353bb15e90
* move vec_env * cleaning up rl_common * tests are passing (but mosts tests are deleted as moved to baselines) * add benchmark runner for smoke tests * removed duplicated algos * route references to rl_algs.a2c to baselines.a2c * route references to rl_algs.a2c to baselines.a2c * unify conftest.py * removing references to duplicated algs from codegen * removing references to duplicated algs from codegen * alex's changes to dummy_vec_env * fixed test_carpole[deepq] testcase by decreasing number of training steps... alex's changes seemed to have fixed the bug and make it train better, but at seed=0 there is a dip in the training curve at 30k steps that fails the test * codegen tests with atol=1e-6 seem to be unstable * rl_common.vec_env -> baselines.common.vec_env mass replace * fixed reference in trpo_mpi * a2c.util references * restored rl_algs.bench in sonic_prob * fix reference in ci/runtests.sh * simplifed expression in baselines/common/cmd_util * further increased rtol to 1e-3 in codegen tests * switched vecenvs to use SimpleImageViewer from gym instead of cv2 * make run.py --play option work with num_envs > 1 * make rosenbrock test reproducible * git subrepo pull (merge) baselines subrepo: subdir: "baselines" merged: "e23524a5" upstream: origin: "git@github.com:openai/baselines.git" branch: "master" commit: "bcde04e7" git-subrepo: version: "0.4.0" origin: "git@github.com:ingydotnet/git-subrepo.git" commit: "74339e8" * updated baselines README (num-timesteps --> num_timesteps) * typo in deepq/README.md
1.6 KiB
1.6 KiB
If you are curious.
Train a Cartpole agent and watch it play once it converges!
Here's a list of commands to run to quickly get a working example:
# Train model and save the results to cartpole_model.pkl
python -m baselines.run --alg=deepq --env=CartPole-v0 --save_path=./cartpole_model.pkl --num_timesteps=1e5
# Load the model saved in cartpole_model.pkl and visualize the learned policy
python -m baselines.run --alg=deepq --env=CartPole-v0 --load_path=./cartpole_model.pkl --num_timesteps=0 --play
If you wish to apply DQN to solve a problem.
Check out our simple agent trained with one stop shop deepq.learn function.
- baselines/deepq/experiments/train_cartpole.py - train a Cartpole agent.
In particular notice that once deepq.learn finishes training it returns act function which can be used to select actions in the environment. Once trained you can easily save it and load at later time. Complimentary file enjoy_cartpole.py loads and visualizes the learned policy.
If you wish to experiment with the algorithm
Check out the examples
- baselines/deepq/experiments/custom_cartpole.py - Cartpole training with more fine grained control over the internals of DQN algorithm.
- baselines/deepq/defaults.py - settings for training on atari. Run
python -m baselines.run --alg=deepq --env=PongNoFrameskip-v4
to train on Atari Pong (see more in repo-wide README.md)