fix: stream logs to Gradio UI using yield instead of logger only

- Added yield statements for all important messages during training and evaluation
- Replaced yield evaluate_saved_model() with for-loop to stream output
- Removed ineffective return log_buffer.read() which had no effect in Gradio

train_abuse_model.py

@@ -60,10 +60,9 @@ logger = logging.getLogger(__name__)
 
 
 # Check versions
-logger.info("Transformers version:", transformers.__version__)
+logger.info(f"Transformers version: {transformers.__version__}")
 
 # Check for GPU availability
-logger.info("Transformers version: %s", torch.__version__)
 logger.info("torch.cuda.is_available(): %s", torch.cuda.is_available())
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
@@ -194,7 +193,8 @@ model_name = "microsoft/deberta-v3-base"
 def run_training(progress=gr.Progress(track_tqdm=True)):
     if os.path.exists("saved_model/"):
         yield "✅ Trained model found! Skipping training...\n"
-        yield evaluate_saved_model()
+        for line in evaluate_saved_model():
+            yield line
         return
     yield "🚀 Starting training...\n"
     try:
@@ -262,6 +262,9 @@ def run_training(progress=gr.Progress(track_tqdm=True)):
         # Start training!
         trainer.train()
 
+        progress(1.0)
+        yield "✅ Progress: 100%\n"
+
         # Save the model and tokenizer
         MODEL_DIR.mkdir(parents=True, exist_ok=True)
         model.save_pretrained(MODEL_DIR)
@@ -277,12 +280,14 @@ def run_training(progress=gr.Progress(track_tqdm=True)):
     # Evaluation
     try:
         if 'trainer' in locals():
-            evaluate_model_with_thresholds(trainer, test_dataset)
+            for line in evaluate_model_with_thresholds(trainer, test_dataset):
+                yield line
             logger.info("Evaluation completed")
+        logger.info("Evaluation completed")
+        yield "📈 Evaluation completed\n"
     except Exception as e:
         logger.exception(f"Evaluation failed: {e}")
-    log_buffer.seek(0)
-    return log_buffer.read()
+    return
 
 def push_model_to_hub():
     try: