Update app.py

app.py

@@ -53,13 +53,12 @@ EMBEDDERS = {
 _CORPUS_CACHE = {}
 
 def ensure_corpus_embeddings(model_name: str, texts: list[str]):
-    """Compute & cache corpus embeddings for a given model name."""
     if model_name in _CORPUS_CACHE:
         return _CORPUS_CACHE[model_name]
     model_id = EMBEDDERS[model_name]
     model = load_sentence_model(model_id)
-    # encode with no progress bar to keep logs clean on Spaces
-    emb = model.encode(texts, show_progress_bar=False, convert_to_numpy=True, normalize_embeddings=True)
+    emb = model.encode(texts, show_progress_bar=False,
+                       convert_to_numpy=True, normalize_embeddings=True)
     _CORPUS_CACHE[model_name] = emb
     return emb
 
@@ -89,7 +88,14 @@ def top3_for_each_model(user_input: str, selected_models: list[str]):
 
 def generate_and_pick_best(prompt: str, n_sequences: int, max_length: int, temperature: float, scorer_model_name: str):
     gen = load_generator()
-    outputs = gen(prompt, max_length=max_length, num_return_sequences=n_sequences, do_sample=True, temperature=temperature)
+    outputs = gen(
+        prompt,
+        max_length=max_length,
+        num_return_sequences=n_sequences,
+        do_sample=True,
+        temperature=temperature,
+        pad_token_id=50256,   # <- added
+    )
     candidates = [o["generated_text"].strip() for o in outputs]
 
     scorer_id = EMBEDDERS[scorer_model_name]
@@ -128,7 +134,7 @@ Small, reliable demo for your final project:
         )
 
     run_btn = gr.Button("🔎 Find Top‑3 Similar Tweets")
-    table_out = gr.Dataframe(interactive=False, wrap=True)
+    table_out = gr.Dataframe(interactive=False, overflow_row_behaviour="paginate")  # <- changed
 
     run_btn.click(top3_for_each_model, inputs=[test_input, models], outputs=table_out)
 
@@ -144,7 +150,7 @@ Small, reliable demo for your final project:
     gen_btn = gr.Button("✨ Generate & Score")
     best_txt = gr.Textbox(label="Best generated tweet")
     best_score = gr.Number(label="Similarity (best)")
-    gen_table = gr.Dataframe(interactive=False, wrap=True)
+    gen_table = gr.Dataframe(interactive=False, overflow_row_behaviour="paginate")  # <- changed
 
     gen_btn.click(generate_and_pick_best,
                   inputs=[test_input, n_seq, max_len, temp, scorer_model],