32 lines
1.1 KiB
Python
32 lines
1.1 KiB
Python
#######################################################
|
|
# Copyright (c) Xuanyi Dong [GitHub D-X-Y], 2021.04 #
|
|
#######################################################
|
|
# Use module in xlayers to construct different models #
|
|
#######################################################
|
|
from typing import List, Text, Dict, Any
|
|
import torch
|
|
|
|
__all__ = ["get_model"]
|
|
|
|
|
|
from xlayers.super_core import SuperSequential
|
|
from xlayers.super_core import SuperSimpleNorm
|
|
from xlayers.super_core import SuperLeakyReLU
|
|
from xlayers.super_core import SuperLinear
|
|
|
|
|
|
def get_model(config: Dict[Text, Any], **kwargs):
|
|
model_type = config.get("model_type", "simple_mlp")
|
|
if model_type == "simple_mlp":
|
|
model = SuperSequential(
|
|
SuperSimpleNorm(kwargs["mean"], kwargs["std"]),
|
|
SuperLinear(kwargs["input_dim"], 200),
|
|
SuperLeakyReLU(),
|
|
SuperLinear(200, 100),
|
|
SuperLeakyReLU(),
|
|
SuperLinear(100, kwargs["output_dim"]),
|
|
)
|
|
else:
|
|
raise TypeError("Unkonwn model type: {:}".format(model_type))
|
|
return model
|