[PEFT] Allow PEFT model dict to be loaded (#25721)

* Allow PEFT model dict to be loaded

* make style

* make style

* Apply suggestions from code review

* address comments

* fixup

* final change

* added tests

* fix test

* better logic for handling if adapter has been loaded

* Update tests/peft_integration/test_peft_integration.py
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

---------
Co-authored-by: younesbelkada <younesbelkada@gmail.com>
Co-authored-by: Younes Belkada <49240599+younesbelkada@users.noreply.github.com>
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

src/transformers/integrations/peft.py

@@ -12,13 +12,14 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import inspect
-from typing import Optional
+from typing import TYPE_CHECKING, Any, Dict, Optional
 
 from ..utils import (
     check_peft_version,
     find_adapter_config_file,
     is_accelerate_available,
     is_peft_available,
+    is_torch_available,
     logging,
 )
 
@@ -30,6 +31,11 @@ if is_accelerate_available():
 # Minimum PEFT version supported for the integration
 MIN_PEFT_VERSION = "0.5.0"
 
+if TYPE_CHECKING:
+    if is_torch_available():
+        import torch
+
+
 logger = logging.get_logger(__name__)
 
 
@@ -61,7 +67,7 @@ class PeftAdapterMixin:
 
     def load_adapter(
         self,
-        peft_model_id: str,
+        peft_model_id: Optional[str] = None,
         adapter_name: Optional[str] = None,
         revision: Optional[str] = None,
         token: Optional[str] = None,
@@ -69,6 +75,8 @@ class PeftAdapterMixin:
         max_memory: Optional[str] = None,
         offload_folder: Optional[str] = None,
         offload_index: Optional[int] = None,
+        peft_config: Dict[str, Any] = None,
+        adapter_state_dict: Optional[Dict[str, "torch.Tensor"]] = None,
     ) -> None:
         """
         Load adapter weights from file or remote Hub folder. If you are not familiar with adapters and PEFT methods, we
@@ -77,7 +85,7 @@ class PeftAdapterMixin:
         Requires peft as a backend to load the adapter weights.
 
         Args:
-            peft_model_id (`str`):
+            peft_model_id (`str`, *optional*):
                 The identifier of the model to look for on the Hub, or a local path to the saved adapter config file
                 and adapter weights.
             adapter_name (`str`, *optional*):
@@ -114,6 +122,12 @@ class PeftAdapterMixin:
                 If the `device_map` contains any value `"disk"`, the folder where we will offload weights.
             offload_index (`int`, `optional`):
                 `offload_index` argument to be passed to `accelerate.dispatch_model` method.
+            peft_config (`Dict[str, Any]`, *optional*):
+                The configuration of the adapter to add, supported adapters are non-prefix tuning and adaption prompts
+                methods. This argument is used in case users directly pass PEFT state dicts
+            adapter_state_dict (`Dict[str, torch.Tensor]`, *optional*):
+                The state dict of the adapter to load. This argument is used in case users directly pass PEFT state
+                dicts
         """
         check_peft_version(min_version=MIN_PEFT_VERSION)
 
@@ -122,11 +136,15 @@ class PeftAdapterMixin:
         from peft import PeftConfig, inject_adapter_in_model, load_peft_weights
         from peft.utils import set_peft_model_state_dict
 
-        if not self._hf_peft_config_loaded:
-            self._hf_peft_config_loaded = True
-        elif adapter_name in self.peft_config:
+        if self._hf_peft_config_loaded and adapter_name in self.peft_config:
             raise ValueError(f"Adapter with name {adapter_name} already exists. Please use a different name.")
 
+        if peft_model_id is None and (adapter_state_dict is None and peft_config is None):
+            raise ValueError(
+                "You should either pass a `peft_model_id` or a `peft_config` and `adapter_state_dict` to load an adapter."
+            )
+
+        if peft_config is None:
             adapter_config_file = find_adapter_config_file(
                 peft_model_id,
                 revision=revision,
@@ -139,15 +157,19 @@ class PeftAdapterMixin:
                     "adapter model."
                 )
 
-        loaded_peft_config = PeftConfig.from_pretrained(
+            peft_config = PeftConfig.from_pretrained(
                 peft_model_id,
                 revision=revision,
                 use_auth_token=token,
             )
 
         # Create and add fresh new adapters into the model.
-        inject_adapter_in_model(loaded_peft_config, self, adapter_name)
+        inject_adapter_in_model(peft_config, self, adapter_name)
+
+        if not self._hf_peft_config_loaded:
+            self._hf_peft_config_loaded = True
 
+        if peft_model_id is not None:
             adapter_state_dict = load_peft_weights(peft_model_id, revision=revision, use_auth_token=token)
 
         # We need to pre-process the state dict to remove unneeded prefixes - for backward compatibility

tests/peft_integration/test_peft_integration.py

@@ -16,6 +16,8 @@ import os
 import tempfile
 import unittest
 
+from huggingface_hub import hf_hub_download
+
 from transformers import AutoModelForCausalLM, OPTForCausalLM
 from transformers.testing_utils import require_peft, require_torch, require_torch_gpu, slow, torch_device
 from transformers.utils import is_torch_available
@@ -300,3 +302,33 @@ class PeftIntegrationTester(unittest.TestCase, PeftTesterMixin):
         for model_id in self.peft_test_model_ids:
             pipe = pipeline("text-generation", model_id)
             _ = pipe("Hello")
+
+    def test_peft_add_adapter_with_state_dict(self):
+        """
+        Simple test that tests the basic usage of PEFT model through `from_pretrained`. This test tests if
+        add_adapter works as expected with a state_dict being passed.
+        """
+        from peft import LoraConfig
+
+        dummy_input = torch.LongTensor([[0, 1, 2, 3, 4, 5, 6, 7]]).to(torch_device)
+
+        for model_id, peft_model_id in zip(self.transformers_test_model_ids, self.peft_test_model_ids):
+            for transformers_class in self.transformers_test_model_classes:
+                model = transformers_class.from_pretrained(model_id).to(torch_device)
+
+                peft_config = LoraConfig(init_lora_weights=False)
+
+                with self.assertRaises(ValueError):
+                    model.load_adapter(peft_model_id=None)
+
+                state_dict_path = hf_hub_download(peft_model_id, "adapter_model.bin")
+
+                dummy_state_dict = torch.load(state_dict_path)
+
+                model.load_adapter(adapter_state_dict=dummy_state_dict, peft_config=peft_config)
+                with self.assertRaises(ValueError):
+                    model.load_adapter(model.load_adapter(adapter_state_dict=dummy_state_dict, peft_config=None))
+                self.assertTrue(self._check_lora_correctly_converted(model))
+
+                # dummy generation
+                _ = model.generate(input_ids=dummy_input)