367778d37f
Policy objects are now parametrised by converting the parameter dataclass instances to kwargs, using some injectable conversions along the way
26 lines
828 B
Python
26 lines
828 B
Python
from abc import ABC, abstractmethod
|
|
|
|
import numpy as np
|
|
import torch
|
|
from torch.optim.lr_scheduler import LRScheduler, LambdaLR
|
|
|
|
from tianshou.highlevel.config import RLSamplingConfig
|
|
|
|
|
|
class LRSchedulerFactory(ABC):
|
|
@abstractmethod
|
|
def create_scheduler(self, optim: torch.optim.Optimizer) -> LRScheduler:
|
|
pass
|
|
|
|
|
|
class LinearLRSchedulerFactory(LRSchedulerFactory):
|
|
def __init__(self, sampling_config: RLSamplingConfig):
|
|
self.sampling_config = sampling_config
|
|
|
|
def create_scheduler(self, optim: torch.optim.Optimizer) -> LRScheduler:
|
|
max_update_num = (
|
|
np.ceil(self.sampling_config.step_per_epoch / self.sampling_config.step_per_collect)
|
|
* self.sampling_config.num_epochs
|
|
)
|
|
return LambdaLR(optim, lr_lambda=lambda epoch: 1 - epoch / max_update_num)
|