Language

unilab.base.torch_env.TorchEnvState¶

class unilab.base.torch_env.TorchEnvState[source]¶

Bases: object

Tensor-native vectorized environment transition.

Terminal pre-reset observations are represented only by final_observation; info never carries duplicate terminal-observation entries.

Parameters:

obs (dict[str, Tensor])
reward (Tensor)
terminated (Tensor)
truncated (Tensor)
info (dict[str, Any])
final_observation (dict[str, Tensor] | None)

Methods

`__init__`(obs, reward, terminated, truncated, ...)
`replace`(**updates)

Attributes

`final_observation`
`obs`
`reward`
`terminated`
`truncated`
`info`

obs: dict[str, Tensor]¶

reward: Tensor¶

terminated: Tensor¶

truncated: Tensor¶

info: dict[str, Any]¶

final_observation: dict[str, Tensor] | None = None¶

replace(**updates)[source]¶

Parameters:: updates (Any)
Return type:: TorchEnvState

__init__(obs, reward, terminated, truncated, info, final_observation=None)¶

Parameters:

obs (dict[str, Tensor])
reward (Tensor)
terminated (Tensor)
truncated (Tensor)
info (dict[str, Any])
final_observation (dict[str, Tensor] | None)