fix threading issue

app.py

@@ -145,23 +145,14 @@ def classify_prompt(prompt: str) -> tuple[str, str]:
     return result_text, label
 
 
-def evaluate_test_set(max_samples: int = None, progress=gr.Progress()) -> str:
-    """Evaluate the model on the test dataset and return metrics.
-    
-    Args:
-        max_samples: Maximum number of samples to evaluate. If None, evaluates on full dataset.
-    """
+def evaluate_test_set(progress=gr.Progress()) -> str:
+    """Evaluate the model on the test dataset and return metrics."""
     if trainer is None or test_tokenized is None:
         return "⚠️ Error: Model or test dataset not loaded."
     
-    # Limit dataset size if specified
+    # Use full test dataset
     eval_dataset = test_tokenized
-    if max_samples is not None and max_samples > 0:
-        max_samples = min(max_samples, len(test_tokenized))
-        eval_dataset = test_tokenized.select(range(max_samples))
-        print(f"Evaluating on {max_samples} samples (out of {len(test_tokenized)} total)")
-    else:
-        print(f"Evaluating on full test set ({len(test_tokenized)} samples)")
+    print(f"Evaluating on full test set ({len(test_tokenized)} samples)")
     
     # Ensure tqdm is enabled for progress tracking
     trainer.args.disable_tqdm = False
@@ -209,12 +200,7 @@ def evaluate_test_set(max_samples: int = None, progress=gr.Progress()) -> str:
     
     # Format results
     output = "## Test Set Evaluation Results\n\n"
-    
-    # Show dataset size info
-    if max_samples is not None and max_samples < len(test_tokenized):
-        output += f"**Note:** Evaluated on {max_samples} samples (out of {len(test_tokenized)} total)\n\n"
-    else:
-        output += f"**Note:** Evaluated on full test set ({len(test_tokenized)} samples)\n\n"
+    output += f"**Note:** Evaluated on full test set ({len(test_tokenized)} samples)\n\n"
     
     # Main metrics
     output += "### Classification Metrics\n\n"
@@ -418,16 +404,8 @@ with app:
         
         # Tab 2: Test Set Evaluation
         with gr.Tab("📊 Evaluate Test Set"):
-            gr.Markdown("### Evaluate the model on the test dataset")
+            gr.Markdown("### Evaluate the model on the full test dataset")
             gr.Markdown("**Note:** Progress percentage will be shown during evaluation.")
-            gr.Markdown("**Tip:** Limit the number of samples for faster evaluation during testing.")
-            
-            max_samples_input = gr.Textbox(
-                label="Maximum samples to evaluate (leave empty for full dataset)",
-                value="",
-                placeholder="Enter a number or leave empty for full dataset",
-                info="Set a limit to evaluate faster. Leave empty to evaluate on the full dataset."
-            )
             
             eval_btn = gr.Button(
                 "Run Evaluation", 
@@ -436,19 +414,9 @@ with app:
             )
             eval_output = gr.Markdown(label="Evaluation Results")
             
-            def run_evaluation(max_samples_str):
+            def run_evaluation():
                 """Run evaluation and return result."""
-                # Handle empty string, None, or invalid values as "use full dataset"
-                max_samples_int = None
-                if max_samples_str and max_samples_str.strip():
-                    try:
-                        max_samples_int = int(max_samples_str.strip())
-                        if max_samples_int < 1:
-                            max_samples_int = None
-                    except (ValueError, TypeError):
-                        max_samples_int = None
-                
-                result = evaluate_test_set(max_samples=max_samples_int)
+                result = evaluate_test_set()
                 return result
             
             def enable_button():
@@ -460,7 +428,6 @@ with app:
                 outputs=eval_btn
             ).then(
                 fn=run_evaluation,
-                inputs=max_samples_input,
                 outputs=eval_output
             ).then(
                 fn=enable_button,