torchrl package

Subpackages

• torchrl.agents package
  • Submodules
  • torchrl.agents.a2c_agent module
  • torchrl.agents.base_agent module
  • torchrl.agents.ddpg_agent module
  • torchrl.agents.dqn_agent module
  • torchrl.agents.gym_random_agent module
  • torchrl.agents.ppo_agent module
  • Module contents
• torchrl.cli package
  • Subpackages
    • torchrl.cli.commands package
      • Submodules
      • torchrl.cli.commands.list module
      • torchrl.cli.commands.resume module
      • torchrl.cli.commands.run module
      • Module contents
  • Submodules
  • torchrl.cli.boot module
  • Module contents
• torchrl.policies package
  • Submodules
  • torchrl.policies.epsilon_greedy module
  • torchrl.policies.ou_noise module
  • Module contents
• torchrl.problems package
  • Submodules
  • torchrl.problems.a2c module
  • torchrl.problems.base_hparams module
  • torchrl.problems.ddpg module
  • torchrl.problems.dqn module
  • torchrl.problems.gym_problem module
  • torchrl.problems.ppo module
  • torchrl.problems.prioritized_dqn module
  • Module contents
• torchrl.registry package
  • Submodules
  • torchrl.registry.problems module
  • torchrl.registry.registry module
  • Module contents
• torchrl.runners package
  • Submodules
  • torchrl.runners.base_runner module
  • torchrl.runners.gym_runner module
  • Module contents
• torchrl.storage package
  • Submodules
  • torchrl.storage.gpu_replay_buffer module
  • torchrl.storage.prioritized_replay_buffer module
  • torchrl.storage.replay_buffer module
  • Module contents
• torchrl.utils package
  • Submodules
  • torchrl.utils.gym_utils module
  • torchrl.utils.misc module
  • torchrl.utils.multi_envs module
  • torchrl.utils.multi_proc_wrapper module
  • torchrl.utils.nop module
  • torchrl.utils.schedule module
  • torchrl.utils.sum_tree module
  • Module contents

Submodules

torchrl.models module

class torchrl.models.A2CNet(input_size, output_size, hidden_size)

Bases: torch.nn.modules.module.Module

forward(obs)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class torchrl.models.ActorCriticNet(input_size, output_size, hidden_size, std=0.0)

Bases: torch.nn.modules.module.Module

forward(obs)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

static init_weights(module)

class torchrl.models.DDPGActorNet(state_size, action_size, hidden_size)

Bases: torch.nn.modules.module.Module

forward(obs)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class torchrl.models.DDPGCriticNet(state_size, action_size, hidden_size)

Bases: torch.nn.modules.module.Module

forward(obs, action)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class torchrl.models.QNet(input_size, output_size)

Bases: torch.nn.modules.module.Module

forward(obs)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Module contents