input_shape=input_shape)(both_input)
encode_left = both_feature.index_select(1, 0)
encode_right = both_feature.index_select(1, 1)
distance = autograd.abs(encode_left - encode_right)
predict = Dense(output_dim=NUM_CLASS_LABEL,
activation="sigmoid",
W_regularizer=L2Regularizer(args.penalty_rate))(distance)
siamese_net = Model(input=both_input, output=predict)
# 声明优化器, 训练并测试模型.
optimizer = Optimizer(model=siamese_net,
training_rdd=train_rdd,
optim_method=Adam(args.learning_rate),
criterion=CrossEntropyCriterion(),
end_trigger=MaxEpoch(args.num_epoch),
batch_size=args.batch_size)
optimizer.set_validation(batch_size=args.batch_size,
val_rdd=test_rdd,
trigger=EveryEpoch(),
val_method=[Top1Accuracy()])
# 设置训练日志, 可用 TensorBoard 查询.
app_name = "logs"
optimizer.set_train_summary(TrainSummary(log_dir=".", app_name=app_name))
optimizer.set_val_summary(ValidationSummary(log_dir=".", app_name=app_name))
optimizer.optimize()
batch_size=args.batch_size)
optimizer.set_validation(batch_size=args.batch_size,
val_rdd=test_rdd,
trigger=EveryEpoch(),
val_method=[Top1Accuracy()])
# Create logs.
app_name = "logs"
optimizer.set_train_summary(TrainSummary(log_dir=".", app_name=app_name))
optimizer.set_val_summary(ValidationSummary(log_dir=".", app_name=app_name))
# Call the optimizer to start training the model.
print('\n\nModel training started!')
print('\n\nPipeline: Intel Analytics Zoo')
print('Starting to train the model on Intel BigDL')
print('Paramaters: Shared\n\n')
pokemon_model = optimizer.optimize()
print('\n\nModel training finished!\n\n')
# Make the predictions.
predictions = pokemon_model.predict(test_rdd).collect()
print('\n\nThe predictions are\n')
print('-------------------------------------------------\n')
print('Encoding - 1\tEncoding - 2\tInference?\n')
print('-------------------------------------------------\n\n')
for i in predictions:
if (abs(i[0] - i[1]) < 0.1):
print(i[0], '\t', i[1], '\tSame Pokemon')
else:
print(i[0], '\t', i[1], '\tDifferent Pokemon')
