edelwace-0.1.0.0: HaskTorch Reinforcement Learning Agents for GACE

Safe Haskell	None
Language	Haskell2010

RPB.RPB

Description

Default / Naive Replay Buffer

Synopsis

data Buffer a = Buffer {
- states :: !a
- actions :: !a
- rewards :: !a
- states' :: !a
- dones :: !a
}
mkBuffer :: Buffer Tensor
empty :: Buffer Tensor
size :: Buffer Tensor -> Int
push :: Int -> Buffer Tensor -> Tensor -> Tensor -> Tensor -> Tensor -> Tensor -> Buffer Tensor
push' :: Int -> Buffer Tensor -> Buffer Tensor -> Buffer Tensor
pop :: Int -> Buffer Tensor -> Buffer Tensor
sample :: Tensor -> Buffer Tensor -> Buffer Tensor
sampleIO :: Int -> Buffer Tensor -> IO (Buffer Tensor)
randomBatches :: Int -> Int -> Buffer Tensor -> IO [Buffer Tensor]
standardizeState :: Float -> Buffer Tensor -> Buffer Tensor

Documentation

data Buffer a Source #

Strict Simple/Naive Replay Buffer

Constructors

Buffer
Fields states :: !a States actions :: !a Actions rewards :: !a Rewards states' :: !a Next States dones :: !a Terminal Mask

Instances

Instances details

Functor Buffer Source #
Instance details Defined in RPB.RPB Methods fmap :: (a -> b) -> Buffer a -> Buffer b # (<$) :: a -> Buffer b -> Buffer a #
Eq a => Eq (Buffer a) Source #
Instance details Defined in RPB.RPB Methods (==) :: Buffer a -> Buffer a -> Bool # (/=) :: Buffer a -> Buffer a -> Bool #
Show a => Show (Buffer a) Source #
Instance details Defined in RPB.RPB Methods showsPrec :: Int -> Buffer a -> ShowS # show :: Buffer a -> String # showList :: [Buffer a] -> ShowS #

mkBuffer :: Buffer Tensor Source #

Create a new, empty Buffer on the CPU

empty :: Buffer Tensor Source #

Create Empty Buffer

size :: Buffer Tensor -> Int Source #

How many Trajectories are currently stored in memory

push :: Int -> Buffer Tensor -> Tensor -> Tensor -> Tensor -> Tensor -> Tensor -> Buffer Tensor Source #

Push new memories into Buffer

push' :: Int -> Buffer Tensor -> Buffer Tensor -> Buffer Tensor Source #

Push one buffer into another one

pop :: Int -> Buffer Tensor -> Buffer Tensor Source #

Pop numElems from Buffer

sample :: Tensor -> Buffer Tensor -> Buffer Tensor Source #

Get the given indices from Buffer

sampleIO :: Int -> Buffer Tensor -> IO (Buffer Tensor) Source #

Uniform random sample from Replay Buffer

randomBatches :: Int -> Int -> Buffer Tensor -> IO [Buffer Tensor] Source #

Generate a list of random batches from a given buffer

standardizeState :: Float -> Buffer Tensor -> Buffer Tensor Source #

Scale and clip states and states'