docarray.math.distance.torch module#

docarray.math.distance.torch.cosine(x_mat, y_mat, eps=1e-07, device='cpu')[source]#

Cosine distance between each row in x_mat and each row in y_mat.

Parameters:

  • x_mat (tensor) – torch with ndim=2

  • y_mat (tensor) – torch with ndim=2

  • eps (float) – a small jitter to avoid divde by zero

  • device (str) – the computational device for embed_model, can be either cpu or cuda.

Return type:

ndarray

Returns:

np.ndarray with ndim=2

docarray.math.distance.torch.euclidean(x_mat, y_mat, device='cpu')[source]#

Euclidean distance between each row in x_mat and each row in y_mat.

Parameters:

  • x_mat (tensor) – torch array with ndim=2

  • y_mat (tensor) – torch array with ndim=2

  • device (str) – the computational device for embed_model, can be either cpu or cuda.

Return type:

ndarray

Returns:

np.ndarray with ndim=2

docarray.math.distance.torch.sqeuclidean(x_mat, y_mat, device='cpu')[source]#

Squared euclidean distance between each row in x_mat and each row in y_mat.

Parameters:

  • x_mat (tensor) – torch array with ndim=2

  • y_mat (tensor) – torch array with ndim=2

  • device (str) – the computational device for embed_model, can be either cpu or cuda.

Return type:

ndarray

Returns:

np.ndarray with ndim=2