if __name__ == "__main__":
# Specify number of particles to use and number of features
nParticles = 60
#nFeatures=51
nFeatures = 47
loader = ModelLoader((nParticles, nFeatures))
model = loader.load()
utils = Utilities(nParticles)
history = Histories()
history.set_up_config(config=config)
history.on_train_begin()
# Build the first training dataset
print("TRAIN_DATA: ", TRAIN_DATA)
X_train, Y, W_train, MVA_train = utils.BuildBatch(indir=TRAIN_DATA)
for epoch in range(100):
pool_local = ThreadPool(processes=1)
# Shuffle loaded datasets and begin
inds = range(len(X_train))
np.random.shuffle(inds)
X_epoch, Y_epoch, W_epoch, MVA_epoch = X_train[inds], Y[inds], W_train[
inds], MVA_train[inds]
# Check that nothing strange happened in the loaded datset
if (np.min(W_train) == np.nan): continue
if (np.min(W_train) == np.inf): continue
##Save the validation:
history.set_mode(mode="train")
# Specify number of particles to use and number of features
nParticles = 60
#nFeatures=51
nFeatures = 47
loader = ModelLoader((nParticles, nFeatures))
## Define Loss for the model:
from Loss.Loss import multi_weighted_logloss
utils = Utilities(nParticles)
#history = Histories()
#history.set_up_config(config=config)
#history.on_train_begin()
# Build the first training dataset
print("TRAIN_DATA: ", TRAIN_DATA)
X_train, Y, W_train, MVA_train = utils.BuildBatch(indir=TRAIN_DATA,
nEvents=50,
nFiles=10)
model = loader.load_multiclass(
ouput_class=4,
loss='categorical_crossentropy') #,weights=class_weight)
for epoch in range(1000):
pool_local = ThreadPool(processes=1)
# Shuffle loaded datasets and begin
inds = range(len(X_train))
np.random.shuffle(inds)
X_epoch, Y_epoch, W_epoch, MVA_epoch = X_train[inds], Y[inds], W_train[
inds], MVA_train[inds]
# Check that nothing strange happened in the loaded datset
if (np.min(W_train) == np.nan): continue
# nFeatures=51
nFeatures = 47
Phi_sizes, F_sizes = (50, 50, 12), (50, 50, 50)
model = PFN(input_dim=nFeatures,
Phi_sizes=Phi_sizes,
F_sizes=F_sizes,
output_dim=1,
output_act='sigmoid',
loss='binary_crossentropy')
plot_model(model, to_file='deepset.png')
utils = Utilities(nParticles)
# Build the first training dataset
X_train, Y, W_train, MVA_train = utils.BuildBatch()
print(MVA_train.shape)
for epoch in range(10000):
# Shuffle loaded datasets and begin
inds = range(len(X_train))
np.random.shuffle(inds)
X_epoch, Y_epoch, W_epoch, MVA_epoch = X_train[inds], Y[inds], W_train[
inds], MVA_epoch[inds]
if (np.min(W_train) == np.nan): continue
if (np.min(W_train) == np.inf): continue
model.fit(X_epoch, Y_epoch, epochs=1, batch_size=4 * 512, verbose=1)
pd.DataFrame(X_epoch).to_csv("X_example.csv", index=False)
pd.DataFrame(Y_epoch).to_csv("Y_example.csv", index=False)
pd.DataFrame(MVA_epoch).to_csv("MVA_example.csv", index=False)
model = loader.load()
utils = Utilities(nParticles)
history = Histories()
history.set_up_config(config=config)
history.on_train_begin()
lr_decay = DecayLR(history.aucs['train'])
lr_decay.set_model(model)
log = Logger(history, TRAINING_RES + "train_" + MODEL_NAME + ".log")
# Build the first training dataset
print("TRAIN_DATA: ", TRAIN_DATA)
#print(X_train[0])
log.on_train_begin()
for epoch in range(int(epoch) + 1, 51):
X_train, Y, W_train, MVA_train = utils.BuildBatch(indir=TRAIN_DATA,
nEvents=250)
pool_local = ThreadPool(processes=1)
# Shuffle loaded datasets and begin
inds = range(len(X_train))
np.random.shuffle(inds)
X_epoch, Y_epoch, W_epoch, MVA_epoch = X_train[inds], Y[inds], W_train[
inds], MVA_train[inds]
# Check that nothing strange happened in the loaded datset
if (np.min(W_train) == np.nan): continue
if (np.min(W_train) == np.inf): continue
##Save the validation:
history.set_mode(mode="train")
