hide dropdown menus when the labels match when labels not matching

app_text_classification.py

@@ -4,7 +4,7 @@ import os
 import time
 import subprocess
 import logging
-import threading
+import collections
 
 import json
 
@@ -198,7 +198,7 @@ def get_demo():
         
         dropdown_placement = [gr.Dropdown(visible=False) for _ in range(MAX_LABELS + MAX_FEATURES)]
             
-        if ppl is None:
+        if ppl is None: # pipeline not found
             gr.Warning("Model not found")
             return (
                 gr.update(visible=False),
@@ -209,7 +209,8 @@ def get_demo():
         model_id2label = ppl.model.config.id2label
         ds_labels, ds_features = get_labels_and_features_from_dataset(dataset_id, dataset_config, dataset_split)
         
-        if ds_labels is None or ds_features is None:
+        # when dataset does not have labels or features
+        if not isinstance(ds_labels, list) or not isinstance(ds_features, list):
             gr.Warning(CONFIRM_MAPPING_DETAILS_FAIL_RAW)
             return (
                 gr.update(visible=False),
@@ -223,7 +224,10 @@ def get_demo():
             ds_features,
             model_id2label,
         )
-        if model_id2label.items() != ds_labels.items():
+
+        # when labels or features are not aligned
+        # show manually column mapping
+        if collections.Counter(model_id2label.items()) != collections.Counter(ds_labels) or ds_features[0] != 'text':
             gr.Warning(CONFIRM_MAPPING_DETAILS_FAIL_RAW)
             return (
                 gr.update(visible=False),
@@ -236,7 +240,7 @@ def get_demo():
         return (
             gr.update(value=prediction_input, visible=True),
             gr.update(value=prediction_output, visible=True),
-            gr.update(open=False),
+            gr.update(visible=True, open=False),
             *column_mappings
         )