What’s the efficient way to convert Python objects of various types to Tensor objects?

1

Hi.
I want to know is there any efficient way to convert Python objects of various types to Tensor objects,like convert_to_tensor in tensorflow, It accepts Tensor objects, numpy arrays, Python lists, and Python scalars.
Although I know that I can determine the type of data and then call the corresponding API, but I think this is too bloated, and I don’t want to write a bunch of ‘if judgment statements’ to determine the type of data, so I would like to ask if there is a corresponding pytorch API or is there any efficient way

Appreciate for any replies.

2

looking forward for your help!

3

The easiest way would be probably to use torch.tensor.
Would that work or are you running into some issues?

1 Like
4

@ptrblck
HI, i have one question…
so i have a list of list of list…sth like this

img1
   bbox
      cls1_bbox_list (100*5 ndarray)
      cls2_bbox_list (95*5 ndarray)
      cls3_bbox_list ([ ] )  ##empty
  masks
      cls1_masks_list  (5*256*256)
       cls2_mask_list ([ ])
img2
  ....

how I can quickly convert them into torch tensor and put into GPU?

I did

 map(torch.cuda.FloatTensor,  the_whole_listoflistoflist)

but it seems a bit slow
any ideas? Thanks in advance :slight_smile:

5

I’m unfortunately not aware of a better approach, so let’s wait for others to chime in :confused:

1 Like