Improve LoRA training UX - Auto-populate and navigation

1. Auto-populate LoRA Adapter Name when dataset is selected
- Generates suggested name from dataset key + timestamp
- Updates automatically on dataset selection change

2. Add 'Load for Training' button in Manage LoRA Adapters
- Select existing LoRA and click to prepare for continued training
- Auto-fills LoRA name with versioned suggestion (e.g., 'model_v2_1234')
- Enables 'Continue training' checkbox
- Selects the base LoRA adapter automatically

3. Improved base LoRA dropdown visibility
- Already toggles on checkbox change
- Now properly populated when using 'Load for Training'

User flow improvement:
- Dataset Training Auto-named LoRA ready to train
- Manage LoRAs Load for Training Training tab ready with all fields set

app.py

@@ -1517,6 +1517,49 @@ def toggle_base_lora(use_existing):
     """Toggle visibility of base LoRA adapter dropdown"""
     return gr.Dropdown(visible=use_existing)
 
+def auto_populate_lora_name(dataset_selection):
+    """Auto-populate LoRA name when dataset is selected"""
+    try:
+        if not dataset_selection or dataset_selection in ["No prepared datasets available", "Error loading datasets"]:
+            return ""
+        
+        # Extract dataset key from display name (format: "dataset_key (N samples)")
+        dataset_key = dataset_selection.split(" (")[0].strip()
+        
+        # Generate suggested LoRA name from dataset
+        import time
+        timestamp = int(time.time() % 10000)  # Last 4 digits
+        suggested_name = f"{dataset_key}_lora_{timestamp}"
+        
+        return suggested_name
+        
+    except Exception as e:
+        logger.error(f"Failed to auto-populate LoRA name: {e}")
+        return ""
+
+def load_lora_for_training(selected_lora):
+    """Load selected LoRA for continued training - navigate to training tab"""
+    try:
+        if not selected_lora:
+            return gr.Textbox(value="❌ No LoRA selected"), gr.Textbox(value=""), gr.Checkbox(value=False), gr.Dropdown(value=None)
+        
+        # Generate suggested name for continued training
+        import time
+        timestamp = int(time.time() % 10000)
+        suggested_name = f"{selected_lora}_v2_{timestamp}"
+        
+        # Return updates: status message, new LoRA name, enable checkbox, select base LoRA
+        return (
+            gr.Textbox(value=f"✅ Ready to continue training '{selected_lora}'"),
+            gr.Textbox(value=suggested_name),
+            gr.Checkbox(value=True),
+            gr.Dropdown(value=selected_lora)
+        )
+        
+    except Exception as e:
+        logger.error(f"Failed to load LoRA for training: {e}")
+        return gr.Textbox(value=f"❌ Error: {str(e)}"), gr.Textbox(value=""), gr.Checkbox(value=False), gr.Dropdown(value=None)
+
 def export_dataset(dataset_key):
     """Export prepared dataset as zip file"""
     try:
@@ -2256,6 +2299,7 @@ with gr.Blocks(
                 )
                 
                 with gr.Row():
+                    load_for_training_btn = gr.Button("🎓 Load for Training", variant="secondary", size="lg", scale=1)
                     download_lora_btn = gr.Button("⬇️ Download LoRA", variant="primary", size="lg", scale=1)
                     delete_lora_btn = gr.Button("🗑️ Delete LoRA", variant="stop", size="lg", scale=1)
                 
@@ -2362,6 +2406,20 @@ with gr.Blocks(
         outputs=[selected_dataset]
     )
     
+    # Auto-populate LoRA name when dataset is selected
+    selected_dataset.change(
+        fn=auto_populate_lora_name,
+        inputs=[selected_dataset],
+        outputs=[lora_name_input]
+    )
+    
+    # Load LoRA for continued training from Manage tab
+    load_for_training_btn.click(
+        fn=load_lora_for_training,
+        inputs=[selected_lora_for_action],
+        outputs=[lora_action_status, lora_name_input, use_existing_lora, base_lora_adapter]
+    )
+    
     start_training_btn.click(
         fn=start_lora_training,
         inputs=[lora_name_input, selected_dataset, batch_size, learning_rate, num_epochs, lora_rank, lora_alpha],