PEFT Integration for Text Encoder to handle multiple alphas/ranks, disable/enable adapters and support for multiple adapters (huggingface#5147)

* more fixes

* up

* up

* style

* add in setup

* oops

* more changes

* v1 rzfactor CI

* Apply suggestions from code review

Co-authored-by: Patrick von Platen <[email protected]>

* few todos

* protect torch import

* style

* fix fuse text encoder

* Update src/diffusers/loaders.py

Co-authored-by: Sayak Paul <[email protected]>

* replace with `recurse_replace_peft_layers`

* keep old modules for BC

* adjustments on `adjust_lora_scale_text_encoder`

* nit

* move tests

* add conversion utils

* remove unneeded methods

* use class method instead

* oops

* use `base_version`

* fix examples

* fix CI

* fix weird error with python 3.8

* fix

* better fix

* style

* Apply suggestions from code review

Co-authored-by: Sayak Paul <[email protected]>
Co-authored-by: Patrick von Platen <[email protected]>

* Apply suggestions from code review

Co-authored-by: Patrick von Platen <[email protected]>

* add comment

* Apply suggestions from code review

Co-authored-by: Sayak Paul <[email protected]>

* conv2d support for recurse remove

* added docstrings

* more docstring

* add deprecate

* revert

* try to fix merge conflicts

* peft integration features for text encoder

1. support multiple rank/alpha values
2. support multiple active adapters
3. support disabling and enabling adapters

* fix bug

* fix code quality

* Apply suggestions from code review

Co-authored-by: Younes Belkada <[email protected]>

* fix bugs

* Apply suggestions from code review

Co-authored-by: Younes Belkada <[email protected]>

* address comments

Co-Authored-By: Benjamin Bossan <[email protected]>
Co-Authored-By: Patrick von Platen <[email protected]>

* fix code quality

* address comments

* address comments

* Apply suggestions from code review

* find and replace

---------

Co-authored-by: younesbelkada <[email protected]>
Co-authored-by: Younes Belkada <[email protected]>
Co-authored-by: Patrick von Platen <[email protected]>
Co-authored-by: Sayak Paul <[email protected]>
Co-authored-by: Benjamin Bossan <[email protected]>

Author: 5 people

src/diffusers/loaders.py

@@ -35,18 +35,23 @@
     convert_state_dict_to_diffusers,
     convert_state_dict_to_peft,
     deprecate,
+    get_adapter_name,
+    get_peft_kwargs,
     is_accelerate_available,
     is_omegaconf_available,
     is_peft_available,
     is_transformers_available,
     logging,
     recurse_remove_peft_layers,
+    scale_lora_layers,
+    set_adapter_layers,
+    set_weights_and_activate_adapters,
 )
 from .utils.import_utils import BACKENDS_MAPPING
 
 
 if is_transformers_available():
-    from transformers import CLIPTextModel, CLIPTextModelWithProjection
+    from transformers import CLIPTextModel, CLIPTextModelWithProjection, PreTrainedModel
 
 if is_accelerate_available():
     from accelerate import init_empty_weights
@@ -1100,7 +1105,9 @@ class LoraLoaderMixin:
     num_fused_loras = 0
     use_peft_backend = USE_PEFT_BACKEND
 
-    def load_lora_weights(self, pretrained_model_name_or_path_or_dict: Union[str, Dict[str, torch.Tensor]], **kwargs):
+    def load_lora_weights(
+        self, pretrained_model_name_or_path_or_dict: Union[str, Dict[str, torch.Tensor]], adapter_name=None, **kwargs
+    ):
         """
         Load LoRA weights specified in `pretrained_model_name_or_path_or_dict` into `self.unet` and
         `self.text_encoder`.
@@ -1120,6 +1127,9 @@ def load_lora_weights(self, pretrained_model_name_or_path_or_dict: Union[str, Di
                 See [`~loaders.LoraLoaderMixin.lora_state_dict`].
             kwargs (`dict`, *optional*):
                 See [`~loaders.LoraLoaderMixin.lora_state_dict`].
+            adapter_name (`str`, *optional*):
+                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
+                `default_{i}` where i is the total number of adapters being loaded.
         """
         # First, ensure that the checkpoint is a compatible one and can be successfully loaded.
         state_dict, network_alphas = self.lora_state_dict(pretrained_model_name_or_path_or_dict, **kwargs)
@@ -1143,6 +1153,7 @@ def load_lora_weights(self, pretrained_model_name_or_path_or_dict: Union[str, Di
             text_encoder=self.text_encoder,
             lora_scale=self.lora_scale,
             low_cpu_mem_usage=low_cpu_mem_usage,
+            adapter_name=adapter_name,
             _pipeline=self,
         )
 
@@ -1500,6 +1511,7 @@ def load_lora_into_text_encoder(
         prefix=None,
         lora_scale=1.0,
         low_cpu_mem_usage=None,
+        adapter_name=None,
         _pipeline=None,
     ):
         """
@@ -1523,6 +1535,9 @@ def load_lora_into_text_encoder(
                 tries to not use more than 1x model size in CPU memory (including peak memory) while loading the model.
                 Only supported for PyTorch >= 1.9.0. If you are using an older version of PyTorch, setting this
                 argument to `True` will raise an error.
+            adapter_name (`str`, *optional*):
+                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
+                `default_{i}` where i is the total number of adapters being loaded.
         """
         low_cpu_mem_usage = low_cpu_mem_usage if low_cpu_mem_usage is not None else _LOW_CPU_MEM_USAGE_DEFAULT
 
@@ -1584,19 +1599,22 @@ def load_lora_into_text_encoder(
                 if cls.use_peft_backend:
                     from peft import LoraConfig
 
-                    lora_rank = list(rank.values())[0]
-                    # By definition, the scale should be alpha divided by rank.
-                    # https://github.com/huggingface/peft/blob/ba0477f2985b1ba311b83459d29895c809404e99/src/peft/tuners/lora/layer.py#L71
-                    alpha = lora_scale * lora_rank
+                    lora_config_kwargs = get_peft_kwargs(rank, network_alphas, text_encoder_lora_state_dict)
 
-                    target_modules = ["q_proj", "k_proj", "v_proj", "out_proj"]
-                    if patch_mlp:
-                        target_modules += ["fc1", "fc2"]
+                    lora_config = LoraConfig(**lora_config_kwargs)
 
-                    # TODO: support multi alpha / rank: https://github.com/huggingface/peft/pull/873
-                    lora_config = LoraConfig(r=lora_rank, target_modules=target_modules, lora_alpha=alpha)
+                    # adapter_name
+                    if adapter_name is None:
+                        adapter_name = get_adapter_name(text_encoder)
 
-                    text_encoder.load_adapter(adapter_state_dict=text_encoder_lora_state_dict, peft_config=lora_config)
+                    # inject LoRA layers and load the state dict
+                    text_encoder.load_adapter(
+                        adapter_name=adapter_name,
+                        adapter_state_dict=text_encoder_lora_state_dict,
+                        peft_config=lora_config,
+                    )
+                    # scale LoRA layers with `lora_scale`
+                    scale_lora_layers(text_encoder, weight=lora_scale)
 
                     is_model_cpu_offload = False
                     is_sequential_cpu_offload = False
@@ -2178,6 +2196,81 @@ def unfuse_text_encoder_lora(text_encoder):
 
         self.num_fused_loras -= 1
 
+    def set_adapter_for_text_encoder(
+        self,
+        adapter_names: Union[List[str], str],
+        text_encoder: Optional[PreTrainedModel] = None,
+        text_encoder_weights: List[float] = None,
+    ):
+        """
+        Sets the adapter layers for the text encoder.
+
+        Args:
+            adapter_names (`List[str]` or `str`):
+                The names of the adapters to use.
+            text_encoder (`torch.nn.Module`, *optional*):
+                The text encoder module to set the adapter layers for. If `None`, it will try to get the `text_encoder`
+                attribute.
+            text_encoder_weights (`List[float]`, *optional*):
+                The weights to use for the text encoder. If `None`, the weights are set to `1.0` for all the adapters.
+        """
+        if not self.use_peft_backend:
+            raise ValueError("PEFT backend is required for this method.")
+
+        def process_weights(adapter_names, weights):
+            if weights is None:
+                weights = [1.0] * len(adapter_names)
+            elif isinstance(weights, float):
+                weights = [weights]
+
+            if len(adapter_names) != len(weights):
+                raise ValueError(
+                    f"Length of adapter names {len(adapter_names)} is not equal to the length of the weights {len(weights)}"
+                )
+            return weights
+
+        adapter_names = [adapter_names] if isinstance(adapter_names, str) else adapter_names
+        text_encoder_weights = process_weights(adapter_names, text_encoder_weights)
+        text_encoder = text_encoder or getattr(self, "text_encoder", None)
+        if text_encoder is None:
+            raise ValueError(
+                "The pipeline does not have a default `pipe.text_encoder` class. Please make sure to pass a `text_encoder` instead."
+            )
+        set_weights_and_activate_adapters(text_encoder, adapter_names, text_encoder_weights)
+
+    def disable_lora_for_text_encoder(self, text_encoder: Optional[PreTrainedModel] = None):
+        """
+        Disables the LoRA layers for the text encoder.
+
+        Args:
+            text_encoder (`torch.nn.Module`, *optional*):
+                The text encoder module to disable the LoRA layers for. If `None`, it will try to get the
+                `text_encoder` attribute.
+        """
+        if not self.use_peft_backend:
+            raise ValueError("PEFT backend is required for this method.")
+
+        text_encoder = text_encoder or getattr(self, "text_encoder", None)
+        if text_encoder is None:
+            raise ValueError("Text Encoder not found.")
+        set_adapter_layers(text_encoder, enabled=False)
+
+    def enable_lora_for_text_encoder(self, text_encoder: Optional[PreTrainedModel] = None):
+        """
+        Enables the LoRA layers for the text encoder.
+
+        Args:
+            text_encoder (`torch.nn.Module`, *optional*):
+                The text encoder module to enable the LoRA layers for. If `None`, it will try to get the `text_encoder`
+                attribute.
+        """
+        if not self.use_peft_backend:
+            raise ValueError("PEFT backend is required for this method.")
+        text_encoder = text_encoder or getattr(self, "text_encoder", None)
+        if text_encoder is None:
+            raise ValueError("Text Encoder not found.")
+        set_adapter_layers(self.text_encoder, enabled=True)
+
 
 class FromSingleFileMixin:
     """

src/diffusers/models/lora.py

@@ -19,19 +19,15 @@
 from torch import nn
 
 from ..loaders import PatchedLoraProjection, text_encoder_attn_modules, text_encoder_mlp_modules
-from ..utils import logging
+from ..utils import logging, scale_lora_layers
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
 def adjust_lora_scale_text_encoder(text_encoder, lora_scale: float = 1.0, use_peft_backend: bool = False):
     if use_peft_backend:
-        from peft.tuners.lora import LoraLayer
-
-        for module in text_encoder.modules():
-            if isinstance(module, LoraLayer):
-                module.scaling[module.active_adapter] = lora_scale
+        scale_lora_layers(text_encoder, weight=lora_scale)
     else:
         for _, attn_module in text_encoder_attn_modules(text_encoder):
             if isinstance(attn_module.q_proj, PatchedLoraProjection):

src/diffusers/utils/__init__.py

@@ -84,7 +84,14 @@
 from .loading_utils import load_image
 from .logging import get_logger
 from .outputs import BaseOutput
-from .peft_utils import recurse_remove_peft_layers
+from .peft_utils import (
+    get_adapter_name,
+    get_peft_kwargs,
+    recurse_remove_peft_layers,
+    scale_lora_layers,
+    set_adapter_layers,
+    set_weights_and_activate_adapters,
+)
 from .pil_utils import PIL_INTERPOLATION, make_image_grid, numpy_to_pil, pt_to_pil
 from .state_dict_utils import convert_state_dict_to_diffusers, convert_state_dict_to_peft
 

src/diffusers/utils/peft_utils.py

@@ -14,6 +14,8 @@
 """
 PEFT utilities: Utilities related to peft library
 """
+import collections
+
 from .import_utils import is_torch_available
 
 
@@ -68,3 +70,98 @@ def recurse_remove_peft_layers(model):
                 torch.cuda.empty_cache()
 
     return model
+
+
+def scale_lora_layers(model, weight):
+    """
+    Adjust the weightage given to the LoRA layers of the model.
+
+    Args:
+        model (`torch.nn.Module`):
+            The model to scale.
+        weight (`float`):
+            The weight to be given to the LoRA layers.
+    """
+    from peft.tuners.tuners_utils import BaseTunerLayer
+
+    for module in model.modules():
+        if isinstance(module, BaseTunerLayer):
+            module.scale_layer(weight)
+
+
+def get_peft_kwargs(rank_dict, network_alpha_dict, peft_state_dict):
+    rank_pattern = {}
+    alpha_pattern = {}
+    r = lora_alpha = list(rank_dict.values())[0]
+    if len(set(rank_dict.values())) > 1:
+        # get the rank occuring the most number of times
+        r = collections.Counter(rank_dict.values()).most_common()[0][0]
+
+        # for modules with rank different from the most occuring rank, add it to the `rank_pattern`
+        rank_pattern = dict(filter(lambda x: x[1] != r, rank_dict.items()))
+        rank_pattern = {k.split(".lora_B.")[0]: v for k, v in rank_pattern.items()}
+
+    if network_alpha_dict is not None and len(set(network_alpha_dict.values())) > 1:
+        # get the alpha occuring the most number of times
+        lora_alpha = collections.Counter(network_alpha_dict.values()).most_common()[0][0]
+
+        # for modules with alpha different from the most occuring alpha, add it to the `alpha_pattern`
+        alpha_pattern = dict(filter(lambda x: x[1] != lora_alpha, network_alpha_dict.items()))
+        alpha_pattern = {".".join(k.split(".down.")[0].split(".")[:-1]): v for k, v in alpha_pattern.items()}
+
+    # layer names without the Diffusers specific
+    target_modules = list({name.split(".lora")[0] for name in peft_state_dict.keys()})
+
+    lora_config_kwargs = {
+        "r": r,
+        "lora_alpha": lora_alpha,
+        "rank_pattern": rank_pattern,
+        "alpha_pattern": alpha_pattern,
+        "target_modules": target_modules,
+    }
+    return lora_config_kwargs
+
+
+def get_adapter_name(model):
+    from peft.tuners.tuners_utils import BaseTunerLayer
+
+    for module in model.modules():
+        if isinstance(module, BaseTunerLayer):
+            return f"default_{len(module.r)}"
+    return "default_0"
+
+
+def set_adapter_layers(model, enabled=True):
+    from peft.tuners.tuners_utils import BaseTunerLayer
+
+    for module in model.modules():
+        if isinstance(module, BaseTunerLayer):
+            # The recent version of PEFT needs to call `enable_adapters` instead
+            if hasattr(module, "enable_adapters"):
+                module.enable_adapters(enabled=False)
+            else:
+                module.disable_adapters = True
+
+
+def set_weights_and_activate_adapters(model, adapter_names, weights):
+    from peft.tuners.tuners_utils import BaseTunerLayer
+
+    # iterate over each adapter, make it active and set the corresponding scaling weight
+    for adapter_name, weight in zip(adapter_names, weights):
+        for module in model.modules():
+            if isinstance(module, BaseTunerLayer):
+                # For backward compatbility with previous PEFT versions
+                if hasattr(module, "set_adapter"):
+                    module.set_adapter(adapter_name)
+                else:
+                    module.active_adapter = adapter_name
+                module.scale_layer(weight)
+
+    # set multiple active adapters
+    for module in model.modules():
+        if isinstance(module, BaseTunerLayer):
+            # For backward compatbility with previous PEFT versions
+            if hasattr(module, "set_adapter"):
+                module.set_adapter(adapter_names)
+            else:
+                module.active_adapter = adapter_names