hacked up

[drisspg]

Stacked PRs:

• ->hacked up #88

---

hacked up

❯ torchfix auto_deprecate.py
auto_deprecate.py:8:15: TOR101 [*] Use of deprecated function torch.nn.functional.soft_margin_loss
--- /home/drisspg/meta/scripts/misc/auto_deprecate.py
+++ /home/drisspg/meta/scripts/misc/auto_deprecate.py
@@ -3,7 +3,7 @@
 
 import torch.nn.functional as F
 
 a = torch.tensor([1, 2, 3])
 b = torch.tensor([4, 5, 6])
-my_loss_fun = F.soft_margin_loss(a, b, size_average=True, reduce="some string")
+my_loss_fun = F.soft_margin_loss(a, b, reduction = 'mean')
 
Finished checking 1 files.
[*] 1 potentially fixable with the --fix option

[kit1980]

This will be complaining about my_loss_fun = F.soft_margin_loss(a, b, reduction = 'mean') ?

Probably better to create another visitor like "torchfix/visitors/deprecated_arguments/" or somethings instead of hacking on "torchfix/visitors/deprecated_symbols/", which is for when whole functions are deprecated.

[kit1980]

Right now TorchFix doesn't depend on or use PyTorch.
Maybe we can add PyTorch to CI, but then which version?
Is it hard to copy or re-implement legacy_get_string?