!7646 eval result only return probs

Merge pull request !7646 from wanghua/master

model_zoo/official/nlp/tinybert/run_task_distill.py

@@ -301,7 +301,7 @@ def do_eval_standalone():
             input_data.append(data[i])
         input_ids, input_mask, token_type_id, label_ids = input_data
         logits = eval_model(input_ids, token_type_id, input_mask)
-        callback.update(logits[3], label_ids)
+        callback.update(logits, label_ids)
     acc = callback.acc_num / callback.total_num
     print("======================================")
     print("============== acc is {}".format(acc))

model_zoo/official/nlp/tinybert/src/tinybert_model.py

@@ -964,7 +964,7 @@ class BertModelCLS(nn.Cell):
     The returned output represents the final logits as the results of log_softmax is propotional to that of softmax.
     """
     def __init__(self, config, is_training, num_labels=2, dropout_prob=0.0,
-                 use_one_hot_embeddings=False, phase_type="teacher"):
+                 use_one_hot_embeddings=False, phase_type="student"):
         super(BertModelCLS, self).__init__()
         self.bert = BertModel(config, is_training, use_one_hot_embeddings)
         self.cast = P.Cast()
@@ -992,4 +992,6 @@ class BertModelCLS(nn.Cell):
             logits = self.dense_1(cls)
         logits = self.cast(logits, self.dtype)
         log_probs = self.log_softmax(logits)
-        return seq_output, att_output, logits, log_probs
+        if self._phase == 'train' or self.phase_type == "teacher":
+            return seq_output, att_output, logits, log_probs
+        return log_probs

model_zoo/official/nlp/tinybert/src/utils.py

@@ -100,7 +100,7 @@ class EvalCallBack(Callback):
                 input_ids, input_mask, token_type_id, label_ids = input_data
                 self.network.set_train(False)
                 logits = self.network(input_ids, token_type_id, input_mask)
-                callback.update(logits[3], label_ids)
+                callback.update(logits, label_ids)
             acc = callback.acc_num / callback.total_num
             with open("./eval.log", "a+") as f:
                 f.write("acc_num {}, total_num{}, accuracy{:.6f}".format(callback.acc_num, callback.total_num,