Update handler.py

handler.py

@@ -35,7 +35,7 @@ class EndpointHandler():
           )
           token_ids, token_mask = tokenized.input_ids.cuda(), tokenized.attention_mask.cuda()
           with torch.no_grad():
-              out = model(token_ids, attention_mask=token_mask)
+              out = self.model(token_ids, attention_mask=token_mask)
           meaned_logits = (out.logits * token_mask.unsqueeze(-1)).sum(1) / token_mask.sum(
               1
           ).unsqueeze(-1)