def DiffSamp(samples):
for samples in samples:
# model = get_model();
model = get_model_activation(act=act)
opt = RMSprop(lr=0.0007, decay=1e-6)
optName = "RMSProp"
initName = "dataInit"
lr = 0.0007
epochs = 50
model.compile(loss='categorical_crossentropy',
optimizer=opt,
metrics=['accuracy'])
model = jan_initialize(model, X_train,
samples) #check the working function :)
hist = model.fit(X_train,
Y_train,
batch_size=batch_size,
epochs=epochs,
validation_data=(X_test, Y_test))
HistoryPath = os.path.join(
save_dir, optName + "_" + initName + "_Hist_" + act +
"_lr_{:1}_epoch_{:1d}_".format(lr, epochs) + model_name)
saveHistory(HistoryPath, hist)
return
def trainOptInit(X_train,
X_test,
Y_train,
Y_test,
initDict,
optDict={'RMSProp': RMSprop(lr=0.0001, decay=1e-6)},
mode='full',
saveHist=True):
if mode == 'test': #Just test over a few values
X_train = X_train[0:200]
Y_train = Y_train[0:200]
X_test = X_test[0:40]
Y_test = Y_test[0:40]
for optName, opt in optDict.items():
#---------------------------------------------- For Different Init and Optimizers paras
for key, initName in initDict.items():
print(" Training for {:s} initialization with {:s} as optimizer".
format(key, optName))
model = get_model(initName,
input_shape=X_train.shape[1:],
num_classes=num_classes)
# lr = 0.0001
# opt = RMSprop(lr=lr, decay=1e-6)
model.compile(loss='categorical_crossentropy',
optimizer=opt,
metrics=['accuracy'])
# hist = model.fit(X_train,Y_train, batch_size=batch_size, epochs=epochs, validation_data=(X_test,Y_test),shuffle=True); #hist will store everything
#earlyStop = EarlyStopping(monitor='val_acc', min_delta=0, patience=10, verbose=0, mode='max')
hist = model.fit(X_train,
Y_train,
batch_size=batch_size,
epochs=epochs,
validation_data=(X_test, Y_test),
shuffle=True,
callbacks=[earlyStop])
#With Callback
if (saveHist == True):
#---------------------------------------------- Save History & Model
model.save(
os.path.join(
save_dir, optName + initName +
"_lr_{:1}_epoch_{:1d}_".format(lr, epochs) +
model_name))
HistoryPath = os.path.join(
save_dir, optName + "_" + initName + "_Hist_" +
"lr_{:1}_epoch_{:1d}_".format(lr, epochs) + model_name)
saveHistory(HistoryPath, hist)
return 1
# To check the model if working
# X_train = X_train[0:200, :] # to reduce the data to ease the computations
# Y_train = Y_train[0:200] # comment these 2 lines while the original execution
load_model_path = os.path.join(save_dir,
str('batch_') + str(idx - 1) + model_name)
print("Model Trained on batch {:1d} is loaded".format(idx - 1))
model = load_model(load_model_path)
filepath = base_dir + "BestModels/FirstModel/" + "batch_{0:01d}_".format(
idx) + "Model - {epoch:02d}-{val_acc:.2f}.hdf5"
checkpointer = ModelCheckpoint(filepath=filepath,
verbose=1,
monitor='categorical_accuracy',
mode='max',
save_best_only=True)
callback_list = [checkpointer]
hist = model.fit(X_train,
Y_train,
batch_size=batch_size,
epochs=epochs,
validation_split=0.2,
shuffle=True,
callbacks=callback_list)
#hist will store everything
model.save(os.path.join(save_dir, str('batch_') + str(idx) + model_name))
saveHistory(os.path.join(save_dir,
str('hist_{0:01d}'.format(idx)) + model_name),
hist=hist) #Save History
model = jan_initialize(model, X_train, samples) #check the working function :)
#
#
hist = model.fit(X_train,
Y_train,
batch_size=batch_size,
epochs=epochs,
validation_data=(X_test, Y_test))
#
optName = "RMSProp"
initName = "dataInit"
HistoryPath = os.path.join(
save_dir, optName + "_" + initName + "_Hist_" +
"lr_{:1}_epoch_{:1d}_Jan15".format(lr, epochs) + str(samples) + "_Samples")
saveHistory(HistoryPath, hist)
#-------------------------------------------------------------- all standard Init with model with different activations
#only import the top and start from here ....
base_dir = "/home/ankit/Desktop/Dataset/MarResults/CIFAR10/RMSProp/eight_layers/tanh"
#save results here...
save_dir = os.path.join(base_dir, 'Mar12')
data_folder = "/home/ankit/Desktop/Dataset/CIFAR10Dataset/cifar-10-batches-py/" #Server
idx = 1
img_size = 32
batch_size = 32
num_classes = 10
model_name = 'Mar12'
mode='max',
save_best_only=True)
callback_list = [checkpointer]
# To check the model if working
# X_train = X_train[0:200, :] # to reduce the data to ease the computations
# Y_train = Y_train[0:200] # comment these 2 lines while the original execution
hist = model.fit(X_train,
Y_train,
batch_size=batch_size,
epochs=epochs,
validation_split=0.2,
shuffle=True,
callbacks=callback_list)
#hist will store everything
saveHistory(os.path.join(save_dir,
str('hist_') + history_name),
hist=hist) #Save History
# prevHist = loadHistory(os.path.join(save_dir,str('hist_') + model_name) #Load Histroy
#-------------------------------------------------------------------------------------------------- Testing
score = model.evaluate(X_train, Y_train, batch_size=32)
model.save(os.path.join(save_dir, model_name))
# model.save(os.path.join(save_dir, str('copy_') + model_name)) #save a copy of the model if you want
#--------------------------------------------------------------------------------------------------- Load Existing Model
load_model_path = os.path.join(save_dir, model_name)
model = load_model(load_model_path)