Current, calling print on a Tensor instance displays all of its scalars in a flat string.
This is not ideal as the default printing behavior because the printed output is prohibitively long for large tensors:
At a minimum, we want:
- Pretty printing. (Printing 1-D element vectors on new lines.)
- Truncation. (Shortening overly long lines, hiding content with “...”.)
- Pretty printing formatter options.
- A mechanism to do full-printing (like how printing is implemented today).
|Basic tensor pretty printing||Closed|