This repository has been archived by the owner on Oct 31, 2023. It is now read-only.

Name		Name	Last commit message	Last commit date
parent directory ..
rl		rl
README.md		README.md
__init__.py		__init__.py
asynchronous.py		asynchronous.py
brax.py		brax.py
dataloader.py		dataloader.py
gym.py		gym.py
gyma.py		gyma.py
remote.py		remote.py
transformers.py		transformers.py
utils.py		utils.py
xformers_transformers.py		xformers_transformers.py

README.md

salina.agents

We propose a list of agents to reuse (see Documentation in the code)

utils

Agents: Execute multiple agents sequentially
TemporalAgent: Execute one agent over multiple timesteps
CopyAgent: An agent to create copies of variables
PrintAgent: An agent that print variables

remote

RemoteAgent: A single agent in a single process
NRemoteAgent: A single agent parallelized over multiple processes

gym

GymAgent: An agent based on an openAI gym environment
AutoResetGymAgent: The same, but with an autoreset when reaching terminal states

brax

AutoResetBraxAgent: An agent based on a BRAX environment with autoreset
NoAutoResetBraxAgent: An agent based on a BRAX environment without autoreset

dataloader

ShuffledDatasetAgent: An agent to read random batches in a torch.utils.data.Dataset
DataLoaderAgent: An agent to do one pass over a complete dataset (based on a DataLoader)

asynchronous

AsynchronousAgent: it is used to execute any agent asynchronously, the agent creating its own workspace at each execution

agent=AsynchronousAgent(my_agent)
agent(**execution_arguments (not workspace provided)
if not agent.is_running(): workspace=agent.get_workspace()