Issue 2350 - support of all padding modes with tensors #2368
Conversation
Codecov Report
@@ Coverage Diff @@
## master #2368 +/- ##
==========================================
+ Coverage 68.49% 68.52% +0.03%
==========================================
Files 93 93
Lines 7655 7670 +15
Branches 1177 1182 +5
==========================================
+ Hits 5243 5256 +13
Misses 2075 2075
- Partials 337 339 +2
Continue to review full report at Codecov.
|
test/test_functional_tensor.py
Outdated
| if pil_tensor.dtype != tensor.dtype: | ||
| pil_tensor = pil_tensor.to(tensor.dtype) |
There was a problem hiding this comment.
Which case led you to have to perform a casting here, is it for the extra float32 and float64 types?
I think it might be preferable to avoid doing implicit casts in this function, and instead perform the cast directly in the caller -- it's more explicit.
There was a problem hiding this comment.
Right, I can cast explicitly before instead of this function 👍
|
|
||
| out_dtype = img.dtype | ||
| need_cast = False | ||
| if img.dtype not in (torch.float32, torch.float64): |
There was a problem hiding this comment.
This could probably be optimized because constant padding supports uint8 types as well.
There was a problem hiding this comment.
I agree, will fix !
test/test_functional_tensor.py
Outdated
| if pil_tensor.dtype != tensor.dtype: | ||
| pil_tensor = pil_tensor.to(tensor.dtype) |
There was a problem hiding this comment.
Right, I can cast explicitly before instead of this function 👍
|
|
||
| out_dtype = img.dtype | ||
| need_cast = False | ||
| if img.dtype not in (torch.float32, torch.float64): |
There was a problem hiding this comment.
I agree, will fix !
* [WIP] functional_tensor supports more padding modes * [WIP] Support all padding modes * Removed wip symmetric mode * Improvements according to the review
Fixes #2350
Description:
functional_tensor.pad