GPT: fix batch predict. (PaddlePaddle#411)

wangxicoding · May 19, 2021 · f2dbd81 · f2dbd81
1 parent c7914e9
commit f2dbd81
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/paddlenlp/transformers/gpt/modeling.py b/paddlenlp/transformers/gpt/modeling.py
@@ -781,14 +781,14 @@ def model(self,
     def forward(self, input_ids, end_id):
         output, cached_kvs = self.model(input_ids, use_cache=True, cache=None)
         src_ids = input_ids
-        nid = paddle.argmax(output[0, -1]).reshape([1, -1])
+        nid = paddle.argmax(output[:, -1, :], axis=-1).reshape([-1, 1])
         src_ids = paddle.concat([src_ids, nid], axis=1)
         cur_len = 0
         while (cur_len < self.max_predict_len):
             output, cached_kvs = self.model(
                 nid, use_cache=True, cache=cached_kvs)
 
-            nid = paddle.argmax(output[0, -1]).reshape([1, -1])
+            nid = paddle.argmax(output[:, -1, :], axis=-1).reshape([-1, 1])
             src_ids = paddle.concat([src_ids, nid], axis=1)
             cur_len += 1
             if paddle.max(nid) == end_id: