Commit 67f0084
authored
Fix padding for int contexts (#227)
*Issue #, if available:* On Linux, the final call to `.to` creates
trouble when input tensors are integer. For example:
```
>>> a = torch.tensor([1])
>>> b = torch.stack([torch.full((1,), torch.nan), a])
>>> b
tensor([[nan],
[1.]])
>>> b.to(a)
tensor([[-9223372036854775808],
[ 1]])
```
By submitting this pull request, I confirm that you can use, modify,
copy, and redistribute this contribution, under the terms of your
choice.1 parent 47cac08 commit 67f0084
4 files changed
+32
-3
lines changed| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
17 | 17 | | |
18 | 18 | | |
19 | 19 | | |
20 | | - | |
| 20 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
1 | 1 | | |
2 | | - | |
| 2 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| 2 | + | |
| 3 | + | |
| 4 | + | |
| 5 | + | |
| 6 | + | |
| 7 | + | |
| 8 | + | |
| 9 | + | |
| 10 | + | |
| 11 | + | |
| 12 | + | |
| 13 | + | |
| 14 | + | |
| 15 | + | |
| 16 | + | |
| 17 | + | |
| 18 | + | |
| 19 | + | |
| 20 | + | |
| 21 | + | |
| 22 | + | |
| 23 | + | |
| 24 | + | |
| 25 | + | |
| 26 | + | |
| 27 | + | |
| 28 | + | |
| 29 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
10 | 10 | | |
11 | 11 | | |
12 | 12 | | |
13 | | - | |
| 13 | + | |
0 commit comments