![]() ![]() Updated Hugging Face Integration page env to vec_env when environment is vectorized Modified get_system_info to avoid issue linked to copy-pasting on GitHub issue Standardized the use of from gym import spaces Monkey-patched np.bool = bool so gym 0.21 is compatible with NumPy 1.24+ Set tensors construction directly on the device (~8% speed boost on GPU) Replaced CartPole-v0 by CartPole-v1 is testsįixed tests/test_distributions.py type hintsįixed stable_baselines3/common/type_aliases.py type hintsįixed stable_baselines3/common/torch_layers.py type hintsįixed stable_baselines3/common/env_util.py type hintsįixed stable_baselines3/common/preprocessing.py type hintsįixed stable_baselines3/common/atari_wrappers.py type hintsįixed stable_baselines3/common/vec_env/vec_check_nan.py type hintsĮxposed modules in _init_.py with the _all_ attribute GitHub CI/setup-python to v4 and checkout to v3 Goal-conditioned environments are now characterized by the availability of the compute_reward method, rather than by their inheritance to gym.GoalEnv Updated the PR template to associate each PR with its peer in RL-Zoo3 and SB3-Contribįixed flake8 config to be compatible with flake8 6+ Used issue forms instead of issue templates You should now explicitely pass a features_extractor parameter when calling extract_features()ĭeprecated shared layers in MlpExtractor ¶ ![]() type annotation of model in evaluate_policyįixed the env checker, the key was not passed when checking images from Dict observation spaceįixed normalize_images which was not passed to parent class in some casesįixed load_from_vector that was broken with newer PyTorch version when passing PyTorch tensor Raise an error when the same gym environment instance is passed as separate environments when creating a vectorized environment with more than one environment. VecNormalize now updates the observation space when normalizing imagesĪdded option to have non-shared features extractor between actor and critic in on-policy algorithms with_bias argument to create_mlpĪdded support for multidimensional spaces.MultiBinary observationsįeatures extractors now properly support unnormalized image-like observations (3D tensor)Īdded normalized_image parameter to NatureCNN and CombinedExtractorįixed a bug in RecurrentPPO where the lstm states where incorrectly reshaped for n_lstm_layers > 1 (thanks RuntimeError: rnn: hx is not contiguous while predicting terminal values for RecurrentPPO when n_lstm_layers > 1Īdded support for python file for configurationįixed ProgressBarCallback under-reporting return type of evaluate_actions in ActorCritcPolicy to reflect that entropy is an optional tensor type annotation of policy in BaseAlgorithm and OffPolicyAlgorithmĪllowed model trained with Python 3.7 to be loaded with Python 3.8+ without the custom_objects workaround ![]() Removed ret attributes in VecNormalize, please use returns instead Removed deprecated sde_net_arch parameter Removed deprecated create_eval_env, eval_env, eval_log_path, n_eval_episodes and eval_freq parameters, You can find more info in issue #1233 Breaking Changes: ¶ To suppress the warning, simply save the model again. Renamed load_parameters to set_parameters documentation about subproc multiprocessing for A2C typo in A2C docstring timesteps to episodes for log_interval description note about gif creation for Atari games information about default network architecture Only use NoopResetEnv and MaxAndSkipEnv when needed in AtariWrapperĪdded support for dict/tuple observations spaces for VecCheckNan, the check is now active in the env_checker() ¶ RL Zoo ¶ Bug Fixes: ¶įixed Atari wrapper that missed the reset condition the argument dtype (default to float32) to the noise for consistency with gym action PPO train/n_updates metric not accounting for early stopping loading of normalized image-based environmentsįixed tests/test_tensorboard.py type hintįixed tests/test_vec_normalize.py type hintįixed stable_baselines3/common/monitor.py type hint You must now explicitely pass a features_extractor parameter when calling extract_features()Īdded repeat_action_probability argument in AtariWrapper. Removed shared layers in mlp_extractor StackedObservations (it now handles dict obs, StackedDictObservations was removed) Changelog ¶ Release 1.8.0a7 (WIP) ¶ Breaking Changes: ¶ ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |