add test case

src/llamafactory/model/model_utils/checkpointing.py

@@ -19,7 +19,7 @@
 # limitations under the License.
 
 import inspect
-from functools import partial
+from functools import partial, wraps
 from types import MethodType
 from typing import TYPE_CHECKING, Any, Dict, Optional, Tuple, Union
 
@@ -73,6 +73,25 @@ class UnslothGradientCheckpointing(torch.autograd.Function):
         return (None, hidden_states.grad) + (None,) * len(ctx.args)
 
 
+def get_custom_gradient_checkpointing_func(gradient_checkpointing_func):
+    r"""
+    Only applies gradient checkpointing to trainable layers.
+    """
+
+    @wraps(gradient_checkpointing_func)
+    def custom_gradient_checkpointing_func(func, *args: Union["torch.Tensor", Any], **kwargs):
+        module: "torch.nn.Module" = func.__self__
+
+        if any(param.requires_grad for param in module.parameters()):
+            for arg in args:
+                if torch.is_tensor(arg) and torch.is_floating_point(arg):
+                    arg.requires_grad_(True)
+
+        return gradient_checkpointing_func(func, *args, **kwargs)
+
+    return custom_gradient_checkpointing_func
+
+
 def _gradient_checkpointing_enable(
     self: "PreTrainedModel",
     gradient_checkpointing_kwargs: Optional[Dict[str, Any]] = None,
@@ -96,22 +115,13 @@ def _gradient_checkpointing_enable(
     else:
         gradient_checkpointing_func = partial(checkpoint, **gradient_checkpointing_kwargs)
 
-    def custom_gradient_checkpointing_func(func, *args: Union["torch.Tensor", Any], **kwargs):
+    gradient_checkpointing_func = get_custom_gradient_checkpointing_func(gradient_checkpointing_func)
-        module: "torch.nn.Module" = func.__self__
-
-        if any(param.requires_grad for param in module.parameters()):
-            for arg in args:
-                if torch.is_tensor(arg) and torch.is_floating_point(arg):
-                    arg.requires_grad_(True)
-
-        return gradient_checkpointing_func(func, *args, **kwargs)
-
     if "value" in inspect.signature(self._set_gradient_checkpointing).parameters:  # old GC format
         self.apply(partial(self._set_gradient_checkpointing, value=True))
         self.enable_input_require_grads()
         logger.warning("You are using the old GC format, some features (e.g. BAdam) will be invalid.")
     else:  # have already enabled input require gradients
-        self._set_gradient_checkpointing(enable=True, gradient_checkpointing_func=custom_gradient_checkpointing_func)
+        self._set_gradient_checkpointing(enable=True, gradient_checkpointing_func=gradient_checkpointing_func)
 
 
 def _fp32_forward_post_hook(

tests/model/model_utils/test_checkpointing.py

@@ -51,6 +51,12 @@ def test_checkpointing_disable():
         assert getattr(module, "gradient_checkpointing") is False
 
 
+def test_unsloth_gradient_checkpointing():
+    model = load_train_model(use_unsloth_gc=True, **TRAIN_ARGS)
+    for module in filter(lambda m: hasattr(m, "gradient_checkpointing"), model.modules()):
+        assert module._gradient_checkpointing_func.__self__.__name__ == "UnslothGradientCheckpointing"
+
+
 def test_upcast_layernorm():
     model = load_train_model(upcast_layernorm=True, **TRAIN_ARGS)
     for name, param in model.named_parameters():