2), # for train and validation metric
dtype=np.float)
weight_tensor = np.zeros(
(len(BATCH_LIST), NUM_FOLDS, EPOCH_QUANTIZER,
NUM_FEATURES + 1), # w1..wN, w0; N = 53
dtype=np.float)
time_tensor = np.zeros(
len(BATCH_LIST),
dtype=np.float) # vector of time values (for each batch size)
sess = Session()
# check different batch sizes
for batch_size_counter, batch_size in enumerate(BATCH_LIST, start=0):
print('=== Current batch size: %d ===' % batch_size)
metrics_tensor[batch_size_counter], weight_tensor[
batch_size_counter], time_tensor[
batch_size_counter] = sess.crossValidation(
linear_regressor, dataset, labels, NUM_EPOCHS,
EPOCH_QUANTIZER, batch_size, NUM_FOLDS, LR)
################
## Write data ##
np.save('../TrainData/metrics.npy', metrics_tensor)
np.save('../TrainData/weights.npy', weight_tensor)
np.save('../TrainData/time.npy', time_tensor)
## Train the model ##
metrics_tensor = np.zeros((len(BATCH_LIST), # write at each batch
NUM_FOLDS, # at each fold
EPOCH_QUANTIZER, # ...at each self._epoch_quantize_param's epoch
1, # for all the metrics to write
2), # for train and validation metric
dtype=np.float)
time_tensor = np.zeros(len(BATCH_LIST), dtype=np.float) # vector of time values (for each batch size)
sess = Session()
# check different batch sizes
for batch_size_counter, batch_size in enumerate(BATCH_LIST, start=0):
print('=== Current batch size: %d ===' % batch_size)
metrics_tensor[batch_size_counter], time_tensor[batch_size_counter] = sess.crossValidation(factor_machine,
dataset, labels,
NUM_FEATURES, NUM_SAMPLES,
NUM_EPOCHS, EPOCH_QUANTIZER,
batch_size,
NUM_FOLDS,
LR)
################
## Write data ##
np.save('../TrainData/{}/metrics.npy'.format(DATASET), metrics_tensor)
np.save('../TrainData/{}/time.npy'.format(DATASET), time_tensor)
