def train_model_chunks(model,
loader,
batch_size=8,
num_epoch=100,
print_every=10,
epoch_done=0):
loss_avg = running_average(100)
acc_avg = running_average(100)
for i in range(num_epoch):
for j in range(loader.num_chunks):
print("\nEpoch no. %s, loading chunk no. %s\n" %
(epoch_done + i + 1, j + 1))
X, Y = loader.next_chunk()
for k in range(len(Y) / batch_size):
loss, accuracy = model.train_on_batch(
X[k * batch_size:(k + 1) * batch_size],
Y[k * batch_size:(k + 1) * batch_size])
curr_avg_loss = loss_avg.upsert(loss)
curr_avg_acc = acc_avg.upsert(accuracy)
if (k + 1) % print_every == 0:
print(
"Epoch: %d, chunk: %d, batch: %d, loss=%.4f, accuracy=%.4f"
% (epoch_done + i + 1, j + 1, k + 1, curr_avg_loss,
curr_avg_acc))
# if (i+1) % 5 == 0:
# model.save_weights('model_weights_cnn_'+str(epoch_done+i+1)+'_'+str(curr_avg_acc)+'.h5',overwrite=True)
return model
def train_model_chunks(model,loader,batch_size=8,num_epoch=100,print_every=10,epoch_done=0):
loss_avg = running_average(100)
acc_avg = running_average(100)
for i in range(num_epoch):
for j in range(loader.num_chunks):
print("\nEpoch no. %s, loading chunk no. %s\n"%(epoch_done+i+1,j+1))
X,Y = loader.next_chunk()
for k in range(len(Y)/batch_size):
loss, accuracy = model.train_on_batch(X[k*batch_size:(k+1)*batch_size], Y[k*batch_size:(k+1)*batch_size])
curr_avg_loss = loss_avg.upsert(loss)
curr_avg_acc = acc_avg.upsert(accuracy)
if (k+1)%print_every == 0:
print("Epoch: %d, chunk: %d, batch: %d, loss=%.4f, accuracy=%.4f"%(epoch_done+i+1,j+1,k+1,curr_avg_loss,curr_avg_acc))
# if (i+1) % 5 == 0:
# model.save_weights('model_weights_cnn_'+str(epoch_done+i+1)+'_'+str(curr_avg_acc)+'.h5',overwrite=True)
return model
def main():
# parse cmd args
if len(sys.argv) > 1:
outfile = sys.argv[1]
else:
outfile = "average.out"
if len(sys.argv) > 2:
col = int(sys.argv[2]) - 1
else:
col = 0
if len(sys.argv) > 3:
width = int(sys.argv[3])
else:
width = 1
# read data from stdin
data = sys.stdin.readlines()
# extract data
values = np.zeros(len(data), dtype=np.float64);
for i, line in enumerate(data):
data[i] = line.split()
values[i] = data[i][col]
# compute running average
average = running_average(values, width, "gaussian")
# save average in file
of = open(outfile, 'w')
for i, line in enumerate(data):
for j in line[:col]:
of.write(j)
of.write("\t")
of.write(str(average[i]))
of.write("\t")
for j in line[col+1:]:
of.write(j)
of.write("\t")
of.write("\n")
of.close()
print "Computed running average of column {} of input data ({} datapoints)".format(col+1, len(average))
def get_data(filename, width, wintype):
datafile = open(filename, "r")
data = datafile.readlines()
datafile.close()
times = np.zeros(len(data), dtype=float)
values = np.zeros(len(data), dtype=float)
for i, line in enumerate(data):
times[i] = float(line.split()[0])
values[i] = float(line.split()[1])
# Convert time to ns
times = times / 1000.0
average = running_average(values, width, win=wintype)
print "Running average computed"
return (average, times)
for i, line in enumerate(boundData):
boundTime[i] = float(line.split()[0])
boundAngle[i] = float(line.split()[1])
for i, line in enumerate(freeData):
freeAngle[i] = float(line.split()[1])
# Convert time to ns
boundTime = boundTime / 1000.0
freeTime = freeTime / 1000.0
# compute running average
width = 100
wintype = "gaussian"
print "Starting average computation"
boundAngleAverage = running_average(boundAngle, width, win=wintype)
freeAngleAverage = running_average(freeAngle, width, win=wintype)
print "Average computation done"
# pickle Data
pickleDataFile = open(pickleDataFilename, 'w')
pickle.dump([boundAngle, freeAngle, boundTime, freeTime, \
boundAngleAverage, freeAngleAverage], pickleDataFile)
pickleDataFile.close()
# if pickle file is newer than data files
else:
# unpickle data
pickleDataFile = open(pickleDataFilename, 'r')
[boundAngle, freeAngle, boundTime, freeTime, boundAngleAverage, freeAngleAverage] = pickle.load(pickleDataFile)
pickleDataFile.close()
