gfn.gym.line

Classes

Line

Mixture of Gaussians Line environment.

Module Contents

class gfn.gym.line.Line(mus, sigmas, init_value, n_sd=4.5, n_steps_per_trajectory=5, device='cpu', debug=False)

Bases: gfn.env.Env

Mixture of Gaussians Line environment.

Parameters:

  • mus (list)

  • sigmas (list)

  • init_value (float)

  • n_sd (float)

  • n_steps_per_trajectory (int)

  • device (Literal['cpu', 'cuda'] | torch.device)

  • debug (bool)

mus

The means of the Gaussians.

sigmas

The standard deviations of the Gaussians.

n_sd

The number of standard deviations to consider for the bounds.

n_steps_per_trajectory

The number of steps per trajectory.

mixture

The mixture of Gaussians.

init_value

The initial value of the state.

backward_step(states, actions)

Performs a backward step in the environment.

Parameters:
Returns:

The previous states.

Return type:

gfn.states.States

init_value
is_action_valid(states, actions, backward=False)

Checks if the actions are valid.

Parameters:
Returns:

True if the actions are valid, False otherwise.

Return type:

bool

log_partition(condition=None)

Returns the log partition of the reward function.

Return type:

torch.Tensor

log_reward(final_states)

Computes the log reward of the environment.

Parameters:

final_states (gfn.states.States) – The final states of the environment.

Returns:

The log reward.

Return type:

torch.Tensor

mixture
mus
n_sd = 4.5
n_steps_per_trajectory = 5
sigmas
step(states, actions)

Performs a step in the environment.

Parameters:
Returns:

The next states.

Return type:

gfn.states.States