def main(args):
config = args
print('loading data')
data, iters, vocab, stats, Q_field = get_data(config)
config.q_size = stats['question_size']
config.a_size = stats['answer_size']
config.c_size = stats['topic_size']
config.keys = ['A', 'c']
config.sizes = {'A': config.a_size, 'c': config.c_size}
if config.verbose:
x = next(iter(iters['val']))
print(x.answers[0].data)
print('loading model')
net, best_acc, config.start_epoch = get_net(config, vocab)
if config.train:
print("Let\'s start Training")
train_all(net, data, iters, config)
else:
print("Let\'s start Testing")
test_and_save(net, data, iters[config.test_iter], config)
rStatesSave = []
mStatesSave = []
for i, fileLists in enumerate(fileNames):
for j, fList in enumerate(fileLists):
dists = []
for f in fList:
print i, j
print f
fbProbs, numSteps, model, statesInRbt, states, logProbs_T, logProbs_O, logProbs, poses, actions, obs, \
actionType, actionSelectionType, numMechanismTypes, numParticles, numRepeats, neff_fract, \
modelNums, realStates, BIAS, FTSD, FOSD, RTSD, ROSD \
= readData.get_data(dirName+f)
# save real
rStatesSave.append(realStates[1:])
# find max state at each step
mStatesSave.append([])
for k, (Ss, lPs) in enumerate(zip(states[1:], logProbs[1:])):
maxProbs = []
maxProbInds = []
for l in relevantMT:
maxProbs.append(max(lPs[l]))
maxProbInds.append(lPs[l].index(maxProbs[-1]))
maxModelInd = maxProbs.index(max(maxProbs))
maxProbInd = maxProbInds[maxModelInd]
rStatesSave = []
mStatesSave = []
for i,fileLists in enumerate(fileNames):
for j,fList in enumerate(fileLists):
dists = []
for f in fList:
print i,j
print f
fbProbs, numSteps, model, statesInRbt, states, logProbs_T, logProbs_O, logProbs, poses, actions, obs, \
actionType, actionSelectionType, numMechanismTypes, numParticles, numRepeats, neff_fract, \
modelNums, realStates, BIAS, FTSD, FOSD, RTSD, ROSD \
= readData.get_data(dirName+f)
# save real
rStatesSave.append(realStates[1:])
# find max state at each step
mStatesSave.append([])
for k,(Ss,lPs) in enumerate(zip(states[1:],logProbs[1:])):
maxProbs = []
maxProbInds = []
for l in relevantMT:
maxProbs.append(max(lPs[l]))
maxProbInds.append(lPs[l].index(maxProbs[-1]))
maxModelInd = maxProbs.index(max(maxProbs))
maxProbInd = maxProbInds[maxModelInd]
def runNet():
np.random.seed(10)
sp.random.seed(10)
net = neuralNet.createNet()
train_ds, test_ds = readData.get_data()
train_model(net, train_ds, test_ds)
mStatesSave = []
fbProbsSave = []
for i, fileLists in enumerate(fileNames):
for j, fList in enumerate(fileLists):
dists = []
fbProbsList = []
for f in fList:
print i, j
print f
fbProbs, numSteps, model, statesInRbt, states, logProbs_T, logProbs_O, logProbs, poses, actions, obs, \
actionType, actionSelectionType, numMechanismTypes, numParticles, numRepeats, neff_fract, \
modelNums, realStates, BIAS, FTSD, FOSD, RTSD, ROSD \
= readData.get_data(f) # used to be dirName+f
# save fbProbs
fbProbsList.append(fbProbs)
# save real
#rStatesSave.append(realStates[1:])
# find max state at each step
#mStatesSave.append([])
#for k,(Ss,lPs) in enumerate(zip(states[1:],logProbs[1:])):
#maxProbs = []
#maxProbInds = []
#for l in relevantMT:
#maxProbs.append(max(lPs[l]))
#maxProbInds.append(lPs[l].index(maxProbs[-1]))
import readData
import sys
f = sys.argv[1]
fbProbs, numSteps, model, statesInRbt, states, logProbs_T, \
logProbs_O, logProbs, poses, actions, obs, \
actionType, actionSelectionType, numMechanismTypes, \
numParticles, numRepeats, neff_fract, \
modelNums, realStates, BIAS, FTSD, FOSD, RTSD, ROSD \
= readData.get_data(f)
print fbProbs
for j in range(numMT):
fileNames[i][j] = files[numMT * numT * i + numT * j : numMT * numT * i + numT * j + numT]
rStatesSave = []
mStatesSave = []
for i, fileLists in enumerate(fileNames):
for j, fList in enumerate(fileLists):
dists = []
for f in fList:
print i, j
print f
fbProbs, numSteps, model, statesInRbt, states, logProbs_T, logProbs_O, logProbs, poses, actions, obs, actionType, actionSelectionType, numMechanismTypes, numParticles, numRepeats, neff_fract, modelNums, realStates, BIAS, FTSD, FOSD, RTSD, ROSD = readData.get_data(
dirName + f
)
# find max state
maxProbs = []
maxProbInds = []
for l in relevantMT:
maxProbs.append(max(logProbs[-1][l]))
maxProbInds.append(logProbs[-1][l].index(maxProbs[-1]))
maxModelInd = maxProbs.index(max(maxProbs))
maxProbInd = maxProbInds[maxModelInd]
maxState = states[-1][relevantMT[maxModelInd]][maxProbInd]
print realStates[-1]
print maxState
rStatesSave = []
mStatesSave = []
for i,fileLists in enumerate(fileNames):
for j,fList in enumerate(fileLists):
dists = []
for f in fList:
print i,j
print f
fbProbs, numSteps, model, statesInRbt, states, logProbs_T, logProbs_O, logProbs, poses, actions, obs, \
actionType, actionSelectionType, numMechanismTypes, numParticles, numRepeats, neff_fract, \
modelNums, realStates, BIAS, FTSD, FOSD, RTSD, ROSD \
= readData.get_data(f) # used to be dirName+f
# save real
rStatesSave.append(realStates[1:])
# find max state at each step
mStatesSave.append([])
for k,(Ss,lPs) in enumerate(zip(states[1:],logProbs[1:])):
maxProbs = []
maxProbInds = []
for l in relevantMT:
maxProbs.append(max(lPs[l]))
maxProbInds.append(lPs[l].index(maxProbs[-1]))
maxModelInd = maxProbs.index(max(maxProbs))
maxProbInd = maxProbInds[maxModelInd]
import readData
import sys
f = sys.argv[1]
fbProbs, numSteps, model, statesInRbt, states, logProbs_T, \
logProbs_O, logProbs, poses, actions, obs, \
actionType, actionSelectionType, numMechanismTypes, \
numParticles, numRepeats, neff_fract, \
modelNums, realStates, BIAS, FTSD, FOSD, RTSD, ROSD \
= readData.get_data(f)
print fbProbs
def runNet():
np.random.seed(10)
sp.random.seed(10)
net = neuralNet.createNet()
train_ds, test_ds = readData.get_data()
train_model(net, train_ds, test_ds)