diffsynth-engine 0.4.3.dev9__py3-none-any.whl → 0.4.3.dev11__py3-none-any.whl

diffsynth_engine/utils/image.py

@@ -1,6 +1,13 @@
 import torch
 import numpy as np
+import math
 from PIL import Image
+from enum import Enum
+from typing import List, Tuple, Optional
+
+from diffsynth_engine.utils import logging
+
+logger = logging.get_logger(__name__)
 
 
 def tensor_to_image(t: torch.Tensor, denormalize: bool = True) -> Image.Image:
@@ -23,3 +30,209 @@ def tensor_to_image(t: torch.Tensor, denormalize: bool = True) -> Image.Image:
     else:
         mode = "RGB"
     return Image.fromarray(t, mode=mode)
+
+
+class ChannelDimension(Enum):
+    FIRST = "channels_first"
+    LAST = "channels_last"
+
+
+def convert_to_rgb(image: Image.Image) -> Image.Image:
+    if not isinstance(image, Image.Image):
+        raise TypeError(f"image must be a PIL.Image.Image, but got {type(image)}")
+    if image.mode == "RGB":
+        return image
+    image = image.convert(mode="RGB")
+    return image
+
+
+def infer_channel_dimension_format(image: np.ndarray) -> ChannelDimension:
+    num_channels = (1, 3)
+    if image.ndim == 3:
+        first_dim, last_dim = 0, 2
+    elif image.ndim == 4:
+        first_dim, last_dim = 1, 3
+    else:
+        raise ValueError(f"Unsupported number of image dimensions: {image.ndim}")
+
+    if image.shape[first_dim] in num_channels and image.shape[last_dim] in num_channels:
+        logger.warning("Image has both first and last dimensions as channels. This may lead to unexpected behavior.")
+        return ChannelDimension.FIRST
+    elif image.shape[first_dim] in num_channels:
+        return ChannelDimension.FIRST
+    elif image.shape[last_dim] in num_channels:
+        return ChannelDimension.LAST
+    raise ValueError("Unable to infer channel dimension format")
+
+
+def get_image_size(image: np.ndarray, channel_dim: Optional[ChannelDimension] = None) -> Tuple[int, int]:
+    """
+    Returns the (height, width) dimensions of the image.
+    """
+    if channel_dim is None:
+        channel_dim = infer_channel_dimension_format(image)
+    if channel_dim == ChannelDimension.FIRST:
+        return image.shape[-2], image.shape[-1]
+    elif channel_dim == ChannelDimension.LAST:
+        return image.shape[-3], image.shape[-2]
+    else:
+        raise ValueError(f"Unsupported channel dimension format: {channel_dim}")
+
+
+def smart_resize(
+    height: int, width: int, factor: int = 28, min_pixels: int = 56 * 56, max_pixels: int = 14 * 14 * 4 * 1280
+) -> Tuple[int, int]:
+    """Rescales the image so that the following conditions are met:
+    1. Both dimensions (height and width) are divisible by 'factor'.
+    2. The total number of pixels is within the range ['min_pixels', 'max_pixels'].
+    3. The aspect ratio of the image is maintained as closely as possible.
+    """
+    abs_aspect_ratio = max(height, width) / min(height, width)
+    if height < factor or width < factor:
+        raise ValueError(f"Image height: {height} and width: {width} must be greater than or equal to factor: {factor}")
+    elif abs_aspect_ratio > 200:
+        raise ValueError(f"absolute aspect ratio must be smaller than 200, got {abs_aspect_ratio}")
+
+    h_bar = round(height / factor) * factor
+    w_bar = round(width / factor) * factor
+    if h_bar * w_bar > max_pixels:
+        beta = math.sqrt(height * width / max_pixels)
+        h_bar = math.floor(height / beta / factor) * factor
+        w_bar = math.floor(width / beta / factor) * factor
+    elif h_bar * w_bar < min_pixels:
+        beta = math.sqrt(min_pixels / (height * width))
+        h_bar = math.ceil(height * beta / factor) * factor
+        w_bar = math.ceil(width * beta / factor) * factor
+    return h_bar, w_bar
+
+
+def to_channel_dimension_format(
+    image: np.ndarray, channel_dim: ChannelDimension, input_channel_dim: Optional[ChannelDimension] = None
+) -> np.ndarray:
+    if not isinstance(image, np.ndarray):
+        raise TypeError(f"Input image must be of type np.ndarray, got {type(image)}")
+    if input_channel_dim is None:
+        input_channel_dim = infer_channel_dimension_format(image)
+    if input_channel_dim == channel_dim:
+        return image
+    if channel_dim == ChannelDimension.FIRST:
+        image = image.transpose((2, 0, 1))
+    elif channel_dim == ChannelDimension.LAST:
+        image = image.transpose((1, 2, 0))
+    else:
+        raise ValueError(f"Unsupported channel dimension format: {channel_dim}")
+    return image
+
+
+def get_channel_dimension_axis(image: np.ndarray, input_data_format: Optional[ChannelDimension] = None) -> int:
+    if input_data_format is None:
+        input_data_format = infer_channel_dimension_format(image)
+    if input_data_format == ChannelDimension.FIRST:
+        return image.ndim - 3
+    elif input_data_format == ChannelDimension.LAST:
+        return image.ndim - 1
+    raise ValueError(f"Unsupported channel dimension format: {input_data_format}")
+
+
+def rescale_image(
+    image: np.ndarray,
+    rescale_factor: float,
+    data_format: Optional[ChannelDimension] = None,
+    input_data_format: Optional[ChannelDimension] = None,
+) -> np.ndarray:
+    rescaled_image = image.astype(np.float64) * rescale_factor
+    if data_format is not None:
+        rescaled_image = to_channel_dimension_format(rescaled_image, data_format, input_data_format)
+    rescaled_image = rescaled_image.astype(np.float32)
+    return rescaled_image
+
+
+def normalize_image(
+    image: np.ndarray,
+    mean: List[float],
+    std: List[float],
+    data_format: Optional[ChannelDimension] = None,
+    input_data_format: Optional[ChannelDimension] = None,
+) -> np.ndarray:
+    if input_data_format is None:
+        input_data_format = infer_channel_dimension_format(image)
+    channel_axis = get_channel_dimension_axis(image, input_data_format)
+    num_channels = image.shape[channel_axis]
+    if len(mean) != num_channels:
+        raise ValueError(f"mean must have {num_channels} elements, but got {len(mean)}")
+    if len(std) != num_channels:
+        raise ValueError(f"std must have {num_channels} elements, but got {len(std)}")
+    if not np.issubdtype(image.dtype, np.floating):
+        image = image.astype(np.float32)
+    mean = np.array(mean, dtype=image.dtype)
+    std = np.array(std, dtype=image.dtype)
+    if input_data_format == ChannelDimension.LAST:
+        image = (image - mean) / std
+    else:
+        image = ((image.T - mean) / std).T
+    if data_format is not None:
+        image = to_channel_dimension_format(image, data_format, input_data_format)
+    return image
+
+
+def to_pil_image(
+    image: np.ndarray,
+    do_rescale: Optional[bool] = None,
+    input_data_format: Optional[ChannelDimension] = None,
+    image_mode: Optional[str] = None,
+) -> Image.Image:
+    image = to_channel_dimension_format(image, ChannelDimension.LAST, input_data_format)
+    image = np.squeeze(image, axis=-1) if image.shape[-1] == 1 else image
+    do_rescale = do_rescale if do_rescale is not None else _need_rescale_pil_conversion(image)
+    if do_rescale:
+        image = rescale_image(image, 255)
+    image = image.astype(np.uint8)
+    return Image.fromarray(image, mode=image_mode)
+
+
+def resize_image(
+    image: np.ndarray,
+    height: int,
+    width: int,
+    resample: Image.Resampling = Image.Resampling.BILINEAR,
+    reducing_gap: Optional[int] = None,
+    input_data_format: Optional[ChannelDimension] = None,
+    data_format: Optional[ChannelDimension] = None,
+) -> np.ndarray:
+    if input_data_format is None:
+        input_data_format = infer_channel_dimension_format(image)
+    data_format = data_format if data_format is not None else input_data_format
+    do_rescale = _need_rescale_pil_conversion(image)
+    pil_image = to_pil_image(image, do_rescale, input_data_format)
+    resized_image = pil_image.resize((width, height), resample=resample, reducing_gap=reducing_gap)
+    resized_image = np.array(resized_image)
+    resized_image = np.expand_dims(resized_image, axis=-1) if resized_image.ndim == 2 else resized_image
+    resized_image = to_channel_dimension_format(resized_image, data_format, ChannelDimension.LAST)
+    resized_image = rescale_image(resized_image, 1 / 255) if do_rescale else resized_image
+    return resized_image
+
+
+def _need_rescale_pil_conversion(image: np.ndarray) -> bool:
+    """
+    Detects whether or not the image needs to be rescaled before being converted to a PIL image.
+    The assumption is that if the image is of type `np.float` and all values are between 0 and 1, it needs to be
+    rescaled.
+    """
+    if image.dtype == np.uint8:
+        do_rescale = False
+    elif np.allclose(image, image.astype(int)):
+        if np.all(0 <= image) and np.all(image <= 255):
+            do_rescale = False
+        else:
+            raise ValueError(
+                "The image to be converted to a PIL image contains value outside the range [0, 255], "
+                f"got [{image.min()}, {image.max()}] which cannot be converted to uint8."
+            )
+    elif np.all(0 <= image) and np.all(image <= 1):
+        do_rescale = True
+    else:
+        raise ValueError(
+            "The image to be converted to PIL image contains values outside the range [0, 1]"
+            f"got [{image.min()}, {image.max()}] which cannot be converted to uint8."
+        )
+    return do_rescale

diffsynth_engine/utils/offload.py

@@ -3,6 +3,7 @@ import torch.nn as nn
 from typing import Dict
 import platform
 
+
 def enable_sequential_cpu_offload(module: nn.Module, device: str = "cuda"):
     module = module.to("cpu")
     if len(list(module.children())) == 0:
@@ -26,13 +27,13 @@ def add_cpu_offload_hook(module: nn.Module, device: str = "cuda", recurse: bool
             for name, buffer in module.named_buffers(recurse=recurse):
                 buffer.data = buffer.data.to(device=device)
             return tuple(x.to(device=device) if isinstance(x, torch.Tensor) else x for x in input_)
-        for name, param in module.named_parameters(recurse=recurse):            
-            if platform.system()  == 'Linux':
+        for name, param in module.named_parameters(recurse=recurse):
+            if platform.system() == "Linux":
                 param.data = param.data.pin_memory()
             offload_param_dict[name] = param.data
             param.data = param.data.to(device=device)
         for name, buffer in module.named_buffers(recurse=recurse):
-            if platform.system()  == 'Linux':            
+            if platform.system() == "Linux":
                 buffer.data = buffer.data.pin_memory()
             offload_param_dict[name] = buffer.data
             buffer.data = buffer.data.to(device=device)
@@ -59,11 +60,11 @@ def offload_model_to_dict(module: nn.Module) -> Dict[str, torch.Tensor]:
     module = module.to("cpu")
     offload_param_dict = {}
     for name, param in module.named_parameters(recurse=True):
-        if platform.system()  == 'Linux':        
+        if platform.system() == "Linux":
             param.data = param.data.pin_memory()
         offload_param_dict[name] = param.data
     for name, buffer in module.named_buffers(recurse=True):
-        if platform.system()  == 'Linux':        
+        if platform.system() == "Linux":
             buffer.data = buffer.data.pin_memory()
         offload_param_dict[name] = buffer.data
     return offload_param_dict

{diffsynth_engine-0.4.3.dev9.dist-info → diffsynth_engine-0.4.3.dev11.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: diffsynth_engine
-Version: 0.4.3.dev9
+Version: 0.4.3.dev11
 Author: MuseAI x ModelScope
 Classifier: Programming Language :: Python :: 3
 Classifier: Operating System :: OS Independent
@@ -29,7 +29,7 @@ Requires-Dist: scikit-image
 Requires-Dist: trimesh
 Provides-Extra: dev
 Requires-Dist: diffusers==0.31.0; extra == "dev"
-Requires-Dist: transformers==4.45.2; extra == "dev"
+Requires-Dist: transformers==4.52.4; extra == "dev"
 Requires-Dist: accelerate; extra == "dev"
 Requires-Dist: build; extra == "dev"
 Requires-Dist: ruff; extra == "dev"

{diffsynth_engine-0.4.3.dev9.dist-info → diffsynth_engine-0.4.3.dev11.dist-info}/RECORD

@@ -30,7 +30,7 @@ diffsynth_engine/conf/models/components/vae.json,sha256=EvlfeZ8hBCquep6JH1zVTcco
 diffsynth_engine/conf/models/flux/flux_dit.json,sha256=DdNjcNH6xI4Uz_sSoTkU0VGxYKhRDa3DdBu0g3M0rWc,5343
 diffsynth_engine/conf/models/flux/flux_text_encoder.json,sha256=AnSsOuOTesiQKW2UIuRWhpBE4Z_L6uFRpXTm3UN67FQ,748
 diffsynth_engine/conf/models/flux/flux_vae.json,sha256=2daLDQPjahufox1mEtsuYEAa0YyKpqvTPdeDYwUKlMQ,19970
-diffsynth_engine/conf/models/qwen_image/qwen2_5_vl_config.json,sha256=jfoXtZiCfom30iItDdfXPKhuH9EkVtXCkYtk5iD9Tr4,604
+diffsynth_engine/conf/models/qwen_image/qwen2_5_vl_config.json,sha256=VnOh5-PrKrOMlI3PohcazU-eZ4AaW_qKNGxNX70UVRE,629
 diffsynth_engine/conf/models/qwen_image/qwen2_5_vl_vision_config.json,sha256=Nkjg1A45MgMKXKfeP3zgIHd0KX_EEwG5ljkVXvEtT0A,364
 diffsynth_engine/conf/models/qwen_image/qwen_image_vae.json,sha256=eVLTSRqbXm3JD8QDkLbM6vFfCdynlS-8QxqCfi4BzrI,815
 diffsynth_engine/conf/models/qwen_image/qwen_image_vae_keymap.json,sha256=u9MJ3yRL45kdqRVoBnYbHkmuUmOseUFtwte-_9ZvdHc,25224
@@ -58,6 +58,7 @@ diffsynth_engine/conf/tokenizers/flux/tokenizer_2/special_tokens_map.json,sha256
 diffsynth_engine/conf/tokenizers/flux/tokenizer_2/spiece.model,sha256=1grLEoz3t_JTbo84pbGKBVNcnhTHo1WQQnDhWwlF6oY,791656
 diffsynth_engine/conf/tokenizers/flux/tokenizer_2/tokenizer.json,sha256=9d_sFjdl4Y4nBTf-iWxJ9frXTbFSVkHZslWjAIuZlZY,2424235
 diffsynth_engine/conf/tokenizers/flux/tokenizer_2/tokenizer_config.json,sha256=Gj0ttkIV7XeFTdQgiqxfg2HBtUccq9GcDvFHLRqJXrA,20817
+diffsynth_engine/conf/tokenizers/qwen_image/qwen2_vl_image_processor.json,sha256=u_Uw_SNp6qET_TWe0Hujm3v75NDEvizHetHrgpVC7CM,644
 diffsynth_engine/conf/tokenizers/qwen_image/tokenizer/added_tokens.json,sha256=2JrIc9odOnjBcOLt7aweUkMEsf3fWQ1kMg0WsMvKZYM,648
 diffsynth_engine/conf/tokenizers/qwen_image/tokenizer/merges.txt,sha256=iDHk8aBERxNA98CoPXvXEwaluGfpX9hw900MUwipBNU,1671853
 diffsynth_engine/conf/tokenizers/qwen_image/tokenizer/special_tokens_map.json,sha256=21yCGVKkEO6UtBk0EeHZDSNEXlBuY1J2Zbc6TZBnEPY,716
@@ -78,13 +79,13 @@ diffsynth_engine/conf/tokenizers/wan/umt5-xxl/tokenizer.json,sha256=bhl7TT29cdoU
 diffsynth_engine/conf/tokenizers/wan/umt5-xxl/tokenizer_config.json,sha256=7Zo6iw-qcacKMoR-BDX-A25uES1N9O23u0ipIeNE3AU,61728
 diffsynth_engine/configs/__init__.py,sha256=7d7DWeWxRlXdU5y3NinEdwlwC4W-V5U3KSBOmUwL_tc,778
 diffsynth_engine/configs/controlnet.py,sha256=EpUkCdRNk2G5uo56syaOzPFdR9g0sDHRXckagmMsgaQ,948
-diffsynth_engine/configs/pipeline.py,sha256=3ZcppVFMfKCk57Z8bYhEr3tvHBvCSMMBHcxyQICixTM,10995
+diffsynth_engine/configs/pipeline.py,sha256=ZBNNJQB3uvle2urEAFn-_cCqqUlJkcpgp-GLepZn24s,11143
 diffsynth_engine/kernels/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 diffsynth_engine/models/__init__.py,sha256=8Ze7cSE8InetgXWTNb0neVA2Q44K7WlE-h7O-02m2sY,119
 diffsynth_engine/models/base.py,sha256=sbyyGP-ENnqicr6cxjEmXRf6dWrmKjCu6k5yamuJ518,2665
 diffsynth_engine/models/utils.py,sha256=r5xLSEog1_ODaFrpqzJvAj3r23PQiEpgivzErClTZTg,1561
 diffsynth_engine/models/basic/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-diffsynth_engine/models/basic/attention.py,sha256=PJ3xQ8ifGeoiEBh-vDRI32-v8Syd8ostuhll3Yo5Nsk,10323
+diffsynth_engine/models/basic/attention.py,sha256=H-LKiVD09GFj6hIwiIb5wiDS1HVCslN_TXgR7FrkUX4,10343
 diffsynth_engine/models/basic/lora.py,sha256=qEh44zfh7ZBblLpjmKzwzAxmTlVyY0wu9IkGsnr7Ih8,10614
 diffsynth_engine/models/basic/relative_position_emb.py,sha256=rCXOweZMcayVnNUVvBcYXMdhHS257B_PC8PZSWxvhNQ,2540
 diffsynth_engine/models/basic/timestep.py,sha256=WJODYqkSXEM0wcS42YkkfrGwxWt0e60zMTkDdUBQqBw,2810
@@ -106,8 +107,8 @@ diffsynth_engine/models/hunyuan3d/moe.py,sha256=FAuUqgrB2ZFb0uGBhI-Afv850HmzDFP5
 diffsynth_engine/models/hunyuan3d/surface_extractor.py,sha256=b15mb1N4PYwAvDk1Gude8qlccRKrSg461xT59RjMEQk,4167
 diffsynth_engine/models/hunyuan3d/volume_decoder.py,sha256=sgflj1a8sIerqGSalBAVQOlyiIihkLOLXYysNbulCoQ,2355
 diffsynth_engine/models/qwen_image/__init__.py,sha256=X5pig621WEsDZ6L7HVkmYspV53-GDfs_la1ncaq_NFw,417
-diffsynth_engine/models/qwen_image/qwen2_5_vl.py,sha256=vpBo6eo_96iLky9YV5MX0nbmOleY2EX97TrJoRBNnw4,56511
-diffsynth_engine/models/qwen_image/qwen_image_dit.py,sha256=qxwuB42rrVy2R-gBNcP_9ykvgKgXP3BuxLhU-0ArPas,16949
+diffsynth_engine/models/qwen_image/qwen2_5_vl.py,sha256=ykq-A7kPZedMNnvsJuHXHEgDgsQoTYhk8HQDTOmxz2Q,55448
+diffsynth_engine/models/qwen_image/qwen_image_dit.py,sha256=RSffbhK2YmcXLdzGilYXZ-URuaKRIfGqoRmWVROsXP4,17698
 diffsynth_engine/models/qwen_image/qwen_image_dit_fbcache.py,sha256=tgCmD4MFNgd3HmLyoYnt8HZCzPBTgQ4zCQjV5qZSW_I,4870
 diffsynth_engine/models/qwen_image/qwen_image_vae.py,sha256=m455iJfJx8KVEsrkinjClokkEqd0RshSZoKZ_QAdRyk,38509
 diffsynth_engine/models/sd/__init__.py,sha256=hjoKRnwoXOLD0wude-w7I6wK5ak7ACMbnbkPuBB2oU0,380
@@ -136,21 +137,23 @@ diffsynth_engine/models/wan/wan_image_encoder.py,sha256=LYwcfCcQmXf9FP08DGaU2bfa
 diffsynth_engine/models/wan/wan_text_encoder.py,sha256=bkphxtqNNwXcEA_OaUrwV9CvICV-s16awu5Z9gjjzsM,10912
 diffsynth_engine/models/wan/wan_vae.py,sha256=AmBuqyPwZCFY0e8lUThlJoNHmpmTm2_dE1XYzXBCaAI,38937
 diffsynth_engine/pipelines/__init__.py,sha256=bEdXa533rXgjySO0aiLlfLkVmxoP6Yy_I4kj3WkpxnI,528
-diffsynth_engine/pipelines/base.py,sha256=goe_UO1LvUXVwP5geUmu0zdFUrSms9iss3OuRyuMjXY,13726
+diffsynth_engine/pipelines/base.py,sha256=RTkVwWaWXr5ujqn5-UBHvdPddYwr-uvChj9-fmoXrms,13729
 diffsynth_engine/pipelines/flux_image.py,sha256=Cwak7zZ86Um63YQB-fsiW87ip8lpzn2kXXS2Btune6o,49176
 diffsynth_engine/pipelines/hunyuan3d_shape.py,sha256=fwNKET54KjCiWDpW2S1Fk-p3nfJreZ-RH7p46VLawEQ,7911
-diffsynth_engine/pipelines/qwen_image.py,sha256=v99ZnERw_g-opET7Sh5XNTFLr_SV9c4wBRfWCVXFayA,18411
-diffsynth_engine/pipelines/sd_image.py,sha256=GhrCadEmAWv4id0NdRpJW_EC2PgItBctXLkfPxq5gDI,18100
-diffsynth_engine/pipelines/sdxl_image.py,sha256=kmidIz8zDtrw9ggLXI3WG7AQq_jmOPVct-O3hGNra_g,21951
+diffsynth_engine/pipelines/qwen_image.py,sha256=QDbiI7xpc8fPgskygz-246zzJOLRLkk-X2ZHK3pRf_Y,23606
+diffsynth_engine/pipelines/sd_image.py,sha256=yreiiQDWaMA5LOgLs_GDoC1hLMVk2LRMk99RLWQ1W90,18076
+diffsynth_engine/pipelines/sdxl_image.py,sha256=-gRkNP9EpPRR0plnCWiSSXVNkIQd2zTVvJgziZuxUQU,21975
 diffsynth_engine/pipelines/utils.py,sha256=lk7sFGEk-fGjgadLpwwppHKG-yZ0RC-4ZmHW7pRRe8A,473
 diffsynth_engine/pipelines/wan_video.py,sha256=l6GUfv7WJ-Hue5MZH9L3eXWHA-I0oOelxCNwUkjMSRg,26132
 diffsynth_engine/processor/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 diffsynth_engine/processor/canny_processor.py,sha256=hV30NlblTkEFUAmF_O-LJrNlGVM2SFrqq6okfF8VpOo,602
 diffsynth_engine/processor/depth_processor.py,sha256=dQvs3JsnyMbz4dyI9QoR8oO-mMFBFAgNvgqeCoaU5jk,1532
-diffsynth_engine/tokenizers/__init__.py,sha256=l4PMxLs6WQ_2ah0LXApkYsQY8-2HQZTHcQHpsWigNnI,294
+diffsynth_engine/tokenizers/__init__.py,sha256=KxTna7UrkptrBU1j3zBYOi_8mVEWlcSMGZwK2ahuHNw,456
 diffsynth_engine/tokenizers/base.py,sha256=JX4C8FX7Y-glpszk39pUlS0QvqRyiKBYr-GWrrpee78,5149
 diffsynth_engine/tokenizers/clip.py,sha256=6yggDSRGZc34CKflO1DwTIisggv53aITe_h-YnsERzc,10695
 diffsynth_engine/tokenizers/qwen2.py,sha256=NDuE0hs1c4WfHOYUZ9KC0L-1vqOQ0Pj7ugWMqp24rys,9263
+diffsynth_engine/tokenizers/qwen2_vl_image_processor.py,sha256=7IBOn2m4AbL-URVrSrFY0k88r4_gkK_nuTQRAxorBes,6239
+diffsynth_engine/tokenizers/qwen2_vl_processor.py,sha256=Zyu8_5ETCjACQ8BX6jvVRWj37nZqJgtI0hesSUGm4-g,4145
 diffsynth_engine/tokenizers/t5.py,sha256=brhRFkXaTzE29hl_wDdcjQ3MCoL0pQslwHIRbMX_bNo,7442
 diffsynth_engine/tokenizers/wan.py,sha256=4bvibHZKNQYHnl0oSyN_pJK5PAxkUC0TWgbNDgckdCQ,2265
 diffsynth_engine/tools/__init__.py,sha256=3sgNSD5sts3cHSyKVH5qziTykJQ6gDRMx_RyrrxSzZ8,388
@@ -160,17 +163,17 @@ diffsynth_engine/tools/flux_reference_tool.py,sha256=wY3ey7s5vZLu9gYBCdZZveapQJY
 diffsynth_engine/tools/flux_replace_tool.py,sha256=ZUpc2T1nMRXOBHGnBN8MImjgs3hoRmJ-SludpZ1u0bM,4557
 diffsynth_engine/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 diffsynth_engine/utils/cache.py,sha256=Ivef22pCuhEq-4H00gSvkLS8ceVZoGis7OSitYL6gH4,2101
-diffsynth_engine/utils/constants.py,sha256=W9znMP8O7td3RF_2ykeIARCWVtIxfro9ytmlXR_GG2c,3121
+diffsynth_engine/utils/constants.py,sha256=csKPtXA9YFoRPnhUBmnw3zf1KlTV963a43Nu4j9TqOE,3354
 diffsynth_engine/utils/download.py,sha256=8O56zQr2taY4BnCBEKVk3YGorTqTTCqTblf7tI9otXA,6721
 diffsynth_engine/utils/env.py,sha256=43x-kBjt5zI2cwZ9G4BOeTbedi2k6TuBzHGOBeFbFvU,280
 diffsynth_engine/utils/flag.py,sha256=6zQLnoEaU69pBEyhavCgydQfP0khw5ppCU7sue4yRqg,1370
 diffsynth_engine/utils/fp8_linear.py,sha256=NosnWMoAr_IpFcLn-OYbAx-vXySphjxutDZqmXLNjJI,4064
 diffsynth_engine/utils/gguf.py,sha256=ZWvw46V4g4uVyAR_oCq-4K5nPdKVrYk3u47uXMgA9lU,14092
-diffsynth_engine/utils/image.py,sha256=_46CVs1Qe7GdZNulWWJISnR_Y6FotC2tZGLKtr04gIE,562
+diffsynth_engine/utils/image.py,sha256=xZ_bEU-DdoSwMPG7jpP1daAro2tsy9ddeXKbXqXaeC0,9335
 diffsynth_engine/utils/loader.py,sha256=Z5v1WNDWFY0OrVubB70j5VU3zeaAfEK_j8c1KrGI4yM,1240
 diffsynth_engine/utils/lock.py,sha256=1Ipgst9eEFfFdViAvD5bxdB6HnHHBcqWYOb__fGaPUI,1601
 diffsynth_engine/utils/logging.py,sha256=XB0xTT8PBN6btkOjFtOvjlrOCRVgDGT8PFAp1vmse28,467
-diffsynth_engine/utils/offload.py,sha256=8dHMEzAQcE-GNbFYotGQg2YBBB0eywoRTIffVmK4A0k,3746
+diffsynth_engine/utils/offload.py,sha256=d8cVb3ToYGbbeNnCd2dSRqR5Rm4lxrviGCTTEZ-cjo0,3703
 diffsynth_engine/utils/onnx.py,sha256=jeWUudJHnESjuiEAHyUZYUZz7dCj34O9aGjHCe8yjWo,1149
 diffsynth_engine/utils/parallel.py,sha256=Z9jqCv4mLV4JyXR3uTHyv1rujPiKU8PSCbAfiN9jkPc,16818
 diffsynth_engine/utils/platform.py,sha256=2lXdw6YkqcRONCeT98n4cyg1Ii8Ybbyj2Ns72Se9tlk,496
@@ -179,8 +182,8 @@ diffsynth_engine/utils/video.py,sha256=Ne0rd2lb59UT1q5EotpjlY7OT8F9oTCFDyo1ST77u
 diffsynth_engine/utils/memory/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 diffsynth_engine/utils/memory/linear_regression.py,sha256=oW_EQEw13oPoyUrxiL8A7Ksa5AuJ2ynI2qhCbfAuZbg,3930
 diffsynth_engine/utils/memory/memory_predcit_model.py,sha256=qYlp0JvZ02IhLD8YVEHrtVPQcGLNHVijpUit0dz4FWE,3896
-diffsynth_engine-0.4.3.dev9.dist-info/licenses/LICENSE,sha256=x7aBqQuVI0IYnftgoTPI_A0I_rjdjPPQkjnU6N2nikM,11346
-diffsynth_engine-0.4.3.dev9.dist-info/METADATA,sha256=sxEnokIrLWZwbL9jkLzpT84VfWXIHDmqrdDma9Yjuxg,1117
-diffsynth_engine-0.4.3.dev9.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-diffsynth_engine-0.4.3.dev9.dist-info/top_level.txt,sha256=6zgbiIzEHLbhgDKRyX0uBJOV3F6VnGGBRIQvSiYYn6w,17
-diffsynth_engine-0.4.3.dev9.dist-info/RECORD,,
+diffsynth_engine-0.4.3.dev11.dist-info/licenses/LICENSE,sha256=x7aBqQuVI0IYnftgoTPI_A0I_rjdjPPQkjnU6N2nikM,11346
+diffsynth_engine-0.4.3.dev11.dist-info/METADATA,sha256=pxJONCO3LVvgJA0OQjR7vUqJWKXlxEpbZe6J6mB5JcI,1118
+diffsynth_engine-0.4.3.dev11.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+diffsynth_engine-0.4.3.dev11.dist-info/top_level.txt,sha256=6zgbiIzEHLbhgDKRyX0uBJOV3F6VnGGBRIQvSiYYn6w,17
+diffsynth_engine-0.4.3.dev11.dist-info/RECORD,,