diff --git a/docs/tutorials/qml/ml_tools.md b/docs/tutorials/qml/ml_tools.md
index 019224ae..5efb6cb7 100644
--- a/docs/tutorials/qml/ml_tools.md
+++ b/docs/tutorials/qml/ml_tools.md
@@ -59,7 +59,7 @@ use `train_with_grad` as example but the code can be used directly with the grad
 As every other training routine commonly used in Machine Learning, it requires
 `model`, `data` and an `optimizer` as input arguments.
 However, in addition, it requires a `loss_fn` and a `TrainConfig`.
-A `loss_fn` is required to be a function which expects both a model and data and returns a tuple of (loss, metrics: `<dict>`), where `metrics` is a dict of scalars which can be customized too.
+A `loss_fn` is required to be a function which expects both a model and data and returns a tuple of (loss, metrics: `<dict>`, ...), where `metrics` is a dict of scalars which can be customized too. It can optionally also return additional values which are utilised by the corresponding user-provided `optimize_step` function inside `train_with_grad`.
 
 ```python exec="on" source="material-block"
 import torch
diff --git a/qadence/ml_tools/train_grad.py b/qadence/ml_tools/train_grad.py
index 8cd32e8d..86c9ed8e 100644
--- a/qadence/ml_tools/train_grad.py
+++ b/qadence/ml_tools/train_grad.py
@@ -48,7 +48,7 @@ def train(
             the model
         optimizer: The optimizer to use.
         config: `TrainConfig` with additional training options.
-        loss_fn: Loss function returning (loss: float, metrics: dict[str, float])
+        loss_fn: Loss function returning (loss: float, metrics: dict[str, float], ...)
         device: String defining device to train on, pass 'cuda' for GPU.
         optimize_step: Customizable optimization callback which is called at every iteration.=
             The function must have the signature `optimize_step(model,
@@ -224,7 +224,7 @@ def loss_fn(model: torch.nn.Module, data: torch.Tensor) -> tuple[torch.Tensor, d
                     if iteration % config.val_every == 0:
                         xs = next(dl_iter_val)
                         xs_to_device = data_to_device(xs, device=device, dtype=data_dtype)
-                        val_loss, _ = loss_fn(model, xs_to_device)
+                        val_loss, *_ = loss_fn(model, xs_to_device)
                         if config.validation_criterion(val_loss, best_val_loss, config.val_epsilon):  # type: ignore[misc]
                             best_val_loss = val_loss
                             if config.folder and config.checkpoint_best_only:
@@ -245,7 +245,7 @@ def loss_fn(model: torch.nn.Module, data: torch.Tensor) -> tuple[torch.Tensor, d
         try:
             xs = next(dl_iter) if dataloader is not None else None  # type: ignore[arg-type]
             xs_to_device = data_to_device(xs, device=device, dtype=data_dtype)
-            loss, metrics = loss_fn(model, xs_to_device)
+            loss, metrics, *_ = loss_fn(model, xs_to_device)
             if iteration % config.print_every == 0 and config.verbose:
                 print_metrics(loss, metrics, iteration)