readd generating samples before training.

Author: entrpn

src/maxdiffusion/trainers/wan_trainer.py

@@ -146,7 +146,7 @@ def start_training(self):
     # del pipeline.vae
 
     # Generate a sample before training to compare against generated sample after training.
-    #pretrained_video_path = generate_sample(self.config, pipeline, filename_prefix="pre-training-")
+    pretrained_video_path = generate_sample(self.config, pipeline, filename_prefix="pre-training-")
     mesh = pipeline.mesh
     data_iterator = self.load_dataset(mesh)