is_masked optimization, rem dep args

app.py

@@ -346,7 +346,8 @@ def predict_gender_pronouns(
         male_gendered_token_ids, 
         female_gendered_token_ids
     )
-    num_preds = torch.sum(tokenized['ids'][0] == MASK_TOKEN_ID).item()
+    initial_is_masked = tokenized['ids'][0] == MASK_TOKEN_ID
+    num_preds = torch.sum(initial_is_masked).item()
 
     female_dfs = []
     male_dfs = []
@@ -359,7 +360,10 @@ def predict_gender_pronouns(
         female_pronoun_preds = []
         male_pronoun_preds = []
         for indie_var_idx in range(len(tokenized['ids'])):
-            is_masked = tokenized['ids'][indie_var_idx] == MASK_TOKEN_ID
+            if dataset == WIKIBIO:
+                is_masked = initial_is_masked  # injected text all same token length
+            else: 
+                is_masked = tokenized['ids'][indie_var_idx] == MASK_TOKEN_ID
 
             ids = tokenized["ids"][indie_var_idx]
             atten_mask = tokenized["atten_mask"][indie_var_idx]
@@ -474,15 +478,17 @@ gr.Interface(
     outputs=[
         gr.outputs.Textbox(
             type="auto", label="Sample target text fed to model"),
-        gr.outputs.Plot(type="auto", label="Plot of softmax probability pronouns predicted female."),
-        gr.outputs.Plot(type="auto", label="Plot of softmax probability pronouns predicted male."),
+        gr.Plot(type="auto", label="Plot of softmax probability pronouns predicted female."),
+        gr.Plot(type="auto", label="Plot of softmax probability pronouns predicted male."),
         gr.outputs.Dataframe(
+            show_label=True,
             overflow_row_behaviour="show_ends",
             label="Table of softmax probability pronouns predicted female",
         ),
         gr.outputs.Dataframe(
+            show_label=True,
             overflow_row_behaviour="show_ends",
             label="Table of softmax probability pronouns predicted male",
         ),
     ],
-).launch(debug=True)
+).launch(debug=True, share=True)