Reviewer comments

src/peft/peft_model.py

@@ -57,7 +57,6 @@
     _prepare_prompt_learning_config,
     _set_adapter,
     _set_trainable,
-    get_modules_to_save_from_config,
     get_peft_model_state_dict,
     id_tensor_storage,
     infer_device,
@@ -954,7 +953,7 @@ def set_additional_trainable_modules(self, peft_config, adapter_name):
             else:
                 self.modules_to_save.update(peft_config.modules_to_save)
             # this may add a new ModulesToSaveWrapper
-            _set_trainable(self, adapter_name, module_names=get_modules_to_save_from_config(peft_config))
+            _set_trainable(self, adapter_name, module_names=getattr(peft_config, "modules_to_save", None))
 
         if getattr(peft_config, "trainable_token_indices", None) is not None:
             if isinstance(peft_config.trainable_token_indices, dict):
@@ -1492,6 +1491,9 @@ def __init__(
             else:
                 peft_config.modules_to_save.extend(classifier_module_names)
 
+        # The modification of peft_config must happen before the init call as the `modules_to_save` information
+        # will be used to guard the target layer matching against matching `modules_to_save` layers. Only the
+        # config is relevant for this, the `modules_to_save` attribute can follow later.
         super().__init__(model, peft_config, adapter_name, **kwargs)
 
         if self.modules_to_save is None:
@@ -1505,7 +1507,7 @@ def __init__(
                 break
 
         # to make sure classifier layer is trainable; this may add a new ModulesToSaveWrapper
-        _set_trainable(self, adapter_name, module_names=get_modules_to_save_from_config(peft_config))
+        _set_trainable(self, adapter_name, module_names=getattr(peft_config, "modules_to_save", None))
 
     def add_adapter(self, adapter_name: str, peft_config: PeftConfig, low_cpu_mem_usage: bool = False) -> None:
         """
@@ -2296,7 +2298,7 @@ def __init__(
                 break
 
         # to make sure classifier layer is trainable; this may add a new ModulesToSaveWrapper
-        _set_trainable(self, adapter_name, module_names=get_modules_to_save_from_config(peft_config))
+        _set_trainable(self, adapter_name, module_names=getattr(peft_config, "modules_to_save", None))
 
     def add_adapter(self, adapter_name: str, peft_config: PeftConfig, low_cpu_mem_usage: bool = False) -> None:
         """
@@ -2517,7 +2519,7 @@ def __init__(
                 break
 
         # to make sure classifier layer is trainable; this may add a new ModulesToSaveWrapper
-        _set_trainable(self, adapter_name, module_names=get_modules_to_save_from_config(peft_config))
+        _set_trainable(self, adapter_name, module_names=getattr(peft_config, "modules_to_save", None))
 
     def add_adapter(self, adapter_name: str, peft_config: PeftConfig, low_cpu_mem_usage: bool = False) -> None:
         """

src/peft/tuners/tuners_utils.py

@@ -38,7 +38,6 @@
     SEQ_CLS_HEAD_NAMES,
 )
 from peft.utils.integrations import init_empty_weights
-from peft.utils.other import get_modules_to_save_from_config
 from peft.utils.peft_types import PeftType, TaskType
 
 from ..config import PeftConfig
@@ -513,7 +512,7 @@ def inject_adapter(
                 # All targeted modules were excluded
                 raise ValueError(
                     "All modules were excluded. This is likely unintended. "
-                    "Check your `target_modules` and `exclude_modules` configuration."
+                    "Check your `target_modules`, `exclude_modules` and `modules_to_save` configuration."
                 )
             elif not excluded_modules and unmatched_modules:
                 # None of the targeted modules matched
@@ -1016,7 +1015,7 @@ def check_target_module_exists(config, key: str) -> bool | re.Match[str] | None:
     # Adapters should never match on modules to save modules as it is a guarantee for conflicts of behavior
     # between `ModulesToSaveWrapper` internals and the potential adapter.
     # TODO extend this to AuxiliaryTrainingWrapper in this PR if possible
-    modules_to_save = get_modules_to_save_from_config(config)
+    modules_to_save = getattr(config, "modules_to_save", None)
     if modules_to_save:
         if any(re.match(rf"(^|.*\.){m}($|\..*)", key) for m in modules_to_save):
             return _ExcludedModule()

src/peft/utils/__init__.py

@@ -42,7 +42,6 @@
     cast_mixed_precision_params,
     get_auto_gptq_quant_linear,
     get_gptqmodel_quant_linear,
-    get_modules_to_save_from_config,
     get_quantization_config,
     id_tensor_storage,
     infer_device,
@@ -84,7 +83,6 @@
     "cast_mixed_precision_params",
     "get_auto_gptq_quant_linear",
     "get_gptqmodel_quant_linear",
-    "get_modules_to_save_from_config",
     "get_peft_model_state_dict",
     "get_quantization_config",
     "id_tensor_storage",

src/peft/utils/other.py

@@ -751,7 +751,7 @@ def _set_trainable(
 
     if not module_names:
         # This is useful for the case that the PEFT config does not have `modules_to_save`, e.g.
-        # in the case of prompt tuning and friends as returned by `get_modules_to_save_from_config`.
+        # in the case of prompt tuning and friends.
         return
 
     trainable_modules = []
@@ -1126,12 +1126,3 @@ def get_pattern_key(pattern_keys: Sequence[str], key_to_match: str) -> str:
         return key
 
     return key_to_match
-
-
-def get_modules_to_save_from_config(peft_config) -> Optional[list[str]]:
-    """Utility for retrieving `modules_to_save` from a PEFT config.
-    This is useful for exceptional tuners like `PromptTuning` which do not have this attribute.
-    """
-    if hasattr(peft_config, "modules_to_save"):
-        return peft_config.modules_to_save
-    return None

tests/test_seq_classifier.py

@@ -13,9 +13,7 @@
 
 import pytest
 import torch
-from transformers import (
-    AutoModelForSequenceClassification,
-)
+from transformers import AutoModelForSequenceClassification
 
 from peft import (
     AdaLoraConfig,
@@ -181,10 +179,8 @@
 
 class TestSequenceClassificationModels(PeftCommonTester):
     r"""
-    Test if the PeftModel behaves as expected. This includes:
-    - test if the model has the expected methods
-
-    We use pytest for debugging purposes to test each model individually.
+    Tests for basic coverage of AutoModelForSequenceClassification and classification-specific cases.
+    Most of the functionality is probably already covered by other tests.
     """
 
     transformers_class = AutoModelForSequenceClassification