make lora target modules configurable and change the default

Author: linoytsaban

examples/advanced_diffusion_training/train_dreambooth_lora_flux_advanced.py

@@ -661,7 +661,8 @@ def parse_args(input_args=None):
         "--lora_blocks",
         type=str,
         default=None,
-        help=('The transformer modules to apply LoRA training on'),
+        help=(
+            'The transformer modules to apply LoRA training on. Please specify the layers in a comma seperated. E.g. - "q_proj,k_proj,v_proj,out_proj" will result in lora training of attention layers only'),
     )
     parser.add_argument(
         "--adam_epsilon",