diff --git a/README.md b/README.md index fc21e37..f394657 100644 --- a/README.md +++ b/README.md @@ -84,6 +84,10 @@ If importance sampling based methods are evaluating to 0, consider visualising t The different kinds of importance samples can also be visualised by querying the ```traj_is_weights``` attribute of a given ```ImportanceSampler``` object. If for example, vanilla importance sampling is being used and the samples are not ```NaN``` or ```Inf``` then visualising the ```traj_is_weights``` may provide insight. In particular, IS weights will tend to inifinity when the evaluation policy places large density on an action in comparison to the behaviour policy. ### Release log + +#### 7.0.1 +* Made the logging location for d3rlpy/FQE callback an optional parameter + #### 7.0.0 (Major API release) * Altered ISEstimator and OPEEstimatorBase APIs to depend on EmpiricalMeanDenomBase and WeightDenomBase * EmpiricalMeanDenomBase and WeightDenomBase seperatly define functions over the dataset value and weights of the individul trajectory weights, respectively. This allows a far greater number of estimators to be flexibly implemented diff --git a/src/offline_rl_ope/_version.py b/src/offline_rl_ope/_version.py index 195960c..9538eee 100644 --- a/src/offline_rl_ope/_version.py +++ b/src/offline_rl_ope/_version.py @@ -1 +1 @@ -__version__ = "7.0.0" \ No newline at end of file +__version__ = "7.0.1" \ No newline at end of file diff --git a/src/offline_rl_ope/api/d3rlpy/Callbacks/DM.py b/src/offline_rl_ope/api/d3rlpy/Callbacks/DM.py index aced7fd..21e714f 100644 --- a/src/offline_rl_ope/api/d3rlpy/Callbacks/DM.py +++ b/src/offline_rl_ope/api/d3rlpy/Callbacks/DM.py @@ -33,7 +33,8 @@ def __init__( model_fit_kwargs:Dict, dataset:MDPDataset, fqe_impl_init:Optional[FQEImplInitArg]=None, - device:DeviceArg = False + device:DeviceArg = False, + logs_loc:Optional[str] = None ) -> None: super().__init__(debug=False, debug_path="") self.__scorers = scorers @@ -41,7 +42,10 @@ def __init__( self.__fqe_cls = fqe_cls self.__model_init_kwargs = model_init_kwargs self.__model_fit_kwargs = model_fit_kwargs - self.__logs_loc = os.path.join(os.getcwd(), "tmp_fqe_logs_loc") + if logs_loc: + self.__logs_loc = logs_loc + else: + self.__logs_loc = os.path.join(os.getcwd(), "tmp_fqe_logs_loc") self.__cur_exp = 0 os.mkdir(self.__logs_loc) self.__fqe_impl_init = fqe_impl_init