Best Practices

For details about the parameters, see GS_OPT_MODEL.

Table 1

Model Parameter	Recommended Value
template_name	rlstm
model_name	The value can be customized, for example, open_ai. The value must meet the unique constraint.
datname	Name of the database to be served, for example, postgres.
ip	IP address of the AI Engine, for example, 127.0.0.1.
port	AI Engine listening port number, for example, 5000.
max_epoch	Iteration times. A large value is recommended to ensure the convergence effect, for example, 2000.
learning_rate	(0, 1] is a floating-point number. A large learning rate is recommended to accelerate convergence.
dim_red	Number of feature values to be reduced. -1: Do not use PCA for dimension reduction. All features are supported. Floating point number in the range of (0, 1]: A smaller value indicates a smaller training dimension and a faster convergence speed, but affects the training accuracy.
hidden_units	If the feature value dimension is high, you are advised to increase the value of this parameter to increase the model complexity. For example, set this parameter to 64, 128, and so on.
batch_size	You are advised to increase the value of this parameter based on the amount of encoded data to accelerate model convergence. For example, set this parameter to 256, 512, and so on.
Other parameters	See GS_OPT_MODEL.

Recommended parameter settings:

INSERT INTO gs_opt_model values('rlstm', 'open_ai', 'postgres', '127.0.0.1', 5000, 2000, 1, -1, 64, 512, 0, false, false, '{S, T}', '{0,0}', '{0,0}', 'Text');