[Fix] Fix attenstion clamp max params (#1034)

2022-09-26 14:12:51 +08:00 · 2022-09-26 14:12:51 +08:00 · 6ebb3f77ad
parent 0b4a67dd31
commit 6ebb3f77ad
1 changed files with 1 additions and 4 deletions
--- a/mmcls/models/utils/attention.py
+++ b/mmcls/models/utils/attention.py
@ -261,10 +261,7 @@ class WindowMSAV2(BaseModule):
        attn = (
            F.normalize(q, dim=-1) @ F.normalize(k, dim=-1).transpose(-2, -1))
        logit_scale = torch.clamp(
-            self.logit_scale,
-            max=torch.log(
-                torch.tensor(1. / 0.01,
-                             device=self.logit_scale.device))).exp()
+            self.logit_scale, max=np.log(1. / 0.01)).exp()
        attn = attn * logit_scale

        relative_position_bias_table = self.cpb_mlp(