diffusers 0.29.0__py3-none-any.whl → 0.29.2__py3-none-any.whl

diffusers/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.29.0"
+__version__ = "0.29.2"
 
 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/loaders/lora.py

@@ -42,7 +42,7 @@ from ..utils import (
     set_adapter_layers,
     set_weights_and_activate_adapters,
 )
-from .lora_conversion_utils import _convert_kohya_lora_to_diffusers, _maybe_map_sgm_blocks_to_diffusers
+from .lora_conversion_utils import _convert_non_diffusers_lora_to_diffusers, _maybe_map_sgm_blocks_to_diffusers
 
 
 if is_transformers_available():
@@ -287,7 +287,7 @@ class LoraLoaderMixin:
             if unet_config is not None:
                 # use unet config to remap block numbers
                 state_dict = _maybe_map_sgm_blocks_to_diffusers(state_dict, unet_config)
-            state_dict, network_alphas = _convert_kohya_lora_to_diffusers(state_dict)
+            state_dict, network_alphas = _convert_non_diffusers_lora_to_diffusers(state_dict)
 
         return state_dict, network_alphas
 
@@ -395,8 +395,7 @@ class LoraLoaderMixin:
         # their prefixes.
         keys = list(state_dict.keys())
         only_text_encoder = all(key.startswith(cls.text_encoder_name) for key in keys)
-
-        if any(key.startswith(cls.unet_name) for key in keys) and not only_text_encoder:
+        if not only_text_encoder:
             # Load the layers corresponding to UNet.
             logger.info(f"Loading {cls.unet_name}.")
             unet.load_attn_procs(

diffusers/loaders/lora_conversion_utils.py

@@ -123,134 +123,76 @@ def _maybe_map_sgm_blocks_to_diffusers(state_dict, unet_config, delimiter="_", b
     return new_state_dict
 
 
-def _convert_kohya_lora_to_diffusers(state_dict, unet_name="unet", text_encoder_name="text_encoder"):
+def _convert_non_diffusers_lora_to_diffusers(state_dict, unet_name="unet", text_encoder_name="text_encoder"):
+    """
+    Converts a non-Diffusers LoRA state dict to a Diffusers compatible state dict.
+
+    Args:
+        state_dict (`dict`): The state dict to convert.
+        unet_name (`str`, optional): The name of the U-Net module in the Diffusers model. Defaults to "unet".
+        text_encoder_name (`str`, optional): The name of the text encoder module in the Diffusers model. Defaults to
+            "text_encoder".
+
+    Returns:
+        `tuple`: A tuple containing the converted state dict and a dictionary of alphas.
+    """
     unet_state_dict = {}
     te_state_dict = {}
     te2_state_dict = {}
     network_alphas = {}
-    is_unet_dora_lora = any("dora_scale" in k and "lora_unet_" in k for k in state_dict)
-    is_te_dora_lora = any("dora_scale" in k and ("lora_te_" in k or "lora_te1_" in k) for k in state_dict)
-    is_te2_dora_lora = any("dora_scale" in k and "lora_te2_" in k for k in state_dict)
 
-    if is_unet_dora_lora or is_te_dora_lora or is_te2_dora_lora:
+    # Check for DoRA-enabled LoRAs.
+    dora_present_in_unet = any("dora_scale" in k and "lora_unet_" in k for k in state_dict)
+    dora_present_in_te = any("dora_scale" in k and ("lora_te_" in k or "lora_te1_" in k) for k in state_dict)
+    dora_present_in_te2 = any("dora_scale" in k and "lora_te2_" in k for k in state_dict)
+    if dora_present_in_unet or dora_present_in_te or dora_present_in_te2:
         if is_peft_version("<", "0.9.0"):
             raise ValueError(
                 "You need `peft` 0.9.0 at least to use DoRA-enabled LoRAs. Please upgrade your installation of `peft`."
             )
 
-    # every down weight has a corresponding up weight and potentially an alpha weight
-    lora_keys = [k for k in state_dict.keys() if k.endswith("lora_down.weight")]
-    for key in lora_keys:
+    # Iterate over all LoRA weights.
+    all_lora_keys = list(state_dict.keys())
+    for key in all_lora_keys:
+        if not key.endswith("lora_down.weight"):
+            continue
+
+        # Extract LoRA name.
         lora_name = key.split(".")[0]
+
+        # Find corresponding up weight and alpha.
         lora_name_up = lora_name + ".lora_up.weight"
         lora_name_alpha = lora_name + ".alpha"
 
+        # Handle U-Net LoRAs.
         if lora_name.startswith("lora_unet_"):
-            diffusers_name = key.replace("lora_unet_", "").replace("_", ".")
-
-            if "input.blocks" in diffusers_name:
-                diffusers_name = diffusers_name.replace("input.blocks", "down_blocks")
-            else:
-                diffusers_name = diffusers_name.replace("down.blocks", "down_blocks")
+            diffusers_name = _convert_unet_lora_key(key)
 
-            if "middle.block" in diffusers_name:
-                diffusers_name = diffusers_name.replace("middle.block", "mid_block")
-            else:
-                diffusers_name = diffusers_name.replace("mid.block", "mid_block")
-            if "output.blocks" in diffusers_name:
-                diffusers_name = diffusers_name.replace("output.blocks", "up_blocks")
-            else:
-                diffusers_name = diffusers_name.replace("up.blocks", "up_blocks")
-
-            diffusers_name = diffusers_name.replace("transformer.blocks", "transformer_blocks")
-            diffusers_name = diffusers_name.replace("to.q.lora", "to_q_lora")
-            diffusers_name = diffusers_name.replace("to.k.lora", "to_k_lora")
-            diffusers_name = diffusers_name.replace("to.v.lora", "to_v_lora")
-            diffusers_name = diffusers_name.replace("to.out.0.lora", "to_out_lora")
-            diffusers_name = diffusers_name.replace("proj.in", "proj_in")
-            diffusers_name = diffusers_name.replace("proj.out", "proj_out")
-            diffusers_name = diffusers_name.replace("emb.layers", "time_emb_proj")
-
-            # SDXL specificity.
-            if "emb" in diffusers_name and "time.emb.proj" not in diffusers_name:
-                pattern = r"\.\d+(?=\D*$)"
-                diffusers_name = re.sub(pattern, "", diffusers_name, count=1)
-            if ".in." in diffusers_name:
-                diffusers_name = diffusers_name.replace("in.layers.2", "conv1")
-            if ".out." in diffusers_name:
-                diffusers_name = diffusers_name.replace("out.layers.3", "conv2")
-            if "downsamplers" in diffusers_name or "upsamplers" in diffusers_name:
-                diffusers_name = diffusers_name.replace("op", "conv")
-            if "skip" in diffusers_name:
-                diffusers_name = diffusers_name.replace("skip.connection", "conv_shortcut")
-
-            # LyCORIS specificity.
-            if "time.emb.proj" in diffusers_name:
-                diffusers_name = diffusers_name.replace("time.emb.proj", "time_emb_proj")
-            if "conv.shortcut" in diffusers_name:
-                diffusers_name = diffusers_name.replace("conv.shortcut", "conv_shortcut")
-
-            # General coverage.
-            if "transformer_blocks" in diffusers_name:
-                if "attn1" in diffusers_name or "attn2" in diffusers_name:
-                    diffusers_name = diffusers_name.replace("attn1", "attn1.processor")
-                    diffusers_name = diffusers_name.replace("attn2", "attn2.processor")
-                    unet_state_dict[diffusers_name] = state_dict.pop(key)
-                    unet_state_dict[diffusers_name.replace(".down.", ".up.")] = state_dict.pop(lora_name_up)
-                elif "ff" in diffusers_name:
-                    unet_state_dict[diffusers_name] = state_dict.pop(key)
-                    unet_state_dict[diffusers_name.replace(".down.", ".up.")] = state_dict.pop(lora_name_up)
-            elif any(key in diffusers_name for key in ("proj_in", "proj_out")):
-                unet_state_dict[diffusers_name] = state_dict.pop(key)
-                unet_state_dict[diffusers_name.replace(".down.", ".up.")] = state_dict.pop(lora_name_up)
-            else:
-                unet_state_dict[diffusers_name] = state_dict.pop(key)
-                unet_state_dict[diffusers_name.replace(".down.", ".up.")] = state_dict.pop(lora_name_up)
+            # Store down and up weights.
+            unet_state_dict[diffusers_name] = state_dict.pop(key)
+            unet_state_dict[diffusers_name.replace(".down.", ".up.")] = state_dict.pop(lora_name_up)
 
-            if is_unet_dora_lora:
+            # Store DoRA scale if present.
+            if dora_present_in_unet:
                 dora_scale_key_to_replace = "_lora.down." if "_lora.down." in diffusers_name else ".lora.down."
                 unet_state_dict[
                     diffusers_name.replace(dora_scale_key_to_replace, ".lora_magnitude_vector.")
                 ] = state_dict.pop(key.replace("lora_down.weight", "dora_scale"))
 
+        # Handle text encoder LoRAs.
         elif lora_name.startswith(("lora_te_", "lora_te1_", "lora_te2_")):
+            diffusers_name = _convert_text_encoder_lora_key(key, lora_name)
+
+            # Store down and up weights for te or te2.
             if lora_name.startswith(("lora_te_", "lora_te1_")):
-                key_to_replace = "lora_te_" if lora_name.startswith("lora_te_") else "lora_te1_"
+                te_state_dict[diffusers_name] = state_dict.pop(key)
+                te_state_dict[diffusers_name.replace(".down.", ".up.")] = state_dict.pop(lora_name_up)
             else:
-                key_to_replace = "lora_te2_"
-
-            diffusers_name = key.replace(key_to_replace, "").replace("_", ".")
-            diffusers_name = diffusers_name.replace("text.model", "text_model")
-            diffusers_name = diffusers_name.replace("self.attn", "self_attn")
-            diffusers_name = diffusers_name.replace("q.proj.lora", "to_q_lora")
-            diffusers_name = diffusers_name.replace("k.proj.lora", "to_k_lora")
-            diffusers_name = diffusers_name.replace("v.proj.lora", "to_v_lora")
-            diffusers_name = diffusers_name.replace("out.proj.lora", "to_out_lora")
-            diffusers_name = diffusers_name.replace("text.projection", "text_projection")
-
-            if "self_attn" in diffusers_name:
-                if lora_name.startswith(("lora_te_", "lora_te1_")):
-                    te_state_dict[diffusers_name] = state_dict.pop(key)
-                    te_state_dict[diffusers_name.replace(".down.", ".up.")] = state_dict.pop(lora_name_up)
-                else:
-                    te2_state_dict[diffusers_name] = state_dict.pop(key)
-                    te2_state_dict[diffusers_name.replace(".down.", ".up.")] = state_dict.pop(lora_name_up)
-            elif "mlp" in diffusers_name:
-                # Be aware that this is the new diffusers convention and the rest of the code might
-                # not utilize it yet.
-                diffusers_name = diffusers_name.replace(".lora.", ".lora_linear_layer.")
-                if lora_name.startswith(("lora_te_", "lora_te1_")):
-                    te_state_dict[diffusers_name] = state_dict.pop(key)
-                    te_state_dict[diffusers_name.replace(".down.", ".up.")] = state_dict.pop(lora_name_up)
-                else:
-                    te2_state_dict[diffusers_name] = state_dict.pop(key)
-                    te2_state_dict[diffusers_name.replace(".down.", ".up.")] = state_dict.pop(lora_name_up)
-            # OneTrainer specificity
-            elif "text_projection" in diffusers_name and lora_name.startswith("lora_te2_"):
                 te2_state_dict[diffusers_name] = state_dict.pop(key)
                 te2_state_dict[diffusers_name.replace(".down.", ".up.")] = state_dict.pop(lora_name_up)
 
-            if (is_te_dora_lora or is_te2_dora_lora) and lora_name.startswith(("lora_te_", "lora_te1_", "lora_te2_")):
+            # Store DoRA scale if present.
+            if dora_present_in_te or dora_present_in_te2:
                 dora_scale_key_to_replace_te = (
                     "_lora.down." if "_lora.down." in diffusers_name else ".lora_linear_layer."
                 )
@@ -263,22 +205,18 @@ def _convert_kohya_lora_to_diffusers(state_dict, unet_name="unet", text_encoder_
                         diffusers_name.replace(dora_scale_key_to_replace_te, ".lora_magnitude_vector.")
                     ] = state_dict.pop(key.replace("lora_down.weight", "dora_scale"))
 
-        # Rename the alphas so that they can be mapped appropriately.
+        # Store alpha if present.
         if lora_name_alpha in state_dict:
             alpha = state_dict.pop(lora_name_alpha).item()
-            if lora_name_alpha.startswith("lora_unet_"):
-                prefix = "unet."
-            elif lora_name_alpha.startswith(("lora_te_", "lora_te1_")):
-                prefix = "text_encoder."
-            else:
-                prefix = "text_encoder_2."
-            new_name = prefix + diffusers_name.split(".lora.")[0] + ".alpha"
-            network_alphas.update({new_name: alpha})
+            network_alphas.update(_get_alpha_name(lora_name_alpha, diffusers_name, alpha))
 
+    # Check if any keys remain.
     if len(state_dict) > 0:
         raise ValueError(f"The following keys have not been correctly renamed: \n\n {', '.join(state_dict.keys())}")
 
-    logger.info("Kohya-style checkpoint detected.")
+    logger.info("Non-diffusers checkpoint detected.")
+
+    # Construct final state dict.
     unet_state_dict = {f"{unet_name}.{module_name}": params for module_name, params in unet_state_dict.items()}
     te_state_dict = {f"{text_encoder_name}.{module_name}": params for module_name, params in te_state_dict.items()}
     te2_state_dict = (
@@ -291,3 +229,100 @@ def _convert_kohya_lora_to_diffusers(state_dict, unet_name="unet", text_encoder_
 
     new_state_dict = {**unet_state_dict, **te_state_dict}
     return new_state_dict, network_alphas
+
+
+def _convert_unet_lora_key(key):
+    """
+    Converts a U-Net LoRA key to a Diffusers compatible key.
+    """
+    diffusers_name = key.replace("lora_unet_", "").replace("_", ".")
+
+    # Replace common U-Net naming patterns.
+    diffusers_name = diffusers_name.replace("input.blocks", "down_blocks")
+    diffusers_name = diffusers_name.replace("down.blocks", "down_blocks")
+    diffusers_name = diffusers_name.replace("middle.block", "mid_block")
+    diffusers_name = diffusers_name.replace("mid.block", "mid_block")
+    diffusers_name = diffusers_name.replace("output.blocks", "up_blocks")
+    diffusers_name = diffusers_name.replace("up.blocks", "up_blocks")
+    diffusers_name = diffusers_name.replace("transformer.blocks", "transformer_blocks")
+    diffusers_name = diffusers_name.replace("to.q.lora", "to_q_lora")
+    diffusers_name = diffusers_name.replace("to.k.lora", "to_k_lora")
+    diffusers_name = diffusers_name.replace("to.v.lora", "to_v_lora")
+    diffusers_name = diffusers_name.replace("to.out.0.lora", "to_out_lora")
+    diffusers_name = diffusers_name.replace("proj.in", "proj_in")
+    diffusers_name = diffusers_name.replace("proj.out", "proj_out")
+    diffusers_name = diffusers_name.replace("emb.layers", "time_emb_proj")
+
+    # SDXL specific conversions.
+    if "emb" in diffusers_name and "time.emb.proj" not in diffusers_name:
+        pattern = r"\.\d+(?=\D*$)"
+        diffusers_name = re.sub(pattern, "", diffusers_name, count=1)
+    if ".in." in diffusers_name:
+        diffusers_name = diffusers_name.replace("in.layers.2", "conv1")
+    if ".out." in diffusers_name:
+        diffusers_name = diffusers_name.replace("out.layers.3", "conv2")
+    if "downsamplers" in diffusers_name or "upsamplers" in diffusers_name:
+        diffusers_name = diffusers_name.replace("op", "conv")
+    if "skip" in diffusers_name:
+        diffusers_name = diffusers_name.replace("skip.connection", "conv_shortcut")
+
+    # LyCORIS specific conversions.
+    if "time.emb.proj" in diffusers_name:
+        diffusers_name = diffusers_name.replace("time.emb.proj", "time_emb_proj")
+    if "conv.shortcut" in diffusers_name:
+        diffusers_name = diffusers_name.replace("conv.shortcut", "conv_shortcut")
+
+    # General conversions.
+    if "transformer_blocks" in diffusers_name:
+        if "attn1" in diffusers_name or "attn2" in diffusers_name:
+            diffusers_name = diffusers_name.replace("attn1", "attn1.processor")
+            diffusers_name = diffusers_name.replace("attn2", "attn2.processor")
+        elif "ff" in diffusers_name:
+            pass
+    elif any(key in diffusers_name for key in ("proj_in", "proj_out")):
+        pass
+    else:
+        pass
+
+    return diffusers_name
+
+
+def _convert_text_encoder_lora_key(key, lora_name):
+    """
+    Converts a text encoder LoRA key to a Diffusers compatible key.
+    """
+    if lora_name.startswith(("lora_te_", "lora_te1_")):
+        key_to_replace = "lora_te_" if lora_name.startswith("lora_te_") else "lora_te1_"
+    else:
+        key_to_replace = "lora_te2_"
+
+    diffusers_name = key.replace(key_to_replace, "").replace("_", ".")
+    diffusers_name = diffusers_name.replace("text.model", "text_model")
+    diffusers_name = diffusers_name.replace("self.attn", "self_attn")
+    diffusers_name = diffusers_name.replace("q.proj.lora", "to_q_lora")
+    diffusers_name = diffusers_name.replace("k.proj.lora", "to_k_lora")
+    diffusers_name = diffusers_name.replace("v.proj.lora", "to_v_lora")
+    diffusers_name = diffusers_name.replace("out.proj.lora", "to_out_lora")
+    diffusers_name = diffusers_name.replace("text.projection", "text_projection")
+
+    if "self_attn" in diffusers_name or "text_projection" in diffusers_name:
+        pass
+    elif "mlp" in diffusers_name:
+        # Be aware that this is the new diffusers convention and the rest of the code might
+        # not utilize it yet.
+        diffusers_name = diffusers_name.replace(".lora.", ".lora_linear_layer.")
+    return diffusers_name
+
+
+def _get_alpha_name(lora_name_alpha, diffusers_name, alpha):
+    """
+    Gets the correct alpha name for the Diffusers model.
+    """
+    if lora_name_alpha.startswith("lora_unet_"):
+        prefix = "unet."
+    elif lora_name_alpha.startswith(("lora_te_", "lora_te1_")):
+        prefix = "text_encoder."
+    else:
+        prefix = "text_encoder_2."
+    new_name = prefix + diffusers_name.split(".lora.")[0] + ".alpha"
+    return {new_name: alpha}

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/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/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/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