HunyuanImage-2.1

diffsynth/configs/model_configs.py

@@ -312,7 +312,58 @@
         "model_hash": "0629116fce1472503a66992f96f3eb1a",
         "model_name": "flux_value_controller",
         "model_class": "diffsynth.models.flux_value_control.SingleValueEncoder",
-    }
+    },
+    {
+        # Example: ModelConfig(model_id="alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Beta", origin_file_pattern="diffusion_pytorch_model.safetensors")
+        "model_hash": "52357cb26250681367488a8954c271e8",
+        "model_name": "flux_controlnet",
+        "model_class": "diffsynth.models.flux_controlnet.FluxControlNet",
+        "state_dict_converter": "diffsynth.utils.state_dict_converters.flux_controlnet.FluxControlNetStateDictConverter",
+        "extra_kwargs": {"num_joint_blocks": 6, "num_single_blocks": 0, "additional_input_dim": 4},
+    },
+    {
+        # Example: ModelConfig(model_id="InstantX/FLUX.1-dev-Controlnet-Union-alpha", origin_file_pattern="diffusion_pytorch_model.safetensors")
+        "model_hash": "78d18b9101345ff695f312e7e62538c0",
+        "model_name": "flux_controlnet",
+        "model_class": "diffsynth.models.flux_controlnet.FluxControlNet",
+        "state_dict_converter": "diffsynth.utils.state_dict_converters.flux_controlnet.FluxControlNetStateDictConverter",
+        "extra_kwargs": {"num_mode": 10, "mode_dict": {"canny": 0, "tile": 1, "depth": 2, "blur": 3, "pose": 4, "gray": 5, "lq": 6}},
+    },
+    {
+        # Example: ModelConfig(model_id="jasperai/Flux.1-dev-Controlnet-Upscaler", origin_file_pattern="diffusion_pytorch_model.safetensors")
+        "model_hash": "b001c89139b5f053c715fe772362dd2a",
+        "model_name": "flux_controlnet",
+        "model_class": "diffsynth.models.flux_controlnet.FluxControlNet",
+        "state_dict_converter": "diffsynth.utils.state_dict_converters.flux_controlnet.FluxControlNetStateDictConverter",
+        "extra_kwargs": {"num_single_blocks": 0},
+    },
+    {
+        # Example: ModelConfig(model_id="ByteDance/InfiniteYou", origin_file_pattern="infu_flux_v1.0/aes_stage2/image_proj_model.bin")
+        "model_hash": "c07c0f04f5ff55e86b4e937c7a40d481",
+        "model_name": "infiniteyou_image_projector",
+        "model_class": "diffsynth.models.flux_infiniteyou.InfiniteYouImageProjector",
+        "state_dict_converter": "diffsynth.utils.state_dict_converters.flux_infiniteyou.FluxInfiniteYouImageProjectorStateDictConverter",
+    },
+    {
+        # Example: ModelConfig(model_id="ByteDance/InfiniteYou", origin_file_pattern="infu_flux_v1.0/aes_stage2/InfuseNetModel/*.safetensors")
+        "model_hash": "7f9583eb8ba86642abb9a21a4b2c9e16",
+        "model_name": "flux_controlnet",
+        "model_class": "diffsynth.models.flux_controlnet.FluxControlNet",
+        "state_dict_converter": "diffsynth.utils.state_dict_converters.flux_controlnet.FluxControlNetStateDictConverter",
+        "extra_kwargs": {"num_joint_blocks": 4, "num_single_blocks": 10},
+    },
+    {
+        # Example: ModelConfig(model_id="DiffSynth-Studio/LoRA-Encoder-FLUX.1-Dev", origin_file_pattern="model.safetensors")
+        "model_hash": "77c2e4dd2440269eb33bfaa0d004f6ab",
+        "model_name": "flux_lora_encoder",
+        "model_class": "diffsynth.models.flux_lora_encoder.FluxLoRAEncoder",
+    },
+    {
+        # Example: ModelConfig(model_id="DiffSynth-Studio/LoRAFusion-preview-FLUX.1-dev", origin_file_pattern="model.safetensors")
+        "model_hash": "30143afb2dea73d1ac580e0787628f8c",
+        "model_name": "flux_lora_patcher",
+        "model_class": "diffsynth.models.flux_lora_patcher.FluxLoraPatcher",
+    },
 ]
 
 MODEL_CONFIGS = qwen_image_series + wan_series + flux_series

diffsynth/models/flux_controlnet.py

@@ -1,9 +1,62 @@
 import torch
 from einops import rearrange, repeat
 from .flux_dit import RoPEEmbedding, TimestepEmbeddings, FluxJointTransformerBlock, FluxSingleTransformerBlock, RMSNorm
-from .utils import hash_state_dict_keys, init_weights_on_device
+# from .utils import hash_state_dict_keys, init_weights_on_device
+from contextlib import contextmanager
 
+def hash_state_dict_keys(state_dict, with_shape=True):
+    keys_str = convert_state_dict_keys_to_single_str(state_dict, with_shape=with_shape)
+    keys_str = keys_str.encode(encoding="UTF-8")
+    return hashlib.md5(keys_str).hexdigest()
 
+@contextmanager
+def init_weights_on_device(device = torch.device("meta"), include_buffers :bool = False):
+
+    old_register_parameter = torch.nn.Module.register_parameter
+    if include_buffers:
+        old_register_buffer = torch.nn.Module.register_buffer
+
+    def register_empty_parameter(module, name, param):
+        old_register_parameter(module, name, param)
+        if param is not None:
+            param_cls = type(module._parameters[name])
+            kwargs = module._parameters[name].__dict__
+            kwargs["requires_grad"] = param.requires_grad
+            module._parameters[name] = param_cls(module._parameters[name].to(device), **kwargs)
+
+    def register_empty_buffer(module, name, buffer, persistent=True):
+        old_register_buffer(module, name, buffer, persistent=persistent)
+        if buffer is not None:
+            module._buffers[name] = module._buffers[name].to(device)
+
+    def patch_tensor_constructor(fn):
+        def wrapper(*args, **kwargs):
+            kwargs["device"] = device
+            return fn(*args, **kwargs)
+
+        return wrapper
+
+    if include_buffers:
+        tensor_constructors_to_patch = {
+            torch_function_name: getattr(torch, torch_function_name)
+            for torch_function_name in ["empty", "zeros", "ones", "full"]
+        }
+    else:
+        tensor_constructors_to_patch = {}
+
+    try:
+        torch.nn.Module.register_parameter = register_empty_parameter
+        if include_buffers:
+            torch.nn.Module.register_buffer = register_empty_buffer
+        for torch_function_name in tensor_constructors_to_patch.keys():
+            setattr(torch, torch_function_name, patch_tensor_constructor(getattr(torch, torch_function_name)))
+        yield
+    finally:
+        torch.nn.Module.register_parameter = old_register_parameter
+        if include_buffers:
+            torch.nn.Module.register_buffer = old_register_buffer
+        for torch_function_name, old_torch_function in tensor_constructors_to_patch.items():
+            setattr(torch, torch_function_name, old_torch_function)
 
 class FluxControlNet(torch.nn.Module):
     def __init__(self, disable_guidance_embedder=False, num_joint_blocks=5, num_single_blocks=10, num_mode=0, mode_dict={}, additional_input_dim=0):
@@ -102,9 +155,9 @@ def forward(
         return controlnet_res_stack, controlnet_single_res_stack
 
 
-    @staticmethod
-    def state_dict_converter():
-        return FluxControlNetStateDictConverter()
+    # @staticmethod
+    # def state_dict_converter():
+    #     return FluxControlNetStateDictConverter()
 
     def quantize(self):
         def cast_to(weight, dtype=None, device=None, copy=False):