Garage

Latest version: v2021.3.0

Safety actively analyzes 682387 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 1 of 2

2020.06.3

- Fixed
* PyTorch 1.7 support ([1934](https://github.com/rlworkgroup/garage/pull/1934))
* `LocalRunner` ignores `worker_cls` attribute of algorithms ([1984](https://github.com/rlworkgroup/garage/pull/1984))
* `mujoco_py` versions greater than v2.0.2.8 are incompatible with some GCC versions in conda ([2000](https://github.com/rlworkgroup/garage/pull/2000))
* MTSAC not learning because it corrupts the termination signal by wrapping with `GarageEnv` twice ([2029](https://github.com/rlworkgroup/garage/pull/2029))
* MTSAC does not respect `max_episode_length_eval` hyperparameter ([2029](https://github.com/rlworkgroup/garage/pull/2029))
* MTSAC MetaWorld examples do not use the correct number of tasks ([2029](https://github.com/rlworkgroup/garage/pull/2029))
* MTSAC now supports a separate `max_episode_length` for evalaution via the `max_episode_length_eval` hyperparameter ([2029](https://github.com/rlworkgroup/garage/pull/2029))
* MTSAC MetaWorld MT50 example used an incorrect `max_episode_length` ([2029](https://github.com/rlworkgroup/garage/pull/2029))

2020.06.2

- Fixed
* Better parameters for example `her_ddpg_fetchreach` ([1763](https://github.com/rlworkgroup/garage/pull/1763))
* Ensure determinism in TensorFlow by using `tfp.SeedStream` ([1821](https://github.com/rlworkgroup/garage/pull/1821))
* Broken rendering of MuJoCo environments to pixels in the NVIDIA Docker container ([1838](https://github.com/rlworkgroup/garage/pull/1838))
* Enable cudnn in the NVIDIA Docker container ([1840](https://github.com/rlworkgroup/garage/pull/1840))
* Bug in `DiscreteQfDerivedPolicy` in which parameters were not returned ([1847](https://github.com/rlworkgroup/garage/pull/1847))
* Populate `TimeLimit.truncated` at every step when using `gym.Env` ([1852](https://github.com/rlworkgroup/garage/pull/1852))
* Bug in which parameters where not copied when TensorFlow primitives are `clone()`ed ([1855](https://github.com/rlworkgroup/garage/pull/1855))
* Typo in the `Makefile` target `run-nvidia` ([1914](https://github.com/rlworkgroup/garage/pull/1914))

2020.06.1

- Fixed
* Pipenv fails to resolve a stable dependency set because of excessively-narrow dependencies in tensorflow-probability ([1721](https://github.com/rlworkgroup/garage/pull/1721))
* Bug which prevented `rollout` from running policies deterministically ([1714](https://github.com/rlworkgroup/garage/pull/1714))

2020.06.0

Added
- Algorithms
* PPO in PyTorch (
[905](https://github.com/rlworkgroup/garage/pull/905),
[1188](https://github.com/rlworkgroup/garage/pull/1188))
* TRPO in PyTorch (
[1018](https://github.com/rlworkgroup/garage/pull/1018),
[1053](https://github.com/rlworkgroup/garage/pull/1053),
[1186](https://github.com/rlworkgroup/garage/pull/1186))
* MAML in PyTorch (
[1128](https://github.com/rlworkgroup/garage/pull/1128),
[1136](https://github.com/rlworkgroup/garage/pull/1136),
[1214](https://github.com/rlworkgroup/garage/pull/1214),
[1234](https://github.com/rlworkgroup/garage/pull/1234),
[1283](https://github.com/rlworkgroup/garage/pull/1283))
* RL2 in TensorFlow (
[1127](https://github.com/rlworkgroup/garage/pull/1127),
[1175](https://github.com/rlworkgroup/garage/pull/1175),
[1189](https://github.com/rlworkgroup/garage/pull/1189),
[1190](https://github.com/rlworkgroup/garage/pull/1190),
[1195](https://github.com/rlworkgroup/garage/pull/1195),
[1231](https://github.com/rlworkgroup/garage/pull/1231))
* PEARL in PyTorch (
[1059](https://github.com/rlworkgroup/garage/pull/1059),
[1124](https://github.com/rlworkgroup/garage/pull/1124),
[1218](https://github.com/rlworkgroup/garage/pull/1218),
[1316](https://github.com/rlworkgroup/garage/pull/1316),
[1374](https://github.com/rlworkgroup/garage/pull/1374))
* SAC in PyTorch ([1235](https://github.com/rlworkgroup/garage/pull/1235))
* MTSAC in PyTorch ([1332](https://github.com/rlworkgroup/garage/pull/1332))
* Task Embeddings in TensorFlow (
[1168](https://github.com/rlworkgroup/garage/pull/1168),
[1169](https://github.com/rlworkgroup/garage/pull/1169),
[1167](https://github.com/rlworkgroup/garage/pull/1167))
- Samplers
* New Sampler API, with efficient multi-env and multi-policy support (
[881](https://github.com/rlworkgroup/garage/pull/881),
[1153](https://github.com/rlworkgroup/garage/pull/1153),
[1319](https://github.com/rlworkgroup/garage/pull/1319))
* `garage.sampler.LocalSampler`, which uses the main process to sample (
[1133](https://github.com/rlworkgroup/garage/pull/1133),
[1156](https://github.com/rlworkgroup/garage/pull/1156))
* Reworked `garage.sampler.RaySampler` to use new API (
[1133](https://github.com/rlworkgroup/garage/pull/1133),
[1134](https://github.com/rlworkgroup/garage/pull/1134))
* `garage.sampler.MultiprocessingSampler` ([1298](https://github.com/rlworkgroup/garage/pull/1298))
* `garage.sampler.VecWorker`, a replacement for VecEnvExecutor ([1311](https://github.com/rlworkgroup/garage/pull/1311))
- APIs
* `garage.TrajectoryBatch` data type (
[1058](https://github.com/rlworkgroup/garage/pull/1058),
[1065](https://github.com/rlworkgroup/garage/pull/1065),
[1132](https://github.com/rlworkgroup/garage/pull/1132),
[1154](https://github.com/rlworkgroup/garage/pull/1154))
* `garage.TimeStep` data type (
[1114](https://github.com/rlworkgroup/garage/pull/1114),
[1221](https://github.com/rlworkgroup/garage/pull/1221))
* `garage.TimeStepBatch` data type ([1529](https://github.com/rlworkgroup/garage/pull/1529))
* `garage.log_performance` (
[1116](https://github.com/rlworkgroup/garage/pull/1116),
[1142](https://github.com/rlworkgroup/garage/pull/1142),
[1159](https://github.com/rlworkgroup/garage/pull/1159))
* `garage.np.algos.MetaRLAlgorithm` ([1142](https://github.com/rlworkgroup/garage/pull/1142))
* `garage.experiment.MetaEvaluator` (
[1142](https://github.com/rlworkgroup/garage/pull/1142),
[1152](https://github.com/rlworkgroup/garage/pull/1152),
[1227](https://github.com/rlworkgroup/garage/pull/1227))
* `garage.log_multitask_performance` ([1192](https://github.com/rlworkgroup/garage/pull/1192))
* `garage.torch.distributions.TanhNormal` ([1140](https://github.com/rlworkgroup/garage/pull/1140))
* `garage.torch.policies.TanhGaussianMLPPolicy` ([1176](https://github.com/rlworkgroup/garage/pull/1176))
* `garage.experiment.wrap_experiment` to replace `run_experiment` with several new features (
[1100](https://github.com/rlworkgroup/garage/pull/1100),
[1155](https://github.com/rlworkgroup/garage/pull/1155),
[1160](https://github.com/rlworkgroup/garage/pull/1160),
[1164](https://github.com/rlworkgroup/garage/pull/1164),
[1249](https://github.com/rlworkgroup/garage/pull/1249),
[1258](https://github.com/rlworkgroup/garage/pull/1258),
[1281](https://github.com/rlworkgroup/garage/pull/1281),
[1396](https://github.com/rlworkgroup/garage/pull/1396),
[1482](https://github.com/rlworkgroup/garage/pull/1482))
* `garage.torch.q_functions.ContinuousCNNQFunction` ([1326](https://github.com/rlworkgroup/garage/pull/1326))
* PyTorch support for non-linearities with parameters ([928](https://github.com/rlworkgroup/garage/pull/928),
* `garage.torch.value_function.GaussianMLPValueFunction` (
[1317](https://github.com/rlworkgroup/garage/pull/1317))
* Simpler PyTorch policy API ([1528](https://github.com/rlworkgroup/garage/pull/1528))
* `garage.envs.TaskOnehotWrapper` ([1157](https://github.com/rlworkgroup/garage/pull/1157))
- HalfCheetah meta environments (
[1108](https://github.com/rlworkgroup/garage/pull/1108),
[1131](https://github.com/rlworkgroup/garage/pull/1131),
[1216](https://github.com/rlworkgroup/garage/pull/1216),
[1385](https://github.com/rlworkgroup/garage/pull/1385))
- PyTorch GPU support ([1182](https://github.com/rlworkgroup/garage/pull/1182))
- PyTorch deterministic support ([1063](https://github.com/rlworkgroup/garage/pull/1063))
- Support for running Meta-RL algorithms on MetaWorld benchmarks (
[1306](https://github.com/rlworkgroup/garage/pull/1306))
- Examples for running MetaWorld benchmarks (
[1010](https://github.com/rlworkgroup/garage/pull/1010),
[1263](https://github.com/rlworkgroup/garage/pull/1263),
[1265](https://github.com/rlworkgroup/garage/pull/1265),
[1265](https://github.com/rlworkgroup/garage/pull/1265),
[1241](https://github.com/rlworkgroup/garage/pull/1241),
[1232](https://github.com/rlworkgroup/garage/pull/1232),
[1327](https://github.com/rlworkgroup/garage/pull/1327),
[1351](https://github.com/rlworkgroup/garage/pull/1351),
[1393](https://github.com/rlworkgroup/garage/pull/1393))
- Improved off-policy evaluation (
[1139](https://github.com/rlworkgroup/garage/pull/1139),
[1279](https://github.com/rlworkgroup/garage/pull/1279),
[1331](https://github.com/rlworkgroup/garage/pull/1331),
[1384](https://github.com/rlworkgroup/garage/pull/1384))

Changed
- Allow TensorFlow 2 (or TF >=1.14) (
[1309](https://github.com/rlworkgroup/garage/pull/1309),
[1563](https://github.com/rlworkgroup/garage/pull/1563))
- Require Torch 1.4.0 (
[1335](https://github.com/rlworkgroup/garage/pull/1335),
[1361](https://github.com/rlworkgroup/garage/pull/1361))
- Ensure TF and torch are optional ([1510](https://github.com/rlworkgroup/garage/pull/1510))
- Update gym to 0.15.4 (
[1098](https://github.com/rlworkgroup/garage/pull/1098),
[1158](https://github.com/rlworkgroup/garage/pull/1158))
- Rename `baseline` to `value_function` ([1275](https://github.com/rlworkgroup/garage/pull/1275))
- Make `runner._sampler` optional ([1394](https://github.com/rlworkgroup/garage/pull/1394))
- Make ExplorationStrategies a type of Policy ([1397](https://github.com/rlworkgroup/garage/pull/1397))
- Use `garage.replay_buffer.PathBuffer` in off-policy algos (
[1173](https://github.com/rlworkgroup/garage/pull/1173),
[1433](https://github.com/rlworkgroup/garage/pull/1433))
- Deprecated `run_experiment` (
[1370](https://github.com/rlworkgroup/garage/pull/1370),
[1412](https://github.com/rlworkgroup/garage/pull/1412))
- Deprecated old-style samplers ([1369](https://github.com/rlworkgroup/garage/pull/1369))
- Refactor TensorFlow to use tfp.distribution (
[1073](https://github.com/rlworkgroup/garage/pull/1073),
[1356](https://github.com/rlworkgroup/garage/pull/1356),
[1357](https://github.com/rlworkgroup/garage/pull/1357),
[1410](https://github.com/rlworkgroup/garage/pull/1410),
[1456](https://github.com/rlworkgroup/garage/pull/1456),
[1444](https://github.com/rlworkgroup/garage/pull/1444),
[1554](https://github.com/rlworkgroup/garage/pull/1554),
[1569](https://github.com/rlworkgroup/garage/pull/1569))
- Set TotalEnvSteps as the default Tensorboard x-axis (
[1017](https://github.com/rlworkgroup/garage/pull/1017),
[1069](https://github.com/rlworkgroup/garage/pull/1069))
- Update dependencies for docs ([1383](https://github.com/rlworkgroup/garage/pull/1383))
- New optimizer_args TensorFlow interface ([1496](https://github.com/rlworkgroup/garage/pull/1496))
- Move LocalTFRunner to garage.experiment ([1513](https://github.com/rlworkgroup/garage/pull/1513))
- Implement HER using PathBuffer (
[1282](https://github.com/rlworkgroup/garage/pull/1282)
[1505](https://github.com/rlworkgroup/garage/pull/1505))
- Change CNN API to use tuples for defining kernels ([1515](https://github.com/rlworkgroup/garage/pull/1515))
- Many documentation improvements (
[1056](https://github.com/rlworkgroup/garage/pull/1056),
[1065](https://github.com/rlworkgroup/garage/pull/1065),
[1120](https://github.com/rlworkgroup/garage/pull/1120),
[1266](https://github.com/rlworkgroup/garage/pull/1266),
[1327](https://github.com/rlworkgroup/garage/pull/1327),
[1413](https://github.com/rlworkgroup/garage/pull/1413),
[1429](https://github.com/rlworkgroup/garage/pull/1429),
[1451](https://github.com/rlworkgroup/garage/pull/1451),
[1481](https://github.com/rlworkgroup/garage/pull/1481),
[1484](https://github.com/rlworkgroup/garage/pull/1484))
- Eliminate use of "base" module name ([1403](https://github.com/rlworkgroup/garage/pull/1403))
- Significant improvements to benchmarking (
[1271](https://github.com/rlworkgroup/garage/pull/1271)
[1291](https://github.com/rlworkgroup/garage/pull/1291),
[1306](https://github.com/rlworkgroup/garage/pull/1306),
[1307](https://github.com/rlworkgroup/garage/pull/1307),
[1310](https://github.com/rlworkgroup/garage/pull/1310),
[1320](https://github.com/rlworkgroup/garage/pull/1320),
[1368](https://github.com/rlworkgroup/garage/pull/1368),
[1380](https://github.com/rlworkgroup/garage/pull/1380),
[1409](https://github.com/rlworkgroup/garage/pull/1409))
- Refactor benchmarks into a separate module (
[1395](https://github.com/rlworkgroup/garage/pull/1395),
[1402](https://github.com/rlworkgroup/garage/pull/1402),
[1400](https://github.com/rlworkgroup/garage/pull/1400),
[1411](https://github.com/rlworkgroup/garage/pull/1411),
[1408](https://github.com/rlworkgroup/garage/pull/1408),
[1416](https://github.com/rlworkgroup/garage/pull/1416),
[1414](https://github.com/rlworkgroup/garage/pull/1414),
[1432](https://github.com/rlworkgroup/garage/pull/1432))

Removed
- Dependencies:
* matplotlib (moved to dev) ([1083](https://github.com/rlworkgroup/garage/pull/1083))
* atari-py ([1194](https://github.com/rlworkgroup/garage/pull/1194))
* gtimer, pandas, rlkit, seaborn (moved to benchmarks) ([1325](https://github.com/rlworkgroup/garage/pull/1325))
* pyprind ([1495](https://github.com/rlworkgroup/garage/pull/1495))
- `RLAlgorithm.get_itr_snapshot` ([1054](https://github.com/rlworkgroup/garage/pull/1054))
- `garage.misc.nb_utils` ([1288](https://github.com/rlworkgroup/garage/pull/1288))
- `garage.np.regressors` ([1493](https://github.com/rlworkgroup/garage/pull/1493))
- `garage.np.BatchPolopt` (
[1486](https://github.com/rlworkgroup/garage/pull/1486),
[1492](https://github.com/rlworkgroup/garage/pull/1492))
- `garage.misc.prog_bar_counter` ([1495](https://github.com/rlworkgroup/garage/pull/1495))
- `garage.tf.envs.TfEnv` ([1443](https://github.com/rlworkgroup/garage/pull/1443))
- `garage.tf.BatchPolopt` ([1504](https://github.com/rlworkgroup/garage/pull/1504))
- `garage.np.OffPolicyRLAlgorithm` ([1552](https://github.com/rlworkgroup/garage/pull/1552))

Fixed
- Bug where `GymEnv` did not pickle ([1029](https://github.com/rlworkgroup/garage/pull/1029))
- Bug where `VecEnvExecutor` conflated terminal state and time limit signal (
[1178](https://github.com/rlworkgroup/garage/pull/1178),
[1570](https://github.com/rlworkgroup/garage/pull/1570))
- Bug where plotter window was opened multiple times ([1253](https://github.com/rlworkgroup/garage/pull/1253))
- Bug where TF plotter used main policy on separate thread ([1270](https://github.com/rlworkgroup/garage/pull/1270))
- Workaround gym timelimit and terminal state conflation ([1118](https://github.com/rlworkgroup/garage/pull/1118))
- Bug where pixels weren't normalized correctly when using CNNs (
[1236](https://github.com/rlworkgroup/garage/pull/1236),
[1419](https://github.com/rlworkgroup/garage/pull/1419))
- Bug where `garage.envs.PointEnv` did not step correctly ([1165](https://github.com/rlworkgroup/garage/pull/1165))
- Bug where sampler workers crashed in non-Deterministic mode ([1567](https://github.com/rlworkgroup/garage/pull/1567))
- Use cloudpickle in old-style samplers to handle lambdas ([1371](https://github.com/rlworkgroup/garage/pull/1371))
- Bug where workers where not shut down after running a resumed algorithm ([1293](https://github.com/rlworkgroup/garage/pull/1293))
- Non-PyPI dependencies, which blocked using pipenv and poetry ([1247](https://github.com/rlworkgroup/garage/pull/1247))
- Bug where TensorFlow paramter setting didn't work across differently named policies ([1355](https://github.com/rlworkgroup/garage/pull/1355))
- Bug where advantages where computed incorrectly in PyTorch ([1197](https://github.com/rlworkgroup/garage/pull/1197))
- Bug where TF plotter was used in LocalRunner ([1267](https://github.com/rlworkgroup/garage/pull/1267))
- Worker processes are no longer started unnecessarily ([1006](https://github.com/rlworkgroup/garage/pull/1006))
- All examples where fixed and are now tested ([1009](https://github.com/rlworkgroup/garage/pull/1009))

2019.10.3

Fixed
- Better parameters for example `her_ddpg_fetchreach` ([1764](https://github.com/rlworkgroup/garage/pull/1764))
- Bug in `DiscreteQfDerivedPolicy` in which parameters were not returned ([1847](https://github.com/rlworkgroup/garage/pull/1847))
- Bug which made it impossible to evaluate stochastic policies deterministically ([1715](https://github.com/rlworkgroup/garage/pull/1715))

2019.10.2

Fixed
- Use a GitHub Token in the CI to retrieve packages to avoid hitting GitHub API rate limit ([1250](https://github.com/rlworkgroup/garage/pull/1250))
- Avoid installing dev extra dependencies during the conda check ([1296](https://github.com/rlworkgroup/garage/pull/1296))
- Install `dm_control` from PyPI ([1406](https://github.com/rlworkgroup/garage/pull/1406))
- Pin tfp to 0.8.x to avoid breaking pipenv ([1480](https://github.com/rlworkgroup/garage/pull/1480))
- Force python 3.5 in CI ([1522](https://github.com/rlworkgroup/garage/pull/1522))
- Separate terminal and completion signal in vectorized sampler ([1581](https://github.com/rlworkgroup/garage/pull/1581))
- Disable certicate check for roboti.us ([1595](https://github.com/rlworkgroup/garage/pull/1595))
- Fix `advantages` shape in `compute_advantage()` in torch tree ([1209](https://github.com/rlworkgroup/garage/pull/1209))
- Fix plotting using tf.plotter ([1292](https://github.com/rlworkgroup/garage/pull/1292))
- Fix duplicate window rendering when using garage.Plotter ([1299](https://github.com/rlworkgroup/garage/pull/1299))
- Fix setting garage.model parameters ([1363](https://github.com/rlworkgroup/garage/pull/1363))
- Fix two example jupyter notebook ([1584](https://github.com/rlworkgroup/garage/pull/1584))
- Fix collecting samples in `RaySampler` ([1583](https://github.com/rlworkgroup/garage/pull/1583))

Page 1 of 2

© 2024 Safety CLI Cybersecurity Inc. All Rights Reserved.