requirements

WavLM_modules.py

@@ -450,9 +450,7 @@ class MultiheadAttention(nn.Module):
         relative_position_bucket = self._relative_positions_bucket(
             relative_position, bidirectional=True
         )
-        # relative_position_bucket = relative_position_bucket.to(
-        #     self.relative_attention_bias.weight.device
-        # )
+        relative_position_bucket.cuda()
         values = self.relative_attention_bias(relative_position_bucket)
         values = values.permute([2, 0, 1])
         return values