# Copyright (c) 2023, Teriks
#
# dgenerate is distributed under the following BSD 3-Clause License
#
# Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
#
# 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
#
# 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in
# the documentation and/or other materials provided with the distribution.
#
# 3. Neither the name of the copyright holder nor the names of its contributors may be used to endorse or promote products derived
# from this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
# HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
# ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
import os.path
import typing
import diffusers
import huggingface_hub
import dgenerate.hfhub as _hfhub
import dgenerate.messages as _messages
import dgenerate.textprocessing as _textprocessing
import dgenerate.types as _types
from dgenerate.pipelinewrapper.uris import exceptions as _exceptions
_lora_uri_parser = _textprocessing.ConceptUriParser(
'LoRA', ['scale', 'revision', 'subfolder', 'weight-name'])
[docs]
class LoRAUri:
"""
Representation of a ``--loras`` uri
"""
# pipelinewrapper.uris.util.get_uri_accepted_args_schema metadata
NAMES = ['LoRA']
[docs]
@staticmethod
def help():
import dgenerate.arguments as _a
return _a.get_raw_help_text('--loras')
FILE_ARGS = {
'model': {'mode': ['in', 'dir'], 'filetypes': [('Models', ['*.safetensors', '*.pt', '*.pth', '*.cpkt', '*.bin'])]}
}
# ===
@property
def model(self) -> str:
"""
Model path, huggingface slug, file path
"""
return self._model
@property
def revision(self) -> _types.OptionalString:
"""
Model repo revision
"""
return self._revision
@property
def subfolder(self) -> _types.OptionalPath:
"""
Model repo subfolder
"""
return self._subfolder
@property
def weight_name(self) -> _types.OptionalName:
"""
Model weight-name
"""
return self._weight_name
@property
def scale(self) -> float:
"""
LoRA scale
"""
return self._scale
[docs]
def __init__(self,
model: str,
revision: _types.OptionalString = None,
subfolder: _types.OptionalPath = None,
weight_name: _types.OptionalName = None,
scale: float = 1.0):
self._model = model
self._scale = scale
self._revision = revision
self._subfolder = subfolder
self._weight_name = weight_name
def __str__(self):
return f'{self.__class__.__name__}({str(_types.get_public_attributes(self))})'
def __repr__(self):
return str(self)
[docs]
@staticmethod
def load_on_pipeline(pipeline: diffusers.DiffusionPipeline,
uris: typing.Iterable[typing.Union["LoRAUri", str]],
fuse_scale: float = 1.0,
use_auth_token: _types.OptionalString = None,
local_files_only: bool = False):
"""
Load LoRA weights on to a pipeline using this URI
:param pipeline: :py:class:`diffusers.DiffusionPipeline`
:param uris: Iterable of :py:class:`LoRAUri` or ``str`` LoRA URIs to load
:param fuse_scale: Global scale for the fused LoRAs, all LoRAs are fused together
using their individual scale value, and then fused into the main model using this scale.
:param use_auth_token: optional huggingface auth token.
:param local_files_only: avoid downloading files and only look for cached files
when the model path is a huggingface slug
:raises dgenerate.ModelNotFoundError: If the model could not be found.
:raises dgenerate.pipelinewrapper.uris.exceptions.InvalidLoRAUriError: On URI parsing errors.
:raises dgenerate.pipelinewrapper.uris.exceptions.LoRAUriLoadError: On loading errors.
"""
def cache_all(e):
if isinstance(e, _exceptions.InvalidLoRAUriError):
raise e
else:
raise _exceptions.LoRAUriLoadError(
f'error loading LoRAs: {e}') from e
with _hfhub.with_hf_errors_as_model_not_found(cache_all):
LoRAUri._load_on_pipeline(pipeline,
uris=uris,
fuse_scale=fuse_scale,
use_auth_token=use_auth_token,
local_files_only=local_files_only)
@staticmethod
def _load_on_pipeline(pipeline: diffusers.DiffusionPipeline,
uris: typing.Iterable[typing.Union["LoRAUri", str]],
fuse_scale: float = 1.0,
use_auth_token: _types.OptionalString = None,
local_files_only: bool = False):
if hasattr(pipeline, 'load_lora_weights'):
adapter_names = []
adapter_weights = []
for adapter_index, lora_uri in enumerate(uris):
if not isinstance(lora_uri, LoRAUri):
lora_uri = LoRAUri.parse(lora_uri)
model_path = _hfhub.download_non_hf_slug_model(lora_uri.model)
weight_name = lora_uri.weight_name
if local_files_only and not os.path.exists(model_path):
subfolder = lora_uri.subfolder if lora_uri.subfolder else ''
probable_path_1 = os.path.join(
subfolder, 'pytorch_lora_weights.safetensors' if
weight_name is None else weight_name)
probable_path_2 = os.path.join(
subfolder, 'pytorch_lora_weights.bin')
file_path = huggingface_hub.try_to_load_from_cache(lora_uri.model,
filename=probable_path_1,
revision=lora_uri.revision)
if not isinstance(file_path, str):
file_path = huggingface_hub.try_to_load_from_cache(lora_uri.model,
filename=probable_path_2,
revision=lora_uri.revision)
if not isinstance(file_path, str):
raise RuntimeError(
f'LoRA model "{lora_uri.model}" '
'was not available in the local huggingface cache.')
else:
weight_name = 'pytorch_lora_weights.bin'
else:
if weight_name is None:
weight_name = 'pytorch_lora_weights.safetensors'
model_path = os.path.dirname(file_path)
adapter_name = str(adapter_index)
adapter_names.append(adapter_name)
adapter_weights.append(lora_uri.scale)
try:
pipeline.load_lora_weights(model_path,
revision=lora_uri.revision,
subfolder=lora_uri.subfolder,
weight_name=weight_name,
local_files_only=local_files_only,
use_safetensors=True,
token=use_auth_token,
adapter_name=adapter_name)
except EnvironmentError:
# brute force, try for .bin files
pipeline.load_lora_weights(model_path,
revision=lora_uri.revision,
subfolder=lora_uri.subfolder,
weight_name=weight_name,
local_files_only=local_files_only,
token=use_auth_token,
adapter_name=adapter_name)
if not pipeline.get_list_adapters():
raise RuntimeError(f'LoRA model "{lora_uri.model}" contained no usable weights.')
_messages.debug_log(f'Added LoRA: "{lora_uri}" to pipeline: "{pipeline.__class__.__name__}"')
_messages.debug_log(f'Fusing all LoRAs into pipeline with global scale: {fuse_scale}')
pipeline.set_adapters(adapter_names, adapter_weights=adapter_weights)
pipeline.fuse_lora(adapter_names=adapter_names, lora_scale=fuse_scale)
else:
raise RuntimeError(f'Pipeline: {pipeline.__class__.__name__} '
f'does not support loading LoRAs.')
[docs]
@staticmethod
def parse(uri: _types.Uri) -> 'LoRAUri':
"""
Parse a ``--loras`` uri and return an object representing its constituents
:param uri: string with ``--loras`` uri syntax
:raise InvalidLoRAUriError:
:return: :py:class:`.LoRAUri`
"""
try:
r = _lora_uri_parser.parse(uri)
scale = r.args.get('scale', 1.0)
try:
scale = float(scale)
except ValueError:
raise _exceptions.InvalidLoRAUriError(
f'LoRA "scale" must be a floating point number, received: {scale}')
return LoRAUri(model=r.concept,
scale=scale,
weight_name=r.args.get('weight-name', None),
revision=r.args.get('revision', None),
subfolder=r.args.get('subfolder', None))
except _textprocessing.ConceptUriParseError as e:
raise _exceptions.InvalidLoRAUriError(e) from e