formatting and minor text changes

app.py

@@ -51,7 +51,8 @@ GENDERED_LIST = [
 
 # %%
 # Fire up the models
-models = {m : pipeline("fill-mask", model=m) for m in MODEL_NAMES if m != OWN_MODEL_NAME}
+models = {m: pipeline("fill-mask", model=m)
+          for m in MODEL_NAMES if m != OWN_MODEL_NAME}
 
 # %%
 # Get the winogender sentences
@@ -60,6 +61,8 @@ occs = sorted(list({sentence_id.split('_')[0]
                     for sentence_id in winogender_sentences}))
 
 # %%
+
+
 def get_gendered_token_ids():
     male_gendered_tokens = [list[0] for list in GENDERED_LIST]
     female_gendered_tokens = [list[1] for list in GENDERED_LIST]
@@ -107,7 +110,8 @@ def get_figure(df, model_name, occ):
     ax.axis('tight')
     ax.set_xlabel("Sentence number")
     ax.set_ylabel("Uncertainty metric")
-    ax.set_title(f"{MODEL_NAME_DICT[model_name]} gender pronoun uncertainty in '{occ}' sentences")
+    ax.set_title(
+        f"{MODEL_NAME_DICT[model_name]} gender pronoun uncertainty in '{occ}' sentences")
     return fig
 
 
@@ -127,8 +131,8 @@ def predict_gender_pronouns(
 
     # For debugging
     print('input_texts', texts)
-    
-    if model_name is None or model_name == '': 
+
+    if model_name is None or model_name == '':
         model_name = MODEL_NAMES[0]
         model = models[model_name]
     elif model_name == OWN_MODEL_NAME:
@@ -213,10 +217,9 @@ with demo:
         we are able to identify likely spurious correlations and exploit them in \
         the scenario of gender underspecified tasks. (Note that introspecting softmax probabilities alone is insufficient, as in the sentences \
         below, LLMs may report a softmax prob of ~0.9 despite the task being underspecified.)")
-
     gr.Markdown("We extend the [Winogender Schemas](https://github.com/rudinger/winogender-schemas) evaluation set to produce\
         eight syntactically similar sentences. However semantically, \
-        only two of the sentences are gender-specified while the rest remain gender-underspecified")
+        only two of the sentences are well-specified while the rest remain underspecified.")
     gr.Markdown("If a model can reliably tell us when it is uncertain about its predictions, one can replace only those uncertain predictions with\
         an appropriate heuristic.")
 
@@ -241,11 +244,11 @@ with demo:
             lines=2,
             label=f"...If you selected '{PICK_YOUR_OWN_LABEL}' above, add your own texts new-line delimited sentences here. Be sure\
             to include a single MASK-ed out pronoun. \
-            If unsure on the required format, click an occupation above instead, to see some example input texts for this round.",
+            If unsure on the required format, click an occupation above instead, to see some example input texts for this round."
         )
 
     with gr.Row():
-        get_text_btn = gr.Button("Load input texts")
+        get_text_btn = gr.Button("1) Load input texts")
 
     get_text_btn.click(
         fn=display_input_texts,
@@ -256,7 +259,7 @@ with demo:
     )
 
     with gr.Row():
-        uncertain_btn = gr.Button("Get uncertainty results!")
+        uncertain_btn = gr.Button("2) Get uncertainty results!")
     gr.Markdown(
         "If there is an * by a sentence number, then at least one top prediction for that sentence was non-gendered.")
 
@@ -277,4 +280,4 @@ with demo:
 
 demo.launch(debug=True)
 
-# %%
+# %%