PathSample#

class flow_matching.path.path_sample.PathSample(x_1: Tensor, x_0: Tensor, t: Tensor, x_t: Tensor, dx_t: Tensor)[source]#

Represents a sample of a conditional-flow generated probability path.

x_1#

the target sample \(X_1\).

Type:

Tensor

x_0#

the source sample \(X_0\).

Type:

Tensor

t#

the time sample \(t\).

Type:

Tensor

x_t#

samples \(X_t \sim p_t(X_t)\), shape (batch_size, …).

Type:

Tensor

dx_t#

conditional target \(\frac{\partial X}{\partial t}\), shape: (batch_size, …).

Type:

Tensor