facebookresearch / facebookresearch/ReAgent
Quickstart Examples Not Working
- Dominant language
- Python
- Stars
- 3.7k
- Forks
- 529
- PR merge metrics
- No merged PRs in 30d
Description
The following commands result in an `AttributeError: module 'reagent.gym.tests.test_gym' has no attribute 'run_test’` because the module `test_gym` does not exist.
```
export CONFIG=reagent/gym/tests/configs/cartpole/discrete_dqn_cartpole_online.yaml
./reagent/workflow/cli.py run reagent.gym.tests.test_gym.run_test $CONFIG
```
See full stacktrace [here](https://gist.github.com/avilay/ed728178aa47d3f9576c8fe08d0a2973).
Even when I try to run one of the existing modules like `run_test_online_episode` I get a `KeyError: 'state_features'...float_features=batch[InputColumn.STATE_FEATURES]`
Here is the full command:
```
./reagent/workflow/cli.py run reagent.gym.tests.test_gym.run_test_online_episode $CONFIG
```
See full stacktrace [here](https://gist.github.com/avilay/0c0ff1ddb045fe0e12d364dbdd3b2c3c)
On further trying the batch RL example the first command running the `offline_gym` module works fine. However, the second command running the `timeline_operator` errors out with `java.lang.StringIndexOutOfBoundsException: begin 0, end 3, length 2`. Here are the full commands that I ran:
```
export CONFIG=reagent/workflow/sample_configs/discrete_dqn_cartpole_offline.yaml
./reagent/workflow/cli.py run reagent.workflow.gym_batch_rl.offline_gym $CONFIG
./reagent/workflow/cli.py run reagent.workflow.gym_batch_rl.timeline_operator $CONFIG
```
See full stacktrace [here](https://gist.github.com/avilay/bd793bc75c970b1f572fccce3c330da8)
Here are my environment details:
* Hardware: Apple M1
* OS: Big Sur 11.2
* Anaconda Python 3.7
* JRE Details
* openjdk 16 2021-03-16
* OpenJDK Runtime Environment Zulu16.28+11-CA (build 16+36)
* OpenJDK 64-Bit Server VM Zulu16.28+11-CA (build 16+36, mixed mode)
* ReAgent master branch cloned on 4/23.
* I had to install a later version of OpenAI gym (0.18.0) because of compatibility issues with M1.
Contributor guide
Assessment
This issue has not been assessed yet.