diffusers 0.29.0__tar.gz → 0.29.1__tar.gz

{diffusers-0.29.0/src/diffusers.egg-info → diffusers-0.29.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: diffusers
-Version: 0.29.0
+Version: 0.29.1
 Summary: State-of-the-art diffusion in PyTorch and JAX.
 Home-page: https://github.com/huggingface/diffusers
 Author: The Hugging Face team (past and future) with the help of all our contributors (https://github.com/huggingface/diffusers/graphs/contributors)
@@ -20,14 +20,88 @@ Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Requires-Python: >=3.8.0
 Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: importlib_metadata
+Requires-Dist: filelock
+Requires-Dist: huggingface-hub>=0.23.2
+Requires-Dist: numpy
+Requires-Dist: regex!=2019.12.17
+Requires-Dist: requests
+Requires-Dist: safetensors>=0.3.1
+Requires-Dist: Pillow
 Provides-Extra: quality
+Requires-Dist: urllib3<=2.0.0; extra == "quality"
+Requires-Dist: isort>=5.5.4; extra == "quality"
+Requires-Dist: ruff==0.1.5; extra == "quality"
+Requires-Dist: hf-doc-builder>=0.3.0; extra == "quality"
 Provides-Extra: docs
+Requires-Dist: hf-doc-builder>=0.3.0; extra == "docs"
 Provides-Extra: training
+Requires-Dist: accelerate>=0.29.3; extra == "training"
+Requires-Dist: datasets; extra == "training"
+Requires-Dist: protobuf<4,>=3.20.3; extra == "training"
+Requires-Dist: tensorboard; extra == "training"
+Requires-Dist: Jinja2; extra == "training"
+Requires-Dist: peft>=0.6.0; extra == "training"
 Provides-Extra: test
+Requires-Dist: compel==0.1.8; extra == "test"
+Requires-Dist: GitPython<3.1.19; extra == "test"
+Requires-Dist: datasets; extra == "test"
+Requires-Dist: Jinja2; extra == "test"
+Requires-Dist: invisible-watermark>=0.2.0; extra == "test"
+Requires-Dist: k-diffusion>=0.0.12; extra == "test"
+Requires-Dist: librosa; extra == "test"
+Requires-Dist: parameterized; extra == "test"
+Requires-Dist: pytest; extra == "test"
+Requires-Dist: pytest-timeout; extra == "test"
+Requires-Dist: pytest-xdist; extra == "test"
+Requires-Dist: requests-mock==1.10.0; extra == "test"
+Requires-Dist: safetensors>=0.3.1; extra == "test"
+Requires-Dist: sentencepiece!=0.1.92,>=0.1.91; extra == "test"
+Requires-Dist: scipy; extra == "test"
+Requires-Dist: torchvision; extra == "test"
+Requires-Dist: transformers>=4.25.1; extra == "test"
 Provides-Extra: torch
+Requires-Dist: torch>=1.4; extra == "torch"
+Requires-Dist: accelerate>=0.29.3; extra == "torch"
 Provides-Extra: flax
+Requires-Dist: jax>=0.4.1; extra == "flax"
+Requires-Dist: jaxlib>=0.4.1; extra == "flax"
+Requires-Dist: flax>=0.4.1; extra == "flax"
 Provides-Extra: dev
-License-File: LICENSE
+Requires-Dist: urllib3<=2.0.0; extra == "dev"
+Requires-Dist: isort>=5.5.4; extra == "dev"
+Requires-Dist: ruff==0.1.5; extra == "dev"
+Requires-Dist: hf-doc-builder>=0.3.0; extra == "dev"
+Requires-Dist: compel==0.1.8; extra == "dev"
+Requires-Dist: GitPython<3.1.19; extra == "dev"
+Requires-Dist: datasets; extra == "dev"
+Requires-Dist: Jinja2; extra == "dev"
+Requires-Dist: invisible-watermark>=0.2.0; extra == "dev"
+Requires-Dist: k-diffusion>=0.0.12; extra == "dev"
+Requires-Dist: librosa; extra == "dev"
+Requires-Dist: parameterized; extra == "dev"
+Requires-Dist: pytest; extra == "dev"
+Requires-Dist: pytest-timeout; extra == "dev"
+Requires-Dist: pytest-xdist; extra == "dev"
+Requires-Dist: requests-mock==1.10.0; extra == "dev"
+Requires-Dist: safetensors>=0.3.1; extra == "dev"
+Requires-Dist: sentencepiece!=0.1.92,>=0.1.91; extra == "dev"
+Requires-Dist: scipy; extra == "dev"
+Requires-Dist: torchvision; extra == "dev"
+Requires-Dist: transformers>=4.25.1; extra == "dev"
+Requires-Dist: accelerate>=0.29.3; extra == "dev"
+Requires-Dist: datasets; extra == "dev"
+Requires-Dist: protobuf<4,>=3.20.3; extra == "dev"
+Requires-Dist: tensorboard; extra == "dev"
+Requires-Dist: Jinja2; extra == "dev"
+Requires-Dist: peft>=0.6.0; extra == "dev"
+Requires-Dist: hf-doc-builder>=0.3.0; extra == "dev"
+Requires-Dist: torch>=1.4; extra == "dev"
+Requires-Dist: accelerate>=0.29.3; extra == "dev"
+Requires-Dist: jax>=0.4.1; extra == "dev"
+Requires-Dist: jaxlib>=0.4.1; extra == "dev"
+Requires-Dist: flax>=0.4.1; extra == "dev"
 
 <!---
 Copyright 2022 - The HuggingFace Team. All rights reserved.

{diffusers-0.29.0 → diffusers-0.29.1}/setup.py

@@ -254,7 +254,7 @@ version_range_max = max(sys.version_info[1], 10) + 1
 
 setup(
     name="diffusers",
-    version="0.29.0",  # expected format is one of x.y.z.dev0, or x.y.z.rc1 or x.y.z (no to dashes, yes to dots)
+    version="0.29.1",  # expected format is one of x.y.z.dev0, or x.y.z.rc1 or x.y.z (no to dashes, yes to dots)
     description="State-of-the-art diffusion in PyTorch and JAX.",
     long_description=open("README.md", "r", encoding="utf-8").read(),
     long_description_content_type="text/markdown",

{diffusers-0.29.0 → diffusers-0.29.1}/src/diffusers/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.29.0"
+__version__ = "0.29.1"
 
 from typing import TYPE_CHECKING
 
@@ -91,6 +91,8 @@ else:
             "MultiAdapter",
             "PixArtTransformer2DModel",
             "PriorTransformer",
+            "SD3ControlNetModel",
+            "SD3MultiControlNetModel",
             "SD3Transformer2DModel",
             "StableCascadeUNet",
             "T2IAdapter",
@@ -278,6 +280,7 @@ else:
             "StableCascadeCombinedPipeline",
             "StableCascadeDecoderPipeline",
             "StableCascadePriorPipeline",
+            "StableDiffusion3ControlNetPipeline",
             "StableDiffusion3Img2ImgPipeline",
             "StableDiffusion3Pipeline",
             "StableDiffusionAdapterPipeline",
@@ -501,6 +504,8 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             MultiAdapter,
             PixArtTransformer2DModel,
             PriorTransformer,
+            SD3ControlNetModel,
+            SD3MultiControlNetModel,
             SD3Transformer2DModel,
             T2IAdapter,
             T5FilmDecoder,
@@ -666,6 +671,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             StableCascadeCombinedPipeline,
             StableCascadeDecoderPipeline,
             StableCascadePriorPipeline,
+            StableDiffusion3ControlNetPipeline,
             StableDiffusion3Img2ImgPipeline,
             StableDiffusion3Pipeline,
             StableDiffusionAdapterPipeline,

{diffusers-0.29.0 → diffusers-0.29.1}/src/diffusers/loaders/single_file.py

@@ -28,9 +28,11 @@ from .single_file_utils import (
     _legacy_load_safety_checker,
     _legacy_load_scheduler,
     create_diffusers_clip_model_from_ldm,
+    create_diffusers_t5_model_from_checkpoint,
     fetch_diffusers_config,
     fetch_original_config,
     is_clip_model_in_single_file,
+    is_t5_in_single_file,
     load_single_file_checkpoint,
 )
 
@@ -118,6 +120,16 @@ def load_single_file_sub_model(
             is_legacy_loading=is_legacy_loading,
         )
 
+    elif is_transformers_model and is_t5_in_single_file(checkpoint):
+        loaded_sub_model = create_diffusers_t5_model_from_checkpoint(
+            class_obj,
+            checkpoint=checkpoint,
+            config=cached_model_config_path,
+            subfolder=name,
+            torch_dtype=torch_dtype,
+            local_files_only=local_files_only,
+        )
+
     elif is_tokenizer and is_legacy_loading:
         loaded_sub_model = _legacy_load_clip_tokenizer(
             class_obj, checkpoint=checkpoint, config=cached_model_config_path, local_files_only=local_files_only

{diffusers-0.29.0 → diffusers-0.29.1}/src/diffusers/loaders/single_file_model.py

@@ -276,16 +276,18 @@ class FromOriginalModelMixin:
 
         if is_accelerate_available():
             unexpected_keys = load_model_dict_into_meta(model, diffusers_format_checkpoint, dtype=torch_dtype)
-            if model._keys_to_ignore_on_load_unexpected is not None:
-                for pat in model._keys_to_ignore_on_load_unexpected:
-                    unexpected_keys = [k for k in unexpected_keys if re.search(pat, k) is None]
 
-            if len(unexpected_keys) > 0:
-                logger.warning(
-                    f"Some weights of the model checkpoint were not used when initializing {cls.__name__}: \n {[', '.join(unexpected_keys)]}"
-                )
         else:
-            model.load_state_dict(diffusers_format_checkpoint)
+            _, unexpected_keys = model.load_state_dict(diffusers_format_checkpoint, strict=False)
+
+        if model._keys_to_ignore_on_load_unexpected is not None:
+            for pat in model._keys_to_ignore_on_load_unexpected:
+                unexpected_keys = [k for k in unexpected_keys if re.search(pat, k) is None]
+
+        if len(unexpected_keys) > 0:
+            logger.warning(
+                f"Some weights of the model checkpoint were not used when initializing {cls.__name__}: \n {[', '.join(unexpected_keys)]}"
+            )
 
         if torch_dtype is not None:
             model.to(torch_dtype)

{diffusers-0.29.0 → diffusers-0.29.1}/src/diffusers/loaders/single_file_utils.py

@@ -252,7 +252,6 @@ LDM_CONTROLNET_KEY = "control_model."
 LDM_CLIP_PREFIX_TO_REMOVE = [
     "cond_stage_model.transformer.",
     "conditioner.embedders.0.transformer.",
-    "text_encoders.clip_l.transformer.",
 ]
 OPEN_CLIP_PREFIX = "conditioner.embedders.0.model."
 LDM_OPEN_CLIP_TEXT_PROJECTION_DIM = 1024
@@ -399,11 +398,14 @@ def is_open_clip_sdxl_model(checkpoint):
 
 
 def is_open_clip_sd3_model(checkpoint):
-    is_open_clip_sdxl_refiner_model(checkpoint)
+    if CHECKPOINT_KEY_NAMES["open_clip_sd3"] in checkpoint:
+        return True
+
+    return False
 
 
 def is_open_clip_sdxl_refiner_model(checkpoint):
-    if CHECKPOINT_KEY_NAMES["open_clip_sd3"] in checkpoint:
+    if CHECKPOINT_KEY_NAMES["open_clip_sdxl_refiner"] in checkpoint:
         return True
 
     return False
@@ -1233,11 +1235,14 @@ def convert_ldm_vae_checkpoint(checkpoint, config):
     return new_checkpoint
 
 
-def convert_ldm_clip_checkpoint(checkpoint):
+def convert_ldm_clip_checkpoint(checkpoint, remove_prefix=None):
     keys = list(checkpoint.keys())
     text_model_dict = {}
 
-    remove_prefixes = LDM_CLIP_PREFIX_TO_REMOVE
+    remove_prefixes = []
+    remove_prefixes.extend(LDM_CLIP_PREFIX_TO_REMOVE)
+    if remove_prefix:
+        remove_prefixes.append(remove_prefix)
 
     for key in keys:
         for prefix in remove_prefixes:
@@ -1263,8 +1268,6 @@ def convert_open_clip_checkpoint(
     else:
         text_proj_dim = LDM_OPEN_CLIP_TEXT_PROJECTION_DIM
 
-    text_model_dict["text_model.embeddings.position_ids"] = text_model.text_model.embeddings.get_buffer("position_ids")
-
     keys = list(checkpoint.keys())
     keys_to_ignore = SD_2_TEXT_ENCODER_KEYS_TO_IGNORE
 
@@ -1313,9 +1316,6 @@ def convert_open_clip_checkpoint(
         else:
             text_model_dict[diffusers_key] = checkpoint.get(key)
 
-    if not (hasattr(text_model, "embeddings") and hasattr(text_model.embeddings.position_ids)):
-        text_model_dict.pop("text_model.embeddings.position_ids", None)
-
     return text_model_dict
 
 
@@ -1376,6 +1376,13 @@ def create_diffusers_clip_model_from_ldm(
     ):
         diffusers_format_checkpoint = convert_ldm_clip_checkpoint(checkpoint)
 
+    elif (
+        is_clip_sd3_model(checkpoint)
+        and checkpoint[CHECKPOINT_KEY_NAMES["clip_sd3"]].shape[-1] == position_embedding_dim
+    ):
+        diffusers_format_checkpoint = convert_ldm_clip_checkpoint(checkpoint, "text_encoders.clip_l.transformer.")
+        diffusers_format_checkpoint["text_projection.weight"] = torch.eye(position_embedding_dim)
+
     elif is_open_clip_model(checkpoint):
         prefix = "cond_stage_model.model."
         diffusers_format_checkpoint = convert_open_clip_checkpoint(model, checkpoint, prefix=prefix)
@@ -1391,26 +1398,28 @@ def create_diffusers_clip_model_from_ldm(
         prefix = "conditioner.embedders.0.model."
         diffusers_format_checkpoint = convert_open_clip_checkpoint(model, checkpoint, prefix=prefix)
 
-    elif is_open_clip_sd3_model(checkpoint):
-        prefix = "text_encoders.clip_g.transformer."
-        diffusers_format_checkpoint = convert_open_clip_checkpoint(model, checkpoint, prefix=prefix)
+    elif (
+        is_open_clip_sd3_model(checkpoint)
+        and checkpoint[CHECKPOINT_KEY_NAMES["open_clip_sd3"]].shape[-1] == position_embedding_dim
+    ):
+        diffusers_format_checkpoint = convert_ldm_clip_checkpoint(checkpoint, "text_encoders.clip_g.transformer.")
 
     else:
         raise ValueError("The provided checkpoint does not seem to contain a valid CLIP model.")
 
     if is_accelerate_available():
         unexpected_keys = load_model_dict_into_meta(model, diffusers_format_checkpoint, dtype=torch_dtype)
-        if model._keys_to_ignore_on_load_unexpected is not None:
-            for pat in model._keys_to_ignore_on_load_unexpected:
-                unexpected_keys = [k for k in unexpected_keys if re.search(pat, k) is None]
+    else:
+        _, unexpected_keys = model.load_state_dict(diffusers_format_checkpoint, strict=False)
 
-        if len(unexpected_keys) > 0:
-            logger.warning(
-                f"Some weights of the model checkpoint were not used when initializing {cls.__name__}: \n {[', '.join(unexpected_keys)]}"
-            )
+    if model._keys_to_ignore_on_load_unexpected is not None:
+        for pat in model._keys_to_ignore_on_load_unexpected:
+            unexpected_keys = [k for k in unexpected_keys if re.search(pat, k) is None]
 
-    else:
-        model.load_state_dict(diffusers_format_checkpoint)
+    if len(unexpected_keys) > 0:
+        logger.warning(
+            f"Some weights of the model checkpoint were not used when initializing {cls.__name__}: \n {[', '.join(unexpected_keys)]}"
+        )
 
     if torch_dtype is not None:
         model.to(torch_dtype)
@@ -1755,7 +1764,7 @@ def convert_sd3_t5_checkpoint_to_diffusers(checkpoint):
     keys = list(checkpoint.keys())
     text_model_dict = {}
 
-    remove_prefixes = ["text_encoders.t5xxl.transformer.encoder."]
+    remove_prefixes = ["text_encoders.t5xxl.transformer."]
 
     for key in keys:
         for prefix in remove_prefixes:
@@ -1799,3 +1808,4 @@ def create_diffusers_t5_model_from_checkpoint(
 
     else:
         model.load_state_dict(diffusers_format_checkpoint)
+    return model

{diffusers-0.29.0 → diffusers-0.29.1}/src/diffusers/models/__init__.py

@@ -33,6 +33,7 @@ if is_torch_available():
     _import_structure["autoencoders.consistency_decoder_vae"] = ["ConsistencyDecoderVAE"]
     _import_structure["autoencoders.vq_model"] = ["VQModel"]
     _import_structure["controlnet"] = ["ControlNetModel"]
+    _import_structure["controlnet_sd3"] = ["SD3ControlNetModel", "SD3MultiControlNetModel"]
     _import_structure["controlnet_xs"] = ["ControlNetXSAdapter", "UNetControlNetXSModel"]
     _import_structure["embeddings"] = ["ImageProjection"]
     _import_structure["modeling_utils"] = ["ModelMixin"]
@@ -74,6 +75,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             VQModel,
         )
         from .controlnet import ControlNetModel
+        from .controlnet_sd3 import SD3ControlNetModel, SD3MultiControlNetModel
         from .controlnet_xs import ControlNetXSAdapter, UNetControlNetXSModel
         from .embeddings import ImageProjection
         from .modeling_utils import ModelMixin