AI/diffusers
1
0
Fork 0
mirror of https://github.com/huggingface/diffusers.git synced 2026-01-29 07:22:12 +03:00
Code Activity

update

Browse Source
This commit is contained in:
DN6
2025-09-18 23:31:07 +05:30
parent d06750a5fd
commit bea02ccba3
3 changed files with 67 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
src/diffusers/loaders/single_file_utils.py
View File

@@ -387,6 +387,14 @@ def is_valid_url(url):
return False
def _validate_single_file_path(pretrained_model_name_or_path):
if os.path.isfile(pretrained_model_name_or_path):
return True
repo_id, weight_name = _extract_repo_id_and_weights_name(pretrained_model_name_or_path)
return bool(repo_id and weight_name)
def _extract_repo_id_and_weights_name(pretrained_model_name_or_path):
if not is_valid_url(pretrained_model_name_or_path):
raise ValueError("Invalid `pretrained_model_name_or_path` provided. Please set it to a valid URL.")
@@ -398,7 +406,6 @@ def _extract_repo_id_and_weights_name(pretrained_model_name_or_path):
pretrained_model_name_or_path = pretrained_model_name_or_path.replace(prefix, "")
match = re.match(pattern, pretrained_model_name_or_path)
if not match:
logger.warning("Unable to identify the repo_id and weights_name from the provided URL.")
return repo_id, weights_name
repo_id = f"{match.group(1)}/{match.group(2)}"

60
src/diffusers/modular_pipelines/modular_pipeline.py
View File

@@ -22,7 +22,7 @@ from dataclasses import dataclass, field
from typing import Any, Dict, List, Optional, Tuple, Union
import torch
from huggingface_hub import create_repo
from huggingface_hub import create_pretrained_model_name_or_path
from huggingface_hub.utils import validate_hf_hub_args
from tqdm.auto import tqdm
from typing_extensions import Self
@@ -325,7 +325,7 @@ class ModularPipelineBlocks(ConfigMixin, PushToHubMixin):
)
if not (has_remote_code and trust_remote_code):
raise ValueError(
"Selected model repository does not happear to have any custom code or does not have a valid `config.json` file."
"Selected model pretrained_model_name_or_pathsitory does not happear to have any custom code or does not have a valid `config.json` file."
)
class_ref = config["auto_map"][cls.__name__]
@@ -366,7 +366,7 @@ class ModularPipelineBlocks(ConfigMixin, PushToHubMixin):
collection: Optional[str] = None,
) -> "ModularPipeline":
"""
create a ModularPipeline, optionally accept modular_repo to load from hub.
create a ModularPipeline, optionally accept modular_pretrained_model_name_or_path to load from hub.
"""
pipeline_class_name = MODULAR_PIPELINE_MAPPING.get(self.model_name, ModularPipeline.__name__)
diffusers_module = importlib.import_module("diffusers")
@@ -1481,7 +1481,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
pretrained_model_name_or_path: Path to a pretrained pipeline configuration. Can be None if the pipeline
does not require any additional loading config. If provided, will first try to load component specs
(only for from_pretrained components) and config values from `modular_model_index.json`, then
fallback to `model_index.json` for compatibility with standard non-modular repositories.
fallback to `model_index.json` for compatibility with standard non-modular pretrained_model_name_or_pathsitories.
components_manager:
Optional ComponentsManager for managing multiple component cross different pipelines and apply
offloading strategies.
@@ -1494,7 +1494,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
pipeline = ModularPipeline(blocks=my_custom_blocks)
# Initialize from pretrained configuration
pipeline = ModularPipeline(blocks=my_blocks, pretrained_model_name_or_path="my-repo/modular-pipeline")
pipeline = ModularPipeline(blocks=my_blocks, pretrained_model_name_or_path="my-pretrained_model_name_or_path/modular-pipeline")
# Initialize with components manager
pipeline = ModularPipeline(
@@ -1528,7 +1528,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
self._component_specs = {spec.name: deepcopy(spec) for spec in self.blocks.expected_components}
self._config_specs = {spec.name: deepcopy(spec) for spec in self.blocks.expected_configs}
# update component_specs and config_specs from modular_repo
# update component_specs and config_specs from modular_pretrained_model_name_or_path
if pretrained_model_name_or_path is not None:
cache_dir = kwargs.pop("cache_dir", None)
force_download = kwargs.pop("force_download", False)
@@ -1573,7 +1573,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
config_dict = DiffusionPipeline.load_config(pretrained_model_name_or_path, **load_config_kwargs)
except EnvironmentError as e:
logger.debug(f" model_index.json not found in the repo: {e}")
logger.debug(f" model_index.json not found in the pretrained_model_name_or_path: {e}")
config_dict = None
# update component_specs and config_specs based on model_index.json
@@ -1582,7 +1582,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
if name in self._component_specs and isinstance(value, (tuple, list)) and len(value) == 2:
library, class_name = value
component_spec_dict = {
"repo": pretrained_model_name_or_path,
"pretrained_model_name_or_path": pretrained_model_name_or_path,
"subfolder": name,
"type_hint": (library, class_name),
}
@@ -1633,13 +1633,13 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
**kwargs,
):
"""
Load a ModularPipeline from a huggingface hub repo.
Load a ModularPipeline from a huggingface hub pretrained_model_name_or_path.
Args:
pretrained_model_name_or_path (`str` or `os.PathLike`, optional):
Path to a pretrained pipeline configuration. It will first try to load config from
`modular_model_index.json`, then fallback to `model_index.json` for compatibility with standard
non-modular repositories. If the repo does not contain any pipeline config, it will be set to None
non-modular pretrained_model_name_or_pathsitories. If the pretrained_model_name_or_path does not contain any pipeline config, it will be set to None
during initialization.
trust_remote_code (`bool`, optional):
Whether to trust remote code when loading the pipeline, need to be set to True if you want to create
@@ -1679,7 +1679,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
# try to load modular_model_index.json
config_dict = cls.load_config(pretrained_model_name_or_path, **load_config_kwargs)
except EnvironmentError as e:
logger.debug(f" modular_model_index.json not found in the repo: {e}")
logger.debug(f" modular_model_index.json not found in the pretrained_model_name_or_path: {e}")
config_dict = None
if config_dict is not None:
@@ -1692,7 +1692,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
config_dict = DiffusionPipeline.load_config(pretrained_model_name_or_path, **load_config_kwargs)
except EnvironmentError as e:
logger.debug(f" model_index.json not found in the repo: {e}")
logger.debug(f" model_index.json not found in the pretrained_model_name_or_path: {e}")
if config_dict is not None:
logger.debug(" try to determine the modular pipeline class from model_index.json")
@@ -1731,11 +1731,15 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
private = kwargs.pop("private", None)
create_pr = kwargs.pop("create_pr", False)
token = kwargs.pop("token", None)
repo_id = kwargs.pop("repo_id", save_directory.split(os.path.sep)[-1])
repo_id = create_repo(repo_id, exist_ok=True, private=private, token=token).repo_id
pretrained_model_name_or_path_id = kwargs.pop(
"pretrained_model_name_or_path_id", save_directory.split(os.path.sep)[-1]
)
pretrained_model_name_or_path_id = create_pretrained_model_name_or_path(
pretrained_model_name_or_path_id, exist_ok=True, private=private, token=token
).pretrained_model_name_or_path_id
# Create a new empty model card and eventually tag it
model_card = load_or_create_model_card(repo_id, token=token, is_pipeline=True)
model_card = load_or_create_model_card(pretrained_model_name_or_path_id, token=token, is_pipeline=True)
model_card = populate_model_card(model_card)
model_card.save(os.path.join(save_directory, "README.md"))
@@ -1745,7 +1749,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
if push_to_hub:
self._upload_folder(
save_directory,
repo_id,
pretrained_model_name_or_path_id,
token=token,
commit_message=commit_message,
create_pr=create_pr,
@@ -1809,7 +1813,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
library, class_name = None, None
# extract the loading spec from the updated component spec that'll be used as part of modular_model_index.json config
# e.g. {"repo": "stabilityai/stable-diffusion-2-1",
# e.g. {"pretrained_model_name_or_path": "stabilityai/stable-diffusion-2-1",
# "type_hint": ("diffusers", "UNet2DConditionModel"),
# "subfolder": "unet",
# "variant": None,
@@ -2113,7 +2117,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
**kwargs: additional kwargs to be passed to `from_pretrained()`.Can be:
- a single value to be applied to all components to be loaded, e.g. torch_dtype=torch.bfloat16
- a dict, e.g. torch_dtype={"unet": torch.bfloat16, "default": torch.float32}
- if potentially override ComponentSpec if passed a different loading field in kwargs, e.g. `repo`,
- if potentially override ComponentSpec if passed a different loading field in kwargs, e.g. `pretrained_model_name_or_path`,
`variant`, `revision`, etc.
"""
@@ -2377,10 +2381,10 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
- "type_hint": Tuple[str, str]
Library name and class name of the component. (e.g. ("diffusers", "UNet2DConditionModel"))
- All loading fields defined by `component_spec.loading_fields()`, typically:
- "repo": Optional[str]
The model repository (e.g., "stabilityai/stable-diffusion-xl").
- "pretrained_model_name_or_path": Optional[str]
The model pretrained_model_name_or_pathsitory (e.g., "stabilityai/stable-diffusion-xl").
- "subfolder": Optional[str]
A subfolder within the repo where this component lives.
A subfolder within the pretrained_model_name_or_path where this component lives.
- "variant": Optional[str]
An optional variant identifier for the model.
- "revision": Optional[str]
@@ -2397,11 +2401,11 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
Example:
>>> from diffusers.pipelines.modular_pipeline_utils import ComponentSpec >>> from diffusers import
UNet2DConditionModel >>> spec = ComponentSpec(
... name="unet", ... type_hint=UNet2DConditionModel, ... config=None, ... repo="path/to/repo", ...
... name="unet", ... type_hint=UNet2DConditionModel, ... config=None, ... pretrained_model_name_or_path="path/to/pretrained_model_name_or_path", ...
subfolder="subfolder", ... variant=None, ... revision=None, ...
default_creation_method="from_pretrained",
... ) >>> ModularPipeline._component_spec_to_dict(spec) {
"type_hint": ("diffusers", "UNet2DConditionModel"), "repo": "path/to/repo", "subfolder": "subfolder",
"type_hint": ("diffusers", "UNet2DConditionModel"), "pretrained_model_name_or_path": "path/to/pretrained_model_name_or_path", "subfolder": "subfolder",
"variant": None, "revision": None,
}
"""
@@ -2431,10 +2435,10 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
- "type_hint": Tuple[str, str]
Library name and class name of the component. (e.g. ("diffusers", "UNet2DConditionModel"))
- All loading fields defined by `component_spec.loading_fields()`, typically:
- "repo": Optional[str]
The model repository (e.g., "stabilityai/stable-diffusion-xl").
- "pretrained_model_name_or_path": Optional[str]
The model pretrained_model_name_or_pathsitory (e.g., "stabilityai/stable-diffusion-xl").
- "subfolder": Optional[str]
A subfolder within the repo where this component lives.
A subfolder within the pretrained_model_name_or_path where this component lives.
- "variant": Optional[str]
An optional variant identifier for the model.
- "revision": Optional[str]
@@ -2451,10 +2455,10 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
ComponentSpec: A reconstructed ComponentSpec object.
Example:
>>> spec_dict = { ... "type_hint": ("diffusers", "UNet2DConditionModel"), ... "repo":
>>> spec_dict = { ... "type_hint": ("diffusers", "UNet2DConditionModel"), ... "pretrained_model_name_or_path":
"stabilityai/stable-diffusion-xl", ... "subfolder": "unet", ... "variant": None, ... "revision": None, ...
} >>> ModularPipeline._dict_to_component_spec("unet", spec_dict) ComponentSpec(
name="unet", type_hint=UNet2DConditionModel, config=None, repo="stabilityai/stable-diffusion-xl",
name="unet", type_hint=UNet2DConditionModel, config=None, pretrained_model_name_or_path="stabilityai/stable-diffusion-xl",
subfolder="unet", variant=None, revision=None, default_creation_method="from_pretrained"
)
"""

43
src/diffusers/modular_pipelines/modular_pipeline_utils.py
View File

@@ -21,6 +21,7 @@ from typing import Any, Dict, List, Literal, Optional, Type, Union
import torch
from ..configuration_utils import ConfigMixin, FrozenDict
from ..loaders.single_file_utils import _validate_single_file_path
from ..utils import is_torch_available, logging
@@ -80,10 +81,10 @@ class ComponentSpec:
type_hint: Type of the component (e.g. UNet2DConditionModel)
description: Optional description of the component
config: Optional config dict for __init__ creation
repo: Optional repo path for from_pretrained creation
subfolder: Optional subfolder in repo
variant: Optional variant in repo
revision: Optional revision in repo
pretrained_model_name_or_path: Optional pretrained_model_name_or_path path for from_pretrained creation
subfolder: Optional subfolder in pretrained_model_name_or_path
variant: Optional variant in pretrained_model_name_or_path
revision: Optional revision in pretrained_model_name_or_path
default_creation_method: Preferred creation method - "from_config" or "from_pretrained"
"""
@@ -92,7 +93,7 @@ class ComponentSpec:
description: Optional[str] = None
config: Optional[FrozenDict] = None
# YiYi Notes: should we change it to pretrained_model_name_or_path for consistency? a bit long for a field name
repo: Optional[Union[str, List[str]]] = field(default=None, metadata={"loading": True})
pretrained_model_name_or_path: Optional[Union[str, List[str]]] = field(default=None, metadata={"loading": True})
subfolder: Optional[str] = field(default="", metadata={"loading": True})
variant: Optional[str] = field(default=None, metadata={"loading": True})
revision: Optional[str] = field(default=None, metadata={"loading": True})
@@ -182,7 +183,7 @@ class ComponentSpec:
@property
def load_id(self) -> str:
"""
Unique identifier for this spec's pretrained load, composed of repo|subfolder|variant|revision (no empty
Unique identifier for this spec's pretrained load, composed of pretrained_model_name_or_path|subfolder|variant|revision (no empty
segments).
"""
if self.default_creation_method == "from_config":
@@ -197,12 +198,12 @@ class ComponentSpec:
Decode a load_id string back into a dictionary of loading fields and values.
Args:
load_id: The load_id string to decode, format: "repo|subfolder|variant|revision"
load_id: The load_id string to decode, format: "pretrained_model_name_or_path|subfolder|variant|revision"
where None values are represented as "null"
Returns:
Dict mapping loading field names to their values. e.g. {
"repo": "path/to/repo", "subfolder": "subfolder", "variant": "variant", "revision": "revision"
"pretrained_model_name_or_path": "path/to/pretrained_model_name_or_path", "subfolder": "subfolder", "variant": "variant", "revision": "revision"
} If a segment value is "null", it's replaced with None. Returns None if load_id is "null" (indicating
component not created with `load` method).
"""
@@ -260,33 +261,43 @@ class ComponentSpec:
def load(self, **kwargs) -> Any:
"""Load component using from_pretrained."""
# select loading fields from kwargs passed from user: e.g. repo, subfolder, variant, revision, note the list could change
# select loading fields from kwargs passed from user: e.g. pretrained_model_name_or_path, subfolder, variant, revision, note the list could change
passed_loading_kwargs = {key: kwargs.pop(key) for key in self.loading_fields() if key in kwargs}
# merge loading field value in the spec with user passed values to create load_kwargs
load_kwargs = {key: passed_loading_kwargs.get(key, getattr(self, key)) for key in self.loading_fields()}
# repo is a required argument for from_pretrained, a.k.a. pretrained_model_name_or_path
repo = load_kwargs.pop("repo", None)
if repo is None:
# pretrained_model_name_or_path is a required argument for from_pretrained, a.k.a. pretrained_model_name_or_path
pretrained_model_name_or_path = load_kwargs.pop("pretrained_model_name_or_path", None)
if pretrained_model_name_or_path is None:
raise ValueError(
"`repo` info is required when using `load` method (you can directly set it in `repo` field of the ComponentSpec or pass it as an argument)"
"`pretrained_model_name_or_path` info is required when using `load` method (you can directly set it in `pretrained_model_name_or_path` field of the ComponentSpec or pass it as an argument)"
)
is_single_file = _validate_single_file_path(pretrained_model_name_or_path)
if is_single_file and self.type_hint is None:
raise ValueError("type_hint is required when loading a single file model")
if self.type_hint is None:
try:
from diffusers import AutoModel
component = AutoModel.from_pretrained(repo, **load_kwargs, **kwargs)
component = AutoModel.from_pretrained(pretrained_model_name_or_path, **load_kwargs, **kwargs)
except Exception as e:
raise ValueError(f"Unable to load {self.name} without `type_hint`: {e}")
# update type_hint if AutoModel load successfully
self.type_hint = component.__class__
else:
# determine load method
load_method = (
getattr(self.type_hint, "from_single_file")
if is_single_file
else getattr(self.type_hint, "from_pretrained")
)
try:
component = self.type_hint.from_pretrained(repo, **load_kwargs, **kwargs)
component = load_method(pretrained_model_name_or_path, **load_kwargs, **kwargs)
except Exception as e:
raise ValueError(f"Unable to load {self.name} using load method: {e}")
self.repo = repo
self.pretrained_model_name_or_path = pretrained_model_name_or_path
for k, v in load_kwargs.items():
setattr(self, k, v)
component._diffusers_load_id = self.load_id