init

feature_extractor/preprocessor_config.json

@@ -0,0 +1,27 @@
+{
+  "crop_size": {
+    "height": 224,
+    "width": 224
+  },
+  "do_center_crop": true,
+  "do_convert_rgb": true,
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.48145466,
+    0.4578275,
+    0.40821073
+  ],
+  "image_processor_type": "CLIPImageProcessor",
+  "image_std": [
+    0.26862954,
+    0.26130258,
+    0.27577711
+  ],
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "shortest_edge": 224
+  }
+}

image_encoder/config.json

@@ -0,0 +1,23 @@
+{
+  "_name_or_path": "Luffuly/unique3d-mvimage-diffuser",
+  "architectures": [
+    "CLIPVisionModelWithProjection"
+  ],
+  "attention_dropout": 0.0,
+  "dropout": 0.0,
+  "hidden_act": "quick_gelu",
+  "hidden_size": 1024,
+  "image_size": 224,
+  "initializer_factor": 1.0,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "model_type": "clip_vision_model",
+  "num_attention_heads": 16,
+  "num_channels": 3,
+  "num_hidden_layers": 24,
+  "patch_size": 14,
+  "projection_dim": 768,
+  "torch_dtype": "float16",
+  "transformers_version": "4.45.2"
+}

image_encoder/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0dd0c1777bb75e2e9b7aa29b799b3ee8ebaaaa731e2c471a9ec589f12542cce7
+size 607980096

model_index.json

@@ -0,0 +1,29 @@
+{
+  "_class_name": "MVDiffusionImagePipeline",
+  "_diffusers_version": "0.30.3",
+  "feature_extractor": [
+    "transformers",
+    "CLIPImageProcessor"
+  ],
+  "image_encoder": [
+    "transformers",
+    "CLIPVisionModelWithProjection"
+  ],
+  "requires_safety_checker": true,
+  "safety_checker": [
+    null,
+    null
+  ],
+  "scheduler": [
+    "diffusers",
+    "DDIMScheduler"
+  ],
+  "unet": [
+    "mv_unet",
+    "UnifieldWrappedUNet"
+  ],
+  "vae": [
+    "diffusers",
+    "AutoencoderKL"
+  ]
+}

scheduler/scheduler_config.json

@@ -0,0 +1,20 @@
+{
+  "_class_name": "DDIMScheduler",
+  "_diffusers_version": "0.30.3",
+  "beta_end": 0.012,
+  "beta_schedule": "scaled_linear",
+  "beta_start": 0.00085,
+  "clip_sample": false,
+  "clip_sample_range": 1.0,
+  "dynamic_thresholding_ratio": 0.995,
+  "num_train_timesteps": 1000,
+  "prediction_type": "epsilon",
+  "rescale_betas_zero_snr": false,
+  "sample_max_value": 1.0,
+  "set_alpha_to_one": false,
+  "skip_prk_steps": true,
+  "steps_offset": 1,
+  "thresholding": false,
+  "timestep_spacing": "leading",
+  "trained_betas": null
+}

unet/config.json

@@ -0,0 +1,69 @@
+{
+  "_class_name": "UnifieldWrappedUNet",
+  "_diffusers_version": "0.30.3",
+  "_name_or_path": "outputs/vroid-mvimage-6view/checkpoint",
+  "act_fn": "silu",
+  "addition_embed_type": null,
+  "addition_embed_type_num_heads": 64,
+  "addition_time_embed_dim": null,
+  "attention_head_dim": 8,
+  "attention_type": "default",
+  "block_out_channels": [
+    320,
+    640,
+    1280,
+    1280
+  ],
+  "center_input_sample": false,
+  "class_embed_type": null,
+  "class_embeddings_concat": false,
+  "conv_in_kernel": 3,
+  "conv_out_kernel": 3,
+  "cross_attention_dim": 768,
+  "cross_attention_norm": null,
+  "down_block_types": [
+    "CrossAttnDownBlock2D",
+    "CrossAttnDownBlock2D",
+    "CrossAttnDownBlock2D",
+    "DownBlock2D"
+  ],
+  "downsample_padding": 1,
+  "dropout": 0.0,
+  "dual_cross_attention": false,
+  "encoder_hid_dim": null,
+  "encoder_hid_dim_type": null,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "in_channels": 8,
+  "layers_per_block": 2,
+  "mid_block_only_cross_attention": null,
+  "mid_block_scale_factor": 1,
+  "mid_block_type": "UNetMidBlock2DCrossAttn",
+  "norm_eps": 1e-05,
+  "norm_num_groups": 32,
+  "num_attention_heads": null,
+  "num_class_embeds": 8,
+  "only_cross_attention": false,
+  "out_channels": 4,
+  "projection_class_embeddings_input_dim": null,
+  "resnet_out_scale_factor": 1.0,
+  "resnet_skip_time_act": false,
+  "resnet_time_scale_shift": "default",
+  "reverse_transformer_layers_per_block": null,
+  "sample_size": 64,
+  "time_cond_proj_dim": null,
+  "time_embedding_act_fn": null,
+  "time_embedding_dim": null,
+  "time_embedding_type": "positional",
+  "timestep_post_act": null,
+  "transformer_layers_per_block": 1,
+  "up_block_types": [
+    "UpBlock2D",
+    "CrossAttnUpBlock2D",
+    "CrossAttnUpBlock2D",
+    "CrossAttnUpBlock2D"
+  ],
+  "n_views": 6,
+  "upcast_attention": false,
+  "use_linear_projection": false
+}

unet/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1134886018ecde11b10f01ae71557789b24762d0c9263524c3fb2bfeff99f02
+size 3438254688

unet/mv_unet.py

@@ -0,0 +1,177 @@
+import torch
+from typing import Optional, Tuple, Union
+from diffusers import UNet2DConditionModel
+from diffusers.models.attention_processor import Attention
+from diffusers.models.unets.unet_2d_condition import UNet2DConditionOutput
+
+
+def switch_multiview_processor(model, enable_filter=lambda x:True):
+    def recursive_add_processors(name: str, module: torch.nn.Module):
+        for sub_name, child in module.named_children():
+            recursive_add_processors(f"{name}.{sub_name}", child)
+
+        if isinstance(module, Attention):
+            processor = module.get_processor()
+            if isinstance(processor, multiviewAttnProc):
+                processor.enabled = enable_filter(f"{name}.processor")
+
+    for name, module in model.named_children():
+        recursive_add_processors(name, module)
+
+
+def add_multiview_processor(model: torch.nn.Module, enable_filter=lambda x:True, **kwargs):
+    return_dict = torch.nn.ModuleDict()
+    def recursive_add_processors(name: str, module: torch.nn.Module):
+        for sub_name, child in module.named_children():
+            if "ref_unet" not in (sub_name + name):
+                recursive_add_processors(f"{name}.{sub_name}", child)
+
+        if isinstance(module, Attention):
+            new_processor = multiviewAttnProc(
+                chained_proc=module.get_processor(),
+                enabled=enable_filter(f"{name}.processor"),
+                name=f"{name}.processor",
+                hidden_states_dim=module.inner_dim,
+                **kwargs
+            )
+            module.set_processor(new_processor)
+            return_dict[f"{name}.processor".replace(".", "__")] = new_processor
+
+    for name, module in model.named_children():
+        recursive_add_processors(name, module)
+
+    return return_dict
+
+
+class multiviewAttnProc(torch.nn.Module):
+    def __init__(
+        self,
+        chained_proc,
+        enabled=False,
+        name=None,
+        hidden_states_dim=None,
+        chain_pos="parralle",     # before or parralle or after
+        num_modalities=1,
+        views=4,
+        base_img_size=64,
+    ) -> None:
+        super().__init__()
+        self.enabled = enabled
+        self.chained_proc = chained_proc
+        self.name = name
+        self.hidden_states_dim = hidden_states_dim
+        self.num_modalities = num_modalities
+        self.views = views
+        self.base_img_size = base_img_size
+        self.chain_pos = chain_pos
+        self.diff_joint_attn = True
+
+    def __call__(
+        self,
+        attn: Attention,
+        hidden_states: torch.FloatTensor,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        **kwargs
+    ) -> torch.Tensor:
+        if not self.enabled:
+            return self.chained_proc(attn, hidden_states, encoder_hidden_states, attention_mask, **kwargs)
+        
+        B, L, C = hidden_states.shape
+        mv = self.views
+        hidden_states = hidden_states.reshape(B // mv, mv, L, C).reshape(-1, mv * L, C)
+        hidden_states = self.chained_proc(attn, hidden_states, encoder_hidden_states, attention_mask, **kwargs)
+        return hidden_states.reshape(B // mv, mv, L, C).reshape(-1, L, C)
+
+
+
+class UnifieldWrappedUNet(UNet2DConditionModel):
+    def __init__(  
+            self,
+            sample_size: Optional[int] = None,
+            in_channels: int = 4,
+            out_channels: int = 4,
+            center_input_sample: bool = False,
+            flip_sin_to_cos: bool = True,
+            freq_shift: int = 0,
+            down_block_types: Tuple[str] = (
+                "CrossAttnDownBlock2D",
+                "CrossAttnDownBlock2D",
+                "CrossAttnDownBlock2D",
+                "DownBlock2D",
+            ),
+            mid_block_type: Optional[str] = "UNetMidBlock2DCrossAttn",
+            up_block_types: Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D"),
+            only_cross_attention: Union[bool, Tuple[bool]] = False,
+            block_out_channels: Tuple[int] = (320, 640, 1280, 1280),
+            layers_per_block: Union[int, Tuple[int]] = 2,
+            downsample_padding: int = 1,
+            mid_block_scale_factor: float = 1,
+            dropout: float = 0.0,
+            act_fn: str = "silu",
+            norm_num_groups: Optional[int] = 32,
+            norm_eps: float = 1e-5,
+            cross_attention_dim: Union[int, Tuple[int]] = 1280,
+            transformer_layers_per_block: Union[int, Tuple[int], Tuple[Tuple]] = 1,
+            reverse_transformer_layers_per_block: Optional[Tuple[Tuple[int]]] = None,
+            encoder_hid_dim: Optional[int] = None,
+            encoder_hid_dim_type: Optional[str] = None,
+            attention_head_dim: Union[int, Tuple[int]] = 8,
+            num_attention_heads: Optional[Union[int, Tuple[int]]] = None,
+            dual_cross_attention: bool = False,
+            use_linear_projection: bool = False,
+            class_embed_type: Optional[str] = None,
+            addition_embed_type: Optional[str] = None,
+            addition_time_embed_dim: Optional[int] = None,
+            num_class_embeds: Optional[int] = None,
+            upcast_attention: bool = False,
+            resnet_time_scale_shift: str = "default",
+            resnet_skip_time_act: bool = False,
+            resnet_out_scale_factor: float = 1.0,
+            time_embedding_type: str = "positional",
+            time_embedding_dim: Optional[int] = None,
+            time_embedding_act_fn: Optional[str] = None,
+            timestep_post_act: Optional[str] = None,
+            time_cond_proj_dim: Optional[int] = None,
+            conv_in_kernel: int = 3,
+            conv_out_kernel: int = 3,
+            projection_class_embeddings_input_dim: Optional[int] = None,
+            attention_type: str = "default",
+            class_embeddings_concat: bool = False,
+            mid_block_only_cross_attention: Optional[bool] = None,
+            cross_attention_norm: Optional[str] = None,
+            addition_embed_type_num_heads: int = 64,
+            multiview_attn_position: str = "attn1", 
+            n_views: int = 4,
+            num_modalities: int = 1, 
+            latent_size: int = 64, 
+            multiview_chain_pose: str = "parralle", 
+            **kwargs
+        ):   
+        super().__init__(**{
+                k: v for k, v in locals().items() if k not in 
+                ["self", "kwargs", "__class__", "multiview_attn_position", "n_views", "num_modalities", "latent_size", "multiview_chain_pose"]
+            })
+        add_multiview_processor(
+                model = self, 
+                enable_filter = lambda name: name.endswith(f"{multiview_attn_position}.processor"),
+                num_modalities = num_modalities,    
+                base_img_size = latent_size,      
+                chain_pos = multiview_chain_pose,
+                views=n_views
+            ) 
+        
+        switch_multiview_processor(self, enable_filter=lambda name: name.endswith(f"{multiview_attn_position}.processor"))
+
+    def __call__(
+        self,
+        sample: torch.Tensor,
+        timestep: Union[torch.Tensor, float, int],
+        encoder_hidden_states: torch.Tensor,
+        condition_latens: torch.Tensor = None,
+        class_labels: Optional[torch.Tensor] = None,  
+    ) -> Union[UNet2DConditionOutput, Tuple]:
+        sample = torch.cat([sample, condition_latens], dim=1)
+        return self.forward( 
+            sample, timestep, encoder_hidden_states, class_labels=class_labels,  
+        ) 

vae/config.json

@@ -0,0 +1,38 @@
+{
+  "_class_name": "AutoencoderKL",
+  "_diffusers_version": "0.30.3",
+  "_name_or_path": "Luffuly/unique3d-mvimage-diffuser",
+  "act_fn": "silu",
+  "block_out_channels": [
+    128,
+    256,
+    512,
+    512
+  ],
+  "down_block_types": [
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D"
+  ],
+  "force_upcast": true,
+  "in_channels": 3,
+  "latent_channels": 4,
+  "latents_mean": null,
+  "latents_std": null,
+  "layers_per_block": 2,
+  "mid_block_add_attention": true,
+  "norm_num_groups": 32,
+  "out_channels": 3,
+  "sample_size": 256,
+  "scaling_factor": 0.18215,
+  "shift_factor": null,
+  "up_block_types": [
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D"
+  ],
+  "use_post_quant_conv": true,
+  "use_quant_conv": true
+}

vae/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e4c08995484ee61270175e9e7a072b66a6e4eeb5f0c266667fe1f45b90daf9a
+size 167335342