Copyright	(c) Sentenai 2017
License	BSD3
Maintainer	sam@sentenai.com
Stability	experimental
Portability	non-portable
Safe Haskell	None
Language	Haskell2010

Environments.Gym.Internal

Description

Underlying implementation to run a Gym environment using the gym-http-client.

Synopsis

Documentation

newtype GymEnvironmentT s a t x Source #

The gym environment, which is shared through all openai/gym instances

Constructors

GymEnvironmentT
Fields getEnvironmentT :: RWST GymConfigs (DList (Event Reward s a)) (LastState s) (ClientT t) x

Instances

(MonadIO t, MonadThrow t) => MonadEnv (EnvironmentT t) State Action Reward Source #
Methods reset :: EnvironmentT t (Initial State) Source # step :: Action -> EnvironmentT t (Obs Reward State) Source #
(MonadThrow t, MonadIO t) => MonadEnv (EnvironmentT t) State Action Reward Source #
Methods reset :: EnvironmentT t (Initial State) Source # step :: Action -> EnvironmentT t (Obs Reward State) Source #
(MonadIO t, MonadThrow t) => MonadEnv (EnvironmentT t) State Action Reward Source #
Methods reset :: EnvironmentT t (Initial State) Source # step :: Action -> EnvironmentT t (Obs Reward State) Source #
(MonadThrow t, MonadIO t) => MonadEnv (EnvironmentT t) StateFL Action Reward Source #
Methods reset :: EnvironmentT t (Initial StateFL) Source # step :: Action -> EnvironmentT t (Obs Reward StateFL) Source #
Monad t => MonadWriter (DList (Event Reward s a)) (GymEnvironmentT s a t) Source #
Methods writer :: (a, DList (Event Reward s a)) -> GymEnvironmentT s a t a # tell :: DList (Event Reward s a) -> GymEnvironmentT s a t () # listen :: GymEnvironmentT s a t a -> GymEnvironmentT s a t (a, DList (Event Reward s a)) # pass :: GymEnvironmentT s a t (a, DList (Event Reward s a) -> DList (Event Reward s a)) -> GymEnvironmentT s a t a #
MonadTrans (GymEnvironmentT s a) Source #
Methods lift :: Monad m => m a -> GymEnvironmentT s a m a #
Monad t => Monad (GymEnvironmentT s a t) Source #
Methods (>>=) :: GymEnvironmentT s a t a -> (a -> GymEnvironmentT s a t b) -> GymEnvironmentT s a t b # (>>) :: GymEnvironmentT s a t a -> GymEnvironmentT s a t b -> GymEnvironmentT s a t b # return :: a -> GymEnvironmentT s a t a # fail :: String -> GymEnvironmentT s a t a #
Functor t => Functor (GymEnvironmentT s a t) Source #
Methods fmap :: (a -> b) -> GymEnvironmentT s a t a -> GymEnvironmentT s a t b # (<$) :: a -> GymEnvironmentT s a t b -> GymEnvironmentT s a t a #
Monad t => Applicative (GymEnvironmentT s a t) Source #
Methods pure :: a -> GymEnvironmentT s a t a # (<>) :: GymEnvironmentT s a t (a -> b) -> GymEnvironmentT s a t a -> GymEnvironmentT s a t b # (>) :: GymEnvironmentT s a t a -> GymEnvironmentT s a t b -> GymEnvironmentT s a t b # (<*) :: GymEnvironmentT s a t a -> GymEnvironmentT s a t b -> GymEnvironmentT s a t a #
MonadIO t => MonadIO (GymEnvironmentT s a t) Source #
Methods liftIO :: IO a -> GymEnvironmentT s a t a #
MonadThrow t => MonadThrow (GymEnvironmentT s a t) Source #
Methods throwM :: Exception e => e -> GymEnvironmentT s a t a #
(MonadIO t, MonadMWCRandom t) => MonadMWCRandom (GymEnvironmentT s a t) Source #
Methods getGen :: GymEnvironmentT s a t GenIO Source #

type RunnerT s a t x = Bool -> GymEnvironmentT s a t x -> t (Either ServantError (DList (Event Reward s a))) Source #

type alias for a runEnvironment type which executes an environment action

type Runner s a x = RunnerT s a IO x Source #

type alias of RunnerT in IO

data GymException Source #

Possible errors we might encounter while interacting with our environment

Constructors

UnexpectedServerResponse String
TypeError String
EnvironmentRequiresReset

Instances

Show GymException Source #
Methods showsPrec :: Int -> GymException -> ShowS # show :: GymException -> String # showList :: [GymException] -> ShowS #
Exception GymException Source #
Methods toException :: GymException -> SomeException # fromException :: SomeException -> Maybe GymException # displayException :: GymException -> String #

runEnvironmentT :: forall o a t x. MonadIO t => GymEnv -> Manager -> BaseUrl -> Bool -> GymEnvironmentT o a t x -> t (Either ServantError (DList (Event Reward o a))) Source #

run an effectful gym to completion and return either an error, or the history of the agent. FIXME: move this into a History monad seperate from the Environment so that we don't blow up memory.

runEnvironment :: GymEnv -> Manager -> BaseUrl -> Bool -> GymEnvironment o a x -> IO (Either ServantError (DList (Event Reward o a))) Source #

same as runEnvironmentT, but with IO as the base monad

runDefaultEnvironmentT :: MonadIO t => GymEnv -> Bool -> GymEnvironmentT o a t x -> t (Either ServantError (DList (Event Reward o a))) Source #

same as runEnvironmentT, however use http-client's default manager settings

runDefaultEnvironment :: GymEnv -> Bool -> GymEnvironment o a x -> IO (Either ServantError (DList (Event Reward o a))) Source #

same as runEnvironment, however use http-client's default manager settings

_reset :: (MonadIO t, MonadThrow t, FromJSON o) => GymEnvironmentT o a t (Initial o) Source #

generic rest function which makes a call to the gym and returns the first observation

_step :: (MonadIO t, MonadThrow t, ToJSON a, r ~ Reward, FromJSON o) => a -> GymEnvironmentT o a t (Obs r o) Source #

generic step function which takes a ToJSONable action and returns a reward and a FromJSONable state