-
- Downloads
Support asynchronous unroll (#1397)
* Support asynchronous unroll Perform unroll and training concurrently. This can be useful in two situations: 1. The environment step is expensive. This can help reducing time spending on unroll 2. Realtime training where the interaction with environment happens in realtime. For this, we can set TrainerConfig.unroll_step_interval to the desired interaction period. * Fix circular import * Address comments * Fix failure for build.sh Need to avoid mess up with commandline arguments * Address comments
Showing
- .ci-cd/build.sh 1 addition, 0 deletions.ci-cd/build.sh
- alf/algorithms/async_unroller.py 289 additions, 0 deletionsalf/algorithms/async_unroller.py
- alf/algorithms/async_unroller_test.py 76 additions, 0 deletionsalf/algorithms/async_unroller_test.py
- alf/algorithms/config.py 38 additions, 0 deletionsalf/algorithms/config.py
- alf/algorithms/rl_algorithm.py 123 additions, 20 deletionsalf/algorithms/rl_algorithm.py
- alf/bin/train.py 3 additions, 1 deletionalf/bin/train.py
- alf/data_structures.py 10 additions, 0 deletionsalf/data_structures.py
- alf/environments/parallel_environment.py 4 additions, 0 deletionsalf/environments/parallel_environment.py
- alf/trainers/evaluator.py 1 addition, 4 deletionsalf/trainers/evaluator.py
- alf/trainers/policy_trainer.py 1 addition, 0 deletionsalf/trainers/policy_trainer.py
- alf/utils/losses.py 2 additions, 2 deletionsalf/utils/losses.py
- alf/utils/summary_utils.py 0 additions, 1 deletionalf/utils/summary_utils.py
Loading
Please register or sign in to comment