fix(run_classifier): move calc_ists_metrics to the top of the main to…

… skip unnecessary script executions

run_classifier.py

@@ -694,6 +694,21 @@ def main(_):
   if FLAGS.save_steps is not None:
     FLAGS.iterations = min(FLAGS.iterations, FLAGS.save_steps)
 
+  if FLAGS.calc_ists_metrics:
+    predictions = []
+
+    if not FLAGS.pred_file:
+      predictions = model_utils.get_predictions(FLAGS.predict_dir)
+    else:
+      predictions = [model_utils.extract_global_step(FLAGS.pred_file[:-4]), FLAGS.pred_file]
+
+    for global_step, pred_file_path in sorted(predictions, key=lambda x: x[0]):
+      f1_scores = calc_ists_metrics(pred_file_path, FLAGS.data_dir + "/test.tsv")
+      tf.logging.info( 'Dataset: ' + FLAGS.data_dir.split("/")[-1] + " Step: " + str(global_step) + ' [F1 Type]: {} \n [F1 Score]: {} \n [F1 T+S]: {}' % f1_scores)
+
+    # End execution after caclulations
+    return None
+
   if FLAGS.do_predict:
     predict_dir = FLAGS.predict_dir
     if not tf.gfile.Exists(predict_dir):
@@ -910,18 +925,6 @@ def tokenize_fn(text):
 
       with tf.gfile.Open(predict_json_path, "w") as fp:
         json.dump(predict_results, fp, indent=4)
-
-  if FLAGS.calc_ists_metrics:
-    predictions = []
-
-    if not FLAGS.pred_file:
-      predictions = model_utils.get_predictions(FLAGS.predict_dir)
-    else:
-      predictions = [model_utils.extract_global_step(FLAGS.pred_file[:-4]), FLAGS.pred_file]
-
-    for global_step, pred_file_path in sorted(predictions, key=lambda x: x[0]):
-      f1_scores = calc_ists_metrics(pred_file_path, FLAGS.data_dir + "/test.tsv")
-      tf.logging.info( 'Dataset: ' + FLAGS.data_dir.split("/")[-1] + " Step: " + str(global_step) + ' [F1 Type]: {} \n [F1 Score]: {} \n [F1 T+S]: {}' % f1_scores)
 
 
 if __name__ == "__main__":