feat: allow offset_noise in dreambooth training example (#2826)

examples/dreambooth/train_dreambooth.py

@@ -417,6 +417,16 @@ def parse_args(input_args=None):
         ),
     )
 
+    parser.add_argument(
+        "--offset_noise",
+        action="store_true",
+        default=False,
+        help=(
+            "Fine-tuning against a modified noise"
+            " See: https://www.crosslabs.org//blog/diffusion-with-offset-noise for more information."
+        ),
+    )
+
     if input_args is not None:
         args = parser.parse_args(input_args)
     else:
@@ -943,7 +953,12 @@ def main(args):
                 latents = latents * vae.config.scaling_factor
 
                 # Sample noise that we'll add to the latents
-                noise = torch.randn_like(latents)
+                if args.offset_noise:
+                    noise = torch.randn_like(latents) + 0.1 * torch.randn(
+                        latents.shape[0], latents.shape[1], 1, 1, device=latents.device
+                    )
+                else:
+                    noise = torch.randn_like(latents)
                 bsz = latents.shape[0]
                 # Sample a random timestep for each image
                 timesteps = torch.randint(0, noise_scheduler.config.num_train_timesteps, (bsz,), device=latents.device)