Added multi model structure wo api key this time

app.py

@@ -126,6 +126,18 @@ def generate(model_name,
 
     return generated_sequences[0]
 
+def show_mode(mode):
+  if mode == 'Single Model':
+    return (
+        gr.update(visible=True),
+        gr.update(visible=False)
+        )
+  if mode == 'Multi-Model':
+    return (
+        gr.update(visible=False),
+        gr.update(visible=True)
+    )
+
 def prepare_dataset(dataset):
   dataset = load_dataset(dataset, split='train')
   return dataset
@@ -252,9 +264,14 @@ def upload_flag(*args):
   if flagging_callback.flag(list(args), flag_option = None):
       return gr.update(visible=True)
 
-with gr.Blocks() as demo:
-  gr.Markdown("# Project Interface proposal")
-  gr.Markdown("### Pick a text generation model below, write a prompt and explore the output")
+CSS = """
+#inside_group {
+  padding-top: 0.6em;
+  padding-bottom: 0.6em;
+}
+"""
+
+with gr.Blocks(css=CSS) as demo:
 
   dataset = gr.Variable(value=DATASET)
   prompts_var = gr.Variable(value=None)
@@ -264,76 +281,106 @@ with gr.Blocks() as demo:
   custom_model_path = gr.Variable(value=None)
   flag_choice = gr.Variable(label = "Flag", value=None)
 
-
   flagging_callback = gr.HuggingFaceDatasetSaver(hf_token = HF_AUTH_TOKEN, 
                                                  dataset_name = "fsdlredteam/flagged_2",
                                                  organization = "fsdlredteam",
                                                  private = True )
+  
+  gr.Markdown("# Project Interface proposal")
+  gr.Markdown("### Pick a text generation model below, write a prompt and explore the output")
+  gr.Markdown("### Or compare multiple models")
+  
+  choose_mode = gr.Radio(choices=['Single Model', "Multi-Model"],
+                         value='Single Model',
+                         interactive=True,
+                         visible=True,
+                         show_label=False)
+
+  with gr.Group() as single_model:
+    with gr.Row():
+
+      with gr.Column(scale=1): # input & prompts dataset exploration 
+        gr.Markdown("### 1. Select a prompt", elem_id="inside_group")
+
+        input_text = gr.Textbox(label="Write your prompt below.", 
+                                interactive=True, 
+                                lines=4, 
+                                elem_id="inside_group")
+        
+        gr.Markdown("— or —", elem_id="inside_group")
+        
+        inspo_button = gr.Button('Click here if you need some inspiration', elem_id="inside_group")
 
-  with gr.Row():
+        prompts_drop = gr.Dropdown(visible=False, elem_id="inside_group")
 
-    with gr.Column(scale=1): # input & prompts dataset exploration 
-      gr.Markdown("### 1. Select a prompt")
+        randomize_button = gr.Button('Show another subset', visible=False, elem_id="inside_group")
 
-      input_text = gr.Textbox(label="Write your prompt below.", interactive=True, lines=4)
+            
+      with gr.Column(scale=1): # Model choice & output 
+        gr.Markdown("### 2. Evaluate output")
 
-      gr.Markdown("— or —")
-      
-      inspo_button = gr.Button('Click here if you need some inspiration')
+        
+        model_radio = gr.Radio(choices=list(CHECKPOINTS.keys()),
+                              label='Model',
+                              interactive=True,
+                              elem_id="inside_group")
+        
+        search_bar = gr.Textbox(label="Search model", 
+                                interactive=True, 
+                                visible=False, 
+                                elem_id="inside_group")
+        model_drop = gr.Dropdown(visible=False)
 
-      prompts_drop = gr.Dropdown(visible=False)
+        generate_button = gr.Button('Submit your prompt')
 
-      randomize_button = gr.Button('Show another subset', visible=False)
+        output_spans = gr.HighlightedText(visible=True, label="Generated text", elem_id="inside_group")
 
-           
-    with gr.Column(scale=1): # Model choice & output 
-      gr.Markdown("### 2. Evaluate output")
+        flag_button = gr.Button("Report output here", visible=False)
 
+    with gr.Row(): # Flagging 
       
-      model_radio = gr.Radio(choices=list(CHECKPOINTS.keys()),
-                             label='Model',
-                             interactive=True)
+      with gr.Column(scale=1): 
+        flag_radio = gr.Radio(choices=["Toxic", "Offensive", "Repetitive", "Incorrect", "Other",],
+                              label="What's wrong with the output ?",
+                              interactive=True,
+                              visible=False,
+                              elem_id="inside_group")
       
-      search_bar = gr.Textbox(label="Search model", interactive=True, visible=False)
-      model_drop = gr.Dropdown(visible=False)
-
-      generate_button = gr.Button('Submit your prompt')
-
-      output_spans = gr.HighlightedText(visible=True, label="Generated text")
+        user_comment = gr.Textbox(label="(Optional) Briefly describe the issue", 
+                                  visible=False, 
+                                  interactive=True,
+                                  elem_id="inside_group")
 
-      flag_button = gr.Button("Report output here", visible=False)
-
-  with gr.Row(): # Flagging 
-    
-    with gr.Column(scale=1): 
-      flag_radio = gr.Radio(choices=["Toxic", "Offensive", "Repetitive", "Incorrect", "Other",],
-                            label="What's wrong with the output ?",
-                            interactive=True,
-                            visible=False)
+      confirm_flag_button = gr.Button("Confirm report", visible=False, elem_id="inside_group")
     
-      user_comment = gr.Textbox(label="(Optional) Briefly describe the issue", 
-                                visible=False, 
-                                interactive=True)
-
-    confirm_flag_button = gr.Button("Confirm report", visible=False)
-  
-  with gr.Row(): # Flagging success
-    success_message = gr.Markdown("Your report has been successfully registered. Thank you!", 
-                                 visible=False,)
-
-  with gr.Row(): # Toxicity buttons
-    toxi_button = gr.Button("Run a toxicity analysis of the model's output", visible=False)
-    toxi_button_compare = gr.Button("Compare toxicity on input and output", visible=False)
-
-  with gr.Row(): # Toxicity scores 
-    toxi_scores_input = gr.JSON(label = "Detoxify classification of your input", 
-                                visible=False)
-    toxi_scores_output = gr.JSON(label="Detoxify classification of the model's output", 
-                                visible=False)
-    toxi_scores_compare = gr.JSON(label = "Percentage change between Input and Output", 
-                                visible=False)
+    with gr.Row(): # Flagging success
+      success_message = gr.Markdown("Your report has been successfully registered. Thank you!",
+                                    visible=False,
+                                    elem_id="inside_group")
+
+    with gr.Row(): # Toxicity buttons
+      toxi_button = gr.Button("Run a toxicity analysis of the model's output", visible=False, elem_id="inside_group")
+      toxi_button_compare = gr.Button("Compare toxicity on input and output", visible=False, elem_id="inside_group")
+
+    with gr.Row(): # Toxicity scores 
+      toxi_scores_input = gr.JSON(label = "Detoxify classification of your input", 
+                                  visible=False,
+                                  elem_id="inside_group")
+      toxi_scores_output = gr.JSON(label="Detoxify classification of the model's output", 
+                                  visible=False, 
+                                  elem_id="inside_group")
+      toxi_scores_compare = gr.JSON(label = "Percentage change between Input and Output", 
+                                  visible=False,
+                                  elem_id="inside_group")
+      
+  with gr.Group() as multi_model:
+    gr.Markdown("Model comparison will be here")
 
 
+  choose_mode.change(fn=show_mode,
+                     inputs=choose_mode,
+                     outputs=[single_model, multi_model])
+  
   inspo_button.click(fn=show_dataset, 
                      inputs=dataset, 
                      outputs=[prompts_drop, randomize_button, prompts_var])