diff --git a/studio/backend/core/training/diffusion_lora_trainer.py b/studio/backend/core/training/diffusion_lora_trainer.py
index 10b6cd89cd..84bbb6bf74 100644
--- a/studio/backend/core/training/diffusion_lora_trainer.py
+++ b/studio/backend/core/training/diffusion_lora_trainer.py
@@ -512,9 +512,7 @@ def run_diffusion_lora_training(
             grad_norm = None
             if cfg.max_grad_norm and cfg.max_grad_norm > 0:
                 # The returned value is the total PRE-clip norm, reported to the UI chart.
-                grad_norm = float(
-                    torch.nn.utils.clip_grad_norm_(lora_params, cfg.max_grad_norm)
-                )
+                grad_norm = float(torch.nn.utils.clip_grad_norm_(lora_params, cfg.max_grad_norm))
             optimizer.step()
             lr_sched.step()
 
diff --git a/studio/backend/core/training/diffusion_training_service.py b/studio/backend/core/training/diffusion_training_service.py
index 8fef55641e..0f57f4d4ca 100644
--- a/studio/backend/core/training/diffusion_training_service.py
+++ b/studio/backend/core/training/diffusion_training_service.py
@@ -139,7 +139,11 @@ def _idle_state() -> dict[str, Any]:
 
 
 def _append_metric(
-    state: dict[str, Any], step: Any, loss: Any, lr: Any, grad_norm: Any = None
+    state: dict[str, Any],
+    step: Any,
+    loss: Any,
+    lr: Any,
+    grad_norm: Any = None,
 ) -> None:
     """Append one (step, loss, lr, grad_norm) point to the bounded history arrays on
     ``state``.