numpy as np\n'), ((4164, 4217), 'numpy.allclose', 'np.allclose', (['stf_poles', '[0.0]'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(stf_poles, [0.0], rtol=1e-05, atol=1e-08)\n', (4175, 4217), True, 'import numpy as np\n'), ((4284, 4334), 'numpy.allclose', 'np.allclose', (['stf_gain', '(1.0)'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(stf_gain, 1.0, rtol=1e-05, atol=1e-08)\n', (4295, 4334), True, 'import numpy as np\n'), ((4357, 4410), 'numpy.allclose', 'np.allclose', (['ntf_poles', '[0.0]'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntf_poles, [0.0], rtol=1e-05, atol=1e-08)\n', (4368, 4410), True, 'import numpy as np\n'), ((4433, 4486), 'numpy.allclose', 'np.allclose', (['ntf_zeros', '[1.0]'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntf_zeros, [1.0], rtol=1e-05, atol=1e-08)\n', (4444, 4486), True, 'import numpy as np\n'), ((4509, 4559), 'numpy.allclose', 'np.allclose', (['ntf_gain', '(1.0)'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntf_gain, 1.0, rtol=1e-05, atol=1e-08)\n', (4520, 4559), True, 'import numpy as np\n'), ((5017, 5071), 'numpy.allclose', 'np.allclose', (['stfs[0][1]', '[0.0]'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(stfs[0][1], [0.0], rtol=1e-05, atol=1e-08)\n', (5028, 5071), True, 'import numpy as np\n'), ((5139, 5191), 'numpy.allclose', 'np.allclose', (['stfs[0][2]', '(1.0)'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(stfs[0][2], 1.0, rtol=1e-05, atol=1e-08)\n', (5150, 5191), True, 'import numpy as np\n'), ((5214, 5268), 'numpy.allclose', 'np.allclose', (['stfs[1][1]', '[0.0]'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(stfs[1][1], [0.0], rtol=1e-05, atol=1e-08)\n', (5225, 5268), True, 'import numpy as np\n'), ((5336, 5388), 'numpy.allclose', 'np.allclose', (['stfs[1][2]', '(1.0)'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(stfs[1][2], 1.0, rtol=1e-05, atol=1e-08)\n', (5347, 5388), True, 'import numpy as np\n'), ((5430, 5487), 'numpy.allclose', 'np.allclose', (['ntfs[0, 0][1]', '[0.0]'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntfs[0, 0][1], [0.0], rtol=1e-05, atol=1e-08)\n', (5441, 5487), True, 'import numpy as np\n'), ((5510, 5567), 'numpy.allclose', 'np.allclose', (['ntfs[0, 0][0]', '[1.0]'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntfs[0, 0][0], [1.0], rtol=1e-05, atol=1e-08)\n', (5521, 5567), True, 'import numpy as np\n'), ((5590, 5645), 'numpy.allclose', 'np.allclose', (['ntfs[0, 0][2]', '(1.0)'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntfs[0, 0][2], 1.0, rtol=1e-05, atol=1e-08)\n', (5601, 5645), True, 'import numpy as np\n'), ((5687, 5744), 'numpy.allclose', 'np.allclose', (['ntfs[1, 0][1]', '[0.0]'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntfs[1, 0][1], [0.0], rtol=1e-05, atol=1e-08)\n', (5698, 5744), True, 'import numpy as np\n'), ((5815, 5871), 'numpy.allclose', 'np.allclose', (['ntfs[1, 0][2]', '(-1.0)'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntfs[1, 0][2], -1.0, rtol=1e-05, atol=1e-08)\n', (5826, 5871), True, 'import numpy as np\n'), ((6009, 6064), 'numpy.allclose', 'np.allclose', (['ntfs[1, 1][2]', '(1.0)'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntfs[1, 1][2], 1.0, rtol=1e-05, atol=1e-08)\n', (6020, 6064), True, 'import numpy as np\n'), ((6106, 6161), 'numpy.allclose', 'np.allclose', (['ntfs[0, 1][2]', '(0.0)'], {'rtol': '(1e-05)', 'atol': '(1e-08)'}), '(ntfs[0, 1][2], 0.0, rtol=1e-05, atol=1e-08)\n', (6117, 6161), True, 'import numpy as np\n')]
import pytest from pymbar import testsystems from reproducibility_project.src.analysis.equilibration import is_equilibrated from reproducibility_project.src.analysis.sampler import ( _decorr_sampling, write_subsampled_values, ) from reproducibility_project.tests.base_test import BaseTest class TestSampler(BaseTest): def test_not_equilibrated(self, correlated_data_tau100_n10000): data = correlated_data_tau100_n10000 with pytest.raises(ValueError): start, stop, step = _decorr_sampling( data, threshold_fraction=0.80, threshold_neff=100 ) def test_equilibrated(self, correlated_data_tau100_n10000): data = correlated_data_tau100_n10000 is_equil, prod_start, ineff, Neff = is_equilibrated( data, threshold_fraction=0.10, threshold_neff=10, nskip=1, ) start, stop, step, neff_samples = _decorr_sampling( data, threshold_fraction=0.10, threshold_neff=10 ) assert start >= prod_start assert step >= ineff assert neff_samples > 1 def test_write_subsampled_incorrect_params(self, tmp_job): with pytest.raises( TypeError, match=r"Expected input \'job\' of type signac\.contrib\.project\.Job", ): write_subsampled_values( "foo", property="density", ) with pytest.raises( ValueError, match=r"Expected \'property\' to be a name of a property", ): write_subsampled_values( tmp_job, property="", ) with pytest.raises( ValueError, match=r"Attempting to overwrite already existing data for property", ): import numpy as np vals = [1, 2, 3, 4, 5, 6] out = f"""foo\n""" for val in vals: out = out + str(val) + "\n" with open(tmp_job.fn("log.txt"), "w") as fp: fp.write(out) tmp_job.data["subsamples/foo"] = np.asarray([1, 2, 3, 4]) tmp_job.doc["sampling_results"] = { "foo": {"start": 1, "stop": 4, "step": 2, "Neff": 2} } write_subsampled_values(tmp_job, property="foo", overwrite=False) def test_file_missing(self, tmp_job): # by default, the tmp job is missing the file with pytest.raises( FileNotFoundError, match=r"File missing\.txt does not exist" ): tmp_job.doc["sampling_results"] = { "foo": {"start": 1, "stop": 4, "step": 2, "Neff": 2} } write_subsampled_values( tmp_job, property="foo", property_filename="missing.txt", overwrite=False, ) def test_correct_samples(self, tmp_job): import numpy as np vals = [1, 2, 3, 4, 5, 6] out = f"""foo\n""" for val in vals: out = out + str(val) + "\n" with open(tmp_job.fn("log.txt"), "w") as fp: fp.write(out) tmp_job.doc["sampling_results"] = { "foo": {"start": 1, "stop": 4, "step": 1, "Neff": 4} } write_subsampled_values(tmp_job, property="foo", overwrite=False) with tmp_job.data: assert len(tmp_job.data["subsamples/foo"]) == 3 np.testing.assert_array_equal( tmp_job.data["subsamples/foo"], [2, 3, 4] )
[ "reproducibility_project.src.analysis.sampler.write_subsampled_values", "numpy.asarray", "pytest.raises", "reproducibility_project.src.analysis.sampler._decorr_sampling", "numpy.testing.assert_array_equal", "reproducibility_project.src.analysis.equilibration.is_equilibrated" ]
[((766, 839), 'reproducibility_project.src.analysis.equilibration.is_equilibrated', 'is_equilibrated', (['data'], {'threshold_fraction': '(0.1)', 'threshold_neff': '(10)', 'nskip': '(1)'}), '(data, threshold_fraction=0.1, threshold_neff=10, nskip=1)\n', (781, 839), False, 'from reproducibility_project.src.analysis.equilibration import is_equilibrated\n'), ((942, 1007), 'reproducibility_project.src.analysis.sampler._decorr_sampling', '_decorr_sampling', (['data'], {'threshold_fraction': '(0.1)', 'threshold_neff': '(10)'}), '(data, threshold_fraction=0.1, threshold_neff=10)\n', (958, 1007), False, 'from reproducibility_project.src.analysis.sampler import _decorr_sampling, write_subsampled_values\n'), ((3308, 3373), 'reproducibility_project.src.analysis.sampler.write_subsampled_values', 'write_subsampled_values', (['tmp_job'], {'property': '"""foo"""', 'overwrite': '(False)'}), "(tmp_job, property='foo', overwrite=False)\n", (3331, 3373), False, 'from reproducibility_project.src.analysis.sampler import _decorr_sampling, write_subsampled_values\n'), ((455, 480), 'pytest.raises', 'pytest.raises', (['ValueError'], {}), '(ValueError)\n', (468, 480), False, 'import pytest\n'), ((514, 580), 'reproducibility_project.src.analysis.sampler._decorr_sampling', '_decorr_sampling', (['data'], {'threshold_fraction': '(0.8)', 'threshold_neff': '(100)'}), '(data, threshold_fraction=0.8, threshold_neff=100)\n', (530, 580), False, 'from reproducibility_project.src.analysis.sampler import _decorr_sampling, write_subsampled_values\n'), ((1204, 1308), 'pytest.raises', 'pytest.raises', (['TypeError'], {'match': '"""Expected input \\\\\'job\\\\\' of type signac\\\\.contrib\\\\.project\\\\.Job"""'}), '(TypeError, match=\n "Expected input \\\\\'job\\\\\' of type signac\\\\.contrib\\\\.project\\\\.Job")\n', (1217, 1308), False, 'import pytest\n'), ((1348, 1398), 'reproducibility_project.src.analysis.sampler.write_subsampled_values', 'write_subsampled_values', (['"""foo"""'], {'property': '"""density"""'}), "('foo', property='density')\n", (1371, 1398), False, 'from reproducibility_project.src.analysis.sampler import _decorr_sampling, write_subsampled_values\n'), ((1460, 1550), 'pytest.raises', 'pytest.raises', (['ValueError'], {'match': '"""Expected \\\\\'property\\\\\' to be a name of a property"""'}), '(ValueError, match=\n "Expected \\\\\'property\\\\\' to be a name of a property")\n', (1473, 1550), False, 'import pytest\n'), ((1593, 1638), 'reproducibility_project.src.analysis.sampler.write_subsampled_values', 'write_subsampled_values', (['tmp_job'], {'property': '""""""'}), "(tmp_job, property='')\n", (1616, 1638), False, 'from reproducibility_project.src.analysis.sampler import _decorr_sampling, write_subsampled_values\n'), ((1700, 1798), 'pytest.raises', 'pytest.raises', (['ValueError'], {'match': '"""Attempting to overwrite already existing data for property"""'}), "(ValueError, match=\n 'Attempting to overwrite already existing data for property')\n", (1713, 1798), False, 'import pytest\n'), ((2137, 2161), 'numpy.asarray', 'np.asarray', (['[1, 2, 3, 4]'], {}), '([1, 2, 3, 4])\n', (2147, 2161), True, 'import numpy as np\n'), ((2306, 2371), 'reproducibility_project.src.analysis.sampler.write_subsampled_values', 'write_subsampled_values', (['tmp_job'], {'property': '"""foo"""', 'overwrite': '(False)'}), "(tmp_job, property='foo', overwrite=False)\n", (2329, 2371), False, 'from reproducibility_project.src.analysis.sampler import _decorr_sampling, write_subsampled_values\n'), ((2482, 2557), 'pytest.raises', 'pytest.raises', (['FileNotFoundError'], {'match': '"""File missing\\\\.txt does not exist"""'}), "(FileNotFoundError, match='File missing\\\\.txt does not exist')\n", (2495, 2557), False, 'import pytest\n'), ((2724, 2827), 'reproducibility_project.src.analysis.sampler.write_subsampled_values', 'write_subsampled_values', (['tmp_job'], {'property': '"""foo"""', 'property_filename': '"""missing.txt"""', 'overwrite': '(False)'}), "(tmp_job, property='foo', property_filename=\n 'missing.txt', overwrite=False)\n", (2747, 2827), False, 'from reproducibility_project.src.analysis.sampler import _decorr_sampling, write_subsampled_values\n'), ((3473, 3545), 'numpy.testing.assert_array_equal', 'np.testing.assert_array_equal', (["tmp_job.data['subsamples/foo']", '[2, 3, 4]'], {}), "(tmp_job.data['subsamples/foo'], [2, 3, 4])\n", (3502, 3545), True, 'import numpy as np\n')]
import logging logging.basicConfig(level=logging.INFO) import numpy as np import functions def model(X_train, Y_train, X_test, Y_test, iterations=2500, learning_rate=0.005): w = np.zeros((X_train.shape[0], 1)) b = 0 parameters, grads, costs = functions.optimize( w, b, X_train, Y_train, iterations=iterations, learning_rate=learning_rate ) w = parameters["w"] b = parameters["b"] Y_prediction_test = functions.predict(w, b, X_test) Y_prediction_train = functions.predict(w, b, X_train) train_accuracy = 100 - np.mean(np.abs(Y_prediction_train - Y_train)) * 100 logging.info(f"train accuracy: {train_accuracy} %") test_accuracy = 100 - np.mean(np.abs(Y_prediction_test - Y_test)) * 100 logging.info(f"test accuracy: {test_accuracy} %") return { "costs": costs, "Y_prediction_test": Y_prediction_test, "Y_prediction_train": Y_prediction_train, "w": w, "b": b, "learning_rate": learning_rate, "iterations": iterations }
[ "logging.basicConfig", "numpy.abs", "functions.predict", "numpy.zeros", "logging.info", "functions.optimize" ]
[((16, 55), 'logging.basicConfig', 'logging.basicConfig', ([], {'level': 'logging.INFO'}), '(level=logging.INFO)\n', (35, 55), False, 'import logging\n'), ((186, 217), 'numpy.zeros', 'np.zeros', (['(X_train.shape[0], 1)'], {}), '((X_train.shape[0], 1))\n', (194, 217), True, 'import numpy as np\n'), ((260, 358), 'functions.optimize', 'functions.optimize', (['w', 'b', 'X_train', 'Y_train'], {'iterations': 'iterations', 'learning_rate': 'learning_rate'}), '(w, b, X_train, Y_train, iterations=iterations,\n learning_rate=learning_rate)\n', (278, 358), False, 'import functions\n'), ((484, 515), 'functions.predict', 'functions.predict', (['w', 'b', 'X_test'], {}), '(w, b, X_test)\n', (501, 515), False, 'import functions\n'), ((541, 573), 'functions.predict', 'functions.predict', (['w', 'b', 'X_train'], {}), '(w, b, X_train)\n', (558, 573), False, 'import functions\n'), ((658, 709), 'logging.info', 'logging.info', (['f"""train accuracy: {train_accuracy} %"""'], {}), "(f'train accuracy: {train_accuracy} %')\n", (670, 709), False, 'import logging\n'), ((791, 840), 'logging.info', 'logging.info', (['f"""test accuracy: {test_accuracy} %"""'], {}), "(f'test accuracy: {test_accuracy} %')\n", (803, 840), False, 'import logging\n'), ((610, 646), 'numpy.abs', 'np.abs', (['(Y_prediction_train - Y_train)'], {}), '(Y_prediction_train - Y_train)\n', (616, 646), True, 'import numpy as np\n'), ((745, 779), 'numpy.abs', 'np.abs', (['(Y_prediction_test - Y_test)'], {}), '(Y_prediction_test - Y_test)\n', (751, 779), True, 'import numpy as np\n')]
''' This script plots the four test predictions of the predictive expectation and variance of BAR-DenseED seen in Figure 13 of the paper. === Distributed by: <NAME> (MIT Liscense) - Associated publication: url: http://www.sciencedirect.com/science/article/pii/S0021999119307612 doi: https://doi.org/10.1016/j.jcp.2019.109056 github: https://github.com/cics-nd/ar-pde-cnn === ''' import sys sys.path.append("..") # Adds higher directory to python modules path. from args import Parser from nn.denseEDcirc import DenseED from nn.bayesNN import BayesNN from nn.swag import SwagNN from utils.utils import mkdirs from utils.burgerLoader import BurgerLoader import matplotlib as mpl import matplotlib.pyplot as plt import matplotlib.cm as cm from matplotlib import rc import matplotlib.gridspec as gridspec import torch import numpy as np import os import time def testSample(args, swag_nn, test_loader, tstep=100, n_samples=10): ''' Tests the samples of the Bayesian SWAG model Args: args (argparse): object with programs arguements model (PyTorch model): DenseED model to be tested test_loader (dataloader): dataloader with test cases (use createTestingLoader) tstep (int): number of timesteps to predict for n_samples (int): number of model samples to draw Returns: u_out (torch.Tensor): [d x nsamples x tstep x nel] predicted quantities of each sample u_target (torch.Tensor): [d x tstep x nel] respective target values loaded from simulator ''' mb_size = int(len(test_loader.dataset)/len(test_loader)) u_out = torch.zeros(mb_size, n_samples, tstep+1, args.nel) betas = torch.zeros(n_samples) for i in range(n_samples): print('Executing model sample {:d}'.format(i)) model = swag_nn.sample(diagCov=True) model.eval() betas[i] = model.model.log_beta.exp() for batch_idx, (input0, uTarget0) in enumerate(test_loader): input = input0.to(args.device) u_target = uTarget0 u_out[:,i,0,:] = input[:,0] # Auto-regress for t_idx in range(tstep): uPred = model(input[:,-args.nic:,:]) u_out[:,i,t_idx+1,:] = uPred[:,0] input = input[:,-int(args.nic-1):,:].detach() input0 = uPred[:,0,:].unsqueeze(1).detach() input = torch.cat([input, input0], dim=1) # Only do the first mini-batch break return u_out, betas, u_target def plotContourGrid(t, xT, uPred, betas, uTarget): ''' Creates grid of 4 different test cases, plots target, prediction, variance and error for each ''' mpl.rcParams['font.family'] = ['serif'] # default is sans-serif rc('text', usetex=False) fig = plt.figure(figsize=(15, 13), dpi=150) outer = gridspec.GridSpec(2, 2, wspace=0.45, hspace=0.2) # Outer grid for i in range(4): # Inner grid inner = gridspec.GridSpecFromSubplotSpec(4, 1, subplot_spec=outer[i], wspace=0, hspace=0.25) ax = [] for j in range(4): ax0 = plt.Subplot(fig, inner[j]) fig.add_subplot(ax0) ax.append(ax0) # Plot specific test case plotPred(fig, ax, t, xT, uPred[i], betas, uTarget[i]) file_dir = '.' # If directory does not exist create it if not os.path.exists(file_dir): os.makedirs(file_dir) file_name = file_dir+"/burger_BAR_pred" plt.savefig(file_name+".png", bbox_inches='tight') plt.savefig(file_name+".pdf", bbox_inches='tight') plt.show() def plotPred(fig, ax, t, xT, uPred, betas, uTarget): ''' Plots specific test case Args: fig: matplotlib figure ax (list): list of four subplot axis t (np.array): [n] array to time values for x axis xT (np.array): [m] array of spacial coordinates for y axis uPred (np.array): [n x m] model predictions uTarget (np.array): [n x m] target field ''' # Start with the target up top cmap = "inferno" c0 = ax[0].imshow(uTarget.T, interpolation='nearest', cmap=cmap, origin='lower', aspect='auto', extent=[t[0],t[-1],xT[0],xT[-1]]) c_max = np.max(uTarget.T) c_min = np.min(uTarget.T) c0.set_clim(vmin=c_min, vmax=c_max) # Plot the mean uPred_mean = np.mean(uPred, axis=0) c0 = ax[1].imshow(uPred_mean.T, interpolation='nearest', cmap=cmap, origin='lower', aspect='auto', extent=[t[0],t[-1],xT[0],xT[-1]]) c0.set_clim(vmin=c_min, vmax=c_max) p0 = ax[0].get_position().get_points().flatten() p1 = ax[1].get_position().get_points().flatten() ax_cbar = fig.add_axes([p1[2]+0.015, p1[1], 0.020, p0[3]-p1[1]]) ticks = np.linspace(0, 1, 5) tickLabels = np.linspace(c_min, c_max, 5) tickLabels = ["{:02.2f}".format(t0) for t0 in tickLabels] cbar = mpl.colorbar.ColorbarBase(ax_cbar, cmap=plt.get_cmap(cmap), orientation='vertical', ticks=ticks) cbar.set_ticklabels(tickLabels) # Variance betas = np.expand_dims(betas, axis=1).repeat(uPred.shape[1], axis=1) # Expand noise parameter betas = np.expand_dims(betas, axis=2).repeat(uPred.shape[2], axis=2) # Expand noise parameter uPred_var = np.mean(1./betas + uPred*uPred, axis=0) - uPred_mean*uPred_mean c0 = ax[2].imshow(uPred_var.T, interpolation='nearest', cmap=cmap, origin='lower', aspect='auto', extent=[t[0],t[-1],xT[0],xT[-1]]) c_max = np.max(uPred_var) c0.set_clim(vmin=0, vmax=c_max) p0 = ax[2].get_position().get_points().flatten() ax_cbar = fig.add_axes([p0[2]+0.015, p0[1], 0.020, p0[3]-p0[1]]) ticks = np.linspace(0, 1, 5) tickLabels = np.linspace(0, c_max, 5) tickLabels = ["{:02.2f}".format(t0) for t0 in tickLabels] cbar = mpl.colorbar.ColorbarBase(ax_cbar, cmap=plt.get_cmap(cmap), orientation='vertical', ticks=ticks) cbar.set_ticklabels(tickLabels) # Mean Error cmap = "viridis" c0 = ax[3].imshow(np.abs(uPred_mean.T - uTarget.T), interpolation='nearest', cmap=cmap, origin='lower', aspect='auto', extent=[t[0],t[-1],xT[0],xT[-1]]) p0 = ax[3].get_position().get_points().flatten() ax_cbar = fig.add_axes([p0[2]+0.015, p0[1], 0.020, p0[3]-p0[1]]) ticks = np.linspace(0, 1, 5) tickLabels = np.linspace(c0.norm.vmin, c0.norm.vmax, 5) tickLabels = ["{:.2e}".format(t0) for t0 in tickLabels] cbar = mpl.colorbar.ColorbarBase(ax_cbar, cmap=plt.get_cmap(cmap), orientation='vertical', ticks=ticks) cbar.set_ticklabels(tickLabels) ax[0].set_ylabel('x', fontsize=14) ax[1].set_ylabel('x', fontsize=14) ax[2].set_ylabel('x', fontsize=14) ax[3].set_ylabel('x', fontsize=14) ax[3].set_xlabel('t', fontsize=14) # Remove some tick labels to help declutter plot ax[0].set_xticklabels([]) ax[1].set_xticklabels([]) ax[2].set_xticklabels([]) for ax0 in ax: ax0.set_yticks([0,0.5,1]) if __name__ == '__main__': # Parse arguements args = Parser().parse(dirs=False) use_cuda = "cpu" if(torch.cuda.is_available()): use_cuda = "cuda" args.device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") print("Torch device:{}".format(args.device)) # Domain settings, matches solver settings x0 = 0 x1 = 1.0 args.dx = (x1 - x0)/args.nel # Create training loader burgerLoader = BurgerLoader(dt=args.dt) # Create training loader test_cases = np.arange(0,5,1).astype(int) testing_loader = burgerLoader.createTestingLoader('../solver/fenics_data_dt0.001_T2.0', test_cases, batch_size=5) # Create DenseED model denseED = DenseED(in_channels=args.nic, out_channels=args.noc, blocks=args.blocks, growth_rate=args.growth_rate, init_features=args.init_features, bn_size=args.bn_size, drop_rate=args.drop_rate, bottleneck=False, out_activation=None).to(args.device) # Bayesian neural network bayes_nn = BayesNN(args, denseED) # Stochastic weighted averages swag_nn = SwagNN(args, bayes_nn, full_cov=True, max_models=args.swag_max) # Load network swag_nn.loadModel(200, file_dir="./networks") with torch.no_grad(): uPred, betas, uTarget = testSample(args, swag_nn, testing_loader, tstep=400, n_samples=30) tTest = np.arange(0, 400*args.dt+1e-8, args.dt) xTest = np.linspace(x0, x1, args.nel+1) plotContourGrid(tTest, xTest, uPred.cpu().numpy(), betas.cpu().numpy(), uTarget.cpu().numpy())
[ "torch.cuda.is_available", "matplotlib.rc", "nn.denseEDcirc.DenseED", "matplotlib.gridspec.GridSpecFromSubplotSpec", "sys.path.append", "numpy.arange", "numpy.mean", "os.path.exists", "numpy.max", "matplotlib.pyplot.Subplot", "matplotlib.gridspec.GridSpec", "numpy.linspace", "numpy.min", "...
[((390, 411), 'sys.path.append', 'sys.path.append', (['""".."""'], {}), "('..')\n", (405, 411), False, 'import sys\n'), ((1596, 1648), 'torch.zeros', 'torch.zeros', (['mb_size', 'n_samples', '(tstep + 1)', 'args.nel'], {}), '(mb_size, n_samples, tstep + 1, args.nel)\n', (1607, 1648), False, 'import torch\n'), ((1659, 1681), 'torch.zeros', 'torch.zeros', (['n_samples'], {}), '(n_samples)\n', (1670, 1681), False, 'import torch\n'), ((2768, 2792), 'matplotlib.rc', 'rc', (['"""text"""'], {'usetex': '(False)'}), "('text', usetex=False)\n", (2770, 2792), False, 'from matplotlib import rc\n'), ((2804, 2841), 'matplotlib.pyplot.figure', 'plt.figure', ([], {'figsize': '(15, 13)', 'dpi': '(150)'}), '(figsize=(15, 13), dpi=150)\n', (2814, 2841), True, 'import matplotlib.pyplot as plt\n'), ((2854, 2902), 'matplotlib.gridspec.GridSpec', 'gridspec.GridSpec', (['(2)', '(2)'], {'wspace': '(0.45)', 'hspace': '(0.2)'}), '(2, 2, wspace=0.45, hspace=0.2)\n', (2871, 2902), True, 'import matplotlib.gridspec as gridspec\n'), ((3497, 3549), 'matplotlib.pyplot.savefig', 'plt.savefig', (["(file_name + '.png')"], {'bbox_inches': '"""tight"""'}), "(file_name + '.png', bbox_inches='tight')\n", (3508, 3549), True, 'import matplotlib.pyplot as plt\n'), ((3552, 3604), 'matplotlib.pyplot.savefig', 'plt.savefig', (["(file_name + '.pdf')"], {'bbox_inches': '"""tight"""'}), "(file_name + '.pdf', bbox_inches='tight')\n", (3563, 3604), True, 'import matplotlib.pyplot as plt\n'), ((3608, 3618), 'matplotlib.pyplot.show', 'plt.show', ([], {}), '()\n', (3616, 3618), True, 'import matplotlib.pyplot as plt\n'), ((4232, 4249), 'numpy.max', 'np.max', (['uTarget.T'], {}), '(uTarget.T)\n', (4238, 4249), True, 'import numpy as np\n'), ((4262, 4279), 'numpy.min', 'np.min', (['uTarget.T'], {}), '(uTarget.T)\n', (4268, 4279), True, 'import numpy as np\n'), ((4358, 4380), 'numpy.mean', 'np.mean', (['uPred'], {'axis': '(0)'}), '(uPred, axis=0)\n', (4365, 4380), True, 'import numpy as np\n'), ((4746, 4766), 'numpy.linspace', 'np.linspace', (['(0)', '(1)', '(5)'], {}), '(0, 1, 5)\n', (4757, 4766), True, 'import numpy as np\n'), ((4784, 4812), 'numpy.linspace', 'np.linspace', (['c_min', 'c_max', '(5)'], {}), '(c_min, c_max, 5)\n', (4795, 4812), True, 'import numpy as np\n'), ((5460, 5477), 'numpy.max', 'np.max', (['uPred_var'], {}), '(uPred_var)\n', (5466, 5477), True, 'import numpy as np\n'), ((5649, 5669), 'numpy.linspace', 'np.linspace', (['(0)', '(1)', '(5)'], {}), '(0, 1, 5)\n', (5660, 5669), True, 'import numpy as np\n'), ((5687, 5711), 'numpy.linspace', 'np.linspace', (['(0)', 'c_max', '(5)'], {}), '(0, c_max, 5)\n', (5698, 5711), True, 'import numpy as np\n'), ((6248, 6268), 'numpy.linspace', 'np.linspace', (['(0)', '(1)', '(5)'], {}), '(0, 1, 5)\n', (6259, 6268), True, 'import numpy as np\n'), ((6286, 6328), 'numpy.linspace', 'np.linspace', (['c0.norm.vmin', 'c0.norm.vmax', '(5)'], {}), '(c0.norm.vmin, c0.norm.vmax, 5)\n', (6297, 6328), True, 'import numpy as np\n'), ((7048, 7073), 'torch.cuda.is_available', 'torch.cuda.is_available', ([], {}), '()\n', (7071, 7073), False, 'import torch\n'), ((7390, 7414), 'utils.burgerLoader.BurgerLoader', 'BurgerLoader', ([], {'dt': 'args.dt'}), '(dt=args.dt)\n', (7402, 7414), False, 'from utils.burgerLoader import BurgerLoader\n'), ((8105, 8127), 'nn.bayesNN.BayesNN', 'BayesNN', (['args', 'denseED'], {}), '(args, denseED)\n', (8112, 8127), False, 'from nn.bayesNN import BayesNN\n'), ((8177, 8240), 'nn.swag.SwagNN', 'SwagNN', (['args', 'bayes_nn'], {'full_cov': '(True)', 'max_models': 'args.swag_max'}), '(args, bayes_nn, full_cov=True, max_models=args.swag_max)\n', (8183, 8240), False, 'from nn.swag import SwagNN\n'), ((8449, 8493), 'numpy.arange', 'np.arange', (['(0)', '(400 * args.dt + 1e-08)', 'args.dt'], {}), '(0, 400 * args.dt + 1e-08, args.dt)\n', (8458, 8493), True, 'import numpy as np\n'), ((8501, 8534), 'numpy.linspace', 'np.linspace', (['x0', 'x1', '(args.nel + 1)'], {}), '(x0, x1, args.nel + 1)\n', (8512, 8534), True, 'import numpy as np\n'), ((2976, 3064), 'matplotlib.gridspec.GridSpecFromSubplotSpec', 'gridspec.GridSpecFromSubplotSpec', (['(4)', '(1)'], {'subplot_spec': 'outer[i]', 'wspace': '(0)', 'hspace': '(0.25)'}), '(4, 1, subplot_spec=outer[i], wspace=0,\n hspace=0.25)\n', (3008, 3064), True, 'import matplotlib.gridspec as gridspec\n'), ((3393, 3417), 'os.path.exists', 'os.path.exists', (['file_dir'], {}), '(file_dir)\n', (3407, 3417), False, 'import os\n'), ((3427, 3448), 'os.makedirs', 'os.makedirs', (['file_dir'], {}), '(file_dir)\n', (3438, 3448), False, 'import os\n'), ((5247, 5291), 'numpy.mean', 'np.mean', (['(1.0 / betas + uPred * uPred)'], {'axis': '(0)'}), '(1.0 / betas + uPred * uPred, axis=0)\n', (5254, 5291), True, 'import numpy as np\n'), ((5979, 6011), 'numpy.abs', 'np.abs', (['(uPred_mean.T - uTarget.T)'], {}), '(uPred_mean.T - uTarget.T)\n', (5985, 6011), True, 'import numpy as np\n'), ((8320, 8335), 'torch.no_grad', 'torch.no_grad', ([], {}), '()\n', (8333, 8335), False, 'import torch\n'), ((3135, 3161), 'matplotlib.pyplot.Subplot', 'plt.Subplot', (['fig', 'inner[j]'], {}), '(fig, inner[j])\n', (3146, 3161), True, 'import matplotlib.pyplot as plt\n'), ((4926, 4944), 'matplotlib.pyplot.get_cmap', 'plt.get_cmap', (['cmap'], {}), '(cmap)\n', (4938, 4944), True, 'import matplotlib.pyplot as plt\n'), ((5047, 5076), 'numpy.expand_dims', 'np.expand_dims', (['betas'], {'axis': '(1)'}), '(betas, axis=1)\n', (5061, 5076), True, 'import numpy as np\n'), ((5145, 5174), 'numpy.expand_dims', 'np.expand_dims', (['betas'], {'axis': '(2)'}), '(betas, axis=2)\n', (5159, 5174), True, 'import numpy as np\n'), ((5825, 5843), 'matplotlib.pyplot.get_cmap', 'plt.get_cmap', (['cmap'], {}), '(cmap)\n', (5837, 5843), True, 'import matplotlib.pyplot as plt\n'), ((6440, 6458), 'matplotlib.pyplot.get_cmap', 'plt.get_cmap', (['cmap'], {}), '(cmap)\n', (6452, 6458), True, 'import matplotlib.pyplot as plt\n'), ((6993, 7001), 'args.Parser', 'Parser', ([], {}), '()\n', (6999, 7001), False, 'from args import Parser\n'), ((7145, 7170), 'torch.cuda.is_available', 'torch.cuda.is_available', ([], {}), '()\n', (7168, 7170), False, 'import torch\n'), ((7461, 7479), 'numpy.arange', 'np.arange', (['(0)', '(5)', '(1)'], {}), '(0, 5, 1)\n', (7470, 7479), True, 'import numpy as np\n'), ((7650, 7886), 'nn.denseEDcirc.DenseED', 'DenseED', ([], {'in_channels': 'args.nic', 'out_channels': 'args.noc', 'blocks': 'args.blocks', 'growth_rate': 'args.growth_rate', 'init_features': 'args.init_features', 'bn_size': 'args.bn_size', 'drop_rate': 'args.drop_rate', 'bottleneck': '(False)', 'out_activation': 'None'}), '(in_channels=args.nic, out_channels=args.noc, blocks=args.blocks,\n growth_rate=args.growth_rate, init_features=args.init_features, bn_size\n =args.bn_size, drop_rate=args.drop_rate, bottleneck=False,\n out_activation=None)\n', (7657, 7886), False, 'from nn.denseEDcirc import DenseED\n'), ((2386, 2419), 'torch.cat', 'torch.cat', (['[input, input0]'], {'dim': '(1)'}), '([input, input0], dim=1)\n', (2395, 2419), False, 'import torch\n')]
#!/usr/bin/env python3 # -*- coding: utf-8 -*- """ Created on Thu Dec 16 08:42:47 2021 @author: <NAME> @name: roi-aselect.py """ import argparse import csv import cv2 import os import sys import numpy as np from pathlib import Path s_default=20 th_default=5.0 threshold=-1 th_under=False th_over=False img_sel=[] cross_col=(255,0,0) decimals=4 roi=[-1,-1,-1,-1] isROI=False def roix(value): return round(value/img_x1,decimals) def roiy(value): return round(value/img_y1,decimals) def getMarkerCoordinates(w,h,x,y,s): top=int(y-s/2) bottom=top+s left=int(x-s/2) right=left+s if top<0: top=0 if bottom>h-1: bottom=h-1 if left<0: left=0 if right>w-1: right=w-1 return(left,right,top,bottom) def getThresholdCoordinate(direction,mean,x,y,s): global img delta_x=0 delta_y=0 sadjust=int(s/2) if direction==0: # Up sx=x-sadjust ex=x+sadjust sy=y-sadjust ey=y-sadjust+1 delta_y=-1 elif direction==1: # Right sx=x+sadjust-1 ex=x+sadjust sy=y-sadjust ey=y+sadjust delta_x=1 elif direction==2: # Down sx=x-sadjust ex=x+sadjust sy=y+sadjust-1 ey=y+sadjust delta_y=1 elif direction==3: # Left sx=x-sadjust ex=x-sadjust+1 sy=y-sadjust ey=y+sadjust delta_x=-1 # Fix values outside ranges if sx<0: sx=0 elif sx>w-1: sx=w-1 if ex<0: ex=0 elif ex>w-1: ex=w-1 if sy<0: sy=0 elif sy>h-1: sy=h-1 if ey<0: ey=0 elif ey>h-1: ey=h-1 # Fix slicing values if ex<=sx: ex=sx+1 if ey<=sy: ey=sy+1 while True: line=img[sy:ey,sx:ex] mean_line=line.mean(axis=(0,1)).mean() if th_under: if mean_line<mean-threshold: break if th_over: if mean_line>mean+threshold: break sx+=delta_x ex+=delta_x sy+=delta_y ey+=delta_y if (sx<0 or sx>w-1) or (ex<0 or ex>w-1): break if (sy<0 or sy>h-1) or (ey<0 or ey>h-1): break if sy<0: sy=0 elif sy>h-1: sy=h-1 if sx<0: sx=0 elif sx>w-1: sx=w-1 if direction==0 or direction==2: return sy else: return sx def click_event(event, x, y, flags, param): global img_clone,img_clone2 global roi,isROI if x<0 or x>=w: return if y<0 or y>=h: return if event == cv2.EVENT_LBUTTONDOWN: (selx1,selx2,sely1,sely2)=getMarkerCoordinates(w,h,x,y,s) if sely1==sely2 or selx1==selx2: return img_sel=img[sely1:sely2,selx1:selx2] mean_bgr=img_sel.mean(axis=(0,1)) neg_bgr=tuple(255-mean_bgr) m=mean_bgr.mean() ysel1=getThresholdCoordinate(0,m,x,y,s) ysel2=getThresholdCoordinate(2,m,x,y,s) xsel1=getThresholdCoordinate(3,m,x,y,s) xsel2=getThresholdCoordinate(1,m,x,y,s) print("("+str(x)+","+str(y)+")") # Draw selection area blk=np.zeros(img.shape,np.uint8) cv2.rectangle(blk, (xsel1,ysel1), (xsel2,ysel2), neg_bgr,-1) img_clone=cv2.addWeighted(img,1,blk,0.25,1) # Draw selection cross sadjust=int(s/2) blk=np.zeros(img.shape,np.uint8) cv2.rectangle(blk,(xsel1,y-sadjust),(xsel2,y+sadjust),cross_col,-1) cv2.rectangle(blk,(x-sadjust,ysel1),(x+sadjust,ysel2),cross_col,-1) img_clone2=cv2.addWeighted(img_clone,1,blk,0.25,1) cv2.imshow('Select ROI',img_clone2) roi=[xsel1,ysel1,xsel2,ysel2] isROI=True elif event == cv2.EVENT_RBUTTONDOWN: # Clear all markers img_clone=img.copy() cv2.imshow('Select ROI', img_clone) roi=[-1,-1,-1,-1] isROI=False parser=argparse.ArgumentParser() parser.add_argument("-f","--file",type=Path, help="specify the image name",required=True) parser.add_argument('-t',type=float,help="threshold (default: "+str(th_default)+") as float ",required=False) parser.add_argument('-s',type=float,help="selection size s x s (default: "+str(s_default)+") as integer ",required=False) parser.add_argument("-u", action="store_true", help="enable under threshold",required=False) parser.add_argument("-o", action="store_true", help="enable over threshold",required=False) args = parser.parse_args() if args.file: stem=args.file.stem fname=str(args.file.name) filename=str(args.file) if not os.path.isfile(filename): print("File "+str(filename)+" not found!") sys.exit(0) if args.t != None: threshold=float(args.t) if threshold<0 or threshold>255: threshold=th_default else: threshold=th_default if args.s != None: s=int(args.s) else: s=s_default if args.u==True: th_under=True if args.o==True: th_over=True if not True in [th_under,th_over]: th_under=True th_over=True stem+="-" print("Auto ROI selection") print("(C) <NAME> 2021") print("") print("Image file: "+filename) print("") print("Current directory:") curdir=os.getcwd() path=Path(curdir) print(curdir) print("") print("1. Select ROI by pressing the mouse left button") print("2. Remove ROI by pressing the mouse right button") print("3. Press any key to accept the selection") img = cv2.imread(filename) if len(img.shape)==2: h,w=img.shape elif len(img.shape)==3: h,w,ch=img.shape img_clone=img.copy() img_clone2=img.copy() cv2.namedWindow("Select ROI", cv2.WINDOW_NORMAL) cv2.imshow("Select ROI",img_clone) cv2.setMouseCallback("Select ROI", click_event) cv2.waitKey(0) print("") if isROI: print("Saving roi.ini") # Build and save a ROI file img_x0=0 img_y0=0 img_x1=w img_y1=h crop_x0=roi[0] crop_x1=roi[2] crop_y0=roi[1] crop_y1=roi[3] roi_x0=roix(crop_x0) roi_w=roix(crop_x1-crop_x0+1) roi_y0=roiy(crop_y0) roi_h=roiy(crop_y1-crop_y0+1) roilist=[] roilist.append(["scale","coordinate name","value"]) roilist.append(["original","img_x0",img_x0]) roilist.append(["original","img_x1",img_x1]) roilist.append(["original","img_y0",img_y0]) roilist.append(["original","img_y1",img_y1]) roilist.append(["original","crop_x0",crop_x0]) roilist.append(["original","crop_x1",crop_x1]) roilist.append(["original","crop_y0",crop_y0]) roilist.append(["original","crop_y1",crop_y1]) roilist.append(["normalized","roi_x0",roi_x0]) roilist.append(["normalized","roi_y0",roi_y0]) roilist.append(["normalized","roi_w",roi_w]) roilist.append(["normalized","roi_h",roi_h]) with open("roi.ini","w",newline="") as csvfile: csvwriter=csv.writer(csvfile,delimiter=";") for s in roilist: csvwriter.writerow(s) # Save ROI images print("Saving ROI image files") cv2.imwrite("roi-patch.jpg",img_clone) cv2.imwrite("roi-selection.jpg",img_clone2) cv2.imwrite("roi.jpg",img[crop_y0:crop_y1,crop_x0:crop_x1]) else: print("ROI not selected.") # All these waitkeys are a hack to get the OpenCV window to close cv2.waitKey(1) cv2.destroyAllWindows() for i in range (1,5): cv2.waitKey(1)
[ "cv2.setMouseCallback", "cv2.imwrite", "cv2.rectangle", "argparse.ArgumentParser", "pathlib.Path", "csv.writer", "os.getcwd", "cv2.imshow", "os.path.isfile", "cv2.waitKey", "numpy.zeros", "cv2.destroyAllWindows", "cv2.addWeighted", "sys.exit", "cv2.imread", "cv2.namedWindow" ]
[((4034, 4059), 'argparse.ArgumentParser', 'argparse.ArgumentParser', ([], {}), '()\n', (4057, 4059), False, 'import argparse\n'), ((5288, 5299), 'os.getcwd', 'os.getcwd', ([], {}), '()\n', (5297, 5299), False, 'import os\n'), ((5305, 5317), 'pathlib.Path', 'Path', (['curdir'], {}), '(curdir)\n', (5309, 5317), False, 'from pathlib import Path\n'), ((5515, 5535), 'cv2.imread', 'cv2.imread', (['filename'], {}), '(filename)\n', (5525, 5535), False, 'import cv2\n'), ((5665, 5713), 'cv2.namedWindow', 'cv2.namedWindow', (['"""Select ROI"""', 'cv2.WINDOW_NORMAL'], {}), "('Select ROI', cv2.WINDOW_NORMAL)\n", (5680, 5713), False, 'import cv2\n'), ((5714, 5749), 'cv2.imshow', 'cv2.imshow', (['"""Select ROI"""', 'img_clone'], {}), "('Select ROI', img_clone)\n", (5724, 5749), False, 'import cv2\n'), ((5749, 5796), 'cv2.setMouseCallback', 'cv2.setMouseCallback', (['"""Select ROI"""', 'click_event'], {}), "('Select ROI', click_event)\n", (5769, 5796), False, 'import cv2\n'), ((5797, 5811), 'cv2.waitKey', 'cv2.waitKey', (['(0)'], {}), '(0)\n', (5808, 5811), False, 'import cv2\n'), ((7325, 7339), 'cv2.waitKey', 'cv2.waitKey', (['(1)'], {}), '(1)\n', (7336, 7339), False, 'import cv2\n'), ((7340, 7363), 'cv2.destroyAllWindows', 'cv2.destroyAllWindows', ([], {}), '()\n', (7361, 7363), False, 'import cv2\n'), ((4698, 4722), 'os.path.isfile', 'os.path.isfile', (['filename'], {}), '(filename)\n', (4712, 4722), False, 'import os\n'), ((4775, 4786), 'sys.exit', 'sys.exit', (['(0)'], {}), '(0)\n', (4783, 4786), False, 'import sys\n'), ((7057, 7096), 'cv2.imwrite', 'cv2.imwrite', (['"""roi-patch.jpg"""', 'img_clone'], {}), "('roi-patch.jpg', img_clone)\n", (7068, 7096), False, 'import cv2\n'), ((7100, 7144), 'cv2.imwrite', 'cv2.imwrite', (['"""roi-selection.jpg"""', 'img_clone2'], {}), "('roi-selection.jpg', img_clone2)\n", (7111, 7144), False, 'import cv2\n'), ((7148, 7209), 'cv2.imwrite', 'cv2.imwrite', (['"""roi.jpg"""', 'img[crop_y0:crop_y1, crop_x0:crop_x1]'], {}), "('roi.jpg', img[crop_y0:crop_y1, crop_x0:crop_x1])\n", (7159, 7209), False, 'import cv2\n'), ((7390, 7404), 'cv2.waitKey', 'cv2.waitKey', (['(1)'], {}), '(1)\n', (7401, 7404), False, 'import cv2\n'), ((3260, 3289), 'numpy.zeros', 'np.zeros', (['img.shape', 'np.uint8'], {}), '(img.shape, np.uint8)\n', (3268, 3289), True, 'import numpy as np\n'), ((3297, 3360), 'cv2.rectangle', 'cv2.rectangle', (['blk', '(xsel1, ysel1)', '(xsel2, ysel2)', 'neg_bgr', '(-1)'], {}), '(blk, (xsel1, ysel1), (xsel2, ysel2), neg_bgr, -1)\n', (3310, 3360), False, 'import cv2\n'), ((3376, 3413), 'cv2.addWeighted', 'cv2.addWeighted', (['img', '(1)', 'blk', '(0.25)', '(1)'], {}), '(img, 1, blk, 0.25, 1)\n', (3391, 3413), False, 'import cv2\n'), ((3487, 3516), 'numpy.zeros', 'np.zeros', (['img.shape', 'np.uint8'], {}), '(img.shape, np.uint8)\n', (3495, 3516), True, 'import numpy as np\n'), ((3524, 3601), 'cv2.rectangle', 'cv2.rectangle', (['blk', '(xsel1, y - sadjust)', '(xsel2, y + sadjust)', 'cross_col', '(-1)'], {}), '(blk, (xsel1, y - sadjust), (xsel2, y + sadjust), cross_col, -1)\n', (3537, 3601), False, 'import cv2\n'), ((3600, 3677), 'cv2.rectangle', 'cv2.rectangle', (['blk', '(x - sadjust, ysel1)', '(x + sadjust, ysel2)', 'cross_col', '(-1)'], {}), '(blk, (x - sadjust, ysel1), (x + sadjust, ysel2), cross_col, -1)\n', (3613, 3677), False, 'import cv2\n'), ((3687, 3730), 'cv2.addWeighted', 'cv2.addWeighted', (['img_clone', '(1)', 'blk', '(0.25)', '(1)'], {}), '(img_clone, 1, blk, 0.25, 1)\n', (3702, 3730), False, 'import cv2\n'), ((3744, 3780), 'cv2.imshow', 'cv2.imshow', (['"""Select ROI"""', 'img_clone2'], {}), "('Select ROI', img_clone2)\n", (3754, 3780), False, 'import cv2\n'), ((6896, 6930), 'csv.writer', 'csv.writer', (['csvfile'], {'delimiter': '""";"""'}), "(csvfile, delimiter=';')\n", (6906, 6930), False, 'import csv\n'), ((3944, 3979), 'cv2.imshow', 'cv2.imshow', (['"""Select ROI"""', 'img_clone'], {}), "('Select ROI', img_clone)\n", (3954, 3979), False, 'import cv2\n')]
import os import json import re import numpy as np from shutil import copyfile from keras.optimizers import SGD import keras.backend as K from AlphaGo.ai import ProbabilisticPolicyPlayer import AlphaGo.go as go from AlphaGo.models.policy import CNNPolicy from AlphaGo.util import flatten_idx def _make_training_pair(st, mv, preprocessor): # Convert move to one-hot st_tensor = preprocessor.state_to_tensor(st) mv_tensor = np.zeros((1, st.size * st.size)) mv_tensor[(0, flatten_idx(mv, st.size))] = 1 return (st_tensor, mv_tensor) def run_n_games(optimizer, learner, opponent, num_games, mock_states=[]): '''Run num_games games to completion, keeping track of each position and move of the learner. (Note: learning cannot happen until all games have completed) ''' board_size = learner.policy.model.input_shape[-1] states = [go.GameState(size=board_size) for _ in range(num_games)] learner_net = learner.policy.model # Allowing injection of a mock state object for testing purposes if mock_states: states = mock_states # Create one list of features (aka state tensors) and one of moves for each game being played. state_tensors = [[] for _ in range(num_games)] move_tensors = [[] for _ in range(num_games)] # List of booleans indicating whether the 'learner' player won. learner_won = [None] * num_games # Start all odd games with moves by 'opponent'. Even games will have 'learner' black. learner_color = [go.BLACK if i % 2 == 0 else go.WHITE for i in range(num_games)] odd_states = states[1::2] moves = opponent.get_moves(odd_states) for st, mv in zip(odd_states, moves): st.do_move(mv) current = learner other = opponent idxs_to_unfinished_states = {i: states[i] for i in range(num_games)} while len(idxs_to_unfinished_states) > 0: # Get next moves by current player for all unfinished states. moves = current.get_moves(idxs_to_unfinished_states.values()) just_finished = [] # Do each move to each state in order. for (idx, state), mv in zip(idxs_to_unfinished_states.iteritems(), moves): # Order is important here. We must get the training pair on the unmodified state before # updating it with do_move. is_learnable = current is learner and mv is not go.PASS_MOVE if is_learnable: (st_tensor, mv_tensor) = _make_training_pair(state, mv, learner.policy.preprocessor) state_tensors[idx].append(st_tensor) move_tensors[idx].append(mv_tensor) state.do_move(mv) if state.is_end_of_game: learner_won[idx] = state.get_winner() == learner_color[idx] just_finished.append(idx) # Remove games that have finished from dict. for idx in just_finished: del idxs_to_unfinished_states[idx] # Swap 'current' and 'other' for next turn. current, other = other, current # Train on each game's results, setting the learning rate negative to 'unlearn' positions from # games where the learner lost. for (st_tensor, mv_tensor, won) in zip(state_tensors, move_tensors, learner_won): optimizer.lr = K.abs(optimizer.lr) * (+1 if won else -1) learner_net.train_on_batch(np.concatenate(st_tensor, axis=0), np.concatenate(mv_tensor, axis=0)) # Return the win ratio. wins = sum(state.get_winner() == pc for (state, pc) in zip(states, learner_color)) return float(wins) / num_games def log_loss(y_true, y_pred): '''Keras 'loss' function for the REINFORCE algorithm, where y_true is the action that was taken, and updates with the negative gradient will make that action more likely. We use the negative gradient because keras expects training data to minimize a loss function. ''' return -y_true * K.log(K.clip(y_pred, K.epsilon(), 1.0 - K.epsilon())) def run_training(cmd_line_args=None): import argparse parser = argparse.ArgumentParser(description='Perform reinforcement learning to improve given policy network. Second phase of pipeline.') # noqa: E501 parser.add_argument("model_json", help="Path to policy model JSON.") parser.add_argument("initial_weights", help="Path to HDF5 file with inital weights (i.e. result of supervised training).") # noqa: E501 parser.add_argument("out_directory", help="Path to folder where the model params and metadata will be saved after each epoch.") # noqa: E501 parser.add_argument("--learning-rate", help="Keras learning rate (Default: 0.001)", type=float, default=0.001) # noqa: E501 parser.add_argument("--policy-temp", help="Distribution temperature of players using policies (Default: 0.67)", type=float, default=0.67) # noqa: E501 parser.add_argument("--save-every", help="Save policy as a new opponent every n batches (Default: 500)", type=int, default=500) # noqa: E501 parser.add_argument("--record-every", help="Save learner's weights every n batches (Default: 1)", type=int, default=1) # noqa: E501 parser.add_argument("--game-batch", help="Number of games per mini-batch (Default: 20)", type=int, default=20) # noqa: E501 parser.add_argument("--move-limit", help="Maximum number of moves per game", type=int, default=500) # noqa: E501 parser.add_argument("--iterations", help="Number of training batches/iterations (Default: 10000)", type=int, default=10000) # noqa: E501 parser.add_argument("--resume", help="Load latest weights in out_directory and resume", default=False, action="store_true") # noqa: E501 parser.add_argument("--verbose", "-v", help="Turn on verbose mode", default=False, action="store_true") # noqa: E501 # Baseline function (TODO) default lambda state: 0 (receives either file # paths to JSON and weights or None, in which case it uses default baseline 0) if cmd_line_args is None: args = parser.parse_args() else: args = parser.parse_args(cmd_line_args) ZEROTH_FILE = "weights.00000.hdf5" if args.resume: if not os.path.exists(os.path.join(args.out_directory, "metadata.json")): raise ValueError("Cannot resume without existing output directory") if not os.path.exists(args.out_directory): if args.verbose: print("creating output directory {}".format(args.out_directory)) os.makedirs(args.out_directory) if not args.resume: # make a copy of weights file, "weights.00000.hdf5" in the output directory copyfile(args.initial_weights, os.path.join(args.out_directory, ZEROTH_FILE)) if args.verbose: print("copied {} to {}".format(args.initial_weights, os.path.join(args.out_directory, ZEROTH_FILE))) player_weights = ZEROTH_FILE iter_start = 1 else: # if resuming, we expect initial_weights to be just a # "weights.#####.hdf5" file, not a full path if not re.match(r"weights\.\d{5}\.hdf5", args.initial_weights): raise ValueError("Expected to resume from weights file with name 'weights.#####.hdf5'") args.initial_weights = os.path.join(args.out_directory, os.path.basename(args.initial_weights)) if not os.path.exists(args.initial_weights): raise ValueError("Cannot resume; weights {} do not exist".format(args.initial_weights)) elif args.verbose: print("Resuming with weights {}".format(args.initial_weights)) player_weights = os.path.basename(args.initial_weights) iter_start = 1 + int(player_weights[8:13]) # Set initial conditions policy = CNNPolicy.load_model(args.model_json) policy.model.load_weights(args.initial_weights) player = ProbabilisticPolicyPlayer(policy, temperature=args.policy_temp, move_limit=args.move_limit) # different opponents come from simply changing the weights of 'opponent.policy.model'. That # is, only 'opp_policy' needs to be changed, and 'opponent' will change. opp_policy = CNNPolicy.load_model(args.model_json) opponent = ProbabilisticPolicyPlayer(opp_policy, temperature=args.policy_temp, move_limit=args.move_limit) if args.verbose: print("created player and opponent with temperature {}".format(args.policy_temp)) if not args.resume: metadata = { "model_file": args.model_json, "init_weights": args.initial_weights, "learning_rate": args.learning_rate, "temperature": args.policy_temp, "game_batch": args.game_batch, "opponents": [ZEROTH_FILE], # which weights from which to sample an opponent each batch "win_ratio": {} # map from player to tuple of (opponent, win ratio) Useful for # validating in lieu of 'accuracy/loss' } else: with open(os.path.join(args.out_directory, "metadata.json"), "r") as f: metadata = json.load(f) # Append args of current run to history of full command args. metadata["cmd_line_args"] = metadata.get("cmd_line_args", []) metadata["cmd_line_args"].append(vars(args)) def save_metadata(): with open(os.path.join(args.out_directory, "metadata.json"), "w") as f: json.dump(metadata, f, sort_keys=True, indent=2) optimizer = SGD(lr=args.learning_rate) player.policy.model.compile(loss=log_loss, optimizer=optimizer) for i_iter in range(iter_start, args.iterations + 1): # Note that player_weights will only be saved as a file every args.record_every iterations. # Regardless, player_weights enters into the metadata to keep track of the win ratio over # time. player_weights = "weights.%05d.hdf5" % i_iter # Randomly choose opponent from pool (possibly self), and playing # game_batch games against them. opp_weights = np.random.choice(metadata["opponents"]) opp_path = os.path.join(args.out_directory, opp_weights) # Load new weights into opponent's network, but keep the same opponent object. opponent.policy.model.load_weights(opp_path) if args.verbose: print("Batch {}\tsampled opponent is {}".format(i_iter, opp_weights)) # Run games (and learn from results). Keep track of the win ratio vs each opponent over # time. win_ratio = run_n_games(optimizer, player, opponent, args.game_batch) metadata["win_ratio"][player_weights] = (opp_weights, win_ratio) # Save intermediate models. if i_iter % args.record_every == 0: player.policy.model.save_weights(os.path.join(args.out_directory, player_weights)) # Add player to batch of oppenents once in a while. if i_iter % args.save_every == 0: metadata["opponents"].append(player_weights) save_metadata() if __name__ == '__main__': run_training()
[ "os.path.exists", "AlphaGo.models.policy.CNNPolicy.load_model", "argparse.ArgumentParser", "os.makedirs", "numpy.random.choice", "os.path.join", "re.match", "keras.backend.epsilon", "AlphaGo.go.GameState", "numpy.zeros", "keras.optimizers.SGD", "AlphaGo.util.flatten_idx", "os.path.basename",...
[((436, 468), 'numpy.zeros', 'np.zeros', (['(1, st.size * st.size)'], {}), '((1, st.size * st.size))\n', (444, 468), True, 'import numpy as np\n'), ((4070, 4208), 'argparse.ArgumentParser', 'argparse.ArgumentParser', ([], {'description': '"""Perform reinforcement learning to improve given policy network. Second phase of pipeline."""'}), "(description=\n 'Perform reinforcement learning to improve given policy network. Second phase of pipeline.'\n )\n", (4093, 4208), False, 'import argparse\n'), ((7785, 7822), 'AlphaGo.models.policy.CNNPolicy.load_model', 'CNNPolicy.load_model', (['args.model_json'], {}), '(args.model_json)\n', (7805, 7822), False, 'from AlphaGo.models.policy import CNNPolicy\n'), ((7888, 7984), 'AlphaGo.ai.ProbabilisticPolicyPlayer', 'ProbabilisticPolicyPlayer', (['policy'], {'temperature': 'args.policy_temp', 'move_limit': 'args.move_limit'}), '(policy, temperature=args.policy_temp, move_limit=\n args.move_limit)\n', (7913, 7984), False, 'from AlphaGo.ai import ProbabilisticPolicyPlayer\n'), ((8211, 8248), 'AlphaGo.models.policy.CNNPolicy.load_model', 'CNNPolicy.load_model', (['args.model_json'], {}), '(args.model_json)\n', (8231, 8248), False, 'from AlphaGo.models.policy import CNNPolicy\n'), ((8264, 8363), 'AlphaGo.ai.ProbabilisticPolicyPlayer', 'ProbabilisticPolicyPlayer', (['opp_policy'], {'temperature': 'args.policy_temp', 'move_limit': 'args.move_limit'}), '(opp_policy, temperature=args.policy_temp,\n move_limit=args.move_limit)\n', (8289, 8363), False, 'from AlphaGo.ai import ProbabilisticPolicyPlayer\n'), ((9553, 9579), 'keras.optimizers.SGD', 'SGD', ([], {'lr': 'args.learning_rate'}), '(lr=args.learning_rate)\n', (9556, 9579), False, 'from keras.optimizers import SGD\n'), ((870, 899), 'AlphaGo.go.GameState', 'go.GameState', ([], {'size': 'board_size'}), '(size=board_size)\n', (882, 899), True, 'import AlphaGo.go as go\n'), ((6313, 6347), 'os.path.exists', 'os.path.exists', (['args.out_directory'], {}), '(args.out_directory)\n', (6327, 6347), False, 'import os\n'), ((6459, 6490), 'os.makedirs', 'os.makedirs', (['args.out_directory'], {}), '(args.out_directory)\n', (6470, 6490), False, 'import os\n'), ((7652, 7690), 'os.path.basename', 'os.path.basename', (['args.initial_weights'], {}), '(args.initial_weights)\n', (7668, 7690), False, 'import os\n'), ((10112, 10151), 'numpy.random.choice', 'np.random.choice', (["metadata['opponents']"], {}), "(metadata['opponents'])\n", (10128, 10151), True, 'import numpy as np\n'), ((10171, 10216), 'os.path.join', 'os.path.join', (['args.out_directory', 'opp_weights'], {}), '(args.out_directory, opp_weights)\n', (10183, 10216), False, 'import os\n'), ((487, 511), 'AlphaGo.util.flatten_idx', 'flatten_idx', (['mv', 'st.size'], {}), '(mv, st.size)\n', (498, 511), False, 'from AlphaGo.util import flatten_idx\n'), ((3272, 3291), 'keras.backend.abs', 'K.abs', (['optimizer.lr'], {}), '(optimizer.lr)\n', (3277, 3291), True, 'import keras.backend as K\n'), ((3349, 3382), 'numpy.concatenate', 'np.concatenate', (['st_tensor'], {'axis': '(0)'}), '(st_tensor, axis=0)\n', (3363, 3382), True, 'import numpy as np\n'), ((3419, 3452), 'numpy.concatenate', 'np.concatenate', (['mv_tensor'], {'axis': '(0)'}), '(mv_tensor, axis=0)\n', (3433, 3452), True, 'import numpy as np\n'), ((6639, 6684), 'os.path.join', 'os.path.join', (['args.out_directory', 'ZEROTH_FILE'], {}), '(args.out_directory, ZEROTH_FILE)\n', (6651, 6684), False, 'import os\n'), ((7067, 7124), 're.match', 're.match', (['"""weights\\\\.\\\\d{5}\\\\.hdf5"""', 'args.initial_weights'], {}), "('weights\\\\.\\\\d{5}\\\\.hdf5', args.initial_weights)\n", (7075, 7124), False, 'import re\n'), ((7332, 7370), 'os.path.basename', 'os.path.basename', (['args.initial_weights'], {}), '(args.initial_weights)\n', (7348, 7370), False, 'import os\n'), ((7387, 7423), 'os.path.exists', 'os.path.exists', (['args.initial_weights'], {}), '(args.initial_weights)\n', (7401, 7423), False, 'import os\n'), ((9174, 9186), 'json.load', 'json.load', (['f'], {}), '(f)\n', (9183, 9186), False, 'import json\n'), ((9487, 9535), 'json.dump', 'json.dump', (['metadata', 'f'], {'sort_keys': '(True)', 'indent': '(2)'}), '(metadata, f, sort_keys=True, indent=2)\n', (9496, 9535), False, 'import json\n'), ((3964, 3975), 'keras.backend.epsilon', 'K.epsilon', ([], {}), '()\n', (3973, 3975), True, 'import keras.backend as K\n'), ((6169, 6218), 'os.path.join', 'os.path.join', (['args.out_directory', '"""metadata.json"""'], {}), "(args.out_directory, 'metadata.json')\n", (6181, 6218), False, 'import os\n'), ((9089, 9138), 'os.path.join', 'os.path.join', (['args.out_directory', '"""metadata.json"""'], {}), "(args.out_directory, 'metadata.json')\n", (9101, 9138), False, 'import os\n'), ((9413, 9462), 'os.path.join', 'os.path.join', (['args.out_directory', '"""metadata.json"""'], {}), "(args.out_directory, 'metadata.json')\n", (9425, 9462), False, 'import os\n'), ((10855, 10903), 'os.path.join', 'os.path.join', (['args.out_directory', 'player_weights'], {}), '(args.out_directory, player_weights)\n', (10867, 10903), False, 'import os\n'), ((3983, 3994), 'keras.backend.epsilon', 'K.epsilon', ([], {}), '()\n', (3992, 3994), True, 'import keras.backend as K\n'), ((6819, 6864), 'os.path.join', 'os.path.join', (['args.out_directory', 'ZEROTH_FILE'], {}), '(args.out_directory, ZEROTH_FILE)\n', (6831, 6864), False, 'import os\n')]
import matplotlib.pyplot as plt import seaborn as sns import numpy as np import pandas as pd import itertools from matplotlib import cm def plot_color_table(df, axis=1, title=None, rev_index=None, color='RdYlGn', prec=1, ): """ Creates color coded comparison table from dataframe values (green high, red low) Parameters ---------- df: pd.DataFrame DataFrame of values axis: int, default=1 axis to normalize data upon title: str title for table rev_index: list(str) list of column names to reverse color coding color: str, default='RdYlGn' color palette for table prec: int, default=1 decimals places to show Returns ------- plot of color coded table """ labels = df.values cdf = df.copy() if axis == 1: cdf -= cdf.mean(axis=0) cdf /= cdf.std(axis=0) else: cdf = cdf.transpose() cdf -= cdf.mean(axis=0) cdf /= cdf.std(axis=0) cdf = cdf.transpose() if rev_index: for i in rev_index: cdf.iloc[:, i] = 1 - cdf.iloc[:, i] plt.figure() if title: plt.title(title) sns.heatmap(cdf, cmap='RdYlGn', linewidths=0.5, annot=labels, fmt=f'0.{prec}f', cbar=False) plt.xticks(rotation=0) plt.yticks(rotation=0) def plot_explained_variance(ev, type='bar', toff=0.02, boff=0.08, yoff=None, **kwargs, ): """ Plot explained variance of PCA Parameters ---------- ev: nd.array explained variance values type: str, default='bar' - 'bar': bar plots for cumulative variance - 'line': line plot for cumulative variance toff: float top x offset for cumulative variance values on plot boff: float bottom x offset for individual variance values on plot yoff: float tottom y offset for individual variance values on plot **kwargs: kwargs for plt.plot() if type==line for cumulative variance Returns ------- plot of explained variance """ ax = pd.Series(100*ev).plot(kind='bar', figsize=(10,5), color='steelblue', fontsize=13) ax.set_ylabel('Explained Variance') ax.set_yticks([0, 20, 40, 60, 80, 100]) ax.set_xlabel('Principal Component') # Create a list to collect the plt.patches data. totals = [] # Find the values and append to list. for i in ax.patches: totals.append(i.get_height()) # Set individual bar lables using patch list. total = sum(totals) yoff_d = {'bar': 0, 'line': 0.2} yoff_d[type] = yoff_d[type] if yoff is None else yoff # Set individual bar lables using patch list. for j, i in enumerate(ax.patches): # Get_x pulls left or right; get_height pushes up or down. if j > 0: ax.text(i.get_x()+boff, i.get_height()+1, str(round(100*ev[j], 1))+'%', fontsize=11, color='dimgrey') ax.text(i.get_x()+toff, 100*np.cumsum(ev)[j]+1+yoff_d[type], str(round(100*np.cumsum(ev)[j], 1))+'%', fontsize=12, color='dimgrey') xlab = [f'PC-{i+1}' for i, _ in enumerate(ax.patches)] if type == 'bar': ax2 = pd.Series(np.cumsum(100*ev)).plot(kind='bar', figsize=(10,5), color='steelblue', fontsize=8, alpha=0.25) for i in ax2.patches: totals.append(i.get_height()) ax2.set_xticklabels(xlab, rotation='horizontal') ax2.xaxis.grid(False) elif type == 'line': plt.plot(np.cumsum(100*ev), ':o', ms=8, c='steelblue', alpha=0.75, **kwargs) ax.xaxis.grid(False) ax.set_xticklabels(xlab, rotation='horizontal') def correlation(df, sort_col=None, plot=False): """ Find correlation of DataFrame, plot results and/or return sorted correlations for a single column of the DataFrame. Parameters ---------- df: pd.DataFrame DataFrame of input values sort_col: str, default=None column to sort correlations on. If provided, function returns DataFrame of results plot: bool, default=False if True plot correlation matrix Returns ------- ret_col: pd.Series results for one column if sort_col is given plot of correlation matrix if plot=True """ corr = df.corr().abs() if plot: # Plot results fig, ax = plt.subplots(figsize=(12, 10)) ax.matshow(corr) cmap = cm.get_cmap('coolwarm', 300) cax = ax.imshow(corr, interpolation="nearest", cmap=cmap) plt.xticks(range(len(corr.columns)), corr.columns) plt.yticks(range(len(corr.columns)), corr.columns) for tick in ax.get_xticklabels(): tick.set_rotation(45) cax.set_clim([0, 1]) fig.colorbar(cax, ticks=[0, .25, .5, .75, 1]) if ret_col: return corr.sort_values([col], ascending=False)[col][1:] def plot_ROC_curve(roc, fpr, tpr, color_palette='muted', colors=None, figsize=(10, 10), title='ROC Curves for Studied Models', ): """ Plot ROC curve with AUC for multliple models. Parameters ---------- roc: dict dictionary of ROC AUC values with model names as keys fpr: dict dictionary of false positive rate values with model names as keys roc: dict dictionary of true positive rate values with model names as keys color_palette: str, defualt='muted' name of seaborn color palette to use colors: list(str), default=None if given, use provided colors instead of a color palette figsize: tuple(int), default=(10,10) figure size title: str figure title Returns ------- plot of ROC curve """ plt.figure(figsize=figsize) sns.set_palette(color_palette, len(roc)) for m in roc.keys(): plt.plot(fpr[m], tpr[m], alpha=0.7, label=f'{m} (area = {roc[m]:0.3f})') plt.plot([0, 1], [0, 1], color='dimgrey', lw=2, linestyle='--') plt.xlim([0.0, 1.0]) plt.ylim([0.0, 1.05]) if title: plt.title(title) plt.xlabel('False Positive Rate') plt.ylabel('True Positive Rate') plt.legend(loc='lower right') def plot_confusion_matrix(cm, classes, prec=0, title='Confusion Matrix', cbar=True, cmap=plt.cm.Blues, ): """ Plots the confusion matrix with count and normalzied values. Parameters ---------- cm: nd.array confusiton matrix count values classes: list(str) class labels for x-axis prec: int, default=0 precision of normalized values title: str, default='Confusion Matrix' figure title cbar: bool, default=True if True include color bar in figure cmap: matplotlib colormap, default=plt.cm.Blues colormap for confusion matrix Returns ------- plot of confusion matrix """ plt.imshow(cm, interpolation='nearest', cmap=cmap) if title: plt.title(title) if cbar: plt.colorbar() tick_marks = np.arange(len(classes)) plt.xticks(tick_marks, classes, rotation=0) plt.yticks(tick_marks, classes) # Normalize counts to % cm_norm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis] thresh = cm.max() / 2. # threshold for text color so it is visible for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])): plt.text(j, i, f'{cm[i, j]:.0f}\n({100*cm_norm[i, j]:.{prec}f}%)', horizontalalignment="center", color="white" if cm[i, j] > thresh else "black") plt.tight_layout() plt.ylabel('True Label') plt.xlabel('Predicted Label') plt.grid(False)
[ "matplotlib.pyplot.grid", "matplotlib.pyplot.ylabel", "matplotlib.cm.astype", "matplotlib.pyplot.imshow", "matplotlib.pyplot.xlabel", "matplotlib.pyplot.plot", "matplotlib.pyplot.yticks", "matplotlib.pyplot.ylim", "matplotlib.cm.get_cmap", "matplotlib.pyplot.xticks", "seaborn.heatmap", "matplo...
[((1250, 1262), 'matplotlib.pyplot.figure', 'plt.figure', ([], {}), '()\n', (1260, 1262), True, 'import matplotlib.pyplot as plt\n'), ((1306, 1402), 'seaborn.heatmap', 'sns.heatmap', (['cdf'], {'cmap': '"""RdYlGn"""', 'linewidths': '(0.5)', 'annot': 'labels', 'fmt': 'f"""0.{prec}f"""', 'cbar': '(False)'}), "(cdf, cmap='RdYlGn', linewidths=0.5, annot=labels, fmt=\n f'0.{prec}f', cbar=False)\n", (1317, 1402), True, 'import seaborn as sns\n'), ((1418, 1440), 'matplotlib.pyplot.xticks', 'plt.xticks', ([], {'rotation': '(0)'}), '(rotation=0)\n', (1428, 1440), True, 'import matplotlib.pyplot as plt\n'), ((1445, 1467), 'matplotlib.pyplot.yticks', 'plt.yticks', ([], {'rotation': '(0)'}), '(rotation=0)\n', (1455, 1467), True, 'import matplotlib.pyplot as plt\n'), ((6237, 6264), 'matplotlib.pyplot.figure', 'plt.figure', ([], {'figsize': 'figsize'}), '(figsize=figsize)\n', (6247, 6264), True, 'import matplotlib.pyplot as plt\n'), ((6420, 6483), 'matplotlib.pyplot.plot', 'plt.plot', (['[0, 1]', '[0, 1]'], {'color': '"""dimgrey"""', 'lw': '(2)', 'linestyle': '"""--"""'}), "([0, 1], [0, 1], color='dimgrey', lw=2, linestyle='--')\n", (6428, 6483), True, 'import matplotlib.pyplot as plt\n'), ((6488, 6508), 'matplotlib.pyplot.xlim', 'plt.xlim', (['[0.0, 1.0]'], {}), '([0.0, 1.0])\n', (6496, 6508), True, 'import matplotlib.pyplot as plt\n'), ((6513, 6534), 'matplotlib.pyplot.ylim', 'plt.ylim', (['[0.0, 1.05]'], {}), '([0.0, 1.05])\n', (6521, 6534), True, 'import matplotlib.pyplot as plt\n'), ((6578, 6611), 'matplotlib.pyplot.xlabel', 'plt.xlabel', (['"""False Positive Rate"""'], {}), "('False Positive Rate')\n", (6588, 6611), True, 'import matplotlib.pyplot as plt\n'), ((6616, 6648), 'matplotlib.pyplot.ylabel', 'plt.ylabel', (['"""True Positive Rate"""'], {}), "('True Positive Rate')\n", (6626, 6648), True, 'import matplotlib.pyplot as plt\n'), ((6653, 6682), 'matplotlib.pyplot.legend', 'plt.legend', ([], {'loc': '"""lower right"""'}), "(loc='lower right')\n", (6663, 6682), True, 'import matplotlib.pyplot as plt\n'), ((7519, 7569), 'matplotlib.pyplot.imshow', 'plt.imshow', (['cm'], {'interpolation': '"""nearest"""', 'cmap': 'cmap'}), "(cm, interpolation='nearest', cmap=cmap)\n", (7529, 7569), True, 'import matplotlib.pyplot as plt\n'), ((7692, 7735), 'matplotlib.pyplot.xticks', 'plt.xticks', (['tick_marks', 'classes'], {'rotation': '(0)'}), '(tick_marks, classes, rotation=0)\n', (7702, 7735), True, 'import matplotlib.pyplot as plt\n'), ((7740, 7771), 'matplotlib.pyplot.yticks', 'plt.yticks', (['tick_marks', 'classes'], {}), '(tick_marks, classes)\n', (7750, 7771), True, 'import matplotlib.pyplot as plt\n'), ((8207, 8225), 'matplotlib.pyplot.tight_layout', 'plt.tight_layout', ([], {}), '()\n', (8223, 8225), True, 'import matplotlib.pyplot as plt\n'), ((8230, 8254), 'matplotlib.pyplot.ylabel', 'plt.ylabel', (['"""True Label"""'], {}), "('True Label')\n", (8240, 8254), True, 'import matplotlib.pyplot as plt\n'), ((8259, 8288), 'matplotlib.pyplot.xlabel', 'plt.xlabel', (['"""Predicted Label"""'], {}), "('Predicted Label')\n", (8269, 8288), True, 'import matplotlib.pyplot as plt\n'), ((8293, 8308), 'matplotlib.pyplot.grid', 'plt.grid', (['(False)'], {}), '(False)\n', (8301, 8308), True, 'import matplotlib.pyplot as plt\n'), ((1285, 1301), 'matplotlib.pyplot.title', 'plt.title', (['title'], {}), '(title)\n', (1294, 1301), True, 'import matplotlib.pyplot as plt\n'), ((4732, 4762), 'matplotlib.pyplot.subplots', 'plt.subplots', ([], {'figsize': '(12, 10)'}), '(figsize=(12, 10))\n', (4744, 4762), True, 'import matplotlib.pyplot as plt\n'), ((4803, 4831), 'matplotlib.cm.get_cmap', 'cm.get_cmap', (['"""coolwarm"""', '(300)'], {}), "('coolwarm', 300)\n", (4814, 4831), False, 'from matplotlib import cm\n'), ((6343, 6415), 'matplotlib.pyplot.plot', 'plt.plot', (['fpr[m]', 'tpr[m]'], {'alpha': '(0.7)', 'label': 'f"""{m} (area = {roc[m]:0.3f})"""'}), "(fpr[m], tpr[m], alpha=0.7, label=f'{m} (area = {roc[m]:0.3f})')\n", (6351, 6415), True, 'import matplotlib.pyplot as plt\n'), ((6557, 6573), 'matplotlib.pyplot.title', 'plt.title', (['title'], {}), '(title)\n', (6566, 6573), True, 'import matplotlib.pyplot as plt\n'), ((7593, 7609), 'matplotlib.pyplot.title', 'plt.title', (['title'], {}), '(title)\n', (7602, 7609), True, 'import matplotlib.pyplot as plt\n'), ((7631, 7645), 'matplotlib.pyplot.colorbar', 'plt.colorbar', ([], {}), '()\n', (7643, 7645), True, 'import matplotlib.pyplot as plt\n'), ((7815, 7833), 'matplotlib.cm.astype', 'cm.astype', (['"""float"""'], {}), "('float')\n", (7824, 7833), False, 'from matplotlib import cm\n'), ((7880, 7888), 'matplotlib.cm.max', 'cm.max', ([], {}), '()\n', (7886, 7888), False, 'from matplotlib import cm\n'), ((8022, 8180), 'matplotlib.pyplot.text', 'plt.text', (['j', 'i', 'f"""{cm[i, j]:.0f}\n({100 * cm_norm[i, j]:.{prec}f}%)"""'], {'horizontalalignment': '"""center"""', 'color': "('white' if cm[i, j] > thresh else 'black')"}), '(j, i, f"""{cm[i, j]:.0f}\n({100 * cm_norm[i, j]:.{prec}f}%)""",\n horizontalalignment=\'center\', color=\'white\' if cm[i, j] > thresh else\n \'black\')\n', (8030, 8180), True, 'import matplotlib.pyplot as plt\n'), ((2374, 2393), 'pandas.Series', 'pd.Series', (['(100 * ev)'], {}), '(100 * ev)\n', (2383, 2393), True, 'import pandas as pd\n'), ((7836, 7850), 'matplotlib.cm.sum', 'cm.sum', ([], {'axis': '(1)'}), '(axis=1)\n', (7842, 7850), False, 'from matplotlib import cm\n'), ((3871, 3890), 'numpy.cumsum', 'np.cumsum', (['(100 * ev)'], {}), '(100 * ev)\n', (3880, 3890), True, 'import numpy as np\n'), ((3556, 3575), 'numpy.cumsum', 'np.cumsum', (['(100 * ev)'], {}), '(100 * ev)\n', (3565, 3575), True, 'import numpy as np\n'), ((3314, 3327), 'numpy.cumsum', 'np.cumsum', (['ev'], {}), '(ev)\n', (3323, 3327), True, 'import numpy as np\n'), ((3377, 3390), 'numpy.cumsum', 'np.cumsum', (['ev'], {}), '(ev)\n', (3386, 3390), True, 'import numpy as np\n')]
#Pre-Proccessed Images Data Collection import cv2 import numpy as np import os photo = cv2.imread("path of image") #photo should be of 640x480 pixels and axis must match. name = input("Enter your name : ") frames = [] outputs = [] frames.append(photo.flatten()) outputs.append([name]) X = np.array(frames) y = np.array(outputs) data = np.hstack([y, X]) f_name = "face_data.npy" if os.path.exists(f_name): old = np.load(f_name) data = np.vstack([old, data]) np.save(f_name, data)
[ "os.path.exists", "numpy.hstack", "numpy.array", "numpy.vstack", "numpy.load", "cv2.imread", "numpy.save" ]
[((91, 118), 'cv2.imread', 'cv2.imread', (['"""path of image"""'], {}), "('path of image')\n", (101, 118), False, 'import cv2\n'), ((305, 321), 'numpy.array', 'np.array', (['frames'], {}), '(frames)\n', (313, 321), True, 'import numpy as np\n'), ((327, 344), 'numpy.array', 'np.array', (['outputs'], {}), '(outputs)\n', (335, 344), True, 'import numpy as np\n'), ((355, 372), 'numpy.hstack', 'np.hstack', (['[y, X]'], {}), '([y, X])\n', (364, 372), True, 'import numpy as np\n'), ((407, 429), 'os.path.exists', 'os.path.exists', (['f_name'], {}), '(f_name)\n', (421, 429), False, 'import os\n'), ((496, 517), 'numpy.save', 'np.save', (['f_name', 'data'], {}), '(f_name, data)\n', (503, 517), True, 'import numpy as np\n'), ((442, 457), 'numpy.load', 'np.load', (['f_name'], {}), '(f_name)\n', (449, 457), True, 'import numpy as np\n'), ((470, 492), 'numpy.vstack', 'np.vstack', (['[old, data]'], {}), '([old, data])\n', (479, 492), True, 'import numpy as np\n')]
import unittest from numba.cuda.testing import CUDATestCase, skip_on_cudasim from numba.tests.support import captured_stdout @skip_on_cudasim("cudasim doesn't support cuda import at non-top-level") class TestMonteCarlo(CUDATestCase): """ Test monte-carlo integration """ def setUp(self): # Prevent output from this test showing up when running the test suite self._captured_stdout = captured_stdout() self._captured_stdout.__enter__() super().setUp() def tearDown(self): # No exception type, value, or traceback self._captured_stdout.__exit__(None, None, None) super().tearDown() def test_ex_montecarlo(self): # ex_montecarlo.import.begin import numba import numpy as np from numba import cuda from numba.cuda.random import ( create_xoroshiro128p_states, xoroshiro128p_uniform_float32, ) # ex_montecarlo.import.end # ex_montecarlo.define.begin # number of samples, higher will lead to a more accurate answer nsamps = 1000000 # ex_montecarlo.define.end # ex_montecarlo.kernel.begin @cuda.jit def mc_integrator_kernel(out, rng_states, lower_lim, upper_lim): """ kernel to draw random samples and evaluate the function to be integrated at those sample values """ size = len(out) gid = cuda.grid(1) if gid < size: # draw a sample between 0 and 1 on this thread samp = xoroshiro128p_uniform_float32(rng_states, gid) # normalize this sample to the limit range samp = samp * (upper_lim - lower_lim) + lower_lim # evaluate the function to be # integrated at the normalized # value of the sample y = func(samp) out[gid] = y # ex_montecarlo.kernel.end # ex_montecarlo.callfunc.begin @cuda.reduce def sum_reduce(a, b): return a + b def mc_integrate(lower_lim, upper_lim, nsamps): """ approximate the definite integral of `func` from `lower_lim` to `upper_lim` """ out = cuda.to_device(np.zeros(nsamps, dtype="float32")) rng_states = create_xoroshiro128p_states(nsamps, seed=42) # jit the function for use in CUDA kernels mc_integrator_kernel.forall(nsamps)( out, rng_states, lower_lim, upper_lim ) # normalization factor to convert # to the average: (b - a)/(N - 1) factor = (upper_lim - lower_lim) / (nsamps - 1) return sum_reduce(out) * factor # ex_montecarlo.callfunc.end # ex_montecarlo.launch.begin # define a function to integrate @numba.jit def func(x): return 1.0 / x mc_integrate(1, 2, nsamps) # array(0.6929643, dtype=float32) mc_integrate(2, 3, nsamps) # array(0.4054021, dtype=float32) # ex_montecarlo.launch.end # values computed independently using maple np.testing.assert_allclose( mc_integrate(1, 2, nsamps), 0.69315, atol=0.001 ) np.testing.assert_allclose( mc_integrate(2, 3, nsamps), 0.4055, atol=0.001 ) if __name__ == "__main__": unittest.main()
[ "numba.cuda.random.create_xoroshiro128p_states", "numba.cuda.random.xoroshiro128p_uniform_float32", "numba.cuda.grid", "unittest.main", "numba.tests.support.captured_stdout", "numpy.zeros", "numba.cuda.testing.skip_on_cudasim" ]
[((129, 200), 'numba.cuda.testing.skip_on_cudasim', 'skip_on_cudasim', (['"""cudasim doesn\'t support cuda import at non-top-level"""'], {}), '("cudasim doesn\'t support cuda import at non-top-level")\n', (144, 200), False, 'from numba.cuda.testing import CUDATestCase, skip_on_cudasim\n'), ((3475, 3490), 'unittest.main', 'unittest.main', ([], {}), '()\n', (3488, 3490), False, 'import unittest\n'), ((419, 436), 'numba.tests.support.captured_stdout', 'captured_stdout', ([], {}), '()\n', (434, 436), False, 'from numba.tests.support import captured_stdout\n'), ((1479, 1491), 'numba.cuda.grid', 'cuda.grid', (['(1)'], {}), '(1)\n', (1488, 1491), False, 'from numba import cuda\n'), ((2403, 2447), 'numba.cuda.random.create_xoroshiro128p_states', 'create_xoroshiro128p_states', (['nsamps'], {'seed': '(42)'}), '(nsamps, seed=42)\n', (2430, 2447), False, 'from numba.cuda.random import create_xoroshiro128p_states, xoroshiro128p_uniform_float32\n'), ((1605, 1651), 'numba.cuda.random.xoroshiro128p_uniform_float32', 'xoroshiro128p_uniform_float32', (['rng_states', 'gid'], {}), '(rng_states, gid)\n', (1634, 1651), False, 'from numba.cuda.random import create_xoroshiro128p_states, xoroshiro128p_uniform_float32\n'), ((2343, 2376), 'numpy.zeros', 'np.zeros', (['nsamps'], {'dtype': '"""float32"""'}), "(nsamps, dtype='float32')\n", (2351, 2376), True, 'import numpy as np\n')]
from typing import List, Optional, Tuple from collections import defaultdict import pickle import json from os import path import click from tqdm import tqdm from sklearn.feature_extraction.text import TfidfVectorizer from flask import Flask, jsonify, request import numpy as np from qanta import util from qanta.dataset import QuizBowlDataset import nltk MODEL_PATH = 'tfidf.pickle' BUZZ_NUM_GUESSES = 5 BUZZ_THRESHOLD = 0.23 W2V_LENGTH = 300 W2V_MODEL = 'full_model.pkl' W2V_LAMBDA = 0.5 IS_MULTI = True IS_BERT = True bert = None guesser = None if IS_BERT: import qanta_bert bert = qanta_bert.qanta_bert() def get_topk(question, docs): arrs = [guesser.answer_docs[d] for d in docs] example = bert.predict(question, arrs) if len(example) <= 0 or len(example[0]) <= 0: return 0 else: return int(example[0][0]["doc_index"]) def guess_and_buzz(model, question_text, idx, multi) -> Tuple[str, bool, int]: if multi: guesses = model.guess([question_text], [idx], BUZZ_NUM_GUESSES)[0] else: guesses = model.guess([question_text], [idx], BUZZ_NUM_GUESSES)[0] scores = [guess[1] for guess in guesses] #print(scores[0] / sum(scores)) buzz = scores[0] / sum(scores) >= BUZZ_THRESHOLD buzz = buzz and idx > 1 if multi: if not IS_BERT: return [guess[0] for guess in guesses], buzz else: topk = get_topk(question_text, [guess[0] for guess in guesses]) buzz = scores[topk] / sum(scores) >= BUZZ_THRESHOLD return guesses[topk][0], buzz else: return guesses[0][0], buzz def batch_guess_and_buzz(model, questions, idxs, multi) -> List[Tuple[str, bool, int]]: if multi: question_guesses = model.guess(questions, idxs, BUZZ_NUM_GUESSES) else: question_guesses = model.guess(questions, idxs, BUZZ_NUM_GUESSES) outputs = [] assert(len(questions) == len(question_guesses)) for i, guesses in tqdm(enumerate(question_guesses)): scores = [guess[1] for guess in guesses] buzz = scores[0] / sum(scores) >= BUZZ_THRESHOLD buzz = buzz and idxs[i] > 1 #print(scores[0] / sum(scores)) if multi: if not IS_BERT: outputs.append(([guess[0] for guess in guesses], buzz)) else: topk = get_topk(questions[i], [guess[0] for guess in guesses]) buzz = scores[topk] / sum(scores) >= BUZZ_THRESHOLD outputs.append((guesses[topk][0], buzz)) else: outputs.append((guesses[0][0], buzz)) return outputs class TfidfGuesser: def __init__(self): self.tfidf_vectorizer = None self.tfidf_matrix = None self.i_to_ans = None self.w2v = pickle.load(open(W2V_MODEL, "rb")) self.answer_docs = defaultdict(str) if IS_BERT: with open("wiki_lookup.json", "r") as f: wiki = json.load(f) for k in tqdm(wiki): self.answer_docs[k] += ' ' + wiki[k]["text"] with open("data/qanta.mapped.2018.04.18.json") as f: dataset = json.load(f) raw_questions = dataset["questions"] GUESSER_TRAIN_FOLD = 'guesstrain' BUZZER_TRAIN_FOLD = 'buzztrain' TRAIN_FOLDS = {GUESSER_TRAIN_FOLD, BUZZER_TRAIN_FOLD} for q in tqdm(raw_questions): if q['fold'] in TRAIN_FOLDS: self.answer_docs[q['page']] += ' ' + q['text'] def train(self, training_data) -> None: questions = training_data[0] answers = training_data[1] answer_docs = defaultdict(str) answer_vecs = defaultdict(lambda: []) for q, ans in tqdm(zip(questions, answers)): text = ' '.join(q) answer_docs[ans] += ' ' + text for s in q: for w in nltk.word_tokenize(s): if w in self.w2v: answer_vecs[ans].append(self.w2v[w]) x_array = [] y_array = [] self.vecs_array = [] for ans, doc in tqdm(answer_docs.items()): x_array.append(doc) y_array.append(ans) if(len(answer_vecs[ans]) == 0): vec = np.zeros(W2V_LENGTH) else: vec = np.sum(answer_vecs[ans], axis = 0) / len(answer_vecs[ans]) vec = vec / np.linalg.norm(vec) self.vecs_array.append(vec) self.vecs_array = np.array(self.vecs_array) self.i_to_ans = {i: ans for i, ans in enumerate(y_array)} print("Fitting") self.tfidf_vectorizer = TfidfVectorizer( ngram_range=(1, 1), min_df=2, max_df=.9 , stop_words = "english").fit(x_array) print("Transform") self.tfidf_matrix = self.tfidf_vectorizer.transform(x_array) def guess(self, questions: List[str], idxs: Optional[int], max_n_guesses: Optional[int]) -> List[List[Tuple[str, float]]]: representations = self.tfidf_vectorizer.transform(questions) guess_matrix = self.tfidf_matrix.dot(representations.T).T ''' print(len(questions), representations.shape) print(self.tfidf_matrix.shape) print(guess_matrix.shape) print(self.vecs_array.shape) ''' vecs_represent = [] for q in questions: temp = [] for w in nltk.word_tokenize(q): if w in self.w2v: temp.append(self.w2v[w]) if not temp: vecs_represent.append(np.zeros(W2V_LENGTH)) else: temp = np.sum(temp, axis = 0) / len(temp) temp = temp / np.linalg.norm(temp) vecs_represent.append(temp) vecs_represent = np.array(vecs_represent) self.vecs_array = np.array(self.vecs_array) for i in range(len(self.vecs_array)): if(type(self.vecs_array[i]) != np.ndarray): print(self.vecs_array[i]) vecs_matrix = self.vecs_array.dot(vecs_represent.T).T guess_matrix = guess_matrix.toarray() + W2V_LAMBDA * vecs_matrix guess_indices = (-guess_matrix).argsort(axis=1)[:, 0:max_n_guesses] guesses = [] for i in range(len(questions)): idxs = guess_indices[i] guesses.append([(self.i_to_ans[j], guess_matrix[i, j]) for j in idxs]) return guesses def save(self): with open(MODEL_PATH, 'wb') as f: pickle.dump({ 'i_to_ans': self.i_to_ans, 'tfidf_vectorizer': self.tfidf_vectorizer, 'tfidf_matrix': self.tfidf_matrix, 'vecs_array': self.vecs_array }, f) @classmethod def load(cls): with open(MODEL_PATH, 'rb') as f: params = pickle.load(f) guesser = TfidfGuesser() guesser.tfidf_vectorizer = params['tfidf_vectorizer'] guesser.tfidf_matrix = params['tfidf_matrix'] guesser.i_to_ans = params['i_to_ans'] guesser.vecs_array = params['vecs_array'] return guesser def create_app(enable_batch=True): global guesser tfidf_guesser = TfidfGuesser.load() guesser = tfidf_guesser app = Flask(__name__) @app.route('/api/1.0/quizbowl/status', methods=['GET']) def status(): return jsonify({ 'batch': enable_batch, 'batch_size': 200, 'ready': True, 'include_wiki_paragraphs': False }) @app.route('/api/1.0/quizbowl/act', methods=['POST']) def act(): idx = request.json['question_idx'] question = request.json['text'] guess, buzz = guess_and_buzz(tfidf_guesser, question, idx, IS_MULTI) return jsonify({'guess': guess, 'buzz': True if buzz else False}) @app.route('/api/1.0/quizbowl/batch_act', methods=['POST']) def batch_act(): idxs = [q['question_idx'] for q in request.json['questions']] questions = [q['text'] for q in request.json['questions']] return jsonify([ {'guess': guess, 'buzz': True if buzz else False} for guess, buzz in batch_guess_and_buzz(tfidf_guesser, questions, idxs, IS_MULTI) ]) return app @click.group() def cli(): pass @cli.command() @click.option('--host', default='0.0.0.0') @click.option('--port', default=4861) @click.option('--disable-batch', default=False, is_flag=True) def web(host, port, disable_batch): """ Start web server wrapping tfidf model """ app = create_app(enable_batch=not disable_batch) app.run(host=host, port=port, debug=False) @cli.command() def train(): """ Train the tfidf model, requires downloaded data and saves to models/ """ dataset = QuizBowlDataset(guesser_train=True) tfidf_guesser = TfidfGuesser() tfidf_guesser.train(dataset.training_data()) tfidf_guesser.save() @cli.command() @click.option('--local-qanta-prefix', default='data/') @click.option('--retrieve-paragraphs', default=False, is_flag=True) def download(local_qanta_prefix, retrieve_paragraphs): """ Run once to download qanta data to data/. Runs inside the docker container, but results save to host machine """ util.download(local_qanta_prefix, retrieve_paragraphs) if __name__ == '__main__': cli()
[ "pickle.dump", "nltk.word_tokenize", "flask.Flask", "click.group", "click.option", "flask.jsonify", "qanta.dataset.QuizBowlDataset", "tqdm.tqdm", "pickle.load", "numpy.array", "numpy.zeros", "qanta_bert.qanta_bert", "collections.defaultdict", "sklearn.feature_extraction.text.TfidfVectorize...
[((8331, 8344), 'click.group', 'click.group', ([], {}), '()\n', (8342, 8344), False, 'import click\n'), ((8383, 8424), 'click.option', 'click.option', (['"""--host"""'], {'default': '"""0.0.0.0"""'}), "('--host', default='0.0.0.0')\n", (8395, 8424), False, 'import click\n'), ((8426, 8462), 'click.option', 'click.option', (['"""--port"""'], {'default': '(4861)'}), "('--port', default=4861)\n", (8438, 8462), False, 'import click\n'), ((8464, 8524), 'click.option', 'click.option', (['"""--disable-batch"""'], {'default': '(False)', 'is_flag': '(True)'}), "('--disable-batch', default=False, is_flag=True)\n", (8476, 8524), False, 'import click\n'), ((9015, 9068), 'click.option', 'click.option', (['"""--local-qanta-prefix"""'], {'default': '"""data/"""'}), "('--local-qanta-prefix', default='data/')\n", (9027, 9068), False, 'import click\n'), ((9070, 9136), 'click.option', 'click.option', (['"""--retrieve-paragraphs"""'], {'default': '(False)', 'is_flag': '(True)'}), "('--retrieve-paragraphs', default=False, is_flag=True)\n", (9082, 9136), False, 'import click\n'), ((598, 621), 'qanta_bert.qanta_bert', 'qanta_bert.qanta_bert', ([], {}), '()\n', (619, 621), False, 'import qanta_bert\n'), ((7320, 7335), 'flask.Flask', 'Flask', (['__name__'], {}), '(__name__)\n', (7325, 7335), False, 'from flask import Flask, jsonify, request\n'), ((8852, 8887), 'qanta.dataset.QuizBowlDataset', 'QuizBowlDataset', ([], {'guesser_train': '(True)'}), '(guesser_train=True)\n', (8867, 8887), False, 'from qanta.dataset import QuizBowlDataset\n'), ((9325, 9379), 'qanta.util.download', 'util.download', (['local_qanta_prefix', 'retrieve_paragraphs'], {}), '(local_qanta_prefix, retrieve_paragraphs)\n', (9338, 9379), False, 'from qanta import util\n'), ((2843, 2859), 'collections.defaultdict', 'defaultdict', (['str'], {}), '(str)\n', (2854, 2859), False, 'from collections import defaultdict\n'), ((3694, 3710), 'collections.defaultdict', 'defaultdict', (['str'], {}), '(str)\n', (3705, 3710), False, 'from collections import defaultdict\n'), ((3733, 3757), 'collections.defaultdict', 'defaultdict', (['(lambda : [])'], {}), '(lambda : [])\n', (3744, 3757), False, 'from collections import defaultdict\n'), ((4543, 4568), 'numpy.array', 'np.array', (['self.vecs_array'], {}), '(self.vecs_array)\n', (4551, 4568), True, 'import numpy as np\n'), ((5836, 5860), 'numpy.array', 'np.array', (['vecs_represent'], {}), '(vecs_represent)\n', (5844, 5860), True, 'import numpy as np\n'), ((5887, 5912), 'numpy.array', 'np.array', (['self.vecs_array'], {}), '(self.vecs_array)\n', (5895, 5912), True, 'import numpy as np\n'), ((7430, 7534), 'flask.jsonify', 'jsonify', (["{'batch': enable_batch, 'batch_size': 200, 'ready': True,\n 'include_wiki_paragraphs': False}"], {}), "({'batch': enable_batch, 'batch_size': 200, 'ready': True,\n 'include_wiki_paragraphs': False})\n", (7437, 7534), False, 'from flask import Flask, jsonify, request\n'), ((7838, 7896), 'flask.jsonify', 'jsonify', (["{'guess': guess, 'buzz': True if buzz else False}"], {}), "({'guess': guess, 'buzz': True if buzz else False})\n", (7845, 7896), False, 'from flask import Flask, jsonify, request\n'), ((2990, 3000), 'tqdm.tqdm', 'tqdm', (['wiki'], {}), '(wiki)\n', (2994, 3000), False, 'from tqdm import tqdm\n'), ((5453, 5474), 'nltk.word_tokenize', 'nltk.word_tokenize', (['q'], {}), '(q)\n', (5471, 5474), False, 'import nltk\n'), ((6547, 6709), 'pickle.dump', 'pickle.dump', (["{'i_to_ans': self.i_to_ans, 'tfidf_vectorizer': self.tfidf_vectorizer,\n 'tfidf_matrix': self.tfidf_matrix, 'vecs_array': self.vecs_array}", 'f'], {}), "({'i_to_ans': self.i_to_ans, 'tfidf_vectorizer': self.\n tfidf_vectorizer, 'tfidf_matrix': self.tfidf_matrix, 'vecs_array': self\n .vecs_array}, f)\n", (6558, 6709), False, 'import pickle\n'), ((6878, 6892), 'pickle.load', 'pickle.load', (['f'], {}), '(f)\n', (6889, 6892), False, 'import pickle\n'), ((2956, 2968), 'json.load', 'json.load', (['f'], {}), '(f)\n', (2965, 2968), False, 'import json\n'), ((3155, 3167), 'json.load', 'json.load', (['f'], {}), '(f)\n', (3164, 3167), False, 'import json\n'), ((3414, 3433), 'tqdm.tqdm', 'tqdm', (['raw_questions'], {}), '(raw_questions)\n', (3418, 3433), False, 'from tqdm import tqdm\n'), ((3933, 3954), 'nltk.word_tokenize', 'nltk.word_tokenize', (['s'], {}), '(s)\n', (3951, 3954), False, 'import nltk\n'), ((4308, 4328), 'numpy.zeros', 'np.zeros', (['W2V_LENGTH'], {}), '(W2V_LENGTH)\n', (4316, 4328), True, 'import numpy as np\n'), ((4692, 4771), 'sklearn.feature_extraction.text.TfidfVectorizer', 'TfidfVectorizer', ([], {'ngram_range': '(1, 1)', 'min_df': '(2)', 'max_df': '(0.9)', 'stop_words': '"""english"""'}), "(ngram_range=(1, 1), min_df=2, max_df=0.9, stop_words='english')\n", (4707, 4771), False, 'from sklearn.feature_extraction.text import TfidfVectorizer\n'), ((4369, 4401), 'numpy.sum', 'np.sum', (['answer_vecs[ans]'], {'axis': '(0)'}), '(answer_vecs[ans], axis=0)\n', (4375, 4401), True, 'import numpy as np\n'), ((4456, 4475), 'numpy.linalg.norm', 'np.linalg.norm', (['vec'], {}), '(vec)\n', (4470, 4475), True, 'import numpy as np\n'), ((5618, 5638), 'numpy.zeros', 'np.zeros', (['W2V_LENGTH'], {}), '(W2V_LENGTH)\n', (5626, 5638), True, 'import numpy as np\n'), ((5681, 5701), 'numpy.sum', 'np.sum', (['temp'], {'axis': '(0)'}), '(temp, axis=0)\n', (5687, 5701), True, 'import numpy as np\n'), ((5746, 5766), 'numpy.linalg.norm', 'np.linalg.norm', (['temp'], {}), '(temp)\n', (5760, 5766), True, 'import numpy as np\n')]
#!/usr/bin/env python import csv import sys import os, os.path import numpy as np from datetime import datetime as dt from scipy import optimize from scripts import signals as sig from scripts import fft_estimator from scripts import optimizing from scripts import utility from scripts import crlb from scripts import cfg try: task = sys.argv[1] except Exception as e: print("No input task provided, exiting. \n Usage: python main.py <task>") exit(1) SNR_dBs =[-10, 0, 10, 20, 30, 40, 50, 60] FFT_Ks = [10, 12, 14, 16, 18, 20] n = len(SNR_dBs) m = len(FFT_Ks) N = 100 # Amount of samples to generate when estimating variance # Generate unique filename for data file output run_number = len([name for name in os.listdir('./data') if os.path.isfile('./data/' + name)]) if task == 'a': filename = 'data/part_a_run_' + str(run_number) + '_N_' + str(N) + '.csv' with open(filename, 'ab') as file: writer = csv.writer(file, delimiter=' ') total_time_begin = dt.now() for i in range(m): K = FFT_Ks[i] M = 2**K for j in range(n): SNR_dB = SNR_dBs[j] w_estimates = np.zeros(N) phi_estimates = np.zeros(N) status_bar_progress = 0 run_time_begin = dt.now() for k in range(N): x_d = sig.x_discrete(SNR_dB) omega_hat, phi_hat, _, _ = fft_estimator.estimator(x_d, M) w_estimates[k] = omega_hat phi_estimates[k] = phi_hat status_bar_progress = utility.print_status_bar(k, status_bar_progress, N) mean_f = np.mean(w_estimates) / (2*np.pi) mean_phi = np.mean(phi_estimates) var_f = np.var(w_estimates) var_phi = np.var(phi_estimates) crlb_f = crlb.omega(SNR_dB) crlb_phi = crlb.phi(SNR_dB) run_time_end = dt.now() print("") utility.print_execution_time(run_time_begin, run_time_end) f_estimate_valid = True phi_estimate_valid = True if var_f < crlb_f: f_estimate_valid = False print("Variance for frequency lower than CRLB!") if var_phi < crlb_phi: phi_estimate_valid = False print("Variance for phi lower than CRLB!") writer.writerow([SNR_dB, K, crlb_f, var_f, f_estimate_valid, crlb_phi, var_phi, phi_estimate_valid, mean_f, mean_phi]) print("CONFIG | SNR [dB]: {}, M: 2^{}, true frequency: {}, true phase: {}".format(SNR_dB, K, cfg.f0, cfg.phi)) print("FREQUENCY | estimated mean: {}, estimated variance: {}, crlb: {}".format(mean_f, var_f, crlb_f)) print("PHASE | estimated mean: {}, estimated variance: {}, crlb: {}".format(mean_phi, var_phi, crlb_phi)) print("") total_time_end = dt.now() utility.print_execution_time(total_time_begin, total_time_end) if task == 'b': filename = 'data/part_b_run_' + str(run_number) + '_N_' + str(N) + '.csv' with open(filename, 'ab') as file: writer = csv.writer(file, delimiter=' ') M = 2**10 total_time_begin = dt.now() for SNR_dB in SNR_dBs: w_estimates = np.zeros(N) phi_estimates = np.zeros(N) status_bar_progress = 0 run_time_begin = dt.now() for i in range(N): x_d = sig.x_discrete(SNR_dB) omega_hat, phi_hat, _, _ = fft_estimator.estimator(x_d, M) omega_opt = optimize.minimize(optimizing.frequency_objective_function, omega_hat, method="Nelder-Mead", args=(M, x_d, phi_hat)) phase_opt = optimize.minimize(optimizing.phase_objective_function, phi_hat, method="Nelder-Mead", args=(x_d, omega_hat)) w_estimates[i] = omega_opt.x[0] phi_estimates[i] = phase_opt.x[0] status_bar_progress = utility.print_status_bar(i, status_bar_progress, N) run_time_end = dt.now() print("") utility.print_execution_time(run_time_begin, run_time_end) mean_f = np.mean(w_estimates) / (2*np.pi) mean_phi = np.mean(phi_estimates) var_f = np.var(w_estimates) var_phi = np.var(phi_estimates) crlb_f = crlb.omega(SNR_dB) crlb_phi = crlb.phi(SNR_dB) f_estimate_valid = True phi_estimate_valid = True if var_f < crlb_f: f_estimate_valid = False print("Variance for f lower than CRLB!") if var_phi < crlb_phi: phi_estimate_valid = False print("Variance for phi lower than CRLB!") writer.writerow([SNR_dB, 10, crlb_f, var_f, f_estimate_valid, crlb_phi, var_phi, phi_estimate_valid, mean_f, mean_phi]) print("CONFIG | SNR [dB]: {}, M: 2^{}, true f: {}, true phase: {}".format(SNR_dB, 10, cfg.f0, cfg.phi)) print("FREQUENCY | estimated mean: {}, estimated variance: {}, crlb: {}".format(mean_f, var_f, crlb_f)) print("PHASE | estimated mean: {}, estimated variance: {}, crlb: {}".format(mean_phi, var_phi, crlb_phi)) print("") total_time_end = dt.now() utility.print_execution_time(total_time_begin, total_time_end)
[ "numpy.mean", "scripts.utility.print_execution_time", "os.listdir", "csv.writer", "scripts.fft_estimator.estimator", "scipy.optimize.minimize", "os.path.isfile", "datetime.datetime.now", "numpy.zeros", "scripts.crlb.phi", "scripts.utility.print_status_bar", "scripts.signals.x_discrete", "num...
[((940, 971), 'csv.writer', 'csv.writer', (['file'], {'delimiter': '""" """'}), "(file, delimiter=' ')\n", (950, 971), False, 'import csv\n'), ((1000, 1008), 'datetime.datetime.now', 'dt.now', ([], {}), '()\n', (1006, 1008), True, 'from datetime import datetime as dt\n'), ((3071, 3079), 'datetime.datetime.now', 'dt.now', ([], {}), '()\n', (3077, 3079), True, 'from datetime import datetime as dt\n'), ((3088, 3150), 'scripts.utility.print_execution_time', 'utility.print_execution_time', (['total_time_begin', 'total_time_end'], {}), '(total_time_begin, total_time_end)\n', (3116, 3150), False, 'from scripts import utility\n'), ((3302, 3333), 'csv.writer', 'csv.writer', (['file'], {'delimiter': '""" """'}), "(file, delimiter=' ')\n", (3312, 3333), False, 'import csv\n'), ((3388, 3396), 'datetime.datetime.now', 'dt.now', ([], {}), '()\n', (3394, 3396), True, 'from datetime import datetime as dt\n'), ((5494, 5502), 'datetime.datetime.now', 'dt.now', ([], {}), '()\n', (5500, 5502), True, 'from datetime import datetime as dt\n'), ((5511, 5573), 'scripts.utility.print_execution_time', 'utility.print_execution_time', (['total_time_begin', 'total_time_end'], {}), '(total_time_begin, total_time_end)\n', (5539, 5573), False, 'from scripts import utility\n'), ((730, 750), 'os.listdir', 'os.listdir', (['"""./data"""'], {}), "('./data')\n", (740, 750), False, 'import os, os.path\n'), ((754, 786), 'os.path.isfile', 'os.path.isfile', (["('./data/' + name)"], {}), "('./data/' + name)\n", (768, 786), False, 'import os, os.path\n'), ((3455, 3466), 'numpy.zeros', 'np.zeros', (['N'], {}), '(N)\n', (3463, 3466), True, 'import numpy as np\n'), ((3495, 3506), 'numpy.zeros', 'np.zeros', (['N'], {}), '(N)\n', (3503, 3506), True, 'import numpy as np\n'), ((3573, 3581), 'datetime.datetime.now', 'dt.now', ([], {}), '()\n', (3579, 3581), True, 'from datetime import datetime as dt\n'), ((4236, 4244), 'datetime.datetime.now', 'dt.now', ([], {}), '()\n', (4242, 4244), True, 'from datetime import datetime as dt\n'), ((4279, 4337), 'scripts.utility.print_execution_time', 'utility.print_execution_time', (['run_time_begin', 'run_time_end'], {}), '(run_time_begin, run_time_end)\n', (4307, 4337), False, 'from scripts import utility\n'), ((4416, 4438), 'numpy.mean', 'np.mean', (['phi_estimates'], {}), '(phi_estimates)\n', (4423, 4438), True, 'import numpy as np\n'), ((4460, 4479), 'numpy.var', 'np.var', (['w_estimates'], {}), '(w_estimates)\n', (4466, 4479), True, 'import numpy as np\n'), ((4502, 4523), 'numpy.var', 'np.var', (['phi_estimates'], {}), '(phi_estimates)\n', (4508, 4523), True, 'import numpy as np\n'), ((4546, 4564), 'scripts.crlb.omega', 'crlb.omega', (['SNR_dB'], {}), '(SNR_dB)\n', (4556, 4564), False, 'from scripts import crlb\n'), ((4588, 4604), 'scripts.crlb.phi', 'crlb.phi', (['SNR_dB'], {}), '(SNR_dB)\n', (4596, 4604), False, 'from scripts import crlb\n'), ((1182, 1193), 'numpy.zeros', 'np.zeros', (['N'], {}), '(N)\n', (1190, 1193), True, 'import numpy as np\n'), ((1226, 1237), 'numpy.zeros', 'np.zeros', (['N'], {}), '(N)\n', (1234, 1237), True, 'import numpy as np\n'), ((1312, 1320), 'datetime.datetime.now', 'dt.now', ([], {}), '()\n', (1318, 1320), True, 'from datetime import datetime as dt\n'), ((1761, 1783), 'numpy.mean', 'np.mean', (['phi_estimates'], {}), '(phi_estimates)\n', (1768, 1783), True, 'import numpy as np\n'), ((1809, 1828), 'numpy.var', 'np.var', (['w_estimates'], {}), '(w_estimates)\n', (1815, 1828), True, 'import numpy as np\n'), ((1855, 1876), 'numpy.var', 'np.var', (['phi_estimates'], {}), '(phi_estimates)\n', (1861, 1876), True, 'import numpy as np\n'), ((1903, 1921), 'scripts.crlb.omega', 'crlb.omega', (['SNR_dB'], {}), '(SNR_dB)\n', (1913, 1921), False, 'from scripts import crlb\n'), ((1949, 1965), 'scripts.crlb.phi', 'crlb.phi', (['SNR_dB'], {}), '(SNR_dB)\n', (1957, 1965), False, 'from scripts import crlb\n'), ((1998, 2006), 'datetime.datetime.now', 'dt.now', ([], {}), '()\n', (2004, 2006), True, 'from datetime import datetime as dt\n'), ((2049, 2107), 'scripts.utility.print_execution_time', 'utility.print_execution_time', (['run_time_begin', 'run_time_end'], {}), '(run_time_begin, run_time_end)\n', (2077, 2107), False, 'from scripts import utility\n'), ((3637, 3659), 'scripts.signals.x_discrete', 'sig.x_discrete', (['SNR_dB'], {}), '(SNR_dB)\n', (3651, 3659), True, 'from scripts import signals as sig\n'), ((3704, 3735), 'scripts.fft_estimator.estimator', 'fft_estimator.estimator', (['x_d', 'M'], {}), '(x_d, M)\n', (3727, 3735), False, 'from scripts import fft_estimator\n'), ((3765, 3884), 'scipy.optimize.minimize', 'optimize.minimize', (['optimizing.frequency_objective_function', 'omega_hat'], {'method': '"""Nelder-Mead"""', 'args': '(M, x_d, phi_hat)'}), "(optimizing.frequency_objective_function, omega_hat,\n method='Nelder-Mead', args=(M, x_d, phi_hat))\n", (3782, 3884), False, 'from scipy import optimize\n'), ((3909, 4022), 'scipy.optimize.minimize', 'optimize.minimize', (['optimizing.phase_objective_function', 'phi_hat'], {'method': '"""Nelder-Mead"""', 'args': '(x_d, omega_hat)'}), "(optimizing.phase_objective_function, phi_hat, method=\n 'Nelder-Mead', args=(x_d, omega_hat))\n", (3926, 4022), False, 'from scipy import optimize\n'), ((4156, 4207), 'scripts.utility.print_status_bar', 'utility.print_status_bar', (['i', 'status_bar_progress', 'N'], {}), '(i, status_bar_progress, N)\n', (4180, 4207), False, 'from scripts import utility\n'), ((4360, 4380), 'numpy.mean', 'np.mean', (['w_estimates'], {}), '(w_estimates)\n', (4367, 4380), True, 'import numpy as np\n'), ((1382, 1404), 'scripts.signals.x_discrete', 'sig.x_discrete', (['SNR_dB'], {}), '(SNR_dB)\n', (1396, 1404), True, 'from scripts import signals as sig\n'), ((1453, 1484), 'scripts.fft_estimator.estimator', 'fft_estimator.estimator', (['x_d', 'M'], {}), '(x_d, M)\n', (1476, 1484), False, 'from scripts import fft_estimator\n'), ((1623, 1674), 'scripts.utility.print_status_bar', 'utility.print_status_bar', (['k', 'status_bar_progress', 'N'], {}), '(k, status_bar_progress, N)\n', (1647, 1674), False, 'from scripts import utility\n'), ((1701, 1721), 'numpy.mean', 'np.mean', (['w_estimates'], {}), '(w_estimates)\n', (1708, 1721), True, 'import numpy as np\n')]
from __future__ import annotations import math from collections import abc from functools import partial from typing import Callable, Iterable, List, Mapping, Optional, Sequence, Tuple, TypeVar, Union, cast import numpy as np import tensorflow as tf from typing_extensions import TypeGuard T1 = TypeVar("T1") T2 = TypeVar("T2") ContainerGeneric = Union[Mapping[str, "ContainerGeneric[T1]"], Sequence["ContainerGeneric[T1]"], T1] ContainerArrays = ContainerGeneric[Union[np.ndarray, np.number]] ContainerTensors = ContainerGeneric[Union[tf.Tensor, tf.SparseTensor]] ContainerAnyTensors = TypeVar("ContainerAnyTensors", ContainerTensors, ContainerArrays) ArrayLike = Union[np.ndarray, np.number, tf.Tensor, tf.SparseTensor, float] ComparisonResultT = Tuple[bool, Optional[List["int | str"]]] DEFAULT_ABSOLUTE_TOLERANCE = 1e-4 def container_fmap(f: Callable[[T1], T2], elements: ContainerGeneric[T1]) -> ContainerGeneric[T2]: if isinstance(elements, abc.Mapping): return {key: container_fmap(f, val) for key, val in elements.items()} if isinstance(elements, tuple): return tuple(map(partial(container_fmap, f), elements)) if isinstance(elements, abc.Sequence) and not isinstance(elements, (str, bytes)): return [container_fmap(f, elem) for elem in elements] return f(elements) def _is_integer_dtype(dtype: np.dtype | tf.dtypes.DType) -> bool: if isinstance(dtype, np.dtype): return np.issubdtype(dtype, np.integer) else: return dtype.is_integer def tensor_equality( tensor1: ArrayLike, tensor2: ArrayLike, atol: float = DEFAULT_ABSOLUTE_TOLERANCE, test_env: Optional[tf.test.TestCase] = None, ) -> bool: """ Checks if two tf tensors/numpy array are equal. For integral tensors checks for exact equality. For float tensors checks for approximate equality. It expects both tensors to have the same dtype and same shape. For tf tensors assumes they are eager tensors. Graph tensors need to be evaluated in session to numpy array. If you pass in tf.test.TestCase it will work with graph tensors that can be evaluated by it. It expects the type to be same shape with one exception, being that scalers can be compared against tensors. """ if isinstance(tensor1, tf.SparseTensor) or isinstance(tensor2, tf.SparseTensor): tensor1 = tf.sparse.to_dense(tensor1) tensor2 = tf.sparse.to_dense(tensor2) if isinstance(tensor1, float) and isinstance(tensor2, float): return math.isclose(tensor1, tensor2, abs_tol=atol) if isinstance(tensor1, (int, float)): if isinstance(tensor2, tf.Tensor): tensor1 = tf.constant(tensor1) else: tensor1 = np.array(tensor1) if isinstance(tensor2, (int, float)): if isinstance(tensor1, tf.Tensor): tensor2 = tf.constant(tensor2) else: tensor2 = np.array(tensor2) assert not isinstance(tensor1, (int, float)) assert not isinstance(tensor2, (int, float)) if type(tensor1) != type(tensor2): return False if tensor1.dtype != tensor2.dtype: return False if tensor1.shape != tensor2.shape: return False if test_env: assert isinstance(tensor1, tf.Tensor) array1 = test_env.evaluate(tensor1) array2 = test_env.evaluate(tensor2) else: if isinstance(tensor1, tf.Tensor) and isinstance(tensor2, tf.Tensor): array1 = tensor1.numpy() array2 = tensor2.numpy() else: assert isinstance(tensor1, (np.ndarray, np.number)) assert isinstance(tensor2, (np.ndarray, np.number)) array1 = tensor1 array2 = tensor2 if _is_integer_dtype(array1.dtype): return bool(np.all(array1 == array2)) else: comparisons = np.isclose(array1, array2, atol=atol, equal_nan=True) return bool(comparisons.all()) def extract_nested_key(data: ContainerGeneric[T1], key_path: Iterable[int | str]) -> T1: result = data for key in key_path: if isinstance(key, int): assert isinstance(result, abc.Sequence) result = result[key] else: assert isinstance(result, abc.Mapping) result = result[key] return cast(T1, result) def _tensor_collection_equality( tensors1: ContainerArrays, tensors2: ContainerArrays, atol: float = DEFAULT_ABSOLUTE_TOLERANCE, debug_path: Optional[List[str | int]] = None, ) -> ComparisonResultT: if debug_path is None: debug_path = [] if not (isinstance(tensors1, type(tensors2)) or isinstance(tensors2, type(tensors1))): return (False, debug_path) # The ors are here for the type checker. While previous line guarantees that the types are the same # the type checker is unable to use that. if not isinstance(tensors1, (np.ndarray, dict, list, tuple, np.number)) or not isinstance( tensors2, (np.ndarray, dict, list, tuple, np.number) ): raise TypeError(f"Unexpected type for tensors1: {type(tensors1)}") if isinstance(tensors1, (np.ndarray, np.number)) or isinstance(tensors2, (np.ndarray, np.number)): result = tensor_equality(tensors1, tensors2, atol=atol) if result: return (result, None) else: return (result, debug_path) if len(tensors1) != len(tensors2): return (False, debug_path) if isinstance(tensors1, dict): assert isinstance(tensors2, dict) for key, value in tensors1.items(): if key not in tensors2: return (False, debug_path + [key]) key_result, key_debug_path = _tensor_collection_equality( value, tensors2[key], atol=atol, debug_path=debug_path + [key] ) # Short circuit if possible. if not key_result: return (key_result, key_debug_path) return (True, None) assert isinstance(tensors2, (list, tuple)) for key, (tensor1, tensor2) in enumerate(zip(tensors1, tensors2)): debug_extension: List[int | str] = [key] key_result, key_debug_path = _tensor_collection_equality( tensor1, tensor2, atol=atol, debug_path=debug_path + debug_extension ) if not key_result: return (key_result, key_debug_path) return (True, None) def evaluate_tensors(tensors: ContainerTensors, test_env: Optional[tf.test.TestCase]) -> ContainerArrays: if test_env: return test_env.evaluate(tensors) else: return container_fmap(lambda tensor: tensor.numpy(), tensors) # type: ignore def _log_tensor_equality_mismatch( arrays1: ContainerArrays, arrays2: ContainerArrays, debug_path: Iterable[str | int] ) -> None: debug_path_str = ",".join(map(str, debug_path)) array1: np.ndarray | np.number array2: np.ndarray | np.number if debug_path_str != "": print(f"Tensor Collection mismatch occurred at {debug_path_str}") array1 = extract_nested_key(arrays1, debug_path) # type: ignore array2 = extract_nested_key(arrays2, debug_path) # type: ignore else: assert isinstance(arrays1, (np.ndarray, np.number)) assert isinstance(arrays2, (np.ndarray, np.number)) array1 = arrays1 array2 = arrays2 print(f"Mismatched tensors") print(array1) print(array2) shape_match = array1.shape == array2.shape dtype_match = array1.dtype == array2.dtype if not shape_match: print(f"Mismatch shapes") print(f"Shape 1: {array1.shape} Shape 2: {array2.shape}") if not dtype_match: print(f"Mismatch dtypes") print(f"Dtype 1: {array1.dtype} Dtype 2: {array2.dtype}") if shape_match and dtype_match: diff = np.absolute(array1 - array2) # type: ignore print(f"Maximum absolute difference: ", diff.max()) print(f"Index Maximum Difference: ", np.unravel_index(diff.argmax(), diff.shape)) def check_container_tensors(tensors: ContainerTensors | ContainerArrays) -> TypeGuard[ContainerTensors]: if isinstance(tensors, (abc.Mapping, abc.Sequence)): if len(tensors) == 0: return True if isinstance(tensors, abc.Mapping): return check_container_tensors(next(iter(tensors.values()))) else: return check_container_tensors(next(iter(tensors))) return isinstance(tensors, (tf.Tensor, tf.SparseTensor)) def tensor_collection_equality( tensors1: ContainerAnyTensors, tensors2: ContainerAnyTensors, atol: float = DEFAULT_ABSOLUTE_TOLERANCE, test_env: Optional[tf.test.TestCase] = None, log_debug_path: bool = True, ) -> bool: """ Compares two collections of tensors. Tensors can either be numpy arrays or tensorflow tensors. The collection should all be of the same type and should not mix numpy/tensorflow. Mixing the two in one collection will error out. Both collections should also be consistent in either being numpy arrays or tensorflow tensors. An assertion error will happen if the two collections are inconsistent type wise. If the tensors are graph tensors then test_env is required. By default if this fails it will print out information about the mismatch. log_debug_path can be disabled if you not want any error messages and just need the return value. Dense and sparse tensors are both suported. Sparse tensors will all be converted to dense tensors. This does not check whether the two collections are consistently sparse or dense. """ assert check_container_tensors(tensors1) == check_container_tensors(tensors2) if check_container_tensors(tensors1): arrays1 = evaluate_tensors(tensors1, test_env) else: arrays1 = cast(ContainerArrays, tensors1) if check_container_tensors(tensors2): arrays2 = evaluate_tensors(tensors2, test_env) else: arrays2 = cast(ContainerArrays, tensors2) result, debug_path = _tensor_collection_equality(arrays1, arrays2, atol=atol) if not result and log_debug_path: assert debug_path is not None _log_tensor_equality_mismatch(arrays1, arrays2, debug_path) return result
[ "numpy.all", "numpy.isclose", "math.isclose", "numpy.absolute", "numpy.issubdtype", "numpy.array", "tensorflow.constant", "functools.partial", "tensorflow.sparse.to_dense", "typing.cast", "typing.TypeVar" ]
[((298, 311), 'typing.TypeVar', 'TypeVar', (['"""T1"""'], {}), "('T1')\n", (305, 311), False, 'from typing import Callable, Iterable, List, Mapping, Optional, Sequence, Tuple, TypeVar, Union, cast\n'), ((317, 330), 'typing.TypeVar', 'TypeVar', (['"""T2"""'], {}), "('T2')\n", (324, 330), False, 'from typing import Callable, Iterable, List, Mapping, Optional, Sequence, Tuple, TypeVar, Union, cast\n'), ((590, 655), 'typing.TypeVar', 'TypeVar', (['"""ContainerAnyTensors"""', 'ContainerTensors', 'ContainerArrays'], {}), "('ContainerAnyTensors', ContainerTensors, ContainerArrays)\n", (597, 655), False, 'from typing import Callable, Iterable, List, Mapping, Optional, Sequence, Tuple, TypeVar, Union, cast\n'), ((4298, 4314), 'typing.cast', 'cast', (['T1', 'result'], {}), '(T1, result)\n', (4302, 4314), False, 'from typing import Callable, Iterable, List, Mapping, Optional, Sequence, Tuple, TypeVar, Union, cast\n'), ((1439, 1471), 'numpy.issubdtype', 'np.issubdtype', (['dtype', 'np.integer'], {}), '(dtype, np.integer)\n', (1452, 1471), True, 'import numpy as np\n'), ((2361, 2388), 'tensorflow.sparse.to_dense', 'tf.sparse.to_dense', (['tensor1'], {}), '(tensor1)\n', (2379, 2388), True, 'import tensorflow as tf\n'), ((2407, 2434), 'tensorflow.sparse.to_dense', 'tf.sparse.to_dense', (['tensor2'], {}), '(tensor2)\n', (2425, 2434), True, 'import tensorflow as tf\n'), ((2517, 2561), 'math.isclose', 'math.isclose', (['tensor1', 'tensor2'], {'abs_tol': 'atol'}), '(tensor1, tensor2, abs_tol=atol)\n', (2529, 2561), False, 'import math\n'), ((3843, 3896), 'numpy.isclose', 'np.isclose', (['array1', 'array2'], {'atol': 'atol', 'equal_nan': '(True)'}), '(array1, array2, atol=atol, equal_nan=True)\n', (3853, 3896), True, 'import numpy as np\n'), ((7818, 7846), 'numpy.absolute', 'np.absolute', (['(array1 - array2)'], {}), '(array1 - array2)\n', (7829, 7846), True, 'import numpy as np\n'), ((9807, 9838), 'typing.cast', 'cast', (['ContainerArrays', 'tensors1'], {}), '(ContainerArrays, tensors1)\n', (9811, 9838), False, 'from typing import Callable, Iterable, List, Mapping, Optional, Sequence, Tuple, TypeVar, Union, cast\n'), ((9965, 9996), 'typing.cast', 'cast', (['ContainerArrays', 'tensors2'], {}), '(ContainerArrays, tensors2)\n', (9969, 9996), False, 'from typing import Callable, Iterable, List, Mapping, Optional, Sequence, Tuple, TypeVar, Union, cast\n'), ((2670, 2690), 'tensorflow.constant', 'tf.constant', (['tensor1'], {}), '(tensor1)\n', (2681, 2690), True, 'import tensorflow as tf\n'), ((2727, 2744), 'numpy.array', 'np.array', (['tensor1'], {}), '(tensor1)\n', (2735, 2744), True, 'import numpy as np\n'), ((2853, 2873), 'tensorflow.constant', 'tf.constant', (['tensor2'], {}), '(tensor2)\n', (2864, 2873), True, 'import tensorflow as tf\n'), ((2910, 2927), 'numpy.array', 'np.array', (['tensor2'], {}), '(tensor2)\n', (2918, 2927), True, 'import numpy as np\n'), ((3785, 3809), 'numpy.all', 'np.all', (['(array1 == array2)'], {}), '(array1 == array2)\n', (3791, 3809), True, 'import numpy as np\n'), ((1110, 1136), 'functools.partial', 'partial', (['container_fmap', 'f'], {}), '(container_fmap, f)\n', (1117, 1136), False, 'from functools import partial\n')]
# threads.py # import time import pickle from datetime import datetime import numpy as np from PyQt5.QtCore import QThread from .. import __version__ from ..fly import Fly from ..ts import FixedCourtshipTrackingSummary from .tracking import * class TrackingThread(QThread): """Worker thread to run tracking algorithm. Parameters ---------- video_settings : list of TrackingSettings Each TrackingSettings object should be completely filled before creating an instance of this object, and running this thread. logger : QTextEdit To store information about video being currently tracked. This may need to be changed to prevent "QObject::connect: Cannot queue arguments of type 'QTextBlock' from being displayed. """ def __init__(self, video_settings, logger, progress, parent=None): super(TrackingThread, self).__init__(parent) self.video_settings = video_settings self.logger = logger self.tracking_progress = progress def run(self): for ix in xrange(len(self.video_settings)): start_time = time.time() settings = self.video_settings[ix] video = settings.video n_frames = video.get_n_frames() timestamps = video.get_all_timestamps() fps = (1. / np.mean(np.diff(timestamps))) male = Fly() female = Fly() tracking_summary = FixedCourtshipTrackingSummary() male.init_params(n_frames) female.init_params(n_frames) male.timestamps = timestamps female.timestamps = timestamps # load video attributes into FixedCourtshipTrackingSummary tracking_summary.video.filename = settings.video_file tracking_summary.video.timestamps = timestamps tracking_summary.video.fps = fps tracking_summary.video.duration_frames = n_frames tracking_summary.video.duration_seconds = (n_frames * 1.) / fps tracking_summary.video.start_time = datetime.fromtimestamp( timestamps[0]).strftime('%Y-%m-%d %H:%M:%S') tracking_summary.video.end_time = datetime.fromtimestamp( timestamps[-1]).strftime('%Y-%m-%d %H:%M:%S') tracking_summary.video.pixels_per_mm = settings.arena.pixels_to_mm # load arena attributes into FixedCourtshipTrackingSummary tracking_summary.arena.shape = 'circular' tracking_summary.arena.center_pixel_rr = settings.arena.center[0] tracking_summary.arena.center_pixel_cc = settings.arena.center[1] tracking_summary.arena.radius_mm = settings.arena.radius tracking_summary.arena.diameter_mm = 2 * settings.arena.radius # load software attributes into FixedCourtshipTrackingSummary tracking_summary.software.tight_threshold = settings.tight_threshold tracking_summary.software.loose_threshold = settings.loose_threshold tracking_summary.software.date_tracked = datetime.today( ).strftime('%Y-%m-%d %H:%M:%S') tracking_summary.software.version = __version__ # set the `group` attribute for the FixedCourtshipTrackingSummary tracking_summary.group = settings.group self.logger.append( 'Tracking started for video: {} \nStart Time: {}'.format( settings.video_file, time.strftime('%H:%M:%S', time.localtime(start_time)) )) # get the location and region properties that define # the fixed female. f_props, f_head, f_rear = find_female( image=settings.arena.background_image, female=settings.female, lp_threshold=settings.tight_threshold ) # update female based on props we just found -- # this ensures that the ellipse used to mask the female is # not biased by variation in user-defined ellipses. tighten_female_ellipse( female=settings.female, female_props=f_props ) # loop through each frame in the video, and find the male. for frame_ix in xrange(n_frames): frame_ix = long(frame_ix) frame, ts = video.get_frame(frame_ix) try: male_props = find_male( image=frame, female=settings.female, arena=settings.arena, lp_threshold=settings.tight_threshold ) except NoPropsDetected as NPD: self.logger.append( '\t' + NPD.message + ' Body @ frame {}'.format(frame_ix) ) continue wing_props = find_wings( image=frame, female=settings.female, arena=settings.arena, male_props=male_props, loose_threshold=settings.loose_threshold, logger=self.logger, frame_ix=frame_ix ) # if wing_props is None: # # male.timestamps[frame_ix] = ts # # female.timestamps[frame_ix] = ts # continue male.body.centroid.row[frame_ix] = male_props.centroid[0] male.body.centroid.col[frame_ix] = male_props.centroid[1] male.body.orientation[frame_ix] = male_props.orientation set_male_props(male, wing_props, frame_ix) set_female_props(female, f_props, f_head, f_rear, frame_ix) percent_complete = (frame_ix + 1.) / n_frames * 100 self.tracking_progress.emit( percent_complete, 'Tracking video {}/{}.'.format( ix + 1, len(self.video_settings))) # update the tracking settings dictionary with male and female items. # tracking_settings.update({'male': male, 'female': female}) # tracking_summary.set_attributes(**tracking_settings) tracking_summary.male = male tracking_summary.female = female save_file = settings.save_file save_type = save_file.split('.')[-1] if save_type == 'xlsx': tracking_summary.to_xlsx(save_file) elif save_type == 'fcts': with open(save_file, 'wb') as SAVE: pickle.dump(tracking_summary, SAVE) end_time = time.time() elapsed_time = end_time - start_time time_hrs = int(elapsed_time / 3600) time_mins = int((elapsed_time - time_hrs * 3600) / 60) time_secs = int(elapsed_time - time_hrs * 3600 - time_mins * 60) self.logger.append( 'End Time: {}\nTotal Time Elapse: {}'.format( time.strftime('%H:%M:%S', time.localtime(end_time)), '{:02d}:{:02d}:{:02d}'.format( time_hrs, time_mins, time_secs)) ) self.logger.append('TRACKING COMPLETE')
[ "datetime.datetime.fromtimestamp", "pickle.dump", "numpy.diff", "datetime.datetime.today", "time.localtime", "time.time" ]
[((1170, 1181), 'time.time', 'time.time', ([], {}), '()\n', (1179, 1181), False, 'import time\n'), ((7048, 7059), 'time.time', 'time.time', ([], {}), '()\n', (7057, 7059), False, 'import time\n'), ((1399, 1418), 'numpy.diff', 'np.diff', (['timestamps'], {}), '(timestamps)\n', (1406, 1418), True, 'import numpy as np\n'), ((2147, 2184), 'datetime.datetime.fromtimestamp', 'datetime.fromtimestamp', (['timestamps[0]'], {}), '(timestamps[0])\n', (2169, 2184), False, 'from datetime import datetime\n'), ((2280, 2318), 'datetime.datetime.fromtimestamp', 'datetime.fromtimestamp', (['timestamps[-1]'], {}), '(timestamps[-1])\n', (2302, 2318), False, 'from datetime import datetime\n'), ((3175, 3191), 'datetime.datetime.today', 'datetime.today', ([], {}), '()\n', (3189, 3191), False, 'from datetime import datetime\n'), ((3634, 3660), 'time.localtime', 'time.localtime', (['start_time'], {}), '(start_time)\n', (3648, 3660), False, 'import time\n'), ((6986, 7021), 'pickle.dump', 'pickle.dump', (['tracking_summary', 'SAVE'], {}), '(tracking_summary, SAVE)\n', (6997, 7021), False, 'import pickle\n'), ((7450, 7474), 'time.localtime', 'time.localtime', (['end_time'], {}), '(end_time)\n', (7464, 7474), False, 'import time\n')]
""" To run this function, we need to change her cprand function because it doesn't return exact error for now """ import numpy as np import tensorly as tl from src._base import random_init_fac from src._comparaison import init_factors from src._hercprand import her_CPRAND import copy def test_nbrestart(i): """ Compute the nb restart for 4 cases : 1 - simple case with exact error 2 - simple case with estimated error 3 - complicated case with exact error 4 - complicated case with estimated error In each case, we simulate nb_rand noised I*J*K rank r random tensors. For each tensor, we do 5 factor matrices initializations. Parameters ---------- i : int choice of case. Returns ------- float mean value of nbrestart """ I=50 J=50 K=50 r=10 # rank n_samples=int(10*r*np.log(r)+1) # nb of randomized samples nb_rand=10 # nb of random initialization list_pct=[] for i in range(nb_rand) : # Random initialization of a noised cp_tensor np.random.seed(i) if i==1 or i==2: fac_true,noise=init_factors(I,J,K,r,scale=False) else : fac_true,noise=init_factors(I,J,K,r,scale=True) t=tl.cp_to_tensor((None,fac_true))+noise for j in range(5): factors=random_init_fac(t,r) if i==1 : # simple case with exact error weights,factors,it,error,_,pct=her_CPRAND(t,r,n_samples,factors=copy.deepcopy(factors),exact_err=True,it_max=500,err_it_max=100) list_pct.append(pct) if i==2: # simple case with estimated error weights,factors,it,_,error,pct=her_CPRAND(t,r,n_samples,factors=copy.deepcopy(factors),exact_err=False,it_max=500,err_it_max=100) list_pct.append(pct) if i==3: # complicated case with exact error weights,factors,it,error,_,pct=her_CPRAND(t,r,n_samples,factors=copy.deepcopy(factors),exact_err=True,it_max=500,err_it_max=100) list_pct.append(pct) if i==4: # complicated case with estimated error weights,factors,it,_,error,pct=her_CPRAND(t,r,n_samples,factors=copy.deepcopy(factors),exact_err=False,it_max=500,err_it_max=400) list_pct.append(pct) return(np.mean(list_pct))
[ "numpy.mean", "src._base.random_init_fac", "numpy.log", "numpy.random.seed", "copy.deepcopy", "tensorly.cp_to_tensor", "src._comparaison.init_factors" ]
[((2404, 2421), 'numpy.mean', 'np.mean', (['list_pct'], {}), '(list_pct)\n', (2411, 2421), True, 'import numpy as np\n'), ((1073, 1090), 'numpy.random.seed', 'np.random.seed', (['i'], {}), '(i)\n', (1087, 1090), True, 'import numpy as np\n'), ((1143, 1180), 'src._comparaison.init_factors', 'init_factors', (['I', 'J', 'K', 'r'], {'scale': '(False)'}), '(I, J, K, r, scale=False)\n', (1155, 1180), False, 'from src._comparaison import init_factors\n'), ((1219, 1255), 'src._comparaison.init_factors', 'init_factors', (['I', 'J', 'K', 'r'], {'scale': '(True)'}), '(I, J, K, r, scale=True)\n', (1231, 1255), False, 'from src._comparaison import init_factors\n'), ((1262, 1295), 'tensorly.cp_to_tensor', 'tl.cp_to_tensor', (['(None, fac_true)'], {}), '((None, fac_true))\n', (1277, 1295), True, 'import tensorly as tl\n'), ((1348, 1369), 'src._base.random_init_fac', 'random_init_fac', (['t', 'r'], {}), '(t, r)\n', (1363, 1369), False, 'from src._base import random_init_fac\n'), ((878, 887), 'numpy.log', 'np.log', (['r'], {}), '(r)\n', (884, 887), True, 'import numpy as np\n'), ((1518, 1540), 'copy.deepcopy', 'copy.deepcopy', (['factors'], {}), '(factors)\n', (1531, 1540), False, 'import copy\n'), ((1773, 1795), 'copy.deepcopy', 'copy.deepcopy', (['factors'], {}), '(factors)\n', (1786, 1795), False, 'import copy\n'), ((2030, 2052), 'copy.deepcopy', 'copy.deepcopy', (['factors'], {}), '(factors)\n', (2043, 2052), False, 'import copy\n'), ((2290, 2312), 'copy.deepcopy', 'copy.deepcopy', (['factors'], {}), '(factors)\n', (2303, 2312), False, 'import copy\n')]
from decouple import config import sys, os from .analyze.analyzeSignal import calcPowers from .analyze.pereiraChangeOfMean import pereiraLikelihood, getChangePoints, cleanLikelihoods from .websocket import wsManager from . import data as dataHp import json import numpy as np from django.http import JsonResponse def indicesInDoubleArray(array2, value, thres): index1 = -1 index2 = -1 minDist = float("inf") for i, array in enumerate(array2): for j, val in enumerate(array): dist = abs(val - value) if dist < thres and dist < minDist: minDist = dist index1 = i index2 = j return index1, index2 def findEvents(power, thres, pre, post, voting, minDist, m): likelihoods = pereiraLikelihood(power, threshold=thres, preEventLength=pre, postEventLength=post, linearFactor=m, verbose=True) # likelihoods = cleanLikelihoods(likelihoods, 5*threshold) # Get change indices changeIndices = getChangePoints(power, likelihoods, windowSize=voting, minDist=minDist) return changeIndices def findUniqueStates(power, changeIndices, thres, minDist): LINE_NOISE = 1.0 # Get State Seuence from all state changes # Handle start state stateSequence = [{'index': 0, 'endIndex': changeIndices[0] if len(changeIndices) > 0 else len(power)}] # Changes in between for i, change in enumerate(changeIndices[:-1]): stateSequence.append({'index': change, 'endIndex': changeIndices[i+1]}) # handle end state if len(changeIndices) > 0: stateSequence.append({'index': changeIndices[-1], 'endIndex': len(power)-1}) # Get Steady states point after each state change for i in range(len(stateSequence)): slice = power[ stateSequence[i]['index'] : stateSequence[i]['endIndex'] ] stateSequence[i]['ssIndex'] = int(stateSequence[i]['index']+minDist ) stateSequence[i]['ssEndIndex'] = int(max(stateSequence[i]['endIndex']-minDist/2, stateSequence[i]['ssIndex']+1)) # Construct mean value of state for i in range(len(stateSequence)): if stateSequence[i]['ssIndex'] is None or stateSequence[i]['ssEndIndex'] is None or stateSequence[i]['ssEndIndex'] - stateSequence[i]['ssIndex'] < 1: stateSequence[i]['mean'] = None else: stateSequence[i]['mean'] = np.mean(power[stateSequence[i]['ssIndex']:stateSequence[i]['ssEndIndex']]) if stateSequence[i]['mean'] <= LINE_NOISE: stateSequence[i]['mean'] = 0 means = sorted([stateSequence[i]['mean'] for i in range(len(stateSequence))]) print(means) cluster = 0 clusters = [0] # lastMean = means[0] # for i in range(1, len(means)): # if abs(lastMean-means[i]) > thres: # lastMean = means[i] # cluster += 1 # # lastMean = np.mean(np.array([means[i], lastMean])) # clusters.append(cluster) for i in range(1, len(means)): if abs(means[i-1]-means[i]) > thres: cluster += 1 clusters.append(cluster) for i in range(len(stateSequence)): stateSequence[i]["stateID"] = clusters[means.index(stateSequence[i]['mean'])] # prevent Self loops # if len(stateSequence) > 1: # newStateSequence = [] # source = stateSequence[0] # for i in range(len(stateSequence)-1): # dest = stateSequence[i+1] # if source["stateID"] == dest["stateID"]: # source['endIndex'] = dest["endIndex"] # source['ssEndIndex'] = dest["ssEndIndex"] # #recalculate mean based on the length of the arrays # source['mean'] = (source['mean'] * (source['endIndex'] - source['index']) + dest["mean"] * (dest['endIndex'] - dest['index']))/(dest['endIndex'] - source['index']) # else: # newStateSequence.append(source) # if dest == stateSequence[-1]: # newStateSequence.append(dest) # source = dest # stateSequence = newStateSequence return stateSequence def autoLabel(request): if request.method != "POST": Http404 response = {} data = json.loads(request.body) parameter = data["parameter"] sessionID = request.session.session_key sessionData = request.session.get('dataInfo', {}) if sessionData["type"] == "fired": wsManager.sendStatus(sessionID, "Loading 50Hz power data...", percent=10) dataDict = dataHp.getSessionData(sessionID, sessionData) usablePower = ["s", "s_l1", "p", "p_l1"] usableKeys = list(set(usablePower) & set(dataDict["measures"])) if len(usableKeys) < 1: if "v" in dataDict["measures"] and "i" in dataDict["measures"]: pass p,q,s = calcPowers(dataDict["data"]["v"], dataDict["data"]["i"], dataDict["samplingrate"]) power = s response["msg"] = "Calculated apparent power using Current and Voltage" else: response["msg"] = "Could not find power, or voltage and current in data. Name it as \"p\",\"s\" or \"v\",\"i\".\n" response["msg"] += "If you have electricity data of multiple supply legs, name it as \"<measure>_l1\", \"<measure>_l2\", ... accordingly." return JsonResponse(response) else: power = list(dataDict["data"][sorted(usableKeys)[-1]]) sr = dataDict["samplingrate"] # We only do this at a max samplingrate of 50 Hz if sr > 50: wsManager.sendStatus(sessionID, text="Resampling to 50Hz...", percent=15) power, timestamps = dataHp.resampleDict(dataDict, sorted(usableKeys)[-1], 50, forceEvenRate=True) #power = dataHp.resample(power, sr, 50) # print(power) sr = 50 newSr = None if "sr" in parameter: newSr = float(parameter["sr"]) if newSr != None and newSr != -1 or "ts" in dataDict: if "ts" in dataDict and newSr is None: newSr = max(1/3.0, dataDict["samplingrate"]) wsManager.sendStatus(sessionID, text="Resampling to "+ str(round(newSr, 2)) + "Hz...", percent=17) power, timestamps = dataHp.resampleDict(dataDict, sorted(usableKeys)[-1], newSr, forceEvenRate=True) #power = dataHp.resample(power, sr, newSr) sr = newSr thres = 5.0 if "thres" in parameter: thres = float(parameter["thres"]) thres = max(thres, 0.1) pre = 1.0*sr if "pre" in parameter: pre = int(float(parameter["pre"])*sr) pre = max(pre, 2) post = 1.0*sr if "post" in parameter: post = int(float(parameter["post"])*sr) post = max(post, 2) voting = 2.0*sr if "voting" in parameter: voting = int(float(parameter["voting"])*sr) voting = max(voting, 1) minDist = 1.0*sr if "minDist" in parameter: minDist = int(float(parameter["minDist"])*sr) minDist = max(minDist, 1) m = 0.005 if "linearCoeff" in parameter: m = float(parameter["linearCoeff"]) print("sr: {}Hz, thres: {}W, pre: {}samples, post: {}:samples, voting: {}samples, minDist: {} samples, m:{}".format(sr, thres, pre, post, voting, minDist, m), flush=True) wsManager.sendStatus(sessionID, "Finding Events...", percent=20) changeIndices = findEvents(power, thres, pre, post, voting, minDist, m) wsManager.sendStatus(sessionID, "Clustering Events...", percent=70) stateSequence = findUniqueStates(power, changeIndices, thres, minDist) if len(changeIndices) == 0: response["msg"] = "No Changes found in signal..." if len(changeIndices) >= 200: response["msg"] = "Too many events found, you may want to change settings" changeIndices = [] wsManager.sendStatus(sessionID, "Generating Labels...") # Convert change indices to timestamps ts = 0 if "timestamp" in dataDict: ts = dataDict["timestamp"] # labels = [{"startTs": ts+(float(i/sr)), "label":""} for i in changeIndices] labels = [{"startTs": ts+(float(i["index"]/sr)), "label":"S" + str(i["stateID"])} for i in stateSequence] response["labels"] = labels return JsonResponse(response)
[ "numpy.mean", "json.loads", "django.http.JsonResponse" ]
[((4216, 4240), 'json.loads', 'json.loads', (['request.body'], {}), '(request.body)\n', (4226, 4240), False, 'import json\n'), ((8077, 8099), 'django.http.JsonResponse', 'JsonResponse', (['response'], {}), '(response)\n', (8089, 8099), False, 'from django.http import JsonResponse\n'), ((2371, 2445), 'numpy.mean', 'np.mean', (["power[stateSequence[i]['ssIndex']:stateSequence[i]['ssEndIndex']]"], {}), "(power[stateSequence[i]['ssIndex']:stateSequence[i]['ssEndIndex']])\n", (2378, 2445), True, 'import numpy as np\n'), ((5311, 5333), 'django.http.JsonResponse', 'JsonResponse', (['response'], {}), '(response)\n', (5323, 5333), False, 'from django.http import JsonResponse\n')]
""" Simplistic implementation of the two-layer neural network. Training method is stochastic (online) gradient descent with momentum. As an example it computes XOR for given input. """ import numpy import time class Dataset: """ TODO: docstring """ def __init__(self, n_in, n_samples): """ TODO: docstring """ self.X = numpy.random.binomial(1, 0.5, (n_samples, n_in)) self.T = self.X^1 class NeuralNet: """ Details: - tanh activation for hidden layer - sigmoid activation for output layer - cross-entropy loss """ def __init__(self, dataset): """ TODO: docstring """ learning_rate = 0.01 momentum = 0.9 n_hidden = 10 n_in = 10 n_out = 10 numpy.random.seed(0) self.V = numpy.random.normal(scale=0.1, size=(n_in, n_hidden)) self.W = numpy.random.normal(scale=0.1, size=(n_hidden, n_out)) self.bv = numpy.zeros(n_hidden) self.bw = numpy.zeros(n_out) params = [self.V, self.W, self.bv, self.bw] X, T = dataset.X, dataset.T # train for epoch in range(100): err, upd = [], [0] * len(params) t0 = time.clock() for i in range(X.shape[0]): loss, grad = self.update(X[i], T[i], *params) for j in range(len(params)): params[j] -= upd[j] for j in range(len(params)): upd[j] = learning_rate * grad[j] + momentum * upd[j] err.append(loss) print('Epoch: {:d}, Loss: {:.8f}, Time: {:.4f}s'.format( epoch, numpy.mean(err), time.clock() - t0)) def __call__(self, x): """ make predictions """ A = numpy.dot(x, self.V) + self.bv B = numpy.dot(numpy.tanh(A), self.W) + self.bw return (self.sigmoid(B) > 0.5).astype(int) def sigmoid(self, x): """ TODO: docstring """ return 1.0 / (1.0 + numpy.exp(-x)) def tanh_prime(self, x): """ TODO: docstring """ return 1 - numpy.tanh(x)**2 def update(self, x, t, V, W, bv, bw): """ use error for each layer as a gradient for biases """ # forward A = numpy.dot(x, V) + bv Z = numpy.tanh(A) B = numpy.dot(Z, W) + bw Y = self.sigmoid(B) # backward Ew = Y - t Ev = self.tanh_prime(A) * numpy.dot(W, Ew) dW = numpy.outer(Z, Ew) dV = numpy.outer(x, Ev) loss = -numpy.mean (t * numpy.log(Y) + (1 - t) * numpy.log(1 - Y)) return loss, (dV, dW, Ev, Ew) if __name__ == '__main__': dataset = Dataset(10, 300) neural_net = NeuralNet(dataset) prediction = neural_net(numpy.random.binomial(1, 0.5, 10)) print('XOR Prediction:', prediction)
[ "numpy.random.normal", "numpy.mean", "time.clock", "numpy.log", "numpy.tanh", "numpy.exp", "numpy.zeros", "numpy.outer", "numpy.dot", "numpy.random.seed", "numpy.random.binomial" ]
[((369, 417), 'numpy.random.binomial', 'numpy.random.binomial', (['(1)', '(0.5)', '(n_samples, n_in)'], {}), '(1, 0.5, (n_samples, n_in))\n', (390, 417), False, 'import numpy\n'), ((809, 829), 'numpy.random.seed', 'numpy.random.seed', (['(0)'], {}), '(0)\n', (826, 829), False, 'import numpy\n'), ((847, 900), 'numpy.random.normal', 'numpy.random.normal', ([], {'scale': '(0.1)', 'size': '(n_in, n_hidden)'}), '(scale=0.1, size=(n_in, n_hidden))\n', (866, 900), False, 'import numpy\n'), ((918, 972), 'numpy.random.normal', 'numpy.random.normal', ([], {'scale': '(0.1)', 'size': '(n_hidden, n_out)'}), '(scale=0.1, size=(n_hidden, n_out))\n', (937, 972), False, 'import numpy\n'), ((991, 1012), 'numpy.zeros', 'numpy.zeros', (['n_hidden'], {}), '(n_hidden)\n', (1002, 1012), False, 'import numpy\n'), ((1031, 1049), 'numpy.zeros', 'numpy.zeros', (['n_out'], {}), '(n_out)\n', (1042, 1049), False, 'import numpy\n'), ((2387, 2400), 'numpy.tanh', 'numpy.tanh', (['A'], {}), '(A)\n', (2397, 2400), False, 'import numpy\n'), ((2564, 2582), 'numpy.outer', 'numpy.outer', (['Z', 'Ew'], {}), '(Z, Ew)\n', (2575, 2582), False, 'import numpy\n'), ((2596, 2614), 'numpy.outer', 'numpy.outer', (['x', 'Ev'], {}), '(x, Ev)\n', (2607, 2614), False, 'import numpy\n'), ((2852, 2885), 'numpy.random.binomial', 'numpy.random.binomial', (['(1)', '(0.5)', '(10)'], {}), '(1, 0.5, 10)\n', (2873, 2885), False, 'import numpy\n'), ((1249, 1261), 'time.clock', 'time.clock', ([], {}), '()\n', (1259, 1261), False, 'import time\n'), ((1826, 1846), 'numpy.dot', 'numpy.dot', (['x', 'self.V'], {}), '(x, self.V)\n', (1835, 1846), False, 'import numpy\n'), ((2354, 2369), 'numpy.dot', 'numpy.dot', (['x', 'V'], {}), '(x, V)\n', (2363, 2369), False, 'import numpy\n'), ((2413, 2428), 'numpy.dot', 'numpy.dot', (['Z', 'W'], {}), '(Z, W)\n', (2422, 2428), False, 'import numpy\n'), ((2534, 2550), 'numpy.dot', 'numpy.dot', (['W', 'Ew'], {}), '(W, Ew)\n', (2543, 2550), False, 'import numpy\n'), ((1879, 1892), 'numpy.tanh', 'numpy.tanh', (['A'], {}), '(A)\n', (1889, 1892), False, 'import numpy\n'), ((2070, 2083), 'numpy.exp', 'numpy.exp', (['(-x)'], {}), '(-x)\n', (2079, 2083), False, 'import numpy\n'), ((2182, 2195), 'numpy.tanh', 'numpy.tanh', (['x'], {}), '(x)\n', (2192, 2195), False, 'import numpy\n'), ((1700, 1715), 'numpy.mean', 'numpy.mean', (['err'], {}), '(err)\n', (1710, 1715), False, 'import numpy\n'), ((2647, 2659), 'numpy.log', 'numpy.log', (['Y'], {}), '(Y)\n', (2656, 2659), False, 'import numpy\n'), ((2672, 2688), 'numpy.log', 'numpy.log', (['(1 - Y)'], {}), '(1 - Y)\n', (2681, 2688), False, 'import numpy\n'), ((1717, 1729), 'time.clock', 'time.clock', ([], {}), '()\n', (1727, 1729), False, 'import time\n')]
import os import numpy as np from PIL import Image import cv2 import pickle face_cascade = cv2.CascadeClassifier("cascade.xml") recognizer = cv2.face.LBPHFaceRecognizer_create() BASE_DIR = os.path.dirname(os.path.abspath(__file__)) image_dir = os.path.join(BASE_DIR, "images") current = 0 names_id_mapping = {} train = [] names = [] for root, dirs, files in os.walk(image_dir): for file in files: if file.endswith("png") or file.endswith("jpg"): path = os.path.join(root, file) label = os.path.basename(os.path.dirname(path)).replace(" ", "-").lower() # print(f"{label} => ${path}") if label in names_id_mapping: pass else: names_id_mapping[label] = current current += 1 id_ = names_id_mapping[label] pillow_image = Image.open(path).convert("L") image_array = np.array(pillow_image, "uint8") faces = face_cascade.detectMultiScale(image_array, scaleFactor=1.5, minNeighbors=5) for (x, y, w, h) in faces: region_of_image = image_array[y: y+h, x: x+w] train.append(region_of_image) names.append(id_) with open("labels.pickle", "wb") as file: pickle.dump(names_id_mapping, file) recognizer.train(train, np.array(names)) recognizer.save("trainer.yml")
[ "PIL.Image.open", "pickle.dump", "os.path.join", "cv2.face.LBPHFaceRecognizer_create", "numpy.array", "os.path.dirname", "os.path.abspath", "cv2.CascadeClassifier", "os.walk" ]
[((92, 128), 'cv2.CascadeClassifier', 'cv2.CascadeClassifier', (['"""cascade.xml"""'], {}), "('cascade.xml')\n", (113, 128), False, 'import cv2\n'), ((142, 178), 'cv2.face.LBPHFaceRecognizer_create', 'cv2.face.LBPHFaceRecognizer_create', ([], {}), '()\n', (176, 178), False, 'import cv2\n'), ((246, 278), 'os.path.join', 'os.path.join', (['BASE_DIR', '"""images"""'], {}), "(BASE_DIR, 'images')\n", (258, 278), False, 'import os\n'), ((362, 380), 'os.walk', 'os.walk', (['image_dir'], {}), '(image_dir)\n', (369, 380), False, 'import os\n'), ((207, 232), 'os.path.abspath', 'os.path.abspath', (['__file__'], {}), '(__file__)\n', (222, 232), False, 'import os\n'), ((1278, 1313), 'pickle.dump', 'pickle.dump', (['names_id_mapping', 'file'], {}), '(names_id_mapping, file)\n', (1289, 1313), False, 'import pickle\n'), ((1339, 1354), 'numpy.array', 'np.array', (['names'], {}), '(names)\n', (1347, 1354), True, 'import numpy as np\n'), ((481, 505), 'os.path.join', 'os.path.join', (['root', 'file'], {}), '(root, file)\n', (493, 505), False, 'import os\n'), ((921, 952), 'numpy.array', 'np.array', (['pillow_image', '"""uint8"""'], {}), "(pillow_image, 'uint8')\n", (929, 952), True, 'import numpy as np\n'), ((865, 881), 'PIL.Image.open', 'Image.open', (['path'], {}), '(path)\n', (875, 881), False, 'from PIL import Image\n'), ((543, 564), 'os.path.dirname', 'os.path.dirname', (['path'], {}), '(path)\n', (558, 564), False, 'import os\n')]
#/usr/bin/python! ##### # unmix_spectra.py performs fully-constrained least-squares # spectral ummixing on an image file # # unmixing implementation described here: # http://pysptools.sourceforge.net/abundance_maps.html # # based on the algorithm desribed here: # <NAME>, <NAME>, and <NAME>. Fully Constrained # Least-Squares Based Linear Unmixing. Althouse. IEEE. 1999. # # c. 2016 <NAME> ##### import os import sys import aei import random import gdal as gdal import numpy as np import pysptools.abundance_maps as abm # create a class to parse out the arguments passed to the main function class parse_args: def __init__(self, arglist): # set up main variables and defaults to parse self.infile = '' self.outfile = '' self.spectral_libs = [] self.n = 20 self.bands = [] self.of = 'GTiff' self.normalize = False # exit if no arguments passed if len(arglist) == 1: usage(exit=True) # read arguments from command line i = 1 while i < len(arglist): arg = arglist[i] # check input flag if arg.lower() == '-i': i += 1 arg = arglist[i] if type(arg) is str: self.infile = arg if not aei.fn.checkFile(self.infile, quiet = True): usage() aei.fn.checkFile(self.infile) sys.exit(1) # check output flag if arg.lower() == '-o': i += 1 arg = arglist[i] if type(arg) is str: self.outfile = arg outpath = os.path.dirname(self.outfile) if outpath == '': outpath = '.' if not os.access(outpath, os.W_OK): usage() print("[ ERROR ]: unable to write to output path: %s" % outpath) sys.exit(1) # check spectral library paths if arg.lower() == "-lib": i += 1 arg = arglist[i] libs = arg.split(" ") # throw an error if only one lib specified if len(libs) == 1: usage() print("[ ERROR ]: unable to unmix with one spectral library: %s" % libs[0]) sys.exit(1) # loop through each lib and update spec_lib list for j in range(len(libs)): if not aei.fn.checkFile(libs[j], quiet=True): usage() aei.fn.checkFile(libs[j]) sys.exit() self.spectral_libs.append(libs[j]) # check number of iterations if arg.lower() == "-n": i += 1 arg = arglist[i] try: self.n = int(arg) except ValueError: usage() print("[ ERROR ]: -n argument is not an integer: %s" % arg) sys.exit(1) # check indices if arg.lower() == "-bands": i += 1 arg = arglist[i] band = arg.split(" ") # loop through and make sure each is a number for j in range(len(band)): try: int(band[j]) except ValueError: usage() print("[ ERROR ]: invalid index set: %s" % ind[j]) sys.exit(1) self.bands.append(int(band[j])) # check normalize flag if arg.lower() == "-normalize": self.normalize = True # check output format if arg.lower() == "-of": i += 1 arg = arglist[i] import gdal_types as gdt if not arg in gdt.list(): usage() print("[ ERROR }: invalid output format: %s" % arg) sys.exit(1) self.of = arg i += 1 def usage(exit=False): """ describes the aeilas.py procedure in case of incorrect parameter calls syntax: usage() """ print( """ $ unmix_spectra.py -lib "lib1 lib2 ... libx" [-n n_random_selections] [-bands "band1 band2 ... bandx"] [-normalize] [-of output_format] [-i] input_file [-o] output_file """ ) if exit: sys.exit(1) def main(): """ the main program for unmix_spectra.py syntax: main() """ # parse the argument list args = parse_args(sys.argv) # load the spectral libraries lib = {} n_libs = len(args.spectral_libs) lnb = np.zeros(n_libs) # get info from each library for i in range(n_libs): # assign libraries to dictionary lib['lib_%s' % i] = aei.read.spectralLib(args.spectral_libs[i]) lnb[i] = (lib['lib_%s' % i].spectra.shape[-1]) # get random indices for each library lib['rnd_%s' % i] = random.sample(range( lib['lib_%s' % i].spectra.shape[0]), args.n) # normalize if set if args.normalize: lib['lib_%s' % i].bn(inds=args.bands) # load the input image and get parameters inf = gdal.Open(args.infile) ns = inf.RasterXSize nl = inf.RasterYSize nb = inf.RasterCount geo = inf.GetGeoTransform() prj = inf.GetProjection() n_bundles = lnb.shape[0] b1 = inf.GetRasterBand(1) # if bands were not set in command line, use all bands if not args.bands: args.bands = range(nb) # check no data param if hasattr(b1, 'GetNoDataValue'): nd = b1.GetNoDataValue() if nd is None: nd = -9999 else: nd = -9999 # check that the spectral libraries match the bands if not 0 in np.where((lnb-nb) != 0)[0].shape: print("[ ERROR ]: number of image bands does not match number of spectral library bands") inf = None sys.exit(1) # report a little print("[ STATUS ]: Beginning fully-constrained least-squares unmixing") print("[ STATUS ]: Input file : %s" % args.infile) print("[ STATUS ]: Output file: %s" % args.outfile) # create an output file ouf = gdal.GetDriverByName(args.of).Create( args.outfile, ns, nl, n_bundles, gdal.GDT_Float32) ouf.SetGeoTransform(geo) ouf.SetProjection(prj) # create an output array arr = np.zeros((nl, ns, n_bundles)) # read the image file and flip dimensions img = inf.ReadAsArray() img = img.transpose([1,2,0]) # find no data vals gd = np.where(img[:,:,0] != nd) # check that the file is not all no-data if gd[0].shape[0] == 0: print("[ ERROR ]: No good-data found in input file") print("[ ERROR ]: Exiting...") sys.exit(1) # subset the image array to good data only img = img[gd[0], gd[1], :] # normalize the data if set if args.normalize: img = aei.fn.bn(img, inds = args.bands) args.bands = range(len(args.bands)) # add a shallow dimension for unmixing algorithm img = np.expand_dims(img[:,args.bands], 0) # set up unmixing class unmixer = abm.FCLS() # loop through each random index and unmix for i in range(args.n): print("[ STATUS ]: Iteration [%s] of [%s]" % ((i + 1), args.n)) # set up the bundles bundles = np.zeros((n_libs,len(args.bands))) for j in range(n_libs): bundles[j,:] = lib['lib_%s' % j].spectra[lib['rnd_%s' % j][i],args.bands] # perform the unmixing arr[gd[0],gd[1],:] += unmixer.map(img, bundles).squeeze() # divide by n iterations to get average response arr /= args.n # report completion print("[ STATUS ]: Completed unmixing iterations") print("[ STATUS ]: Writing data to file") # and write the results to the output file for i in range(n_libs): band = ouf.GetRasterBand(i+1) band.WriteArray(arr[:,:,i]) band.SetNoDataValue(0.0) band.FlushCache() if __name__ == "__main__": main()
[ "aei.read.spectralLib", "gdal.Open", "aei.fn.checkFile", "gdal.GetDriverByName", "numpy.where", "os.access", "gdal_types.list", "os.path.dirname", "numpy.zeros", "numpy.expand_dims", "sys.exit", "aei.fn.bn", "pysptools.abundance_maps.FCLS" ]
[((5283, 5299), 'numpy.zeros', 'np.zeros', (['n_libs'], {}), '(n_libs)\n', (5291, 5299), True, 'import numpy as np\n'), ((5882, 5904), 'gdal.Open', 'gdal.Open', (['args.infile'], {}), '(args.infile)\n', (5891, 5904), True, 'import gdal as gdal\n'), ((7114, 7143), 'numpy.zeros', 'np.zeros', (['(nl, ns, n_bundles)'], {}), '((nl, ns, n_bundles))\n', (7122, 7143), True, 'import numpy as np\n'), ((7294, 7322), 'numpy.where', 'np.where', (['(img[:, :, 0] != nd)'], {}), '(img[:, :, 0] != nd)\n', (7302, 7322), True, 'import numpy as np\n'), ((7826, 7863), 'numpy.expand_dims', 'np.expand_dims', (['img[:, args.bands]', '(0)'], {}), '(img[:, args.bands], 0)\n', (7840, 7863), True, 'import numpy as np\n'), ((7910, 7920), 'pysptools.abundance_maps.FCLS', 'abm.FCLS', ([], {}), '()\n', (7918, 7920), True, 'import pysptools.abundance_maps as abm\n'), ((5015, 5026), 'sys.exit', 'sys.exit', (['(1)'], {}), '(1)\n', (5023, 5026), False, 'import sys\n'), ((5444, 5487), 'aei.read.spectralLib', 'aei.read.spectralLib', (['args.spectral_libs[i]'], {}), '(args.spectral_libs[i])\n', (5464, 5487), False, 'import aei\n'), ((6637, 6648), 'sys.exit', 'sys.exit', (['(1)'], {}), '(1)\n', (6645, 6648), False, 'import sys\n'), ((7507, 7518), 'sys.exit', 'sys.exit', (['(1)'], {}), '(1)\n', (7515, 7518), False, 'import sys\n'), ((7676, 7707), 'aei.fn.bn', 'aei.fn.bn', (['img'], {'inds': 'args.bands'}), '(img, inds=args.bands)\n', (7685, 7707), False, 'import aei\n'), ((6910, 6939), 'gdal.GetDriverByName', 'gdal.GetDriverByName', (['args.of'], {}), '(args.of)\n', (6930, 6939), True, 'import gdal as gdal\n'), ((1809, 1838), 'os.path.dirname', 'os.path.dirname', (['self.outfile'], {}), '(self.outfile)\n', (1824, 1838), False, 'import os\n'), ((2583, 2594), 'sys.exit', 'sys.exit', (['(1)'], {}), '(1)\n', (2591, 2594), False, 'import sys\n'), ((4533, 4544), 'sys.exit', 'sys.exit', (['(1)'], {}), '(1)\n', (4541, 4544), False, 'import sys\n'), ((6478, 6501), 'numpy.where', 'np.where', (['(lnb - nb != 0)'], {}), '(lnb - nb != 0)\n', (6486, 6501), True, 'import numpy as np\n'), ((1382, 1423), 'aei.fn.checkFile', 'aei.fn.checkFile', (['self.infile'], {'quiet': '(True)'}), '(self.infile, quiet=True)\n', (1398, 1423), False, 'import aei\n'), ((1483, 1512), 'aei.fn.checkFile', 'aei.fn.checkFile', (['self.infile'], {}), '(self.infile)\n', (1499, 1512), False, 'import aei\n'), ((1537, 1548), 'sys.exit', 'sys.exit', (['(1)'], {}), '(1)\n', (1545, 1548), False, 'import sys\n'), ((1942, 1969), 'os.access', 'os.access', (['outpath', 'os.W_OK'], {}), '(outpath, os.W_OK)\n', (1951, 1969), False, 'import os\n'), ((2116, 2127), 'sys.exit', 'sys.exit', (['(1)'], {}), '(1)\n', (2124, 2127), False, 'import sys\n'), ((2747, 2784), 'aei.fn.checkFile', 'aei.fn.checkFile', (['libs[j]'], {'quiet': '(True)'}), '(libs[j], quiet=True)\n', (2763, 2784), False, 'import aei\n'), ((2842, 2867), 'aei.fn.checkFile', 'aei.fn.checkFile', (['libs[j]'], {}), '(libs[j])\n', (2858, 2867), False, 'import aei\n'), ((2892, 2902), 'sys.exit', 'sys.exit', ([], {}), '()\n', (2900, 2902), False, 'import sys\n'), ((3376, 3387), 'sys.exit', 'sys.exit', (['(1)'], {}), '(1)\n', (3384, 3387), False, 'import sys\n'), ((4401, 4411), 'gdal_types.list', 'gdt.list', ([], {}), '()\n', (4409, 4411), True, 'import gdal_types as gdt\n'), ((3946, 3957), 'sys.exit', 'sys.exit', (['(1)'], {}), '(1)\n', (3954, 3957), False, 'import sys\n')]
import numpy as np from numpy import zeros from dipy.segment.threshold import upper_bound_by_percent, upper_bound_by_rate from numpy.testing import assert_equal, run_module_suite def test_adjustment(): imga = zeros([128, 128]) for y in range(128): for x in range(128): if y > 10 and y < 115 and x > 10 and x < 115: imga[x, y] = 100 if y > 39 and y < 88 and x > 39 and x < 88: imga[x, y] = 150 if y > 59 and y < 69 and x > 59 and x < 69: imga[x, y] = 255 high_1 = upper_bound_by_rate(imga) high_2 = upper_bound_by_percent(imga) vol1 = np.interp(imga, xp=[imga.min(), high_1], fp=[0, 255]) vol2 = np.interp(imga, xp=[imga.min(), high_2], fp=[0, 255]) count2 = (88 - 40) * (88 - 40) count1 = (114 - 10) * (114 - 10) count1_test = 0 count2_test = 0 count2_upper = (88 - 40) * (88 - 40) count1_upper = (114 - 10) * (114 - 10) count1_upper_test = 0 count2_upper_test = 0 value1 = np.unique(vol1) value2 = np.unique(vol2) for i in range(128): for j in range(128): if vol1[i][j] > value1[1]: count2_test = count2_test + 1 if vol1[i][j] > 0: count1_test = count1_test + 1 for i in range(128): for j in range(128): if vol2[i][j] > value2[1]: count2_upper_test = count2_upper_test + 1 if vol2[i][j] > 0: count1_upper_test = count1_upper_test + 1 assert_equal(count2, count2_test) assert_equal(count1, count1_test) assert_equal(count2_upper, count2_upper_test) assert_equal(count1_upper, count1_upper_test) if __name__ == '__main__': run_module_suite()
[ "numpy.unique", "numpy.testing.assert_equal", "dipy.segment.threshold.upper_bound_by_percent", "dipy.segment.threshold.upper_bound_by_rate", "numpy.zeros", "numpy.testing.run_module_suite" ]
[((216, 233), 'numpy.zeros', 'zeros', (['[128, 128]'], {}), '([128, 128])\n', (221, 233), False, 'from numpy import zeros\n'), ((572, 597), 'dipy.segment.threshold.upper_bound_by_rate', 'upper_bound_by_rate', (['imga'], {}), '(imga)\n', (591, 597), False, 'from dipy.segment.threshold import upper_bound_by_percent, upper_bound_by_rate\n'), ((611, 639), 'dipy.segment.threshold.upper_bound_by_percent', 'upper_bound_by_percent', (['imga'], {}), '(imga)\n', (633, 639), False, 'from dipy.segment.threshold import upper_bound_by_percent, upper_bound_by_rate\n'), ((1035, 1050), 'numpy.unique', 'np.unique', (['vol1'], {}), '(vol1)\n', (1044, 1050), True, 'import numpy as np\n'), ((1064, 1079), 'numpy.unique', 'np.unique', (['vol2'], {}), '(vol2)\n', (1073, 1079), True, 'import numpy as np\n'), ((1544, 1577), 'numpy.testing.assert_equal', 'assert_equal', (['count2', 'count2_test'], {}), '(count2, count2_test)\n', (1556, 1577), False, 'from numpy.testing import assert_equal, run_module_suite\n'), ((1582, 1615), 'numpy.testing.assert_equal', 'assert_equal', (['count1', 'count1_test'], {}), '(count1, count1_test)\n', (1594, 1615), False, 'from numpy.testing import assert_equal, run_module_suite\n'), ((1621, 1666), 'numpy.testing.assert_equal', 'assert_equal', (['count2_upper', 'count2_upper_test'], {}), '(count2_upper, count2_upper_test)\n', (1633, 1666), False, 'from numpy.testing import assert_equal, run_module_suite\n'), ((1671, 1716), 'numpy.testing.assert_equal', 'assert_equal', (['count1_upper', 'count1_upper_test'], {}), '(count1_upper, count1_upper_test)\n', (1683, 1716), False, 'from numpy.testing import assert_equal, run_module_suite\n'), ((1749, 1767), 'numpy.testing.run_module_suite', 'run_module_suite', ([], {}), '()\n', (1765, 1767), False, 'from numpy.testing import assert_equal, run_module_suite\n')]
#!/usr/bin/env python3 import sys from pathlib import Path from .posterization_gui import * from .simplepalettes import * import numpy as np import cv2 try: from PyQt5.QtCore import * from PyQt5.QtGui import * from PyQt5.QtWidgets import * except ImportError: from PySide2.QtCore import * from PySide2.QtGui import * from PySide2.QtWidgets import * from skimage.transform import rescale from qtwidgets import Toggle from PIL import Image # In command line: "pip install opencv-python-headless" to avoid qt complaining two set of binaries ### This sometimes happens: 'qt.gui.icc: fromIccProfile: failed minimal tag size sanity' class TimerMessageBox( QMessageBox ): def __init__( self, timeout = 3, parent = None ): super( TimerMessageBox, self ).__init__( parent ) self.setWindowTitle( "Algorithm is processing your image. Please hold on." ) self.time_to_wait = timeout self.setText( "Algorithm is processing your image. Please hold on." ) self.setStandardButtons( QMessageBox.NoButton ) self.timer = QTimer() self.timer.setInterval( 100 ) self.timer.timeout.connect( self.changeContent ) self.timer.start() def changeContent( self ): self.time_to_wait -= 1 if self.time_to_wait <= 0: self.close() def closeEvent( self, event ): self.timer.stop() event.accept() class MainWindow( QWidget ): def __init__( self ): super().__init__() self.title = 'Posterization' self.x = 300 self.y = 0 self.width = 100 self.height = 300 self.initUI() def initUI( self ): self.setWindowTitle( self.title ) self.setGeometry( self.x, self.y, self.width, self.height ) #self.setStyleSheet("background-color: white;") # Set the welcome icon in GIF self.welcome_img_path = str( Path( __file__ ).parent / "car.jpg" ) #self.welcome_img_path = "car.jpg" self.welcome = QPixmap( self.welcome_img_path ) self.imageLabel = QLabel() self.imageLabel.setPixmap( self.welcome ) # Toggle for switching to downsampled version self.switch = Toggle() self.switch.setMaximumWidth( 200 ) ### palette images self.paletteLabel = QLabel() self.paletteLabel.setPixmap( QPixmap() ) #### Variables self.imagePath = "" self.palette = None # used for recoloring self.weights_per_pixel_smooth = None self.weights_per_pixel = None self.palette_recolor = None self.palette_og = None self.waitingtime = 1 self.show_palette = 1 self.show_input = 0 self.live_smoothing = True self.blur_window_slider_val = 7 # default self.blur_slider_val = 0.1 # default self.binary_slider_val = 0.8 # default self.cluster_slider_val = 20 # default self.palette_slider_val = 6 # default self.blend_slider_val = 3 # default self.current_image_indx = -1 # Track the current image index in the image list self.imageList = [] self.add_to_imageList( cv2.cvtColor( cv2.imread( self.welcome_img_path ), cv2.COLOR_BGR2RGB ) ) self.paletteList = [-1 * np.ones( ( 1, 1, 1 ) )] self.input_image = None # Store it as np array self.saliency_map = None self.posterized_image_wo_smooth = -1 * np.ones( ( 1, 1, 1 ) ) # Store it as np array self.posterized_image_w_smooth = -1 * np.ones( ( 1, 1, 1 ) ) # Store it as np array #### BOXES btns_io_box = QHBoxLayout() # set bottons' box for I/O # algorithm_btns_box = QVBoxLayout() # set bottons' box for algorithms sld_box_palette = QHBoxLayout() sld_box_blend = QHBoxLayout() sld_box_cluster = QHBoxLayout() sld_box_binary = QHBoxLayout() toggle_box = QHBoxLayout() sld_box_blur = QHBoxLayout() sld_box_window = QHBoxLayout() btns_posterize_box = QHBoxLayout() # set bottons' box for posterization and reset sld_r_recolor = QHBoxLayout() sld_g_recolor = QHBoxLayout() sld_b_recolor = QHBoxLayout() blur_box = QHBoxLayout() img_box = QHBoxLayout() # set image's box pages_box = QVBoxLayout() # set next-previous box show_hide_box = QVBoxLayout() combo_recolor_box = QHBoxLayout() recolor_btn_box = QHBoxLayout() #### BUTTONS # button for selecting an input image self.img_btn = QPushButton( 'Choose Image' ) self.img_btn.clicked.connect( self.get_image ) self.img_btn.setToolTip( 'Press the button to <b>select</b> an image.' ) self.img_btn.setMaximumWidth( 150 ) # button for posterizing the given image self.posterize_btn = QPushButton( 'Posterize' ) self.posterize_btn.clicked.connect( self.posterize ) self.posterize_btn.setToolTip( 'Press the button to <b>posterize</b> your image.' ) self.posterize_btn.setMaximumWidth( 110 ) # button for reseting posterization parameters self.reset_posterize_btn = QPushButton( 'Reset' ) self.reset_posterize_btn.clicked.connect( self.reset_posterize ) self.reset_posterize_btn.setToolTip( 'Press the button to <b>reset</b> all posterization parameters.' ) self.reset_posterize_btn.setMaximumWidth( 110 ) # button for re-smoothing the posterized image self.smooth_btn = QPushButton( 'Re-Smooth' ) self.smooth_btn.clicked.connect( self.smooth ) self.smooth_btn.setToolTip( 'Press the button to <b>re-smooth</b> your posterized image.' ) self.smooth_btn.setMaximumWidth( 150 ) # button for loading the saliency map self.map_btn = QPushButton( 'Smooth with Custom Map' ) self.map_btn.clicked.connect( self.pop_up_load_saliency_map ) self.map_btn.setToolTip( 'Press the button to <b>load</b> your own map to blur.' ) self.map_btn.setMaximumWidth( 180 ) # button for saving the posterized image self.save_btn = QPushButton( 'Save Current Image' ) self.save_btn.clicked.connect( self.save_current_image ) self.save_btn.setToolTip( 'Press the button to <b>save</b> your current image.' ) self.save_btn.setMaximumWidth( 150 ) # button for saving the palette self.save_palette_btn = QPushButton( 'Save Palette' ) self.save_palette_btn.clicked.connect( self.save_current_palette ) self.save_palette_btn.setToolTip( 'Press the button to <b>save</b> the palette.' ) self.save_palette_btn.setMaximumWidth( 150 ) #### Previous-next buttons self.previous_btn = QPushButton( 'Previous Posterization' ) self.previous_btn.clicked.connect( self.paste_previous_image ) self.previous_btn.setToolTip( 'Press the button to see your <b>previous</b> image in the gallory.' ) self.previous_btn.setMinimumWidth( 165 ) self.previous_btn.setMaximumWidth( 165 ) self.next_btn = QPushButton( 'Next Posterization' ) self.next_btn.clicked.connect( self.paste_next_image ) self.next_btn.setToolTip( 'Press the button to see your <b>next</b> image in the gallory.' ) self.next_btn.setMinimumWidth( 165 ) self.next_btn.setMaximumWidth( 165 ) #### Show/Hide buttons self.palette_btn = QPushButton( 'Show/Hide Palette' ) self.palette_btn.clicked.connect( self.show_hide_palette ) self.palette_btn.setToolTip( 'Press the button to <b>show</b> or <b>hide</b> the palette.' ) self.palette_btn.setMinimumWidth( 165 ) self.palette_btn.setMaximumWidth( 165 ) self.og_img_btn = QPushButton( 'Show/Hide Input Image' ) self.og_img_btn.clicked.connect( self.show_hide_input_image ) self.og_img_btn.setToolTip( 'Press the button to <b>show</b> your input image.' ) self.og_img_btn.setMinimumWidth( 165 ) self.og_img_btn.setMaximumWidth( 165 ) #### SLIDERS # slider for palette size self.blend_sld = QSlider( Qt.Horizontal ) self.blend_sld.setRange( 0, 15 ) self.blend_sld.setFocusPolicy( Qt.NoFocus ) self.blend_sld.setSliderPosition( self.blend_slider_val ) self.blend_sld.setPageStep( 1 ) self.blend_sld.setToolTip( 'Fine-tune the slider to get your desired main palette size.' ) self.blend_sld.setMinimumWidth( 150 ) self.blend_sld.setMaximumWidth( 200 ) self.blend_sld.valueChanged.connect( self.blend_change_slider ) # slider for palette size self.palette_sld = QSlider( Qt.Horizontal ) self.palette_sld.setRange( 4, 15 ) self.palette_sld.setFocusPolicy( Qt.NoFocus ) self.palette_sld.setSliderPosition( self.palette_slider_val ) self.palette_sld.setPageStep( 1 ) self.palette_sld.setToolTip( 'Fine-tune the slider to get your desired main palette size.' ) self.palette_sld.setMinimumWidth( 150 ) self.palette_sld.setMaximumWidth( 200 ) self.palette_sld.valueChanged.connect( self.palette_change_slider ) # slider for number of clusters for Kmeans self.cluster_sld = QSlider( Qt.Horizontal ) self.cluster_sld.setRange( 15, 50 ) self.cluster_sld.setFocusPolicy( Qt.NoFocus ) self.cluster_sld.setSliderPosition( self.cluster_slider_val ) self.cluster_sld.setPageStep( 1 ) self.cluster_sld.setToolTip( 'Fine-tune the slider to get your desired threshold for outlier colors.' ) self.cluster_sld.setMinimumWidth( 150 ) self.cluster_sld.setMaximumWidth( 200 ) self.cluster_sld.valueChanged.connect( self.cluster_change_slider ) # slider for binary penalization self.binary_sld = QSlider( Qt.Horizontal ) self.binary_sld.setRange( 1, 200 ) self.binary_sld.setFocusPolicy( Qt.NoFocus ) self.binary_sld.setSliderPosition( int( 100 * self.binary_slider_val ) ) self.binary_sld.setPageStep( 1 ) self.binary_sld.setToolTip( 'Fine-tune the slider to get your desired penalization on binary term.' ) self.binary_sld.setMinimumWidth( 150 ) self.binary_sld.setMaximumWidth( 200 ) self.binary_sld.valueChanged.connect( self.binary_change_slider ) # slider for blurring threshold self.blur_sld = QSlider( Qt.Horizontal ) self.blur_sld.setRange( 0, 100 ) self.blur_sld.setFocusPolicy( Qt.NoFocus ) self.blur_sld.setSliderPosition( int( 100 * self.blur_slider_val ) ) self.blur_sld.setPageStep( 1 ) self.blur_sld.setToolTip( 'Fine-tune the slider to get your desired blurring threshold.' ) self.blur_sld.setMinimumWidth( 150 ) self.blur_sld.setMaximumWidth( 200 ) self.blur_sld.valueChanged.connect( self.blur_change_slider ) # slider for blurring threshold self.blur_window_sld = QSlider( Qt.Horizontal ) self.blur_window_sld.setRange( 0, 3 ) self.blur_window_sld.setFocusPolicy( Qt.NoFocus ) self.blur_window_sld.setSliderPosition( ( self.blur_window_slider_val - 3 ) / 2 ) self.blur_window_sld.setPageStep( 1 ) self.blur_window_sld.setToolTip( 'Fine-tune the slider to get your desired blurring window size.' ) self.blur_window_sld.setMinimumWidth( 150 ) self.blur_window_sld.setMaximumWidth( 200 ) self.blur_window_sld.valueChanged.connect( self.blur_window_change_slider ) ### LABELS self.switch_text = QLabel( 'Downsampled Version? ' ) self.switch_text.setAlignment( Qt.AlignLeft ) # labels self.blur_window_text = QLabel( 'Boundary smoothess (Default: 7):' ) self.blur_text = QLabel( 'Detail abstraction (Default: 0.1): ' ) self.binary_text = QLabel( 'Region clumpiness (Default: 0.8): ' ) self.cluster_text = QLabel( 'Rare color suppression (Default: 20):' ) self.palette_text = QLabel( 'Palette size (Default: 6): ' ) self.blend_text = QLabel( 'Palette blends (Default: 3): ' ) self.blur_window_text.setMaximumWidth( 250 ) self.blur_text.setMaximumWidth( 250 ) self.binary_text.setMaximumWidth( 250 ) self.cluster_text.setMaximumWidth( 250 ) self.palette_text.setMaximumWidth( 250 ) self.blend_text.setMaximumWidth( 250 ) # label text for blur slider self.blur_window_sld_label = QLabel( '7' ) self.blur_window_sld_label.setAlignment( Qt.AlignLeft ) self.blur_window_sld_label.setMinimumWidth( 80 ) self.blur_sld_label = QLabel( '0.1' ) self.blur_sld_label.setAlignment( Qt.AlignLeft ) self.blur_sld_label.setMinimumWidth( 80 ) # label text for binary penalization slider self.binary_sld_label = QLabel( '0.8' ) self.binary_sld_label.setAlignment( Qt.AlignLeft ) self.binary_sld_label.setMinimumWidth( 80 ) # label text for kmeans cluster slider self.cluster_sld_label = QLabel( '20' ) self.cluster_sld_label.setAlignment( Qt.AlignLeft ) self.cluster_sld_label.setMinimumWidth( 80 ) # label text for palette size slider self.palette_sld_label = QLabel( '6' ) self.palette_sld_label.setAlignment( Qt.AlignLeft ) self.palette_sld_label.setMinimumWidth( 80 ) # label text for blending way slider self.blend_sld_label = QLabel( '3' ) self.blend_sld_label.setAlignment( Qt.AlignLeft ) self.blend_sld_label.setMinimumWidth( 80 ) #### ### combo boxes for recoloring #### self.combobox = QComboBox(self) self.combobox.setMaximumWidth(100) self.combotext = QLabel( 'Choose color: ' ) self.r_slider_val = 0 self.g_slider_val = 0 self.b_slider_val = 0 self.rgb_text = QLabel( 'Recolor the image via its palette:' ) self.rgb_text.setMaximumWidth( 250 ) self.r_sld_label = QLabel( '0' ) self.r_sld_label.setAlignment( Qt.AlignLeft ) self.r_sld_label.setMinimumWidth( 80 ) self.r_sld_text_label = QLabel( 'R:' ) self.r_sld_text_label.setAlignment( Qt.AlignLeft ) self.g_sld_label = QLabel( '0' ) self.g_sld_label.setAlignment( Qt.AlignLeft ) self.g_sld_label.setMinimumWidth( 80 ) self.g_sld_text_label = QLabel( 'G:' ) self.g_sld_text_label.setAlignment( Qt.AlignRight ) self.b_sld_label = QLabel( '0' ) self.b_sld_label.setAlignment( Qt.AlignLeft ) self.b_sld_label.setMinimumWidth( 80 ) self.b_sld_text_label = QLabel( 'B:' ) self.b_sld_text_label.setAlignment( Qt.AlignRight ) # slider for palette recoloring self.r_sld = QSlider( Qt.Horizontal ) self.r_sld.setRange( 0, 255 ) self.r_sld.setFocusPolicy( Qt.NoFocus ) self.r_sld.setSliderPosition( self.r_slider_val ) self.r_sld.setPageStep( 1 ) self.r_sld.setToolTip( 'Fine-tune the slider to get your desired recoloring for R-channel.' ) self.r_sld.setMinimumWidth( 150 ) self.r_sld.setMaximumWidth( 200 ) self.r_sld.valueChanged.connect( self.r_change_slider ) self.g_sld = QSlider( Qt.Horizontal ) self.g_sld.setRange( 0, 255 ) self.g_sld.setFocusPolicy( Qt.NoFocus ) self.g_sld.setSliderPosition( self.g_slider_val ) self.g_sld.setPageStep( 1 ) self.g_sld.setToolTip( 'Fine-tune the slider to get your desired recoloring for G-channel.' ) self.g_sld.setMinimumWidth( 150 ) self.g_sld.setMaximumWidth( 200 ) self.g_sld.valueChanged.connect( self.g_change_slider ) self.b_sld = QSlider( Qt.Horizontal ) self.b_sld.setRange( 0, 255 ) self.b_sld.setFocusPolicy( Qt.NoFocus ) self.b_sld.setSliderPosition( self.b_slider_val ) self.b_sld.setPageStep( 1 ) self.b_sld.setToolTip( 'Fine-tune the slider to get your desired recoloring for B-channel.' ) self.b_sld.setMinimumWidth( 150 ) self.b_sld.setMaximumWidth( 200 ) self.b_sld.valueChanged.connect( self.b_change_slider ) self.recolor_btn = QPushButton( 'Reset Current Color' ) self.recolor_btn.clicked.connect( self.reset_current_recoloring ) self.recolor_btn.setToolTip( 'Press the button to <b>reset</b> the current palette color.' ) self.recolor_btn.setMinimumWidth( 150 ) self.recolor_btn.setMaximumWidth( 150 ) self.undo_recolor_btn = QPushButton( 'Reset All Colors' ) self.undo_recolor_btn.clicked.connect( self.reset_all_recoloring ) self.undo_recolor_btn.setToolTip( 'Press the button to <b>undo</b> your all previous recolorings.' ) self.undo_recolor_btn.setMinimumWidth( 150 ) self.undo_recolor_btn.setMaximumWidth( 150 ) ### BOX FRAMES btns_io_box.addWidget( self.img_btn ) btns_io_box.addWidget( self.save_btn ) btns_io_box.addWidget( self.save_palette_btn ) btns_io_box.addStretch(40) # Separate boxes for parameters sld_box_palette.addWidget( self.palette_text ) sld_box_palette.addWidget( self.palette_sld ) sld_box_palette.addWidget( self.palette_sld_label ) sld_box_palette.addStretch(8) sld_box_blend.addWidget( self.blend_text ) sld_box_blend.addWidget( self.blend_sld ) sld_box_blend.addWidget( self.blend_sld_label ) sld_box_blend.addStretch(8) sld_box_cluster.addWidget( self.cluster_text ) sld_box_cluster.addWidget( self.cluster_sld ) sld_box_cluster.addWidget( self.cluster_sld_label ) sld_box_cluster.addStretch(8) sld_box_binary.addWidget( self.binary_text ) sld_box_binary.addWidget( self.binary_sld ) sld_box_binary.addWidget( self.binary_sld_label ) sld_box_binary.addStretch(8) toggle_box.addWidget( self.switch_text ) toggle_box.addWidget( self.switch ) toggle_box.addStretch(8) btns_posterize_box.addWidget( self.posterize_btn ) btns_posterize_box.addWidget( self.reset_posterize_btn ) btns_posterize_box.addStretch(8) sld_box_blur.addWidget( self.blur_text ) sld_box_blur.addWidget( self.blur_sld ) sld_box_blur.addWidget( self.blur_sld_label ) sld_box_blur.addStretch(8) sld_box_window.addWidget( self.blur_window_text ) sld_box_window.addWidget( self.blur_window_sld ) sld_box_window.addWidget( self.blur_window_sld_label ) sld_box_window.addStretch(8) # blur box for re-smooth and smooth by map blur_box.addWidget( self.smooth_btn ) blur_box.addWidget( self.map_btn ) blur_box.addStretch(8) # recoloring box combo_recolor_box.addWidget( self.combotext ) combo_recolor_box.addWidget( self.combobox ) combo_recolor_box.addStretch(8) sld_r_recolor.addWidget( self.r_sld_text_label ) sld_r_recolor.addWidget( self.r_sld ) sld_r_recolor.addWidget( self.r_sld_label ) sld_r_recolor.addStretch(8) sld_g_recolor.addWidget( self.g_sld_text_label ) sld_g_recolor.addWidget( self.g_sld ) sld_g_recolor.addWidget( self.g_sld_label ) sld_g_recolor.addStretch(8) sld_b_recolor.addWidget( self.b_sld_text_label ) sld_b_recolor.addWidget( self.b_sld ) sld_b_recolor.addWidget( self.b_sld_label ) sld_b_recolor.addStretch(8) recolor_btn_box.addWidget( self.recolor_btn ) recolor_btn_box.addWidget( self.undo_recolor_btn ) recolor_btn_box.addStretch(8) # Image box img_box.addStretch(1) img_box.addWidget( self.paletteLabel ) img_box.addStretch(1) img_box.addWidget( self.imageLabel ) img_box.addStretch(4) # Previous-next box pages_box.addWidget( self.previous_btn ) show_hide_box.addWidget( self.next_btn ) # Show-hide box pages_box.addWidget( self.palette_btn ) show_hide_box.addWidget( self.og_img_btn ) # Set grid layout grid = QGridLayout() grid.setSpacing(12) grid.addLayout( btns_io_box, 0, 0 ) ### parameters for posterization grid.addLayout( sld_box_palette, 1, 0 ) grid.addLayout( sld_box_blend, 2, 0 ) grid.addLayout( sld_box_cluster, 3, 0 ) grid.addLayout( sld_box_binary, 4, 0 ) grid.addLayout( toggle_box, 5, 0 ) grid.addLayout( btns_posterize_box, 6, 0 ) ### parameters for smoothing grid.addLayout( sld_box_blur, 8, 0 ) grid.addLayout( sld_box_window, 9, 0 ) grid.addLayout( blur_box, 10, 0 ) ### boxes for previous/next and show/hide grid.addLayout( pages_box, 0, 10 ) grid.addLayout( show_hide_box, 0, 11 ) ### sliders for recoloring grid.addWidget( self.rgb_text, 12, 0 ) grid.addLayout( combo_recolor_box, 13, 0 ) grid.addLayout( sld_r_recolor, 14, 0 ) grid.addLayout( sld_g_recolor, 15, 0 ) grid.addLayout( sld_b_recolor, 16, 0 ) grid.addLayout( recolor_btn_box, 17, 0 ) grid.addLayout( img_box, 1, 1, 19, 19 ) self.setLayout(grid) self.show() ### Recoloring functions def set_rgb_slider( self, color ): color = color * 255. self.r_change_slider( int( color[0] ) ) self.g_change_slider( int( color[1] ) ) self.b_change_slider( int( color[2] ) ) self.r_sld.setSliderPosition( int( color[0] ) ) self.g_sld.setSliderPosition( int( color[1] ) ) self.b_sld.setSliderPosition( int( color[2] ) ) def onActivated( self, text ): color_indx = int( text ) - 1 color = self.palette_recolor[ color_indx ] self.set_rgb_slider( color ) def set_combo_icon( self ): self.combobox.clear() # reset combo box for i in range( len( self.palette_recolor ) ): self.combobox.addItem( str( i + 1 ) ) self.combobox.activated[str].connect( self.onActivated ) default_color = self.palette_recolor[0] self.set_rgb_slider( default_color ) def get_recolor_img_and_palette( self ): #recolor_img = ( self.weights_per_pixel @ self.palette_recolor ).reshape( self.input_image.shape ) # fix it for odd resolution when downsampled version applied if self.input_image.shape[0] % 2 == 1 and self.switch.isChecked(): w = self.input_image.shape[0] + 1 else: w = self.input_image.shape[0] if self.input_image.shape[1] % 2 == 1 and self.switch.isChecked(): h = self.input_image.shape[1] + 1 else: h = self.input_image.shape[1] recolor_smooth_img = np.clip( 0, 255, ( self.weights_per_pixel_smooth @ self.palette_recolor ).reshape( ( w, h, 3 ) ) * 255. ).astype( np.uint8 ) #recolor_smooth_img = post_smoothing( PIL.Image.fromarray( np.clip( 0, 255, recolor_img * 255. ).astype( np.uint8 ), 'RGB' ), #self.blur_slider_val, blur_window = self.blur_window_slider_val ) new_palette = np.ascontiguousarray( np.clip( 0, 255, simplepalettes.palette2swatch( self.palette_recolor ) * 255. ).astype( np.uint8 ).transpose( ( 1, 0, 2 ) ) ) return recolor_smooth_img, new_palette def recolor_via_palette( self ): color_indx = int( self.combobox.currentText() ) - 1 r_value = self.r_sld.value() g_value = self.g_sld.value() b_value = self.b_sld.value() self.palette_recolor[ color_indx ] = np.array([ r_value, g_value, b_value ]) / 255. recolor_img, new_palette = self.get_recolor_img_and_palette() self.add_to_paletteList( new_palette ) self.add_to_imageList( recolor_img ) self.set_image( self.imageLabel, self.imageList[-1] ) self.set_image( self.paletteLabel, self.paletteList[-1] ) # update current index position self.current_image_indx = len( self.imageList ) - 1 def reset_current_recoloring( self ): if self.posterized_image_wo_smooth[0][0][0] == -1: QMessageBox.warning( self, 'Warning', 'Please posterize your image first' ) else: # visualization for current combox text color_indx = int( self.combobox.currentText() ) - 1 current_color = self.palette_og[ color_indx ] self.set_rgb_slider( current_color ) self.palette_recolor[ color_indx ] = self.palette_og[ color_indx ].copy() recolor_img, new_palette = self.get_recolor_img_and_palette() self.add_to_paletteList( new_palette ) self.add_to_imageList( recolor_img ) self.set_image( self.imageLabel, self.imageList[-1] ) self.set_image( self.paletteLabel, self.paletteList[-1] ) # update current index position self.current_image_indx = len( self.imageList ) - 1 def reset_all_recoloring( self ): if self.posterized_image_wo_smooth[0][0][0] == -1: QMessageBox.warning( self, 'Warning', 'Please posterize your image first' ) else: # visualization for current combox text color_indx = int( self.combobox.currentText() ) - 1 current_color = self.palette_og[ color_indx ] self.set_rgb_slider( current_color ) self.palette_recolor = self.palette_og.copy() self.add_to_paletteList( self.palette ) self.add_to_imageList( self.posterized_image_w_smooth ) self.set_image( self.imageLabel, self.imageList[-1] ) self.set_image( self.paletteLabel, self.paletteList[-1] ) # update current index position self.current_image_indx = len( self.imageList ) - 1 ### Reset for posterization parameters def reset_posterize( self ): self.binary_change_slider( 80 ) self.cluster_change_slider( 20 ) self.palette_change_slider( 6 ) self.blend_change_slider( 3 ) self.binary_sld.setSliderPosition( 80 ) self.cluster_sld.setSliderPosition( 20 ) self.palette_sld.setSliderPosition( 6 ) self.blend_sld.setSliderPosition( 3 ) self.binary_sld.repaint() self.cluster_sld.repaint() self.palette_sld.repaint() self.blend_sld.repaint() ### Slider functions def r_change_slider(self, value): self.r_slider_val = value self.r_sld_label.setText( str( value ) ) if self.live_smoothing: self.recolor_via_palette() def g_change_slider(self, value): self.g_slider_val = value self.g_sld_label.setText( str( value ) ) if self.live_smoothing: self.recolor_via_palette() def b_change_slider(self, value): self.b_slider_val = value self.b_sld_label.setText( str( value ) ) if self.live_smoothing: self.recolor_via_palette() def blur_window_change_slider(self, value): self.blur_window_slider_val = 2 * value + 3 self.blur_window_sld_label.setText( str( 2 * value + 3 ) ) if self.live_smoothing: self.smooth() def blur_change_slider(self, value): self.blur_slider_val = value / 100 self.blur_sld_label.setText( str( value / 100 ) ) if self.live_smoothing: self.smooth() def binary_change_slider(self, value): self.binary_slider_val = value / 100 self.binary_sld_label.setText( str( value / 100 ) ) def cluster_change_slider(self, value): self.cluster_slider_val = value self.cluster_sld_label.setText( str( value ) ) def palette_change_slider(self, value): self.palette_slider_val = value self.palette_sld_label.setText( str( value ) ) def blend_change_slider(self, value): self.blend_slider_val = value self.blend_sld_label.setText( str( value ) ) # Function for selecting an input image def get_image( self ): img = QFileDialog.getOpenFileName( self, 'Select file' ) if img: path = img[0] self.load_image( path ) else: QMessageBox.warning( self, 'Warning' , 'No file selected.' ) def paste_previous_image( self ): self.current_image_indx -= 1 if self.current_image_indx == -2: QMessageBox.warning( self,'Warning','Please select an image first.' ) self.current_image_indx += 1 elif self.current_image_indx == -1: QMessageBox.warning( self,'Warning','No more previous image.' ) self.current_image_indx += 1 else: if self.current_image_indx != 0 and self.show_palette == 1: self.set_image( self.paletteLabel, self.paletteList[self.current_image_indx] ) else: # input image has no palette, so place a blank self.paletteLabel.setPixmap( QPixmap() ) self.paletteLabel.repaint() self.set_image( self.imageLabel, self.imageList[self.current_image_indx] ) def paste_next_image( self ): self.current_image_indx += 1 if self.current_image_indx == 0: QMessageBox.warning( self,'Warning','Please select an image first.' ) self.current_image_indx -= 1 elif self.current_image_indx == len( self.imageList ): QMessageBox.warning( self,'Warning','No more next image.' ) self.current_image_indx -= 1 else: if self.current_image_indx != 0 and self.show_palette == 1: self.set_image( self.paletteLabel, self.paletteList[self.current_image_indx] ) else: # input image has no palette, so place a blank self.paletteLabel.setPixmap( QPixmap() ) self.paletteLabel.repaint() self.set_image( self.imageLabel, self.imageList[self.current_image_indx] ) #Load new image function def set_image( self, panel, image ): #Load the image into the label height, width, dim = image.shape qim = QImage( image.data, width, height, 3 * width, QImage.Format_RGB888 ) panel.setPixmap( QPixmap( qim ) ) panel.repaint() def add_to_imageList( self, image ): self.imageList.append( np.asarray( image ) ) def add_to_paletteList( self, palette ): self.paletteList.append( np.asarray( palette ) ) def load_image( self, path ): print ( "Loading Image." ) self.imageList = [] # initialized back to empty when giving another input image self.paletteList = [-1 * np.ones( ( 1, 1, 1 ) )] self.paletteLabel.setPixmap( QPixmap() ) # push input image in the list self.current_image_indx += 1 self.input_image = cv2.cvtColor( cv2.imread( path ), cv2.COLOR_BGR2RGB ) self.add_to_imageList( self.input_image ) self.imageLabel.setPixmap( QPixmap( path ) ) self.imagePath = path def show_hide_palette( self ): #if self.imagePath == "": # QMessageBox.warning( self, 'Warning', 'Please select an image first.' ) if self.paletteList[-1][0, 0, 0] == -1: QMessageBox.warning( self, 'Warning', 'You do not have palette. Please posterize the image first.' ) else: self.show_palette = 1 - self.show_palette if self.current_image_indx != 0 and self.show_palette == 1: self.set_image( self.paletteLabel, self.paletteList[self.current_image_indx] ) else: # input image has no palette, so place a blank self.paletteLabel.setPixmap( QPixmap() ) def show_hide_input_image( self ): #if self.imagePath == "": # QMessageBox.warning( self, 'Warning', 'Please select an image first.' ) if self.posterized_image_wo_smooth[0][0][0] == -1: QMessageBox.warning( self, 'Warning', 'This is your input image.' ) else: self.show_input = 1 - self.show_input if self.show_input == 1: self.set_image( self.imageLabel, self.imageList[0] ) else: self.set_image( self.imageLabel, self.imageList[self.current_image_indx] ) # posterization def posterize( self ): #if self.imagePath == "": # QMessageBox.warning( self, 'Warning', 'Please select an image first.' ) #else: if self.imagePath == "": img_arr = np.asfarray( PIL.Image.open( self.welcome_img_path ).convert( 'RGB' ) ) / 255. self.input_image = img_arr path = self.welcome_img_path else: img_arr = np.asfarray( PIL.Image.open( self.imagePath ).convert( 'RGB' ) ) / 255. path = self.imagePath width, height, dim = img_arr.shape length = max( width, height ) self.message = "This image has size " + str( height ) + ' x ' + str( width ) + '.\n\n' if length >= 1800: self.message += 'This is a large image and may take more than 8 mins to process.\n' + 'We suggest you posterize a downsized version to select appropriate parameters or vectorize the output.\n\n' else: if 500 < length < 600: self.waitingtime = 2 elif 600 < length < 1000: self.waitingtime = 3 elif 1000 <= length: self.waitingtime = 4 self.message += 'This will take roughly ' + str( self.waitingtime ) + ' minutes to process.\n\n' reply = QMessageBox.question( self, 'Message', self.message + 'Do you want to proceed and posterize the image?', QMessageBox.Yes | QMessageBox.No, QMessageBox.No ) if reply == QMessageBox.Yes: print( "Start posterizing." ) # algorithm starts start = time.time() messagebox = TimerMessageBox( 1, self ) messagebox.open() # if downsampled version is selected, downsize the input and divide the penality by 2 if self.switch.isChecked(): print( 'Downsampled version selected.' ) img_arr = rescale( img_arr, 0.5, order=0, multichannel=True , anti_aliasing=False ) self.binary_slider_val /= 2 # K-means img_arr_re = img_arr.reshape( ( -1, 3 ) ) img_arr_cluster = get_kmeans_cluster_image( self.cluster_slider_val, img_arr_re, img_arr.shape[0], img_arr.shape[1] ) # MLO post_img, final_colors, add_mix_layers, palette = \ posterization( path, img_arr, img_arr_cluster, self.palette_slider_val, self.blend_slider_val, self.binary_slider_val ) if self.switch.isChecked(): ### 'Greyscale' might fail in this case since the palette size for greyscale is 2. new_am = add_mix_layers.reshape( ( post_img.shape[0], post_img.shape[1], self.palette_slider_val ) ) self.weights_per_pixel = rescale( new_am, 2, order=0, multichannel=True, anti_aliasing=False ).reshape( -1, self.palette_slider_val ) else: self.weights_per_pixel = add_mix_layers # save weight list per pixel # save palette # 'ascontiguousarray' to make a C contiguous copy self.palette = np.ascontiguousarray( np.clip( 0, 255, simplepalettes.palette2swatch( palette ) * 255. ).astype( np.uint8 ).transpose( ( 1, 0, 2 ) ) ) if self.switch.isChecked(): post_img = rescale( post_img, 2, order=0, multichannel=True, anti_aliasing=False ) self.posterized_image_wo_smooth = np.clip( 0, 255, post_img*255. ).astype( np.uint8 ) self.posterized_image_wo_smooth = cv2.medianBlur( self.posterized_image_wo_smooth, 5 ) else: self.posterized_image_wo_smooth = np.clip( 0, 255, post_img * 255. ).astype( np.uint8 ) # convert to uint8 format #self.posterized_image_wo_smooth = np.clip( 0, 255, post_img * 255. ).astype( np.uint8 ) # make a map from unique colors to weights unique_colors, unique_indices = np.unique( self.posterized_image_wo_smooth.reshape( -1, 3 ), return_index = True, axis = 0 ) color2weights = {} for col, index in zip( unique_colors, unique_indices ): weights = self.weights_per_pixel[ index ] color2weights[ tuple( col ) ] = weights # post-smoothing self.posterized_image_w_smooth = post_smoothing( PIL.Image.fromarray( self.posterized_image_wo_smooth, 'RGB' ), self.blur_slider_val, blur_window = self.blur_window_slider_val ) # pass smoothing along to the weights self.weights_per_pixel_smooth = self.weights_per_pixel.copy() for col, weights in color2weights.items(): #color_mask = ( self.posterized_image_w_smooth.reshape( -1, 3 ) == np.array( col ) [None,:] ).all() color_mask = np.where( np.all( self.posterized_image_w_smooth.reshape( -1, 3 ) == np.array( col ), axis = 1 ) )[0] self.weights_per_pixel_smooth[ color_mask ] = weights self.weights_per_pixel_smooth.shape = self.weights_per_pixel.shape ### setting for recoloring self.palette_recolor = palette # save for palette recoloring self.palette_og = self.palette_recolor.copy() self.set_combo_icon() end = time.time() print( "Finished. Total time: ", end - start ) self.add_to_paletteList( self.palette ) self.add_to_imageList( self.posterized_image_w_smooth ) self.set_image( self.imageLabel, self.imageList[-1] ) self.set_image( self.paletteLabel, self.paletteList[-1] ) # update current index position self.current_image_indx = len( self.imageList ) - 1 else: pass # re-smooth the image def smooth( self ): #if self.imagePath == "": # QMessageBox.warning( self,'Warning','Please select an image first!' ) #else: if self.posterized_image_wo_smooth[0][0][0] == -1: QMessageBox.warning( self, 'Warning', 'Please posterize your image first' ) else: print( "Start smoothing." ) #messagebox = TimerMessageBox( 1, self ) #messagebox.open() self.posterized_image_w_smooth = post_smoothing( PIL.Image.fromarray( self.posterized_image_wo_smooth, 'RGB' ), self.blur_slider_val, blur_window = self.blur_window_slider_val ) print( "Smoothing Finished." ) self.add_to_paletteList( self.paletteList[-1] ) self.add_to_imageList( self.posterized_image_w_smooth ) self.set_image( self.imageLabel, self.imageList[-1] ) # update current index position self.current_image_indx = len( self.imageList ) - 1 # function to save current image def save_current_image( self ): #if self.imagePath == "": # QMessageBox.warning( self,'Warning','Please select an image first.' ) #else: if self.posterized_image_wo_smooth[0][0][0] == -1: QMessageBox.warning( self, 'Warning', 'Please posterize your image first.' ) else: reply = QMessageBox.question( self, 'Message', "Are you sure to save your current image on this panel?", QMessageBox.Yes | QMessageBox.No, QMessageBox.No ) if reply == QMessageBox.Yes: image_name = QFileDialog.getSaveFileName( self, 'Save Image' ) if not image_name: return if image_name[0][-4:] in ['.jpg', '.png']: path_name = image_name[0] else: path_name = image_name[0] + '.png' Image.fromarray( self.imageList[self.current_image_indx] ).save( path_name ) else: pass # function to save current image def save_current_palette( self ): #if self.imagePath == "": # QMessageBox.warning( self,'Warning','Please select an image first.' ) #else: if self.posterized_image_wo_smooth[0][0][0] == -1: QMessageBox.warning( self, 'Warning', 'Please posterize your image first.' ) else: reply = QMessageBox.question( self, 'Message', "Are you sure to save your current palette on this panel?", QMessageBox.Yes | QMessageBox.No, QMessageBox.No ) if reply == QMessageBox.Yes: image_name = QFileDialog.getSaveFileName( self, 'Save Palette' ) if not image_name: return if image_name[0][-4:] in ['.jpg', '.png']: path_name = image_name[0] else: path_name = image_name[0] + '.png' Image.fromarray( self.paletteList[self.current_image_indx] ).save( path_name ) else: pass # load user's own blurring map def pop_up_load_saliency_map( self ): #if self.imagePath == "": # QMessageBox.warning( self,'Warning','Please select an image first.' ) #else: if self.posterized_image_wo_smooth[0][0][0] == -1: QMessageBox.warning( self, 'Warning', 'Please posterize your image first.' ) else: reply = QMessageBox.question( self, 'Message', "Do you have your own blurring map (in grayscale and in .jpg/.png extension)?", QMessageBox.Yes | QMessageBox.No, QMessageBox.No ) if reply == QMessageBox.Yes: map = QFileDialog.getOpenFileName( self, 'Select file' ) map_path = map[0] if map_path[-4:] not in ['.jpg', '.png']: QMessageBox.warning( self, 'Warning', 'Please upload your map with .jpg or .png extension.' ) return self.saliency_map = cv2.imread( map[0] ) / 255 h_s, w_s, dim_s = self.saliency_map.shape h_i, w_i, dim_i = self.input_image.shape if ( h_i, w_i ) != ( h_s, w_s ): QMessageBox.warning( self, 'Warning', 'Please upload your map with size:\n\n ' + ' ' + str( h_i ) + ' x ' + str( w_i ) + '\n\n' + 'You upload the map with size:\n\n ' + ' ' + str( h_s ) + ' x ' + str( w_s ) ) return if not np.array_equal( self.saliency_map[:,:,0], self.saliency_map[:,:,1] ) or not np.array_equal( self.saliency_map[:,:,1], self.saliency_map[:,:,2] ): QMessageBox.warning( self, 'Warning', 'Please upload your map with grayscale.' ) return print( "Start smoothing." ) messagebox = TimerMessageBox( 1, self ) messagebox.open() self.posterized_image_w_smooth = post_smoothing( PIL.Image.fromarray( self.posterized_image_wo_smooth, 'RGB' ), self.blur_slider_val, blur_window = self.blur_window_slider_val, blur_map = self.saliency_map[:, :, 0] ) print( "Smoothing Finished." ) self.add_to_paletteList( self.paletteList[-1] ) self.add_to_imageList( self.posterized_image_w_smooth ) self.set_image( self.imageLabel, self.imageList[-1] ) # update current index position self.current_image_indx = len( self.imageList ) - 1 # Function if users tend to close the app def closeEvent( self, event ): reply = QMessageBox.question( self, 'Message', "Are you sure you want to quit?", QMessageBox.Yes | QMessageBox.No, QMessageBox.No ) if reply == QMessageBox.Yes: event.accept() else: event.ignore() def main(): app = QApplication( sys.argv ) ex = MainWindow() sys.exit( app.exec_() ) if __name__ == '__main__': main()
[ "numpy.clip", "PIL.Image.fromarray", "numpy.ones", "pathlib.Path", "numpy.asarray", "cv2.medianBlur", "numpy.array", "numpy.array_equal", "qtwidgets.Toggle", "skimage.transform.rescale", "cv2.imread" ]
[((2289, 2297), 'qtwidgets.Toggle', 'Toggle', ([], {}), '()\n', (2295, 2297), False, 'from qtwidgets import Toggle\n'), ((3647, 3665), 'numpy.ones', 'np.ones', (['(1, 1, 1)'], {}), '((1, 1, 1))\n', (3654, 3665), True, 'import numpy as np\n'), ((3744, 3762), 'numpy.ones', 'np.ones', (['(1, 1, 1)'], {}), '((1, 1, 1))\n', (3751, 3762), True, 'import numpy as np\n'), ((24825, 24862), 'numpy.array', 'np.array', (['[r_value, g_value, b_value]'], {}), '([r_value, g_value, b_value])\n', (24833, 24862), True, 'import numpy as np\n'), ((31821, 31838), 'numpy.asarray', 'np.asarray', (['image'], {}), '(image)\n', (31831, 31838), True, 'import numpy as np\n'), ((31930, 31949), 'numpy.asarray', 'np.asarray', (['palette'], {}), '(palette)\n', (31940, 31949), True, 'import numpy as np\n'), ((32374, 32390), 'cv2.imread', 'cv2.imread', (['path'], {}), '(path)\n', (32384, 32390), False, 'import cv2\n'), ((3370, 3403), 'cv2.imread', 'cv2.imread', (['self.welcome_img_path'], {}), '(self.welcome_img_path)\n', (3380, 3403), False, 'import cv2\n'), ((3471, 3489), 'numpy.ones', 'np.ones', (['(1, 1, 1)'], {}), '((1, 1, 1))\n', (3478, 3489), True, 'import numpy as np\n'), ((32166, 32184), 'numpy.ones', 'np.ones', (['(1, 1, 1)'], {}), '((1, 1, 1))\n', (32173, 32184), True, 'import numpy as np\n'), ((35856, 35926), 'skimage.transform.rescale', 'rescale', (['img_arr', '(0.5)'], {'order': '(0)', 'multichannel': '(True)', 'anti_aliasing': '(False)'}), '(img_arr, 0.5, order=0, multichannel=True, anti_aliasing=False)\n', (35863, 35926), False, 'from skimage.transform import rescale\n'), ((37293, 37362), 'skimage.transform.rescale', 'rescale', (['post_img', '(2)'], {'order': '(0)', 'multichannel': '(True)', 'anti_aliasing': '(False)'}), '(post_img, 2, order=0, multichannel=True, anti_aliasing=False)\n', (37300, 37362), False, 'from skimage.transform import rescale\n'), ((37518, 37568), 'cv2.medianBlur', 'cv2.medianBlur', (['self.posterized_image_wo_smooth', '(5)'], {}), '(self.posterized_image_wo_smooth, 5)\n', (37532, 37568), False, 'import cv2\n'), ((1982, 1996), 'pathlib.Path', 'Path', (['__file__'], {}), '(__file__)\n', (1986, 1996), False, 'from pathlib import Path\n'), ((44178, 44196), 'cv2.imread', 'cv2.imread', (['map[0]'], {}), '(map[0])\n', (44188, 44196), False, 'import cv2\n'), ((36734, 36801), 'skimage.transform.rescale', 'rescale', (['new_am', '(2)'], {'order': '(0)', 'multichannel': '(True)', 'anti_aliasing': '(False)'}), '(new_am, 2, order=0, multichannel=True, anti_aliasing=False)\n', (36741, 36801), False, 'from skimage.transform import rescale\n'), ((37416, 37449), 'numpy.clip', 'np.clip', (['(0)', '(255)', '(post_img * 255.0)'], {}), '(0, 255, post_img * 255.0)\n', (37423, 37449), True, 'import numpy as np\n'), ((37639, 37672), 'numpy.clip', 'np.clip', (['(0)', '(255)', '(post_img * 255.0)'], {}), '(0, 255, post_img * 255.0)\n', (37646, 37672), True, 'import numpy as np\n'), ((41955, 42011), 'PIL.Image.fromarray', 'Image.fromarray', (['self.imageList[self.current_image_indx]'], {}), '(self.imageList[self.current_image_indx])\n', (41970, 42011), False, 'from PIL import Image\n'), ((43052, 43110), 'PIL.Image.fromarray', 'Image.fromarray', (['self.paletteList[self.current_image_indx]'], {}), '(self.paletteList[self.current_image_indx])\n', (43067, 43110), False, 'from PIL import Image\n'), ((44698, 44768), 'numpy.array_equal', 'np.array_equal', (['self.saliency_map[:, :, 0]', 'self.saliency_map[:, :, 1]'], {}), '(self.saliency_map[:, :, 0], self.saliency_map[:, :, 1])\n', (44712, 44768), True, 'import numpy as np\n'), ((44774, 44844), 'numpy.array_equal', 'np.array_equal', (['self.saliency_map[:, :, 1]', 'self.saliency_map[:, :, 2]'], {}), '(self.saliency_map[:, :, 1], self.saliency_map[:, :, 2])\n', (44788, 44844), True, 'import numpy as np\n'), ((38893, 38906), 'numpy.array', 'np.array', (['col'], {}), '(col)\n', (38901, 38906), True, 'import numpy as np\n')]
import pandas as pd import re from IPython.display import display import matplotlib.pyplot as plt import numpy as np import os from pathlib import Path from glob import glob import black import natsort import operator # from matplotlib.pyplot import cm import matplotlib as mpl from constants import * from matplotlib.lines import Line2D plt.style.use("style.mplstyle") #matplotlib style sheet avg_parameter = 3 background_colour = ["r","b","k", "m"] double_size = 8 layer_start, layer_end = 1,5 # setting of layers for graphs from layer_name output_files = "output_files" layer_name = ["Serial", "MPIIO", "PHDF5", "ADIOS2_HDF5", "ADIOS2_BP4"] def dict_output( filename, layer_name, array_size ): # function to return array, time, rate from CSV file mydata = pd.read_csv( filename, index_col=False, skiprows=0, names=[ "Fields", "Layername", "ArraySize", "NumNodes", "AverageTime", "AverageRate" ] ) """ Set array filter """ num_nodes = mydata.NumNodes[(mydata.Layername == layer_name) & (mydata.ArraySize == array_size)] # number of nodes rate = mydata.AverageRate[(mydata.Layername == layer_name) & (mydata.ArraySize == array_size)] # Gb/s return rate, num_nodes def average_dir(dir, min_size, max_size): rate = [] rate_avg = [] ranks = [] Global_size = [] rate_persize = [] """ Obtain paths for subdirectories """ dirFiles = glob( f"{os.getcwd()}/output_dirs/{dir}/*/" ) # assign full output directories for csv file output_dirs = natsort.natsorted(dirFiles) # N_size = 2 ** 8 # only checking 2^8 cube len for layers in range(layer_start, layer_end): for size in range(min_size,max_size): cube_len = 2 ** size for i in range(len(output_dirs)): trail = os.path.basename( os.path.normpath(output_dirs[i]) ) # gives name of cores, and processors if "_v1" in trail: # start with first version of file, iterate from then. rate_avg = 0 avgd_num = 0 for job_num in range(1,avg_parameter+1): #for v1,v2,v3 avgd_dir = output_dirs[i].replace("v1",f"v{job_num}") # replace v1 with v2,v3? Preceeding path can stay same. avgd_file = f"{avgd_dir}output.csv" if os.path.isfile(avgd_file): #check if v1, v2, v3 exist for that node config avgd_num += 1 # accurate avg parameter, iterates for every version that actually exits. rate, ranks = dict_output(avgd_file,layer_name[layers],cube_len) rate_avg += rate rate_avg = rate_avg/avgd_num Global_size = ((cube_len ** 3) * double_size * ranks.values) / (10 ** 9) # global array size from cube length rate_persize.append([layer_name[layers], ranks.values[0], cube_len, Global_size, rate_avg.values[0]]) # append plot output data to rate_persize # # output of rate_persize list array to csv file output_data = pd.DataFrame(rate_persize, columns=[ 'LayerName','Ranks', 'CubeLen', 'GlobalSize', 'Rate']) out = output_data.to_csv(f"output_csv/{dir}_avg_output.csv", index=False) def plot_rate_v_ranks(target_dir, layer_start, layer_end, min_size, max_size, ax1): max_rate, max_rank = 0, 0 # outputdata_layer_v_size(f"{target_dir}", min_size, max_size) # data processing function, outputs everything to output.csv average_dir(f"{target_dir}", min_size, max_size) # data processing function, outputs everything to output.csv input_data = pd.read_csv( # read from output.csv f"output_csv/{target_dir}_avg_output.csv", index_col=False, skiprows=0 ) for x in range(layer_start, layer_end): for y in range(min_size, max_size): size_t = 2 ** y label1 = layer_name[x] ranks = input_data.Ranks[ (input_data.LayerName == layer_name[x]) & (input_data.CubeLen == size_t) ] rate = input_data.Rate[ ( input_data.LayerName == layer_name[x]) & (input_data.CubeLen == size_t ) ] ax1.plot(ranks, rate, marker_desc[x], label=label1, c=background_colour[x-layer_start]) if max_rate < max(rate): max_rate = max(rate) max_rank = ranks[rate.idxmax()] return max_rate, max_rank def plotting_mult_dirs(target_dir, min_size, max_size, param): """ Plot formatting """ # fig1 = plt.figure(figsize=(8,6)) fig1 = plt.figure(figsize=(6,5.5)) ax1 = plt.axes() """ Inits """ max_rate, max_rank, local_rate, local_rank = 0,0,0,0 # init variable for finding max rate and corresponding ranks for x in target_dir: local_rate, local_rank = plot_rate_v_ranks(x, layer_start, layer_end, min_size, max_size, ax1) if max_rate < local_rate: #in case multiple set of directories max_rate = local_rate max_rank = local_rank text = "Rank={:.0f},Max Rate={:.3f}GB/s".format(max_rank,max_rate) ax1.annotate(text, xy=(max_rank, max_rate), xytext=(0.95, 0.96), **kw) legend_elements = [ Line2D([0], [0], marker = "D", color='r', lw=1, label='MPIIO'), Line2D([0], [0], marker= "o", color='b', lw=1, label='PHDF5'), Line2D([0], [0], marker= "v", color='k', lw=1,label='ADIOS2/HDF5'), Line2D([0], [0], marker= "^", color='m', lw=1, label='ADIOS2/BP4')] ax1.legend(handles=legend_elements) ax1.set_xlabel("MPI ranks") ax1.set_ylabel("Average Rate (GB/s)") ax1.set_yscale("log") ax1.set_xscale("log") ax1.set_ybound(0,4) fig1.suptitle(f"I/O rate comparison b/w NextGenIO & Fulhame - {param} striping") fig1.tight_layout() def speedup(dir): """ Inits """ double_size = 8 layer_start = 3 # setting of layers for graphs from layer_name layer_end = 5 rate_hdf5 = [] num_nodes_hdf5 = [] array_hdf5 = [] """ Plot formatting """ # fig1 = plt.figure(figsize=(8,6)) fig1 = plt.figure(figsize=(8,6)) # plt.rcParams["font.size"] = "10" ax1 = plt.axes() """ Obtain paths for subdirectories """ dirFiles = glob( f"{os.getcwd()}/{dir}/*/" ) # assign full output directories for csv file output_dirs = natsort.natsorted(dirFiles) num_dir = len(output_dirs) # number of outputs # marker_desc = ["-o", "-*", "-+", "-D", "-^", "-<", "-s", "-.", "-p", "-h"] background_colour = ["r","b","k","m","y"] """ Plots """ max_rate = 0 max_time = 0 max_time_ar = 0 max_rate_ar = 0 new_rate = [] marker_desc = ["-*","-o","--*","--o"] iter = 0 for x in range(layer_start, layer_end): for i in range(num_dir): # number of tails for same proc trail = os.path.basename( os.path.normpath(output_dirs[i]) ) # gives name of core_no.ofproc filename = output_dirs[i] + "output.csv" if f"_v{1}" in trail: # select particular v. rate, num_nodes, array = output_speedup(filename, layer_name[x]) rate_hdf5, num_nodes_hdf5, array_hdf5 = output_speedup(filename, layer_name[2]) # obtain hdf5 parameters for comparison new_rate = speedup_comp(rate_hdf5, rate) layer = f"{layer_name[x]}/{num_nodes.values[x]}" if num_nodes.values[0] == 1: # hacky way of selecting num_nodes. ax1.plot(array, new_rate, marker_desc[iter], c=background_colour[iter], label=layer) iter += 1 if num_nodes.values[0] == 384: # hacky way of selecting num_nodes. ax1.plot(array, new_rate, marker_desc[iter], c=background_colour[iter], label=layer) iter += 1 ax1.legend(loc = "upper right")# Don't allow the axis tdo be on top of your data ax1.set_xlabel("Global Size (GB)") ax1.set_ylabel("Speedup compared to HDF5") ax1.set_yscale("log") ax1.set_xscale("log") # fig1.suptitle("Benchmarking speedup results w.r.t. I/O rates for HDF5") fig1.tight_layout() def output_speedup( filename, layer_name ): # function to return array, time, rate from CSV file double_size = 8 mydata = pd.read_csv( filename, index_col=False, skiprows=0, names=[ "Fields", "Layername", "ArraySize", "NumNodes", "AverageTime", "AverageRate" ] ) """ Set array filter """ array = mydata.ArraySize[(mydata.Layername == layer_name) ] # N1 local array dimension num_nodes = mydata.NumNodes[(mydata.Layername == layer_name)] # number of nodes Global_size = ((array** 3) * double_size * num_nodes.values[0]) / (10 ** 9) # global array size from cube length # time = mydata.AverageTime[(mydata.Layername == layer_name) & (mydata.ArraySize == array_size)] # units sec rate = mydata.AverageRate[(mydata.Layername == layer_name)] # Gb/s return rate, num_nodes, Global_size def speedup_comp(rate_hdf5,rate): new_rate = [] for x in range(len(rate_hdf5)): new_rate.append(rate.values[x]/rate_hdf5.values[x]) return new_rate def xcompact(): """ Plot formatting """ fig1 = plt.figure(figsize=(6,5.5)) # fig1 = plt.figure(figsize=(8,6)) # plt.rcParams["font.size"] = "10" ax1 = plt.axes() xcom = pd.read_csv( "xcompact.csv", index_col=False, skiprows=0, names=[ "Rank", "TotalTime", "WriteTime", "ComputeTime", "BW" ] ) arrow_dim = 300 point_dim = 550 plt.arrow(x=point_dim+arrow_dim, y=60, dx=-arrow_dim, dy=0, width=1.5, head_length=30, facecolor='red') plt.annotate('I/O bottleneck', xy = (point_dim+arrow_dim+50, 58)) ax1.plot(xcom.Rank,xcom.TotalTime, "-^", c="r", label="Total") ax1.plot(xcom.Rank,xcom.WriteTime,"-o", c="b",label = "Write" ) ax1.plot(xcom.Rank,xcom.ComputeTime, "-<", c="k", label = "Compute") ax1.set_xlabel("MPI ranks") ax1.set_ylabel("Time taken (s)") ax1.legend(loc = "upper right") ax1.set_yscale("log") ax1.set_xscale("log") plt.axvline(x=512, color='k', linestyle='--') # fig1.suptitle("Benchmarking for XCompact3D") def outputdata_layer_v_size(filedir, cube_len_start, cube_len_end): """ Inits """ layer_name = ["Serial", "MPIIO", "PHDF5", "ADIOS2_HDF5", "ADIOS2_BP4"] rate_persize = [] target_dir = f"{os.getcwd()}/output_dirs/{filedir}/*/" layer_start, layer_end = 1, len(layer_name) """ Obtain paths for subdirectories """ dirFiles = glob( target_dir ) # assign full output directories for csv file output_dirs = natsort.natsorted(dirFiles) """ Find data points recursively from the target dir, selecting based on layer name """ for l in range(layer_start, layer_end): # select layer name for i in range(len(output_dirs)): filename = output_dirs[i] + "output.csv" trail = os.path.basename( os.path.normpath(output_dirs[i]) ) # gives name of core_no.ofproc for x in range (cube_len_start,cube_len_end): # specify max array size parameter N_size = 2 ** x rate, ranks = dict_output(f"{output_dirs[i]}output.csv",layer_name[l],N_size) # info from specified directory, against matching layername and cube length Global_size = ((N_size ** 3) * double_size * ranks.values[0]) / (10 ** 9) # global array size from cube length rate_persize.append([layer_name[l], ranks.values[0], N_size, Global_size, rate.values[0]]) # append plot output data to rate_persize # output of rate_persize list array to csv file output_data = pd.DataFrame(rate_persize, columns=[ 'LayerName','Ranks', 'CubeLen', 'GlobalSize', 'Rate']) out = output_data.to_csv(f"output_csv/{filedir}_output.csv", index=False) def compare_benchio_f(): # bar plot to compare benchio with benchmark_c """ Both run with 1 I/O runs and 1 rank. Array size = 250*250*250. Can increase this to multiple for better accuracy. """ """ Plot formatting """ # fig1 = plt.figure(figsize=(8,6)) fig1 = plt.figure(figsize=(6,5.5)) ax1 = plt.axes() """Data initialisation""" rate_c = np.empty(3) array = ( 256 * 256 * 256 * double_size / (10 ** 9) ) # for array size in GB with size of double = 8bytes """Data from benchio fortran""" # avg rate from benchio fortran computed in B/s f = open(f"{output_files}/nvram_benchiof.out", 'r') avg_rates_f = re.findall("avgrate\s=\s+(\d+\.\d+)", f.read()) rate_f = [x * (2 ** 20) / (10 ** 9) for x in [float(i) for i in avg_rates_f]] #convert units from MiB/s to GB/s """Data from benchio c""" for i in range(0, 3): # to compare serial, mpi, hdf5 rate_c[i], num_nodes_temp = dict_output(f"{output_files}/nvram.csv", layer_name[i], 256) """Bar plot""" index = np.arange(3) bar_width = 0.35 benchio_c = ax1.bar(index, rate_c, bar_width, label="benchmark_c") benchio_f = ax1.bar(index + bar_width, rate_f, bar_width, label="benchio") ax1.set_xlabel("Layers") ax1.set_ylabel("Averate Rate (GB/s)") ax1.legend() ax1.set_xticks(index + bar_width / 2) ax1.set_xticklabels(layer_name[0:3]) fig1.suptitle("NVRAM verification") fig1.tight_layout()
[ "pandas.read_csv", "numpy.arange", "matplotlib.pyplot.style.use", "os.getcwd", "os.path.normpath", "matplotlib.pyplot.annotate", "matplotlib.pyplot.figure", "matplotlib.pyplot.axvline", "matplotlib.pyplot.axes", "natsort.natsorted", "numpy.empty", "os.path.isfile", "pandas.DataFrame", "mat...
[((340, 371), 'matplotlib.pyplot.style.use', 'plt.style.use', (['"""style.mplstyle"""'], {}), "('style.mplstyle')\n", (353, 371), True, 'import matplotlib.pyplot as plt\n'), ((782, 922), 'pandas.read_csv', 'pd.read_csv', (['filename'], {'index_col': '(False)', 'skiprows': '(0)', 'names': "['Fields', 'Layername', 'ArraySize', 'NumNodes', 'AverageTime', 'AverageRate']"}), "(filename, index_col=False, skiprows=0, names=['Fields',\n 'Layername', 'ArraySize', 'NumNodes', 'AverageTime', 'AverageRate'])\n", (793, 922), True, 'import pandas as pd\n'), ((1656, 1683), 'natsort.natsorted', 'natsort.natsorted', (['dirFiles'], {}), '(dirFiles)\n', (1673, 1683), False, 'import natsort\n'), ((3323, 3418), 'pandas.DataFrame', 'pd.DataFrame', (['rate_persize'], {'columns': "['LayerName', 'Ranks', 'CubeLen', 'GlobalSize', 'Rate']"}), "(rate_persize, columns=['LayerName', 'Ranks', 'CubeLen',\n 'GlobalSize', 'Rate'])\n", (3335, 3418), True, 'import pandas as pd\n'), ((3871, 3958), 'pandas.read_csv', 'pd.read_csv', (['f"""output_csv/{target_dir}_avg_output.csv"""'], {'index_col': '(False)', 'skiprows': '(0)'}), "(f'output_csv/{target_dir}_avg_output.csv', index_col=False,\n skiprows=0)\n", (3882, 3958), True, 'import pandas as pd\n'), ((4795, 4823), 'matplotlib.pyplot.figure', 'plt.figure', ([], {'figsize': '(6, 5.5)'}), '(figsize=(6, 5.5))\n', (4805, 4823), True, 'import matplotlib.pyplot as plt\n'), ((4833, 4843), 'matplotlib.pyplot.axes', 'plt.axes', ([], {}), '()\n', (4841, 4843), True, 'import matplotlib.pyplot as plt\n'), ((6317, 6343), 'matplotlib.pyplot.figure', 'plt.figure', ([], {'figsize': '(8, 6)'}), '(figsize=(8, 6))\n', (6327, 6343), True, 'import matplotlib.pyplot as plt\n'), ((6397, 6407), 'matplotlib.pyplot.axes', 'plt.axes', ([], {}), '()\n', (6405, 6407), True, 'import matplotlib.pyplot as plt\n'), ((6587, 6614), 'natsort.natsorted', 'natsort.natsorted', (['dirFiles'], {}), '(dirFiles)\n', (6604, 6614), False, 'import natsort\n'), ((8639, 8779), 'pandas.read_csv', 'pd.read_csv', (['filename'], {'index_col': '(False)', 'skiprows': '(0)', 'names': "['Fields', 'Layername', 'ArraySize', 'NumNodes', 'AverageTime', 'AverageRate']"}), "(filename, index_col=False, skiprows=0, names=['Fields',\n 'Layername', 'ArraySize', 'NumNodes', 'AverageTime', 'AverageRate'])\n", (8650, 8779), True, 'import pandas as pd\n'), ((9698, 9726), 'matplotlib.pyplot.figure', 'plt.figure', ([], {'figsize': '(6, 5.5)'}), '(figsize=(6, 5.5))\n', (9708, 9726), True, 'import matplotlib.pyplot as plt\n'), ((9819, 9829), 'matplotlib.pyplot.axes', 'plt.axes', ([], {}), '()\n', (9827, 9829), True, 'import matplotlib.pyplot as plt\n'), ((9842, 9965), 'pandas.read_csv', 'pd.read_csv', (['"""xcompact.csv"""'], {'index_col': '(False)', 'skiprows': '(0)', 'names': "['Rank', 'TotalTime', 'WriteTime', 'ComputeTime', 'BW']"}), "('xcompact.csv', index_col=False, skiprows=0, names=['Rank',\n 'TotalTime', 'WriteTime', 'ComputeTime', 'BW'])\n", (9853, 9965), True, 'import pandas as pd\n'), ((10095, 10204), 'matplotlib.pyplot.arrow', 'plt.arrow', ([], {'x': '(point_dim + arrow_dim)', 'y': '(60)', 'dx': '(-arrow_dim)', 'dy': '(0)', 'width': '(1.5)', 'head_length': '(30)', 'facecolor': '"""red"""'}), "(x=point_dim + arrow_dim, y=60, dx=-arrow_dim, dy=0, width=1.5,\n head_length=30, facecolor='red')\n", (10104, 10204), True, 'import matplotlib.pyplot as plt\n'), ((10204, 10271), 'matplotlib.pyplot.annotate', 'plt.annotate', (['"""I/O bottleneck"""'], {'xy': '(point_dim + arrow_dim + 50, 58)'}), "('I/O bottleneck', xy=(point_dim + arrow_dim + 50, 58))\n", (10216, 10271), True, 'import matplotlib.pyplot as plt\n'), ((10645, 10690), 'matplotlib.pyplot.axvline', 'plt.axvline', ([], {'x': '(512)', 'color': '"""k"""', 'linestyle': '"""--"""'}), "(x=512, color='k', linestyle='--')\n", (10656, 10690), True, 'import matplotlib.pyplot as plt\n'), ((11113, 11129), 'glob.glob', 'glob', (['target_dir'], {}), '(target_dir)\n', (11117, 11129), False, 'from glob import glob\n'), ((11209, 11236), 'natsort.natsorted', 'natsort.natsorted', (['dirFiles'], {}), '(dirFiles)\n', (11226, 11236), False, 'import natsort\n'), ((12295, 12390), 'pandas.DataFrame', 'pd.DataFrame', (['rate_persize'], {'columns': "['LayerName', 'Ranks', 'CubeLen', 'GlobalSize', 'Rate']"}), "(rate_persize, columns=['LayerName', 'Ranks', 'CubeLen',\n 'GlobalSize', 'Rate'])\n", (12307, 12390), True, 'import pandas as pd\n'), ((12750, 12778), 'matplotlib.pyplot.figure', 'plt.figure', ([], {'figsize': '(6, 5.5)'}), '(figsize=(6, 5.5))\n', (12760, 12778), True, 'import matplotlib.pyplot as plt\n'), ((12788, 12798), 'matplotlib.pyplot.axes', 'plt.axes', ([], {}), '()\n', (12796, 12798), True, 'import matplotlib.pyplot as plt\n'), ((12843, 12854), 'numpy.empty', 'np.empty', (['(3)'], {}), '(3)\n', (12851, 12854), True, 'import numpy as np\n'), ((13525, 13537), 'numpy.arange', 'np.arange', (['(3)'], {}), '(3)\n', (13534, 13537), True, 'import numpy as np\n'), ((5433, 5493), 'matplotlib.lines.Line2D', 'Line2D', (['[0]', '[0]'], {'marker': '"""D"""', 'color': '"""r"""', 'lw': '(1)', 'label': '"""MPIIO"""'}), "([0], [0], marker='D', color='r', lw=1, label='MPIIO')\n", (5439, 5493), False, 'from matplotlib.lines import Line2D\n'), ((5501, 5561), 'matplotlib.lines.Line2D', 'Line2D', (['[0]', '[0]'], {'marker': '"""o"""', 'color': '"""b"""', 'lw': '(1)', 'label': '"""PHDF5"""'}), "([0], [0], marker='o', color='b', lw=1, label='PHDF5')\n", (5507, 5561), False, 'from matplotlib.lines import Line2D\n'), ((5568, 5634), 'matplotlib.lines.Line2D', 'Line2D', (['[0]', '[0]'], {'marker': '"""v"""', 'color': '"""k"""', 'lw': '(1)', 'label': '"""ADIOS2/HDF5"""'}), "([0], [0], marker='v', color='k', lw=1, label='ADIOS2/HDF5')\n", (5574, 5634), False, 'from matplotlib.lines import Line2D\n'), ((5640, 5705), 'matplotlib.lines.Line2D', 'Line2D', (['[0]', '[0]'], {'marker': '"""^"""', 'color': '"""m"""', 'lw': '(1)', 'label': '"""ADIOS2/BP4"""'}), "([0], [0], marker='^', color='m', lw=1, label='ADIOS2/BP4')\n", (5646, 5705), False, 'from matplotlib.lines import Line2D\n'), ((10958, 10969), 'os.getcwd', 'os.getcwd', ([], {}), '()\n', (10967, 10969), False, 'import os\n'), ((1550, 1561), 'os.getcwd', 'os.getcwd', ([], {}), '()\n', (1559, 1561), False, 'import os\n'), ((6493, 6504), 'os.getcwd', 'os.getcwd', ([], {}), '()\n', (6502, 6504), False, 'import os\n'), ((7153, 7185), 'os.path.normpath', 'os.path.normpath', (['output_dirs[i]'], {}), '(output_dirs[i])\n', (7169, 7185), False, 'import os\n'), ((11554, 11586), 'os.path.normpath', 'os.path.normpath', (['output_dirs[i]'], {}), '(output_dirs[i])\n', (11570, 11586), False, 'import os\n'), ((1998, 2030), 'os.path.normpath', 'os.path.normpath', (['output_dirs[i]'], {}), '(output_dirs[i])\n', (2014, 2030), False, 'import os\n'), ((2554, 2579), 'os.path.isfile', 'os.path.isfile', (['avgd_file'], {}), '(avgd_file)\n', (2568, 2579), False, 'import os\n')]
#!/usr/bin/env python # Copyright (c) 2015 SnapDisco Pty Ltd, Australia. # All rights reserved. # # This source code is licensed under the terms of the MIT license # found in the "LICENSE" file in the root directory of this source tree. import _refcount import numpy as np import sys def refcount(obj): # Subtract 3 from the refcount, because each function call adds 1. # (+1 for this func, +1 for `sys.getrefcount`, +1 for some other reason? # Perhaps the 3rd +1 is for the function implementing the `-` operator?) return sys.getrefcount(obj) - 3 def check_refcount(varname, var, expected_rc): rc = refcount(var) - 2 # and -2 more for this function... print("Get refcount(%s) -> %d" % (varname, rc)) assertion = "refcount(%s) == %d" % (varname, expected_rc) assert (rc == expected_rc), assertion print("TEST PASSED: %s" % assertion) a = np.arange(20, dtype=np.int32).reshape((5, 4)) print("a =") print(a) check_refcount("a", a, 1) print("\nb = _refcount.createsome(a)") print("> Crossing into Nim...") b = _refcount.createsome(a) print("\n... and we're back in Python") check_refcount("a", a, 1) check_refcount("b", b, 1) print("b =") print(b) print("\n------------\n") c = np.arange(20, 40, dtype=np.int32).reshape((5, 4)) print("c =") print(c) check_refcount("c", c, 1) print("\nd = _refcount.identity(c)") print("> Crossing into Nim...") d = _refcount.identity(c) print("\n... and we're back in Python") check_refcount("c", c, 2) check_refcount("d", d, 2) print("\n------------\n") e = np.arange(20, 40, dtype=np.int32).reshape((5, 4)) print("e =") print(e) check_refcount("e", e, 1) print("\n_refcount.identity(e)") print("> Crossing into Nim...") _refcount.identity(c) print("\n... and we're back in Python") check_refcount("e", e, 1) print("\n------------\n") print("f = _refcount.identity(np.arange(...))") print("> Crossing into Nim...") f = _refcount.identity(np.arange(40, 60, dtype=np.int32).reshape((5, 4))) print("\n... and we're back in Python") check_refcount("f", f, 1) print("\n------------\n") g = np.arange(60, 80, dtype=np.int32).reshape((5, 4)) h = np.arange(80, 100, dtype=np.int32).reshape((5, 4)) print("g =") print(g) check_refcount("g", g, 1) print("\nh =") print(h) check_refcount("h", h, 1) print("\ni = _refcount.twogoinonecomesout(g, h)") print("> Crossing into Nim...") i = _refcount.twogoinonecomesout(g, h) print("\n... and we're back in Python") check_refcount("g", g, 1) check_refcount("h", h, 2) check_refcount("i", i, 2) print("i =") print(i)
[ "_refcount.twogoinonecomesout", "_refcount.identity", "_refcount.createsome", "sys.getrefcount", "numpy.arange" ]
[((1052, 1075), '_refcount.createsome', '_refcount.createsome', (['a'], {}), '(a)\n', (1072, 1075), False, 'import _refcount\n'), ((1392, 1413), '_refcount.identity', '_refcount.identity', (['c'], {}), '(c)\n', (1410, 1413), False, 'import _refcount\n'), ((1700, 1721), '_refcount.identity', '_refcount.identity', (['c'], {}), '(c)\n', (1718, 1721), False, 'import _refcount\n'), ((2355, 2389), '_refcount.twogoinonecomesout', '_refcount.twogoinonecomesout', (['g', 'h'], {}), '(g, h)\n', (2383, 2389), False, 'import _refcount\n'), ((543, 563), 'sys.getrefcount', 'sys.getrefcount', (['obj'], {}), '(obj)\n', (558, 563), False, 'import sys\n'), ((883, 912), 'numpy.arange', 'np.arange', (['(20)'], {'dtype': 'np.int32'}), '(20, dtype=np.int32)\n', (892, 912), True, 'import numpy as np\n'), ((1221, 1254), 'numpy.arange', 'np.arange', (['(20)', '(40)'], {'dtype': 'np.int32'}), '(20, 40, dtype=np.int32)\n', (1230, 1254), True, 'import numpy as np\n'), ((1537, 1570), 'numpy.arange', 'np.arange', (['(20)', '(40)'], {'dtype': 'np.int32'}), '(20, 40, dtype=np.int32)\n', (1546, 1570), True, 'import numpy as np\n'), ((2066, 2099), 'numpy.arange', 'np.arange', (['(60)', '(80)'], {'dtype': 'np.int32'}), '(60, 80, dtype=np.int32)\n', (2075, 2099), True, 'import numpy as np\n'), ((2120, 2154), 'numpy.arange', 'np.arange', (['(80)', '(100)'], {'dtype': 'np.int32'}), '(80, 100, dtype=np.int32)\n', (2129, 2154), True, 'import numpy as np\n'), ((1918, 1951), 'numpy.arange', 'np.arange', (['(40)', '(60)'], {'dtype': 'np.int32'}), '(40, 60, dtype=np.int32)\n', (1927, 1951), True, 'import numpy as np\n')]
""" Unit and regression test for the measure module. """ # Import package, test suite, and other packages as needed import molecool import pytest import sys import numpy as np def test_calculate_distance(): """test will pass so long as `calculate_distance` function within `measure` module run as expected""" r1 = np.array([1, 0, 0]) r2 = np.array([0, 0, 0]) expected_distance = 1.0 calculated_distance = molecool.calculate_distance(r1, r2) assert calculated_distance == expected_distance def test_calculate_distance_typeerror(): """test will pass so long as calling `calculate_distance` function within `measure` module does not raise TypeError""" r1 = [1, 0, 0] r2 = [0, 0, 0] with pytest.raises(TypeError): calculated_distance = molecool.calculate_distance(r1, r2) def test_calculate_angle(): """test will pass so long as `calculate_angle` function within `measure` module run as expected""" r1 = np.array([1, 0, 0]) r2 = np.array([0, 0, 0]) r3 = np.array([0, 1, 0]) expected_angle_in_degrees = 90 calculated_angle_in_degrees = molecool.calculate_angle(r1, r2, r3, degrees=True) assert calculated_angle_in_degrees == expected_angle_in_degrees @pytest.mark.parametrize("p1, p2, p3, expected_angle_in_degrees", [ (np.array([np.sqrt(2)/2, np.sqrt(2)/2, 0]), np.array([0, 0, 0]), np.array([0, 1, 0]), 45), (np.array([1, 0, 0]), np.array([0, 0, 0]), np.array([0, 1, 0]), 90) ]) def test_calculate_angle_many(p1, p2, p3, expected_angle_in_degrees): """test will pass so long as `calculate_angle` function within `measure` module run as expected""" calculated_angle_in_degrees = molecool.calculate_angle(p1, p2, p3, degrees=True) assert calculated_angle_in_degrees == expected_angle_in_degrees
[ "molecool.calculate_distance", "numpy.sqrt", "molecool.calculate_angle", "numpy.array", "pytest.raises" ]
[((330, 349), 'numpy.array', 'np.array', (['[1, 0, 0]'], {}), '([1, 0, 0])\n', (338, 349), True, 'import numpy as np\n'), ((359, 378), 'numpy.array', 'np.array', (['[0, 0, 0]'], {}), '([0, 0, 0])\n', (367, 378), True, 'import numpy as np\n'), ((434, 469), 'molecool.calculate_distance', 'molecool.calculate_distance', (['r1', 'r2'], {}), '(r1, r2)\n', (461, 469), False, 'import molecool\n'), ((978, 997), 'numpy.array', 'np.array', (['[1, 0, 0]'], {}), '([1, 0, 0])\n', (986, 997), True, 'import numpy as np\n'), ((1007, 1026), 'numpy.array', 'np.array', (['[0, 0, 0]'], {}), '([0, 0, 0])\n', (1015, 1026), True, 'import numpy as np\n'), ((1036, 1055), 'numpy.array', 'np.array', (['[0, 1, 0]'], {}), '([0, 1, 0])\n', (1044, 1055), True, 'import numpy as np\n'), ((1126, 1176), 'molecool.calculate_angle', 'molecool.calculate_angle', (['r1', 'r2', 'r3'], {'degrees': '(True)'}), '(r1, r2, r3, degrees=True)\n', (1150, 1176), False, 'import molecool\n'), ((1688, 1738), 'molecool.calculate_angle', 'molecool.calculate_angle', (['p1', 'p2', 'p3'], {'degrees': '(True)'}), '(p1, p2, p3, degrees=True)\n', (1712, 1738), False, 'import molecool\n'), ((740, 764), 'pytest.raises', 'pytest.raises', (['TypeError'], {}), '(TypeError)\n', (753, 764), False, 'import pytest\n'), ((796, 831), 'molecool.calculate_distance', 'molecool.calculate_distance', (['r1', 'r2'], {}), '(r1, r2)\n', (823, 831), False, 'import molecool\n'), ((1358, 1377), 'numpy.array', 'np.array', (['[0, 0, 0]'], {}), '([0, 0, 0])\n', (1366, 1377), True, 'import numpy as np\n'), ((1379, 1398), 'numpy.array', 'np.array', (['[0, 1, 0]'], {}), '([0, 1, 0])\n', (1387, 1398), True, 'import numpy as np\n'), ((1406, 1425), 'numpy.array', 'np.array', (['[1, 0, 0]'], {}), '([1, 0, 0])\n', (1414, 1425), True, 'import numpy as np\n'), ((1427, 1446), 'numpy.array', 'np.array', (['[0, 0, 0]'], {}), '([0, 0, 0])\n', (1435, 1446), True, 'import numpy as np\n'), ((1448, 1467), 'numpy.array', 'np.array', (['[0, 1, 0]'], {}), '([0, 1, 0])\n', (1456, 1467), True, 'import numpy as np\n'), ((1325, 1335), 'numpy.sqrt', 'np.sqrt', (['(2)'], {}), '(2)\n', (1332, 1335), True, 'import numpy as np\n'), ((1339, 1349), 'numpy.sqrt', 'np.sqrt', (['(2)'], {}), '(2)\n', (1346, 1349), True, 'import numpy as np\n')]
#<NAME> #Last Modified: 12/15/2015 import numpy as np import matplotlib as mpl from ..calculations import HexDhCal from scipy.integrate import trapz Tbulk = np.array([1,0]) def Tbulk(z,FluxPro,Tbulkin,NFuel,qlin,Cp,Uinlet,rho,HaF): for i in range(2,8): Tbulk = Tbulkin + (np.array(trapz(z,None,i,0),FluxPro(np.array([0,i])))*NFuel*qlin)/(Cp*Uinlet*rho*HexDhCal.HaF(HexDhCal.Ha(Ac),NFuel,FoCD,WoD)) #Bulk Temperature of Coolant - C return Tbulk print(Tbulk)
[ "numpy.array", "scipy.integrate.trapz" ]
[((173, 189), 'numpy.array', 'np.array', (['[1, 0]'], {}), '([1, 0])\n', (181, 189), True, 'import numpy as np\n'), ((309, 329), 'scipy.integrate.trapz', 'trapz', (['z', 'None', 'i', '(0)'], {}), '(z, None, i, 0)\n', (314, 329), False, 'from scipy.integrate import trapz\n'), ((335, 351), 'numpy.array', 'np.array', (['[0, i]'], {}), '([0, i])\n', (343, 351), True, 'import numpy as np\n')]
""" formation evaluation Author: <NAME> Email: <EMAIL> """ import lasio import numpy as np import matplotlib.pyplot as plt import pandas as pd import matplotlib.ticker as ticker import warnings plt.style.use('ggplot') warnings.filterwarnings("ignore") class formation_eval: """ Evaluates the formation and determines formation characteristic such as shale volume, reservoir and non-reservoir zones. args:: datapath: LAS datapath mnemonics: list of well log mnemonics, if 'None', density, neutron, Gamma ray, SP and resistivity logs are passed if available. """ def __init__(self, datapath: str = None, mnemonics: list = None): """ :type datapath: str :type mnemonics: list """ if mnemonics is None: self.mnemonics = mnemonics if datapath is None: pass else: self.datapath = datapath def read_lasio(self): """ reads the .LAS file and converts to a dataframe :return: log data description and log dataframe """ path = self.datapath lasfile = lasio.read(path) df = lasfile.df() if 'D' or 'E' or 'P' in df.index.name: df.reset_index(inplace=True) return lasfile.header, df def well_logs(self, dataframe: pd.DataFrame): """ filters the dataframe and returns a dataframe with the specified mnemonics or necessary and available mnemonics :param dataframe: pandas dataframe object :return: well log dataframe """ if self.mnemonics is None: logs = ['DEPTH', 'ROP', 'GR', 'SP', 'CALI', 'BS', 'RD', 'RT', 'RM', 'RS', 'NPHI', 'CNL', 'RHOB', 'DRHO', 'PHID', 'DT', 'PEF'] logsnew = [] for i in logs: if i in dataframe.columns: logsnew.append(i) else: logs = self.mnemonics logsnew = [] for i in logs: if i in dataframe.columns: logsnew.append(i) well_logs = logsnew if 'GR' in well_logs: dataframe['GR'] = np.where(dataframe['GR'] > 150, 150, dataframe['GR']) dataframe['GR'] = np.where(dataframe['GR'] < 0, 0, dataframe['GR']) if 'NPHI' in well_logs: dataframe['NPHI'] = np.where(dataframe['NPHI'] > 0.50, 0.50, dataframe['NPHI']) dataframe['NPHI'] = np.where(dataframe['NPHI'] < -0.15, -0.15, dataframe['NPHI']) return dataframe[well_logs] @staticmethod def lognan_cleaning(dataframe: pd.DataFrame, fill_value: int or float = None): """ This fills the missing numbers in the dataframe :param fill_value: value to replace missing number with, if 'None', mean is used. :param dataframe: pandas dataframe :rtype: pd.Dataframe """ df = dataframe.copy() if fill_value is None: df.fillna(np.mean(df), inplace=True) else: df.fillna(fill_value, axis=1, inplace=True) return df @staticmethod def log_viz(data, min_depth: float or int or None = None, max_depth: float or int or None = None, plotsize: tuple = None): """ well log plots :param data: well logs dataframe :param min_depth: top of reservoir, closer to the surface (length units) :param max_depth: bottom of reservoir, closer to the subsurface (length units) :param plotsize: the plot figsize in tuple form """ if plotsize is None: plotsize = (18, 15) logs = data.columns fig, ax = plt.subplots(nrows=1, ncols=6, figsize=plotsize, sharey=False) fig.suptitle("Logs Visualization", fontsize=22, y=1.02) # General setting for all axis for axes in ax: axes.get_xaxis().set_visible(False) axes.invert_yaxis() axes.spines['left'].set_color('k') axes.spines['right'].set_color('k') axes.minorticks_on() if min_depth and max_depth is not None: axes.set_ylim(max_depth, min_depth) else: axes.set_ylim(data['DEPTH'].max(), data['DEPTH'].min()) # 1st track: CALI, BS if 'CALI' and 'BS' not in logs: fig.delaxes(ax=ax[0]) ax[1].set_ylim(max_depth, min_depth) else: ax[0].minorticks_on() ax[0].grid(b=True, which='major', color='black', linestyle='--') ax[0].grid(b=True, which='minor', color='grey', linestyle=':') if 'CALI' in logs: cali = ax[0].twiny() cali.minorticks_on() cali.set_xlim(6, 26) cali.plot(data.CALI, data.DEPTH, label='CALI[in]', color='red') cali.spines['top'].set_position(('outward', 20)) cali.spines['top'].set_color('r') cali.set_xlabel('CALI[in]', color='red') cali.tick_params(axis='x', colors='red') cali.grid(b=True, which='major', color='k', linestyle='--') cali.grid(b=True, which='minor', color='grey', linestyle=':') else: pass if 'BS' in logs: bs = ax[0].twiny() bs.minorticks_on() bs.set_xlim(6, 26) bs.plot(data.BS, data.DEPTH, label='BS[in]', color='y') bs.spines['top'].set_position(('outward', 60)) bs.spines['top'].set_color('y') bs.set_xlabel('BS[in]', color='y') bs.tick_params(axis='x', colors='y') bs.grid(b=True, which='major', color='k', linestyle='--') bs.grid(b=True, which='minor', color='grey', linestyle=':') else: pass # 2nd track: GR, SP if 'CALI' and 'BS' not in logs: ax[1].set_ylim(max_depth, min_depth) ax[1].minorticks_on() ax[1].grid(b=True, which='major', color='black', linestyle='--') ax[1].grid(b=True, which='minor', color='grey', linestyle=':') if 'GR' in logs: gr = ax[1].twiny() gr.minorticks_on() gr.set_xlim(0, 150) gr.plot(data.GR, data.DEPTH, label='GR[api]', color='green') gr.spines['top'].set_position(('outward', 20)) gr.spines['top'].set_color('g') gr.set_xlabel('GR[api]', color='green') gr.tick_params(axis='x', colors='green') gr.grid(b=True, which='major', color='k', linestyle='--') gr.grid(b=True, which='minor', color='grey', linestyle=':') if 'SP' in logs: sp = ax[1].twiny() sp.minorticks_on() sp.set_xlim(data['SP'].min(), data.max()) sp.plot(data.GR, data.DEPTH, label='SP[mV]', color='b') sp.spines['top'].set_position(('outward', 60)) sp.spines['top'].set_color('b') sp.set_xlabel('GR[api]', color='b') sp.tick_params(axis='x', colors='b') sp.grid(b=True, which='major', color='k', linestyle='--') sp.grid(b=True, which='minor', color='grey', linestyle=':') # 3rd track: resistivity track ax[2].grid(b=True, which='major', color='black', linestyle='--') ax[2].grid(b=True, which='minor', color='grey', linestyle=':') ax[2].minorticks_on() if 'RD' in logs: rd = ax[2].twiny() rd.minorticks_on() rd.set_xlim(0.2, 2500) rd.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=3)) rd.spines['top'].set_position(('outward', 10)) rd.spines['top'].set_color('k') rd.semilogx(data.RD, data.DEPTH, '--', linewidth=1, c='black') rd.set_xlabel('RD [ohm.m]', color='black') rd.tick_params(axis='x', colors='black') rd.grid(b=True, which='major', color='black', linestyle='--') rd.grid(b=True, which='minor', color='grey', linestyle=':') if 'RS' in logs: rs = ax[2].twiny() rs.minorticks_on() rs.set_xlim(0.2, 2500) rs.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=3)) rs.spines['top'].set_position(('outward', 50)) rs.spines['top'].set_color('b') rs.semilogx(data.RS, data.DEPTH, linewidth=1, c='b', ) rs.set_xlabel('RS [ohm.m]', color='b') rs.tick_params(axis='x', colors='b') rs.grid(b=True, which='major', color='black', linestyle='--') rs.grid(b=True, which='minor', color='grey', linestyle=':') if 'RM' in logs: rm = ax[2].twiny() rm.minorticks_on() rm.set_xlim(0.2, 2500) rm.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=3)) rm.spines['top'].set_position(('outward', 90)) rm.spines['top'].set_color('gold') rm.semilogx(data.RM, data.DEPTH, linewidth=1, c='gold', ) rm.set_xlabel('RM [ohm.m]', color='gold') rm.tick_params(axis='x', colors='gold') rm.grid(b=True, which='major', color='black', linestyle='--') rm.grid(b=True, which='minor', color='grey', linestyle=':') if 'RT' in logs: rt = ax[2].twiny() rt.minorticks_on() rt.set_xlim(0.2, 2500) rt.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=3)) rt.spines['top'].set_position(('outward', 130)) rt.spines['top'].set_color('brown') rt.semilogx(data.RT, data.DEPTH, linewidth=1, c='brown', ) rt.set_xlabel('RT [ohm.m]', color='brown') rt.tick_params(axis='x', colors='brown') rt.grid(b=True, which='major', color='black', linestyle='--') rt.grid(b=True, which='minor', color='grey', linestyle=':') if 'RXO' in logs: rxo = ax[2].twiny() rxo.minorticks_on() rxo.set_xlim(0.2, 2500) rxo.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=3)) rxo.spines['top'].set_position(('outward', 170)) rxo.spines['top'].set_color('c') rxo.semilogx(data.RXO, data.DEPTH, linewidth=1, c='c', ) rxo.set_xlabel('RXO [ohm.m]', color='c') rxo.tick_params(axis='x', colors='c') rxo.grid(b=True, which='major', color='black', linestyle='--') rxo.grid(b=True, which='minor', color='grey', linestyle=':') # 4th track NPHI, DPHI, RHOB ax[3].minorticks_on() ax[3].grid(b=True, which='major', color='black', linestyle='--') ax[3].grid(b=True, which='minor', color='grey', linestyle=':') if 'NPHI' in logs: nphi = ax[3].twiny() nphi.minorticks_on() nphi.set_xlim(0.45, -0.15) nphi.spines['top'].set_position(('outward', 20)) nphi.spines['top'].set_color('blue') nphi.set_xlabel("v/v") nphi.plot(data.NPHI, data.DEPTH, linewidth=1, label='v/v', color='blue') nphi.set_xlabel('NPHI [v/v]', color='blue') nphi.tick_params(axis='x', colors='blue') nphi.xaxis.set_major_locator(plt.MultipleLocator(0.2)) nphi.grid(b=True, which='major', color='black', linestyle='--') nphi.grid(b=True, which='minor', color='grey', linestyle=':') if 'RHOB' in logs: rhob = ax[3].twiny() rhob.set_xlim(1.95, 2.95) rhob.plot(data.RHOB, data.DEPTH, '--', linewidth=1, label='g/cm^3', color='red') rhob.spines['top'].set_position(('outward', 60)) rhob.spines['top'].set_color('red') rhob.set_xlabel('RHOB [g/cm^3]', color='red') rhob.tick_params(axis='x', colors='red') rhob.xaxis.set_major_locator(plt.MultipleLocator(0.4)) elif 'PHID' in logs: phid = ax[3].twiny() phid.set_xlim(0.45, -0.15) phid.plot(data.PHID, data.DEPTH, '--', linewidth=1, label='%', color='red') phid.spines['top'].set_position(('outward', 60)) phid.spines['top'].set_color('red') phid.set_xlabel('PHID [%]', color='red') phid.tick_params(axis='x', colors='red') phid.xaxis.set_major_locator(plt.MultipleLocator(0.4)) if 'NPHI' and 'RHOB' in logs: # https://stackoverflow.com/questions/57766457/how-to-plot-fill-betweenx-to-fill-the-area-between-y1-and-y2-with-different-scal x2p, _ = (rhob.transData + nphi.transData.inverted()).transform(np.c_[data.RHOB, data.DEPTH]).T nphi.autoscale(False) nphi.fill_betweenx(data.DEPTH, data.NPHI, x2p, color="goldenrod", alpha=0.4, where=(x2p > data.NPHI)) nphi.fill_betweenx(data.DEPTH, data.NPHI, x2p, color="turquoise", alpha=0.4, where=(x2p < data.NPHI)) # 5th DT and PEF if 'PEF' and 'DT' not in logs: fig.delaxes(ax=ax[4]) else: ax[4].minorticks_on() ax[4].grid(b=True, which='major', color='black', linestyle='--') ax[4].grid(b=True, which='minor', color='grey', linestyle=':') if 'DT' in logs: dt = ax[4].twiny() dt.minorticks_on() dt.set_xlim(200, 40) dt.spines['top'].set_position(('outward', 20)) dt.spines['top'].set_color('c') dt.plot(data.DT, data.DEPTH, linewidth=1, label="US/F", color='c') dt.set_xlabel("DT", color='c') dt.tick_params(axis='x', colors='c') dt.grid(b=True, which='major', color='black', linestyle='--') dt.grid(b=True, which='minor', color='grey', linestyle=':') else: pass if 'PEF' in logs: pef = ax[4].twiny() pef.plot(data.PEF, data.DEPTH, '--', linewidth=1, label="b/elc", color='lime') pef.spines['top'].set_position(('outward', 60)) pef.spines['top'].set_color('lime') pef.set_xlabel("PEF", color='lime') pef.tick_params(axis='x', colors='lime') pef.grid(b=True, which='major', color='black', linestyle='--') pef.grid(b=True, which='minor', color='grey', linestyle=':') else: pass # 6th track: vsh_larionov, vsh_linear if 'vsh_linear' and 'vsh_larionov' not in logs: fig.delaxes(ax=ax[5]) else: ax[5].minorticks_on() ax[5].grid(b=True, which='major', color='black', linestyle='--') ax[5].grid(b=True, which='minor', color='grey', linestyle=':') if 'vsh_linear' in logs: vsh_linear = ax[5].twiny() vsh_linear.minorticks_on() vsh_linear.plot(data.vsh_linear, data.DEPTH, label='CALI[in]', color='k') vsh_linear.spines['top'].set_position(('outward', 20)) vsh_linear.spines['top'].set_color('k') vsh_linear.set_xlabel('vsh_linear[%]', color='k') vsh_linear.tick_params(axis='x', colors='k') vsh_linear.grid(b=True, which='major', color='black', linestyle='--') vsh_linear.grid(b=True, which='minor', color='grey', linestyle=':') else: pass if 'vsh_larionov' in logs: vsh_larionov = ax[5].twiny() vsh_larionov.minorticks_on() vsh_larionov.plot(data.vsh_larionov, data.DEPTH, label='BS[in]', color='brown') vsh_larionov.spines['top'].set_position(('outward', 60)) vsh_larionov.spines['top'].set_color('brown') vsh_larionov.set_xlabel('vsh_larionov[%]', color='brown') vsh_larionov.tick_params(axis='x', colors='brown') vsh_larionov.grid(b=True, which='major', color='black', linestyle='--') vsh_larionov.grid(b=True, which='minor', color='grey', linestyle=':') else: pass if 'PHIE' in logs: phie = ax[5].twiny() phie.minorticks_on() phie.plot(data.PHIE, data.DEPTH, linewidth=1, label='%', color='indigo') phie.set_xlim(1, 0.0) phie.spines['top'].set_position(('outward', 100)) phie.spines['top'].set_color('indigo') phie.set_xlabel('PHIE [%]', color='indigo') phie.tick_params(axis='x', colors='indigo') else: pass plt.tight_layout(pad=2, h_pad=10, w_pad=2) @staticmethod def triple_combo_plot(data, min_depth: float or int or None, max_depth: float or int or None, plotsize: tuple = None): """ This gives the 'sp-gr', 'resistivity' and 'neutron-density' log plot :param data: well logs dataframe :param min_depth: top of reservoir, closer to the surface (length units) :param max_depth: bottom of reservoir, closer to the subsurface (length units) :param plotsize: the plot figsize in tuple form, default is (14,22) """ if plotsize is None: plotsize = (12, 15) logs = data.columns fig, ax = plt.subplots(nrows=1, ncols=3, figsize=plotsize, sharey='all') fig.suptitle("Triple-Combo Plot", fontsize=22, y=1.02) # General setting for all axis for axes in ax: axes.get_xaxis().set_visible(False) axes.invert_yaxis() axes.spines['left'].set_color('k') axes.spines['right'].set_color('k') if min_depth is not None: axes.set_ylim(max_depth, min_depth) else: axes.set_ylim(data['DEPTH'].max(), data['DEPTH'].min()) # 1st track: GR, SP ax[0].minorticks_on() ax[0].grid(b=True, which='major', color='black', linestyle='--') ax[0].grid(b=True, which='minor', color='grey', linestyle=':') # 1st track: GR, SP if 'GR' in logs: gr = ax[0].twiny() gr.minorticks_on() gr.set_xlim(0, 150) gr.plot(data.GR, data.DEPTH, label='GR[api]', color='green') gr.spines['top'].set_position(('outward', 20)) gr.spines['top'].set_color('g') gr.set_xlabel('GR[api]', color='green') gr.tick_params(axis='x', colors='green') gr.grid(b=True, which='major', color='k', linestyle='--') gr.grid(b=True, which='minor', color='grey', linestyle=':') if 'SP' in logs: sp = ax[0].twiny() sp.minorticks_on() sp.set_xlim(data['SP'].min(), data.max()) sp.plot(data.GR, data.DEPTH, label='SP[mV]', color='b') sp.spines['top'].set_position(('outward', 60)) sp.spines['top'].set_color('b') sp.set_xlabel('GR[api]', color='b') sp.tick_params(axis='x', colors='b') sp.grid(b=True, which='major', color='k', linestyle='--') sp.grid(b=True, which='minor', color='grey', linestyle=':') # 2nd track: resistivity track ax[1].minorticks_on() ax[1].grid(b=True, which='major', color='black', linestyle='--') ax[1].grid(b=True, which='minor', color='grey', linestyle=':') if 'RD' in logs: rd = ax[1].twiny() rd.set_xlim(0.2, 2500) rd.spines['top'].set_position(('outward', 10)) rd.spines['top'].set_color('y') rd.semilogx(data.RD, data.DEPTH, '--', linewidth=1, c='y') rd.set_xlabel('RD [ohm.m]', color='y') rd.tick_params(axis='x', colors='y') rd.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=3)) rd.grid(b=True, which='major', color='black', linestyle='--') rd.grid(b=True, which='minor', color='grey', linestyle=':') if 'RS' in logs: rs = ax[1].twiny() rs.set_xlim(0.2, 2500) rs.spines['top'].set_position(('outward', 50)) rs.spines['top'].set_color('m') rs.semilogx(data.RS, data.DEPTH, linewidth=1, c='m', ) rs.set_xlabel('RS [ohm.m]', color='m') rs.tick_params(axis='x', colors='m') rs.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=3)) rs.grid(b=True, which='major', color='black', linestyle='--') rs.grid(b=True, which='minor', color='grey', linestyle=':') if 'RM' in logs: rm = ax[1].twiny() rm.set_xlim(0.2, 2500) rm.spines['top'].set_position(('outward', 90)) rm.spines['top'].set_color('C1') rm.semilogx(data.RM, data.DEPTH, linewidth=1, c='C1', ) rm.set_xlabel('RM [ohm.m]', color='C1') rm.tick_params(axis='x', colors='C1') rm.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=3)) rm.grid(b=True, which='major', color='black', linestyle='--') rm.grid(b=True, which='minor', color='grey', linestyle=':') if 'RT' in logs: rt = ax[1].twiny() rt.set_xlim(0.2, 2500) rt.spines['top'].set_position(('outward', 130)) rt.spines['top'].set_color('brown') rt.semilogx(data.RT, data.DEPTH, linewidth=1, c='brown', ) rt.set_xlabel('RT [ohm.m]', color='brown') rt.tick_params(axis='x', colors='brown') rt.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=3)) if 'RXO' in logs: rxo = ax[1].twiny() rxo.set_xlim(0.2, 2500) rxo.spines['top'].set_position(('outward', 170)) rxo.spines['top'].set_color('c') rxo.semilogx(data.RXO, data.DEPTH, linewidth=1, c='c', ) rxo.set_xlabel('RXO [ohm.m]', color='c') rxo.tick_params(axis='x', colors='c') rxo.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=3)) # 3rd track NPHI, DPHI, RHOB ax[2].minorticks_on() ax[2].grid(b=True, which='major', color='black', linestyle='--') ax[2].grid(b=True, which='minor', color='grey', linestyle=':') if 'NPHI' in logs: nphi = ax[2].twiny() nphi.minorticks_on() nphi.set_xlim(0.45, -0.15) nphi.spines['top'].set_position(('outward', 20)) nphi.spines['top'].set_color('blue') nphi.set_xlabel("v/v") nphi.plot(data.NPHI, data.DEPTH, linewidth=1, label='v/v', color='blue') nphi.set_xlabel('NPHI [v/v] ', color='blue') nphi.tick_params(axis='x', colors='blue') nphi.xaxis.set_major_locator(plt.MultipleLocator(0.2)) nphi.grid(b=True, which='major', color='black', linestyle='--') nphi.grid(b=True, which='minor', color='grey', linestyle=':') if 'RHOB' in logs: rhob = ax[2].twiny() rhob.set_xlim(1.95, 2.95) rhob.plot(data.RHOB, data.DEPTH, '--', linewidth=1, label='g/cm^3', color='red') rhob.spines['top'].set_position(('outward', 60)) rhob.spines['top'].set_color('red') rhob.set_xlabel('RHOB [g/cm^3]', color='red') rhob.tick_params(axis='x', colors='red') rhob.xaxis.set_major_locator(plt.MultipleLocator(0.4)) elif 'DPHI' in logs: dphi = ax[2].twiny() dphi.set_xlim(0.45, -0.15) dphi.plot(data.DPHI, data.DEPTH, '--', linewidth=1, label='%', color='red') dphi.spines['top'].set_position(('outward', 60)) dphi.spines['top'].set_color('red') dphi.set_xlabel('DPHI [%]', color='red') dphi.tick_params(axis='x', colors='red') dphi.xaxis.set_major_locator(plt.MultipleLocator(0.4)) if 'NPHI' and 'RHOB' in logs: # https://stackoverflow.com/questions/57766457/how-to-plot-fill-betweenx-to-fill-the-area-between-y1-and-y2-with-different-scal x2p, _ = (rhob.transData + nphi.transData.inverted()).transform(np.c_[data.RHOB, data.DEPTH]).T nphi.autoscale(False) nphi.fill_betweenx(data.DEPTH, data.NPHI, x2p, color="y", alpha=0.4, where=(x2p > data.NPHI)) nphi.fill_betweenx(data.DEPTH, data.NPHI, x2p, color="turquoise", alpha=0.4, where=(x2p < data.NPHI)) plt.tight_layout(pad=2, h_pad=10, w_pad=2) @staticmethod def doub_logplot(data, logs: list, reslog: list or None = None, min_depth: float or int or None = None, max_depth: float or int or None = None, plotsize: tuple = None): """ This gives a combination plot of your choice :param logs: name of logs to plot in a list :param reslog: name of resistivity logs to plot in a list :param data: well logs dataframe :param min_depth: top of reservoir, closer to the surface (length units) :param max_depth: bottom of reservoir, closer to the subsurface (length units) :param plotsize: the plot figsize in tuple form, default is (14, 22) """ if plotsize is None: plotsize = (17, 15) if reslog is None: reslog = [] total = len(logs) + len(reslog) total_logs = logs + reslog # create the subplots; ncols equals the number of logs fig, ax = plt.subplots(nrows=1, ncols=total, figsize=plotsize) # General setting for all axis for axes in ax: axes.invert_yaxis() for xtick in axes.get_xticklabels(): xtick.set_fontsize(10) for ytick in axes.get_yticklabels(): ytick.set_fontsize(10) if min_depth and max_depth is not None: axes.set_ylim(max_depth, min_depth) else: axes.set_ylim(data['DEPTH'].max(), data['DEPTH'].min()) colors = ['k', 'brown', 'r', 'purple', 'y', 'orange', 'c', 'gold', 'b', 'plum', 'navy', 'm', 'sienna', 'teal', 'g'] for i in range(len(total_logs)): if i < len(logs): # for non-resistivity, normal plot colrs = np.random.choice(colors) ax[i].minorticks_on() ax[i].grid(b=True, which='major', color='black', linestyle='--') ax[i].grid(b=True, which='minor', color='grey', linestyle=':') ax[i].plot(data[total_logs[i]], data['DEPTH'], color=colrs) ax[i].set_xlim(data[total_logs[i]].min(), data[total_logs[i]].max()) ax[i].set_title(total_logs[i], size=20) ax[i].grid(b=True, which='major', color='black', linestyle='--') ax[i].grid(b=True, which='minor', color='grey', linestyle=':') colors.remove(colrs) if logs[i] == 'NPHI': ax[i].set_xlim(0.45, -0.15) else: # for resistivity, semilog plot colrs = np.random.choice(colors) ax[i].minorticks_on() ax[i].grid(b=True, which='major', color='black', linestyle='--') ax[i].grid(b=True, which='minor', color='grey', linestyle=':') ax[i].semilogx(data[total_logs[i]], data['DEPTH'], color=colrs) ax[i].set_xlim(0.2, 2500) ax[i].set_title(total_logs[i], size=20) ax[i].grid(b=True, which='major', color='black', linestyle='--') ax[i].grid(b=True, which='minor', color='grey', linestyle=':') colors.remove(colrs) plt.tight_layout(1.1)
[ "numpy.mean", "lasio.read", "matplotlib.ticker.LogLocator", "numpy.where", "numpy.random.choice", "matplotlib.pyplot.style.use", "matplotlib.pyplot.MultipleLocator", "matplotlib.pyplot.subplots", "matplotlib.pyplot.tight_layout", "warnings.filterwarnings" ]
[((207, 230), 'matplotlib.pyplot.style.use', 'plt.style.use', (['"""ggplot"""'], {}), "('ggplot')\n", (220, 230), True, 'import matplotlib.pyplot as plt\n'), ((232, 265), 'warnings.filterwarnings', 'warnings.filterwarnings', (['"""ignore"""'], {}), "('ignore')\n", (255, 265), False, 'import warnings\n'), ((1213, 1229), 'lasio.read', 'lasio.read', (['path'], {}), '(path)\n', (1223, 1229), False, 'import lasio\n'), ((3829, 3891), 'matplotlib.pyplot.subplots', 'plt.subplots', ([], {'nrows': '(1)', 'ncols': '(6)', 'figsize': 'plotsize', 'sharey': '(False)'}), '(nrows=1, ncols=6, figsize=plotsize, sharey=False)\n', (3841, 3891), True, 'import matplotlib.pyplot as plt\n'), ((17246, 17288), 'matplotlib.pyplot.tight_layout', 'plt.tight_layout', ([], {'pad': '(2)', 'h_pad': '(10)', 'w_pad': '(2)'}), '(pad=2, h_pad=10, w_pad=2)\n', (17262, 17288), True, 'import matplotlib.pyplot as plt\n'), ((17967, 18029), 'matplotlib.pyplot.subplots', 'plt.subplots', ([], {'nrows': '(1)', 'ncols': '(3)', 'figsize': 'plotsize', 'sharey': '"""all"""'}), "(nrows=1, ncols=3, figsize=plotsize, sharey='all')\n", (17979, 18029), True, 'import matplotlib.pyplot as plt\n'), ((25305, 25347), 'matplotlib.pyplot.tight_layout', 'plt.tight_layout', ([], {'pad': '(2)', 'h_pad': '(10)', 'w_pad': '(2)'}), '(pad=2, h_pad=10, w_pad=2)\n', (25321, 25347), True, 'import matplotlib.pyplot as plt\n'), ((26337, 26389), 'matplotlib.pyplot.subplots', 'plt.subplots', ([], {'nrows': '(1)', 'ncols': 'total', 'figsize': 'plotsize'}), '(nrows=1, ncols=total, figsize=plotsize)\n', (26349, 26389), True, 'import matplotlib.pyplot as plt\n'), ((28615, 28636), 'matplotlib.pyplot.tight_layout', 'plt.tight_layout', (['(1.1)'], {}), '(1.1)\n', (28631, 28636), True, 'import matplotlib.pyplot as plt\n'), ((2281, 2334), 'numpy.where', 'np.where', (["(dataframe['GR'] > 150)", '(150)', "dataframe['GR']"], {}), "(dataframe['GR'] > 150, 150, dataframe['GR'])\n", (2289, 2334), True, 'import numpy as np\n'), ((2366, 2415), 'numpy.where', 'np.where', (["(dataframe['GR'] < 0)", '(0)', "dataframe['GR']"], {}), "(dataframe['GR'] < 0, 0, dataframe['GR'])\n", (2374, 2415), True, 'import numpy as np\n'), ((2482, 2539), 'numpy.where', 'np.where', (["(dataframe['NPHI'] > 0.5)", '(0.5)', "dataframe['NPHI']"], {}), "(dataframe['NPHI'] > 0.5, 0.5, dataframe['NPHI'])\n", (2490, 2539), True, 'import numpy as np\n'), ((2575, 2636), 'numpy.where', 'np.where', (["(dataframe['NPHI'] < -0.15)", '(-0.15)', "dataframe['NPHI']"], {}), "(dataframe['NPHI'] < -0.15, -0.15, dataframe['NPHI'])\n", (2583, 2636), True, 'import numpy as np\n'), ((3113, 3124), 'numpy.mean', 'np.mean', (['df'], {}), '(df)\n', (3120, 3124), True, 'import numpy as np\n'), ((7893, 7933), 'matplotlib.ticker.LogLocator', 'ticker.LogLocator', ([], {'base': '(10.0)', 'numticks': '(3)'}), '(base=10.0, numticks=3)\n', (7910, 7933), True, 'import matplotlib.ticker as ticker\n'), ((8542, 8582), 'matplotlib.ticker.LogLocator', 'ticker.LogLocator', ([], {'base': '(10.0)', 'numticks': '(3)'}), '(base=10.0, numticks=3)\n', (8559, 8582), True, 'import matplotlib.ticker as ticker\n'), ((9175, 9215), 'matplotlib.ticker.LogLocator', 'ticker.LogLocator', ([], {'base': '(10.0)', 'numticks': '(3)'}), '(base=10.0, numticks=3)\n', (9192, 9215), True, 'import matplotlib.ticker as ticker\n'), ((9820, 9860), 'matplotlib.ticker.LogLocator', 'ticker.LogLocator', ([], {'base': '(10.0)', 'numticks': '(3)'}), '(base=10.0, numticks=3)\n', (9837, 9860), True, 'import matplotlib.ticker as ticker\n'), ((10475, 10515), 'matplotlib.ticker.LogLocator', 'ticker.LogLocator', ([], {'base': '(10.0)', 'numticks': '(3)'}), '(base=10.0, numticks=3)\n', (10492, 10515), True, 'import matplotlib.ticker as ticker\n'), ((11693, 11717), 'matplotlib.pyplot.MultipleLocator', 'plt.MultipleLocator', (['(0.2)'], {}), '(0.2)\n', (11712, 11717), True, 'import matplotlib.pyplot as plt\n'), ((12336, 12360), 'matplotlib.pyplot.MultipleLocator', 'plt.MultipleLocator', (['(0.4)'], {}), '(0.4)\n', (12355, 12360), True, 'import matplotlib.pyplot as plt\n'), ((20520, 20560), 'matplotlib.ticker.LogLocator', 'ticker.LogLocator', ([], {'base': '(10.0)', 'numticks': '(3)'}), '(base=10.0, numticks=3)\n', (20537, 20560), True, 'import matplotlib.ticker as ticker\n'), ((21121, 21161), 'matplotlib.ticker.LogLocator', 'ticker.LogLocator', ([], {'base': '(10.0)', 'numticks': '(3)'}), '(base=10.0, numticks=3)\n', (21138, 21161), True, 'import matplotlib.ticker as ticker\n'), ((21726, 21766), 'matplotlib.ticker.LogLocator', 'ticker.LogLocator', ([], {'base': '(10.0)', 'numticks': '(3)'}), '(base=10.0, numticks=3)\n', (21743, 21766), True, 'import matplotlib.ticker as ticker\n'), ((22344, 22384), 'matplotlib.ticker.LogLocator', 'ticker.LogLocator', ([], {'base': '(10.0)', 'numticks': '(3)'}), '(base=10.0, numticks=3)\n', (22361, 22384), True, 'import matplotlib.ticker as ticker\n'), ((22809, 22849), 'matplotlib.ticker.LogLocator', 'ticker.LogLocator', ([], {'base': '(10.0)', 'numticks': '(3)'}), '(base=10.0, numticks=3)\n', (22826, 22849), True, 'import matplotlib.ticker as ticker\n'), ((23595, 23619), 'matplotlib.pyplot.MultipleLocator', 'plt.MultipleLocator', (['(0.2)'], {}), '(0.2)\n', (23614, 23619), True, 'import matplotlib.pyplot as plt\n'), ((24238, 24262), 'matplotlib.pyplot.MultipleLocator', 'plt.MultipleLocator', (['(0.4)'], {}), '(0.4)\n', (24257, 24262), True, 'import matplotlib.pyplot as plt\n'), ((27168, 27192), 'numpy.random.choice', 'np.random.choice', (['colors'], {}), '(colors)\n', (27184, 27192), True, 'import numpy as np\n'), ((27997, 28021), 'numpy.random.choice', 'np.random.choice', (['colors'], {}), '(colors)\n', (28013, 28021), True, 'import numpy as np\n'), ((12818, 12842), 'matplotlib.pyplot.MultipleLocator', 'plt.MultipleLocator', (['(0.4)'], {}), '(0.4)\n', (12837, 12842), True, 'import matplotlib.pyplot as plt\n'), ((24720, 24744), 'matplotlib.pyplot.MultipleLocator', 'plt.MultipleLocator', (['(0.4)'], {}), '(0.4)\n', (24739, 24744), True, 'import matplotlib.pyplot as plt\n')]
from kfp.components import InputPath, OutputPath, create_component_from_func def xgboost_predict( data_path: InputPath('CSV'), # Also supports LibSVM model_path: InputPath('XGBoostModel'), predictions_path: OutputPath('Predictions'), label_column: int = None, ): '''Make predictions using a trained XGBoost model. Args: data_path: Path for the feature data in CSV format. model_path: Path for the trained model in binary XGBoost format. predictions_path: Output path for the predictions. label_column: Column containing the label data. Annotations: author: <NAME> <<EMAIL>> ''' from pathlib import Path import numpy import pandas import xgboost df = pandas.read_csv( data_path, ) if label_column is not None: df = df.drop(columns=[df.columns[label_column]]) testing_data = xgboost.DMatrix( data=df, ) model = xgboost.Booster(model_file=model_path) predictions = model.predict(testing_data) Path(predictions_path).parent.mkdir(parents=True, exist_ok=True) numpy.savetxt(predictions_path, predictions) if __name__ == '__main__': create_component_from_func( xgboost_predict, output_component_file='component.yaml', base_image='python:3.7', packages_to_install=[ 'xgboost==1.1.1', 'pandas==1.0.5', ], annotations={ "author": "<NAME> <<EMAIL>>", }, )
[ "pandas.read_csv", "pathlib.Path", "kfp.components.create_component_from_func", "kfp.components.InputPath", "xgboost.Booster", "numpy.savetxt", "xgboost.DMatrix", "kfp.components.OutputPath" ]
[((748, 774), 'pandas.read_csv', 'pandas.read_csv', (['data_path'], {}), '(data_path)\n', (763, 774), False, 'import pandas\n'), ((901, 925), 'xgboost.DMatrix', 'xgboost.DMatrix', ([], {'data': 'df'}), '(data=df)\n', (916, 925), False, 'import xgboost\n'), ((954, 992), 'xgboost.Booster', 'xgboost.Booster', ([], {'model_file': 'model_path'}), '(model_file=model_path)\n', (969, 992), False, 'import xgboost\n'), ((1114, 1158), 'numpy.savetxt', 'numpy.savetxt', (['predictions_path', 'predictions'], {}), '(predictions_path, predictions)\n', (1127, 1158), False, 'import numpy\n'), ((1192, 1415), 'kfp.components.create_component_from_func', 'create_component_from_func', (['xgboost_predict'], {'output_component_file': '"""component.yaml"""', 'base_image': '"""python:3.7"""', 'packages_to_install': "['xgboost==1.1.1', 'pandas==1.0.5']", 'annotations': "{'author': '<NAME> <<EMAIL>>'}"}), "(xgboost_predict, output_component_file=\n 'component.yaml', base_image='python:3.7', packages_to_install=[\n 'xgboost==1.1.1', 'pandas==1.0.5'], annotations={'author':\n '<NAME> <<EMAIL>>'})\n", (1218, 1415), False, 'from kfp.components import InputPath, OutputPath, create_component_from_func\n'), ((114, 130), 'kfp.components.InputPath', 'InputPath', (['"""CSV"""'], {}), "('CSV')\n", (123, 130), False, 'from kfp.components import InputPath, OutputPath, create_component_from_func\n'), ((172, 197), 'kfp.components.InputPath', 'InputPath', (['"""XGBoostModel"""'], {}), "('XGBoostModel')\n", (181, 197), False, 'from kfp.components import InputPath, OutputPath, create_component_from_func\n'), ((221, 246), 'kfp.components.OutputPath', 'OutputPath', (['"""Predictions"""'], {}), "('Predictions')\n", (231, 246), False, 'from kfp.components import InputPath, OutputPath, create_component_from_func\n'), ((1045, 1067), 'pathlib.Path', 'Path', (['predictions_path'], {}), '(predictions_path)\n', (1049, 1067), False, 'from pathlib import Path\n')]
#training the NN import sys sys.path.append("F:/Riku_Ka/AudioTestCode") import shutil import os from pydub import AudioSegment import numpy as np from keras import Sequential from keras.layers import Dense, Activation from keras.optimizers import adam from keras.models import load_model import time from keras.models import model_from_json startTime = time.time() #produce training data set. trainingDIR = "F:/Riku_Ka/Training_Dementia/Training Data Set" modelDIR = "F:/Riku_Ka/Training_Dementia" modelFILE = "Model_for Dementia1.json" weightFILE = "Model_for Dementia1.h5" trainingMatrix = np.zeros(shape=(1,14)) MFCCfiles = os.listdir(trainingDIR) for f in MFCCfiles: tempFile = trainingDIR + '/' + f print(tempFile) tempMatrix = np.loadtxt(tempFile) print(tempMatrix.shape) trainingMatrix = np.vstack((trainingMatrix,tempMatrix)) trainingDate = trainingMatrix[:,1:] print("The shape of training data matrix is: ") print(trainingDate.shape) trainingLabels = trainingMatrix[:,0] print("Time point 1 is " + str(time.time() - startTime)) np.random.seed(10) myModel = Sequential() myModel.add(Dense(units=64,input_dim=13, activation='relu')) myModel.add(Dense(units=8, activation='relu')) myModel.add(Dense(1, activation='sigmoid')) myModel.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy']) myModel.fit(x=trainingDate, y=trainingLabels, epochs=150, batch_size=32 ) # Save the trained model myModel_json = myModel.to_json() with open(modelDIR + '/' + modelFILE, 'w') as json_file: json_file.write(myModel_json) myModel.save_weights(modelDIR + '/' + weightFILE) print("The trained model has been saved.") print("Time point 3 is " + str(time.time() - startTime)) #produce test data set. testDIR = "F:/Riku_Ka/Training_Dementia/Test Data Set" MFCCfiles = os.listdir(testDIR) for f in MFCCfiles: tempFile = testDIR + '/' + f print(tempFile) tempMatrix = np.loadtxt(tempFile) testData = tempMatrix[:,1:] testLabels = tempMatrix[:,0] scores = myModel.evaluate(x=testData, y=testLabels) print("The accuracy for file {0} is {1}".format(f,scores[1]*100)) print("Time point 2 is " + str(time.time() - startTime)) #Select files according to special WORDS import os, random, shutil resDIR = "F:/Riku_Ka/Training_Dementia/MFCC with Labels for Health" dstDIR = "F:/Riku_Ka/Training_Dementia/Test Data Set" import sys sys.path.append("F:/Riku_Ka/AudioTestCode") import shutil import os from pydub import AudioSegment import numpy as np import FHC_Audio sourceDIR = "F:/Riku_Ka/PROMPT-UnzipAudioFile/音声データ_003" dstDIR = "F:/Riku_Ka/Training_Dementia/MFCC for Dementia-Pt only" subDIRs = os.listdir(sourceDIR) #listHealth = ['BH', 'KH', 'OH', 'PKH', 'POH', 'PTH', 'TH'] #listDepression = ['AD', 'BD', 'GD', 'KD', 'MD', 'ND', 'OD', 'PKD', 'POD', 'PTD', 'SD', 'TD'] listDementias = ['AC','GC','KC','OC','PKC','SC','TC'] print(listDementias) for dirs in subDIRs: tempDIR = sourceDIR + '/' + dirs files = os.listdir(tempDIR) for f in files: if any(x in f for x in listDementias) and f.find('Pt') != -1: tempFILE = tempDIR + '/' + f print("For the file {}: ".format(f)) try: featureMFCC = FHC_Audio.getMFCC4SingleFile(tempFILE, 13) featureFILE = dstDIR + '/' + os.path.splitext(f)[0] +"_MFCC.txt" np.savetxt(featureFILE, featureMFCC) except ValueError: print("There is a ValueError.") #select data randomly for TEST and TRAINING respectively import os, random, shutil sourceDIR = "F:/Riku_Ka/PROMPT-UnzipAudioFile/音声データ_003" dstDIR = "F:/Riku_Ka/Training_Dementia/MFCC for Dementia-Pt only" files = os.listdir(sourceDIR) numOfTest = int(len(files)/3) trainingSamples = random.sample(files, numOfTest) for sample in trainingSamples: shutil.move(resDIR + '/' + sample, dstDIR + '/' + sample) #Give the labels to file resDIR = "F:/Riku_Ka/Training_Dementia/MFCC for Dementia-Pt only" dstDIR = "F:/Riku_Ka/Training_Dementia/MFCC with Labels for Dementia" #myDIR = "F:/Riku_Ka/Training/MFCC for Health Data for Training" MFCCfiles = os.listdir(resDIR) for f in MFCCfiles: tempFile = resDIR + '/' + f tempMatrix = np.loadtxt(tempFile) n,m = tempMatrix.shape print("The size of {0} is {1} * {2}".format(f,n,m)) newCol = np.ones((n,1)) newMatrix = np.hstack((newCol,tempMatrix)) newFile = dstDIR + '/1_' + f np.savetxt(newFile,newMatrix)
[ "keras.Sequential", "random.sample", "os.listdir", "numpy.ones", "shutil.move", "numpy.hstack", "os.path.splitext", "numpy.zeros", "numpy.random.seed", "numpy.vstack", "time.time", "keras.layers.Dense", "numpy.savetxt", "numpy.loadtxt", "FHC_Audio.getMFCC4SingleFile", "sys.path.append"...
[((30, 73), 'sys.path.append', 'sys.path.append', (['"""F:/Riku_Ka/AudioTestCode"""'], {}), "('F:/Riku_Ka/AudioTestCode')\n", (45, 73), False, 'import sys\n'), ((371, 382), 'time.time', 'time.time', ([], {}), '()\n', (380, 382), False, 'import time\n'), ((620, 643), 'numpy.zeros', 'np.zeros', ([], {'shape': '(1, 14)'}), '(shape=(1, 14))\n', (628, 643), True, 'import numpy as np\n'), ((658, 681), 'os.listdir', 'os.listdir', (['trainingDIR'], {}), '(trainingDIR)\n', (668, 681), False, 'import os, random, shutil\n'), ((1109, 1127), 'numpy.random.seed', 'np.random.seed', (['(10)'], {}), '(10)\n', (1123, 1127), True, 'import numpy as np\n'), ((1141, 1153), 'keras.Sequential', 'Sequential', ([], {}), '()\n', (1151, 1153), False, 'from keras import Sequential\n'), ((1887, 1906), 'os.listdir', 'os.listdir', (['testDIR'], {}), '(testDIR)\n', (1897, 1906), False, 'import os, random, shutil\n'), ((2499, 2542), 'sys.path.append', 'sys.path.append', (['"""F:/Riku_Ka/AudioTestCode"""'], {}), "('F:/Riku_Ka/AudioTestCode')\n", (2514, 2542), False, 'import sys\n'), ((2780, 2801), 'os.listdir', 'os.listdir', (['sourceDIR'], {}), '(sourceDIR)\n', (2790, 2801), False, 'import os, random, shutil\n'), ((3844, 3865), 'os.listdir', 'os.listdir', (['sourceDIR'], {}), '(sourceDIR)\n', (3854, 3865), False, 'import os, random, shutil\n'), ((3916, 3947), 'random.sample', 'random.sample', (['files', 'numOfTest'], {}), '(files, numOfTest)\n', (3929, 3947), False, 'import os, random, shutil\n'), ((4290, 4308), 'os.listdir', 'os.listdir', (['resDIR'], {}), '(resDIR)\n', (4300, 4308), False, 'import os, random, shutil\n'), ((780, 800), 'numpy.loadtxt', 'np.loadtxt', (['tempFile'], {}), '(tempFile)\n', (790, 800), True, 'import numpy as np\n'), ((852, 891), 'numpy.vstack', 'np.vstack', (['(trainingMatrix, tempMatrix)'], {}), '((trainingMatrix, tempMatrix))\n', (861, 891), True, 'import numpy as np\n'), ((1169, 1217), 'keras.layers.Dense', 'Dense', ([], {'units': '(64)', 'input_dim': '(13)', 'activation': '"""relu"""'}), "(units=64, input_dim=13, activation='relu')\n", (1174, 1217), False, 'from keras.layers import Dense, Activation\n'), ((1231, 1264), 'keras.layers.Dense', 'Dense', ([], {'units': '(8)', 'activation': '"""relu"""'}), "(units=8, activation='relu')\n", (1236, 1264), False, 'from keras.layers import Dense, Activation\n'), ((1279, 1309), 'keras.layers.Dense', 'Dense', (['(1)'], {'activation': '"""sigmoid"""'}), "(1, activation='sigmoid')\n", (1284, 1309), False, 'from keras.layers import Dense, Activation\n'), ((2001, 2021), 'numpy.loadtxt', 'np.loadtxt', (['tempFile'], {}), '(tempFile)\n', (2011, 2021), True, 'import numpy as np\n'), ((3110, 3129), 'os.listdir', 'os.listdir', (['tempDIR'], {}), '(tempDIR)\n', (3120, 3129), False, 'import os, random, shutil\n'), ((3985, 4042), 'shutil.move', 'shutil.move', (["(resDIR + '/' + sample)", "(dstDIR + '/' + sample)"], {}), "(resDIR + '/' + sample, dstDIR + '/' + sample)\n", (3996, 4042), False, 'import os, random, shutil\n'), ((4383, 4403), 'numpy.loadtxt', 'np.loadtxt', (['tempFile'], {}), '(tempFile)\n', (4393, 4403), True, 'import numpy as np\n'), ((4503, 4518), 'numpy.ones', 'np.ones', (['(n, 1)'], {}), '((n, 1))\n', (4510, 4518), True, 'import numpy as np\n'), ((4535, 4566), 'numpy.hstack', 'np.hstack', (['(newCol, tempMatrix)'], {}), '((newCol, tempMatrix))\n', (4544, 4566), True, 'import numpy as np\n'), ((4605, 4635), 'numpy.savetxt', 'np.savetxt', (['newFile', 'newMatrix'], {}), '(newFile, newMatrix)\n', (4615, 4635), True, 'import numpy as np\n'), ((1078, 1089), 'time.time', 'time.time', ([], {}), '()\n', (1087, 1089), False, 'import time\n'), ((1763, 1774), 'time.time', 'time.time', ([], {}), '()\n', (1772, 1774), False, 'import time\n'), ((2263, 2274), 'time.time', 'time.time', ([], {}), '()\n', (2272, 2274), False, 'import time\n'), ((3363, 3405), 'FHC_Audio.getMFCC4SingleFile', 'FHC_Audio.getMFCC4SingleFile', (['tempFILE', '(13)'], {}), '(tempFILE, 13)\n', (3391, 3405), False, 'import FHC_Audio\n'), ((3505, 3541), 'numpy.savetxt', 'np.savetxt', (['featureFILE', 'featureMFCC'], {}), '(featureFILE, featureMFCC)\n', (3515, 3541), True, 'import numpy as np\n'), ((3452, 3471), 'os.path.splitext', 'os.path.splitext', (['f'], {}), '(f)\n', (3468, 3471), False, 'import os, random, shutil\n')]
""" TextCNN Model for Sentiment Analysis =============================================== This example shows how to use convolutional neural networks (textCNN) for sentiment analysis on various datasets. <NAME>. (2014). Convolutional neural networks for sentence classification. arXiv preprint arXiv:1408.5882. """ # coding: utf-8 # Licensed to the Apache Software Foundation (ASF) under one # or more contributor license agreements. See the NOTICE file # distributed with this work for additional information # regarding copyright ownership. The ASF licenses this file # to you under the Apache License, Version 2.0 (the # "License"); you may not use this file except in compliance # with the License. You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, # software distributed under the License is distributed on an # "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY # KIND, either express or implied. See the License for the # specific language governing permissions and limitations # under the License. import argparse import time import random import numpy as np import mxnet as mx from mxnet import nd, gluon, autograd from mxnet.gluon.data import DataLoader import process_data import text_cnn np.random.seed(3435) random.seed(3435) mx.random.seed(3435) parser = argparse.ArgumentParser(description='Sentiment analysis with the textCNN model on\ various datasets.') parser.add_argument('--data_name', choices=['MR', 'SST-1', 'SST-2', 'Subj', 'TREC'], default='MR', help='name of the data set') parser.add_argument('--model_mode', choices=['rand', 'static', 'non-static', 'multichannel'], default='multichannel', help='Variants of the textCNN model (see the paper:\ Convolutional Neural Networks for Sentence Classification).') parser.add_argument('--lr', type=float, default=2.5E-3, help='initial learning rate') parser.add_argument('--epochs', type=int, default=20, help='upper epoch limit') parser.add_argument('--batch_size', type=int, default=50, metavar='N', help='batch size') parser.add_argument('--dropout', type=float, default=.5, help='dropout applied to layers (0 = no dropout)') parser.add_argument('--log-interval', type=int, default=30, metavar='N', help='report interval') parser.add_argument('--save-prefix', type=str, default='sa-model', help='path to save the final model') parser.add_argument('--gpu', type=int, default=None, help='id of the gpu to use. Set it to empty means to use cpu.') args = parser.parse_args() print(args) if args.gpu is None: print('Use cpu') context = mx.cpu() else: print('Use gpu%d' % args.gpu) context = mx.gpu(args.gpu) if args.data_name == 'MR' or args.data_name == 'Subj': vocab, max_len, output_size, train_dataset, train_data_lengths \ = process_data.load_dataset(args.data_name) else: vocab, max_len, output_size, train_dataset, train_data_lengths, \ test_dataset, test_data_lengths = process_data.load_dataset(args.data_name) model = text_cnn.model(args.dropout, vocab, args.model_mode, output_size) print(model) loss = gluon.loss.SoftmaxCrossEntropyLoss() def evaluate(net, dataloader): """Evaluate network on the specified dataset""" total_L = 0.0 total_sample_num = 0 total_correct_num = 0 start_log_interval_time = time.time() print('Begin Testing...') for i, (data, label) in enumerate(dataloader): data = mx.nd.transpose(data.as_in_context(context)) label = label.as_in_context(context) output = net(data) L = loss(output, label) pred = nd.argmax(output, axis=1) total_L += L.sum().asscalar() total_sample_num += label.shape[0] total_correct_num += (pred.astype('int') == label).sum().asscalar() if (i + 1) % args.log_interval == 0: print('[Batch {}/{}] elapsed {:.2f} s'.format( i + 1, len(dataloader), time.time() - start_log_interval_time)) start_log_interval_time = time.time() avg_L = total_L / float(total_sample_num) acc = total_correct_num / float(total_sample_num) return avg_L, acc def train(net, train_data, test_data): """Train textCNN model for sentiment analysis.""" start_pipeline_time = time.time() net, trainer = text_cnn.init(net, vocab, args.model_mode, context, args.lr) random.shuffle(train_data) sp = int(len(train_data)*0.9) train_dataloader = DataLoader(dataset=train_data[:sp], batch_size=args.batch_size, shuffle=True) val_dataloader = DataLoader(dataset=train_data[sp:], batch_size=args.batch_size, shuffle=False) test_dataloader = DataLoader(dataset=test_data, batch_size=args.batch_size, shuffle=False) # Training/Testing. best_val_acc = 0 for epoch in range(args.epochs): # Epoch training stats. start_epoch_time = time.time() epoch_L = 0.0 epoch_sent_num = 0 epoch_wc = 0 # Log interval training stats. start_log_interval_time = time.time() log_interval_wc = 0 log_interval_sent_num = 0 log_interval_L = 0.0 for i, (data, label) in enumerate(train_dataloader): data = mx.nd.transpose(data.as_in_context(context)) label = label.as_in_context(context) wc = max_len log_interval_wc += wc epoch_wc += wc log_interval_sent_num += data.shape[1] epoch_sent_num += data.shape[1] with autograd.record(): output = net(data) L = loss(output, label).mean() L.backward() # Update parameter. trainer.step(1) log_interval_L += L.asscalar() epoch_L += L.asscalar() if (i + 1) % args.log_interval == 0: print('[Epoch %d Batch %d/%d] avg loss %g, throughput %gK wps' % ( epoch, i + 1, len(train_dataloader), log_interval_L / log_interval_sent_num, log_interval_wc / 1000 / (time.time() - start_log_interval_time))) # Clear log interval training stats. start_log_interval_time = time.time() log_interval_wc = 0 log_interval_sent_num = 0 log_interval_L = 0 end_epoch_time = time.time() val_avg_L, val_acc = evaluate(net, val_dataloader) print('[Epoch %d] train avg loss %g, ' 'test acc %.4f, test avg loss %g, throughput %gK wps' % ( epoch, epoch_L / epoch_sent_num, val_acc, val_avg_L, epoch_wc / 1000 / (end_epoch_time - start_epoch_time))) if val_acc >= best_val_acc: print('Observed Improvement.') best_val_acc = val_acc test_avg_L, test_acc = evaluate(net, test_dataloader) print('Test loss %g, test acc %.4f'%(test_avg_L, test_acc)) print('Total time cost %.2fs'%(time.time()-start_pipeline_time)) return test_acc def k_fold_cross_valid(k, net, all_dataset): test_acc = [] fold_size = len(all_dataset) // k random.shuffle(all_dataset) for test_i in range(10): test_data = all_dataset[test_i * fold_size: (test_i + 1) * fold_size] train_data = all_dataset[: test_i * fold_size] + all_dataset[(test_i + 1) * fold_size:] print(len(train_data), len(test_data)) test_acc.append(train(net, train_data, test_data)) print(sum(test_acc) / k) if __name__ == '__main__': if args.data_name != 'MR' and args.data_name != 'Subj': train(model, train_dataset, test_dataset) else: k_fold_cross_valid(10, model, train_dataset)
[ "mxnet.gluon.loss.SoftmaxCrossEntropyLoss", "mxnet.nd.argmax", "mxnet.autograd.record", "random.shuffle", "argparse.ArgumentParser", "mxnet.cpu", "text_cnn.init", "random.seed", "mxnet.gpu", "numpy.random.seed", "mxnet.random.seed", "mxnet.gluon.data.DataLoader", "process_data.load_dataset",...
[((1317, 1337), 'numpy.random.seed', 'np.random.seed', (['(3435)'], {}), '(3435)\n', (1331, 1337), True, 'import numpy as np\n'), ((1338, 1355), 'random.seed', 'random.seed', (['(3435)'], {}), '(3435)\n', (1349, 1355), False, 'import random\n'), ((1356, 1376), 'mxnet.random.seed', 'mx.random.seed', (['(3435)'], {}), '(3435)\n', (1370, 1376), True, 'import mxnet as mx\n'), ((1387, 1530), 'argparse.ArgumentParser', 'argparse.ArgumentParser', ([], {'description': '"""Sentiment analysis with the textCNN model on various datasets."""'}), "(description=\n 'Sentiment analysis with the textCNN model on various datasets.'\n )\n", (1410, 1530), False, 'import argparse\n'), ((3280, 3345), 'text_cnn.model', 'text_cnn.model', (['args.dropout', 'vocab', 'args.model_mode', 'output_size'], {}), '(args.dropout, vocab, args.model_mode, output_size)\n', (3294, 3345), False, 'import text_cnn\n'), ((3367, 3403), 'mxnet.gluon.loss.SoftmaxCrossEntropyLoss', 'gluon.loss.SoftmaxCrossEntropyLoss', ([], {}), '()\n', (3401, 3403), False, 'from mxnet import nd, gluon, autograd\n'), ((2862, 2870), 'mxnet.cpu', 'mx.cpu', ([], {}), '()\n', (2868, 2870), True, 'import mxnet as mx\n'), ((2925, 2941), 'mxnet.gpu', 'mx.gpu', (['args.gpu'], {}), '(args.gpu)\n', (2931, 2941), True, 'import mxnet as mx\n'), ((3073, 3114), 'process_data.load_dataset', 'process_data.load_dataset', (['args.data_name'], {}), '(args.data_name)\n', (3098, 3114), False, 'import process_data\n'), ((3229, 3270), 'process_data.load_dataset', 'process_data.load_dataset', (['args.data_name'], {}), '(args.data_name)\n', (3254, 3270), False, 'import process_data\n'), ((3587, 3598), 'time.time', 'time.time', ([], {}), '()\n', (3596, 3598), False, 'import time\n'), ((4518, 4529), 'time.time', 'time.time', ([], {}), '()\n', (4527, 4529), False, 'import time\n'), ((4549, 4609), 'text_cnn.init', 'text_cnn.init', (['net', 'vocab', 'args.model_mode', 'context', 'args.lr'], {}), '(net, vocab, args.model_mode, context, args.lr)\n', (4562, 4609), False, 'import text_cnn\n'), ((4614, 4640), 'random.shuffle', 'random.shuffle', (['train_data'], {}), '(train_data)\n', (4628, 4640), False, 'import random\n'), ((4698, 4775), 'mxnet.gluon.data.DataLoader', 'DataLoader', ([], {'dataset': 'train_data[:sp]', 'batch_size': 'args.batch_size', 'shuffle': '(True)'}), '(dataset=train_data[:sp], batch_size=args.batch_size, shuffle=True)\n', (4708, 4775), False, 'from mxnet.gluon.data import DataLoader\n'), ((4865, 4943), 'mxnet.gluon.data.DataLoader', 'DataLoader', ([], {'dataset': 'train_data[sp:]', 'batch_size': 'args.batch_size', 'shuffle': '(False)'}), '(dataset=train_data[sp:], batch_size=args.batch_size, shuffle=False)\n', (4875, 4943), False, 'from mxnet.gluon.data import DataLoader\n'), ((5030, 5102), 'mxnet.gluon.data.DataLoader', 'DataLoader', ([], {'dataset': 'test_data', 'batch_size': 'args.batch_size', 'shuffle': '(False)'}), '(dataset=test_data, batch_size=args.batch_size, shuffle=False)\n', (5040, 5102), False, 'from mxnet.gluon.data import DataLoader\n'), ((7581, 7608), 'random.shuffle', 'random.shuffle', (['all_dataset'], {}), '(all_dataset)\n', (7595, 7608), False, 'import random\n'), ((3859, 3884), 'mxnet.nd.argmax', 'nd.argmax', (['output'], {'axis': '(1)'}), '(output, axis=1)\n', (3868, 3884), False, 'from mxnet import nd, gluon, autograd\n'), ((5310, 5321), 'time.time', 'time.time', ([], {}), '()\n', (5319, 5321), False, 'import time\n'), ((5465, 5476), 'time.time', 'time.time', ([], {}), '()\n', (5474, 5476), False, 'import time\n'), ((6787, 6798), 'time.time', 'time.time', ([], {}), '()\n', (6796, 6798), False, 'import time\n'), ((4264, 4275), 'time.time', 'time.time', ([], {}), '()\n', (4273, 4275), False, 'import time\n'), ((5941, 5958), 'mxnet.autograd.record', 'autograd.record', ([], {}), '()\n', (5956, 5958), False, 'from mxnet import nd, gluon, autograd\n'), ((6637, 6648), 'time.time', 'time.time', ([], {}), '()\n', (6646, 6648), False, 'import time\n'), ((7421, 7432), 'time.time', 'time.time', ([], {}), '()\n', (7430, 7432), False, 'import time\n'), ((4186, 4197), 'time.time', 'time.time', ([], {}), '()\n', (4195, 4197), False, 'import time\n'), ((6501, 6512), 'time.time', 'time.time', ([], {}), '()\n', (6510, 6512), False, 'import time\n')]
from functools import lru_cache import os import shutil import struct import numpy as np import torch import re from fairseq.data.datautils import utf8_to_uxxxx, uxxxx_to_utf8 import cv2 from fairseq.data import FairseqDataset import json import lmdb import logging LOG = logging.getLogger(__name__) class OcrLmdbDataset(FairseqDataset): """Takes a text file as input and binarizes it in memory at instantiation. Original lines are also kept in memory""" def __init__( self, split, data_dir, dictionary, transforms, image_height, max_allowed_width, ): LOG.info("...OcrLmdbDataset %s", data_dir) self.data_dir = data_dir self.split = split self.dictionary = dictionary self.preprocess = transforms self.image_height = image_height self.max_allowed_width = max_allowed_width with open(os.path.join(self.data_dir, "desc.json"), "r") as fh: self.data_desc = json.load(fh) self.sizes = [] for entry in self.data_desc[self.split]: self.sizes.append(len(entry["trans"].split())) self.sizes = np.array(self.sizes) self.lmdb_env = lmdb.Environment( os.path.join(self.data_dir, "line-images.lmdb"), map_size=1e6, readonly=True, lock=False, ) self.lmdb_txn = self.lmdb_env.begin(buffers=True) self.size_group_limits = [150, 200, 300, 350, 450, 600, np.inf] self.size_group_keys = self.size_group_limits self.size_groups = dict() self.size_groups_dict = dict() for cur_limit in self.size_group_limits: self.size_groups[cur_limit] = [] self.size_groups_dict[cur_limit] = dict() for idx, entry in enumerate(self.data_desc[self.split]): width_orig, height_orig = entry["width"], entry["height"] normalized_width = width_orig * (self.image_height / height_orig) for cur_limit in self.size_group_limits: if ( normalized_width < cur_limit and normalized_width < self.max_allowed_width ): self.size_groups[cur_limit].append(idx) self.size_groups_dict[cur_limit][idx] = 1 break # Now get final size (might have dropped large entries!) self.nentries = 0 self.max_index = 0 for cur_limit in self.size_group_limits: self.nentries += len(self.size_groups[cur_limit]) if len(self.size_groups[cur_limit]) > 0: cur_max = max(self.size_groups[cur_limit]) if cur_max > self.max_index: self.max_index = cur_max print("...finished loading, size {}".format(self.nentries)) print("count by group") total_group_cnt = 0 for cur_limit in self.size_group_limits: print("group", cur_limit, len(self.size_groups[cur_limit])) total_group_cnt += len(self.size_groups[cur_limit]) print("TOTAL...", total_group_cnt) def __getitem__(self, index): entry = self.data_desc[self.split][index] max_width = 0 for cur_limit in self.size_group_limits: if index in self.size_groups_dict[cur_limit]: max_width = cur_limit break group_id = max_width image_name = entry["id"] img_bytes = np.asarray( self.lmdb_txn.get(entry["id"].encode("ascii")), dtype=np.uint8 ) line_image = cv2.imdecode(img_bytes, cv2.IMREAD_COLOR) # -1) # Do a check for RGBA images; if found get rid of alpha channel if len(line_image.shape) == 3 and line_image.shape[2] == 4: line_image = cv2.cvtColor(line_image, cv2.COLOR_BGRA2BGR) line_image = self.preprocess(line_image) # Sanity check: make sure width@30px lh is long enough not to crash our model; we pad to at least 15px wide # Need to do this and change the "real" image size so that pack_padded doens't complain if line_image.size(2) < 15: line_image_ = torch.ones( line_image.size(0), line_image.size(1), 15) line_image_[:, :, : line_image.size(2)] = line_image line_image = line_image_ # Add padding up to max-width, so that we have consistent size for cudnn.benchmark to work with original_width = line_image.size(2) original_height = line_image.size(1) transcription = [] for char in entry["trans"].split(): transcription.append(self.dictionary.index(char)) src_metadata = { "target": transcription, # "target_len": len(transcription), "uxxx_trans": entry["trans"], "utf8_trans": uxxxx_to_utf8(entry["trans"]), "width": original_width, "height": original_height, "group": group_id, "image_name": image_name, "image": line_image, "id": index, } return src_metadata def __len__(self): return self.nentries
[ "logging.getLogger", "fairseq.data.datautils.uxxxx_to_utf8", "os.path.join", "numpy.array", "cv2.imdecode", "cv2.cvtColor", "json.load" ]
[((278, 305), 'logging.getLogger', 'logging.getLogger', (['__name__'], {}), '(__name__)\n', (295, 305), False, 'import logging\n'), ((1134, 1154), 'numpy.array', 'np.array', (['self.sizes'], {}), '(self.sizes)\n', (1142, 1154), True, 'import numpy as np\n'), ((3594, 3635), 'cv2.imdecode', 'cv2.imdecode', (['img_bytes', 'cv2.IMREAD_COLOR'], {}), '(img_bytes, cv2.IMREAD_COLOR)\n', (3606, 3635), False, 'import cv2\n'), ((966, 979), 'json.load', 'json.load', (['fh'], {}), '(fh)\n', (975, 979), False, 'import json\n'), ((1210, 1257), 'os.path.join', 'os.path.join', (['self.data_dir', '"""line-images.lmdb"""'], {}), "(self.data_dir, 'line-images.lmdb')\n", (1222, 1257), False, 'import os\n'), ((3808, 3852), 'cv2.cvtColor', 'cv2.cvtColor', (['line_image', 'cv2.COLOR_BGRA2BGR'], {}), '(line_image, cv2.COLOR_BGRA2BGR)\n', (3820, 3852), False, 'import cv2\n'), ((4859, 4888), 'fairseq.data.datautils.uxxxx_to_utf8', 'uxxxx_to_utf8', (["entry['trans']"], {}), "(entry['trans'])\n", (4872, 4888), False, 'from fairseq.data.datautils import utf8_to_uxxxx, uxxxx_to_utf8\n'), ((883, 923), 'os.path.join', 'os.path.join', (['self.data_dir', '"""desc.json"""'], {}), "(self.data_dir, 'desc.json')\n", (895, 923), False, 'import os\n')]
import numpy as np import pytest from lagom.core.transform import Clip from lagom.core.transform import Centralize from lagom.core.transform import Normalize from lagom.core.transform import Standardize from lagom.core.transform import ExpFactorCumSum from lagom.core.transform import RunningMeanStd from lagom.core.transform import RankTransform class TestTransform(object): def test_clip(self): clip = Clip() # Test scalar assert clip(x=2, a_min=0, a_max=1) == 1 assert clip(x=0.5, a_min=0, a_max=1) == 0.5 assert clip(x=-1, a_min=0, a_max=1) == 0 # Test numpy scalar assert clip(x=np.array(2), a_min=0, a_max=1) == 1 assert clip(x=np.array(0.5), a_min=0, a_max=1) == 0.5 assert clip(x=np.array(-1), a_min=0, a_max=1) == 0 # # Test vector # def _test_vec(x): assert np.alltrue(clip(x=x, a_min=2, a_max=3) == [2, 2, 3, 3]) # Tuple a = (1, 2, 3, 4) _test_vec(a) # List b = [1, 2, 3, 4] _test_vec(b) # ndarray c = np.array([1, 2, 3, 4]) _test_vec(c) # # Test exceptions # # ndarray more than 1-dim is not allowed d = np.array([[1, 2, 3, 4]]) with pytest.raises(ValueError): clip(x=d, a_min=2, a_max=3) def test_centralize(self): centralize = Centralize() # Test scalar assert centralize(x=1) == 1 assert centralize(x=0) == 0 assert centralize(x=2) == 2 assert centralize(x=1, mean=1) == 1 assert centralize(x=0, mean=1) == 0 assert centralize(x=2, mean=1) == 2 # Test numpy scalar assert centralize(x=np.array(1)) == 1 assert centralize(x=np.array(0)) == 0 assert centralize(x=np.array(2)) == 2 assert centralize(x=np.array(1), mean=1) == 1 assert centralize(x=np.array(0), mean=1) == 0 assert centralize(x=np.array(2), mean=1) == 2 # # Test vector # def _test_vec(x): assert np.alltrue(centralize(x=x) == [-1.5, -0.5, 0.5, 1.5]) assert np.alltrue(centralize(x=x, mean=1) == [0, 1, 2, 3]) # Tuple a = (1, 2, 3, 4) _test_vec(a) # List b = [1, 2, 3, 4] _test_vec(b) # ndarray c = np.array([1, 2, 3, 4]) _test_vec(c) # # Test exceptions # # ndarray more than 1-dim is not allowed d = np.array([[1, 2, 3, 4]]) with pytest.raises(ValueError): centralize(x=d) def test_normalize(self): normalize = Normalize(eps=1.1920929e-07) # Test scalar assert normalize(x=-1) == 0 assert normalize(x=0.5) == 0.5 assert normalize(x=2) == 1 assert normalize(x=-1, min_val=0, max_val=1) == 0 assert normalize(x=0.5, min_val=0, max_val=1) == 0.5 assert normalize(x=2, min_val=0, max_val=1) == 1 # Test numpy scalar assert normalize(x=np.array(-1)) == 0 assert normalize(x=np.array(0.5)) == 0.5 assert normalize(x=np.array(2)) == 1 assert normalize(x=np.array(-1), min_val=0, max_val=1) == 0 assert normalize(x=np.array(0.5), min_val=0, max_val=1) == 0.5 assert normalize(x=np.array(2), min_val=0, max_val=1) == 1 # # Test vector # def _test_vec(x): assert np.allclose(normalize(x=x), [0. , 0.33333332, 0.66666664, 0.99999996]) assert np.allclose(normalize(x=x, min_val=0, max_val=1), [0.99999988, 1.99999976, 2.99999964, 3.99999952]) # Tuple a = (1, 2, 3, 4) _test_vec(a) # List b = [1, 2, 3, 4] _test_vec(b) # ndarray c = np.array([1, 2, 3, 4]) _test_vec(c) # # Test exceptions # # ndarray more than 1-dim is not allowed d = np.array([[1, 2, 3, 4]]) with pytest.raises(ValueError): normalize(x=d) def test_standardize(self): standardize = Standardize(eps=1.1920929e-07) # Test scalar assert standardize(x=-1) == -1 assert standardize(x=0) == 0 assert standardize(x=1) == 1 assert standardize(x=-1, mean=0, std=1) == -1 assert standardize(x=0, mean=0, std=1) == 0 assert standardize(x=1, mean=0, std=1) == 1 # Test numpy scalar assert standardize(x=np.array(-1)) == -1 assert standardize(x=np.array(0)) == 0 assert standardize(x=np.array(1)) == 1 assert standardize(x=np.array(-1), mean=0, std=1) == -1 assert standardize(x=np.array(0), mean=0, std=1) == 0 assert standardize(x=np.array(1), mean=0, std=1) == 1 # # Test vector # def _test_vec(x): assert np.allclose(standardize(x=x), [-1.34164064, -0.44721355, 0.44721355, 1.34164064]) assert np.allclose(standardize(x=x, mean=0, std=1), [0.99999988, 1.99999976, 2.99999964, 3.99999952]) # Tuple a = (1, 2, 3, 4) _test_vec(a) # List b = [1, 2, 3, 4] _test_vec(b) # ndarray c = np.array([1, 2, 3, 4]) _test_vec(c) # # Test exceptions # # ndarray more than 1-dim is not allowed d = np.array([[1, 2, 3, 4]]) with pytest.raises(ValueError): standardize(x=d) def test_expfactorcumsum(self): expfactorcumsum = ExpFactorCumSum(alpha=0.1) # # Test vector # def _test_vec(x): assert np.allclose(expfactorcumsum(x=x), [1.23, 2.3, 3.0]) # Tuple a = (1, 2, 3) _test_vec(a) # List b = [1, 2, 3] _test_vec(b) # ndarray c = np.array([1, 2, 3]) _test_vec(c) # # Test exceptions # # Scalar is not allowed with pytest.raises(AssertionError): expfactorcumsum(x=1) # ndarray more than 1-dim is not allowed d = np.array([[1, 2, 3]]) with pytest.raises(ValueError): expfactorcumsum(x=d) def test_runningmeanstd(self): def _test_moments(runningmeanstd, x): assert np.allclose(runningmeanstd.mu, np.mean(x)) assert np.allclose(runningmeanstd.sigma, np.std(x)) a = [1, 2, 3, 4] # Scalar runningmeanstd = RunningMeanStd() [runningmeanstd(i) for i in a] _test_moments(runningmeanstd=runningmeanstd, x=a) # Vector runningmeanstd = RunningMeanStd() runningmeanstd(a) _test_moments(runningmeanstd=runningmeanstd, x=a) # n-dim array b = np.array([[1, 10, 100], [2, 20, 200], [3, 30, 300], [4, 40, 400]]) runningmeanstd = RunningMeanStd() runningmeanstd(b) assert np.allclose(runningmeanstd.mu, b.mean(0)) assert np.allclose(runningmeanstd.sigma, b.std(0)) def test_rank_transform(self): rank_transform = RankTransform() # List a = [3, 14, 1] assert np.allclose(rank_transform(a, centered=False), [1, 2, 0]) assert np.allclose(rank_transform(a), [0, 0.5, -0.5]) # ndarray b = np.array([3, 14, 1]) assert np.allclose(rank_transform(b, centered=False), [1, 2, 0]) assert np.allclose(rank_transform(b), [0, 0.5, -0.5]) # # Test exceptions # # Scalar is not allowed with pytest.raises(AssertionError): rank_transform(5) # ndarray more than 1-dim is not allowed c = np.array([[3, 14, 1]]) with pytest.raises(ValueError): rank_transform(c)
[ "numpy.mean", "lagom.core.transform.ExpFactorCumSum", "lagom.core.transform.Normalize", "lagom.core.transform.Clip", "lagom.core.transform.Centralize", "lagom.core.transform.RunningMeanStd", "numpy.array", "lagom.core.transform.RankTransform", "pytest.raises", "numpy.std", "lagom.core.transform....
[((421, 427), 'lagom.core.transform.Clip', 'Clip', ([], {}), '()\n', (425, 427), False, 'from lagom.core.transform import Clip\n'), ((1156, 1178), 'numpy.array', 'np.array', (['[1, 2, 3, 4]'], {}), '([1, 2, 3, 4])\n', (1164, 1178), True, 'import numpy as np\n'), ((1316, 1340), 'numpy.array', 'np.array', (['[[1, 2, 3, 4]]'], {}), '([[1, 2, 3, 4]])\n', (1324, 1340), True, 'import numpy as np\n'), ((1482, 1494), 'lagom.core.transform.Centralize', 'Centralize', ([], {}), '()\n', (1492, 1494), False, 'from lagom.core.transform import Centralize\n'), ((2522, 2544), 'numpy.array', 'np.array', (['[1, 2, 3, 4]'], {}), '([1, 2, 3, 4])\n', (2530, 2544), True, 'import numpy as np\n'), ((2682, 2706), 'numpy.array', 'np.array', (['[[1, 2, 3, 4]]'], {}), '([[1, 2, 3, 4]])\n', (2690, 2706), True, 'import numpy as np\n'), ((2834, 2862), 'lagom.core.transform.Normalize', 'Normalize', ([], {'eps': '(1.1920929e-07)'}), '(eps=1.1920929e-07)\n', (2843, 2862), False, 'from lagom.core.transform import Normalize\n'), ((4118, 4140), 'numpy.array', 'np.array', (['[1, 2, 3, 4]'], {}), '([1, 2, 3, 4])\n', (4126, 4140), True, 'import numpy as np\n'), ((4278, 4302), 'numpy.array', 'np.array', (['[[1, 2, 3, 4]]'], {}), '([[1, 2, 3, 4]])\n', (4286, 4302), True, 'import numpy as np\n'), ((4433, 4463), 'lagom.core.transform.Standardize', 'Standardize', ([], {'eps': '(1.1920929e-07)'}), '(eps=1.1920929e-07)\n', (4444, 4463), False, 'from lagom.core.transform import Standardize\n'), ((5699, 5721), 'numpy.array', 'np.array', (['[1, 2, 3, 4]'], {}), '([1, 2, 3, 4])\n', (5707, 5721), True, 'import numpy as np\n'), ((5859, 5883), 'numpy.array', 'np.array', (['[[1, 2, 3, 4]]'], {}), '([[1, 2, 3, 4]])\n', (5867, 5883), True, 'import numpy as np\n'), ((6024, 6050), 'lagom.core.transform.ExpFactorCumSum', 'ExpFactorCumSum', ([], {'alpha': '(0.1)'}), '(alpha=0.1)\n', (6039, 6050), False, 'from lagom.core.transform import ExpFactorCumSum\n'), ((6409, 6428), 'numpy.array', 'np.array', (['[1, 2, 3]'], {}), '([1, 2, 3])\n', (6417, 6428), True, 'import numpy as np\n'), ((6684, 6705), 'numpy.array', 'np.array', (['[[1, 2, 3]]'], {}), '([[1, 2, 3]])\n', (6692, 6705), True, 'import numpy as np\n'), ((7068, 7084), 'lagom.core.transform.RunningMeanStd', 'RunningMeanStd', ([], {}), '()\n', (7082, 7084), False, 'from lagom.core.transform import RunningMeanStd\n'), ((7225, 7241), 'lagom.core.transform.RunningMeanStd', 'RunningMeanStd', ([], {}), '()\n', (7239, 7241), False, 'from lagom.core.transform import RunningMeanStd\n'), ((7361, 7427), 'numpy.array', 'np.array', (['[[1, 10, 100], [2, 20, 200], [3, 30, 300], [4, 40, 400]]'], {}), '([[1, 10, 100], [2, 20, 200], [3, 30, 300], [4, 40, 400]])\n', (7369, 7427), True, 'import numpy as np\n'), ((7453, 7469), 'lagom.core.transform.RunningMeanStd', 'RunningMeanStd', ([], {}), '()\n', (7467, 7469), False, 'from lagom.core.transform import RunningMeanStd\n'), ((7681, 7696), 'lagom.core.transform.RankTransform', 'RankTransform', ([], {}), '()\n', (7694, 7696), False, 'from lagom.core.transform import RankTransform\n'), ((7918, 7938), 'numpy.array', 'np.array', (['[3, 14, 1]'], {}), '([3, 14, 1])\n', (7926, 7938), True, 'import numpy as np\n'), ((8309, 8331), 'numpy.array', 'np.array', (['[[3, 14, 1]]'], {}), '([[3, 14, 1]])\n', (8317, 8331), True, 'import numpy as np\n'), ((1354, 1379), 'pytest.raises', 'pytest.raises', (['ValueError'], {}), '(ValueError)\n', (1367, 1379), False, 'import pytest\n'), ((2720, 2745), 'pytest.raises', 'pytest.raises', (['ValueError'], {}), '(ValueError)\n', (2733, 2745), False, 'import pytest\n'), ((4316, 4341), 'pytest.raises', 'pytest.raises', (['ValueError'], {}), '(ValueError)\n', (4329, 4341), False, 'import pytest\n'), ((5897, 5922), 'pytest.raises', 'pytest.raises', (['ValueError'], {}), '(ValueError)\n', (5910, 5922), False, 'import pytest\n'), ((6550, 6579), 'pytest.raises', 'pytest.raises', (['AssertionError'], {}), '(AssertionError)\n', (6563, 6579), False, 'import pytest\n'), ((6719, 6744), 'pytest.raises', 'pytest.raises', (['ValueError'], {}), '(ValueError)\n', (6732, 6744), False, 'import pytest\n'), ((8174, 8203), 'pytest.raises', 'pytest.raises', (['AssertionError'], {}), '(AssertionError)\n', (8187, 8203), False, 'import pytest\n'), ((8345, 8370), 'pytest.raises', 'pytest.raises', (['ValueError'], {}), '(ValueError)\n', (8358, 8370), False, 'import pytest\n'), ((6923, 6933), 'numpy.mean', 'np.mean', (['x'], {}), '(x)\n', (6930, 6933), True, 'import numpy as np\n'), ((6988, 6997), 'numpy.std', 'np.std', (['x'], {}), '(x)\n', (6994, 6997), True, 'import numpy as np\n'), ((667, 678), 'numpy.array', 'np.array', (['(2)'], {}), '(2)\n', (675, 678), True, 'import numpy as np\n'), ((725, 738), 'numpy.array', 'np.array', (['(0.5)'], {}), '(0.5)\n', (733, 738), True, 'import numpy as np\n'), ((787, 799), 'numpy.array', 'np.array', (['(-1)'], {}), '(-1)\n', (795, 799), True, 'import numpy as np\n'), ((1840, 1851), 'numpy.array', 'np.array', (['(1)'], {}), '(1)\n', (1848, 1851), True, 'import numpy as np\n'), ((1886, 1897), 'numpy.array', 'np.array', (['(0)'], {}), '(0)\n', (1894, 1897), True, 'import numpy as np\n'), ((1932, 1943), 'numpy.array', 'np.array', (['(2)'], {}), '(2)\n', (1940, 1943), True, 'import numpy as np\n'), ((1987, 1998), 'numpy.array', 'np.array', (['(1)'], {}), '(1)\n', (1995, 1998), True, 'import numpy as np\n'), ((2041, 2052), 'numpy.array', 'np.array', (['(0)'], {}), '(0)\n', (2049, 2052), True, 'import numpy as np\n'), ((2095, 2106), 'numpy.array', 'np.array', (['(2)'], {}), '(2)\n', (2103, 2106), True, 'import numpy as np\n'), ((3253, 3265), 'numpy.array', 'np.array', (['(-1)'], {}), '(-1)\n', (3261, 3265), True, 'import numpy as np\n'), ((3299, 3312), 'numpy.array', 'np.array', (['(0.5)'], {}), '(0.5)\n', (3307, 3312), True, 'import numpy as np\n'), ((3348, 3359), 'numpy.array', 'np.array', (['(2)'], {}), '(2)\n', (3356, 3359), True, 'import numpy as np\n'), ((3402, 3414), 'numpy.array', 'np.array', (['(-1)'], {}), '(-1)\n', (3410, 3414), True, 'import numpy as np\n'), ((3470, 3483), 'numpy.array', 'np.array', (['(0.5)'], {}), '(0.5)\n', (3478, 3483), True, 'import numpy as np\n'), ((3541, 3552), 'numpy.array', 'np.array', (['(2)'], {}), '(2)\n', (3549, 3552), True, 'import numpy as np\n'), ((4841, 4853), 'numpy.array', 'np.array', (['(-1)'], {}), '(-1)\n', (4849, 4853), True, 'import numpy as np\n'), ((4890, 4901), 'numpy.array', 'np.array', (['(0)'], {}), '(0)\n', (4898, 4901), True, 'import numpy as np\n'), ((4937, 4948), 'numpy.array', 'np.array', (['(1)'], {}), '(1)\n', (4945, 4948), True, 'import numpy as np\n'), ((4993, 5005), 'numpy.array', 'np.array', (['(-1)'], {}), '(-1)\n', (5001, 5005), True, 'import numpy as np\n'), ((5057, 5068), 'numpy.array', 'np.array', (['(0)'], {}), '(0)\n', (5065, 5068), True, 'import numpy as np\n'), ((5119, 5130), 'numpy.array', 'np.array', (['(1)'], {}), '(1)\n', (5127, 5130), True, 'import numpy as np\n')]
import numpy as np from ipywidgets import widgets import matplotlib.pyplot as plt from itk import binary_dilate_image_filter,binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter class OpenCloseWidget(): def __init__(self): self.image = np.zeros((50,50), dtype=np.uint8) self.image[13, 13] = 255 self.image[20:30, 20:23] = 255 self.image[30, 21] = 255 self.image[31:40, 20:23] = 255 self.image_copy = self.image reset_button = widgets.Button(description='Reset Image') open_button = widgets.Button(description='Open') close_button = widgets.Button(description='Close') reset_button.on_click(self.reset) open_button.on_click(self.opening) close_button.on_click(self.closing) display(widgets.HBox([open_button, close_button, reset_button])) self.fig = plt.figure(figsize=(5,5)) self.img_obj = plt.imshow(self.image, origin='lower') plt.clim((0,255)) def opening(self, event): itk_image = GetImageFromArray(self.image) self.image = GetArrayFromImage(binary_morphological_opening_image_filter(itk_image)) self.img_obj.set_data(self.image) def closing(self, event): itk_image = GetImageFromArray(self.image) self.image = GetArrayFromImage(binary_morphological_closing_image_filter(itk_image)) self.img_obj.set_data(self.image) def reset(self, event): self.image = self.image_copy self.img_obj.set_data(self.image_copy) class DilateErodeWidget(): def __init__(self): self.image = np.zeros((50,50), dtype=np.uint8) self.image[13, 13] = 255 self.image[20:30, 20:23] = 255 self.image[30, 21] = 255 self.image[31:40, 20:23] = 255 self.image_copy = self.image reset_button = widgets.Button(description='Reset Image') dilate_button = widgets.Button(description='Dilate') erode_button = widgets.Button(description='Erode') dilate_button.on_click(self.dilate) erode_button.on_click(self.erode) reset_button.on_click(self.reset) display(widgets.HBox([dilate_button, erode_button, reset_button])) self.fig = plt.figure(figsize=(5,5)) self.img_obj = plt.imshow(self.image, origin='lower') plt.clim((0,255)) def dilate(self, event): itk_image = GetImageFromArray(self.image) self.image = GetArrayFromImage(binary_dilate_image_filter(itk_image)) self.img_obj.set_data(self.image) def erode(self, event): itk_image = GetImageFromArray(self.image) self.image = GetArrayFromImage(binary_erode_image_filter(itk_image)) self.img_obj.set_data(self.image) def reset(self, event): self.image = self.image_copy self.img_obj.set_data(self.image_copy) class Drawer(): def __init__(self, paint_width=1, paint_value = 255, erase_value=0): self.drawing = False self.paint_width = paint_width self.paint_value = paint_value self.erase_value = erase_value self.image = self.create_image() dilate_button = widgets.Button(description='Dilate') erode_button = widgets.Button(description='Erode') open_button = widgets.Button(description='Open') close_button = widgets.Button(description='Close') reset_button = widgets.Button(description='Reset Image') dilate_button.on_click(self.dilate) erode_button.on_click(self.erode) open_button.on_click(self.opening) close_button.on_click(self.closing) reset_button.on_click(self.reset) display(widgets.HBox([dilate_button,erode_button, open_button, close_button, reset_button])) self.fig = plt.figure(figsize=(5,5)) self.img_obj = plt.imshow(self.image, origin='lower') plt.clim((0,255)) plt.show() self.fig.canvas.mpl_connect('button_press_event', self.onclick) self.fig.canvas.mpl_connect('button_release_event', self.onrelease) self.fig.canvas.mpl_connect('motion_notify_event', self.onmove) def create_image(self): return np.zeros((100,100), dtype=np.uint8) def dilate(self, event): itk_image = GetImageFromArray(self.image) self.image = GetArrayFromImage(binary_dilate_image_filter(itk_image)) self.img_obj.set_data(self.image) def erode(self, event): itk_image = GetImageFromArray(self.image) self.image = GetArrayFromImage(binary_erode_image_filter(itk_image)) self.img_obj.set_data(self.image) def reset(self, event): self.image = self.create_image() self.img_obj.set_data(self.image) def opening(self, event): itk_image = GetImageFromArray(self.image) self.image = GetArrayFromImage(binary_morphological_opening_image_filter(itk_image)) self.img_obj.set_data(self.image) def closing(self, event): itk_image = GetImageFromArray(self.image) self.image = GetArrayFromImage(binary_morphological_closing_image_filter(itk_image)) self.img_obj.set_data(self.image) def onclick(self, event): self.drawing = True if event.button == 1: self.draw_point(int(event.xdata), int(event.ydata), self.paint_value) elif event.button == 3: self.draw_point(int(event.xdata), int(event.ydata), self.erase_value) def onmove(self, event): if event.button == 1: self.draw_point(int(event.xdata), int(event.ydata), self.paint_value) elif event.button == 3: self.draw_point(int(event.xdata), int(event.ydata), self.erase_value) def draw_point(self, ix, iy, value): if self.drawing == True: if self.paint_width == 0: self.image[iy, ix] = value self.img_obj._A.data[iy, ix] = value else: self.image[iy-self.paint_width : iy+self.paint_width, ix-self.paint_width : ix+self.paint_width] = value self.img_obj._A.data[iy-self.paint_width : iy+self.paint_width, ix-self.paint_width : ix+self.paint_width] = value plt.draw() def onrelease(self, event): self.drawing=False
[ "matplotlib.pyplot.imshow", "matplotlib.pyplot.clim", "matplotlib.pyplot.draw", "itk.binary_erode_image_filter", "itk.binary_morphological_opening_image_filter", "ipywidgets.widgets.Button", "numpy.zeros", "matplotlib.pyplot.figure", "ipywidgets.widgets.HBox", "itk.GetImageFromArray", "itk.binar...
[((386, 420), 'numpy.zeros', 'np.zeros', (['(50, 50)'], {'dtype': 'np.uint8'}), '((50, 50), dtype=np.uint8)\n', (394, 420), True, 'import numpy as np\n'), ((626, 667), 'ipywidgets.widgets.Button', 'widgets.Button', ([], {'description': '"""Reset Image"""'}), "(description='Reset Image')\n", (640, 667), False, 'from ipywidgets import widgets\n'), ((690, 724), 'ipywidgets.widgets.Button', 'widgets.Button', ([], {'description': '"""Open"""'}), "(description='Open')\n", (704, 724), False, 'from ipywidgets import widgets\n'), ((748, 783), 'ipywidgets.widgets.Button', 'widgets.Button', ([], {'description': '"""Close"""'}), "(description='Close')\n", (762, 783), False, 'from ipywidgets import widgets\n'), ((1016, 1042), 'matplotlib.pyplot.figure', 'plt.figure', ([], {'figsize': '(5, 5)'}), '(figsize=(5, 5))\n', (1026, 1042), True, 'import matplotlib.pyplot as plt\n'), ((1065, 1103), 'matplotlib.pyplot.imshow', 'plt.imshow', (['self.image'], {'origin': '"""lower"""'}), "(self.image, origin='lower')\n", (1075, 1103), True, 'import matplotlib.pyplot as plt\n'), ((1112, 1130), 'matplotlib.pyplot.clim', 'plt.clim', (['(0, 255)'], {}), '((0, 255))\n', (1120, 1130), True, 'import matplotlib.pyplot as plt\n'), ((1181, 1210), 'itk.GetImageFromArray', 'GetImageFromArray', (['self.image'], {}), '(self.image)\n', (1198, 1210), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((1402, 1431), 'itk.GetImageFromArray', 'GetImageFromArray', (['self.image'], {}), '(self.image)\n', (1419, 1431), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((1754, 1788), 'numpy.zeros', 'np.zeros', (['(50, 50)'], {'dtype': 'np.uint8'}), '((50, 50), dtype=np.uint8)\n', (1762, 1788), True, 'import numpy as np\n'), ((1994, 2035), 'ipywidgets.widgets.Button', 'widgets.Button', ([], {'description': '"""Reset Image"""'}), "(description='Reset Image')\n", (2008, 2035), False, 'from ipywidgets import widgets\n'), ((2060, 2096), 'ipywidgets.widgets.Button', 'widgets.Button', ([], {'description': '"""Dilate"""'}), "(description='Dilate')\n", (2074, 2096), False, 'from ipywidgets import widgets\n'), ((2120, 2155), 'ipywidgets.widgets.Button', 'widgets.Button', ([], {'description': '"""Erode"""'}), "(description='Erode')\n", (2134, 2155), False, 'from ipywidgets import widgets\n'), ((2389, 2415), 'matplotlib.pyplot.figure', 'plt.figure', ([], {'figsize': '(5, 5)'}), '(figsize=(5, 5))\n', (2399, 2415), True, 'import matplotlib.pyplot as plt\n'), ((2438, 2476), 'matplotlib.pyplot.imshow', 'plt.imshow', (['self.image'], {'origin': '"""lower"""'}), "(self.image, origin='lower')\n", (2448, 2476), True, 'import matplotlib.pyplot as plt\n'), ((2485, 2503), 'matplotlib.pyplot.clim', 'plt.clim', (['(0, 255)'], {}), '((0, 255))\n', (2493, 2503), True, 'import matplotlib.pyplot as plt\n'), ((2553, 2582), 'itk.GetImageFromArray', 'GetImageFromArray', (['self.image'], {}), '(self.image)\n', (2570, 2582), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((2760, 2789), 'itk.GetImageFromArray', 'GetImageFromArray', (['self.image'], {}), '(self.image)\n', (2777, 2789), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((3340, 3376), 'ipywidgets.widgets.Button', 'widgets.Button', ([], {'description': '"""Dilate"""'}), "(description='Dilate')\n", (3354, 3376), False, 'from ipywidgets import widgets\n'), ((3400, 3435), 'ipywidgets.widgets.Button', 'widgets.Button', ([], {'description': '"""Erode"""'}), "(description='Erode')\n", (3414, 3435), False, 'from ipywidgets import widgets\n'), ((3458, 3492), 'ipywidgets.widgets.Button', 'widgets.Button', ([], {'description': '"""Open"""'}), "(description='Open')\n", (3472, 3492), False, 'from ipywidgets import widgets\n'), ((3516, 3551), 'ipywidgets.widgets.Button', 'widgets.Button', ([], {'description': '"""Close"""'}), "(description='Close')\n", (3530, 3551), False, 'from ipywidgets import widgets\n'), ((3575, 3616), 'ipywidgets.widgets.Button', 'widgets.Button', ([], {'description': '"""Reset Image"""'}), "(description='Reset Image')\n", (3589, 3616), False, 'from ipywidgets import widgets\n'), ((3955, 3981), 'matplotlib.pyplot.figure', 'plt.figure', ([], {'figsize': '(5, 5)'}), '(figsize=(5, 5))\n', (3965, 3981), True, 'import matplotlib.pyplot as plt\n'), ((4004, 4042), 'matplotlib.pyplot.imshow', 'plt.imshow', (['self.image'], {'origin': '"""lower"""'}), "(self.image, origin='lower')\n", (4014, 4042), True, 'import matplotlib.pyplot as plt\n'), ((4051, 4069), 'matplotlib.pyplot.clim', 'plt.clim', (['(0, 255)'], {}), '((0, 255))\n', (4059, 4069), True, 'import matplotlib.pyplot as plt\n'), ((4088, 4098), 'matplotlib.pyplot.show', 'plt.show', ([], {}), '()\n', (4096, 4098), True, 'import matplotlib.pyplot as plt\n'), ((4377, 4413), 'numpy.zeros', 'np.zeros', (['(100, 100)'], {'dtype': 'np.uint8'}), '((100, 100), dtype=np.uint8)\n', (4385, 4413), True, 'import numpy as np\n'), ((4463, 4492), 'itk.GetImageFromArray', 'GetImageFromArray', (['self.image'], {}), '(self.image)\n', (4480, 4492), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((4670, 4699), 'itk.GetImageFromArray', 'GetImageFromArray', (['self.image'], {}), '(self.image)\n', (4687, 4699), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((4986, 5015), 'itk.GetImageFromArray', 'GetImageFromArray', (['self.image'], {}), '(self.image)\n', (5003, 5015), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((5207, 5236), 'itk.GetImageFromArray', 'GetImageFromArray', (['self.image'], {}), '(self.image)\n', (5224, 5236), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((939, 994), 'ipywidgets.widgets.HBox', 'widgets.HBox', (['[open_button, close_button, reset_button]'], {}), '([open_button, close_button, reset_button])\n', (951, 994), False, 'from ipywidgets import widgets\n'), ((1250, 1302), 'itk.binary_morphological_opening_image_filter', 'binary_morphological_opening_image_filter', (['itk_image'], {}), '(itk_image)\n', (1291, 1302), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((1471, 1523), 'itk.binary_morphological_closing_image_filter', 'binary_morphological_closing_image_filter', (['itk_image'], {}), '(itk_image)\n', (1512, 1523), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((2310, 2367), 'ipywidgets.widgets.HBox', 'widgets.HBox', (['[dilate_button, erode_button, reset_button]'], {}), '([dilate_button, erode_button, reset_button])\n', (2322, 2367), False, 'from ipywidgets import widgets\n'), ((2622, 2659), 'itk.binary_dilate_image_filter', 'binary_dilate_image_filter', (['itk_image'], {}), '(itk_image)\n', (2648, 2659), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((2829, 2865), 'itk.binary_erode_image_filter', 'binary_erode_image_filter', (['itk_image'], {}), '(itk_image)\n', (2854, 2865), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((3850, 3938), 'ipywidgets.widgets.HBox', 'widgets.HBox', (['[dilate_button, erode_button, open_button, close_button, reset_button]'], {}), '([dilate_button, erode_button, open_button, close_button,\n reset_button])\n', (3862, 3938), False, 'from ipywidgets import widgets\n'), ((4532, 4569), 'itk.binary_dilate_image_filter', 'binary_dilate_image_filter', (['itk_image'], {}), '(itk_image)\n', (4558, 4569), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((4739, 4775), 'itk.binary_erode_image_filter', 'binary_erode_image_filter', (['itk_image'], {}), '(itk_image)\n', (4764, 4775), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((5055, 5107), 'itk.binary_morphological_opening_image_filter', 'binary_morphological_opening_image_filter', (['itk_image'], {}), '(itk_image)\n', (5096, 5107), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((5276, 5328), 'itk.binary_morphological_closing_image_filter', 'binary_morphological_closing_image_filter', (['itk_image'], {}), '(itk_image)\n', (5317, 5328), False, 'from itk import binary_dilate_image_filter, binary_morphological_closing_image_filter, binary_morphological_opening_image_filter, binary_erode_image_filter, GetArrayFromImage, GetImageFromArray, label_image_to_shape_label_map_filter\n'), ((6485, 6495), 'matplotlib.pyplot.draw', 'plt.draw', ([], {}), '()\n', (6493, 6495), True, 'import matplotlib.pyplot as plt\n')]
import random import numpy as np import scipy import time import json import os import pdb import pickle import pandas from progressbar import * from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot from keras.models import Sequential, load_model, Model from keras.optimizers import RMSprop, Adam, SGD from keras import backend as K from keras import regularizers from keras.utils.np_utils import to_categorical from utils import convnet_vgg, convnet_mod, convnet_ori, convnet_com def softmax(x): e_x = np.exp(x - np.max(x)) return e_x / e_x.sum() #return x / np.linalg.norm(x) def makeFunc(x): return lambda y:y[:,x] class BaseListenerNetwork(object): def __init__(self, modelname, optfilename, lr, entropy_coefficient, config_dict): self.modelname = modelname self.optfilename = optfilename self.lr = lr self.entropy_coefficient = entropy_coefficient assert config_dict, "config_dict does not exist" self.config = config_dict self.initialize_model() self.build_train_fn() def rebuild_train_fn(self, entropy_coefficient=None, lr=None): if entropy_coefficient: self.entropy_coefficient = entropy_coefficient if lr: self.lr = lr self.build_train_fn() def save(self): self.listener_model.save(self.modelname) def load(self): self.listener_model = load_model(self.modelname) def save_weights(self): self.listener_model.save_weights(self.modelname) def load_weights(self): self.listener_model.load_weights(self.modelname) def save_opt(self): symbolic_weights = self.opt.weights weight_values = K.batch_get_value(symbolic_weights) with open(self.optfilename, 'wb') as f: pickle.dump(weight_values, f) def load_opt(self): with open(self.optfilename, 'rb') as f: weight_values = pickle.load(f) self.opt.set_weights(weight_values) def save_memory(self): self.memory_model_weights = self.listener_model.get_weights() def load_memory(self): self.listener_model.set_weights(self.memory_model_weights) class PaperListenerNetwork(BaseListenerNetwork): def __init__(self, modelname, optfilename, lr, entropy_coefficient, config_dict): super(PaperListenerNetwork, self).__init__(modelname, optfilename, lr, entropy_coefficient, config_dict) self.batch_speaker_message = [] self.batch_action = [] self.batch_candidates = [] self.batch_reward = [] def initialize_model(self): """ Batch input and output. """ if not os.path.exists(self.modelname): ## Define model t_input = Input(shape=(self.config['max_message_length'],)) #Speakers Message, shape(bs, max_message_length) c_inputs_all = Input(shape=(self.config['n_classes'], self.config['speaker_input_dim'])) #Candidates, shape(bs, n_class, speaker_input_dim) inputs = [t_input, c_inputs_all] z = Dense(self.config['speaker_input_dim'], activation='sigmoid')(t_input) #shape(bs, speaker_input_dim) ts = [] us = [] for _ in range(self.config['n_classes']): #c_input = Input(shape=(self.config['speaker_input_dim'],)) #shape(bs, speaker_input_dim) c_input = Lambda(makeFunc(_))(c_inputs_all) #shape(bs, speaker_input_dim) #t = Lambda(lambda x: K.expand_dims(K.sum(-K.square(x), axis=1)))(add([t_trans, Lambda(lambda x: -x)(c_input)])) #shape(bs, 1) t = Dot(1, False)([z, c_input]) #shape(bs, 1) ts.append(t) us.append(c_input) U = concatenate(ts) #shape(bs, n_classes) us = concatenate(us) final_output = Lambda(lambda x: K.softmax(x))(U) #shape(bs, n_classes) #final_output = Dense(self.n_classes, activation='softmax', kernel_initializer='identity')(U) #final_output = Dense(self.n_classes, activation='softmax')(U) #f1 = Dense(50)(U) #f2 = Lambda(lambda x: K.square(x))(f1) #final_output = Dense(self.n_classes, activation='softmax')(f2) self.listener_model = Model(inputs=inputs, outputs=[final_output, U, z, us]) #self.listener_model.compile(loss="categorical_crossentropy", optimizer=RMSprop(lr=self.config['listener_lr'])) else: self.load() #check!!! def build_train_fn(self): """ Batch input and output. """ #direct prob input!!! action_prob_placeholder = self.listener_model.output[0] #(bs, n_classes) action_onehot_placeholder = K.placeholder(shape=(None, self.config['n_classes']), name="action_onehot") #(bs, n_classes) reward_placeholder = K.placeholder(shape=(None,), name="reward") #(?) action_prob = K.sum(action_prob_placeholder * action_onehot_placeholder, axis=1) log_action_prob = K.log(action_prob) loss = - log_action_prob * reward_placeholder entropy = K.sum(action_prob_placeholder * K.log(action_prob_placeholder + 1e-10), axis=1) #entropy = K.sum(entropy) loss = loss + self.entropy_coefficient * entropy loss = K.mean(loss) self.opt = Adam(lr=self.lr) self.updates = self.opt.get_updates(params=self.listener_model.trainable_weights, loss=loss) if os.path.exists(self.optfilename): self.load_opt() self.train_fn = K.function( inputs = self.listener_model.input + [action_onehot_placeholder, reward_placeholder], outputs=[loss, loss], updates=self.updates) def reshape_message_candidates(self, speaker_message, candidates): assert len(speaker_message.shape)==1 and speaker_message.shape[0]==self.config['max_message_length'] assert len(candidates.shape)==2 and candidates.shape[0]==self.config['n_classes'] and candidates.shape[1]==self.config['speaker_input_dim'] speaker_message = np.expand_dims(speaker_message, axis=0) #shape(1, max_message_length) #X = [speaker_message] + [c.reshape([1,-1]) for c in candidates] X = [speaker_message, np.expand_dims(candidates, axis=0)] return X def sample_from_listener_policy(self, speaker_message, candidates): """ Input and output are all just one instance. No bs dimensize. """ X = self.reshape_message_candidates(speaker_message, candidates) listener_output= self.listener_model.predict_on_batch(X) y, U, z = listener_output[:3] #us = listener_output[3] listener_probs = y listener_probs = np.squeeze(listener_probs) #shape(n_class) listener_action = np.random.choice(np.arange(self.config['n_classes']), p=listener_probs) #int U = np.squeeze(U) return listener_action, listener_probs, U def infer_from_listener_policy(self, speaker_message, candidates): """ Input and output are all just one instance. No bs dimensize. """ X = self.reshape_message_candidates(speaker_message, candidates) listener_output= self.listener_model.predict_on_batch(X) y, U, z = listener_output[:3] #us = listener_output[3] listener_probs = y listener_probs = np.squeeze(listener_probs) #shape(n_class) listener_action = np.argmax(listener_probs) #int U = np.squeeze(U) return listener_action, listener_probs, U def train_listener_policy_on_batch(self): """ Train as a batch. Loss is an float for a batch """ action_onehot = to_categorical(self.batch_action, num_classes=self.config['n_classes']) #self.batch_candidates = np.array(self.batch_candidates).transpose([1, 0, 2]).tolist() #shape(num_classes, bs, speaker_input_dim) #self.batch_candidates = np.swapaxes(np.array(self.batch_candidates), 0, 1).tolist() #shape(num_classes, bs, speaker_input_dim) #self.batch_candidates = np.swapaxes(np.array(self.batch_candidates), 0, 1).astype('float32').tolist() #shape(num_classes, bs, speaker_input_dim) #self.batch_candidates = [np.array(_) for _ in self.batch_candidates] #_loss, _entropy = self.train_fn([self.batch_speaker_message] + self.batch_candidates + [action_onehot, self.batch_reward] ) _loss, _entropy = self.train_fn([np.array(self.batch_speaker_message), self.batch_candidates, action_onehot, self.batch_reward] ) #print("Listener loss: ", _loss) self.batch_speaker_message = [] #shape(bs, max_message_length) self.batch_action = [] #shape(bs) self.batch_candidates = [] #shape(bs, n_classes, speaker_input_dim) self.batch_reward = [] #shape(bs) def remember_listener_training_details(self, speaker_message, action, action_probs, target, candidates, reward): """ Inputs are just one instance. No bs dimensize. """ self.batch_speaker_message.append(speaker_message) self.batch_action.append(action) self.batch_candidates.append(candidates) self.batch_reward.append(reward) class PaperListenerNetwork_rnn(PaperListenerNetwork): def reshape_message_candidates(self, speaker_message, candidates): #if not self.config['fixed_length']: # assert len(speaker_message.shape)==1 and speaker_message.shape[0]<=self.config['max_message_length'] #else: # assert len(speaker_message.shape)==1 and speaker_message.shape[0]==self.config['max_message_length'] assert len(speaker_message.shape)==1 and speaker_message.shape[0]<=self.config['max_message_length'] assert len(candidates.shape)==2 and candidates.shape[0]==self.config['n_classes'] and candidates.shape[1]==self.config['speaker_input_dim'] speaker_message = np.expand_dims(to_categorical(speaker_message, self.config['alphabet_size']), axis=0) #shape(1, message_length, alphabet_size) #X = [speaker_message] + [c.reshape([1,-1]) for c in candidates] X = [speaker_message, np.expand_dims(candidates, axis=0)] return X def initialize_model(self): """ Batch input and output. """ ## Define model if not os.path.exists(self.modelname): t_input = Input(shape=(None, self.config['alphabet_size'],)) #Speakers Message, shape(bs, message_length, alphabet_size) #c_inputs_all = Input(shape=(self.config['n_classes'], self.config['speaker_input_dim'])) #Candidates, shape(bs, n_classes, speaker_input_dim) c_inputs_all = Input(shape=(None, self.config['speaker_input_dim'])) #Candidates, shape(bs, n_classes, speaker_input_dim) inputs = [t_input, c_inputs_all] lstm = LSTM(self.config['listener_dim'], activation='tanh', return_sequences=False, return_state=True) o, sh, sc = lstm(t_input) z = Dense(self.config['listener_dim'], activation='sigmoid')(o) #shape(bs, listener_dim) ts = [] us = [] u = Dense(self.config['listener_dim'], activation='sigmoid') for _ in range(self.config['n_classes']): #c_input = Input(shape=(self.config['speaker_input_dim'],)) #shape(bs, speaker_input_dim) c_input = Lambda(makeFunc(_))(c_inputs_all) uc = u(c_input) t = Lambda(lambda x: K.expand_dims(K.sum(-K.square(x), axis=1)))(add([z, Lambda(lambda x: -x)(uc)])) #shape(bs, 1) #t = Dot(1, False)([z,uc]) #shape(bs, 1) ts.append(t) us.append(uc) U = concatenate(ts) #shape(bs, n_classes) us = concatenate(us) final_output = Lambda(lambda x: K.softmax(x))(U) #shape(bs, n_classes) self.listener_model = Model(inputs=inputs, outputs=[final_output, U, z, us]) #self.listener_model.compile(loss="categorical_crossentropy", optimizer=RMSprop(lr=self.config['listener_lr'])) else: self.load() #check!!! def set_updates(self): self.opt = Adam(lr=self.lr) #adam = RMSprop(lr=self.lr) self.updates = self.opt.get_updates(params=self.listener_model.trainable_weights, loss=self.loss) if os.path.exists(self.optfilename): self.load_opt() def build_train_fn(self): """ Batch input and output. """ #direct prob input!!! action_prob_placeholder = self.listener_model.output[0] #(bs, n_classes) #action_onehot_placeholder = K.placeholder(shape=(None, self.config['n_classes']), name="action_onehot") #(bs, n_classes) action_onehot_placeholder = K.placeholder(shape=(None, None), name="action_onehot") #(bs, n_classes) reward_placeholder = K.placeholder(shape=(None,), name="reward") #(?) action_prob = K.sum(action_prob_placeholder*action_onehot_placeholder, axis=1) log_action_prob = K.log(action_prob) loss = - log_action_prob*reward_placeholder entropy = K.sum(action_prob_placeholder * K.log(action_prob_placeholder + 1e-10), axis=1) #entropy = K.sum(entropy) loss = loss + self.entropy_coefficient * entropy loss = K.mean(loss) self.loss =loss self.set_updates() self.train_fn = K.function( inputs = self.listener_model.input + [action_onehot_placeholder, reward_placeholder], outputs=[loss, loss], updates=self.updates) def remember_listener_training_details(self, speaker_message, action, action_probs, target, candidates, reward): """ Inputs are just one instance. No bs dimensize. """ #if not self.config['fixed_length']: toadd = self.config['max_message_length'] - len(speaker_message) for _ in range(toadd): speaker_message = np.append(speaker_message, -1) speaker_message = to_categorical(speaker_message, self.config['alphabet_size']) #shape(message_length, alphabet_size) self.batch_speaker_message.append(speaker_message) self.batch_action.append(action) self.batch_candidates.append(candidates) self.batch_reward.append(reward) class PaperListenerNetwork_rnn_conv(PaperListenerNetwork_rnn): def __init__(self, modelname, optfilename, lr, entropy_coefficient, pretrain_convmodel_file, traincnn, config): self.pretrain_convmodel_file = pretrain_convmodel_file self.traincnn = traincnn super(PaperListenerNetwork_rnn_conv, self).__init__(modelname, optfilename, lr, entropy_coefficient, config) def initialize_model(self): """ Batch input and output. """ if not os.path.exists(self.modelname): ## Define model self.conv_model = convnet_com(self.config['speaker_input_w'], self.config['speaker_input_h'], 3, preloadfile=self.pretrain_convmodel_file, name='conv_model_l') t_input = Input(shape=(None, self.config['alphabet_size'],)) #Speakers Message, shape(bs, message_length, alphabet_size) c_inputs_all = Input(shape=(self.config['n_classes'], self.config['speaker_input_w'], self.config['speaker_input_h'], 3), name='image_l') #Candidates, shape(bs, speaker_input_w, speaker_input_h, 3) inputs = [t_input, c_inputs_all] lstm = LSTM(self.config['listener_dim'], activation='tanh', return_sequences=False, return_state=True) o, sh, sc = lstm(t_input) z = Dense(self.config['listener_dim'], activation='sigmoid')(o) #shape(bs, listener_dim) #u = Dense(self.config['listener_dim'], activation='sigmoid',kernel_regularizer=regularizers.l2(0.01)) u = Dense(self.config['listener_dim'], activation='sigmoid') ts = [] us = [] for _ in range(self.config['n_classes']): #c_input = Input(shape=(self.config['speaker_input_w'],self.config['speaker_input_h'],3)) #speaker_model.input[0], shape(bs, speaker_input_w, speaker_input_h, 3) #c_input = Lambda(lambda x: x[:, _])(c_inputs_all) c_input = Lambda(makeFunc(_))(c_inputs_all) conv_outputs = self.conv_model(c_input) uc = u(conv_outputs) t = Lambda(lambda x: K.expand_dims(K.sum(-K.square(x),axis=1)))(add([z, Lambda(lambda x: -x)(uc)])) #shape(bs, 1) #t = Dot(1, False)([z,uc]) #shape(bs, 1) ts.append(t) us.append(uc) U = concatenate(ts) #shape(bs, n_classes) us = concatenate(us) final_output = Lambda(lambda x: K.softmax(x))(U) #shape(bs, n_classes) self.listener_model = Model(inputs=inputs, outputs=[final_output, U, z, us]) #self.listener_model.compile(loss="categorical_crossentropy", optimizer=RMSprop(lr=self.config['listener_lr'])) else: self.load() #check!!! self.conv_model = [l for l in self.listener_model.layers if l.name=='conv_model_l'][0] #self.listener_model.layers[6].kernel_regularizer = None #self.internal_model = Model(inputs=self.listener_model.inputs, outputs=[self.listener_model.layers[7].get_output_at(_) for _ in range(2)] + [self.listener_model.layers[6].output, self.listener_model.layers[-2].output]) #dot #self.internal_model = Model(inputs=self.listener_model.inputs, outputs=[self.listener_model.layers[6].get_output_at(_) for _ in range(2)] + [self.listener_model.layers[7].output, self.listener_model.layers[-2].output]) #euc self.trainable_weights_others = [] self.trainable_weights_conv = [] for layer in self.listener_model.layers: if layer.name!='conv_model_l': self.trainable_weights_others.extend(layer.trainable_weights) else: self.trainable_weights_conv.extend(layer.trainable_weights) def set_updates(self): self.opt = Adam(lr=self.lr) #self.opt = RMSprop(lr=self.lr) #opt = SGD(lr=self.lr, momentum=0.9, decay=1e-6, nesterov=True) if not self.traincnn: #self.updates = self.opt.get_updates(params=self.trainable_weights_others+self.trainable_weights_rnn, loss=self.loss) self.updates = self.opt.get_updates(params=self.trainable_weights_others, loss=self.loss) else: self.updates = self.opt.get_updates(params=self.listener_model.trainable_weights, loss=self.loss) if os.path.exists(self.optfilename): self.load_opt() def reshape_message_candidates(self, speaker_message, candidates): #if not self.config['fixed_length']: # assert len(speaker_message.shape)==1 and speaker_message.shape[0]<=self.config['max_message_length'] #else: # assert len(speaker_message.shape)==1 and speaker_message.shape[0]==self.config['max_message_length'] assert len(speaker_message.shape)==1 and speaker_message.shape[0]<=self.config['max_message_length'] assert len(candidates.shape)==4 and candidates.shape[0]==self.config['n_classes'] and candidates.shape[1]==self.config['speaker_input_w'] and candidates.shape[2]==self.config['speaker_input_h'] speaker_message = np.expand_dims(to_categorical(speaker_message, self.config['alphabet_size']), axis=0) #shape(1, ?, alphabet_size) X = [speaker_message, np.expand_dims(candidates, axis=0)] return X ''' class PaperListenerNetwork_rnn_conv_color(PaperListenerNetwork_rnn): def initialize_model(self): """ Batch input and output. """ if not os.path.exists(self.modelname): ## Define model t_input = Input(shape=(None, self.config['alphabet_size'],)) #Speakers Message, shape(bs, message_length, alphabet_size) c_inputs_all = Input(shape=(self.config['n_classes'], 8)) inputs = [t_input, c_inputs_all] lstm = LSTM(self.config['listener_dim'], activation='tanh', return_sequences=False, return_state=True) o, sh, sc = lstm(t_input) z = Dense(self.config['listener_dim'], activation='sigmoid')(o) #shape(bs, listener_dim) u = Dense(self.config['listener_dim'], activation='sigmoid') ts = [] for _ in range(self.config['n_classes']): #c_input = Input(shape=(self.config['speaker_input_w'],self.config['speaker_input_h'],3)) #speaker_model.input[0], shape(bs, speaker_input_w, speaker_input_h, 3) #c_input = Lambda(lambda x: x[:, _])(c_inputs_all) c_input = Lambda(makeFunc(_))(c_inputs_all) #conv_outputs = conv_model(c_input) #conv_outputs = c_input uc = u(c_input) t = Lambda(lambda x: K.expand_dims(K.sum(-K.square(x),axis=1)))(add([z, Lambda(lambda x: -x)(uc)])) #shape(bs, 1) ts.append(t) U = concatenate(ts) #shape(bs, n_classes) final_output = Lambda(lambda x: K.softmax(x))(U) #shape(bs, n_classes) self.listener_model = Model(inputs=inputs, outputs=[final_output, z, U]) #self.listener_model.compile(loss="categorical_crossentropy", optimizer=RMSprop(lr=self.config['listener_lr'])) else: self.load() #check!!! self.trainable_weights_rnn = self.listener_model.trainable_weights[:3] self.trainable_weights_others = self.listener_model.trainable_weights[3:] def set_updates(self): self.opt = Adam(lr=self.lr) #opt = RMSprop(lr=self.lr) #opt = SGD(lr=self.lr, momentum=0.9, decay=1e-6, nesterov=True) self.updates = self.opt.get_updates(params=self.listener_model.trainable_weights, loss=self.loss) if os.path.exists(self.optfilename): self.load_opt() def reshape_message_candidates(self, speaker_message, candidates): #if not self.config['fixed_length']: # assert len(speaker_message.shape)==1 and speaker_message.shape[0]<=self.config['max_message_length'] #else: # assert len(speaker_message.shape)==1 and speaker_message.shape[0]==self.config['max_message_length'] #pdb.set_trace() assert len(speaker_message.shape)==1 and speaker_message.shape[0]<=self.config['max_message_length'] assert len(candidates.shape)==2 and candidates.shape[0]==self.config['n_classes'] and candidates.shape[1]==8 speaker_message = np.expand_dims(to_categorical(speaker_message, self.config['alphabet_size']), axis=0) #shape(1, ?, alphabet_size) X = [speaker_message, np.expand_dims(candidates, axis=0)] return X class PaperListenerNetwork_direct(BaseListenerNetwork): def __init__(self, modelname, config_dict): assert False #TOMODIFY super(PaperListenerNetwork_direct, self).__init__(modelname, config_dict) self.batch_speaker_message = [] self.batch_action = [] self.batch_candidates = [] self.batch_reward = [] def initialize_model(self): """ Batch input and output. """ if not os.path.exists(self.modelname): ## Define model ## Speakers Message t_input = Input(shape=(self.config['max_message_length'],)) #shape(bs, max_message_length) t_trans = Dense(self.config['speaker_input_dim'], #kernel_initializer=keras.initializers.Identity(gain=1.0), #bias_initializer='zeros', activation='sigmoid')(t_input) #shape(bs, speaker_input_dim) inputs = [t_input] ts = [] for _ in range(self.config['n_classes']): c_input = Input(shape=(self.config['speaker_input_dim'],)) #shape(bs, speaker_input_dim) t = Lambda(lambda x: K.expand_dims(K.sum(-K.square(x),axis=1)))(add([t_trans, Lambda(lambda x: -x)(c_input)])) #shape(bs, 1) inputs.append(c_input) ts.append(t) U = concatenate(ts) #shape(bs, n_classes) listener_probs = U #listener_probs = Lambda(lambda x: K.softmax(x))(U) #shape(bs, n_classes) listener_infer_action = Lambda(lambda x: K.argmax(x))(U) #shape(bs) target_onehot_placeholder = Input(shape=(self.config['n_classes'],), name="action_onehot") #(bs, n_classes) listener_prob_2 = dot([listener_probs, target_onehot_placeholder], axes=1) listener_prob_2 = Lambda(lambda x:K.squeeze(x, axis=1))(listener_prob_2) self.listener_model = Model(inputs=inputs + [target_onehot_placeholder], outputs=[listener_probs, listener_infer_action, t_trans, listener_prob_2]) else: self.load() #check!!! def build_train_fn(self): """ Batch input and output. """ #direct prob input!!! #reward_placeholder = K.placeholder(shape=(None,), name="reward") #(?) action_prob = self.listener_model.output[3] #loss = K.log(-action_prob)*reward_placeholder #loss = - action_prob * reward_placeholder loss = - action_prob loss = K.mean(loss) self.opt = Adam(lr=self.config['listener_lr']) self.updates = self.opt.get_updates(params=self.listener_model.trainable_weights,loss=loss) #if os.path.exists(self.optfilename): # self.load_opt() self.train_fn = K.function( #inputs = self.listener_model.input + [reward_placeholder], inputs = self.listener_model.input, outputs=[loss, loss], updates=self.updates) def sample_from_listener_policy(self, speaker_message, candidates): """ Input and output are all just one instance. No bs dimensize. """ X = self.reshape_message_candidates(speaker_message, candidates) + [np.zeros([1, self.config['n_classes']])] listener_probs, listener_infer_action, _t_trans, _lp2 = self.listener_model.predict_on_batch(X) listener_probs = np.squeeze(listener_probs) #shape(n_class) #listener_probs = scipy.special.softmax(listener_probs) listener_probs = softmax(listener_probs) #pdb.set_trace() #???norm??? listener_action = np.random.choice(np.arange(self.config['n_classes']), p=listener_probs) #int return listener_action, listener_probs def infer_from_listener_policy(self, speaker_message, candidates): """ Input and output are all just one instance. No bs dimensize. """ X = self.reshape_message_candidates(speaker_message, candidates) + [np.zeros([1, self.config['n_classes']])] listener_probs, listener_infer_action, _t_trans, _lp2 = self.listener_model.predict_on_batch(X) listener_probs = np.squeeze(listener_probs) #shape(n_class) listener_probs = softmax(listener_probs) listener_action = np.squeeze(listener_infer_action).tolist() #int return listener_action, listener_probs def train_listener_policy_on_batch(self): """ Train as a batch. Loss is an float for a batch """ self.batch_candidates = np.array(self.batch_candidates).transpose([1, 0, 2]).tolist() #shape(num_classes, bs, speaker_input_dim #_loss, _entropy = self.train_fn([self.batch_speaker_message] + self.batch_candidates + [self.batch_action, self.batch_reward] ) _loss, _entropy = self.train_fn([self.batch_speaker_message] + self.batch_candidates + [self.batch_action] ) #print("Listener loss: ", _loss) self.batch_speaker_message = [] #shape(bs, max_message_length) self.batch_action = [] #shape(bs, n_classes) self.batch_candidates = [] #shape(bs, n_classes, speaker_input_dim) self.batch_reward = [] #shape(bs) def remember_listener_training_details(self, speaker_message, action, action_probs, target, candidates, reward): """ Inputs are just one instance. No bs dimensize. """ #action_onehot = np.zeros(self.config['n_classes']) #action_onehot[action] = 1 action_onehot = np.ones(self.config['n_classes']) * np.all(target==candidates, axis=1) self.batch_action.append(action_onehot) self.batch_speaker_message.append(speaker_message) self.batch_candidates.append(candidates) self.batch_reward.append(reward) '''
[ "keras.backend.sum", "numpy.array", "keras.layers.Dense", "numpy.arange", "os.path.exists", "keras.backend.square", "keras.backend.placeholder", "keras.layers.Dot", "numpy.max", "keras.layers.LSTM", "keras.backend.batch_get_value", "keras.layers.concatenate", "keras.models.Model", "keras.o...
[((1362, 1388), 'keras.models.load_model', 'load_model', (['self.modelname'], {}), '(self.modelname)\n', (1372, 1388), False, 'from keras.models import Sequential, load_model, Model\n'), ((1633, 1668), 'keras.backend.batch_get_value', 'K.batch_get_value', (['symbolic_weights'], {}), '(symbolic_weights)\n', (1650, 1668), True, 'from keras import backend as K\n'), ((4340, 4415), 'keras.backend.placeholder', 'K.placeholder', ([], {'shape': "(None, self.config['n_classes'])", 'name': '"""action_onehot"""'}), "(shape=(None, self.config['n_classes']), name='action_onehot')\n", (4353, 4415), True, 'from keras import backend as K\n'), ((4457, 4500), 'keras.backend.placeholder', 'K.placeholder', ([], {'shape': '(None,)', 'name': '"""reward"""'}), "(shape=(None,), name='reward')\n", (4470, 4500), True, 'from keras import backend as K\n'), ((4523, 4589), 'keras.backend.sum', 'K.sum', (['(action_prob_placeholder * action_onehot_placeholder)'], {'axis': '(1)'}), '(action_prob_placeholder * action_onehot_placeholder, axis=1)\n', (4528, 4589), True, 'from keras import backend as K\n'), ((4611, 4629), 'keras.backend.log', 'K.log', (['action_prob'], {}), '(action_prob)\n', (4616, 4629), True, 'from keras import backend as K\n'), ((4865, 4877), 'keras.backend.mean', 'K.mean', (['loss'], {}), '(loss)\n', (4871, 4877), True, 'from keras import backend as K\n'), ((4894, 4910), 'keras.optimizers.Adam', 'Adam', ([], {'lr': 'self.lr'}), '(lr=self.lr)\n', (4898, 4910), False, 'from keras.optimizers import RMSprop, Adam, SGD\n'), ((5013, 5045), 'os.path.exists', 'os.path.exists', (['self.optfilename'], {}), '(self.optfilename)\n', (5027, 5045), False, 'import os\n'), ((5086, 5228), 'keras.backend.function', 'K.function', ([], {'inputs': '(self.listener_model.input + [action_onehot_placeholder, reward_placeholder])', 'outputs': '[loss, loss]', 'updates': 'self.updates'}), '(inputs=self.listener_model.input + [action_onehot_placeholder,\n reward_placeholder], outputs=[loss, loss], updates=self.updates)\n', (5096, 5228), True, 'from keras import backend as K\n'), ((5576, 5615), 'numpy.expand_dims', 'np.expand_dims', (['speaker_message'], {'axis': '(0)'}), '(speaker_message, axis=0)\n', (5590, 5615), True, 'import numpy as np\n'), ((6168, 6194), 'numpy.squeeze', 'np.squeeze', (['listener_probs'], {}), '(listener_probs)\n', (6178, 6194), True, 'import numpy as np\n'), ((6316, 6329), 'numpy.squeeze', 'np.squeeze', (['U'], {}), '(U)\n', (6326, 6329), True, 'import numpy as np\n'), ((6755, 6781), 'numpy.squeeze', 'np.squeeze', (['listener_probs'], {}), '(listener_probs)\n', (6765, 6781), True, 'import numpy as np\n'), ((6819, 6844), 'numpy.argmax', 'np.argmax', (['listener_probs'], {}), '(listener_probs)\n', (6828, 6844), True, 'import numpy as np\n'), ((6857, 6870), 'numpy.squeeze', 'np.squeeze', (['U'], {}), '(U)\n', (6867, 6870), True, 'import numpy as np\n'), ((7047, 7118), 'keras.utils.np_utils.to_categorical', 'to_categorical', (['self.batch_action'], {'num_classes': "self.config['n_classes']"}), "(self.batch_action, num_classes=self.config['n_classes'])\n", (7061, 7118), False, 'from keras.utils.np_utils import to_categorical\n'), ((11147, 11163), 'keras.optimizers.Adam', 'Adam', ([], {'lr': 'self.lr'}), '(lr=self.lr)\n', (11151, 11163), False, 'from keras.optimizers import RMSprop, Adam, SGD\n'), ((11302, 11334), 'os.path.exists', 'os.path.exists', (['self.optfilename'], {}), '(self.optfilename)\n', (11316, 11334), False, 'import os\n'), ((11684, 11739), 'keras.backend.placeholder', 'K.placeholder', ([], {'shape': '(None, None)', 'name': '"""action_onehot"""'}), "(shape=(None, None), name='action_onehot')\n", (11697, 11739), True, 'from keras import backend as K\n'), ((11781, 11824), 'keras.backend.placeholder', 'K.placeholder', ([], {'shape': '(None,)', 'name': '"""reward"""'}), "(shape=(None,), name='reward')\n", (11794, 11824), True, 'from keras import backend as K\n'), ((11847, 11913), 'keras.backend.sum', 'K.sum', (['(action_prob_placeholder * action_onehot_placeholder)'], {'axis': '(1)'}), '(action_prob_placeholder * action_onehot_placeholder, axis=1)\n', (11852, 11913), True, 'from keras import backend as K\n'), ((11933, 11951), 'keras.backend.log', 'K.log', (['action_prob'], {}), '(action_prob)\n', (11938, 11951), True, 'from keras import backend as K\n'), ((12187, 12199), 'keras.backend.mean', 'K.mean', (['loss'], {}), '(loss)\n', (12193, 12199), True, 'from keras import backend as K\n'), ((12264, 12406), 'keras.backend.function', 'K.function', ([], {'inputs': '(self.listener_model.input + [action_onehot_placeholder, reward_placeholder])', 'outputs': '[loss, loss]', 'updates': 'self.updates'}), '(inputs=self.listener_model.input + [action_onehot_placeholder,\n reward_placeholder], outputs=[loss, loss], updates=self.updates)\n', (12274, 12406), True, 'from keras import backend as K\n'), ((12813, 12874), 'keras.utils.np_utils.to_categorical', 'to_categorical', (['speaker_message', "self.config['alphabet_size']"], {}), "(speaker_message, self.config['alphabet_size'])\n", (12827, 12874), False, 'from keras.utils.np_utils import to_categorical\n'), ((16508, 16524), 'keras.optimizers.Adam', 'Adam', ([], {'lr': 'self.lr'}), '(lr=self.lr)\n', (16512, 16524), False, 'from keras.optimizers import RMSprop, Adam, SGD\n'), ((16991, 17023), 'os.path.exists', 'os.path.exists', (['self.optfilename'], {}), '(self.optfilename)\n', (17005, 17023), False, 'import os\n'), ((558, 567), 'numpy.max', 'np.max', (['x'], {}), '(x)\n', (564, 567), True, 'import numpy as np\n'), ((1716, 1745), 'pickle.dump', 'pickle.dump', (['weight_values', 'f'], {}), '(weight_values, f)\n', (1727, 1745), False, 'import pickle\n'), ((1833, 1847), 'pickle.load', 'pickle.load', (['f'], {}), '(f)\n', (1844, 1847), False, 'import pickle\n'), ((2513, 2543), 'os.path.exists', 'os.path.exists', (['self.modelname'], {}), '(self.modelname)\n', (2527, 2543), False, 'import os\n'), ((2579, 2628), 'keras.layers.Input', 'Input', ([], {'shape': "(self.config['max_message_length'],)"}), "(shape=(self.config['max_message_length'],))\n", (2584, 2628), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((2697, 2770), 'keras.layers.Input', 'Input', ([], {'shape': "(self.config['n_classes'], self.config['speaker_input_dim'])"}), "(shape=(self.config['n_classes'], self.config['speaker_input_dim']))\n", (2702, 2770), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((3454, 3469), 'keras.layers.concatenate', 'concatenate', (['ts'], {}), '(ts)\n', (3465, 3469), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((3501, 3516), 'keras.layers.concatenate', 'concatenate', (['us'], {}), '(us)\n', (3512, 3516), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((3925, 3979), 'keras.models.Model', 'Model', ([], {'inputs': 'inputs', 'outputs': '[final_output, U, z, us]'}), '(inputs=inputs, outputs=[final_output, U, z, us])\n', (3930, 3979), False, 'from keras.models import Sequential, load_model, Model\n'), ((5739, 5773), 'numpy.expand_dims', 'np.expand_dims', (['candidates'], {'axis': '(0)'}), '(candidates, axis=0)\n', (5753, 5773), True, 'import numpy as np\n'), ((6249, 6284), 'numpy.arange', 'np.arange', (["self.config['n_classes']"], {}), "(self.config['n_classes'])\n", (6258, 6284), True, 'import numpy as np\n'), ((9143, 9204), 'keras.utils.np_utils.to_categorical', 'to_categorical', (['speaker_message', "self.config['alphabet_size']"], {}), "(speaker_message, self.config['alphabet_size'])\n", (9157, 9204), False, 'from keras.utils.np_utils import to_categorical\n'), ((9348, 9382), 'numpy.expand_dims', 'np.expand_dims', (['candidates'], {'axis': '(0)'}), '(candidates, axis=0)\n', (9362, 9382), True, 'import numpy as np\n'), ((9498, 9528), 'os.path.exists', 'os.path.exists', (['self.modelname'], {}), '(self.modelname)\n', (9512, 9528), False, 'import os\n'), ((9544, 9593), 'keras.layers.Input', 'Input', ([], {'shape': "(None, self.config['alphabet_size'])"}), "(shape=(None, self.config['alphabet_size']))\n", (9549, 9593), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((9821, 9874), 'keras.layers.Input', 'Input', ([], {'shape': "(None, self.config['speaker_input_dim'])"}), "(shape=(None, self.config['speaker_input_dim']))\n", (9826, 9874), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((9981, 10080), 'keras.layers.LSTM', 'LSTM', (["self.config['listener_dim']"], {'activation': '"""tanh"""', 'return_sequences': '(False)', 'return_state': '(True)'}), "(self.config['listener_dim'], activation='tanh', return_sequences=False,\n return_state=True)\n", (9985, 10080), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((10237, 10293), 'keras.layers.Dense', 'Dense', (["self.config['listener_dim']"], {'activation': '"""sigmoid"""'}), "(self.config['listener_dim'], activation='sigmoid')\n", (10242, 10293), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((10722, 10737), 'keras.layers.concatenate', 'concatenate', (['ts'], {}), '(ts)\n', (10733, 10737), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((10769, 10784), 'keras.layers.concatenate', 'concatenate', (['us'], {}), '(us)\n', (10780, 10784), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((10896, 10950), 'keras.models.Model', 'Model', ([], {'inputs': 'inputs', 'outputs': '[final_output, U, z, us]'}), '(inputs=inputs, outputs=[final_output, U, z, us])\n', (10901, 10950), False, 'from keras.models import Sequential, load_model, Model\n'), ((12757, 12787), 'numpy.append', 'np.append', (['speaker_message', '(-1)'], {}), '(speaker_message, -1)\n', (12766, 12787), True, 'import numpy as np\n'), ((13546, 13576), 'os.path.exists', 'os.path.exists', (['self.modelname'], {}), '(self.modelname)\n', (13560, 13576), False, 'import os\n'), ((13620, 13765), 'utils.convnet_com', 'convnet_com', (["self.config['speaker_input_w']", "self.config['speaker_input_h']", '(3)'], {'preloadfile': 'self.pretrain_convmodel_file', 'name': '"""conv_model_l"""'}), "(self.config['speaker_input_w'], self.config['speaker_input_h'],\n 3, preloadfile=self.pretrain_convmodel_file, name='conv_model_l')\n", (13631, 13765), False, 'from utils import convnet_vgg, convnet_mod, convnet_ori, convnet_com\n'), ((13778, 13827), 'keras.layers.Input', 'Input', ([], {'shape': "(None, self.config['alphabet_size'])"}), "(shape=(None, self.config['alphabet_size']))\n", (13783, 13827), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((13909, 14036), 'keras.layers.Input', 'Input', ([], {'shape': "(self.config['n_classes'], self.config['speaker_input_w'], self.config[\n 'speaker_input_h'], 3)", 'name': '"""image_l"""'}), "(shape=(self.config['n_classes'], self.config['speaker_input_w'], self\n .config['speaker_input_h'], 3), name='image_l')\n", (13914, 14036), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((14145, 14244), 'keras.layers.LSTM', 'LSTM', (["self.config['listener_dim']"], {'activation': '"""tanh"""', 'return_sequences': '(False)', 'return_state': '(True)'}), "(self.config['listener_dim'], activation='tanh', return_sequences=False,\n return_state=True)\n", (14149, 14244), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((14481, 14537), 'keras.layers.Dense', 'Dense', (["self.config['listener_dim']"], {'activation': '"""sigmoid"""'}), "(self.config['listener_dim'], activation='sigmoid')\n", (14486, 14537), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((15178, 15193), 'keras.layers.concatenate', 'concatenate', (['ts'], {}), '(ts)\n', (15189, 15193), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((15225, 15240), 'keras.layers.concatenate', 'concatenate', (['us'], {}), '(us)\n', (15236, 15240), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((15347, 15401), 'keras.models.Model', 'Model', ([], {'inputs': 'inputs', 'outputs': '[final_output, U, z, us]'}), '(inputs=inputs, outputs=[final_output, U, z, us])\n', (15352, 15401), False, 'from keras.models import Sequential, load_model, Model\n'), ((17720, 17781), 'keras.utils.np_utils.to_categorical', 'to_categorical', (['speaker_message', "self.config['alphabet_size']"], {}), "(speaker_message, self.config['alphabet_size'])\n", (17734, 17781), False, 'from keras.utils.np_utils import to_categorical\n'), ((17844, 17878), 'numpy.expand_dims', 'np.expand_dims', (['candidates'], {'axis': '(0)'}), '(candidates, axis=0)\n', (17858, 17878), True, 'import numpy as np\n'), ((2871, 2932), 'keras.layers.Dense', 'Dense', (["self.config['speaker_input_dim']"], {'activation': '"""sigmoid"""'}), "(self.config['speaker_input_dim'], activation='sigmoid')\n", (2876, 2932), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((4726, 4764), 'keras.backend.log', 'K.log', (['(action_prob_placeholder + 1e-10)'], {}), '(action_prob_placeholder + 1e-10)\n', (4731, 4764), True, 'from keras import backend as K\n'), ((7775, 7811), 'numpy.array', 'np.array', (['self.batch_speaker_message'], {}), '(self.batch_speaker_message)\n', (7783, 7811), True, 'import numpy as np\n'), ((10115, 10171), 'keras.layers.Dense', 'Dense', (["self.config['listener_dim']"], {'activation': '"""sigmoid"""'}), "(self.config['listener_dim'], activation='sigmoid')\n", (10120, 10171), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((12046, 12084), 'keras.backend.log', 'K.log', (['(action_prob_placeholder + 1e-10)'], {}), '(action_prob_placeholder + 1e-10)\n', (12051, 12084), True, 'from keras import backend as K\n'), ((14279, 14335), 'keras.layers.Dense', 'Dense', (["self.config['listener_dim']"], {'activation': '"""sigmoid"""'}), "(self.config['listener_dim'], activation='sigmoid')\n", (14284, 14335), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((3357, 3370), 'keras.layers.Dot', 'Dot', (['(1)', '(False)'], {}), '(1, False)\n', (3360, 3370), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((3553, 3565), 'keras.backend.softmax', 'K.softmax', (['x'], {}), '(x)\n', (3562, 3565), True, 'from keras import backend as K\n'), ((10821, 10833), 'keras.backend.softmax', 'K.softmax', (['x'], {}), '(x)\n', (10830, 10833), True, 'from keras import backend as K\n'), ((15277, 15289), 'keras.backend.softmax', 'K.softmax', (['x'], {}), '(x)\n', (15286, 15289), True, 'from keras import backend as K\n'), ((10583, 10603), 'keras.layers.Lambda', 'Lambda', (['(lambda x: -x)'], {}), '(lambda x: -x)\n', (10589, 10603), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((15040, 15060), 'keras.layers.Lambda', 'Lambda', (['(lambda x: -x)'], {}), '(lambda x: -x)\n', (15046, 15060), False, 'from keras.layers import Input, Dense, LSTM, Lambda, concatenate, add, Dot\n'), ((10552, 10563), 'keras.backend.square', 'K.square', (['x'], {}), '(x)\n', (10560, 10563), True, 'from keras import backend as K\n'), ((15010, 15021), 'keras.backend.square', 'K.square', (['x'], {}), '(x)\n', (15018, 15021), True, 'from keras import backend as K\n')]
import argparse import torch import random from torchvision import utils from torchvision import transforms from sklearn.decomposition import FastICA import numpy as np import random from PIL import Image, ImageDraw, ImageFont from model import Generator def ica_single_img( directions, ckpt, degree=5, channel_multiplier=2, latent=512, n_mlp=8, max_channel_size=512, size=256, truncation=0.7, device='cuda', full_model=False, initial_latent=None, resolution=64, start_component=0, end_component=None, num_of_columns=5, col=None, row=None, no_index=False, seed=None, need_PIL=False ): if row is None and need_PIL: assert "If you need a gif, please select a row!" print("Loading checkpoints...") ckpt = torch.load(ckpt) if full_model: state_dict = ckpt.state_dict() g = ckpt.to(device) else: state_dict = ckpt["g_ema"] g = Generator( size, latent, n_mlp, channel_multiplier=channel_multiplier, max_channel_size=max_channel_size ).to(device) g.load_state_dict(state_dict) components = directions num_of_components = directions.shape[1] print("Generating images..") if seed: torch.manual_seed(seed) np.random.seed(seed) random.seed(seed) trunc = g.mean_latent(4) else: trunc = g.mean_latent(4096) w_plus = False if initial_latent: proj = torch.load(initial_latent) key = list(proj.keys()) latent = proj[key[0]]['latent'].detach().to(device) # print(proj[key[0]]['noise']) noise = proj[key[0]]['noise'] if len(list(latent.shape)) == 2: w_plus = True else: latent = trunc noise = None alpha = range(-num_of_columns // 2 + 1, num_of_columns // 2 + 1) resize_transoform = transforms.Resize(resolution) to_pil_transoform = transforms.ToPILImage() to_tensor_transoform = transforms.ToTensor() directional_results = [] if row: d_range = range(num_of_components)[row:row + 1] else: d_range = range(num_of_components)[start_component:end_component] if need_PIL: PIL_list = [] for d in d_range: txt = Image.new("RGB", (48, 48), (255, 255, 255)) draw = ImageDraw.Draw(txt) draw.text((0, 20), "i = " + str(d), fill=(0, 0, 0)) txt = txt.resize((resolution, resolution)) txt = to_tensor_transoform(txt).to(device).unsqueeze(0) imgs = [txt] direction = degree * components[:, d].T if col: imgs = [] i_range = range(num_of_components)[col:col + 1] else: imgs = [txt] i_range = range(num_of_columns) if no_index: imgs = [] for i in i_range: if w_plus: target_latent = torch.unsqueeze(latent, 0).clone() target_latent[0] = target_latent[0] + alpha[i] * direction else: target_latent = latent + alpha[i] * direction img, _ = g( [target_latent], input_is_latent=True, noise=noise ) img = resize_transoform(img) imgs += [img] if need_PIL: PIL_gird = utils.make_grid( img[0, :, :, :], pad_value=1, normalize=True, range=(-1, 1), nrow=1, ) PIL_img = to_pil_transoform(PIL_gird) PIL_list.append(PIL_img) final_image = torch.cat(imgs).unsqueeze(0) final_image = torch.transpose(final_image, 0, 1) directional_results += [final_image] if col: nrow = 1 else: nrow = num_of_columns + 1 final_image = torch.cat(directional_results, 0) final_image = final_image.reshape(final_image.shape[0] * final_image.shape[1], final_image.shape[2], final_image.shape[3], final_image.shape[4]) grid = utils.make_grid( final_image, pad_value=1, normalize=True, range=(-1, 1), nrow=nrow, ) if need_PIL: pil_result = PIL_list else: pil_result = None return to_pil_transoform(grid), pil_result if __name__ == "__main__": torch.manual_seed(1) np.random.seed(1) random.seed(1) torch.set_grad_enabled(False) parser = argparse.ArgumentParser(description="Apply closed form factorization") parser.add_argument( "-d", "--degree", type=float, default=4, help="scalar factors for moving latent vectors along eigenvector", ) parser.add_argument( "--channel_multiplier", type=int, default=2, help='channel multiplier factor. config-f = 2, else = 1', ) parser.add_argument( "--latent", type=int, default=512, help="demension of the latent", ) parser.add_argument( "--n_mlp", type=int, default=8, help="n_mlp", ) parser.add_argument( "--max_channel_size", type=int, default=512, help="max channel size", ) parser.add_argument("--ckpt", type=str, required=True, help="stylegan2 checkpoints") parser.add_argument( "--size", type=int, default=256, help="output image size of the generator" ) parser.add_argument( "--truncation", type=float, default=0.7, help="truncation factor" ) parser.add_argument( "--device", type=str, default="cuda", help="device to run the model" ) parser.add_argument('--full_model', default=False, action='store_true') parser.add_argument("--initial_latent", type=str, required=False, default=None) parser.add_argument("--resolution", type=int, default=64, help="resolution") parser.add_argument("--start_component", type=int, default=0, help="start_component") parser.add_argument("--end_component", type=int, default=None, help="end_component") parser.add_argument("--num_of_columns", type=int, default=5, help="num_of_columns") parser.add_argument("--col", type=int, default=None, help="column") parser.add_argument("--row", type=int, default=None, help="row") parser.add_argument('--no_index', default=False, action='store_true') parser.add_argument("--random_seed", type=int, default=None, help="random seed") parser.add_argument("--factor", type=str, default=None, required=True, help="factor") parser.add_argument('--gif', default=False, action='store_true') parser.add_argument('--prename', default=None, type=str) args = parser.parse_args() directions = torch.load(args.factor)["eigvec"].to(args.device) grid, pil_result = ica_single_img( directions, args.ckpt, degree=args.degree, channel_multiplier=args.channel_multiplier, latent=args.latent, n_mlp=args.n_mlp, max_channel_size=args.max_channel_size, size=args.size, truncation=args.truncation, full_model=args.full_model, initial_latent=args.initial_latent, resolution=args.resolution, start_component=args.start_component, end_component=args.end_component, num_of_columns=args.num_of_columns, col=args.col, row=args.row, no_index=args.no_index, seed=args.random_seed, need_PIL=args.gif, ) if args.prename: if args.random_seed: nm = args.prename + str(args.row) + str(args.random_seed) else: nm = args.prename + str(args.row) else: nm = str(args.row) grid.save(nm + ".png") pil_result = pil_result + list(reversed(pil_result)) pil_result[0].save(nm + '.gif', save_all=True, append_images=pil_result[1:], duration=100, loop=0)
[ "torch.manual_seed", "torchvision.transforms.ToPILImage", "torch.set_grad_enabled", "argparse.ArgumentParser", "PIL.Image.new", "torch.load", "torch.unsqueeze", "random.seed", "torch.transpose", "PIL.ImageDraw.Draw", "model.Generator", "numpy.random.seed", "torchvision.transforms.Resize", ...
[((895, 911), 'torch.load', 'torch.load', (['ckpt'], {}), '(ckpt)\n', (905, 911), False, 'import torch\n'), ((1988, 2017), 'torchvision.transforms.Resize', 'transforms.Resize', (['resolution'], {}), '(resolution)\n', (2005, 2017), False, 'from torchvision import transforms\n'), ((2042, 2065), 'torchvision.transforms.ToPILImage', 'transforms.ToPILImage', ([], {}), '()\n', (2063, 2065), False, 'from torchvision import transforms\n'), ((2093, 2114), 'torchvision.transforms.ToTensor', 'transforms.ToTensor', ([], {}), '()\n', (2112, 2114), False, 'from torchvision import transforms\n'), ((4001, 4034), 'torch.cat', 'torch.cat', (['directional_results', '(0)'], {}), '(directional_results, 0)\n', (4010, 4034), False, 'import torch\n'), ((4235, 4322), 'torchvision.utils.make_grid', 'utils.make_grid', (['final_image'], {'pad_value': '(1)', 'normalize': '(True)', 'range': '(-1, 1)', 'nrow': 'nrow'}), '(final_image, pad_value=1, normalize=True, range=(-1, 1),\n nrow=nrow)\n', (4250, 4322), False, 'from torchvision import utils\n'), ((4532, 4552), 'torch.manual_seed', 'torch.manual_seed', (['(1)'], {}), '(1)\n', (4549, 4552), False, 'import torch\n'), ((4557, 4574), 'numpy.random.seed', 'np.random.seed', (['(1)'], {}), '(1)\n', (4571, 4574), True, 'import numpy as np\n'), ((4579, 4593), 'random.seed', 'random.seed', (['(1)'], {}), '(1)\n', (4590, 4593), False, 'import random\n'), ((4598, 4627), 'torch.set_grad_enabled', 'torch.set_grad_enabled', (['(False)'], {}), '(False)\n', (4620, 4627), False, 'import torch\n'), ((4641, 4711), 'argparse.ArgumentParser', 'argparse.ArgumentParser', ([], {'description': '"""Apply closed form factorization"""'}), "(description='Apply closed form factorization')\n", (4664, 4711), False, 'import argparse\n'), ((1360, 1383), 'torch.manual_seed', 'torch.manual_seed', (['seed'], {}), '(seed)\n', (1377, 1383), False, 'import torch\n'), ((1392, 1412), 'numpy.random.seed', 'np.random.seed', (['seed'], {}), '(seed)\n', (1406, 1412), True, 'import numpy as np\n'), ((1421, 1438), 'random.seed', 'random.seed', (['seed'], {}), '(seed)\n', (1432, 1438), False, 'import random\n'), ((1577, 1603), 'torch.load', 'torch.load', (['initial_latent'], {}), '(initial_latent)\n', (1587, 1603), False, 'import torch\n'), ((2376, 2419), 'PIL.Image.new', 'Image.new', (['"""RGB"""', '(48, 48)', '(255, 255, 255)'], {}), "('RGB', (48, 48), (255, 255, 255))\n", (2385, 2419), False, 'from PIL import Image, ImageDraw, ImageFont\n'), ((2435, 2454), 'PIL.ImageDraw.Draw', 'ImageDraw.Draw', (['txt'], {}), '(txt)\n', (2449, 2454), False, 'from PIL import Image, ImageDraw, ImageFont\n'), ((3828, 3862), 'torch.transpose', 'torch.transpose', (['final_image', '(0)', '(1)'], {}), '(final_image, 0, 1)\n', (3843, 3862), False, 'import torch\n'), ((1055, 1163), 'model.Generator', 'Generator', (['size', 'latent', 'n_mlp'], {'channel_multiplier': 'channel_multiplier', 'max_channel_size': 'max_channel_size'}), '(size, latent, n_mlp, channel_multiplier=channel_multiplier,\n max_channel_size=max_channel_size)\n', (1064, 1163), False, 'from model import Generator\n'), ((3455, 3543), 'torchvision.utils.make_grid', 'utils.make_grid', (['img[0, :, :, :]'], {'pad_value': '(1)', 'normalize': '(True)', 'range': '(-1, 1)', 'nrow': '(1)'}), '(img[0, :, :, :], pad_value=1, normalize=True, range=(-1, 1),\n nrow=1)\n', (3470, 3543), False, 'from torchvision import utils\n'), ((3777, 3792), 'torch.cat', 'torch.cat', (['imgs'], {}), '(imgs)\n', (3786, 3792), False, 'import torch\n'), ((6924, 6947), 'torch.load', 'torch.load', (['args.factor'], {}), '(args.factor)\n', (6934, 6947), False, 'import torch\n'), ((3009, 3035), 'torch.unsqueeze', 'torch.unsqueeze', (['latent', '(0)'], {}), '(latent, 0)\n', (3024, 3035), False, 'import torch\n')]
"""This module models the water level graphs to an appropiate degree polynomial. """ import matplotlib import numpy as np def polyfit(dates, levels, p): # convert dates to nummbers datenums = matplotlib.dates.date2num(dates) # Find coefficients of best-fit polynomial f(x) of degree p p_coeff = np.polyfit(datenums - datenums[0], levels, p) # Convert coefficient into a polynomial that can be evaluated, poly = np.poly1d(p_coeff) return poly, datenums[0] def floodwarning(station,dates,levels,p): datenums = matplotlib.dates.date2num(dates) p_coeff = np.polyfit(datenums - datenums[0], levels, p) poly = np.poly1d(p_coeff) polyd1 = np.polyder(poly) high = station.typical_range[1] low = station.typical_range[0] latest_level = levels[0] d1_latest = polyd1(0) # print("Station,Latest,Highest,d1,d2,:", station.name,latest_level,highest_level,d1_latest,d2_latest,) # default level as to not araise either too much worry or too little. fw = "moderate" # above highest level and increasing => severe if (d1_latest > 0 and latest_level > high): fw = "severe" # above highest level and decreasing => high if (d1_latest < 0 and latest_level > high): fw = "high" # near highest level and increasing => high if (d1_latest > 0 and (abs(latest_level - high) > abs(latest_level - low))): fw = "high" # near highest level and decreasing => moderate if (d1_latest < 0 and (abs(latest_level - high) > abs(latest_level - low))): fw = "moderate" # near lowest level and increasing => moderate if (d1_latest > 0 and (abs(latest_level - high) < abs(latest_level - low))): fw = "moderate" # near lowest level and decreasing => low if (d1_latest < 0 and (abs(latest_level - high) < abs(latest_level - low))): fw = "low" # below lowest level and decreasing => low if (d1_latest < 0 and latest_level < low): fw = "low" return fw
[ "matplotlib.dates.date2num", "numpy.polyder", "numpy.poly1d", "numpy.polyfit" ]
[((204, 236), 'matplotlib.dates.date2num', 'matplotlib.dates.date2num', (['dates'], {}), '(dates)\n', (229, 236), False, 'import matplotlib\n'), ((316, 361), 'numpy.polyfit', 'np.polyfit', (['(datenums - datenums[0])', 'levels', 'p'], {}), '(datenums - datenums[0], levels, p)\n', (326, 361), True, 'import numpy as np\n'), ((441, 459), 'numpy.poly1d', 'np.poly1d', (['p_coeff'], {}), '(p_coeff)\n', (450, 459), True, 'import numpy as np\n'), ((548, 580), 'matplotlib.dates.date2num', 'matplotlib.dates.date2num', (['dates'], {}), '(dates)\n', (573, 580), False, 'import matplotlib\n'), ((595, 640), 'numpy.polyfit', 'np.polyfit', (['(datenums - datenums[0])', 'levels', 'p'], {}), '(datenums - datenums[0], levels, p)\n', (605, 640), True, 'import numpy as np\n'), ((652, 670), 'numpy.poly1d', 'np.poly1d', (['p_coeff'], {}), '(p_coeff)\n', (661, 670), True, 'import numpy as np\n'), ((684, 700), 'numpy.polyder', 'np.polyder', (['poly'], {}), '(poly)\n', (694, 700), True, 'import numpy as np\n')]
import argparse import csv from collections import defaultdict import numpy as np import sys def main(args): missing_info_scores = defaultdict(list) usefulness_scores = defaultdict(list) missing_info_score_histogram = defaultdict(int) usefulness_score_histogram = defaultdict(int) with open(args.mturk_results_file) as csv_results_file: csv_reader = csv.DictReader(csv_results_file) for row in csv_reader: hit_id = row['HITId'] for i in range(1, 6): if row['Answer.missing_info_%d.%d' %(i, i)] == 'true': # import pdb; pdb.set_trace() missing_info_scores[hit_id].append(i) missing_info_score_histogram[i] += 1 for i in range(6): if row['Answer.usefulness_%d.%d' % (i, i)] == 'true': usefulness_scores[hit_id].append(i) usefulness_score_histogram[i] += 1 missing_info_score_var = 0 usefulness_score_var = 0 missing_info_agreement = 0 usefulness_agreement = 0 for hit_id in usefulness_scores: missing_info_score_var += np.var(missing_info_scores[hit_id]) usefulness_score_var += np.var(usefulness_scores[hit_id]) for i in range(1, 6): if missing_info_scores[hit_id].count(i) > 2: missing_info_agreement += 1 break for i in range(6): if usefulness_scores[hit_id].count(i) > 1: # if usefulness_scores[hit_id].count(i) > 2 or (usefulness_scores[hit_id].count(i) + usefulness_scores[hit_id].count(i-1)) > 2 or (usefulness_scores[hit_id].count(i) + usefulness_scores[hit_id].count(i+1)) > 2: usefulness_agreement += 1 break print('Missing info score variance %.2f' % (missing_info_score_var*1./len(missing_info_scores))) print('Usefulness score variance %.2f' % (usefulness_score_var*1./len(usefulness_scores))) print('Missing info agreement: %d out of %d' % (missing_info_agreement, len(missing_info_scores))) print('Usefulness agreement: %d out of %d' % (usefulness_agreement, len(usefulness_scores))) print('Missing info score histogram') print(missing_info_score_histogram) print('Usefulness score histogram') print(usefulness_score_histogram) if __name__ == "__main__": argparser = argparse.ArgumentParser(sys.argv[0]) argparser.add_argument("--mturk_results_file", type = str) args = argparser.parse_args() print(args) main(args)
[ "numpy.var", "csv.DictReader", "collections.defaultdict", "argparse.ArgumentParser" ]
[((136, 153), 'collections.defaultdict', 'defaultdict', (['list'], {}), '(list)\n', (147, 153), False, 'from collections import defaultdict\n'), ((178, 195), 'collections.defaultdict', 'defaultdict', (['list'], {}), '(list)\n', (189, 195), False, 'from collections import defaultdict\n'), ((231, 247), 'collections.defaultdict', 'defaultdict', (['int'], {}), '(int)\n', (242, 247), False, 'from collections import defaultdict\n'), ((281, 297), 'collections.defaultdict', 'defaultdict', (['int'], {}), '(int)\n', (292, 297), False, 'from collections import defaultdict\n'), ((2379, 2415), 'argparse.ArgumentParser', 'argparse.ArgumentParser', (['sys.argv[0]'], {}), '(sys.argv[0])\n', (2402, 2415), False, 'import argparse\n'), ((380, 412), 'csv.DictReader', 'csv.DictReader', (['csv_results_file'], {}), '(csv_results_file)\n', (394, 412), False, 'import csv\n'), ((1152, 1187), 'numpy.var', 'np.var', (['missing_info_scores[hit_id]'], {}), '(missing_info_scores[hit_id])\n', (1158, 1187), True, 'import numpy as np\n'), ((1220, 1253), 'numpy.var', 'np.var', (['usefulness_scores[hit_id]'], {}), '(usefulness_scores[hit_id])\n', (1226, 1253), True, 'import numpy as np\n')]
#!/usr/bin/env python import argparse from glob import iglob as glob from functools import partial from itertools import starmap import h5py import pandas as pd import numpy as np from numba import jit from scipy import linalg from dautil.util import map_parallel IDX = pd.IndexSlice _REAL_TO_SIM = 1e-12 __version__ = '0.2' def get_inputs(df, df_theory, df_filter, f_modecoupling, l_common, spectrum, sub_split, null_split, compute_nl=False): '''``spectrum``: one of TT, EE, BB, TE, TE, EB ''' df_Cl_sim = df.loc[IDX['Cl', spectrum, sub_split, null_split, :], l_common] df_Cl_sim.reset_index(level=(0, 1, 2, 3), inplace=True, drop=True) if compute_nl: df_Nl_sim = df.loc[IDX['Nl', spectrum, sub_split, null_split, :], l_common] df_Nl_sim.reset_index(level=(0, 1, 2, 3), inplace=True, drop=True) df_Cl_sim += 1.j * df_Nl_sim del df_Nl_sim M = f_modecoupling['{0}{0}'.format(spectrum)][:][l_common][:, l_common] # auto-spectra if spectrum[0] == spectrum[1]: F = df_filter.loc[IDX['{0}{0}'.format(spectrum), sub_split, null_split], l_common].values.real # cross-spectra else: F = df_filter.loc[IDX['{0}{0}{0}{0}'.format(spectrum[0]), sub_split, null_split], l_common].values.real F *= df_filter.loc[IDX['{0}{0}{0}{0}'.format(spectrum[1]), sub_split, null_split], l_common].values.real F = np.sqrt(F) Cl_th = df_theory.loc[l_common, spectrum].values if spectrum in ('TT', 'EE', 'BB', 'TE') else None return M, np.ascontiguousarray(F), df_Cl_sim, Cl_th def matrix_reduce(M, b): '''reduce the resolution of the matrix M by bin-width ``b`` and devided by ``b`` ''' if b == 1: return M else: m, n = M.shape return np.einsum('ijkl->ik', M.reshape(m // b, b, n // b, b)) / b def matrix_reduce_row(M, b): '''reduce the resolution of the matrix M by bin-width ``b`` and devided by ``b`` ''' if b == 1: return M else: m, n = M.shape return np.einsum('ijk->ik', M.reshape(m // b, b, n)) / b def spectra_reduce(Cls, b): '''reduce the resolution across the second axis by bin-width ``b`` and devided by ``b`` ''' if b == 1: return Cls else: m, n = Cls.shape return np.einsum('ijk->ij', Cls.reshape(m, n // b, b)) / b @jit(nopython=True, nogil=True) def solve_K_l(M, F, B_2): return F.reshape(-1, 1) * M * B_2.reshape(1, -1) def solve(K_l, Cl_sim, bin_width, return_w=False): K_b = matrix_reduce(K_l, bin_width) Cl_sim_binned = spectra_reduce(Cl_sim, bin_width) Cl = linalg.solve(K_b, Cl_sim_binned.T).T if return_w: P_bl_K_ll = matrix_reduce_row(K_l, bin_width) w_bl = linalg.solve(K_b, P_bl_K_ll) return Cl, w_bl else: return Cl def solve_spectra(df_pseudo, df_theory, df_filter, f_modecoupling, B_2, bin_width, l_common, l_common_binned, spectrum, sub_split, null_split, compute_nl=False, return_w=False): M, F, df_Cl_sim, Cl_th = get_inputs(df_pseudo, df_theory, df_filter, f_modecoupling, l_common, spectrum, sub_split, null_split, compute_nl=compute_nl) K_l = solve_K_l(M, F, B_2) del M, F if Cl_th is not None: df_Cl_sim.loc[IDX['theory', 0], :] = K_l @ Cl_th del Cl_th res = solve(K_l, df_Cl_sim.values, bin_width, return_w=return_w) del K_l if return_w: Cl, w_bl = res else: Cl = res df = pd.DataFrame( Cl, index=df_Cl_sim.index, columns=l_common_binned ) if return_w: return df, w_bl else: return df def main(pseudospectra, theory, filter_transfer, modecoupling, beam, bin_width, l_min, l_max, processes, compute_nl=False, return_w=False, l_boundary=600, l_lower=50, l_upper=4300): ''' `l_boundary`: the pivot point of l bins. e.g. given a bin_width, 600 is always the boundary between 2 bins. `l_lower`: lowest l we trust. e.g. 50 `l_upper`: highest l we trust. e.g. 4300 due to F_TT can becomes negative above that. ''' if l_min < 0: l_min = l_boundary - (l_boundary - l_lower) // bin_width * bin_width print(f'auto l-min at {l_min}') if l_max < 0: l_max = l_boundary + (l_upper - l_boundary) // bin_width * bin_width print(f'auto l-max at {l_max}') l_common = np.arange(l_min, l_max) l_common_binned = l_common.reshape(-1, bin_width).mean(axis=1) # Reading df_beam = pd.read_hdf(beam)['all'] B_2 = np.square(np.interp(l_common, df_beam.index, df_beam.values.real)) del df_beam df = pd.concat( map_parallel(pd.read_hdf, glob(pseudospectra), mode='multithreading', processes=processes) ) df_theory = pd.read_hdf(theory) * _REAL_TO_SIM df_filter = pd.read_hdf(filter_transfer) # mapping all cases index = pd.MultiIndex.from_product( ( df.index.levels[1], df.index.levels[2], df.index.levels[3] ), names=df.index.names[1:4] ) if return_w: index_w = pd.MultiIndex.from_product( ( df.index.levels[1], df.index.levels[2], df.index.levels[3], l_common_binned ), names=df.index.names[1:4] + ['b'] ) with h5py.File(modecoupling, 'r') as f: res = list(starmap( partial(solve_spectra, df, df_theory, df_filter, f, B_2, bin_width, l_common, l_common_binned, compute_nl=compute_nl, return_w=return_w), index )) if return_w: Cls, w_bls = list(map(list, zip(*res))) else: Cls = res df_spectra = pd.concat( Cls, keys=index ) del l_common_binned, B_2, df, df_theory, df_filter df_spectra.index.names = index.names + df_spectra.index.names[-2:] del index df_spectra.sort_index(inplace=True) if return_w: df_window = pd.DataFrame( np.concatenate(w_bls, axis=0), index=index_w, columns=l_common ) df_window.sort_index(inplace=True) return df_spectra, df_window else: return df_spectra def cli(): parser = argparse.ArgumentParser(description="Obtain final spectra from pseudo-spectra, mode-coupling, filter transfer function, etc.") parser.add_argument('-o', '--output', required=True, help='Output HDF5 filename.') parser.add_argument('--modecoupling', required=True, help='Input modecoupling HDF5 file.') parser.add_argument('--pseudospectra', required=True, help='Input pseudospectra DataFrame in HDF5. Can be a glob pattern.') parser.add_argument('--theory', required=True, help='Input theory DataFrame in HDF5.') parser.add_argument('--beam', required=True, help='Input beam DataFrame in HDF5.') parser.add_argument('--filter-transfer', required=True, help='Input filter transfer function in HDF5.') parser.add_argument('--compute-nl', action='store_true', help='If specified, compute Nl and store as the imaginary part of the spectra DataFrame.') parser.add_argument('--return-w', action='store_true', help='If specified, return the band-power window function w.') parser.add_argument('--bin-width', default=300, type=int, help='bin width. Default: 300') parser.add_argument('--l-min', type=int, default=-1, help='Minimum l. Default: auto-calculated. Lowest value: 2.') parser.add_argument('--l-max', type=int, default=-1, help='maximum l (exclusive). Default: auto-calculated. Highest value: 4901') parser.add_argument('-c', '--compress-level', default=9, type=int, help='compress level of gzip algorithm. Default: 9.') parser.add_argument('-v', '--version', action='version', version='%(prog)s {}'.format(__version__)) parser.add_argument('-p', '--processes', type=int, default=1, help='No. of parallel processes.') args = parser.parse_args() res = main( args.pseudospectra, args.theory, args.filter_transfer, args.modecoupling, args.beam, args.bin_width, args.l_min, args.l_max, args.processes, compute_nl=args.compute_nl, return_w=args.return_w ) if args.return_w: df_spectra, df_window = res else: df_spectra = res df_spectra.to_hdf( args.output, 'spectra', mode='w', format='table', complevel=args.compress_level, fletcher32=True, ) if args.return_w: df_window.to_hdf( args.output, 'window', mode='a', format='table', complevel=args.compress_level, fletcher32=True, ) if __name__ == "__main__": cli()
[ "pandas.MultiIndex.from_product", "numpy.sqrt", "argparse.ArgumentParser", "numpy.arange", "glob.iglob", "scipy.linalg.solve", "numpy.ascontiguousarray", "h5py.File", "numba.jit", "functools.partial", "numpy.interp", "numpy.concatenate", "pandas.DataFrame", "pandas.concat", "pandas.read_...
[((2369, 2399), 'numba.jit', 'jit', ([], {'nopython': '(True)', 'nogil': '(True)'}), '(nopython=True, nogil=True)\n', (2372, 2399), False, 'from numba import jit\n'), ((3486, 3550), 'pandas.DataFrame', 'pd.DataFrame', (['Cl'], {'index': 'df_Cl_sim.index', 'columns': 'l_common_binned'}), '(Cl, index=df_Cl_sim.index, columns=l_common_binned)\n', (3498, 3550), True, 'import pandas as pd\n'), ((4381, 4404), 'numpy.arange', 'np.arange', (['l_min', 'l_max'], {}), '(l_min, l_max)\n', (4390, 4404), True, 'import numpy as np\n'), ((4813, 4841), 'pandas.read_hdf', 'pd.read_hdf', (['filter_transfer'], {}), '(filter_transfer)\n', (4824, 4841), True, 'import pandas as pd\n'), ((4879, 4999), 'pandas.MultiIndex.from_product', 'pd.MultiIndex.from_product', (['(df.index.levels[1], df.index.levels[2], df.index.levels[3])'], {'names': 'df.index.names[1:4]'}), '((df.index.levels[1], df.index.levels[2], df.\n index.levels[3]), names=df.index.names[1:4])\n', (4905, 4999), True, 'import pandas as pd\n'), ((5715, 5741), 'pandas.concat', 'pd.concat', (['Cls'], {'keys': 'index'}), '(Cls, keys=index)\n', (5724, 5741), True, 'import pandas as pd\n'), ((6249, 6385), 'argparse.ArgumentParser', 'argparse.ArgumentParser', ([], {'description': '"""Obtain final spectra from pseudo-spectra, mode-coupling, filter transfer function, etc."""'}), "(description=\n 'Obtain final spectra from pseudo-spectra, mode-coupling, filter transfer function, etc.'\n )\n", (6272, 6385), False, 'import argparse\n'), ((1404, 1414), 'numpy.sqrt', 'np.sqrt', (['F'], {}), '(F)\n', (1411, 1414), True, 'import numpy as np\n'), ((1533, 1556), 'numpy.ascontiguousarray', 'np.ascontiguousarray', (['F'], {}), '(F)\n', (1553, 1556), True, 'import numpy as np\n'), ((2635, 2669), 'scipy.linalg.solve', 'linalg.solve', (['K_b', 'Cl_sim_binned.T'], {}), '(K_b, Cl_sim_binned.T)\n', (2647, 2669), False, 'from scipy import linalg\n'), ((2758, 2786), 'scipy.linalg.solve', 'linalg.solve', (['K_b', 'P_bl_K_ll'], {}), '(K_b, P_bl_K_ll)\n', (2770, 2786), False, 'from scipy import linalg\n'), ((4501, 4518), 'pandas.read_hdf', 'pd.read_hdf', (['beam'], {}), '(beam)\n', (4512, 4518), True, 'import pandas as pd\n'), ((4546, 4601), 'numpy.interp', 'np.interp', (['l_common', 'df_beam.index', 'df_beam.values.real'], {}), '(l_common, df_beam.index, df_beam.values.real)\n', (4555, 4601), True, 'import numpy as np\n'), ((4762, 4781), 'pandas.read_hdf', 'pd.read_hdf', (['theory'], {}), '(theory)\n', (4773, 4781), True, 'import pandas as pd\n'), ((5098, 5243), 'pandas.MultiIndex.from_product', 'pd.MultiIndex.from_product', (['(df.index.levels[1], df.index.levels[2], df.index.levels[3], l_common_binned)'], {'names': "(df.index.names[1:4] + ['b'])"}), "((df.index.levels[1], df.index.levels[2], df.\n index.levels[3], l_common_binned), names=df.index.names[1:4] + ['b'])\n", (5124, 5243), True, 'import pandas as pd\n'), ((5361, 5389), 'h5py.File', 'h5py.File', (['modecoupling', '"""r"""'], {}), "(modecoupling, 'r')\n", (5370, 5389), False, 'import h5py\n'), ((4674, 4693), 'glob.iglob', 'glob', (['pseudospectra'], {}), '(pseudospectra)\n', (4678, 4693), True, 'from glob import iglob as glob\n'), ((6010, 6039), 'numpy.concatenate', 'np.concatenate', (['w_bls'], {'axis': '(0)'}), '(w_bls, axis=0)\n', (6024, 6039), True, 'import numpy as np\n'), ((5436, 5576), 'functools.partial', 'partial', (['solve_spectra', 'df', 'df_theory', 'df_filter', 'f', 'B_2', 'bin_width', 'l_common', 'l_common_binned'], {'compute_nl': 'compute_nl', 'return_w': 'return_w'}), '(solve_spectra, df, df_theory, df_filter, f, B_2, bin_width,\n l_common, l_common_binned, compute_nl=compute_nl, return_w=return_w)\n', (5443, 5576), False, 'from functools import partial\n')]
import numpy as np import os,sys sys.path.insert(0,'scripts/') import shutil import kaldi_data as kd import argparse parser = argparse.ArgumentParser(description="Shuttling data", add_help=True) parser.add_argument("--source", type=str, default="data/test_segments/utt2lang_sorted",help="source data") parser.add_argument("--target", type=str, default="result_test_sorted.csv", help="target data") parser.add_argument("--utt2spk", action='store_true', help="for utt2spk file") parser.add_argument("--utt2lang", action='store_true', help="for utt2lang file") args = parser.parse_known_args()[0] SOURCE_FOLDER = args.source TARGET_FOLDER = args.target if not os.path.exists(TARGET_FOLDER): os.mkdir(TARGET_FOLDER) wavlist,utt_label,spk_label = kd.read_data_list(SOURCE_FOLDER, utt2spk=args.utt2spk, utt2lang=args.utt2lang) idx = range(len(wavlist)) np.random.shuffle(idx) wavlist = wavlist[idx] utt_label = utt_label[idx] spk_label = spk_label[idx] kd.write_data(TARGET_FOLDER,wavlist,utt_label,spk_label)
[ "os.path.exists", "sys.path.insert", "argparse.ArgumentParser", "kaldi_data.read_data_list", "os.mkdir", "kaldi_data.write_data", "numpy.random.shuffle" ]
[((33, 63), 'sys.path.insert', 'sys.path.insert', (['(0)', '"""scripts/"""'], {}), "(0, 'scripts/')\n", (48, 63), False, 'import os, sys\n'), ((127, 195), 'argparse.ArgumentParser', 'argparse.ArgumentParser', ([], {'description': '"""Shuttling data"""', 'add_help': '(True)'}), "(description='Shuttling data', add_help=True)\n", (150, 195), False, 'import argparse\n'), ((754, 832), 'kaldi_data.read_data_list', 'kd.read_data_list', (['SOURCE_FOLDER'], {'utt2spk': 'args.utt2spk', 'utt2lang': 'args.utt2lang'}), '(SOURCE_FOLDER, utt2spk=args.utt2spk, utt2lang=args.utt2lang)\n', (771, 832), True, 'import kaldi_data as kd\n'), ((860, 882), 'numpy.random.shuffle', 'np.random.shuffle', (['idx'], {}), '(idx)\n', (877, 882), True, 'import numpy as np\n'), ((961, 1020), 'kaldi_data.write_data', 'kd.write_data', (['TARGET_FOLDER', 'wavlist', 'utt_label', 'spk_label'], {}), '(TARGET_FOLDER, wavlist, utt_label, spk_label)\n', (974, 1020), True, 'import kaldi_data as kd\n'), ((660, 689), 'os.path.exists', 'os.path.exists', (['TARGET_FOLDER'], {}), '(TARGET_FOLDER)\n', (674, 689), False, 'import os, sys\n'), ((695, 718), 'os.mkdir', 'os.mkdir', (['TARGET_FOLDER'], {}), '(TARGET_FOLDER)\n', (703, 718), False, 'import os, sys\n')]
import numpy as np import pandas as pd import patsy FILE_PATH_CENSUS80_EXTRACT = "data/QOB.txt" FILE_PATH_FULL_CENSUS7080 = "data/NEW7080.dta" def get_df_census80(): cols = [0, 1, 3, 4, 5, 8, 9, 10, 11, 12, 15, 16, 17, 18, 19, 20, 23, 24, 26] cols_names = [ "AGE", "AGEQ", "EDUC", "ENOCENT", "ESOCENT", "LWKLYWGE", "MARRIED", "MIDATL", "MT", "NEWENG", "CENSUS", "STATE", "QOB", "RACE", "SMSA", "SOATL", "WNOCENT", "WSOCENT", "YOB", ] df = pd.read_csv(FILE_PATH_CENSUS80_EXTRACT, sep=" ", usecols=cols, names=cols_names) # correct AGEQ df.loc[df["CENSUS"] == 80, "AGEQ"] = df["AGEQ"] - 1900 return df def get_df_census70(): cols = [ "v1", "v2", "v4", "v5", "v6", "v9", "v10", "v11", "v12", "v13", "v16", "v17", "v18", "v19", "v20", "v21", "v24", "v25", "v27", ] cols_names = [ "AGE", "AGEQ", "EDUC", "ENOCENT", "ESOCENT", "LWKLYWGE", "MARRIED", "MIDATL", "MT", "NEWENG", "CENSUS", "STATE", "QOB", "RACE", "SMSA", "SOATL", "WNOCENT", "WSOCENT", "YOB", ] df = pd.read_stata(FILE_PATH_FULL_CENSUS7080, columns=cols) df = df.rename(columns=dict(zip(cols, cols_names))) return df.loc[df["CENSUS"] == 70] def get_df_census70_census_80(): cols = [ "v1", "v2", "v4", "v5", "v6", "v9", "v10", "v11", "v12", "v13", "v16", "v17", "v18", "v19", "v20", "v21", "v24", "v25", "v27", ] cols_names = [ "AGE", "AGEQ", "EDUC", "ENOCENT", "ESOCENT", "LWKLYWGE", "MARRIED", "MIDATL", "MT", "NEWENG", "CENSUS", "STATE", "QOB", "RACE", "SMSA", "SOATL", "WNOCENT", "WSOCENT", "YOB", ] df = pd.read_stata(FILE_PATH_FULL_CENSUS7080, columns=cols) df = df.rename(columns=dict(zip(cols, cols_names))) return df def prepare_census_data( df, const=True, qob=True, yob=True, age=True, state_of_birth=False, qob_x_yob=False, qob_x_state=False, ): if const: df = add_constant(df) if qob or qob_x_yob or qob_x_state: df = add_quarter_of_birth_dummies(df) if yob or qob_x_yob: df = add_year_of_birth_dummies(df) if age: df = add_age_squared(df) if state_of_birth or qob_x_state: df = add_state_of_birth_dummies(df) if qob_x_yob: df = add_qob_yob_interactions(df) if qob_x_state: df = add_qob_state_interactions(df, qob_x_state) return df def add_constant(df): df["CONST"] = 1 df["CONST"] = df["CONST"].astype(np.uint8) return df def get_constant_name(): return ["CONST"] def add_quarter_of_birth_dummies(df): return pd.concat((df, pd.get_dummies(df["QOB"], prefix="DUMMY_QOB")), axis=1) def get_quarter_of_birth_dummy_names(start=1, end=3): return [f"DUMMY_QOB_{j}" for j in range(start, end + 1)] def add_year_of_birth_dummies(df): return pd.concat((df, pd.get_dummies(df["YOB"] % 10, prefix="DUMMY_YOB")), axis=1) def get_year_of_birth_dummy_names(start=0, end=8): return [f"DUMMY_YOB_{i}" for i in range(start, end + 1)] def add_age_squared(df): df["AGESQ"] = df["AGEQ"].pow(2) return df def get_age_control_names(ageq=True, agesq=True): lst = [] if ageq: lst.append("AGEQ") if agesq: lst.append("AGESQ") return lst def add_state_of_birth_dummies(df): return pd.concat((df, pd.get_dummies(df["STATE"], prefix="DUMMY_STATE")), axis=1) def get_state_of_birth_dummy_names(state_list): return [f"DUMMY_STATE_{i}" for i in state_list] def get_state_list(df, rm_state=1): state_list = set(df["STATE"]) state_list.remove(rm_state) return state_list def add_qob_yob_interactions(df): interact_qob_yob = patsy.dmatrix( " + ".join(get_qob_yob_interaction_names()), df, return_type="dataframe" ) interact_qob_yob.drop("Intercept", axis=1, inplace=True) return pd.concat((df, interact_qob_yob.astype(np.uint8)), axis=1) def get_qob_yob_interaction_names(qob_start=1, qob_end=3, yob_start=0, yob_end=9): return [ f"DUMMY_YOB_{i}:DUMMY_QOB_{j}" for j in range(qob_start, qob_end + 1) for i in range(yob_start, yob_end + 1) ] def add_qob_state_interactions(df, state_list): interact_qob_state = patsy.dmatrix( " + ".join(get_qob_state_of_birth_interaction_names(state_list)), df, return_type="dataframe", ) interact_qob_state.drop("Intercept", axis=1, inplace=True) return pd.concat((df, interact_qob_state.astype(np.uint8)), axis=1) def get_qob_state_of_birth_interaction_names(state_list): return [f"DUMMY_STATE_{i}:DUMMY_QOB_{j}" for j in range(1, 4) for i in state_list] def get_further_exogenous_regressors(race=True, smsa=True, married=True): lst = [] if race: lst.append("RACE") if smsa: lst.append("SMSA") if married: lst.append("MARRIED") return lst def get_region_of_residence_dummies(): return ["NEWENG", "MIDATL", "ENOCENT", "WNOCENT", "SOATL", "ESOCENT", "WSOCENT", "MT"] def get_education_name(): return ["EDUC"] def get_log_weekly_wage_name(): return ["LWKLYWGE"] def add_education_dummies(df): # dummy variable high school degree (12 or more years of education) df["DUMMY_HIGH_SCHOOL"] = [1 if x >= 12 else 0 for x in df["EDUC"]] # dummy variable college degree (16 or more years of education) df["DUMMY_COLLEGE"] = [1 if x >= 16 else 0 for x in df["EDUC"]] # dummy variable master's degree (18 or more years of education) df["DUMMY_MASTER"] = [1 if x >= 18 else 0 for x in df["EDUC"]] # dummy variable doctoral degree (20 or more years of education) df["DUMMY_DOCTOR"] = [1 if x >= 20 else 0 for x in df["EDUC"]] return df def add_detrended_educational_variables(df, educ_vars=("EDUC")): for ev in educ_vars: mean_ev = df.groupby(["YOB", "QOB"])[ev].mean().to_frame() mean_ev["MV_AVG"] = two_sided_moving_average(mean_ev.values) for yob in set(df["YOB"]): for qob in set(df["QOB"]): df.loc[(df["YOB"] == yob) & (df["QOB"] == qob), f"MV_AVG_{ev}"] = mean_ev.loc[ (yob, qob), "MV_AVG" ] df[f"DTRND_{ev}"] = df[ev] - df[f"MV_AVG_{ev}"] return df def two_sided_moving_average(x): ma = np.full_like(x, np.nan) for i in range(2, len(x) - 2): ma[i] = (x[i - 2] + x[i - 1] + x[i + 1] + x[i + 2]) / 4 return ma
[ "pandas.get_dummies", "pandas.read_stata", "numpy.full_like", "pandas.read_csv" ]
[((613, 698), 'pandas.read_csv', 'pd.read_csv', (['FILE_PATH_CENSUS80_EXTRACT'], {'sep': '""" """', 'usecols': 'cols', 'names': 'cols_names'}), "(FILE_PATH_CENSUS80_EXTRACT, sep=' ', usecols=cols, names=cols_names\n )\n", (624, 698), True, 'import pandas as pd\n'), ((1474, 1528), 'pandas.read_stata', 'pd.read_stata', (['FILE_PATH_FULL_CENSUS7080'], {'columns': 'cols'}), '(FILE_PATH_FULL_CENSUS7080, columns=cols)\n', (1487, 1528), True, 'import pandas as pd\n'), ((2321, 2375), 'pandas.read_stata', 'pd.read_stata', (['FILE_PATH_FULL_CENSUS7080'], {'columns': 'cols'}), '(FILE_PATH_FULL_CENSUS7080, columns=cols)\n', (2334, 2375), True, 'import pandas as pd\n'), ((6997, 7020), 'numpy.full_like', 'np.full_like', (['x', 'np.nan'], {}), '(x, np.nan)\n', (7009, 7020), True, 'import numpy as np\n'), ((3311, 3356), 'pandas.get_dummies', 'pd.get_dummies', (["df['QOB']"], {'prefix': '"""DUMMY_QOB"""'}), "(df['QOB'], prefix='DUMMY_QOB')\n", (3325, 3356), True, 'import pandas as pd\n'), ((3547, 3597), 'pandas.get_dummies', 'pd.get_dummies', (["(df['YOB'] % 10)"], {'prefix': '"""DUMMY_YOB"""'}), "(df['YOB'] % 10, prefix='DUMMY_YOB')\n", (3561, 3597), True, 'import pandas as pd\n'), ((4028, 4077), 'pandas.get_dummies', 'pd.get_dummies', (["df['STATE']"], {'prefix': '"""DUMMY_STATE"""'}), "(df['STATE'], prefix='DUMMY_STATE')\n", (4042, 4077), True, 'import pandas as pd\n')]
# MIT License # # Copyright (C) The Adversarial Robustness Toolbox (ART) Authors 2020 # # Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated # documentation files (the "Software"), to deal in the Software without restriction, including without limitation the # rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit # persons to whom the Software is furnished to do so, subject to the following conditions: # # The above copyright notice and this permission notice shall be included in all copies or substantial portions of the # Software. # # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE # WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE # AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, # TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE # SOFTWARE. from __future__ import absolute_import, division, print_function, unicode_literals import logging import numpy as np import pytest from numpy.testing import assert_allclose, assert_array_equal from scipy.io.wavfile import read from art.estimators.speech_recognition.speech_recognizer import SpeechRecognizerMixin from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR from art.estimators.tensorflow import TensorFlowV2Estimator from art.utils import get_file from tests.utils import ARTTestException logger = logging.getLogger(__name__) class TestTensorFlowLingvoASR: """ Test the TensorFlowLingvoASR estimator. """ @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") def test_is_subclass(self, art_warning): try: assert issubclass(TensorFlowLingvoASR, (SpeechRecognizerMixin, TensorFlowV2Estimator)) except ARTTestException as e: art_warning(e) @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") def test_implements_abstract_methods(self, art_warning): try: import tensorflow.compat.v1 as tf1 TensorFlowLingvoASR() except ARTTestException as e: art_warning(e) @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") def test_load_model(self, art_warning): try: import tensorflow.compat.v1 as tf1 TensorFlowLingvoASR() graph = tf1.get_default_graph() assert graph.get_operations() except ARTTestException as e: art_warning(e) @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") def test_create_decoder_input(self, art_warning, audio_batch_padded): try: import tensorflow.compat.v1 as tf1 test_input, test_mask_frequency = audio_batch_padded test_target_dummy = np.array(["DUMMY"] * test_input.shape[0]) lingvo = TensorFlowLingvoASR() decoder_input_tf = lingvo._create_decoder_input(lingvo._x_padded, lingvo._y_target, lingvo._mask_frequency) decoder_input = lingvo._sess.run( decoder_input_tf, { lingvo._x_padded: test_input, lingvo._y_target: test_target_dummy, lingvo._mask_frequency: test_mask_frequency, }, ) assert set(decoder_input.keys()).issuperset({"src", "tgt", "sample_ids"}) assert set(decoder_input.src.keys()).issuperset({"src_inputs", "paddings"}) assert set(decoder_input.tgt.keys()).issuperset({"ids", "labels", "paddings", "weights"}) except ARTTestException as e: art_warning(e) @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") def test_create_log_mel_features(self, art_warning, audio_batch_padded): try: import tensorflow.compat.v1 as tf1 test_input, _ = audio_batch_padded lingvo = TensorFlowLingvoASR() features_tf = lingvo._create_log_mel_features(lingvo._x_padded) features = lingvo._sess.run(features_tf, {lingvo._x_padded: test_input}) assert features.shape[2] == 80 assert len(features.shape) == 4 except ARTTestException as e: art_warning(e) @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") def test_pad_audio_input(self, art_warning): try: import tensorflow.compat.v1 as tf1 test_input = np.array([np.array([1]), np.array([2] * 480)], dtype=object) test_mask = np.array([[True] + [False] * 479, [True] * 480]) test_output = np.array([[1] + [0] * 479, [2] * 480]) lingvo = TensorFlowLingvoASR() output, mask, mask_freq = lingvo._pad_audio_input(test_input) assert_array_equal(test_output, output) assert_array_equal(test_mask, mask) except ARTTestException as e: art_warning(e) @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") def test_predict_batch(self, art_warning, audio_batch_padded): try: import tensorflow.compat.v1 as tf1 test_input, test_mask_frequency = audio_batch_padded test_target_dummy = np.array(["DUMMY"] * test_input.shape[0]) lingvo = TensorFlowLingvoASR() feed_dict = { lingvo._x_padded: test_input, lingvo._y_target: test_target_dummy, lingvo._mask_frequency: test_mask_frequency, } predictions = lingvo._sess.run(lingvo._predict_batch_op, feed_dict) assert set(predictions.keys()).issuperset( { "target_ids", "target_labels", "target_weights", "target_paddings", "transcripts", "topk_decoded", "topk_ids", "topk_lens", "topk_scores", "norm_wer_errors", "norm_wer_words", "utt_id", } ) except ARTTestException as e: art_warning(e) @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") def test_predict(self, art_warning, audio_data): try: import tensorflow.compat.v1 as tf1 test_input = audio_data lingvo = TensorFlowLingvoASR() predictions = lingvo.predict(test_input, batch_size=2) assert predictions.shape == test_input.shape assert isinstance(predictions[0], np.str_) except ARTTestException as e: art_warning(e) @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") def test_loss_gradient_tensor(self, art_warning, audio_batch_padded): try: import tensorflow.compat.v1 as tf1 test_input, test_mask_frequency = audio_batch_padded test_target_dummy = np.array(["DUMMY"] * test_input.shape[0]) lingvo = TensorFlowLingvoASR() feed_dict = { lingvo._x_padded: test_input, lingvo._y_target: test_target_dummy, lingvo._mask_frequency: test_mask_frequency, } loss_gradient = lingvo._sess.run(lingvo._loss_gradient_op, feed_dict) assert test_input.shape == loss_gradient.shape assert loss_gradient.sum() == 0.0 except ARTTestException as e: art_warning(e) @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") @pytest.mark.parametrize("batch_mode", [True, False]) def test_loss_gradient_batch_mode(self, art_warning, batch_mode, audio_data): try: import tensorflow.compat.v1 as tf1 test_input = audio_data test_target = np.array(["This", "is", "a dummy", "a dummy"]) lingvo = TensorFlowLingvoASR() if batch_mode: gradients = lingvo._loss_gradient_per_batch(test_input, test_target) else: gradients = lingvo._loss_gradient_per_sequence(test_input, test_target) gradients_abs_sum = np.array([np.abs(g).sum() for g in gradients], dtype=object) # test shape, equal inputs have equal gradients, non-zero inputs have non-zero gradient sums assert [x.shape for x in test_input] == [g.shape for g in gradients] assert_allclose(np.abs(gradients[2]).sum(), np.abs(gradients[3]).sum(), rtol=1e-01) assert_array_equal(gradients_abs_sum > 0, [False, True, True, True]) except ARTTestException as e: art_warning(e) class TestTensorFlowLingvoASRLibriSpeechSamples: # specify LibriSpeech samples for download and with transcriptions samples = { "3575-170457-0013.wav": { "uri": ( "https://github.com/tensorflow/cleverhans/blob/6ef939059172901db582c7702eb803b7171e3db5/" "examples/adversarial_asr/LibriSpeech/test-clean/3575/170457/3575-170457-0013.wav?raw=true" ), "transcript": ( "THE MORE SHE IS ENGAGED IN HER PROPER DUTIES THE LAST LEISURE WILL SHE HAVE FOR IT EVEN AS" " AN ACCOMPLISHMENT AND A RECREATION" ), }, "5105-28241-0006.wav": { "uri": ( "https://github.com/tensorflow/cleverhans/blob/6ef939059172901db582c7702eb803b7171e3db5/" "examples/adversarial_asr/LibriSpeech/test-clean/5105/28241/5105-28241-0006.wav?raw=true" ), "transcript": ( "THE LOG AND THE COMPASS THEREFORE WERE ABLE TO BE CALLED UPON TO DO THE WORK OF THE SEXTANT WHICH" " HAD BECOME UTTERLY USELESS" ), }, "2300-131720-0015.wav": { "uri": ( "https://github.com/tensorflow/cleverhans/blob/6ef939059172901db582c7702eb803b7171e3db5/" "examples/adversarial_asr/LibriSpeech/test-clean/2300/131720/2300-131720-0015.wav?raw=true" ), "transcript": ( "HE OBTAINED THE DESIRED SPEED AND LOAD WITH A FRICTION BRAKE ALSO REGULATOR OF SPEED BUT WAITED FOR AN" " INDICATOR TO VERIFY IT" ), }, } @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") def test_predict(self, art_warning): try: import tensorflow.compat.v1 as tf1 transcripts = list() audios = list() for filename, sample in self.samples.items(): file_path = get_file(filename, sample["uri"]) _, audio = read(file_path) audios.append(audio) transcripts.append(sample["transcript"]) audio_batch = np.array(audios, dtype=object) lingvo = TensorFlowLingvoASR() prediction = lingvo.predict(audio_batch, batch_size=1) assert prediction[0] == transcripts[0] except ARTTestException as e: art_warning(e) @pytest.mark.skip_module("lingvo") @pytest.mark.skip_framework("pytorch", "tensorflow1", "tensorflow2", "mxnet", "kerastf", "non_dl_frameworks") @pytest.mark.xfail(reason="Known issue that needs further investigation") def test_loss_gradient(self, art_warning): try: import tensorflow.compat.v1 as tf1 transcripts = list() audios = list() for filename, sample in self.samples.items(): file_path = get_file(filename, sample["uri"]) _, audio = read(file_path) audios.append(audio) transcripts.append(sample["transcript"]) audio_batch = np.array(audios, dtype=object) target_batch = np.array(transcripts) lingvo = TensorFlowLingvoASR() gradient_batch = lingvo._loss_gradient_per_batch(audio_batch, target_batch) gradient_sequence = lingvo._loss_gradient_per_sequence(audio_batch, target_batch) gradient_batch_sum = np.array([np.abs(gb).sum() for gb in gradient_batch], dtype=object) gradient_sequence_sum = np.array([np.abs(gs).sum() for gs in gradient_sequence], dtype=object) # test loss gradients per batch and per sequence are the same assert_allclose(gradient_sequence_sum, gradient_batch_sum, rtol=1e-05) # test gradient_batch, gradient_sequence and audios items have same shapes assert ( [gb.shape for gb in gradient_batch] == [gs.shape for gs in gradient_sequence] == [a.shape for a in audios] ) except ARTTestException as e: art_warning(e)
[ "logging.getLogger", "numpy.abs", "art.utils.get_file", "art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR", "pytest.mark.xfail", "numpy.testing.assert_allclose", "pytest.mark.skip_module", "pytest.mark.skip_framework", "pytest.mark.parametrize", "numpy.array", "scipy.io.wa...
[((1677, 1704), 'logging.getLogger', 'logging.getLogger', (['__name__'], {}), '(__name__)\n', (1694, 1704), False, 'import logging\n'), ((1804, 1837), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (1827, 1837), False, 'import pytest\n'), ((1843, 1955), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (1869, 1955), False, 'import pytest\n'), ((2180, 2213), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (2203, 2213), False, 'import pytest\n'), ((2219, 2331), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (2245, 2331), False, 'import pytest\n'), ((2555, 2588), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (2578, 2588), False, 'import pytest\n'), ((2594, 2706), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (2620, 2706), False, 'import pytest\n'), ((2999, 3032), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (3022, 3032), False, 'import pytest\n'), ((3038, 3150), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (3064, 3150), False, 'import pytest\n'), ((4236, 4269), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (4259, 4269), False, 'import pytest\n'), ((4275, 4387), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (4301, 4387), False, 'import pytest\n'), ((4932, 4965), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (4955, 4965), False, 'import pytest\n'), ((4971, 5083), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (4997, 5083), False, 'import pytest\n'), ((5703, 5736), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (5726, 5736), False, 'import pytest\n'), ((5742, 5854), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (5768, 5854), False, 'import pytest\n'), ((7044, 7077), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (7067, 7077), False, 'import pytest\n'), ((7083, 7195), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (7109, 7195), False, 'import pytest\n'), ((7636, 7669), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (7659, 7669), False, 'import pytest\n'), ((7675, 7787), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (7701, 7787), False, 'import pytest\n'), ((8560, 8593), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (8583, 8593), False, 'import pytest\n'), ((8599, 8711), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (8625, 8711), False, 'import pytest\n'), ((8713, 8765), 'pytest.mark.parametrize', 'pytest.mark.parametrize', (['"""batch_mode"""', '[True, False]'], {}), "('batch_mode', [True, False])\n", (8736, 8765), False, 'import pytest\n'), ((11452, 11485), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (11475, 11485), False, 'import pytest\n'), ((11491, 11603), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (11517, 11603), False, 'import pytest\n'), ((12311, 12344), 'pytest.mark.skip_module', 'pytest.mark.skip_module', (['"""lingvo"""'], {}), "('lingvo')\n", (12334, 12344), False, 'import pytest\n'), ((12350, 12462), 'pytest.mark.skip_framework', 'pytest.mark.skip_framework', (['"""pytorch"""', '"""tensorflow1"""', '"""tensorflow2"""', '"""mxnet"""', '"""kerastf"""', '"""non_dl_frameworks"""'], {}), "('pytorch', 'tensorflow1', 'tensorflow2', 'mxnet',\n 'kerastf', 'non_dl_frameworks')\n", (12376, 12462), False, 'import pytest\n'), ((12464, 12536), 'pytest.mark.xfail', 'pytest.mark.xfail', ([], {'reason': '"""Known issue that needs further investigation"""'}), "(reason='Known issue that needs further investigation')\n", (12481, 12536), False, 'import pytest\n'), ((2462, 2483), 'art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR', 'TensorFlowLingvoASR', ([], {}), '()\n', (2481, 2483), False, 'from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR\n'), ((2820, 2841), 'art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR', 'TensorFlowLingvoASR', ([], {}), '()\n', (2839, 2841), False, 'from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR\n'), ((2862, 2885), 'tensorflow.compat.v1.get_default_graph', 'tf1.get_default_graph', ([], {}), '()\n', (2883, 2885), True, 'import tensorflow.compat.v1 as tf1\n'), ((3379, 3420), 'numpy.array', 'np.array', (["(['DUMMY'] * test_input.shape[0])"], {}), "(['DUMMY'] * test_input.shape[0])\n", (3387, 3420), True, 'import numpy as np\n'), ((3443, 3464), 'art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR', 'TensorFlowLingvoASR', ([], {}), '()\n', (3462, 3464), False, 'from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR\n'), ((4590, 4611), 'art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR', 'TensorFlowLingvoASR', ([], {}), '()\n', (4609, 4611), False, 'from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR\n'), ((5300, 5348), 'numpy.array', 'np.array', (['[[True] + [False] * 479, [True] * 480]'], {}), '([[True] + [False] * 479, [True] * 480])\n', (5308, 5348), True, 'import numpy as np\n'), ((5375, 5413), 'numpy.array', 'np.array', (['[[1] + [0] * 479, [2] * 480]'], {}), '([[1] + [0] * 479, [2] * 480])\n', (5383, 5413), True, 'import numpy as np\n'), ((5436, 5457), 'art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR', 'TensorFlowLingvoASR', ([], {}), '()\n', (5455, 5457), False, 'from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR\n'), ((5544, 5583), 'numpy.testing.assert_array_equal', 'assert_array_equal', (['test_output', 'output'], {}), '(test_output, output)\n', (5562, 5583), False, 'from numpy.testing import assert_allclose, assert_array_equal\n'), ((5596, 5631), 'numpy.testing.assert_array_equal', 'assert_array_equal', (['test_mask', 'mask'], {}), '(test_mask, mask)\n', (5614, 5631), False, 'from numpy.testing import assert_allclose, assert_array_equal\n'), ((6076, 6117), 'numpy.array', 'np.array', (["(['DUMMY'] * test_input.shape[0])"], {}), "(['DUMMY'] * test_input.shape[0])\n", (6084, 6117), True, 'import numpy as np\n'), ((6140, 6161), 'art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR', 'TensorFlowLingvoASR', ([], {}), '()\n', (6159, 6161), False, 'from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR\n'), ((7364, 7385), 'art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR', 'TensorFlowLingvoASR', ([], {}), '()\n', (7383, 7385), False, 'from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR\n'), ((8016, 8057), 'numpy.array', 'np.array', (["(['DUMMY'] * test_input.shape[0])"], {}), "(['DUMMY'] * test_input.shape[0])\n", (8024, 8057), True, 'import numpy as np\n'), ((8080, 8101), 'art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR', 'TensorFlowLingvoASR', ([], {}), '()\n', (8099, 8101), False, 'from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR\n'), ((8971, 9017), 'numpy.array', 'np.array', (["['This', 'is', 'a dummy', 'a dummy']"], {}), "(['This', 'is', 'a dummy', 'a dummy'])\n", (8979, 9017), True, 'import numpy as np\n'), ((9040, 9061), 'art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR', 'TensorFlowLingvoASR', ([], {}), '()\n', (9059, 9061), False, 'from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR\n'), ((9669, 9737), 'numpy.testing.assert_array_equal', 'assert_array_equal', (['(gradients_abs_sum > 0)', '[False, True, True, True]'], {}), '(gradients_abs_sum > 0, [False, True, True, True])\n', (9687, 9737), False, 'from numpy.testing import assert_allclose, assert_array_equal\n'), ((12047, 12077), 'numpy.array', 'np.array', (['audios'], {'dtype': 'object'}), '(audios, dtype=object)\n', (12055, 12077), True, 'import numpy as np\n'), ((12100, 12121), 'art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR', 'TensorFlowLingvoASR', ([], {}), '()\n', (12119, 12121), False, 'from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR\n'), ((12990, 13020), 'numpy.array', 'np.array', (['audios'], {'dtype': 'object'}), '(audios, dtype=object)\n', (12998, 13020), True, 'import numpy as np\n'), ((13048, 13069), 'numpy.array', 'np.array', (['transcripts'], {}), '(transcripts)\n', (13056, 13069), True, 'import numpy as np\n'), ((13092, 13113), 'art.estimators.speech_recognition.tensorflow_lingvo.TensorFlowLingvoASR', 'TensorFlowLingvoASR', ([], {}), '()\n', (13111, 13113), False, 'from art.estimators.speech_recognition.tensorflow_lingvo import TensorFlowLingvoASR\n'), ((13592, 13662), 'numpy.testing.assert_allclose', 'assert_allclose', (['gradient_sequence_sum', 'gradient_batch_sum'], {'rtol': '(1e-05)'}), '(gradient_sequence_sum, gradient_batch_sum, rtol=1e-05)\n', (13607, 13662), False, 'from numpy.testing import assert_allclose, assert_array_equal\n'), ((11849, 11882), 'art.utils.get_file', 'get_file', (['filename', "sample['uri']"], {}), "(filename, sample['uri'])\n", (11857, 11882), False, 'from art.utils import get_file\n'), ((11910, 11925), 'scipy.io.wavfile.read', 'read', (['file_path'], {}), '(file_path)\n', (11914, 11925), False, 'from scipy.io.wavfile import read\n'), ((12792, 12825), 'art.utils.get_file', 'get_file', (['filename', "sample['uri']"], {}), "(filename, sample['uri'])\n", (12800, 12825), False, 'from art.utils import get_file\n'), ((12853, 12868), 'scipy.io.wavfile.read', 'read', (['file_path'], {}), '(file_path)\n', (12857, 12868), False, 'from scipy.io.wavfile import read\n'), ((5225, 5238), 'numpy.array', 'np.array', (['[1]'], {}), '([1])\n', (5233, 5238), True, 'import numpy as np\n'), ((5240, 5259), 'numpy.array', 'np.array', (['([2] * 480)'], {}), '([2] * 480)\n', (5248, 5259), True, 'import numpy as np\n'), ((9589, 9609), 'numpy.abs', 'np.abs', (['gradients[2]'], {}), '(gradients[2])\n', (9595, 9609), True, 'import numpy as np\n'), ((9617, 9637), 'numpy.abs', 'np.abs', (['gradients[3]'], {}), '(gradients[3])\n', (9623, 9637), True, 'import numpy as np\n'), ((9323, 9332), 'numpy.abs', 'np.abs', (['g'], {}), '(g)\n', (9329, 9332), True, 'import numpy as np\n'), ((13340, 13350), 'numpy.abs', 'np.abs', (['gb'], {}), '(gb)\n', (13346, 13350), True, 'import numpy as np\n'), ((13444, 13454), 'numpy.abs', 'np.abs', (['gs'], {}), '(gs)\n', (13450, 13454), True, 'import numpy as np\n')]
import os import imageio import numpy as np from metrics.tests_utils_cy import time_func from metrics.ms_ssim_m.ms_ssim import ms_ssim from metrics.ms_ssim_m.ms_ssim_cy import ms_ssim as ms_ssim_cy from metrics.ms_ssim_m.ms_ssim_tf import MS_SSIM_TF image = imageio.imread(os.path.join(__file__, '../..', 'corgi.jpg')).astype(np.int32) noisy = (image + np.random.uniform(-5, 5, image.shape)).astype(np.int32) obj = MS_SSIM_TF() image = np.broadcast_to(image, (4,) + image.shape) noisy = np.broadcast_to(noisy, (4,) + noisy.shape) time_func(ms_ssim, 'ms_ssim', image, noisy) time_func(ms_ssim_cy, 'ms_ssim_cy cython', image, noisy) time_func(obj.ms_ssim, 'ms_ssim_tf', image, noisy)
[ "os.path.join", "metrics.tests_utils_cy.time_func", "numpy.random.uniform", "numpy.broadcast_to", "metrics.ms_ssim_m.ms_ssim_tf.MS_SSIM_TF" ]
[((418, 430), 'metrics.ms_ssim_m.ms_ssim_tf.MS_SSIM_TF', 'MS_SSIM_TF', ([], {}), '()\n', (428, 430), False, 'from metrics.ms_ssim_m.ms_ssim_tf import MS_SSIM_TF\n'), ((439, 481), 'numpy.broadcast_to', 'np.broadcast_to', (['image', '((4,) + image.shape)'], {}), '(image, (4,) + image.shape)\n', (454, 481), True, 'import numpy as np\n'), ((490, 532), 'numpy.broadcast_to', 'np.broadcast_to', (['noisy', '((4,) + noisy.shape)'], {}), '(noisy, (4,) + noisy.shape)\n', (505, 532), True, 'import numpy as np\n'), ((534, 577), 'metrics.tests_utils_cy.time_func', 'time_func', (['ms_ssim', '"""ms_ssim"""', 'image', 'noisy'], {}), "(ms_ssim, 'ms_ssim', image, noisy)\n", (543, 577), False, 'from metrics.tests_utils_cy import time_func\n'), ((578, 634), 'metrics.tests_utils_cy.time_func', 'time_func', (['ms_ssim_cy', '"""ms_ssim_cy cython"""', 'image', 'noisy'], {}), "(ms_ssim_cy, 'ms_ssim_cy cython', image, noisy)\n", (587, 634), False, 'from metrics.tests_utils_cy import time_func\n'), ((635, 685), 'metrics.tests_utils_cy.time_func', 'time_func', (['obj.ms_ssim', '"""ms_ssim_tf"""', 'image', 'noisy'], {}), "(obj.ms_ssim, 'ms_ssim_tf', image, noisy)\n", (644, 685), False, 'from metrics.tests_utils_cy import time_func\n'), ((276, 320), 'os.path.join', 'os.path.join', (['__file__', '"""../.."""', '"""corgi.jpg"""'], {}), "(__file__, '../..', 'corgi.jpg')\n", (288, 320), False, 'import os\n'), ((356, 393), 'numpy.random.uniform', 'np.random.uniform', (['(-5)', '(5)', 'image.shape'], {}), '(-5, 5, image.shape)\n', (373, 393), True, 'import numpy as np\n')]
# Copyright 2019-2022 The <NAME> at the California Institute of # Technology (Caltech), with support from the Paul Allen Family Foundation, # Google, & National Institutes of Health (NIH) under Grant U24CA224309-01. # All rights reserved. # # Licensed under a modified Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.github.com/vanvalenlab/deepcell-spots/LICENSE # # The Work provided may be used for non-commercial academic purposes only. # For any other use of the Work, including commercial use, please contact: # <EMAIL> # # Neither the name of Caltech nor the names of its contributors may be used # to endorse or promote products derived from this software without specific # prior written permission. # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== """Tests for expectation maximization cluster visualization""" from itertools import combinations import numpy as np from scipy.spatial import distance from tensorflow.python.platform import test from deepcell_spots.cluster_vis import ca_to_adjacency_matrix, jitter class TestClusterVis(test.TestCase): def test_jitter(self): coords = np.zeros((10, 2)) size = 5 noisy_coords = jitter(coords, size) self.assertEqual(np.shape(coords), np.shape(noisy_coords)) self.assertNotEqual(coords.all(), noisy_coords.all()) def test_ca_to_adjacency_matrix(self): num_clusters = 10 num_annotators = 3 ca_matrix = np.ones((num_clusters, num_annotators)) A = ca_to_adjacency_matrix(ca_matrix) self.assertEqual(np.shape(A)[0], np.shape(A)[1], ca_matrix[0]) if __name__ == '__main__': test.main()
[ "numpy.shape", "numpy.ones", "deepcell_spots.cluster_vis.ca_to_adjacency_matrix", "numpy.zeros", "deepcell_spots.cluster_vis.jitter", "tensorflow.python.platform.test.main" ]
[((2101, 2112), 'tensorflow.python.platform.test.main', 'test.main', ([], {}), '()\n', (2110, 2112), False, 'from tensorflow.python.platform import test\n'), ((1585, 1602), 'numpy.zeros', 'np.zeros', (['(10, 2)'], {}), '((10, 2))\n', (1593, 1602), True, 'import numpy as np\n'), ((1643, 1663), 'deepcell_spots.cluster_vis.jitter', 'jitter', (['coords', 'size'], {}), '(coords, size)\n', (1649, 1663), False, 'from deepcell_spots.cluster_vis import ca_to_adjacency_matrix, jitter\n'), ((1910, 1949), 'numpy.ones', 'np.ones', (['(num_clusters, num_annotators)'], {}), '((num_clusters, num_annotators))\n', (1917, 1949), True, 'import numpy as np\n'), ((1962, 1995), 'deepcell_spots.cluster_vis.ca_to_adjacency_matrix', 'ca_to_adjacency_matrix', (['ca_matrix'], {}), '(ca_matrix)\n', (1984, 1995), False, 'from deepcell_spots.cluster_vis import ca_to_adjacency_matrix, jitter\n'), ((1689, 1705), 'numpy.shape', 'np.shape', (['coords'], {}), '(coords)\n', (1697, 1705), True, 'import numpy as np\n'), ((1707, 1729), 'numpy.shape', 'np.shape', (['noisy_coords'], {}), '(noisy_coords)\n', (1715, 1729), True, 'import numpy as np\n'), ((2022, 2033), 'numpy.shape', 'np.shape', (['A'], {}), '(A)\n', (2030, 2033), True, 'import numpy as np\n'), ((2038, 2049), 'numpy.shape', 'np.shape', (['A'], {}), '(A)\n', (2046, 2049), True, 'import numpy as np\n')]
import logging import numpy as np from ..Dataset import Dataset def crop(jets, pileup=False): #logging.warning("Cropping...") if pileup: logging.warning("pileup") pt_min, pt_max, m_min, m_max = 300, 365, 150, 220 else: pt_min, pt_max, m_min, m_max = 250, 300, 50, 110 good_jets = [] bad_jets = [] #good_indices = [] for i, j in enumerate(jets): if pt_min < j.pt < pt_max and m_min < j.mass < m_max: good_jets.append(j) else: bad_jets.append(j) # Weights for flatness in pt w = np.zeros(len(good_jets)) y_ = np.array([jet.y for jet in good_jets]) jets_0 = [jet for jet in good_jets if jet.y == 0] pdf, edges = np.histogram([j.pt for j in jets_0], density=True, range=[pt_min, pt_max], bins=50) pts = [j.pt for j in jets_0] indices = np.searchsorted(edges, pts) - 1 inv_w = 1. / pdf[indices] inv_w /= inv_w.sum() #w[y_==0] = inv_w for i, (iw, jet) in enumerate(zip(inv_w, good_jets)): if jet.y == 0: w[i] = iw jets_1 = [jet for jet in good_jets if jet.y == 1] pdf, edges = np.histogram([j.pt for j in jets_1], density=True, range=[pt_min, pt_max], bins=50) pts = [j.pt for j in jets_1] indices = np.searchsorted(edges, pts) - 1 inv_w = 1. / pdf[indices] inv_w /= inv_w.sum() #w[y_==1] = inv_w for i, (iw, jet) in enumerate(zip(inv_w, good_jets)): if jet.y == 1: w[i] = iw return good_jets, bad_jets, w def crop_dataset(dataset): logging.info(dataset.subproblem) pileup = (dataset.subproblem == 'pileup') good_jets, bad_jets, w = crop(dataset.jets, pileup) cropped_dataset = Dataset(bad_jets) new_dataset = Dataset(good_jets, w) return new_dataset, cropped_dataset
[ "numpy.histogram", "numpy.searchsorted", "logging.warning", "numpy.array", "logging.info" ]
[((615, 653), 'numpy.array', 'np.array', (['[jet.y for jet in good_jets]'], {}), '([jet.y for jet in good_jets])\n', (623, 653), True, 'import numpy as np\n'), ((726, 813), 'numpy.histogram', 'np.histogram', (['[j.pt for j in jets_0]'], {'density': '(True)', 'range': '[pt_min, pt_max]', 'bins': '(50)'}), '([j.pt for j in jets_0], density=True, range=[pt_min, pt_max],\n bins=50)\n', (738, 813), True, 'import numpy as np\n'), ((1141, 1228), 'numpy.histogram', 'np.histogram', (['[j.pt for j in jets_1]'], {'density': '(True)', 'range': '[pt_min, pt_max]', 'bins': '(50)'}), '([j.pt for j in jets_1], density=True, range=[pt_min, pt_max],\n bins=50)\n', (1153, 1228), True, 'import numpy as np\n'), ((1552, 1584), 'logging.info', 'logging.info', (['dataset.subproblem'], {}), '(dataset.subproblem)\n', (1564, 1584), False, 'import logging\n'), ((154, 179), 'logging.warning', 'logging.warning', (['"""pileup"""'], {}), "('pileup')\n", (169, 179), False, 'import logging\n'), ((857, 884), 'numpy.searchsorted', 'np.searchsorted', (['edges', 'pts'], {}), '(edges, pts)\n', (872, 884), True, 'import numpy as np\n'), ((1272, 1299), 'numpy.searchsorted', 'np.searchsorted', (['edges', 'pts'], {}), '(edges, pts)\n', (1287, 1299), True, 'import numpy as np\n')]
from pathlib import Path import numpy import sys from keras.layers import Dense from keras.models import load_model, Sequential from numpy import loadtxt import pickle import os from sklearn.externals import joblib featureVectorSize = 251 def wider_deep_model(): # create model model = Sequential() model.add(Dense(featureVectorSize+20, input_dim=featureVectorSize, kernel_initializer='normal', activation='relu')) model.add(Dense(55, kernel_initializer='normal', activation='relu')) model.add(Dense(1, kernel_initializer='normal')) # Compile model model.compile(loss='mean_squared_error', optimizer='adam') return model def main(): #inputFile = loadtxt("planVectorsSGD2-kmeans-simword-opportuneWordcount.txt", comments="#", delimiter=" ", unpack=False) currentDirPath =os.path.dirname(os.path.realpath(__file__)) dirPath = str(Path.home()) model="nn" if (len(sys.argv)>=2): model = sys.argv[1] if (len(sys.argv)>=3): inputFile = loadtxt(sys.argv[2], comments="#", delimiter=" ", unpack=False) else: inputFile = loadtxt(os.path.join(dirPath,".rheem","mlModelVectors.txt"), comments="#", delimiter=" ", unpack=False) #size = 146; #start = 13; #size = 213 size=251 start = 0 dimInputFile = inputFile.ndim if(dimInputFile==1): inputFile = numpy.reshape(inputFile, (-1,inputFile.size)) x_test = inputFile[:,0:size] y_test = inputFile[start:,size] # x_train = inputFile[:,0:size] # y_train = inputFile[:,size] # # x_test = inputFile[:,0:size] # y_test = inputFile[:,size] # load the model from disk if(model=="forest"): # load the model from disk filename = os.path.join(currentDirPath, "model-forest.sav") print("Loading model: "+filename) model = pickle.load(open(filename, 'rb')) elif(model=="nn"): filename = os.path.join(currentDirPath,'nn.pkl') print("Loading model: "+filename) # Load the pipeline first: model = joblib.load(filename) # Then, load the Keras model: model.named_steps['mlp'].model = load_model(os.path.join(currentDirPath,'keras_model.h5')) # fix random seed for reproducibility seed = 7 numpy.random.seed(seed) #kfold = KFold(n_splits=10, random_state=seed) #results = cross_val_score(regr, x_train, y_train, cv=kfold) #accuracy_score(prediction,y_train) #print("Results: %.2f (%.2f) MSE" % (results.mean(), results.std())) prediction = model.predict(x_test) # for num in range(1,min([34,len(x_test)])): # if num % 2 == 0: # print("estimated time for " + str(x_test[num][size-2]) + "-" + str(x_test[num][size-1]) + " in java : " + str( # prediction[num]) + "(real " + str(y_test[num]) + ")") # else: # print("estimated time for " + str(x_test[num][size-2]) + "-" + str(x_test[num][size-1]) + " in spark : " + str( # prediction[num]) + "(real " + str(y_test[num]) + ")") # print results to text if (len(sys.argv) >= 4): saveLocation = loadtxt(sys.argv[3], comments="#", delimiter=" ", unpack=False) else: saveLocation = os.path.join(dirPath, ".rheem", "estimates.txt") # delete first if(os._exists(saveLocation)): os.remove(saveLocation) text_file = open(saveLocation, "w") # print estimates dimResults = prediction.ndim if (dimResults == 0): text_file.write("%d" % prediction) text_file.write("\n") else: for num in range(0, prediction.size): t = prediction[num] text_file.write("%d" % prediction[num]) text_file.write("\n") text_file.close() print("estimation done!") if __name__ == "__main__": main()
[ "numpy.reshape", "os._exists", "pathlib.Path.home", "sklearn.externals.joblib.load", "os.path.join", "keras.models.Sequential", "os.path.realpath", "numpy.random.seed", "keras.layers.Dense", "numpy.loadtxt", "os.remove" ]
[((299, 311), 'keras.models.Sequential', 'Sequential', ([], {}), '()\n', (309, 311), False, 'from keras.models import load_model, Sequential\n'), ((2280, 2303), 'numpy.random.seed', 'numpy.random.seed', (['seed'], {}), '(seed)\n', (2297, 2303), False, 'import numpy\n'), ((3314, 3338), 'os._exists', 'os._exists', (['saveLocation'], {}), '(saveLocation)\n', (3324, 3338), False, 'import os\n'), ((326, 436), 'keras.layers.Dense', 'Dense', (['(featureVectorSize + 20)'], {'input_dim': 'featureVectorSize', 'kernel_initializer': '"""normal"""', 'activation': '"""relu"""'}), "(featureVectorSize + 20, input_dim=featureVectorSize,\n kernel_initializer='normal', activation='relu')\n", (331, 436), False, 'from keras.layers import Dense\n'), ((446, 503), 'keras.layers.Dense', 'Dense', (['(55)'], {'kernel_initializer': '"""normal"""', 'activation': '"""relu"""'}), "(55, kernel_initializer='normal', activation='relu')\n", (451, 503), False, 'from keras.layers import Dense\n'), ((519, 556), 'keras.layers.Dense', 'Dense', (['(1)'], {'kernel_initializer': '"""normal"""'}), "(1, kernel_initializer='normal')\n", (524, 556), False, 'from keras.layers import Dense\n'), ((834, 860), 'os.path.realpath', 'os.path.realpath', (['__file__'], {}), '(__file__)\n', (850, 860), False, 'import os\n'), ((881, 892), 'pathlib.Path.home', 'Path.home', ([], {}), '()\n', (890, 892), False, 'from pathlib import Path\n'), ((1014, 1077), 'numpy.loadtxt', 'loadtxt', (['sys.argv[2]'], {'comments': '"""#"""', 'delimiter': '""" """', 'unpack': '(False)'}), "(sys.argv[2], comments='#', delimiter=' ', unpack=False)\n", (1021, 1077), False, 'from numpy import loadtxt\n'), ((1370, 1416), 'numpy.reshape', 'numpy.reshape', (['inputFile', '(-1, inputFile.size)'], {}), '(inputFile, (-1, inputFile.size))\n', (1383, 1416), False, 'import numpy\n'), ((1743, 1791), 'os.path.join', 'os.path.join', (['currentDirPath', '"""model-forest.sav"""'], {}), "(currentDirPath, 'model-forest.sav')\n", (1755, 1791), False, 'import os\n'), ((3141, 3204), 'numpy.loadtxt', 'loadtxt', (['sys.argv[3]'], {'comments': '"""#"""', 'delimiter': '""" """', 'unpack': '(False)'}), "(sys.argv[3], comments='#', delimiter=' ', unpack=False)\n", (3148, 3204), False, 'from numpy import loadtxt\n'), ((3238, 3286), 'os.path.join', 'os.path.join', (['dirPath', '""".rheem"""', '"""estimates.txt"""'], {}), "(dirPath, '.rheem', 'estimates.txt')\n", (3250, 3286), False, 'import os\n'), ((3349, 3372), 'os.remove', 'os.remove', (['saveLocation'], {}), '(saveLocation)\n', (3358, 3372), False, 'import os\n'), ((1116, 1169), 'os.path.join', 'os.path.join', (['dirPath', '""".rheem"""', '"""mlModelVectors.txt"""'], {}), "(dirPath, '.rheem', 'mlModelVectors.txt')\n", (1128, 1169), False, 'import os\n'), ((1926, 1964), 'os.path.join', 'os.path.join', (['currentDirPath', '"""nn.pkl"""'], {}), "(currentDirPath, 'nn.pkl')\n", (1938, 1964), False, 'import os\n'), ((2057, 2078), 'sklearn.externals.joblib.load', 'joblib.load', (['filename'], {}), '(filename)\n', (2068, 2078), False, 'from sklearn.externals import joblib\n'), ((2170, 2216), 'os.path.join', 'os.path.join', (['currentDirPath', '"""keras_model.h5"""'], {}), "(currentDirPath, 'keras_model.h5')\n", (2182, 2216), False, 'import os\n')]
import matplotlib.pyplot as plt import numpy as np from brancher.variables import RootVariable, RandomVariable, ProbabilisticModel from brancher.standard_variables import NormalVariable, DeterministicVariable, MultivariateNormalVariable from brancher import inference import brancher.functions as BF N_itr = 250 N_smpl = 1000 optimizer = "SGD" lr = 0.001 #0.0001 # Probabilistic model # T = 30 dt = 0.1 driving_noise = 0.1 measure_noise = 0.15 x0 = NormalVariable(0., driving_noise, 'x0') y0 = NormalVariable(x0, measure_noise, 'y0') x = [x0] y = [y0] x_names = ["x0"] y_names = ["y0"] y_range = [t for t in range(T) if (t < 0 or t > 28)] for t in range(1, T): x_names.append("x{}".format(t)) x.append(NormalVariable(x[t - 1], np.sqrt(dt)*driving_noise, x_names[t])) if t in y_range: y_name = "y{}".format(t) y_names.append(y_name) y.append(NormalVariable(x[t], measure_noise, y_name)) AR_model = ProbabilisticModel(x + y) # Generate data # data = AR_model._get_sample(number_samples=1) time_series = [float(data[yt].data) for yt in y] ground_truth = [float(data[xt].data) for xt in x] #true_b = data[b].data #print("The true coefficient is: {}".format(float(true_b))) # Observe data # [yt.observe(data[yt][:, 0, :]) for yt in y] # get time series #plt.plot([data[xt][:, 0, :] for xt in x]) #plt.scatter(y_range, time_series, c="k") #plt.show() # Structured variational distribution # Qx = [NormalVariable(0., 1., 'x0', learnable=True)] Qx_mean = [RootVariable(0., 'x0_mean', learnable=True)] Qlambda = [RootVariable(-1., 'x0_lambda', learnable=True)] for t in range(1, T): if t in y_range: l = 0. else: l = 1. Qx_mean.append(RootVariable(0, x_names[t] + "_mean", learnable=True)) Qlambda.append(RootVariable(l, x_names[t] + "_lambda", learnable=True)) Qx.append(NormalVariable(BF.sigmoid(Qlambda[t])*Qx[t - 1] + (1 - BF.sigmoid(Qlambda[t]))*Qx_mean[t], np.sqrt(dt)*driving_noise, x_names[t], learnable=True)) variational_posterior = ProbabilisticModel(Qx) AR_model.set_posterior_model(variational_posterior) # Inference # inference.perform_inference(AR_model, number_iterations=N_itr, number_samples=N_smpl, optimizer=optimizer, lr=lr) loss_list1 = AR_model.diagnostics["loss curve"] samples_PE = AR_model.posterior_model.get_sample(1000) # # # Plot posterior # #from brancher.visualizations import plot_density # #plot_density(AR_model.posterior_model, variables=["x0", "x1"]) # # # ELBO # N_ELBO_smpl = 5000 # ELBO1 = AR_model.estimate_log_model_evidence(N_ELBO_smpl) # print("PE: {}".format(ELBO1)) # # # Statistics # posterior_samples1 = AR_model._get_posterior_sample(2000) # #b_posterior_samples1 = posterior_samples1[b].detach().numpy().flatten() # #b_mean1 = np.mean(b_posterior_samples1) # #b_sd1 = np.sqrt(np.var(b_posterior_samples1)) # # x_mean1 = [] # lower_bound1 = [] # upper_bound1 = [] # for xt in x: # x_posterior_samples1 = posterior_samples1[xt].detach().numpy().flatten() # mean1 = np.mean(x_posterior_samples1) # sd1 = np.sqrt(np.var(x_posterior_samples1)) # x_mean1.append(mean1) # lower_bound1.append(mean1 - sd1) # upper_bound1.append(mean1 + sd1) # #print("The estimated coefficient is: {} +- {}".format(b_mean1, b_sd1)) # # Two subplots, unpack the axes array immediately # f, (ax1, ax2, ax3) = plt.subplots(1, 3) # ax1.plot(range(T), x_mean1, color="b", label="PE") # ax1.scatter(y_range, time_series, color="k") # ax1.plot(range(T), ground_truth, color="k", ls ="--", lw=1.5) # ax1.fill_between(range(T), lower_bound1, upper_bound1, color="b", alpha=0.25) # ax1.set_title("Time series") # ax2.plot(np.array(loss_list1), color="b") # ax2.set_title("Convergence") # ax2.set_xlabel("Iteration") # ax3.hist(b_posterior_samples1, 25, color="b", alpha=0.25) # ax3.set_title("Posterior samples (b)") # ax3.set_xlim(0, 1) # plt.show() # Mean-field variational distribution # #Qb = BetaVariable(8., 1., "b", learnable=True) Qx = [NormalVariable(0., 1., 'x0', learnable=True)] for t in range(1, T): Qx.append(NormalVariable(0, 2., x_names[t], learnable=True)) variational_posterior = ProbabilisticModel(Qx) AR_model.set_posterior_model(variational_posterior) # Inference # inference.perform_inference(AR_model, number_iterations=N_itr, number_samples=N_smpl, optimizer=optimizer, lr=lr) loss_list2 = AR_model.diagnostics["loss curve"] #Plot posterior from brancher.visualizations import plot_density plot_density(AR_model.posterior_model, variables=["x0", "x1"]) # ELBO ELBO2 = AR_model.estimate_log_model_evidence(N_ELBO_smpl) print("MF: {}".format(ELBO2)) # # samples_MF = AR_model.posterior_model.get_sample(1000) # # # Statistics # posterior_samples2 = AR_model._get_posterior_sample(2000) # #b_posterior_samples2 = posterior_samples2[b].detach().numpy().flatten() # #b_mean2 = np.mean(b_posterior_samples2) # #b_sd2 = np.sqrt(np.var(b_posterior_samples2)) # # x_mean2 = [] # lower_bound2 = [] # upper_bound2 = [] # for xt in x: # x_posterior_samples2 = posterior_samples2[xt].detach().numpy().flatten() # mean2 = np.mean(x_posterior_samples2) # sd2 = np.sqrt(np.var(x_posterior_samples2)) # x_mean2.append(mean2) # lower_bound2.append(mean2 - sd2) # upper_bound2.append(mean2 + sd2) #print("The estimated coefficient is: {} +- {}".format(b_mean2, b_sd2)) # # Multivariate normal variational distribution # # QV = MultivariateNormalVariable(loc=np.zeros((T,)), # covariance_matrix=2*np.identity(T), # learnable=True) # #Qb = BetaVariable(8., 1., "b", learnable=True) # Qx = [NormalVariable(QV[0], 0.1, 'x0', learnable=True)] # # for t in range(1, T): # Qx.append(NormalVariable(QV[t], 0.1, x_names[t], learnable=True)) # variational_posterior = ProbabilisticModel(Qx) # AR_model.set_posterior_model(variational_posterior) # # # Inference # # inference.perform_inference(AR_model, # number_iterations=N_itr, # number_samples=N_smpl, # optimizer=optimizer, # lr=lr) # # loss_list3 = AR_model.diagnostics["loss curve"] # # # Plot posterior # #plot_density(AR_model.posterior_model, variables=["x0", "x1"]) # # # ELBO # ELBO3 = AR_model.estimate_log_model_evidence(N_ELBO_smpl) # print("MN: {}".format(ELBO3)) # # samples_MN = AR_model.posterior_model.get_sample(1000) # # # Statistics # posterior_samples3 = AR_model._get_posterior_sample(2000) # #b_posterior_samples3 = posterior_samples3[b].detach().numpy().flatten() # #b_mean3 = np.mean(b_posterior_samples3) # #b_sd3 = np.sqrt(np.var(b_posterior_samples3)) # # x_mean3 = [] # lower_bound3 = [] # upper_bound3 = [] # for xt in x: # x_posterior_samples3 = posterior_samples3[xt].detach().numpy().flatten() # mean3 = np.mean(x_posterior_samples3) # sd3 = np.sqrt(np.var(x_posterior_samples3)) # x_mean3.append(mean3) # lower_bound3.append(mean3 - sd3) # upper_bound3.append(mean3 + sd3) # #print("The estimated coefficient is: {} +- {}".format(b_mean3, b_sd3)) # # # Structured NN distribution # # latent_size = 10 # hidden_size = 10 # #Qb = BetaVariable(8., 1., "b", learnable=True) # Qepsilon = NormalVariable(np.zeros((10,1)), np.ones((10,)), 'epsilon', learnable=True) # W1 = RootVariable(np.random.normal(0, 0.1, (hidden_size, latent_size)), "W1", learnable=True) # W2 = RootVariable(np.random.normal(0, 0.1, (T, hidden_size)), "W2", learnable=True) # pre_x = BF.matmul(W2, BF.sigmoid(BF.matmul(W1, Qepsilon))) # Qx = [] # for t in range(0, T): # pre_x_t = DeterministicVariable(pre_x[t], "x{}_mean".format(t), learnable=True) # Qx.append(NormalVariable(pre_x_t, 1., x_names[t], learnable=True)) # variational_posterior = ProbabilisticModel(Qx) # AR_model.set_posterior_model(variational_posterior) # # # Inference # # inference.perform_inference(AR_model, # number_iterations=N_itr, # number_samples=N_smpl, # optimizer=optimizer, # lr=lr) # # loss_list4 = AR_model.diagnostics["loss curve"] #plot_density(AR_model.posterior_model, variables=["x0", "x1"]) #plt.show() ## ELBO #ELBO4 = AR_model.estimate_log_model_evidence(N_ELBO_smpl) #print("NN: {}".format(ELBO4)) # # samples_NN = AR_model.posterior_model.get_sample(1000) # # # Statistics # posterior_samples4 = AR_model._get_posterior_sample(2000) # #b_posterior_samples4 = posterior_samples4[b].detach().numpy().flatten() # #b_mean4 = np.mean(b_posterior_samples4) # #b_sd4 = np.sqrt(np.var(b_posterior_samples2)) # # x_mean4 = [] # lower_bound4 = [] # upper_bound4 = [] # for xt in x: # x_posterior_samples4 = posterior_samples4[xt].detach().numpy().flatten() # mean4 = np.mean(x_posterior_samples4) # sd4 = np.sqrt(np.var(x_posterior_samples4)) # x_mean4.append(mean4) # lower_bound4.append(mean4 - sd4) # upper_bound4.append(mean4 + sd4) # #print("The estimated coefficient is: {} +- {}".format(b_mean4, b_sd4)) # # # Densities # from brancher.visualizations import plot_multiple_samples # #plot_multiple_samples([samples_PE, samples_MF, samples_NN], variables=["x0", "x1"], labels=["PE","MF", "NN"]) # #plot_multiple_samples([samples_PE, samples_MF, samples_MN, samples_NN], variables=["x0", "x1"], labels=["PE","MF", "MN", "NN"]) # plot_multiple_samples([samples_PE, samples_NN], variables=["x0", "x1"], labels=["PE", "NN"]) # plt.show() # # # Two subplots, unpack the axes array immediately # f, (ax1, ax2) = plt.subplots(1, 2) # ax1.plot(range(T), x_mean1, color="b", label="PE") # ax1.plot(range(T), x_mean2, color="r", label="MF") # ax1.plot(range(T), x_mean3, color="g", label="MV") # ax1.plot(range(T), x_mean4, color="m", label="NN") # #ax1.scatter(y_range, time_series, color="k") # ax1.plot(range(T), ground_truth, color="k", ls ="--", lw=1.5) # ax1.fill_between(range(T), lower_bound1, upper_bound1, color="b", alpha=0.25) # ax1.fill_between(range(T), lower_bound2, upper_bound2, color="r", alpha=0.25) # ax1.fill_between(range(T), lower_bound3, upper_bound3, color="g", alpha=0.25) # ax1.fill_between(range(T), lower_bound4, upper_bound4, color="m", alpha=0.25) # ax1.set_title("Time series") # ax2.plot(np.array(loss_list1), color="b") # ax2.plot(np.array(loss_list2), color="r") # ax2.plot(np.array(loss_list3), color="g") # ax2.plot(np.array(loss_list4), color="m") # ax2.set_title("Convergence") # ax2.set_xlabel("Iteration") # plt.show()
[ "brancher.visualizations.plot_density", "numpy.sqrt", "brancher.functions.sigmoid", "brancher.variables.ProbabilisticModel", "brancher.variables.RootVariable", "brancher.standard_variables.NormalVariable", "brancher.inference.perform_inference" ]
[((452, 492), 'brancher.standard_variables.NormalVariable', 'NormalVariable', (['(0.0)', 'driving_noise', '"""x0"""'], {}), "(0.0, driving_noise, 'x0')\n", (466, 492), False, 'from brancher.standard_variables import NormalVariable, DeterministicVariable, MultivariateNormalVariable\n'), ((497, 536), 'brancher.standard_variables.NormalVariable', 'NormalVariable', (['x0', 'measure_noise', '"""y0"""'], {}), "(x0, measure_noise, 'y0')\n", (511, 536), False, 'from brancher.standard_variables import NormalVariable, DeterministicVariable, MultivariateNormalVariable\n'), ((937, 962), 'brancher.variables.ProbabilisticModel', 'ProbabilisticModel', (['(x + y)'], {}), '(x + y)\n', (955, 962), False, 'from brancher.variables import RootVariable, RandomVariable, ProbabilisticModel\n'), ((2045, 2067), 'brancher.variables.ProbabilisticModel', 'ProbabilisticModel', (['Qx'], {}), '(Qx)\n', (2063, 2067), False, 'from brancher.variables import RootVariable, RandomVariable, ProbabilisticModel\n'), ((2135, 2252), 'brancher.inference.perform_inference', 'inference.perform_inference', (['AR_model'], {'number_iterations': 'N_itr', 'number_samples': 'N_smpl', 'optimizer': 'optimizer', 'lr': 'lr'}), '(AR_model, number_iterations=N_itr,\n number_samples=N_smpl, optimizer=optimizer, lr=lr)\n', (2162, 2252), False, 'from brancher import inference\n'), ((4261, 4283), 'brancher.variables.ProbabilisticModel', 'ProbabilisticModel', (['Qx'], {}), '(Qx)\n', (4279, 4283), False, 'from brancher.variables import RootVariable, RandomVariable, ProbabilisticModel\n'), ((4351, 4468), 'brancher.inference.perform_inference', 'inference.perform_inference', (['AR_model'], {'number_iterations': 'N_itr', 'number_samples': 'N_smpl', 'optimizer': 'optimizer', 'lr': 'lr'}), '(AR_model, number_iterations=N_itr,\n number_samples=N_smpl, optimizer=optimizer, lr=lr)\n', (4378, 4468), False, 'from brancher import inference\n'), ((4692, 4754), 'brancher.visualizations.plot_density', 'plot_density', (['AR_model.posterior_model'], {'variables': "['x0', 'x1']"}), "(AR_model.posterior_model, variables=['x0', 'x1'])\n", (4704, 4754), False, 'from brancher.visualizations import plot_density\n'), ((1435, 1481), 'brancher.standard_variables.NormalVariable', 'NormalVariable', (['(0.0)', '(1.0)', '"""x0"""'], {'learnable': '(True)'}), "(0.0, 1.0, 'x0', learnable=True)\n", (1449, 1481), False, 'from brancher.standard_variables import NormalVariable, DeterministicVariable, MultivariateNormalVariable\n'), ((1492, 1536), 'brancher.variables.RootVariable', 'RootVariable', (['(0.0)', '"""x0_mean"""'], {'learnable': '(True)'}), "(0.0, 'x0_mean', learnable=True)\n", (1504, 1536), False, 'from brancher.variables import RootVariable, RandomVariable, ProbabilisticModel\n'), ((1548, 1595), 'brancher.variables.RootVariable', 'RootVariable', (['(-1.0)', '"""x0_lambda"""'], {'learnable': '(True)'}), "(-1.0, 'x0_lambda', learnable=True)\n", (1560, 1595), False, 'from brancher.variables import RootVariable, RandomVariable, ProbabilisticModel\n'), ((4103, 4149), 'brancher.standard_variables.NormalVariable', 'NormalVariable', (['(0.0)', '(1.0)', '"""x0"""'], {'learnable': '(True)'}), "(0.0, 1.0, 'x0', learnable=True)\n", (4117, 4149), False, 'from brancher.standard_variables import NormalVariable, DeterministicVariable, MultivariateNormalVariable\n'), ((1700, 1753), 'brancher.variables.RootVariable', 'RootVariable', (['(0)', "(x_names[t] + '_mean')"], {'learnable': '(True)'}), "(0, x_names[t] + '_mean', learnable=True)\n", (1712, 1753), False, 'from brancher.variables import RootVariable, RandomVariable, ProbabilisticModel\n'), ((1774, 1829), 'brancher.variables.RootVariable', 'RootVariable', (['l', "(x_names[t] + '_lambda')"], {'learnable': '(True)'}), "(l, x_names[t] + '_lambda', learnable=True)\n", (1786, 1829), False, 'from brancher.variables import RootVariable, RandomVariable, ProbabilisticModel\n'), ((4186, 4236), 'brancher.standard_variables.NormalVariable', 'NormalVariable', (['(0)', '(2.0)', 'x_names[t]'], {'learnable': '(True)'}), '(0, 2.0, x_names[t], learnable=True)\n', (4200, 4236), False, 'from brancher.standard_variables import NormalVariable, DeterministicVariable, MultivariateNormalVariable\n'), ((881, 924), 'brancher.standard_variables.NormalVariable', 'NormalVariable', (['x[t]', 'measure_noise', 'y_name'], {}), '(x[t], measure_noise, y_name)\n', (895, 924), False, 'from brancher.standard_variables import NormalVariable, DeterministicVariable, MultivariateNormalVariable\n'), ((739, 750), 'numpy.sqrt', 'np.sqrt', (['dt'], {}), '(dt)\n', (746, 750), True, 'import numpy as np\n'), ((1965, 1976), 'numpy.sqrt', 'np.sqrt', (['dt'], {}), '(dt)\n', (1972, 1976), True, 'import numpy as np\n'), ((1860, 1882), 'brancher.functions.sigmoid', 'BF.sigmoid', (['Qlambda[t]'], {}), '(Qlambda[t])\n', (1870, 1882), True, 'import brancher.functions as BF\n'), ((1900, 1922), 'brancher.functions.sigmoid', 'BF.sigmoid', (['Qlambda[t]'], {}), '(Qlambda[t])\n', (1910, 1922), True, 'import brancher.functions as BF\n')]
import pandas as pd import numpy as np import matplotlib.pyplot as plt plt.rcParams.update({'font.size': 22}) import math plt.rcParams.update({'font.size': 22}) name = "Pl" data = pd.read_csv(name + ".csv", names=["l", "P"]) K = 0.2 data["P1"] = 9.80665 * K * data["P"] X = data["l"].values sigma_X = 0.4 Y = data["P"].values sigma_Y = 0.7 A = np.vstack([X[1:], np.ones(len(X[1:]))]).T k, b = np.linalg.lstsq(A, Y[1:], rcond=None)[0] fig = plt.figure(figsize=(12, 7)) ax = fig.gca() plt.scatter(X, Y, marker=".") plt.errorbar(X, Y, xerr=sigma_X, yerr=sigma_Y, linestyle="None") delta_x = (X.max() - X.min()) / len(X) delta_y = (Y.max() - Y.min()) / len(Y) ax.set_xlim(X.min() - delta_x/2, X.max() + delta_x/2) ax.set_ylim((Y.min() - delta_y/2), Y.max() + delta_y/2) plt.xlabel("$l, мм$") plt.ylabel("$P, 10^{-3} Па$") plt.plot(X, (k*X + b), 'r', label='Fitted line') plt.grid(True) plt.savefig("./" + name + ".png")
[ "matplotlib.pyplot.grid", "matplotlib.pyplot.savefig", "pandas.read_csv", "matplotlib.pyplot.ylabel", "matplotlib.pyplot.xlabel", "matplotlib.pyplot.plot", "matplotlib.pyplot.rcParams.update", "matplotlib.pyplot.figure", "matplotlib.pyplot.scatter", "matplotlib.pyplot.errorbar", "numpy.linalg.ls...
[((71, 109), 'matplotlib.pyplot.rcParams.update', 'plt.rcParams.update', (["{'font.size': 22}"], {}), "({'font.size': 22})\n", (90, 109), True, 'import matplotlib.pyplot as plt\n'), ((122, 160), 'matplotlib.pyplot.rcParams.update', 'plt.rcParams.update', (["{'font.size': 22}"], {}), "({'font.size': 22})\n", (141, 160), True, 'import matplotlib.pyplot as plt\n'), ((180, 224), 'pandas.read_csv', 'pd.read_csv', (["(name + '.csv')"], {'names': "['l', 'P']"}), "(name + '.csv', names=['l', 'P'])\n", (191, 224), True, 'import pandas as pd\n'), ((440, 467), 'matplotlib.pyplot.figure', 'plt.figure', ([], {'figsize': '(12, 7)'}), '(figsize=(12, 7))\n', (450, 467), True, 'import matplotlib.pyplot as plt\n'), ((483, 512), 'matplotlib.pyplot.scatter', 'plt.scatter', (['X', 'Y'], {'marker': '"""."""'}), "(X, Y, marker='.')\n", (494, 512), True, 'import matplotlib.pyplot as plt\n'), ((513, 577), 'matplotlib.pyplot.errorbar', 'plt.errorbar', (['X', 'Y'], {'xerr': 'sigma_X', 'yerr': 'sigma_Y', 'linestyle': '"""None"""'}), "(X, Y, xerr=sigma_X, yerr=sigma_Y, linestyle='None')\n", (525, 577), True, 'import matplotlib.pyplot as plt\n'), ((766, 787), 'matplotlib.pyplot.xlabel', 'plt.xlabel', (['"""$l, мм$"""'], {}), "('$l, мм$')\n", (776, 787), True, 'import matplotlib.pyplot as plt\n'), ((788, 817), 'matplotlib.pyplot.ylabel', 'plt.ylabel', (['"""$P, 10^{-3} Па$"""'], {}), "('$P, 10^{-3} Па$')\n", (798, 817), True, 'import matplotlib.pyplot as plt\n'), ((818, 866), 'matplotlib.pyplot.plot', 'plt.plot', (['X', '(k * X + b)', '"""r"""'], {'label': '"""Fitted line"""'}), "(X, k * X + b, 'r', label='Fitted line')\n", (826, 866), True, 'import matplotlib.pyplot as plt\n'), ((867, 881), 'matplotlib.pyplot.grid', 'plt.grid', (['(True)'], {}), '(True)\n', (875, 881), True, 'import matplotlib.pyplot as plt\n'), ((882, 915), 'matplotlib.pyplot.savefig', 'plt.savefig', (["('./' + name + '.png')"], {}), "('./' + name + '.png')\n", (893, 915), True, 'import matplotlib.pyplot as plt\n'), ((393, 430), 'numpy.linalg.lstsq', 'np.linalg.lstsq', (['A', 'Y[1:]'], {'rcond': 'None'}), '(A, Y[1:], rcond=None)\n', (408, 430), True, 'import numpy as np\n')]
import numpy as np from pandas import read_csv # an example of asia bayesian net: # https://www.eecis.udel.edu/~shatkay/Course/papers/Lauritzen1988.pdf class BayesianNet(object): def __init__(self, names, edges, tables=None): self.n_nodes = len(names) if tables is None: tables = [[0]] * self.n_nodes self.nodes = [{'name': name, 'table': np.array( table)} for name, table in zip(names, tables)] self.name2idx = {k: v for v, k in enumerate(names)} self.graph = np.zeros((self.n_nodes, self.n_nodes)) for edge in edges: self.graph[self.name2idx[edge[1]], self.name2idx[edge[0]]] = 1 self.binary = np.array( [1 << self.n_nodes - 1 - i for i in range(self.n_nodes)]) def fit(self, data): data_size = len(data) for i, node in enumerate(self.nodes): table = [] parents = self.graph[i] == 1 marginal = data[:, parents] index = np.zeros(data.shape[0]) if marginal.shape[1] > 0: index = ( marginal * self.binary[-marginal.shape[1]:]).sum(axis=1) for j in range(2**parents.sum()): table.append(data[(index == j), i].sum() / (index == j).sum()) node['table'] = np.array(table) def joint_p(self, values): p = 1 for i in range(self.n_nodes): index = 0 parents = self.graph[i] == 1 if parents.sum() > 0: index = np.dot(values[parents], self.binary[-parents.sum():]) p *= (1 - values[i]) + (2 * values[i] - 1) * \ self.nodes[i]['table'][int(index)] return p def marginal_p(self, condition): p = 0 values = -np.ones(self.n_nodes) for v in condition: values[self.name2idx[v[1]]] = int(v[0] != '~') mask = np.arange(self.n_nodes)[(values == -1)] n_unkowns = self.n_nodes - len(condition) for i in range(2**n_unkowns): values[mask] = np.array( [int(x) for x in '{:0{size}b}'.format(i, size=n_unkowns)]) p += self.joint_p(values) return p def query(self, v, condition): p_pos = self.marginal_p([f'+{v}'] + condition) / self.marginal_p(condition) return [1 - p_pos, p_pos] def get_asia_data(url): return read_csv(url).apply(lambda x: x == 'yes').astype(int).values def main(): names = 'ATSLBEXD' edges = ['AT', 'SL', 'SB', 'TE', 'LE', 'BD', 'EX', 'ED'] #tables = [[0.01], [0.01, 0.05], [0.5], [0.01, 0.1], [0.3, 0.6], [0, 1, 1, 1], [0.05, 0.98], [0.1, 0.7, 0.8, 0.9]] # also can use predefined conditional tables bn = BayesianNet(list(names), edges) asia_url = 'http://www.ccd.pitt.edu/wiki/images/ASIA10k.csv' bn.fit(get_asia_data(asia_url)) print(bn.nodes) for condition in [[], ['+A', '~S'], ['+A', '~S', '~D', '+X']]: for c in ['T', 'L', 'B', 'E']: print('p({}|{})={}'.format(c, ','.join( condition), bn.query(c, condition))) if __name__ == "__main__": main()
[ "numpy.ones", "pandas.read_csv", "numpy.array", "numpy.zeros", "numpy.arange" ]
[((532, 570), 'numpy.zeros', 'np.zeros', (['(self.n_nodes, self.n_nodes)'], {}), '((self.n_nodes, self.n_nodes))\n', (540, 570), True, 'import numpy as np\n'), ((1001, 1024), 'numpy.zeros', 'np.zeros', (['data.shape[0]'], {}), '(data.shape[0])\n', (1009, 1024), True, 'import numpy as np\n'), ((1319, 1334), 'numpy.array', 'np.array', (['table'], {}), '(table)\n', (1327, 1334), True, 'import numpy as np\n'), ((1791, 1812), 'numpy.ones', 'np.ones', (['self.n_nodes'], {}), '(self.n_nodes)\n', (1798, 1812), True, 'import numpy as np\n'), ((1915, 1938), 'numpy.arange', 'np.arange', (['self.n_nodes'], {}), '(self.n_nodes)\n', (1924, 1938), True, 'import numpy as np\n'), ((382, 397), 'numpy.array', 'np.array', (['table'], {}), '(table)\n', (390, 397), True, 'import numpy as np\n'), ((2401, 2414), 'pandas.read_csv', 'read_csv', (['url'], {}), '(url)\n', (2409, 2414), False, 'from pandas import read_csv\n')]
# -*- coding: utf-8 -*- import numpy as np class Population: """ A Population is a measure over the the cells at given timepoint. Parameters ---------- time : int or float The time at which the cells where measured. p : 1-D array-like Measure over the cells at the given timepoint. name : str Optional population name. """ def __init__(self, time, p, name=None): self.time = time self.p = np.asarray(p, dtype=np.float64) self.name = name def normalize(self): """ Make the measure sum to 1, i.e. be a probability distribution over cells. """ self.p = self.p / self.p.sum() def make_binary(self): """ Set non-zero values to 1. """ p = np.zeros(len(self.p)) p[self.p > 0.0] = 1.0 self.p = p @staticmethod def get_missing_population(*populations): initial_p_sum = np.array([pop.p for pop in populations]).T.sum(axis=1) missing_cells = np.where(initial_p_sum == 0)[0] if len(missing_cells) > 0: missing_cells_p = np.zeros_like(populations[0].p) missing_cells_p[missing_cells] = 1.0 return Population(populations[0].time, missing_cells_p, 'Other') return None @staticmethod def copy(*populations, normalize=None, add_missing=False): populations_copy = [] for pop in populations: pop_copy = Population(pop.time, pop.p, pop.name) populations_copy.append(pop_copy) populations = populations_copy if add_missing: # add "other" population if any cells are missing across all populations missing_pop = Population.get_missing_population(*populations) if missing_pop is not None: populations.append(missing_pop) if normalize or normalize == False: for pop in populations: if normalize: pop.normalize() else: pop.make_binary() return populations
[ "numpy.where", "numpy.array", "numpy.zeros_like", "numpy.asarray" ]
[((470, 501), 'numpy.asarray', 'np.asarray', (['p'], {'dtype': 'np.float64'}), '(p, dtype=np.float64)\n', (480, 501), True, 'import numpy as np\n'), ((1035, 1063), 'numpy.where', 'np.where', (['(initial_p_sum == 0)'], {}), '(initial_p_sum == 0)\n', (1043, 1063), True, 'import numpy as np\n'), ((1132, 1163), 'numpy.zeros_like', 'np.zeros_like', (['populations[0].p'], {}), '(populations[0].p)\n', (1145, 1163), True, 'import numpy as np\n'), ((956, 996), 'numpy.array', 'np.array', (['[pop.p for pop in populations]'], {}), '([pop.p for pop in populations])\n', (964, 996), True, 'import numpy as np\n')]