Fix label access on clustering model (#3030)

* Fix label access on clustering model

* format code

* update predict

* add default shape for label

Author: lhw362950217

go/executor/executor.go

@@ -25,10 +25,11 @@ import (
 	"path"
 	"path/filepath"
 	"regexp"
-	"sqlflow.org/sqlflow/go/codegen/experimental"
 	"strings"
 	"sync"
 
+	"sqlflow.org/sqlflow/go/codegen/experimental"
+
 	"sqlflow.org/sqlflow/go/verifier"
 
 	"sqlflow.org/sqlflow/go/codegen/optimize"

go/ir/ir_generator.go

@@ -114,7 +114,8 @@ func GenerateTrainStmt(slct *parser.SQLFlowSelectStmt) (*TrainStmt, error) {
 	}
 	label := &NumericColumn{
 		FieldDesc: &FieldDesc{
-			Name: tc.Label,
+			Name:  tc.Label,
+			Shape: []int{1},
 		}}
 
 	vslct, _ := parseValidationSelect(attrList)

go/sql/executor_ir_test.go

@@ -188,7 +188,6 @@ func TestExecutorTrainAndPredictDNN(t *testing.T) {
 }
 
 func TestExecutorTrainAndPredictClusteringLocalFS(t *testing.T) {
-	t.Skip("fix random nan loss error then re-enable this test")
 	a := assert.New(t)
 	modelDir, e := ioutil.TempDir("/tmp", "sqlflow_models")
 	a.Nil(e)

python/runtime/local/create_result_table.py

@@ -32,10 +32,11 @@ def create_predict_table(conn, select, result_table, train_label_desc,
     """
     name_and_types = db.selected_columns_and_types(conn, select)
     train_label_index = -1
-    for i, (name, _) in enumerate(name_and_types):
-        if name == train_label_desc.name:
-            train_label_index = i
-            break
+    if train_label_desc:
+        for i, (name, _) in enumerate(name_and_types):
+            if name == train_label_desc.name:
+                train_label_index = i
+                break
 
     if train_label_index >= 0:
         del name_and_types[train_label_index]
@@ -45,10 +46,12 @@ def create_predict_table(conn, select, result_table, train_label_desc,
         column_strs.append("%s %s" %
                            (name, db.to_db_field_type(conn.driver, typ)))
 
-    if train_label_desc.format == DataFormat.PLAIN:
+    if train_label_desc and train_label_desc.format == DataFormat.PLAIN:
         train_label_field_type = DataType.to_db_field_type(
             conn.driver, train_label_desc.dtype)
     else:
+        # if no train lable description is provided (clustering),
+        # we treat the column type as string
         train_label_field_type = DataType.to_db_field_type(
             conn.driver, DataType.STRING)
 

python/runtime/local/tensorflow_submitter/predict.py

@@ -46,8 +46,9 @@ def pred(datasource, select, result_table, pred_label_name, model):
 
     model_params = model.get_meta("attributes")
     train_fc_map = model.get_meta("features")
-    train_label_desc = model.get_meta("label").get_field_desc()[0]
-    train_label_name = train_label_desc.name
+    label_meta = model.get_meta("label")
+    train_label_desc = label_meta.get_field_desc()[0] if label_meta else None
+    train_label_name = train_label_desc.name if train_label_desc else None
     estimator_string = model.get_meta("class_name")
     save = "model_save"
 

python/runtime/step/tensorflow/train.py

@@ -80,7 +80,12 @@ def train_step(original_sql,
     feature_column_names = [fd.name for fd in field_descs]
     feature_metas = dict([(fd.name, fd.to_dict(dtype_to_string=True))
                           for fd in field_descs])
-    label_meta = fc_label_ir.get_field_desc()[0].to_dict(dtype_to_string=True)
+
+    # no label for clustering model
+    label_meta = None
+    if fc_label_ir:
+        label_meta = fc_label_ir.get_field_desc()[0].to_dict(
+            dtype_to_string=True)
 
     feature_column_names_map = dict()
     for target in fc_map_ir: