refresh text box in gradio

app.py

@@ -77,7 +77,8 @@ def run_full_evaluation_gradio():
         return
 
     try:
-        yield "Starting full evaluation... This will process 25,000 samples and may take 10-20 minutes. Please be patient."
+        accumulated_text = "Starting full evaluation... This will process 25,000 samples and may take 10-20 minutes. Please be patient.\n"
+        yield accumulated_text
         
         device = sentiment_inferer.device
         model = sentiment_inferer.model
@@ -87,7 +88,8 @@ def run_full_evaluation_gradio():
 
         yield "Loading IMDB test dataset (this might take a moment)..."
         imdb_test_full = load_dataset("imdb", split="test")
-        yield f"IMDB test dataset loaded ({len(imdb_test_full)} samples). Tokenizing dataset..."
+        accumulated_text += f"IMDB test dataset loaded ({len(imdb_test_full)} samples). Tokenizing dataset...\n"
+        yield accumulated_text
 
         def tokenize_function(examples):
             tokenized_output = tokenizer(examples["text"], truncation=True, padding="max_length", max_length=max_length)
@@ -100,11 +102,11 @@ def run_full_evaluation_gradio():
         tokenized_imdb_test_full.set_format("torch", columns=["input_ids", "attention_mask", "labels", "lengths"])
 
         test_dataloader_full = DataLoader(tokenized_imdb_test_full, batch_size=batch_size)
-        yield "Dataset tokenized and DataLoader prepared. Starting model evaluation on the test set..."
+        accumulated_text += "Dataset tokenized and DataLoader prepared. Starting model evaluation on the test set...\n"
+        yield accumulated_text
 
         # The 'evaluate' function from evaluation.py is now a generator.
-        # Iterate through its yielded updates and results.
-        final_results_str = ""
+        # Iterate through its yielded updates and results, accumulating text.
         for update in evaluate(model, test_dataloader_full, device):
             if isinstance(update, dict):
                 # This is the final results dictionary
@@ -115,16 +117,13 @@ def run_full_evaluation_gradio():
                     else:
                         results_str += f"{key.capitalize()}: {value}\n"
                 results_str += "\nEvaluation finished."
-                final_results_str = results_str # Store to yield last
-                yield results_str # Optionally yield intermediate dict if needed, or just final string
+                accumulated_text += results_str
+                yield accumulated_text 
                 break # Stop after getting the results dict
             else:
                 # This is a progress string
-                yield str(update) + "\n" # Append newline to each progress string
-        
-        # Ensure the final formatted results string is yielded if not already (e.g., if loop broke early)
-        # However, the logic above should yield it before breaking.
-        # If evaluate could end without yielding a dict, this might be needed.
+                accumulated_text += str(update) + "\n" # Append newline to each progress string
+                yield accumulated_text
 
     except Exception as e:
         import traceback