Hi Matias,

A Tensor is a general name for a n-dimensional array. The term has risen in popularity ever since TensorFlow has gained traction as the most popular neural network framework in Python.

In the documentation tensors are not shown as an array data type, but rather as an abstraction just like tables, see here. Following this discussion on github: Wes has mentioned: "Tensors aren't part of the Arrow columnar format. Let's continue the discussion on the mailing list or JIRA". Therefore, they are not included as a data type and thus not usable as array format.

Regarding your question of the Plasma object store, from my perspective Plasma is a generic byte object storage and thus not connected to the discussion above.


Op vr 19 mrt. 2021 om 14:50 schreef Matias Guijarro <matias.guijarro@free.fr>:
Hi Fernando,

Thanks for the information about Tensor (I wonder where the name comes
from ?).

Are tensors considered apart or is it possible to put them in a column
like normal ?

On 19/03/2021 10:20, Fernando Herrera wrote:
> However, I don't think the data stored in the tensor will be
> compressed. It will be
> orderly stored so you can share the tensors with other processes.

How do you share them with other processes ? Is it via Plasma ?

(sorry for the noob questions - as said before I am investigating Arrow
now, and
essentially collecting information before trying anything)

Thanks a lot,

Met vriendelijke groeten / Kind regards, 

Tom Scheffers
+316 233 728 15

Young Bulls