Update hapi predict interface (#28180)

* update hapi predict interface

* fix code style

* fix docs

* fix docs

* fix docs

* update docs

* fix codes style

* fix unittest

* fix unittest

* fix coverage

python/paddle/fluid/tests/unittests/test_rnn_decode_api.py

@@ -628,7 +628,7 @@ class ModuleApiTest(unittest.TestCase):
             model.prepare()
             if self.param_states:
                 model.load(self.param_states, optim_state=None)
-            return model.test_batch(self.inputs)
+            return model.predict_batch(self.inputs)
 
     def check_output_with_place(self, place, mode="test"):
         dygraph_output = self._calc_output(place, mode, dygraph=True)

python/paddle/hapi/model.py

@@ -261,7 +261,7 @@ class StaticGraphAdapter(object):
         self.mode = 'eval'
         return self._run(inputs, labels)
 
-    def test_batch(self, inputs):
+    def predict_batch(self, inputs):
         self.mode = 'test'
         return self._run(inputs, None)
 
@@ -723,7 +723,7 @@ class DynamicGraphAdapter(object):
         else:
             return metrics
 
-    def test_batch(self, inputs):
+    def predict_batch(self, inputs):
         self.model.network.eval()
         self.mode = 'test'
         inputs = [to_variable(x) for x in to_list(inputs)]
@@ -894,10 +894,13 @@ class Model(object):
         Run one training step on a batch of data.
 
         Args:
-            inputs (list): A list of numpy.ndarray, each is a batch of
-                input data.
-            labels (list): A list of numpy.ndarray, each is a batch of
-                input label. If has no labels, set None. Default is None.
+            inputs (numpy.ndarray|Tensor|list): Batch of input data. It could 
+                be a numpy array or paddle.Tensor, or a list of arrays or 
+                tensors (in case the model has multiple inputs).
+            labels (numpy.ndarray|Tensor|list): Batch of labels. It could be 
+                a numpy array or paddle.Tensor, or a list of arrays or tensors 
+                (in case the model has multiple labels). If has no labels, 
+                set None. Default is None.
 
         Returns:
             A list of scalar training loss if the model has no metrics,
@@ -941,10 +944,13 @@ class Model(object):
         Run one evaluating step on a batch of data.
 
         Args:
-            inputs (list): A list of numpy.ndarray, each is a batch of
-                input data.
-            labels (list): A list of numpy.ndarray, each is a batch of
-                input label. If has no labels, set None. Default is None.
+            inputs (numpy.ndarray|Tensor|list): Batch of input data. It could 
+                be a numpy array or paddle.Tensor, or a list of arrays or 
+                tensors (in case the model has multiple inputs).
+            labels (numpy.ndarray|Tensor|list): Batch of labels. It could be 
+                a numpy array or paddle.Tensor, or a list of arrays or tensors 
+                (in case the model has multiple labels). If has no labels, 
+                set None. Default is None.
 
         Returns:
             A list of scalar testing loss if the model has no metrics,
@@ -984,13 +990,14 @@ class Model(object):
             self._update_inputs()
         return loss
 
-    def test_batch(self, inputs):
+    def predict_batch(self, inputs):
         """
-        Run one testing step on a batch of data.
+        Run one predicting step on a batch of data.
 
         Args:
-            inputs (list): A list of numpy.ndarray, each is a batch of
-                input data.
+            inputs (numpy.ndarray|Tensor|list): Batch of input data. It could 
+                be a numpy array or paddle.Tensor, or a list of arrays or 
+                tensors (in case the model has multiple inputs).
 
         Returns:
             A list of numpy.ndarray of predictions, that is the outputs
@@ -1019,10 +1026,10 @@ class Model(object):
               model = paddle.Model(net, input, label)
               model.prepare()
               data = np.random.random(size=(4,784)).astype(np.float32)
-              out = model.test_batch([data])
+              out = model.predict_batch([data])
               print(out)
         """
-        loss = self._adapter.test_batch(inputs)
+        loss = self._adapter.predict_batch(inputs)
         if fluid.in_dygraph_mode() and self._input_shapes is None:
             self._update_inputs()
         return loss
@@ -1847,10 +1854,9 @@ class Model(object):
                     logs[k] = v
             else:
                 if self._inputs is not None:
-                    outs = getattr(self,
-                                   mode + '_batch')(data[:len(self._inputs)])
+                    outs = self.predict_batch(data[:len(self._inputs)])
                 else:
-                    outs = getattr(self, mode + '_batch')(data)
+                    outs = self.predict_batch(data)
 
                 outputs.append(outs)
 

python/paddle/tests/test_model.py

@@ -284,6 +284,17 @@ class TestModel(unittest.TestCase):
 
         fluid.disable_dygraph() if dynamic else None
 
+    def test_predict_without_inputs(self):
+        fluid.enable_dygraph(self.device)
+        model = Model(LeNet())
+        model.prepare()
+        model.load(self.weight_path)
+        model._inputs = None
+        output = model.predict(
+            self.test_dataset, batch_size=64, stack_outputs=True)
+        np.testing.assert_equal(output[0].shape[0], len(self.test_dataset))
+        fluid.disable_dygraph()
+
 
 class MyModel(paddle.nn.Layer):
     def __init__(self):
@@ -370,7 +381,7 @@ class TestModelFunction(unittest.TestCase):
             inputs = [InputSpec([None, dim], 'float32', 'x')]
             model = Model(net, inputs)
             model.prepare()
-            out, = model.test_batch([data])
+            out, = model.predict_batch([data])
 
             np.testing.assert_allclose(out, ref, rtol=1e-6)
             fluid.disable_dygraph() if dynamic else None
@@ -546,7 +557,7 @@ class TestModelFunction(unittest.TestCase):
                 np.random.random((1, 1, 28, 28)), dtype=np.float32)
 
             model.save(save_dir, training=False)
-            ori_results = model.test_batch(tensor_img)
+            ori_results = model.predict_batch(tensor_img)
             fluid.disable_dygraph() if dynamic else None
 
             place = fluid.CPUPlace() if not fluid.is_compiled_with_cuda(
@@ -569,7 +580,7 @@ class TestModelFunction(unittest.TestCase):
         mnist_data = MnistDataset(mode='train')
         paddle.disable_static()
         # without inputs
-        for initial in ["fit", "train_batch", "eval_batch", "test_batch"]:
+        for initial in ["fit", "train_batch", "eval_batch", "predict_batch"]:
             save_dir = tempfile.mkdtemp()
             if not os.path.exists(save_dir):
                 os.makedirs(save_dir)
@@ -590,7 +601,7 @@ class TestModelFunction(unittest.TestCase):
                 elif initial == "eval_batch":
                     model.eval_batch([img], [label])
                 else:
-                    model.test_batch([img])
+                    model.predict_batch([img])
 
             model.save(save_dir, training=False)
             shutil.rmtree(save_dir)

python/paddle/tests/test_pretrained_model.py

@@ -40,10 +40,10 @@ class TestPretrainedModel(unittest.TestCase):
 
             if dygraph:
                 model.save(path)
-                res['dygraph'] = model.test_batch(x)
+                res['dygraph'] = model.predict_batch(x)
             else:
                 model.load(path)
-                res['static'] = model.test_batch(x)
+                res['static'] = model.predict_batch(x)
 
             if not dygraph:
                 paddle.disable_static()

python/paddle/tests/test_transforms.py

@@ -205,6 +205,12 @@ class TestTransformsCV2(unittest.TestCase):
         assert isinstance(tensor, paddle.Tensor)
         np.testing.assert_equal(tensor.shape, (3, 50, 100))
 
+    def test_keys(self):
+        fake_img1 = self.create_image((200, 150, 3))
+        fake_img2 = self.create_image((200, 150, 3))
+        trans_pad = transforms.Pad(10, keys=("image", ))
+        fake_img_padded = trans_pad((fake_img1, fake_img2))
+
     def test_exception(self):
         trans = transforms.Compose([transforms.Resize(-1)])
 

python/paddle/tests/test_vision_models.py

@@ -33,7 +33,7 @@ class TestVisonModels(unittest.TestCase):
         model = paddle.Model(net, input)
         model.prepare()
 
-        model.test_batch(x)
+        model.predict_batch(x)
 
     def test_mobilenetv2_pretrained(self):
         self.models_infer('mobilenet_v2', pretrained=False)
@@ -77,7 +77,7 @@ class TestVisonModels(unittest.TestCase):
         lenet.prepare()
 
         x = np.array(np.random.random((2, 1, 28, 28)), dtype=np.float32)
-        lenet.test_batch(x)
+        lenet.predict_batch(x)
 
 
 if __name__ == '__main__':

python/paddle/vision/transforms/transforms.py

@@ -272,7 +272,7 @@ class BaseTransform(object):
             else:
                 outputs.append(apply_func(inputs[i]))
         if len(inputs) > len(self.keys):
-            outputs.extend(input[len(self.keys):])
+            outputs.extend(inputs[len(self.keys):])
 
         if len(outputs) == 1:
             outputs = outputs[0]