Remove is pai for evaluate (#2749)

* remove is_pai for evaluate

* update

Author: typhoonzero

go/codegen/pai/template_tf.go

@@ -273,7 +273,7 @@ if os.environ.get('DISPLAY', '') == '':
 import json
 import types
 import sys
-from runtime.tensorflow import evaluate
+from runtime.pai.tensorflow import evaluate
 
 try:
     tf.enable_eager_execution()
@@ -307,7 +307,7 @@ if is_estimator:
 else:
     oss.load_file("{{.OSSModelDir}}", "model_save")
 
-evaluate.evaluate(datasource="{{.DataSource}}",
+evaluate._evaluate(datasource="{{.DataSource}}",
                   estimator_string=estimator,
                   select="""{{.Select}}""",
                   result_table="{{.ResultTable}}",
@@ -321,6 +321,5 @@ evaluate.evaluate(datasource="{{.DataSource}}",
                   batch_size=1,
                   validation_steps=None,
                   verbose=0,
-                  is_pai="{{.IsPAI}}" == "true",
                   pai_table="{{.PAITable}}")
 `

go/codegen/tensorflow/template_evaluate.go

@@ -104,7 +104,5 @@ evaluate(datasource="{{.DataSource}}",
          hdfs_namenode_addr="{{.HDFSNameNodeAddr}}",
          hive_location="{{.HiveLocation}}",
          hdfs_user="{{.HDFSUser}}",
-         hdfs_pass="{{.HDFSPass}}",
-         is_pai="{{.IsPAI}}" == "true",
-         pai_table="{{.PAIEvaluateTable}}")
+         hdfs_pass="{{.HDFSPass}}")
 `

python/runtime/pai/tensorflow/evaluate.py

@@ -21,7 +21,12 @@
 from runtime import oss
 from runtime.pai.pai_distributed import define_tf_flags
 from runtime.tensorflow import is_tf_estimator
-from runtime.tensorflow.evaluate import evaluate as _evaluate
+from runtime.tensorflow.evaluate import (estimator_evaluate, keras_evaluate,
+                                         write_result_metrics)
+from runtime.tensorflow.input_fn import get_dataset_fn
+from runtime.tensorflow.keras_with_feature_column_input import \
+    init_model_with_feature_column
+from runtime.tensorflow.set_log_level import set_log_level
 from tensorflow.estimator import (BoostedTreesClassifier,
                                   BoostedTreesRegressor, DNNClassifier,
                                   DNNLinearCombinedClassifier,
@@ -88,3 +93,57 @@ def evaluate(datasource, select, data_table, result_table, oss_model_path,
               verbose=0,
               is_pai=True,
               pai_table=data_table)
+
+
+def _evaluate(datasource,
+              estimator_string,
+              select,
+              result_table,
+              feature_columns,
+              feature_column_names,
+              feature_metas={},
+              label_meta={},
+              model_params={},
+              validation_metrics=["Accuracy"],
+              save="",
+              batch_size=1,
+              validation_steps=None,
+              verbose=0,
+              pai_table=""):
+    runtime.import_model_def(estimator_string, globals())
+    estimator_cls = eval(estimator_string)
+    is_estimator = is_tf_estimator(estimator_cls)
+    set_log_level(verbose, is_estimator)
+    eval_dataset = get_dataset_fn(select,
+                                  datasource,
+                                  feature_column_names,
+                                  feature_metas,
+                                  label_meta,
+                                  is_pai=True,
+                                  pai_table=pai_table,
+                                  batch_size=batch_size)
+
+    model_params.update(feature_columns)
+    if is_estimator:
+        FLAGS = tf.app.flags.FLAGS
+        model_params["model_dir"] = FLAGS.checkpointDir
+        estimator = estimator_cls(**model_params)
+        result_metrics = estimator_evaluate(estimator, eval_dataset,
+                                            validation_metrics)
+    else:
+        keras_model = init_model_with_feature_column(estimator, model_params)
+        keras_model_pkg = sys.modules[estimator_cls.__module__]
+        result_metrics = keras_evaluate(keras_model, eval_dataset, save,
+                                        keras_model_pkg, validation_metrics)
+
+    if result_table:
+        metric_name_list = ["loss"] + validation_metrics
+        write_result_metrics(result_metrics,
+                             metric_name_list,
+                             result_table,
+                             "pai_maxcompute",
+                             None,
+                             hdfs_namenode_addr="",
+                             hive_location="",
+                             hdfs_user="",
+                             hdfs_pass="")

python/runtime/tensorflow/evaluate.py

@@ -53,28 +53,23 @@ def evaluate(datasource,
              hdfs_namenode_addr="",
              hive_location="",
              hdfs_user="",
-             hdfs_pass="",
-             is_pai=False,
-             pai_table=""):
+             hdfs_pass=""):
     runtime.import_model_def(estimator_string, globals())
     estimator_cls = eval(estimator_string)
-
     is_estimator = is_tf_estimator(estimator_cls)
-
     set_log_level(verbose, is_estimator)
-
-    eval_dataset = get_dataset_fn(select, datasource, feature_column_names,
-                                  feature_metas, label_meta, is_pai, pai_table,
-                                  batch_size)
+    eval_dataset = get_dataset_fn(select,
+                                  datasource,
+                                  feature_column_names,
+                                  feature_metas,
+                                  label_meta,
+                                  is_pai=False,
+                                  pai_table="",
+                                  batch_size=batch_size)
 
     model_params.update(feature_columns)
     if is_estimator:
-        if is_pai:
-            FLAGS = tf.app.flags.FLAGS
-            model_params["model_dir"] = FLAGS.checkpointDir
-        else:
-            model_params["model_dir"] = save
-        # tf estimator always have feature_column argument
+        model_params["model_dir"] = save
         estimator = estimator_cls(**model_params)
         result_metrics = estimator_evaluate(estimator, eval_dataset,
                                             validation_metrics)
@@ -85,13 +80,8 @@ def evaluate(datasource,
                                         keras_model_pkg, validation_metrics)
 
     # write result metrics to a table
-    if is_pai:
-        driver = "pai_maxcompute"
-        conn = None
-    else:
-        conn = connect_with_data_source(datasource)
-        driver = conn.driver
-
+    conn = connect_with_data_source(datasource)
+    driver = conn.driver
     if result_table:
         metric_name_list = ["loss"] + validation_metrics
         write_result_metrics(result_metrics,