Respect `is_rec` when eval (#3187)

ppcls/engine/evaluation/classification.py

@@ -56,7 +56,10 @@ def classification_eval(engine, epoch_id=0):
 
         # image input
         with engine.auto_cast(is_eval=True):
-            out = engine.model(batch[0], batch[1])
+            if engine.is_rec:
+                out = engine.model(batch[0], batch[1])
+            else:
+                out = engine.model(batch[0])
 
         # just for DistributedBatchSampler issue: repeat sampling
         current_samples = batch_size * paddle.distributed.get_world_size()

ppcls/engine/evaluation/retrieval.py

@@ -137,7 +137,10 @@ def compute_feature(engine, name="gallery"):
             has_camera = True
             batch[2] = batch[2].reshape([-1, 1]).astype("int64")
         with engine.auto_cast(is_eval=True):
-            out = engine.model(batch[0], batch[1])
+            if engine.is_rec:
+                out = engine.model(batch[0], batch[1])
+            else:
+                out = engine.model(batch[0])
         if "Student" in out:
             out = out["Student"]