import torch
import torch_optimizer as optim
from typing import List, Tuple, Dict, Union, Any
from finol.utils import load_config
[docs]class OptimizerSelector:
"""
Class to select optimizer for portfolio selection.
"""
def __init__(self, model) -> None:
self.config = load_config()
self.model = model
self.optimizer_dict = {
"Adadelta": torch.optim.Adadelta,
"Adagrad": torch.optim.Adagrad,
"Adam": torch.optim.Adam,
"AdamW": torch.optim.AdamW,
"SparseAdam": torch.optim.SparseAdam,
"Adamax": torch.optim.Adamax,
"ASGD": torch.optim.ASGD,
"SGD": torch.optim.SGD,
"RAdam": torch.optim.RAdam,
"Rprop": torch.optim.Rprop,
"RMSprop": torch.optim.RMSprop,
"NAdam": torch.optim.NAdam,
"LBFGS": torch.optim.LBFGS,
"A2GradExp": optim.A2GradExp,
"A2GradInc": optim.A2GradInc,
"A2GradUni": optim.A2GradUni,
"AccSGD": optim.AccSGD,
"AdaBelief": optim.AdaBelief,
"AdaBound": optim.AdaBound,
"AdaMod": optim.AdaMod,
"Adafactor": optim.Adafactor,
"Adahessian": optim.Adahessian,
"AdamP": optim.AdamP,
"AggMo": optim.AggMo,
"Apollo": optim.Apollo,
"DiffGrad": optim.DiffGrad,
"LARS": optim.LARS,
"Lamb": optim.Lamb,
"MADGRAD": optim.MADGRAD,
"NovoGrad": optim.NovoGrad,
"PID": optim.PID,
"QHAdam": optim.QHAdam,
"QHM": optim.QHM,
"Ranger": optim.Ranger,
"RangerQH": optim.RangerQH,
"RangerVA": optim.RangerVA,
"SGDP": optim.SGDP,
"SGDW": optim.SGDW,
"SWATS": optim.SWATS,
"Shampoo": optim.Shampoo,
"Yogi": optim.Yogi
}
[docs] def select_optimizer(self, sampled_lr: Union[Dict[str, Any], None] = None, sampled_optimizer: Union[Dict[str, Any], None] = None) -> object:
"""
Selects an optimizer based on the configuration.
:param sampled_lr: Sampled learning rate or None. Required only when optimizing the learning rate selection.
:param sampled_optimizer: Sampled optimizer or None. Required only when optimizing the optimizer selection.
:return: The selected optimizer for the model.
:raises ValueError: If the optimizer name is invalid.
"""
if sampled_lr!= None:
lr = sampled_lr
else:
lr = self.config["LEARNING_RATE"]
if sampled_optimizer!= None:
optimizer_name = sampled_optimizer
else:
optimizer_name = self.config["OPTIMIZER_NAME"]
optimizer_cls = self.optimizer_dict.get(optimizer_name, None)
if optimizer_cls is None:
raise ValueError(f"Invalid optimizer name: {optimizer_name}. Supported optimizers are: {self.optimizer_dict}")
return optimizer_cls(self.model.parameters(), lr=lr)
