bugfix

app.py

@@ -5,9 +5,12 @@ import numpy as np
 import random
 import spaces
 import cv2
+from diffusers import DiffusionPipeline
+from diffusers import FluxInpaintPipeline
 import torch
 from PIL import Image, ImageFilter
 from huggingface_hub import login
+from diffusers import AutoencoderTiny, AutoencoderKL
 from huggingface_hub import hf_hub_download, HfFileSystem, ModelCard, snapshot_download
 import copy
 import random
@@ -16,56 +19,60 @@ import boto3
 from io import BytesIO
 from datetime import datetime
 from diffusers.utils import load_image, make_image_grid
-import json
-from controlnet_flux import FluxControlNetModel
-from transformer_flux import FluxTransformer2DModel
-from pipeline_flux_controlnet_inpaint import FluxControlNetInpaintingPipeline
 
+import json
+from preprocessor import Preprocessor
+from diffusers import FluxControlNetInpaintPipeline
+from diffusers.models import FluxControlNetModel
 
 HF_TOKEN = os.environ.get("HF_TOKEN")
 
 login(token=HF_TOKEN)
 
 MAX_SEED = np.iinfo(np.int32).max
-IMAGE_SIZE = 1024
+IMAGE_SIZE = 512
 
 # init
 device = "cuda" if torch.cuda.is_available() else "cpu"
 base_model = "black-forest-labs/FLUX.1-dev"
 
-controlnet = FluxControlNetModel.from_pretrained("alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Alpha", torch_dtype=torch.bfloat16)
-transformer = FluxTransformer2DModel.from_pretrained("black-forest-labs/FLUX.1-dev", subfolder='transformer', torch_dytpe=torch.bfloat16)
+controlnet_model = 'InstantX/FLUX.1-dev-Controlnet-Union'
+controlnet = FluxControlNetModel.from_pretrained(controlnet_model, torch_dtype=torch.bfloat16)
+
+
+pipe = FluxControlNetInpaintPipeline.from_pretrained(base_model, controlnet=controlnet, torch_dtype=torch.bfloat16).to(device)
 
-pipe = FluxControlNetInpaintingPipeline.from_pretrained(
-    base_model, 
-    controlnet=controlnet, 
-    transformer=transformer,
-    torch_dtype=torch.bfloat16).to(device)
+# pipe.enable_model_cpu_offload() # for saving memory
 
+control_mode_ids = {
+    "canny": 0,        # supported
+    "tile": 1,         # supported
+    "depth": 2,        # supported
+    "blur": 3,         # supported
+    "pose": 4,         # supported
+    "gray": 5,         # supported
+    "lq": 6,           # supported
+}
 
 def clear_cuda_cache():
     torch.cuda.empty_cache()
 
+
 class calculateDuration:
     def __init__(self, activity_name=""):
         self.activity_name = activity_name
 
     def __enter__(self):
         self.start_time = time.time()
-        self.start_time_formatted = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(self.start_time))
-        print(f"Activity: {self.activity_name}, Start time: {self.start_time_formatted}")
         return self
     
     def __exit__(self, exc_type, exc_value, traceback):
         self.end_time = time.time()
         self.elapsed_time = self.end_time - self.start_time
-        self.end_time_formatted = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(self.end_time))
-        
         if self.activity_name:
             print(f"Elapsed time for {self.activity_name}: {self.elapsed_time:.6f} seconds")
         else:
             print(f"Elapsed time: {self.elapsed_time:.6f} seconds")
-        
 
 
 def calculate_image_dimensions_for_flux(
@@ -140,6 +147,8 @@ def upload_image_to_r2(image, account_id, access_key, secret_key, bucket_name):
 def run_flux(
     image: Image.Image,
     mask: Image.Image,
+    control_image: Image.Image,
+    control_mode: int,
     prompt: str,
     seed_slicer: int,
     randomize_seed_checkbox: bool,
@@ -148,37 +157,41 @@ def run_flux(
     resolution_wh: Tuple[int, int],
     progress
 ) -> Image.Image:
+    print("Running FLUX...")
+    width, height = resolution_wh
+    if randomize_seed_checkbox:
+        seed_slicer = random.randint(0, MAX_SEED)
+    generator = torch.Generator().manual_seed(seed_slicer)
 
     with calculateDuration("run pipe"):
-        print("start to run pipe", prompt, control_mode)
-        # pipe.to(device)
-        width, height = resolution_wh
-        if randomize_seed_checkbox:
-            seed_slicer = random.randint(0, MAX_SEED)
-        generator = torch.Generator().manual_seed(seed_slicer)
+        print("start to run pipe", prompt)
+       
         with torch.inference_mode():
             generated_image = pipe(
                 prompt=prompt,
+                image=image,
                 mask_image=mask,
-                control_image=image,
-                controlnet_conditioning_scale=0.9,
+                control_image=control_image,
+                control_mode=control_mode,
+                controlnet_conditioning_scale=0.55,
                 width=width,
                 height=height,
-                strength=0.7,
-                guidance_scale=3.5,
+                strength=strength_slider,
                 generator=generator,
-                num_inference_steps=28,
+                num_inference_steps=num_inference_steps_slider,
             ).images[0]
     progress(99, "Generate image success!")
     return generated_image
 
 
 def load_loras(lora_strings_json:str):
+    lora_configs = None
     if lora_strings_json:
         try:
             lora_configs = json.loads(lora_strings_json)
         except:
-            lora_configs = None
+            print("parse lora failed")
+
     if lora_configs:
         with calculateDuration("Loading LoRA weights"):
             pipe.unload_lora_weights()
@@ -198,12 +211,43 @@ def load_loras(lora_strings_json:str):
             pipe.set_adapters(adapter_names, adapter_weights=adapter_weights)
 
 
+def generate_control_image(image, mask, control_mode, width, height):
+        # generated control_
+    with calculateDuration("Generate control image"):
+        preprocessor = Preprocessor()
+        if control_mode == "depth":
+            preprocessor.load("Midas")
+            control_image = preprocessor(
+                image=image,
+                image_resolution=width,
+                detect_resolution=512,
+            )
+        if control_mode == "pose":
+            preprocessor.load("Openpose")
+            control_image = preprocessor(
+                image=image,
+                hand_and_face=False,
+                image_resolution=width,
+                detect_resolution=512,
+            )
+        if control_mode == "canny":
+            preprocessor.load("Canny")
+            control_image = preprocessor(
+                image=image,
+                image_resolution=width,
+                detect_resolution=512,
+            )
+    
+    control_image = control_image.resize((width, height), Image.LANCZOS)    
+    return control_image
+
 def process(
     image_url: str,
     mask_url: str,
     inpainting_prompt_text: str,
     mask_inflation_slider: int,
     mask_blur_slider: int,
+    control_mode: str,
     seed_slicer: int,
     randomize_seed_checkbox: bool,
     strength_slider: float,
@@ -245,13 +289,18 @@ def process(
         mask = mask.resize((width, height), Image.LANCZOS)
         mask = process_mask(mask, mask_inflation=mask_inflation_slider, mask_blur=mask_blur_slider)
 
-    # load loras
+    control_image = generate_control_image(image, mask, control_mode, width, height)
+    control_mode_id = control_mode_ids[control_mode]
+    clear_cuda_cache()
+
     load_loras(lora_strings_json=lora_strings_json)
     
     try:
         generated_image = run_flux(
             image=image,
             mask=mask,
+            control_image=control_image, 
+            control_mode=control_mode_id,
             prompt=inpainting_prompt_text,
             seed_slicer=seed_slicer,
             randomize_seed_checkbox=randomize_seed_checkbox,
@@ -422,4 +471,4 @@ with gr.Blocks() as demo:
     )
 
 demo.queue(api_open=False)
-demo.launch()
+demo.launch()

controlnet_flux.py

@@ -1,418 +0,0 @@
-from dataclasses import dataclass
-from typing import Any, Dict, List, Optional, Tuple, Union
-
-import torch
-import torch.nn as nn
-
-from diffusers.configuration_utils import ConfigMixin, register_to_config
-from diffusers.loaders import PeftAdapterMixin
-from diffusers.models.modeling_utils import ModelMixin
-from diffusers.models.attention_processor import AttentionProcessor
-from diffusers.utils import (
-    USE_PEFT_BACKEND,
-    is_torch_version,
-    logging,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
-from diffusers.models.controlnet import BaseOutput, zero_module
-from diffusers.models.embeddings import (
-    CombinedTimestepGuidanceTextProjEmbeddings,
-    CombinedTimestepTextProjEmbeddings,
-)
-from diffusers.models.modeling_outputs import Transformer2DModelOutput
-from transformer_flux import (
-    EmbedND,
-    FluxSingleTransformerBlock,
-    FluxTransformerBlock,
-)
-
-
-logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
-
-
-@dataclass
-class FluxControlNetOutput(BaseOutput):
-    controlnet_block_samples: Tuple[torch.Tensor]
-    controlnet_single_block_samples: Tuple[torch.Tensor]
-
-
-class FluxControlNetModel(ModelMixin, ConfigMixin, PeftAdapterMixin):
-    _supports_gradient_checkpointing = True
-
-    @register_to_config
-    def __init__(
-        self,
-        patch_size: int = 1,
-        in_channels: int = 64,
-        num_layers: int = 19,
-        num_single_layers: int = 38,
-        attention_head_dim: int = 128,
-        num_attention_heads: int = 24,
-        joint_attention_dim: int = 4096,
-        pooled_projection_dim: int = 768,
-        guidance_embeds: bool = False,
-        axes_dims_rope: List[int] = [16, 56, 56],
-        extra_condition_channels: int = 1 * 4,
-    ):
-        super().__init__()
-        self.out_channels = in_channels
-        self.inner_dim = num_attention_heads * attention_head_dim
-
-        self.pos_embed = EmbedND(
-            dim=self.inner_dim, theta=10000, axes_dim=axes_dims_rope
-        )
-        text_time_guidance_cls = (
-            CombinedTimestepGuidanceTextProjEmbeddings
-            if guidance_embeds
-            else CombinedTimestepTextProjEmbeddings
-        )
-        self.time_text_embed = text_time_guidance_cls(
-            embedding_dim=self.inner_dim, pooled_projection_dim=pooled_projection_dim
-        )
-
-        self.context_embedder = nn.Linear(joint_attention_dim, self.inner_dim)
-        self.x_embedder = nn.Linear(in_channels, self.inner_dim)
-
-        self.transformer_blocks = nn.ModuleList(
-            [
-                FluxTransformerBlock(
-                    dim=self.inner_dim,
-                    num_attention_heads=num_attention_heads,
-                    attention_head_dim=attention_head_dim,
-                )
-                for _ in range(num_layers)
-            ]
-        )
-
-        self.single_transformer_blocks = nn.ModuleList(
-            [
-                FluxSingleTransformerBlock(
-                    dim=self.inner_dim,
-                    num_attention_heads=num_attention_heads,
-                    attention_head_dim=attention_head_dim,
-                )
-                for _ in range(num_single_layers)
-            ]
-        )
-
-        # controlnet_blocks
-        self.controlnet_blocks = nn.ModuleList([])
-        for _ in range(len(self.transformer_blocks)):
-            self.controlnet_blocks.append(
-                zero_module(nn.Linear(self.inner_dim, self.inner_dim))
-            )
-
-        self.controlnet_single_blocks = nn.ModuleList([])
-        for _ in range(len(self.single_transformer_blocks)):
-            self.controlnet_single_blocks.append(
-                zero_module(nn.Linear(self.inner_dim, self.inner_dim))
-            )
-
-        self.controlnet_x_embedder = zero_module(
-            torch.nn.Linear(in_channels + extra_condition_channels, self.inner_dim)
-        )
-
-        self.gradient_checkpointing = False
-
-    @property
-    # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.attn_processors
-    def attn_processors(self):
-        r"""
-        Returns:
-            `dict` of attention processors: A dictionary containing all attention processors used in the model with
-            indexed by its weight name.
-        """
-        # set recursively
-        processors = {}
-
-        def fn_recursive_add_processors(name: str, module: torch.nn.Module, processors: Dict[str, AttentionProcessor]):
-            if hasattr(module, "get_processor"):
-                processors[f"{name}.processor"] = module.get_processor()
-
-            for sub_name, child in module.named_children():
-                fn_recursive_add_processors(f"{name}.{sub_name}", child, processors)
-
-            return processors
-
-        for name, module in self.named_children():
-            fn_recursive_add_processors(name, module, processors)
-
-        return processors
-
-    # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.set_attn_processor
-    def set_attn_processor(self, processor):
-        r"""
-        Sets the attention processor to use to compute attention.
-
-        Parameters:
-            processor (`dict` of `AttentionProcessor` or only `AttentionProcessor`):
-                The instantiated processor class or a dictionary of processor classes that will be set as the processor
-                for **all** `Attention` layers.
-
-                If `processor` is a dict, the key needs to define the path to the corresponding cross attention
-                processor. This is strongly recommended when setting trainable attention processors.
-
-        """
-        count = len(self.attn_processors.keys())
-
-        if isinstance(processor, dict) and len(processor) != count:
-            raise ValueError(
-                f"A dict of processors was passed, but the number of processors {len(processor)} does not match the"
-                f" number of attention layers: {count}. Please make sure to pass {count} processor classes."
-            )
-
-        def fn_recursive_attn_processor(name: str, module: torch.nn.Module, processor):
-            if hasattr(module, "set_processor"):
-                if not isinstance(processor, dict):
-                    module.set_processor(processor)
-                else:
-                    module.set_processor(processor.pop(f"{name}.processor"))
-
-            for sub_name, child in module.named_children():
-                fn_recursive_attn_processor(f"{name}.{sub_name}", child, processor)
-
-        for name, module in self.named_children():
-            fn_recursive_attn_processor(name, module, processor)
-
-    def _set_gradient_checkpointing(self, module, value=False):
-        if hasattr(module, "gradient_checkpointing"):
-            module.gradient_checkpointing = value
-
-    @classmethod
-    def from_transformer(
-        cls,
-        transformer,
-        num_layers: int = 4,
-        num_single_layers: int = 10,
-        attention_head_dim: int = 128,
-        num_attention_heads: int = 24,
-        load_weights_from_transformer=True,
-    ):
-        config = transformer.config
-        config["num_layers"] = num_layers
-        config["num_single_layers"] = num_single_layers
-        config["attention_head_dim"] = attention_head_dim
-        config["num_attention_heads"] = num_attention_heads
-
-        controlnet = cls(**config)
-
-        if load_weights_from_transformer:
-            controlnet.pos_embed.load_state_dict(transformer.pos_embed.state_dict())
-            controlnet.time_text_embed.load_state_dict(
-                transformer.time_text_embed.state_dict()
-            )
-            controlnet.context_embedder.load_state_dict(
-                transformer.context_embedder.state_dict()
-            )
-            controlnet.x_embedder.load_state_dict(transformer.x_embedder.state_dict())
-            controlnet.transformer_blocks.load_state_dict(
-                transformer.transformer_blocks.state_dict(), strict=False
-            )
-            controlnet.single_transformer_blocks.load_state_dict(
-                transformer.single_transformer_blocks.state_dict(), strict=False
-            )
-
-            controlnet.controlnet_x_embedder = zero_module(
-                controlnet.controlnet_x_embedder
-            )
-
-        return controlnet
-
-    def forward(
-        self,
-        hidden_states: torch.Tensor,
-        controlnet_cond: torch.Tensor,
-        conditioning_scale: float = 1.0,
-        encoder_hidden_states: torch.Tensor = None,
-        pooled_projections: torch.Tensor = None,
-        timestep: torch.LongTensor = None,
-        img_ids: torch.Tensor = None,
-        txt_ids: torch.Tensor = None,
-        guidance: torch.Tensor = None,
-        joint_attention_kwargs: Optional[Dict[str, Any]] = None,
-        return_dict: bool = True,
-    ) -> Union[torch.FloatTensor, Transformer2DModelOutput]:
-        """
-        The [`FluxTransformer2DModel`] forward method.
-
-        Args:
-            hidden_states (`torch.FloatTensor` of shape `(batch size, channel, height, width)`):
-                Input `hidden_states`.
-            encoder_hidden_states (`torch.FloatTensor` of shape `(batch size, sequence_len, embed_dims)`):
-                Conditional embeddings (embeddings computed from the input conditions such as prompts) to use.
-            pooled_projections (`torch.FloatTensor` of shape `(batch_size, projection_dim)`): Embeddings projected
-                from the embeddings of input conditions.
-            timestep ( `torch.LongTensor`):
-                Used to indicate denoising step.
-            block_controlnet_hidden_states: (`list` of `torch.Tensor`):
-                A list of tensors that if specified are added to the residuals of transformer blocks.
-            joint_attention_kwargs (`dict`, *optional*):
-                A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
-                `self.processor` in
-                [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
-            return_dict (`bool`, *optional*, defaults to `True`):
-                Whether or not to return a [`~models.transformer_2d.Transformer2DModelOutput`] instead of a plain
-                tuple.
-
-        Returns:
-            If `return_dict` is True, an [`~models.transformer_2d.Transformer2DModelOutput`] is returned, otherwise a
-            `tuple` where the first element is the sample tensor.
-        """
-        if joint_attention_kwargs is not None:
-            joint_attention_kwargs = joint_attention_kwargs.copy()
-            lora_scale = joint_attention_kwargs.pop("scale", 1.0)
-        else:
-            lora_scale = 1.0
-
-        if USE_PEFT_BACKEND:
-            # weight the lora layers by setting `lora_scale` for each PEFT layer
-            scale_lora_layers(self, lora_scale)
-        else:
-            if (
-                joint_attention_kwargs is not None
-                and joint_attention_kwargs.get("scale", None) is not None
-            ):
-                logger.warning(
-                    "Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective."
-                )
-        hidden_states = self.x_embedder(hidden_states)
-
-        # add condition
-        hidden_states = hidden_states + self.controlnet_x_embedder(controlnet_cond)
-
-        timestep = timestep.to(hidden_states.dtype) * 1000
-        if guidance is not None:
-            guidance = guidance.to(hidden_states.dtype) * 1000
-        else:
-            guidance = None
-        temb = (
-            self.time_text_embed(timestep, pooled_projections)
-            if guidance is None
-            else self.time_text_embed(timestep, guidance, pooled_projections)
-        )
-        encoder_hidden_states = self.context_embedder(encoder_hidden_states)
-
-        txt_ids = txt_ids.expand(img_ids.size(0), -1, -1)
-        ids = torch.cat((txt_ids, img_ids), dim=1)
-        image_rotary_emb = self.pos_embed(ids)
-
-        block_samples = ()
-        for _, block in enumerate(self.transformer_blocks):
-            if self.training and self.gradient_checkpointing:
-
-                def create_custom_forward(module, return_dict=None):
-                    def custom_forward(*inputs):
-                        if return_dict is not None:
-                            return module(*inputs, return_dict=return_dict)
-                        else:
-                            return module(*inputs)
-
-                    return custom_forward
-
-                ckpt_kwargs: Dict[str, Any] = (
-                    {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
-                )
-                (
-                    encoder_hidden_states,
-                    hidden_states,
-                ) = torch.utils.checkpoint.checkpoint(
-                    create_custom_forward(block),
-                    hidden_states,
-                    encoder_hidden_states,
-                    temb,
-                    image_rotary_emb,
-                    **ckpt_kwargs,
-                )
-
-            else:
-                encoder_hidden_states, hidden_states = block(
-                    hidden_states=hidden_states,
-                    encoder_hidden_states=encoder_hidden_states,
-                    temb=temb,
-                    image_rotary_emb=image_rotary_emb,
-                )
-            block_samples = block_samples + (hidden_states,)
-
-        hidden_states = torch.cat([encoder_hidden_states, hidden_states], dim=1)
-
-        single_block_samples = ()
-        for _, block in enumerate(self.single_transformer_blocks):
-            if self.training and self.gradient_checkpointing:
-
-                def create_custom_forward(module, return_dict=None):
-                    def custom_forward(*inputs):
-                        if return_dict is not None:
-                            return module(*inputs, return_dict=return_dict)
-                        else:
-                            return module(*inputs)
-
-                    return custom_forward
-
-                ckpt_kwargs: Dict[str, Any] = (
-                    {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
-                )
-                hidden_states = torch.utils.checkpoint.checkpoint(
-                    create_custom_forward(block),
-                    hidden_states,
-                    temb,
-                    image_rotary_emb,
-                    **ckpt_kwargs,
-                )
-
-            else:
-                hidden_states = block(
-                    hidden_states=hidden_states,
-                    temb=temb,
-                    image_rotary_emb=image_rotary_emb,
-                )
-            single_block_samples = single_block_samples + (
-                hidden_states[:, encoder_hidden_states.shape[1] :],
-            )
-
-        # controlnet block
-        controlnet_block_samples = ()
-        for block_sample, controlnet_block in zip(
-            block_samples, self.controlnet_blocks
-        ):
-            block_sample = controlnet_block(block_sample)
-            controlnet_block_samples = controlnet_block_samples + (block_sample,)
-
-        controlnet_single_block_samples = ()
-        for single_block_sample, controlnet_block in zip(
-            single_block_samples, self.controlnet_single_blocks
-        ):
-            single_block_sample = controlnet_block(single_block_sample)
-            controlnet_single_block_samples = controlnet_single_block_samples + (
-                single_block_sample,
-            )
-
-        # scaling
-        controlnet_block_samples = [
-            sample * conditioning_scale for sample in controlnet_block_samples
-        ]
-        controlnet_single_block_samples = [
-            sample * conditioning_scale for sample in controlnet_single_block_samples
-        ]
-
-        #
-        controlnet_block_samples = (
-            None if len(controlnet_block_samples) == 0 else controlnet_block_samples
-        )
-        controlnet_single_block_samples = (
-            None
-            if len(controlnet_single_block_samples) == 0
-            else controlnet_single_block_samples
-        )
-
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
-        if not return_dict:
-            return (controlnet_block_samples, controlnet_single_block_samples)
-
-        return FluxControlNetOutput(
-            controlnet_block_samples=controlnet_block_samples,
-            controlnet_single_block_samples=controlnet_single_block_samples,
-        )

depth_anything_v2/dinov2.py

@@ -0,0 +1,415 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+# References:
+#   https://github.com/facebookresearch/dino/blob/main/vision_transformer.py
+#   https://github.com/rwightman/pytorch-image-models/tree/master/timm/models/vision_transformer.py
+
+from functools import partial
+import math
+import logging
+from typing import Sequence, Tuple, Union, Callable
+
+import torch
+import torch.nn as nn
+import torch.utils.checkpoint
+from torch.nn.init import trunc_normal_
+
+from .dinov2_layers import Mlp, PatchEmbed, SwiGLUFFNFused, MemEffAttention, NestedTensorBlock as Block
+
+
+logger = logging.getLogger("dinov2")
+
+
+def named_apply(fn: Callable, module: nn.Module, name="", depth_first=True, include_root=False) -> nn.Module:
+    if not depth_first and include_root:
+        fn(module=module, name=name)
+    for child_name, child_module in module.named_children():
+        child_name = ".".join((name, child_name)) if name else child_name
+        named_apply(fn=fn, module=child_module, name=child_name, depth_first=depth_first, include_root=True)
+    if depth_first and include_root:
+        fn(module=module, name=name)
+    return module
+
+
+class BlockChunk(nn.ModuleList):
+    def forward(self, x):
+        for b in self:
+            x = b(x)
+        return x
+
+
+class DinoVisionTransformer(nn.Module):
+    def __init__(
+        self,
+        img_size=224,
+        patch_size=16,
+        in_chans=3,
+        embed_dim=768,
+        depth=12,
+        num_heads=12,
+        mlp_ratio=4.0,
+        qkv_bias=True,
+        ffn_bias=True,
+        proj_bias=True,
+        drop_path_rate=0.0,
+        drop_path_uniform=False,
+        init_values=None,  # for layerscale: None or 0 => no layerscale
+        embed_layer=PatchEmbed,
+        act_layer=nn.GELU,
+        block_fn=Block,
+        ffn_layer="mlp",
+        block_chunks=1,
+        num_register_tokens=0,
+        interpolate_antialias=False,
+        interpolate_offset=0.1,
+    ):
+        """
+        Args:
+            img_size (int, tuple): input image size
+            patch_size (int, tuple): patch size
+            in_chans (int): number of input channels
+            embed_dim (int): embedding dimension
+            depth (int): depth of transformer
+            num_heads (int): number of attention heads
+            mlp_ratio (int): ratio of mlp hidden dim to embedding dim
+            qkv_bias (bool): enable bias for qkv if True
+            proj_bias (bool): enable bias for proj in attn if True
+            ffn_bias (bool): enable bias for ffn if True
+            drop_path_rate (float): stochastic depth rate
+            drop_path_uniform (bool): apply uniform drop rate across blocks
+            weight_init (str): weight init scheme
+            init_values (float): layer-scale init values
+            embed_layer (nn.Module): patch embedding layer
+            act_layer (nn.Module): MLP activation layer
+            block_fn (nn.Module): transformer block class
+            ffn_layer (str): "mlp", "swiglu", "swiglufused" or "identity"
+            block_chunks: (int) split block sequence into block_chunks units for FSDP wrap
+            num_register_tokens: (int) number of extra cls tokens (so-called "registers")
+            interpolate_antialias: (str) flag to apply anti-aliasing when interpolating positional embeddings
+            interpolate_offset: (float) work-around offset to apply when interpolating positional embeddings
+        """
+        super().__init__()
+        norm_layer = partial(nn.LayerNorm, eps=1e-6)
+
+        self.num_features = self.embed_dim = embed_dim  # num_features for consistency with other models
+        self.num_tokens = 1
+        self.n_blocks = depth
+        self.num_heads = num_heads
+        self.patch_size = patch_size
+        self.num_register_tokens = num_register_tokens
+        self.interpolate_antialias = interpolate_antialias
+        self.interpolate_offset = interpolate_offset
+
+        self.patch_embed = embed_layer(img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim)
+        num_patches = self.patch_embed.num_patches
+
+        self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))
+        self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + self.num_tokens, embed_dim))
+        assert num_register_tokens >= 0
+        self.register_tokens = (
+            nn.Parameter(torch.zeros(1, num_register_tokens, embed_dim)) if num_register_tokens else None
+        )
+
+        if drop_path_uniform is True:
+            dpr = [drop_path_rate] * depth
+        else:
+            dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]  # stochastic depth decay rule
+
+        if ffn_layer == "mlp":
+            logger.info("using MLP layer as FFN")
+            ffn_layer = Mlp
+        elif ffn_layer == "swiglufused" or ffn_layer == "swiglu":
+            logger.info("using SwiGLU layer as FFN")
+            ffn_layer = SwiGLUFFNFused
+        elif ffn_layer == "identity":
+            logger.info("using Identity layer as FFN")
+
+            def f(*args, **kwargs):
+                return nn.Identity()
+
+            ffn_layer = f
+        else:
+            raise NotImplementedError
+
+        blocks_list = [
+            block_fn(
+                dim=embed_dim,
+                num_heads=num_heads,
+                mlp_ratio=mlp_ratio,
+                qkv_bias=qkv_bias,
+                proj_bias=proj_bias,
+                ffn_bias=ffn_bias,
+                drop_path=dpr[i],
+                norm_layer=norm_layer,
+                act_layer=act_layer,
+                ffn_layer=ffn_layer,
+                init_values=init_values,
+            )
+            for i in range(depth)
+        ]
+        if block_chunks > 0:
+            self.chunked_blocks = True
+            chunked_blocks = []
+            chunksize = depth // block_chunks
+            for i in range(0, depth, chunksize):
+                # this is to keep the block index consistent if we chunk the block list
+                chunked_blocks.append([nn.Identity()] * i + blocks_list[i : i + chunksize])
+            self.blocks = nn.ModuleList([BlockChunk(p) for p in chunked_blocks])
+        else:
+            self.chunked_blocks = False
+            self.blocks = nn.ModuleList(blocks_list)
+
+        self.norm = norm_layer(embed_dim)
+        self.head = nn.Identity()
+
+        self.mask_token = nn.Parameter(torch.zeros(1, embed_dim))
+
+        self.init_weights()
+
+    def init_weights(self):
+        trunc_normal_(self.pos_embed, std=0.02)
+        nn.init.normal_(self.cls_token, std=1e-6)
+        if self.register_tokens is not None:
+            nn.init.normal_(self.register_tokens, std=1e-6)
+        named_apply(init_weights_vit_timm, self)
+
+    def interpolate_pos_encoding(self, x, w, h):
+        previous_dtype = x.dtype
+        npatch = x.shape[1] - 1
+        N = self.pos_embed.shape[1] - 1
+        if npatch == N and w == h:
+            return self.pos_embed
+        pos_embed = self.pos_embed.float()
+        class_pos_embed = pos_embed[:, 0]
+        patch_pos_embed = pos_embed[:, 1:]
+        dim = x.shape[-1]
+        w0 = w // self.patch_size
+        h0 = h // self.patch_size
+        # we add a small number to avoid floating point error in the interpolation
+        # see discussion at https://github.com/facebookresearch/dino/issues/8
+        # DINOv2 with register modify the interpolate_offset from 0.1 to 0.0
+        w0, h0 = w0 + self.interpolate_offset, h0 + self.interpolate_offset
+        # w0, h0 = w0 + 0.1, h0 + 0.1
+        
+        sqrt_N = math.sqrt(N)
+        sx, sy = float(w0) / sqrt_N, float(h0) / sqrt_N
+        patch_pos_embed = nn.functional.interpolate(
+            patch_pos_embed.reshape(1, int(sqrt_N), int(sqrt_N), dim).permute(0, 3, 1, 2),
+            scale_factor=(sx, sy),
+            # (int(w0), int(h0)), # to solve the upsampling shape issue
+            mode="bicubic",
+            antialias=self.interpolate_antialias
+        )
+        
+        assert int(w0) == patch_pos_embed.shape[-2]
+        assert int(h0) == patch_pos_embed.shape[-1]
+        patch_pos_embed = patch_pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)
+        return torch.cat((class_pos_embed.unsqueeze(0), patch_pos_embed), dim=1).to(previous_dtype)
+
+    def prepare_tokens_with_masks(self, x, masks=None):
+        B, nc, w, h = x.shape
+        x = self.patch_embed(x)
+        if masks is not None:
+            x = torch.where(masks.unsqueeze(-1), self.mask_token.to(x.dtype).unsqueeze(0), x)
+
+        x = torch.cat((self.cls_token.expand(x.shape[0], -1, -1), x), dim=1)
+        x = x + self.interpolate_pos_encoding(x, w, h)
+
+        if self.register_tokens is not None:
+            x = torch.cat(
+                (
+                    x[:, :1],
+                    self.register_tokens.expand(x.shape[0], -1, -1),
+                    x[:, 1:],
+                ),
+                dim=1,
+            )
+
+        return x
+
+    def forward_features_list(self, x_list, masks_list):
+        x = [self.prepare_tokens_with_masks(x, masks) for x, masks in zip(x_list, masks_list)]
+        for blk in self.blocks:
+            x = blk(x)
+
+        all_x = x
+        output = []
+        for x, masks in zip(all_x, masks_list):
+            x_norm = self.norm(x)
+            output.append(
+                {
+                    "x_norm_clstoken": x_norm[:, 0],
+                    "x_norm_regtokens": x_norm[:, 1 : self.num_register_tokens + 1],
+                    "x_norm_patchtokens": x_norm[:, self.num_register_tokens + 1 :],
+                    "x_prenorm": x,
+                    "masks": masks,
+                }
+            )
+        return output
+
+    def forward_features(self, x, masks=None):
+        if isinstance(x, list):
+            return self.forward_features_list(x, masks)
+
+        x = self.prepare_tokens_with_masks(x, masks)
+
+        for blk in self.blocks:
+            x = blk(x)
+
+        x_norm = self.norm(x)
+        return {
+            "x_norm_clstoken": x_norm[:, 0],
+            "x_norm_regtokens": x_norm[:, 1 : self.num_register_tokens + 1],
+            "x_norm_patchtokens": x_norm[:, self.num_register_tokens + 1 :],
+            "x_prenorm": x,
+            "masks": masks,
+        }
+
+    def _get_intermediate_layers_not_chunked(self, x, n=1):
+        x = self.prepare_tokens_with_masks(x)
+        # If n is an int, take the n last blocks. If it's a list, take them
+        output, total_block_len = [], len(self.blocks)
+        blocks_to_take = range(total_block_len - n, total_block_len) if isinstance(n, int) else n
+        for i, blk in enumerate(self.blocks):
+            x = blk(x)
+            if i in blocks_to_take:
+                output.append(x)
+        assert len(output) == len(blocks_to_take), f"only {len(output)} / {len(blocks_to_take)} blocks found"
+        return output
+
+    def _get_intermediate_layers_chunked(self, x, n=1):
+        x = self.prepare_tokens_with_masks(x)
+        output, i, total_block_len = [], 0, len(self.blocks[-1])
+        # If n is an int, take the n last blocks. If it's a list, take them
+        blocks_to_take = range(total_block_len - n, total_block_len) if isinstance(n, int) else n
+        for block_chunk in self.blocks:
+            for blk in block_chunk[i:]:  # Passing the nn.Identity()
+                x = blk(x)
+                if i in blocks_to_take:
+                    output.append(x)
+                i += 1
+        assert len(output) == len(blocks_to_take), f"only {len(output)} / {len(blocks_to_take)} blocks found"
+        return output
+
+    def get_intermediate_layers(
+        self,
+        x: torch.Tensor,
+        n: Union[int, Sequence] = 1,  # Layers or n last layers to take
+        reshape: bool = False,
+        return_class_token: bool = False,
+        norm=True
+    ) -> Tuple[Union[torch.Tensor, Tuple[torch.Tensor]]]:
+        if self.chunked_blocks:
+            outputs = self._get_intermediate_layers_chunked(x, n)
+        else:
+            outputs = self._get_intermediate_layers_not_chunked(x, n)
+        if norm:
+            outputs = [self.norm(out) for out in outputs]
+        class_tokens = [out[:, 0] for out in outputs]
+        outputs = [out[:, 1 + self.num_register_tokens:] for out in outputs]
+        if reshape:
+            B, _, w, h = x.shape
+            outputs = [
+                out.reshape(B, w // self.patch_size, h // self.patch_size, -1).permute(0, 3, 1, 2).contiguous()
+                for out in outputs
+            ]
+        if return_class_token:
+            return tuple(zip(outputs, class_tokens))
+        return tuple(outputs)
+
+    def forward(self, *args, is_training=False, **kwargs):
+        ret = self.forward_features(*args, **kwargs)
+        if is_training:
+            return ret
+        else:
+            return self.head(ret["x_norm_clstoken"])
+
+
+def init_weights_vit_timm(module: nn.Module, name: str = ""):
+    """ViT weight initialization, original timm impl (for reproducibility)"""
+    if isinstance(module, nn.Linear):
+        trunc_normal_(module.weight, std=0.02)
+        if module.bias is not None:
+            nn.init.zeros_(module.bias)
+
+
+def vit_small(patch_size=16, num_register_tokens=0, **kwargs):
+    model = DinoVisionTransformer(
+        patch_size=patch_size,
+        embed_dim=384,
+        depth=12,
+        num_heads=6,
+        mlp_ratio=4,
+        block_fn=partial(Block, attn_class=MemEffAttention),
+        num_register_tokens=num_register_tokens,
+        **kwargs,
+    )
+    return model
+
+
+def vit_base(patch_size=16, num_register_tokens=0, **kwargs):
+    model = DinoVisionTransformer(
+        patch_size=patch_size,
+        embed_dim=768,
+        depth=12,
+        num_heads=12,
+        mlp_ratio=4,
+        block_fn=partial(Block, attn_class=MemEffAttention),
+        num_register_tokens=num_register_tokens,
+        **kwargs,
+    )
+    return model
+
+
+def vit_large(patch_size=16, num_register_tokens=0, **kwargs):
+    model = DinoVisionTransformer(
+        patch_size=patch_size,
+        embed_dim=1024,
+        depth=24,
+        num_heads=16,
+        mlp_ratio=4,
+        block_fn=partial(Block, attn_class=MemEffAttention),
+        num_register_tokens=num_register_tokens,
+        **kwargs,
+    )
+    return model
+
+
+def vit_giant2(patch_size=16, num_register_tokens=0, **kwargs):
+    """
+    Close to ViT-giant, with embed-dim 1536 and 24 heads => embed-dim per head 64
+    """
+    model = DinoVisionTransformer(
+        patch_size=patch_size,
+        embed_dim=1536,
+        depth=40,
+        num_heads=24,
+        mlp_ratio=4,
+        block_fn=partial(Block, attn_class=MemEffAttention),
+        num_register_tokens=num_register_tokens,
+        **kwargs,
+    )
+    return model
+
+
+def DINOv2(model_name):
+    model_zoo = {
+        "vits": vit_small, 
+        "vitb": vit_base, 
+        "vitl": vit_large, 
+        "vitg": vit_giant2
+    }
+    
+    return model_zoo[model_name](
+        img_size=518,
+        patch_size=14,
+        init_values=1.0,
+        ffn_layer="mlp" if model_name != "vitg" else "swiglufused",
+        block_chunks=0,
+        num_register_tokens=0,
+        interpolate_antialias=False,
+        interpolate_offset=0.1
+    )

depth_anything_v2/dinov2_layers/__init__.py

@@ -0,0 +1,11 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+from .mlp import Mlp
+from .patch_embed import PatchEmbed
+from .swiglu_ffn import SwiGLUFFN, SwiGLUFFNFused
+from .block import NestedTensorBlock
+from .attention import MemEffAttention

depth_anything_v2/dinov2_layers/attention.py

@@ -0,0 +1,81 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+# References:
+#   https://github.com/facebookresearch/dino/blob/master/vision_transformer.py
+#   https://github.com/rwightman/pytorch-image-models/tree/master/timm/models/vision_transformer.py
+
+import logging
+
+from torch import Tensor
+from torch import nn
+
+
+logger = logging.getLogger("dinov2")
+
+
+try:
+    from xformers.ops import memory_efficient_attention, unbind, fmha
+
+    XFORMERS_AVAILABLE = True
+except ImportError:
+    logger.warning("xFormers not available")
+    XFORMERS_AVAILABLE = False
+
+
+class Attention(nn.Module):
+    def __init__(
+        self,
+        dim: int,
+        num_heads: int = 8,
+        qkv_bias: bool = False,
+        proj_bias: bool = True,
+        attn_drop: float = 0.0,
+        proj_drop: float = 0.0,
+    ) -> None:
+        super().__init__()
+        self.num_heads = num_heads
+        head_dim = dim // num_heads
+        self.scale = head_dim**-0.5
+
+        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
+        self.attn_drop = nn.Dropout(attn_drop)
+        self.proj = nn.Linear(dim, dim, bias=proj_bias)
+        self.proj_drop = nn.Dropout(proj_drop)
+
+    def forward(self, x: Tensor) -> Tensor:
+        B, N, C = x.shape
+        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
+
+        q, k, v = qkv[0] * self.scale, qkv[1], qkv[2]
+        attn = q @ k.transpose(-2, -1)
+
+        attn = attn.softmax(dim=-1)
+        attn = self.attn_drop(attn)
+
+        x = (attn @ v).transpose(1, 2).reshape(B, N, C)
+        x = self.proj(x)
+        x = self.proj_drop(x)
+        return x
+
+
+class MemEffAttention(Attention):
+    def forward(self, x: Tensor, attn_bias=None) -> Tensor:
+        if not XFORMERS_AVAILABLE:
+            assert attn_bias is None, "xFormers is required for nested tensors usage"
+            return super().forward(x)
+
+        B, N, C = x.shape
+        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads)
+
+        q, k, v = unbind(qkv, 2)
+
+        x = memory_efficient_attention(q, k, v, attn_bias=attn_bias)
+        x = x.reshape([B, N, C])
+
+        x = self.proj(x)
+        x = self.proj_drop(x)
+        return x

depth_anything_v2/dinov2_layers/block.py

@@ -0,0 +1,252 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+# References:
+#   https://github.com/facebookresearch/dino/blob/master/vision_transformer.py
+#   https://github.com/rwightman/pytorch-image-models/tree/master/timm/layers/patch_embed.py
+
+import logging
+from typing import Callable, List, Any, Tuple, Dict
+
+import torch
+from torch import nn, Tensor
+
+from .attention import Attention, MemEffAttention
+from .drop_path import DropPath
+from .layer_scale import LayerScale
+from .mlp import Mlp
+
+
+logger = logging.getLogger("dinov2")
+
+
+try:
+    from xformers.ops import fmha
+    from xformers.ops import scaled_index_add, index_select_cat
+
+    XFORMERS_AVAILABLE = True
+except ImportError:
+    logger.warning("xFormers not available")
+    XFORMERS_AVAILABLE = False
+
+
+class Block(nn.Module):
+    def __init__(
+        self,
+        dim: int,
+        num_heads: int,
+        mlp_ratio: float = 4.0,
+        qkv_bias: bool = False,
+        proj_bias: bool = True,
+        ffn_bias: bool = True,
+        drop: float = 0.0,
+        attn_drop: float = 0.0,
+        init_values=None,
+        drop_path: float = 0.0,
+        act_layer: Callable[..., nn.Module] = nn.GELU,
+        norm_layer: Callable[..., nn.Module] = nn.LayerNorm,
+        attn_class: Callable[..., nn.Module] = Attention,
+        ffn_layer: Callable[..., nn.Module] = Mlp,
+    ) -> None:
+        super().__init__()
+        # print(f"biases: qkv: {qkv_bias}, proj: {proj_bias}, ffn: {ffn_bias}")
+        self.norm1 = norm_layer(dim)
+        self.attn = attn_class(
+            dim,
+            num_heads=num_heads,
+            qkv_bias=qkv_bias,
+            proj_bias=proj_bias,
+            attn_drop=attn_drop,
+            proj_drop=drop,
+        )
+        self.ls1 = LayerScale(dim, init_values=init_values) if init_values else nn.Identity()
+        self.drop_path1 = DropPath(drop_path) if drop_path > 0.0 else nn.Identity()
+
+        self.norm2 = norm_layer(dim)
+        mlp_hidden_dim = int(dim * mlp_ratio)
+        self.mlp = ffn_layer(
+            in_features=dim,
+            hidden_features=mlp_hidden_dim,
+            act_layer=act_layer,
+            drop=drop,
+            bias=ffn_bias,
+        )
+        self.ls2 = LayerScale(dim, init_values=init_values) if init_values else nn.Identity()
+        self.drop_path2 = DropPath(drop_path) if drop_path > 0.0 else nn.Identity()
+
+        self.sample_drop_ratio = drop_path
+
+    def forward(self, x: Tensor) -> Tensor:
+        def attn_residual_func(x: Tensor) -> Tensor:
+            return self.ls1(self.attn(self.norm1(x)))
+
+        def ffn_residual_func(x: Tensor) -> Tensor:
+            return self.ls2(self.mlp(self.norm2(x)))
+
+        if self.training and self.sample_drop_ratio > 0.1:
+            # the overhead is compensated only for a drop path rate larger than 0.1
+            x = drop_add_residual_stochastic_depth(
+                x,
+                residual_func=attn_residual_func,
+                sample_drop_ratio=self.sample_drop_ratio,
+            )
+            x = drop_add_residual_stochastic_depth(
+                x,
+                residual_func=ffn_residual_func,
+                sample_drop_ratio=self.sample_drop_ratio,
+            )
+        elif self.training and self.sample_drop_ratio > 0.0:
+            x = x + self.drop_path1(attn_residual_func(x))
+            x = x + self.drop_path1(ffn_residual_func(x))  # FIXME: drop_path2
+        else:
+            x = x + attn_residual_func(x)
+            x = x + ffn_residual_func(x)
+        return x
+
+
+def drop_add_residual_stochastic_depth(
+    x: Tensor,
+    residual_func: Callable[[Tensor], Tensor],
+    sample_drop_ratio: float = 0.0,
+) -> Tensor:
+    # 1) extract subset using permutation
+    b, n, d = x.shape
+    sample_subset_size = max(int(b * (1 - sample_drop_ratio)), 1)
+    brange = (torch.randperm(b, device=x.device))[:sample_subset_size]
+    x_subset = x[brange]
+
+    # 2) apply residual_func to get residual
+    residual = residual_func(x_subset)
+
+    x_flat = x.flatten(1)
+    residual = residual.flatten(1)
+
+    residual_scale_factor = b / sample_subset_size
+
+    # 3) add the residual
+    x_plus_residual = torch.index_add(x_flat, 0, brange, residual.to(dtype=x.dtype), alpha=residual_scale_factor)
+    return x_plus_residual.view_as(x)
+
+
+def get_branges_scales(x, sample_drop_ratio=0.0):
+    b, n, d = x.shape
+    sample_subset_size = max(int(b * (1 - sample_drop_ratio)), 1)
+    brange = (torch.randperm(b, device=x.device))[:sample_subset_size]
+    residual_scale_factor = b / sample_subset_size
+    return brange, residual_scale_factor
+
+
+def add_residual(x, brange, residual, residual_scale_factor, scaling_vector=None):
+    if scaling_vector is None:
+        x_flat = x.flatten(1)
+        residual = residual.flatten(1)
+        x_plus_residual = torch.index_add(x_flat, 0, brange, residual.to(dtype=x.dtype), alpha=residual_scale_factor)
+    else:
+        x_plus_residual = scaled_index_add(
+            x, brange, residual.to(dtype=x.dtype), scaling=scaling_vector, alpha=residual_scale_factor
+        )
+    return x_plus_residual
+
+
+attn_bias_cache: Dict[Tuple, Any] = {}
+
+
+def get_attn_bias_and_cat(x_list, branges=None):
+    """
+    this will perform the index select, cat the tensors, and provide the attn_bias from cache
+    """
+    batch_sizes = [b.shape[0] for b in branges] if branges is not None else [x.shape[0] for x in x_list]
+    all_shapes = tuple((b, x.shape[1]) for b, x in zip(batch_sizes, x_list))
+    if all_shapes not in attn_bias_cache.keys():
+        seqlens = []
+        for b, x in zip(batch_sizes, x_list):
+            for _ in range(b):
+                seqlens.append(x.shape[1])
+        attn_bias = fmha.BlockDiagonalMask.from_seqlens(seqlens)
+        attn_bias._batch_sizes = batch_sizes
+        attn_bias_cache[all_shapes] = attn_bias
+
+    if branges is not None:
+        cat_tensors = index_select_cat([x.flatten(1) for x in x_list], branges).view(1, -1, x_list[0].shape[-1])
+    else:
+        tensors_bs1 = tuple(x.reshape([1, -1, *x.shape[2:]]) for x in x_list)
+        cat_tensors = torch.cat(tensors_bs1, dim=1)
+
+    return attn_bias_cache[all_shapes], cat_tensors
+
+
+def drop_add_residual_stochastic_depth_list(
+    x_list: List[Tensor],
+    residual_func: Callable[[Tensor, Any], Tensor],
+    sample_drop_ratio: float = 0.0,
+    scaling_vector=None,
+) -> Tensor:
+    # 1) generate random set of indices for dropping samples in the batch
+    branges_scales = [get_branges_scales(x, sample_drop_ratio=sample_drop_ratio) for x in x_list]
+    branges = [s[0] for s in branges_scales]
+    residual_scale_factors = [s[1] for s in branges_scales]
+
+    # 2) get attention bias and index+concat the tensors
+    attn_bias, x_cat = get_attn_bias_and_cat(x_list, branges)
+
+    # 3) apply residual_func to get residual, and split the result
+    residual_list = attn_bias.split(residual_func(x_cat, attn_bias=attn_bias))  # type: ignore
+
+    outputs = []
+    for x, brange, residual, residual_scale_factor in zip(x_list, branges, residual_list, residual_scale_factors):
+        outputs.append(add_residual(x, brange, residual, residual_scale_factor, scaling_vector).view_as(x))
+    return outputs
+
+
+class NestedTensorBlock(Block):
+    def forward_nested(self, x_list: List[Tensor]) -> List[Tensor]:
+        """
+        x_list contains a list of tensors to nest together and run
+        """
+        assert isinstance(self.attn, MemEffAttention)
+
+        if self.training and self.sample_drop_ratio > 0.0:
+
+            def attn_residual_func(x: Tensor, attn_bias=None) -> Tensor:
+                return self.attn(self.norm1(x), attn_bias=attn_bias)
+
+            def ffn_residual_func(x: Tensor, attn_bias=None) -> Tensor:
+                return self.mlp(self.norm2(x))
+
+            x_list = drop_add_residual_stochastic_depth_list(
+                x_list,
+                residual_func=attn_residual_func,
+                sample_drop_ratio=self.sample_drop_ratio,
+                scaling_vector=self.ls1.gamma if isinstance(self.ls1, LayerScale) else None,
+            )
+            x_list = drop_add_residual_stochastic_depth_list(
+                x_list,
+                residual_func=ffn_residual_func,
+                sample_drop_ratio=self.sample_drop_ratio,
+                scaling_vector=self.ls2.gamma if isinstance(self.ls1, LayerScale) else None,
+            )
+            return x_list
+        else:
+
+            def attn_residual_func(x: Tensor, attn_bias=None) -> Tensor:
+                return self.ls1(self.attn(self.norm1(x), attn_bias=attn_bias))
+
+            def ffn_residual_func(x: Tensor, attn_bias=None) -> Tensor:
+                return self.ls2(self.mlp(self.norm2(x)))
+
+            attn_bias, x = get_attn_bias_and_cat(x_list)
+            x = x + attn_residual_func(x, attn_bias=attn_bias)
+            x = x + ffn_residual_func(x)
+            return attn_bias.split(x)
+
+    def forward(self, x_or_x_list):
+        if isinstance(x_or_x_list, Tensor):
+            return super().forward(x_or_x_list)
+        elif isinstance(x_or_x_list, list):
+            assert XFORMERS_AVAILABLE, "Please install xFormers for nested tensors usage"
+            return self.forward_nested(x_or_x_list)
+        else:
+            raise AssertionError

depth_anything_v2/dinov2_layers/drop_path.py

@@ -0,0 +1,35 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+# References:
+#   https://github.com/facebookresearch/dino/blob/master/vision_transformer.py
+#   https://github.com/rwightman/pytorch-image-models/tree/master/timm/layers/drop.py
+
+
+from torch import nn
+
+
+def drop_path(x, drop_prob: float = 0.0, training: bool = False):
+    if drop_prob == 0.0 or not training:
+        return x
+    keep_prob = 1 - drop_prob
+    shape = (x.shape[0],) + (1,) * (x.ndim - 1)  # work with diff dim tensors, not just 2D ConvNets
+    random_tensor = x.new_empty(shape).bernoulli_(keep_prob)
+    if keep_prob > 0.0:
+        random_tensor.div_(keep_prob)
+    output = x * random_tensor
+    return output
+
+
+class DropPath(nn.Module):
+    """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks)."""
+
+    def __init__(self, drop_prob=None):
+        super(DropPath, self).__init__()
+        self.drop_prob = drop_prob
+
+    def forward(self, x):
+        return drop_path(x, self.drop_prob, self.training)

depth_anything_v2/dinov2_layers/layer_scale.py

@@ -0,0 +1,28 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+# Modified from: https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/vision_transformer.py#L103-L110
+
+from typing import Union
+
+import torch
+from torch import Tensor
+from torch import nn
+
+
+class LayerScale(nn.Module):
+    def __init__(
+        self,
+        dim: int,
+        init_values: Union[float, Tensor] = 1e-5,
+        inplace: bool = False,
+    ) -> None:
+        super().__init__()
+        self.inplace = inplace
+        self.gamma = nn.Parameter(init_values * torch.ones(dim))
+
+    def forward(self, x: Tensor) -> Tensor:
+        return x.mul_(self.gamma) if self.inplace else x * self.gamma

depth_anything_v2/dinov2_layers/mlp.py

@@ -0,0 +1,41 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+# References:
+#   https://github.com/facebookresearch/dino/blob/master/vision_transformer.py
+#   https://github.com/rwightman/pytorch-image-models/tree/master/timm/layers/mlp.py
+
+
+from typing import Callable, Optional
+
+from torch import Tensor, nn
+
+
+class Mlp(nn.Module):
+    def __init__(
+        self,
+        in_features: int,
+        hidden_features: Optional[int] = None,
+        out_features: Optional[int] = None,
+        act_layer: Callable[..., nn.Module] = nn.GELU,
+        drop: float = 0.0,
+        bias: bool = True,
+    ) -> None:
+        super().__init__()
+        out_features = out_features or in_features
+        hidden_features = hidden_features or in_features
+        self.fc1 = nn.Linear(in_features, hidden_features, bias=bias)
+        self.act = act_layer()
+        self.fc2 = nn.Linear(hidden_features, out_features, bias=bias)
+        self.drop = nn.Dropout(drop)
+
+    def forward(self, x: Tensor) -> Tensor:
+        x = self.fc1(x)
+        x = self.act(x)
+        x = self.drop(x)
+        x = self.fc2(x)
+        x = self.drop(x)
+        return x

depth_anything_v2/dinov2_layers/patch_embed.py

@@ -0,0 +1,88 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+# References:
+#   https://github.com/facebookresearch/dino/blob/master/vision_transformer.py
+#   https://github.com/rwightman/pytorch-image-models/tree/master/timm/layers/patch_embed.py
+
+from typing import Callable, Optional, Tuple, Union
+
+from torch import Tensor
+import torch.nn as nn
+
+
+def make_2tuple(x):
+    if isinstance(x, tuple):
+        assert len(x) == 2
+        return x
+
+    assert isinstance(x, int)
+    return (x, x)
+
+
+class PatchEmbed(nn.Module):
+    """
+    2D image to patch embedding: (B,C,H,W) -> (B,N,D)
+    Args:
+        img_size: Image size.
+        patch_size: Patch token size.
+        in_chans: Number of input image channels.
+        embed_dim: Number of linear projection output channels.
+        norm_layer: Normalization layer.
+    """
+
+    def __init__(
+        self,
+        img_size: Union[int, Tuple[int, int]] = 224,
+        patch_size: Union[int, Tuple[int, int]] = 16,
+        in_chans: int = 3,
+        embed_dim: int = 768,
+        norm_layer: Optional[Callable] = None,
+        flatten_embedding: bool = True,
+    ) -> None:
+        super().__init__()
+
+        image_HW = make_2tuple(img_size)
+        patch_HW = make_2tuple(patch_size)
+        patch_grid_size = (
+            image_HW[0] // patch_HW[0],
+            image_HW[1] // patch_HW[1],
+        )
+
+        self.img_size = image_HW
+        self.patch_size = patch_HW
+        self.patches_resolution = patch_grid_size
+        self.num_patches = patch_grid_size[0] * patch_grid_size[1]
+
+        self.in_chans = in_chans
+        self.embed_dim = embed_dim
+
+        self.flatten_embedding = flatten_embedding
+
+        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_HW, stride=patch_HW)
+        self.norm = norm_layer(embed_dim) if norm_layer else nn.Identity()
+
+    def forward(self, x: Tensor) -> Tensor:
+        _, _, H, W = x.shape
+        patch_H, patch_W = self.patch_size
+
+        assert H % patch_H == 0, f"Input image height {H} is not a multiple of patch height {patch_H}"
+        assert W % patch_W == 0, f"Input image width {W} is not a multiple of patch width: {patch_W}"
+
+        x = self.proj(x)  # B C H W
+        H, W = x.size(2), x.size(3)
+        x = x.flatten(2).transpose(1, 2)  # B HW C
+        x = self.norm(x)
+        if not self.flatten_embedding:
+            x = x.reshape(-1, H, W, self.embed_dim)  # B H W C
+        return x
+
+    def flops(self) -> float:
+        Ho, Wo = self.patches_resolution
+        flops = Ho * Wo * self.embed_dim * self.in_chans * (self.patch_size[0] * self.patch_size[1])
+        if self.norm is not None:
+            flops += Ho * Wo * self.embed_dim
+        return flops

depth_anything_v2/dinov2_layers/swiglu_ffn.py

@@ -0,0 +1,63 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+from typing import Callable, Optional
+
+from torch import Tensor, nn
+import torch.nn.functional as F
+
+
+class SwiGLUFFN(nn.Module):
+    def __init__(
+        self,
+        in_features: int,
+        hidden_features: Optional[int] = None,
+        out_features: Optional[int] = None,
+        act_layer: Callable[..., nn.Module] = None,
+        drop: float = 0.0,
+        bias: bool = True,
+    ) -> None:
+        super().__init__()
+        out_features = out_features or in_features
+        hidden_features = hidden_features or in_features
+        self.w12 = nn.Linear(in_features, 2 * hidden_features, bias=bias)
+        self.w3 = nn.Linear(hidden_features, out_features, bias=bias)
+
+    def forward(self, x: Tensor) -> Tensor:
+        x12 = self.w12(x)
+        x1, x2 = x12.chunk(2, dim=-1)
+        hidden = F.silu(x1) * x2
+        return self.w3(hidden)
+
+
+try:
+    from xformers.ops import SwiGLU
+
+    XFORMERS_AVAILABLE = True
+except ImportError:
+    SwiGLU = SwiGLUFFN
+    XFORMERS_AVAILABLE = False
+
+
+class SwiGLUFFNFused(SwiGLU):
+    def __init__(
+        self,
+        in_features: int,
+        hidden_features: Optional[int] = None,
+        out_features: Optional[int] = None,
+        act_layer: Callable[..., nn.Module] = None,
+        drop: float = 0.0,
+        bias: bool = True,
+    ) -> None:
+        out_features = out_features or in_features
+        hidden_features = hidden_features or in_features
+        hidden_features = (int(hidden_features * 2 / 3) + 7) // 8 * 8
+        super().__init__(
+            in_features=in_features,
+            hidden_features=hidden_features,
+            out_features=out_features,
+            bias=bias,
+        )

depth_anything_v2/dpt.py

@@ -0,0 +1,221 @@
+import cv2
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torchvision.transforms import Compose
+
+from .dinov2 import DINOv2
+from .util.blocks import FeatureFusionBlock, _make_scratch
+from .util.transform import Resize, NormalizeImage, PrepareForNet
+
+
+def _make_fusion_block(features, use_bn, size=None):
+    return FeatureFusionBlock(
+        features,
+        nn.ReLU(False),
+        deconv=False,
+        bn=use_bn,
+        expand=False,
+        align_corners=True,
+        size=size,
+    )
+
+
+class ConvBlock(nn.Module):
+    def __init__(self, in_feature, out_feature):
+        super().__init__()
+        
+        self.conv_block = nn.Sequential(
+            nn.Conv2d(in_feature, out_feature, kernel_size=3, stride=1, padding=1),
+            nn.BatchNorm2d(out_feature),
+            nn.ReLU(True)
+        )
+    
+    def forward(self, x):
+        return self.conv_block(x)
+
+
+class DPTHead(nn.Module):
+    def __init__(
+        self, 
+        in_channels, 
+        features=256, 
+        use_bn=False, 
+        out_channels=[256, 512, 1024, 1024], 
+        use_clstoken=False
+    ):
+        super(DPTHead, self).__init__()
+        
+        self.use_clstoken = use_clstoken
+        
+        self.projects = nn.ModuleList([
+            nn.Conv2d(
+                in_channels=in_channels,
+                out_channels=out_channel,
+                kernel_size=1,
+                stride=1,
+                padding=0,
+            ) for out_channel in out_channels
+        ])
+        
+        self.resize_layers = nn.ModuleList([
+            nn.ConvTranspose2d(
+                in_channels=out_channels[0],
+                out_channels=out_channels[0],
+                kernel_size=4,
+                stride=4,
+                padding=0),
+            nn.ConvTranspose2d(
+                in_channels=out_channels[1],
+                out_channels=out_channels[1],
+                kernel_size=2,
+                stride=2,
+                padding=0),
+            nn.Identity(),
+            nn.Conv2d(
+                in_channels=out_channels[3],
+                out_channels=out_channels[3],
+                kernel_size=3,
+                stride=2,
+                padding=1)
+        ])
+        
+        if use_clstoken:
+            self.readout_projects = nn.ModuleList()
+            for _ in range(len(self.projects)):
+                self.readout_projects.append(
+                    nn.Sequential(
+                        nn.Linear(2 * in_channels, in_channels),
+                        nn.GELU()))
+        
+        self.scratch = _make_scratch(
+            out_channels,
+            features,
+            groups=1,
+            expand=False,
+        )
+        
+        self.scratch.stem_transpose = None
+        
+        self.scratch.refinenet1 = _make_fusion_block(features, use_bn)
+        self.scratch.refinenet2 = _make_fusion_block(features, use_bn)
+        self.scratch.refinenet3 = _make_fusion_block(features, use_bn)
+        self.scratch.refinenet4 = _make_fusion_block(features, use_bn)
+        
+        head_features_1 = features
+        head_features_2 = 32
+        
+        self.scratch.output_conv1 = nn.Conv2d(head_features_1, head_features_1 // 2, kernel_size=3, stride=1, padding=1)
+        self.scratch.output_conv2 = nn.Sequential(
+            nn.Conv2d(head_features_1 // 2, head_features_2, kernel_size=3, stride=1, padding=1),
+            nn.ReLU(True),
+            nn.Conv2d(head_features_2, 1, kernel_size=1, stride=1, padding=0),
+            nn.ReLU(True),
+            nn.Identity(),
+        )
+    
+    def forward(self, out_features, patch_h, patch_w):
+        out = []
+        for i, x in enumerate(out_features):
+            if self.use_clstoken:
+                x, cls_token = x[0], x[1]
+                readout = cls_token.unsqueeze(1).expand_as(x)
+                x = self.readout_projects[i](torch.cat((x, readout), -1))
+            else:
+                x = x[0]
+            
+            x = x.permute(0, 2, 1).reshape((x.shape[0], x.shape[-1], patch_h, patch_w))
+            
+            x = self.projects[i](x)
+            x = self.resize_layers[i](x)
+            
+            out.append(x)
+        
+        layer_1, layer_2, layer_3, layer_4 = out
+        
+        layer_1_rn = self.scratch.layer1_rn(layer_1)
+        layer_2_rn = self.scratch.layer2_rn(layer_2)
+        layer_3_rn = self.scratch.layer3_rn(layer_3)
+        layer_4_rn = self.scratch.layer4_rn(layer_4)
+        
+        path_4 = self.scratch.refinenet4(layer_4_rn, size=layer_3_rn.shape[2:])        
+        path_3 = self.scratch.refinenet3(path_4, layer_3_rn, size=layer_2_rn.shape[2:])
+        path_2 = self.scratch.refinenet2(path_3, layer_2_rn, size=layer_1_rn.shape[2:])
+        path_1 = self.scratch.refinenet1(path_2, layer_1_rn)
+        
+        out = self.scratch.output_conv1(path_1)
+        out = F.interpolate(out, (int(patch_h * 14), int(patch_w * 14)), mode="bilinear", align_corners=True)
+        out = self.scratch.output_conv2(out)
+        
+        return out
+
+
+class DepthAnythingV2(nn.Module):
+    def __init__(
+        self, 
+        encoder='vitl', 
+        features=256, 
+        out_channels=[256, 512, 1024, 1024], 
+        use_bn=False, 
+        use_clstoken=False
+    ):
+        super(DepthAnythingV2, self).__init__()
+        
+        self.intermediate_layer_idx = {
+            'vits': [2, 5, 8, 11],
+            'vitb': [2, 5, 8, 11], 
+            'vitl': [4, 11, 17, 23], 
+            'vitg': [9, 19, 29, 39]
+        }
+        
+        self.encoder = encoder
+        self.pretrained = DINOv2(model_name=encoder)
+        
+        self.depth_head = DPTHead(self.pretrained.embed_dim, features, use_bn, out_channels=out_channels, use_clstoken=use_clstoken)
+    
+    def forward(self, x):
+        patch_h, patch_w = x.shape[-2] // 14, x.shape[-1] // 14
+        
+        features = self.pretrained.get_intermediate_layers(x, self.intermediate_layer_idx[self.encoder], return_class_token=True)
+        
+        depth = self.depth_head(features, patch_h, patch_w)
+        depth = F.relu(depth)
+        
+        return depth.squeeze(1)
+    
+    @torch.no_grad()
+    def infer_image(self, raw_image, input_size=518):
+        image, (h, w) = self.image2tensor(raw_image, input_size)
+        
+        depth = self.forward(image)
+        
+        depth = F.interpolate(depth[:, None], (h, w), mode="bilinear", align_corners=True)[0, 0]
+        
+        return depth.cpu().numpy()
+    
+    def image2tensor(self, raw_image, input_size=518):        
+        transform = Compose([
+            Resize(
+                width=input_size,
+                height=input_size,
+                resize_target=False,
+                keep_aspect_ratio=True,
+                ensure_multiple_of=14,
+                resize_method='lower_bound',
+                image_interpolation_method=cv2.INTER_CUBIC,
+            ),
+            NormalizeImage(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
+            PrepareForNet(),
+        ])
+        
+        h, w = raw_image.shape[:2]
+        
+        image = cv2.cvtColor(raw_image, cv2.COLOR_BGR2RGB) / 255.0
+        
+        image = transform({'image': image})['image']
+        image = torch.from_numpy(image).unsqueeze(0)
+        
+        DEVICE = 'cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu'
+        image = image.to(DEVICE)
+        
+        return image, (h, w)

depth_anything_v2/util/blocks.py

@@ -0,0 +1,144 @@
+import torch.nn as nn
+
+
+def _make_scratch(in_shape, out_shape, groups=1, expand=False):
+    scratch = nn.Module()
+
+    out_shape1 = out_shape
+    out_shape2 = out_shape
+    out_shape3 = out_shape
+    if len(in_shape) >= 4:
+        out_shape4 = out_shape
+
+    if expand:
+        out_shape1 = out_shape
+        out_shape2 = out_shape * 2
+        out_shape3 = out_shape * 4
+        if len(in_shape) >= 4:
+            out_shape4 = out_shape * 8
+
+    scratch.layer1_rn = nn.Conv2d(in_shape[0], out_shape1, kernel_size=3, stride=1, padding=1, bias=False, groups=groups)
+    scratch.layer2_rn = nn.Conv2d(in_shape[1], out_shape2, kernel_size=3, stride=1, padding=1, bias=False, groups=groups)
+    scratch.layer3_rn = nn.Conv2d(in_shape[2], out_shape3, kernel_size=3, stride=1, padding=1, bias=False, groups=groups)
+    if len(in_shape) >= 4:
+        scratch.layer4_rn = nn.Conv2d(in_shape[3], out_shape4, kernel_size=3, stride=1, padding=1, bias=False, groups=groups)
+
+    return scratch
+
+
+class ResidualConvUnit(nn.Module):
+    """Residual convolution module.
+    """
+
+    def __init__(self, features, activation, bn):
+        """Init.
+        Args:
+            features (int): number of features
+        """
+        super().__init__()
+
+        self.bn = bn
+
+        self.groups=1
+
+        self.conv1 = nn.Conv2d(features, features, kernel_size=3, stride=1, padding=1, bias=True, groups=self.groups)
+        
+        self.conv2 = nn.Conv2d(features, features, kernel_size=3, stride=1, padding=1, bias=True, groups=self.groups)
+
+        if self.bn == True:
+            self.bn1 = nn.BatchNorm2d(features)
+            self.bn2 = nn.BatchNorm2d(features)
+
+        self.activation = activation
+
+        self.skip_add = nn.quantized.FloatFunctional()
+
+    def forward(self, x):
+        """Forward pass.
+        Args:
+            x (tensor): input
+        Returns:
+            tensor: output
+        """
+        
+        out = self.activation(x)
+        out = self.conv1(out)
+        if self.bn == True:
+            out = self.bn1(out)
+       
+        out = self.activation(out)
+        out = self.conv2(out)
+        if self.bn == True:
+            out = self.bn2(out)
+
+        if self.groups > 1:
+            out = self.conv_merge(out)
+
+        return self.skip_add.add(out, x)
+
+
+class FeatureFusionBlock(nn.Module):
+    """Feature fusion block.
+    """
+
+    def __init__(
+        self, 
+        features, 
+        activation, 
+        deconv=False, 
+        bn=False, 
+        expand=False, 
+        align_corners=True,
+        size=None
+    ):
+        """Init.
+        
+        Args:
+            features (int): number of features
+        """
+        super(FeatureFusionBlock, self).__init__()
+
+        self.deconv = deconv
+        self.align_corners = align_corners
+
+        self.groups=1
+
+        self.expand = expand
+        out_features = features
+        if self.expand == True:
+            out_features = features // 2
+        
+        self.out_conv = nn.Conv2d(features, out_features, kernel_size=1, stride=1, padding=0, bias=True, groups=1)
+
+        self.resConfUnit1 = ResidualConvUnit(features, activation, bn)
+        self.resConfUnit2 = ResidualConvUnit(features, activation, bn)
+        
+        self.skip_add = nn.quantized.FloatFunctional()
+
+        self.size=size
+
+    def forward(self, *xs, size=None):
+        """Forward pass.
+        Returns:
+            tensor: output
+        """
+        output = xs[0]
+
+        if len(xs) == 2:
+            res = self.resConfUnit1(xs[1])
+            output = self.skip_add.add(output, res)
+
+        output = self.resConfUnit2(output)
+
+        if (size is None) and (self.size is None):
+            modifier = {"scale_factor": 2}
+        elif size is None:
+            modifier = {"size": self.size}
+        else:
+            modifier = {"size": size}
+
+        output = nn.functional.interpolate(output, **modifier, mode="bilinear", align_corners=self.align_corners)
+        
+        output = self.out_conv(output)
+
+        return output

depth_anything_v2/util/transform.py

@@ -0,0 +1,157 @@
+import numpy as np
+import cv2
+
+
+class Resize(object):
+    """Resize sample to given size (width, height).
+    """
+
+    def __init__(
+        self,
+        width,
+        height,
+        resize_target=True,
+        keep_aspect_ratio=False,
+        ensure_multiple_of=1,
+        resize_method="lower_bound",
+        image_interpolation_method=cv2.INTER_AREA,
+    ):
+        """Init.
+        Args:
+            width (int): desired output width
+            height (int): desired output height
+            resize_target (bool, optional):
+                True: Resize the full sample (image, mask, target).
+                False: Resize image only.
+                Defaults to True.
+            keep_aspect_ratio (bool, optional):
+                True: Keep the aspect ratio of the input sample.
+                Output sample might not have the given width and height, and
+                resize behaviour depends on the parameter 'resize_method'.
+                Defaults to False.
+            ensure_multiple_of (int, optional):
+                Output width and height is constrained to be multiple of this parameter.
+                Defaults to 1.
+            resize_method (str, optional):
+                "lower_bound": Output will be at least as large as the given size.
+                "upper_bound": Output will be at max as large as the given size. (Output size might be smaller than given size.)
+                "minimal": Scale as least as possible.  (Output size might be smaller than given size.)
+                Defaults to "lower_bound".
+        """
+        self.__width = width
+        self.__height = height
+
+        self.__resize_target = resize_target
+        self.__keep_aspect_ratio = keep_aspect_ratio
+        self.__multiple_of = ensure_multiple_of
+        self.__resize_method = resize_method
+        self.__image_interpolation_method = image_interpolation_method
+
+    def constrain_to_multiple_of(self, x, min_val=0, max_val=None):
+        y = (np.round(x / self.__multiple_of) * self.__multiple_of).astype(int)
+
+        if max_val is not None and y > max_val:
+            y = (np.floor(x / self.__multiple_of) * self.__multiple_of).astype(int)
+
+        if y < min_val:
+            y = (np.ceil(x / self.__multiple_of) * self.__multiple_of).astype(int)
+
+        return y
+
+    def get_size(self, width, height):
+        # determine new height and width
+        scale_height = self.__height / height
+        scale_width = self.__width / width
+
+        if self.__keep_aspect_ratio:
+            if self.__resize_method == "lower_bound":
+                # scale such that output size is lower bound
+                if scale_width > scale_height:
+                    # fit width
+                    scale_height = scale_width
+                else:
+                    # fit height
+                    scale_width = scale_height
+            elif self.__resize_method == "upper_bound":
+                # scale such that output size is upper bound
+                if scale_width < scale_height:
+                    # fit width
+                    scale_height = scale_width
+                else:
+                    # fit height
+                    scale_width = scale_height
+            elif self.__resize_method == "minimal":
+                # scale as least as possbile
+                if abs(1 - scale_width) < abs(1 - scale_height):
+                    # fit width
+                    scale_height = scale_width
+                else:
+                    # fit height
+                    scale_width = scale_height
+            else:
+                raise ValueError(f"resize_method {self.__resize_method} not implemented")
+
+        if self.__resize_method == "lower_bound":
+            new_height = self.constrain_to_multiple_of(scale_height * height, min_val=self.__height)
+            new_width = self.constrain_to_multiple_of(scale_width * width, min_val=self.__width)
+        elif self.__resize_method == "upper_bound":
+            new_height = self.constrain_to_multiple_of(scale_height * height, max_val=self.__height)
+            new_width = self.constrain_to_multiple_of(scale_width * width, max_val=self.__width)
+        elif self.__resize_method == "minimal":
+            new_height = self.constrain_to_multiple_of(scale_height * height)
+            new_width = self.constrain_to_multiple_of(scale_width * width)
+        else:
+            raise ValueError(f"resize_method {self.__resize_method} not implemented")
+
+        return (new_width, new_height)
+
+    def __call__(self, sample):
+        width, height = self.get_size(sample["image"].shape[1], sample["image"].shape[0])
+        
+        # resize sample
+        sample["image"] = cv2.resize(sample["image"], (width, height), interpolation=self.__image_interpolation_method)
+
+        if self.__resize_target:
+            if "depth" in sample:
+                sample["depth"] = cv2.resize(sample["depth"], (width, height), interpolation=cv2.INTER_NEAREST)
+                
+            if "mask" in sample:
+                sample["mask"] = cv2.resize(sample["mask"].astype(np.float32), (width, height), interpolation=cv2.INTER_NEAREST)
+        
+        return sample
+
+
+class NormalizeImage(object):
+    """Normlize image by given mean and std.
+    """
+
+    def __init__(self, mean, std):
+        self.__mean = mean
+        self.__std = std
+
+    def __call__(self, sample):
+        sample["image"] = (sample["image"] - self.__mean) / self.__std
+
+        return sample
+
+
+class PrepareForNet(object):
+    """Prepare sample for usage as network input.
+    """
+
+    def __init__(self):
+        pass
+
+    def __call__(self, sample):
+        image = np.transpose(sample["image"], (2, 0, 1))
+        sample["image"] = np.ascontiguousarray(image).astype(np.float32)
+
+        if "depth" in sample:
+            depth = sample["depth"].astype(np.float32)
+            sample["depth"] = np.ascontiguousarray(depth)
+        
+        if "mask" in sample:
+            sample["mask"] = sample["mask"].astype(np.float32)
+            sample["mask"] = np.ascontiguousarray(sample["mask"])
+        
+        return sample

pipeline_flux_controlnet_inpaint.py

@@ -1,1046 +0,0 @@
-import inspect
-from typing import Any, Callable, Dict, List, Optional, Union
-
-import numpy as np
-import torch
-from transformers import (
-    CLIPTextModel,
-    CLIPTokenizer,
-    T5EncoderModel,
-    T5TokenizerFast,
-)
-
-from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
-from diffusers.loaders import FluxLoraLoaderMixin
-from diffusers.models.autoencoders import AutoencoderKL
-
-from diffusers.schedulers import FlowMatchEulerDiscreteScheduler
-from diffusers.utils import (
-    USE_PEFT_BACKEND,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
-from diffusers.utils.torch_utils import randn_tensor
-from diffusers.pipelines.pipeline_utils import DiffusionPipeline
-from diffusers.pipelines.flux.pipeline_output import FluxPipelineOutput
-
-from transformer_flux import FluxTransformer2DModel
-from controlnet_flux import FluxControlNetModel
-
-if is_torch_xla_available():
-    import torch_xla.core.xla_model as xm
-
-    XLA_AVAILABLE = True
-else:
-    XLA_AVAILABLE = False
-
-
-logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
-
-EXAMPLE_DOC_STRING = """
-    Examples:
-        ```py
-        >>> import torch
-        >>> from diffusers.utils import load_image
-        >>> from diffusers import FluxControlNetPipeline
-        >>> from diffusers import FluxControlNetModel
-
-        >>> controlnet_model = "InstantX/FLUX.1-dev-controlnet-canny-alpha"
-        >>> controlnet = FluxControlNetModel.from_pretrained(controlnet_model, torch_dtype=torch.bfloat16)
-        >>> pipe = FluxControlNetPipeline.from_pretrained(
-        ...     base_model, controlnet=controlnet, torch_dtype=torch.bfloat16
-        ... )
-        >>> pipe.to("cuda")
-        >>> control_image = load_image("https://huggingface.co/InstantX/SD3-Controlnet-Canny/resolve/main/canny.jpg")
-        >>> control_mask = load_image("https://huggingface.co/InstantX/SD3-Controlnet-Canny/resolve/main/canny.jpg")
-        >>> prompt = "A girl in city, 25 years old, cool, futuristic"
-        >>> image = pipe(
-        ...     prompt,
-        ...     control_image=control_image,
-        ...     controlnet_conditioning_scale=0.6,
-        ...     num_inference_steps=28,
-        ...     guidance_scale=3.5,
-        ... ).images[0]
-        >>> image.save("flux.png")
-        ```
-"""
-
-
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.16,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    """
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError(
-            "Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values"
-        )
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(
-            inspect.signature(scheduler.set_timesteps).parameters.keys()
-        )
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(
-            inspect.signature(scheduler.set_timesteps).parameters.keys()
-        )
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class FluxControlNetInpaintingPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
-    r"""
-    The Flux pipeline for text-to-image generation.
-
-    Reference: https://blackforestlabs.ai/announcing-black-forest-labs/
-
-    Args:
-        transformer ([`FluxTransformer2DModel`]):
-            Conditional Transformer (MMDiT) architecture to denoise the encoded image latents.
-        scheduler ([`FlowMatchEulerDiscreteScheduler`]):
-            A scheduler to be used in combination with `transformer` to denoise the encoded image latents.
-        vae ([`AutoencoderKL`]):
-            Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
-        text_encoder ([`CLIPTextModel`]):
-            [CLIP](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), specifically
-            the [clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) variant.
-        text_encoder_2 ([`T5EncoderModel`]):
-            [T5](https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5EncoderModel), specifically
-            the [google/t5-v1_1-xxl](https://huggingface.co/google/t5-v1_1-xxl) variant.
-        tokenizer (`CLIPTokenizer`):
-            Tokenizer of class
-            [CLIPTokenizer](https://huggingface.co/docs/transformers/en/model_doc/clip#transformers.CLIPTokenizer).
-        tokenizer_2 (`T5TokenizerFast`):
-            Second Tokenizer of class
-            [T5TokenizerFast](https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5TokenizerFast).
-    """
-
-    model_cpu_offload_seq = "text_encoder->text_encoder_2->transformer->vae"
-    _optional_components = []
-    _callback_tensor_inputs = ["latents", "prompt_embeds"]
-
-    def __init__(
-        self,
-        scheduler: FlowMatchEulerDiscreteScheduler,
-        vae: AutoencoderKL,
-        text_encoder: CLIPTextModel,
-        tokenizer: CLIPTokenizer,
-        text_encoder_2: T5EncoderModel,
-        tokenizer_2: T5TokenizerFast,
-        transformer: FluxTransformer2DModel,
-        controlnet: FluxControlNetModel,
-    ):
-        super().__init__()
-
-        self.register_modules(
-            vae=vae,
-            text_encoder=text_encoder,
-            text_encoder_2=text_encoder_2,
-            tokenizer=tokenizer,
-            tokenizer_2=tokenizer_2,
-            transformer=transformer,
-            scheduler=scheduler,
-            controlnet=controlnet,
-        )
-        self.vae_scale_factor = (
-            2 ** (len(self.vae.config.block_out_channels))
-            if hasattr(self, "vae") and self.vae is not None
-            else 16
-        )
-        self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_resize=True, do_convert_rgb=True, do_normalize=True)
-        self.mask_processor = VaeImageProcessor(
-            vae_scale_factor=self.vae_scale_factor,
-            do_resize=True,
-            do_convert_grayscale=True,
-            do_normalize=False,
-            do_binarize=True,
-        )
-        self.tokenizer_max_length = (
-            self.tokenizer.model_max_length
-            if hasattr(self, "tokenizer") and self.tokenizer is not None
-            else 77
-        )
-        self.default_sample_size = 64
-    
-    @property
-    def do_classifier_free_guidance(self):
-        return self._guidance_scale > 1
-
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(
-            prompt, padding="longest", return_tensors="pt"
-        ).input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-            text_input_ids, untruncated_ids
-        ):
-            removed_text = self.tokenizer_2.batch_decode(
-                untruncated_ids[:, self.tokenizer_max_length - 1 : -1]
-            )
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(
-            text_input_ids.to(device), output_hidden_states=False
-        )[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(
-            batch_size * num_images_per_prompt, seq_len, -1
-        )
-
-        return prompt_embeds
-
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(
-            prompt, padding="longest", return_tensors="pt"
-        ).input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-            text_input_ids, untruncated_ids
-        ):
-            removed_text = self.tokenizer.batch_decode(
-                untruncated_ids[:, self.tokenizer_max_length - 1 : -1]
-            )
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(
-            text_input_ids.to(device), output_hidden_states=False
-        )
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Union[str, List[str]],
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        do_classifier_free_guidance: bool = True,
-        negative_prompt: Optional[Union[str, List[str]]] = None,
-        negative_prompt_2: Optional[Union[str, List[str]]] = None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier-free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                negative prompt to be encoded
-            negative_prompt_2 (`str` or `List[str]`, *optional*):
-                negative prompt to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `negative_prompt` is
-                used in all text-encoders
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        if prompt is not None:
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if do_classifier_free_guidance:
-            # 处理 negative prompt
-            negative_prompt = negative_prompt or ""
-            negative_prompt_2 = negative_prompt_2 or negative_prompt
-            
-            negative_pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                negative_prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            negative_prompt_embeds = self._get_t5_prompt_embeds(
-                negative_prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        text_ids = torch.zeros(batch_size, prompt_embeds.shape[1], 3).to(
-            device=device, dtype=self.text_encoder.dtype
-        )
-
-        return prompt_embeds, pooled_prompt_embeds, negative_prompt_embeds, negative_pooled_prompt_embeds,text_ids
-
-    def check_inputs(
-        self,
-        prompt,
-        prompt_2,
-        height,
-        width,
-        prompt_embeds=None,
-        pooled_prompt_embeds=None,
-        callback_on_step_end_tensor_inputs=None,
-        max_sequence_length=None,
-    ):
-        if height % 8 != 0 or width % 8 != 0:
-            raise ValueError(
-                f"`height` and `width` have to be divisible by 8 but are {height} and {width}."
-            )
-
-        if callback_on_step_end_tensor_inputs is not None and not all(
-            k in self._callback_tensor_inputs
-            for k in callback_on_step_end_tensor_inputs
-        ):
-            raise ValueError(
-                f"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}"
-            )
-
-        if prompt is not None and prompt_embeds is not None:
-            raise ValueError(
-                f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
-                " only forward one of the two."
-            )
-        elif prompt_2 is not None and prompt_embeds is not None:
-            raise ValueError(
-                f"Cannot forward both `prompt_2`: {prompt_2} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
-                " only forward one of the two."
-            )
-        elif prompt is None and prompt_embeds is None:
-            raise ValueError(
-                "Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined."
-            )
-        elif prompt is not None and (
-            not isinstance(prompt, str) and not isinstance(prompt, list)
-        ):
-            raise ValueError(
-                f"`prompt` has to be of type `str` or `list` but is {type(prompt)}"
-            )
-        elif prompt_2 is not None and (
-            not isinstance(prompt_2, str) and not isinstance(prompt_2, list)
-        ):
-            raise ValueError(
-                f"`prompt_2` has to be of type `str` or `list` but is {type(prompt_2)}"
-            )
-
-        if prompt_embeds is not None and pooled_prompt_embeds is None:
-            raise ValueError(
-                "If `prompt_embeds` are provided, `pooled_prompt_embeds` also have to be passed. Make sure to generate `pooled_prompt_embeds` from the same text encoder that was used to generate `prompt_embeds`."
-            )
-
-        if max_sequence_length is not None and max_sequence_length > 512:
-            raise ValueError(
-                f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}"
-            )
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux._prepare_latent_image_ids
-    @staticmethod
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height // 2, width // 2, 3)
-        latent_image_ids[..., 1] = (
-            latent_image_ids[..., 1] + torch.arange(height // 2)[:, None]
-        )
-        latent_image_ids[..., 2] = (
-            latent_image_ids[..., 2] + torch.arange(width // 2)[None, :]
-        )
-
-        (
-            latent_image_id_height,
-            latent_image_id_width,
-            latent_image_id_channels,
-        ) = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids[None, :].repeat(batch_size, 1, 1, 1)
-        latent_image_ids = latent_image_ids.reshape(
-            batch_size,
-            latent_image_id_height * latent_image_id_width,
-            latent_image_id_channels,
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux._pack_latents
-    @staticmethod
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(
-            batch_size, num_channels_latents, height // 2, 2, width // 2, 2
-        )
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(
-            batch_size, (height // 2) * (width // 2), num_channels_latents * 4
-        )
-
-        return latents
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux._unpack_latents
-    @staticmethod
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        height = height // vae_scale_factor
-        width = width // vae_scale_factor
-
-        latents = latents.view(batch_size, height, width, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(
-            batch_size, channels // (2 * 2), height * 2, width * 2
-        )
-
-        return latents
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.prepare_latents
-    def prepare_latents(
-        self,
-        batch_size,
-        num_channels_latents,
-        height,
-        width,
-        dtype,
-        device,
-        generator,
-        latents=None,
-    ):
-        height = 2 * (int(height) // self.vae_scale_factor)
-        width = 2 * (int(width) // self.vae_scale_factor)
-
-        shape = (batch_size, num_channels_latents, height, width)
-
-        if latents is not None:
-            latent_image_ids = self._prepare_latent_image_ids(
-                batch_size, height, width, device, dtype
-            )
-            return latents.to(device=device, dtype=dtype), latent_image_ids
-
-        if isinstance(generator, list) and len(generator) != batch_size:
-            raise ValueError(
-                f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
-                f" size of {batch_size}. Make sure the batch size matches the length of the generators."
-            )
-
-        latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
-        latents = self._pack_latents(
-            latents, batch_size, num_channels_latents, height, width
-        )
-
-        latent_image_ids = self._prepare_latent_image_ids(
-            batch_size, height, width, device, dtype
-        )
-
-        return latents, latent_image_ids
-
-    # Copied from diffusers.pipelines.controlnet.pipeline_controlnet.StableDiffusionControlNetPipeline.prepare_image
-    def prepare_image(
-        self,
-        image,
-        width,
-        height,
-        batch_size,
-        num_images_per_prompt,
-        device,
-        dtype,
-    ):
-        if isinstance(image, torch.Tensor):
-            pass
-        else:
-            image = self.image_processor.preprocess(image, height=height, width=width)
-
-        image_batch_size = image.shape[0]
-
-        if image_batch_size == 1:
-            repeat_by = batch_size
-        else:
-            # image batch size is the same as prompt batch size
-            repeat_by = num_images_per_prompt
-
-        image = image.repeat_interleave(repeat_by, dim=0)
-
-        image = image.to(device=device, dtype=dtype)
-
-        return image
-
-    def prepare_image_with_mask(
-        self,
-        image,
-        mask,
-        width,
-        height,
-        batch_size,
-        num_images_per_prompt,
-        device,
-        dtype,
-        do_classifier_free_guidance = False,
-    ):
-        # Prepare image
-        if isinstance(image, torch.Tensor):
-            pass
-        else:
-            image = self.image_processor.preprocess(image, height=height, width=width)
-
-        image_batch_size = image.shape[0]
-        if image_batch_size == 1:
-            repeat_by = batch_size
-        else:
-            # image batch size is the same as prompt batch size
-            repeat_by = num_images_per_prompt
-        image = image.repeat_interleave(repeat_by, dim=0)
-        image = image.to(device=device, dtype=dtype)
-
-        # Prepare mask
-        if isinstance(mask, torch.Tensor):
-            pass
-        else:
-            mask = self.mask_processor.preprocess(mask, height=height, width=width)
-        mask = mask.repeat_interleave(repeat_by, dim=0)
-        mask = mask.to(device=device, dtype=dtype)
-
-        # Get masked image
-        masked_image = image.clone()
-        masked_image[(mask > 0.5).repeat(1, 3, 1, 1)] = -1
-
-        # Encode to latents
-        image_latents = self.vae.encode(masked_image.to(self.vae.dtype)).latent_dist.sample()
-        image_latents = (
-            image_latents - self.vae.config.shift_factor
-        ) * self.vae.config.scaling_factor
-        image_latents = image_latents.to(dtype)
-
-        mask = torch.nn.functional.interpolate(
-            mask, size=(height // self.vae_scale_factor * 2, width // self.vae_scale_factor * 2)
-        )
-        mask = 1 - mask
-
-        control_image = torch.cat([image_latents, mask], dim=1)
-
-        # Pack cond latents
-        packed_control_image = self._pack_latents(
-            control_image,
-            batch_size * num_images_per_prompt,
-            control_image.shape[1],
-            control_image.shape[2],
-            control_image.shape[3],
-        )
-        
-        if do_classifier_free_guidance:
-            packed_control_image = torch.cat([packed_control_image] * 2)
-
-        return packed_control_image, height, width
-
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
-    @torch.no_grad()
-    @replace_example_docstring(EXAMPLE_DOC_STRING)
-    def __call__(
-        self,
-        prompt: Union[str, List[str]] = None,
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        height: Optional[int] = None,
-        width: Optional[int] = None,
-        num_inference_steps: int = 28,
-        timesteps: List[int] = None,
-        guidance_scale: float = 7.0,
-        true_guidance_scale: float = 3.5 ,
-        negative_prompt: Optional[Union[str, List[str]]] = None,
-        negative_prompt_2: Optional[Union[str, List[str]]] = None,
-        control_image: PipelineImageInput = None,
-        control_mask: PipelineImageInput = None,
-        controlnet_conditioning_scale: Union[float, List[float]] = 1.0,
-        num_images_per_prompt: Optional[int] = 1,
-        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        output_type: Optional[str] = "pil",
-        return_dict: bool = True,
-        joint_attention_kwargs: Optional[Dict[str, Any]] = None,
-        callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
-        callback_on_step_end_tensor_inputs: List[str] = ["latents"],
-        max_sequence_length: int = 512,
-    ):
-        r"""
-        Function invoked when calling the pipeline for generation.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
-                instead.
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                will be used instead
-            height (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):
-                The height in pixels of the generated image. This is set to 1024 by default for the best results.
-            width (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):
-                The width in pixels of the generated image. This is set to 1024 by default for the best results.
-            num_inference_steps (`int`, *optional*, defaults to 50):
-                The number of denoising steps. More denoising steps usually lead to a higher quality image at the
-                expense of slower inference.
-            timesteps (`List[int]`, *optional*):
-                Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument
-                in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is
-                passed will be used. Must be in descending order.
-            guidance_scale (`float`, *optional*, defaults to 7.0):
-                Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
-                `guidance_scale` is defined as `w` of equation 2. of [Imagen
-                Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
-                1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
-                usually at the expense of lower image quality.
-            num_images_per_prompt (`int`, *optional*, defaults to 1):
-                The number of images to generate per prompt.
-            generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
-                One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html)
-                to make generation deterministic.
-            latents (`torch.FloatTensor`, *optional*):
-                Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
-                generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
-                tensor will ge generated by sampling using the supplied random `generator`.
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            output_type (`str`, *optional*, defaults to `"pil"`):
-                The output format of the generate image. Choose between
-                [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
-            return_dict (`bool`, *optional*, defaults to `True`):
-                Whether or not to return a [`~pipelines.flux.FluxPipelineOutput`] instead of a plain tuple.
-            joint_attention_kwargs (`dict`, *optional*):
-                A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
-                `self.processor` in
-                [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
-            callback_on_step_end (`Callable`, *optional*):
-                A function that calls at the end of each denoising steps during the inference. The function is called
-                with the following arguments: `callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int,
-                callback_kwargs: Dict)`. `callback_kwargs` will include a list of all tensors as specified by
-                `callback_on_step_end_tensor_inputs`.
-            callback_on_step_end_tensor_inputs (`List`, *optional*):
-                The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list
-                will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the
-                `._callback_tensor_inputs` attribute of your pipeline class.
-            max_sequence_length (`int` defaults to 512): Maximum sequence length to use with the `prompt`.
-
-        Examples:
-
-        Returns:
-            [`~pipelines.flux.FluxPipelineOutput`] or `tuple`: [`~pipelines.flux.FluxPipelineOutput`] if `return_dict`
-            is True, otherwise a `tuple`. When returning a tuple, the first element is a list with the generated
-            images.
-        """
-
-        height = height or self.default_sample_size * self.vae_scale_factor
-        width = width or self.default_sample_size * self.vae_scale_factor
-
-        # 1. Check inputs. Raise error if not correct
-        self.check_inputs(
-            prompt,
-            prompt_2,
-            height,
-            width,
-            prompt_embeds=prompt_embeds,
-            pooled_prompt_embeds=pooled_prompt_embeds,
-            callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs,
-            max_sequence_length=max_sequence_length,
-        )
-
-        self._guidance_scale = true_guidance_scale
-        self._joint_attention_kwargs = joint_attention_kwargs
-        self._interrupt = False
-
-        # 2. Define call parameters
-        if prompt is not None and isinstance(prompt, str):
-            batch_size = 1
-        elif prompt is not None and isinstance(prompt, list):
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        device = self._execution_device
-        dtype = self.transformer.dtype
-
-        lora_scale = (
-            self.joint_attention_kwargs.get("scale", None)
-            if self.joint_attention_kwargs is not None
-            else None
-        )
-        (            
-            prompt_embeds,
-            pooled_prompt_embeds,
-            negative_prompt_embeds,
-            negative_pooled_prompt_embeds,
-            text_ids
-        ) = self.encode_prompt(
-            prompt=prompt,
-            prompt_2=prompt_2,
-            prompt_embeds=prompt_embeds,
-            pooled_prompt_embeds=pooled_prompt_embeds,
-            do_classifier_free_guidance = self.do_classifier_free_guidance,
-            negative_prompt = negative_prompt,
-            negative_prompt_2 = negative_prompt_2,
-            device=device,
-            num_images_per_prompt=num_images_per_prompt,
-            max_sequence_length=max_sequence_length,
-            lora_scale=lora_scale,
-        )
-        
-        # 在 encode_prompt 之后
-        if self.do_classifier_free_guidance:
-            prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds], dim = 0)
-            pooled_prompt_embeds = torch.cat([negative_pooled_prompt_embeds, pooled_prompt_embeds], dim = 0)
-            text_ids = torch.cat([text_ids, text_ids], dim = 0)
-
-        # 3. Prepare control image
-        num_channels_latents = self.transformer.config.in_channels // 4
-        if isinstance(self.controlnet, FluxControlNetModel):
-            control_image, height, width = self.prepare_image_with_mask(
-                image=control_image,
-                mask=control_mask,
-                width=width,
-                height=height,
-                batch_size=batch_size * num_images_per_prompt,
-                num_images_per_prompt=num_images_per_prompt,
-                device=device,
-                dtype=dtype,
-                do_classifier_free_guidance=self.do_classifier_free_guidance,
-            )
-
-        # 4. Prepare latent variables
-        num_channels_latents = self.transformer.config.in_channels // 4
-        latents, latent_image_ids = self.prepare_latents(
-            batch_size * num_images_per_prompt,
-            num_channels_latents,
-            height,
-            width,
-            prompt_embeds.dtype,
-            device,
-            generator,
-            latents,
-        )
-        
-        if self.do_classifier_free_guidance:
-            latent_image_ids = torch.cat([latent_image_ids] * 2)
-
-        # 5. Prepare timesteps
-        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
-        image_seq_len = latents.shape[1]
-        mu = calculate_shift(
-            image_seq_len,
-            self.scheduler.config.base_image_seq_len,
-            self.scheduler.config.max_image_seq_len,
-            self.scheduler.config.base_shift,
-            self.scheduler.config.max_shift,
-        )
-        timesteps, num_inference_steps = retrieve_timesteps(
-            self.scheduler,
-            num_inference_steps,
-            device,
-            timesteps,
-            sigmas,
-            mu=mu,
-        )
-
-        num_warmup_steps = max(
-            len(timesteps) - num_inference_steps * self.scheduler.order, 0
-        )
-        self._num_timesteps = len(timesteps)
-
-        # 6. Denoising loop
-        with self.progress_bar(total=num_inference_steps) as progress_bar:
-            for i, t in enumerate(timesteps):
-                if self.interrupt:
-                    continue
-                
-                latent_model_input = torch.cat([latents] * 2) if self.do_classifier_free_guidance else latents
-
-                # broadcast to batch dimension in a way that's compatible with ONNX/Core ML
-                timestep = t.expand(latent_model_input.shape[0]).to(latent_model_input.dtype)
-
-                # handle guidance
-                if self.transformer.config.guidance_embeds:
-                    guidance = torch.tensor([guidance_scale], device=device)
-                    guidance = guidance.expand(latent_model_input.shape[0])
-                else:
-                    guidance = None
-
-                # controlnet
-                (
-                    controlnet_block_samples,
-                    controlnet_single_block_samples,
-                ) = self.controlnet(
-                    hidden_states=latent_model_input,
-                    controlnet_cond=control_image,
-                    conditioning_scale=controlnet_conditioning_scale,
-                    timestep=timestep / 1000,
-                    guidance=guidance,
-                    pooled_projections=pooled_prompt_embeds,
-                    encoder_hidden_states=prompt_embeds,
-                    txt_ids=text_ids,
-                    img_ids=latent_image_ids,
-                    joint_attention_kwargs=self.joint_attention_kwargs,
-                    return_dict=False,
-                )
-
-                noise_pred = self.transformer(
-                    hidden_states=latent_model_input,
-                    # YiYi notes: divide it by 1000 for now because we scale it by 1000 in the transforme rmodel (we should not keep it but I want to keep the inputs same for the model for testing)
-                    timestep=timestep / 1000,
-                    guidance=guidance,
-                    pooled_projections=pooled_prompt_embeds,
-                    encoder_hidden_states=prompt_embeds,
-                    controlnet_block_samples=[
-                        sample.to(dtype=self.transformer.dtype)
-                        for sample in controlnet_block_samples
-                    ],
-                    controlnet_single_block_samples=[
-                        sample.to(dtype=self.transformer.dtype)
-                        for sample in controlnet_single_block_samples
-                    ] if controlnet_single_block_samples is not None else controlnet_single_block_samples,
-                    txt_ids=text_ids,
-                    img_ids=latent_image_ids,
-                    joint_attention_kwargs=self.joint_attention_kwargs,
-                    return_dict=False,
-                )[0]
-                
-                # 在生成循环中
-                if self.do_classifier_free_guidance:
-                    noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
-                    noise_pred = noise_pred_uncond + true_guidance_scale * (noise_pred_text - noise_pred_uncond)
-
-                # compute the previous noisy sample x_t -> x_t-1
-                latents_dtype = latents.dtype
-                latents = self.scheduler.step(
-                    noise_pred, t, latents, return_dict=False
-                )[0]
-
-                if latents.dtype != latents_dtype:
-                    if torch.backends.mps.is_available():
-                        # some platforms (eg. apple mps) misbehave due to a pytorch bug: https://github.com/pytorch/pytorch/pull/99272
-                        latents = latents.to(latents_dtype)
-
-                if callback_on_step_end is not None:
-                    callback_kwargs = {}
-                    for k in callback_on_step_end_tensor_inputs:
-                        callback_kwargs[k] = locals()[k]
-                    callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)
-
-                    latents = callback_outputs.pop("latents", latents)
-                    prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds)
-
-                # call the callback, if provided
-                if i == len(timesteps) - 1 or (
-                    (i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0
-                ):
-                    progress_bar.update()
-
-                if XLA_AVAILABLE:
-                    xm.mark_step()
-
-        if output_type == "latent":
-            image = latents
-
-        else:
-            latents = self._unpack_latents(
-                latents, height, width, self.vae_scale_factor
-            )
-            latents = (
-                latents / self.vae.config.scaling_factor
-            ) + self.vae.config.shift_factor
-            latents = latents.to(self.vae.dtype)
-
-            image = self.vae.decode(latents, return_dict=False)[0]
-            image = self.image_processor.postprocess(image, output_type=output_type)
-
-        # Offload all models
-        self.maybe_free_model_hooks()
-
-        if not return_dict:
-            return (image,)
-
-        return FluxPipelineOutput(images=image)

preprocessor.py

@@ -81,4 +81,4 @@ class Preprocessor:
             image = np.array(image)
             image = HWC3(image)
             image = resize_image(image, resolution=image_resolution)
-            return PIL.Image.fromarray(image)
+            return PIL.Image.fromarray(image)

requirements.txt

@@ -7,7 +7,7 @@ einops
 spaces
 gradio
 opencv-python
-diffusers==0.30.2
+git+https://github.com/huggingface/diffusers.git
 boto3
 sentencepiece
 peft

transformer_flux.py

@@ -1,525 +0,0 @@
-from typing import Any, Dict, List, Optional, Union
-
-import numpy as np
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-
-from diffusers.configuration_utils import ConfigMixin, register_to_config
-from diffusers.loaders import FromOriginalModelMixin, PeftAdapterMixin
-from diffusers.models.attention import FeedForward
-from diffusers.models.attention_processor import (
-    Attention,
-    FluxAttnProcessor2_0,
-    FluxSingleAttnProcessor2_0,
-)
-from diffusers.models.modeling_utils import ModelMixin
-from diffusers.models.normalization import (
-    AdaLayerNormContinuous,
-    AdaLayerNormZero,
-    AdaLayerNormZeroSingle,
-)
-from diffusers.utils import (
-    USE_PEFT_BACKEND,
-    is_torch_version,
-    logging,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
-from diffusers.utils.torch_utils import maybe_allow_in_graph
-from diffusers.models.embeddings import (
-    CombinedTimestepGuidanceTextProjEmbeddings,
-    CombinedTimestepTextProjEmbeddings,
-)
-from diffusers.models.modeling_outputs import Transformer2DModelOutput
-
-
-logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
-
-
-# YiYi to-do: refactor rope related functions/classes
-def rope(pos: torch.Tensor, dim: int, theta: int) -> torch.Tensor:
-    assert dim % 2 == 0, "The dimension must be even."
-
-    scale = torch.arange(0, dim, 2, dtype=torch.float64, device=pos.device) / dim
-    omega = 1.0 / (theta**scale)
-
-    batch_size, seq_length = pos.shape
-    out = torch.einsum("...n,d->...nd", pos, omega)
-    cos_out = torch.cos(out)
-    sin_out = torch.sin(out)
-
-    stacked_out = torch.stack([cos_out, -sin_out, sin_out, cos_out], dim=-1)
-    out = stacked_out.view(batch_size, -1, dim // 2, 2, 2)
-    return out.float()
-
-
-# YiYi to-do: refactor rope related functions/classes
-class EmbedND(nn.Module):
-    def __init__(self, dim: int, theta: int, axes_dim: List[int]):
-        super().__init__()
-        self.dim = dim
-        self.theta = theta
-        self.axes_dim = axes_dim
-
-    def forward(self, ids: torch.Tensor) -> torch.Tensor:
-        n_axes = ids.shape[-1]
-        emb = torch.cat(
-            [rope(ids[..., i], self.axes_dim[i], self.theta) for i in range(n_axes)],
-            dim=-3,
-        )
-        return emb.unsqueeze(1)
-
-
-@maybe_allow_in_graph
-class FluxSingleTransformerBlock(nn.Module):
-    r"""
-    A Transformer block following the MMDiT architecture, introduced in Stable Diffusion 3.
-
-    Reference: https://arxiv.org/abs/2403.03206
-
-    Parameters:
-        dim (`int`): The number of channels in the input and output.
-        num_attention_heads (`int`): The number of heads to use for multi-head attention.
-        attention_head_dim (`int`): The number of channels in each head.
-        context_pre_only (`bool`): Boolean to determine if we should add some blocks associated with the
-            processing of `context` conditions.
-    """
-
-    def __init__(self, dim, num_attention_heads, attention_head_dim, mlp_ratio=4.0):
-        super().__init__()
-        self.mlp_hidden_dim = int(dim * mlp_ratio)
-
-        self.norm = AdaLayerNormZeroSingle(dim)
-        self.proj_mlp = nn.Linear(dim, self.mlp_hidden_dim)
-        self.act_mlp = nn.GELU(approximate="tanh")
-        self.proj_out = nn.Linear(dim + self.mlp_hidden_dim, dim)
-
-        processor = FluxSingleAttnProcessor2_0()
-        self.attn = Attention(
-            query_dim=dim,
-            cross_attention_dim=None,
-            dim_head=attention_head_dim,
-            heads=num_attention_heads,
-            out_dim=dim,
-            bias=True,
-            processor=processor,
-            qk_norm="rms_norm",
-            eps=1e-6,
-            pre_only=True,
-        )
-
-    def forward(
-        self,
-        hidden_states: torch.FloatTensor,
-        temb: torch.FloatTensor,
-        image_rotary_emb=None,
-    ):
-        residual = hidden_states
-        norm_hidden_states, gate = self.norm(hidden_states, emb=temb)
-        mlp_hidden_states = self.act_mlp(self.proj_mlp(norm_hidden_states))
-
-        attn_output = self.attn(
-            hidden_states=norm_hidden_states,
-            image_rotary_emb=image_rotary_emb,
-        )
-
-        hidden_states = torch.cat([attn_output, mlp_hidden_states], dim=2)
-        gate = gate.unsqueeze(1)
-        hidden_states = gate * self.proj_out(hidden_states)
-        hidden_states = residual + hidden_states
-        if hidden_states.dtype == torch.float16:
-            hidden_states = hidden_states.clip(-65504, 65504)
-
-        return hidden_states
-
-
-@maybe_allow_in_graph
-class FluxTransformerBlock(nn.Module):
-    r"""
-    A Transformer block following the MMDiT architecture, introduced in Stable Diffusion 3.
-
-    Reference: https://arxiv.org/abs/2403.03206
-
-    Parameters:
-        dim (`int`): The number of channels in the input and output.
-        num_attention_heads (`int`): The number of heads to use for multi-head attention.
-        attention_head_dim (`int`): The number of channels in each head.
-        context_pre_only (`bool`): Boolean to determine if we should add some blocks associated with the
-            processing of `context` conditions.
-    """
-
-    def __init__(
-        self, dim, num_attention_heads, attention_head_dim, qk_norm="rms_norm", eps=1e-6
-    ):
-        super().__init__()
-
-        self.norm1 = AdaLayerNormZero(dim)
-
-        self.norm1_context = AdaLayerNormZero(dim)
-
-        if hasattr(F, "scaled_dot_product_attention"):
-            processor = FluxAttnProcessor2_0()
-        else:
-            raise ValueError(
-                "The current PyTorch version does not support the `scaled_dot_product_attention` function."
-            )
-        self.attn = Attention(
-            query_dim=dim,
-            cross_attention_dim=None,
-            added_kv_proj_dim=dim,
-            dim_head=attention_head_dim,
-            heads=num_attention_heads,
-            out_dim=dim,
-            context_pre_only=False,
-            bias=True,
-            processor=processor,
-            qk_norm=qk_norm,
-            eps=eps,
-        )
-
-        self.norm2 = nn.LayerNorm(dim, elementwise_affine=False, eps=1e-6)
-        self.ff = FeedForward(dim=dim, dim_out=dim, activation_fn="gelu-approximate")
-
-        self.norm2_context = nn.LayerNorm(dim, elementwise_affine=False, eps=1e-6)
-        self.ff_context = FeedForward(
-            dim=dim, dim_out=dim, activation_fn="gelu-approximate"
-        )
-
-        # let chunk size default to None
-        self._chunk_size = None
-        self._chunk_dim = 0
-
-    def forward(
-        self,
-        hidden_states: torch.FloatTensor,
-        encoder_hidden_states: torch.FloatTensor,
-        temb: torch.FloatTensor,
-        image_rotary_emb=None,
-    ):
-        norm_hidden_states, gate_msa, shift_mlp, scale_mlp, gate_mlp = self.norm1(
-            hidden_states, emb=temb
-        )
-
-        (
-            norm_encoder_hidden_states,
-            c_gate_msa,
-            c_shift_mlp,
-            c_scale_mlp,
-            c_gate_mlp,
-        ) = self.norm1_context(encoder_hidden_states, emb=temb)
-
-        # Attention.
-        attn_output, context_attn_output = self.attn(
-            hidden_states=norm_hidden_states,
-            encoder_hidden_states=norm_encoder_hidden_states,
-            image_rotary_emb=image_rotary_emb,
-        )
-
-        # Process attention outputs for the `hidden_states`.
-        attn_output = gate_msa.unsqueeze(1) * attn_output
-        hidden_states = hidden_states + attn_output
-
-        norm_hidden_states = self.norm2(hidden_states)
-        norm_hidden_states = (
-            norm_hidden_states * (1 + scale_mlp[:, None]) + shift_mlp[:, None]
-        )
-
-        ff_output = self.ff(norm_hidden_states)
-        ff_output = gate_mlp.unsqueeze(1) * ff_output
-
-        hidden_states = hidden_states + ff_output
-
-        # Process attention outputs for the `encoder_hidden_states`.
-
-        context_attn_output = c_gate_msa.unsqueeze(1) * context_attn_output
-        encoder_hidden_states = encoder_hidden_states + context_attn_output
-
-        norm_encoder_hidden_states = self.norm2_context(encoder_hidden_states)
-        norm_encoder_hidden_states = (
-            norm_encoder_hidden_states * (1 + c_scale_mlp[:, None])
-            + c_shift_mlp[:, None]
-        )
-
-        context_ff_output = self.ff_context(norm_encoder_hidden_states)
-        encoder_hidden_states = (
-            encoder_hidden_states + c_gate_mlp.unsqueeze(1) * context_ff_output
-        )
-        if encoder_hidden_states.dtype == torch.float16:
-            encoder_hidden_states = encoder_hidden_states.clip(-65504, 65504)
-
-        return encoder_hidden_states, hidden_states
-
-
-class FluxTransformer2DModel(
-    ModelMixin, ConfigMixin, PeftAdapterMixin, FromOriginalModelMixin
-):
-    """
-    The Transformer model introduced in Flux.
-
-    Reference: https://blackforestlabs.ai/announcing-black-forest-labs/
-
-    Parameters:
-        patch_size (`int`): Patch size to turn the input data into small patches.
-        in_channels (`int`, *optional*, defaults to 16): The number of channels in the input.
-        num_layers (`int`, *optional*, defaults to 18): The number of layers of MMDiT blocks to use.
-        num_single_layers (`int`, *optional*, defaults to 18): The number of layers of single DiT blocks to use.
-        attention_head_dim (`int`, *optional*, defaults to 64): The number of channels in each head.
-        num_attention_heads (`int`, *optional*, defaults to 18): The number of heads to use for multi-head attention.
-        joint_attention_dim (`int`, *optional*): The number of `encoder_hidden_states` dimensions to use.
-        pooled_projection_dim (`int`): Number of dimensions to use when projecting the `pooled_projections`.
-        guidance_embeds (`bool`, defaults to False): Whether to use guidance embeddings.
-    """
-
-    _supports_gradient_checkpointing = True
-
-    @register_to_config
-    def __init__(
-        self,
-        patch_size: int = 1,
-        in_channels: int = 64,
-        num_layers: int = 19,
-        num_single_layers: int = 38,
-        attention_head_dim: int = 128,
-        num_attention_heads: int = 24,
-        joint_attention_dim: int = 4096,
-        pooled_projection_dim: int = 768,
-        guidance_embeds: bool = False,
-        axes_dims_rope: List[int] = [16, 56, 56],
-    ):
-        super().__init__()
-        self.out_channels = in_channels
-        self.inner_dim = (
-            self.config.num_attention_heads * self.config.attention_head_dim
-        )
-
-        self.pos_embed = EmbedND(
-            dim=self.inner_dim, theta=10000, axes_dim=axes_dims_rope
-        )
-        text_time_guidance_cls = (
-            CombinedTimestepGuidanceTextProjEmbeddings
-            if guidance_embeds
-            else CombinedTimestepTextProjEmbeddings
-        )
-        self.time_text_embed = text_time_guidance_cls(
-            embedding_dim=self.inner_dim,
-            pooled_projection_dim=self.config.pooled_projection_dim,
-        )
-
-        self.context_embedder = nn.Linear(
-            self.config.joint_attention_dim, self.inner_dim
-        )
-        self.x_embedder = torch.nn.Linear(self.config.in_channels, self.inner_dim)
-
-        self.transformer_blocks = nn.ModuleList(
-            [
-                FluxTransformerBlock(
-                    dim=self.inner_dim,
-                    num_attention_heads=self.config.num_attention_heads,
-                    attention_head_dim=self.config.attention_head_dim,
-                )
-                for i in range(self.config.num_layers)
-            ]
-        )
-
-        self.single_transformer_blocks = nn.ModuleList(
-            [
-                FluxSingleTransformerBlock(
-                    dim=self.inner_dim,
-                    num_attention_heads=self.config.num_attention_heads,
-                    attention_head_dim=self.config.attention_head_dim,
-                )
-                for i in range(self.config.num_single_layers)
-            ]
-        )
-
-        self.norm_out = AdaLayerNormContinuous(
-            self.inner_dim, self.inner_dim, elementwise_affine=False, eps=1e-6
-        )
-        self.proj_out = nn.Linear(
-            self.inner_dim, patch_size * patch_size * self.out_channels, bias=True
-        )
-
-        self.gradient_checkpointing = False
-
-    def _set_gradient_checkpointing(self, module, value=False):
-        if hasattr(module, "gradient_checkpointing"):
-            module.gradient_checkpointing = value
-
-    def forward(
-        self,
-        hidden_states: torch.Tensor,
-        encoder_hidden_states: torch.Tensor = None,
-        pooled_projections: torch.Tensor = None,
-        timestep: torch.LongTensor = None,
-        img_ids: torch.Tensor = None,
-        txt_ids: torch.Tensor = None,
-        guidance: torch.Tensor = None,
-        joint_attention_kwargs: Optional[Dict[str, Any]] = None,
-        controlnet_block_samples=None,
-        controlnet_single_block_samples=None,
-        return_dict: bool = True,
-    ) -> Union[torch.FloatTensor, Transformer2DModelOutput]:
-        """
-        The [`FluxTransformer2DModel`] forward method.
-
-        Args:
-            hidden_states (`torch.FloatTensor` of shape `(batch size, channel, height, width)`):
-                Input `hidden_states`.
-            encoder_hidden_states (`torch.FloatTensor` of shape `(batch size, sequence_len, embed_dims)`):
-                Conditional embeddings (embeddings computed from the input conditions such as prompts) to use.
-            pooled_projections (`torch.FloatTensor` of shape `(batch_size, projection_dim)`): Embeddings projected
-                from the embeddings of input conditions.
-            timestep ( `torch.LongTensor`):
-                Used to indicate denoising step.
-            block_controlnet_hidden_states: (`list` of `torch.Tensor`):
-                A list of tensors that if specified are added to the residuals of transformer blocks.
-            joint_attention_kwargs (`dict`, *optional*):
-                A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
-                `self.processor` in
-                [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
-            return_dict (`bool`, *optional*, defaults to `True`):
-                Whether or not to return a [`~models.transformer_2d.Transformer2DModelOutput`] instead of a plain
-                tuple.
-
-        Returns:
-            If `return_dict` is True, an [`~models.transformer_2d.Transformer2DModelOutput`] is returned, otherwise a
-            `tuple` where the first element is the sample tensor.
-        """
-        if joint_attention_kwargs is not None:
-            joint_attention_kwargs = joint_attention_kwargs.copy()
-            lora_scale = joint_attention_kwargs.pop("scale", 1.0)
-        else:
-            lora_scale = 1.0
-
-        if USE_PEFT_BACKEND:
-            # weight the lora layers by setting `lora_scale` for each PEFT layer
-            scale_lora_layers(self, lora_scale)
-        else:
-            if (
-                joint_attention_kwargs is not None
-                and joint_attention_kwargs.get("scale", None) is not None
-            ):
-                logger.warning(
-                    "Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective."
-                )
-        hidden_states = self.x_embedder(hidden_states)
-
-        timestep = timestep.to(hidden_states.dtype) * 1000
-        if guidance is not None:
-            guidance = guidance.to(hidden_states.dtype) * 1000
-        else:
-            guidance = None
-        temb = (
-            self.time_text_embed(timestep, pooled_projections)
-            if guidance is None
-            else self.time_text_embed(timestep, guidance, pooled_projections)
-        )
-        encoder_hidden_states = self.context_embedder(encoder_hidden_states)
-
-        txt_ids = txt_ids.expand(img_ids.size(0), -1, -1)
-        ids = torch.cat((txt_ids, img_ids), dim=1)
-        image_rotary_emb = self.pos_embed(ids)
-
-        for index_block, block in enumerate(self.transformer_blocks):
-            if self.training and self.gradient_checkpointing:
-
-                def create_custom_forward(module, return_dict=None):
-                    def custom_forward(*inputs):
-                        if return_dict is not None:
-                            return module(*inputs, return_dict=return_dict)
-                        else:
-                            return module(*inputs)
-
-                    return custom_forward
-
-                ckpt_kwargs: Dict[str, Any] = (
-                    {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
-                )
-                (
-                    encoder_hidden_states,
-                    hidden_states,
-                ) = torch.utils.checkpoint.checkpoint(
-                    create_custom_forward(block),
-                    hidden_states,
-                    encoder_hidden_states,
-                    temb,
-                    image_rotary_emb,
-                    **ckpt_kwargs,
-                )
-
-            else:
-                encoder_hidden_states, hidden_states = block(
-                    hidden_states=hidden_states,
-                    encoder_hidden_states=encoder_hidden_states,
-                    temb=temb,
-                    image_rotary_emb=image_rotary_emb,
-                )
-
-            # controlnet residual
-            if controlnet_block_samples is not None:
-                interval_control = len(self.transformer_blocks) / len(
-                    controlnet_block_samples
-                )
-                interval_control = int(np.ceil(interval_control))
-                hidden_states = (
-                    hidden_states
-                    + controlnet_block_samples[index_block // interval_control]
-                )
-
-        hidden_states = torch.cat([encoder_hidden_states, hidden_states], dim=1)
-
-        for index_block, block in enumerate(self.single_transformer_blocks):
-            if self.training and self.gradient_checkpointing:
-
-                def create_custom_forward(module, return_dict=None):
-                    def custom_forward(*inputs):
-                        if return_dict is not None:
-                            return module(*inputs, return_dict=return_dict)
-                        else:
-                            return module(*inputs)
-
-                    return custom_forward
-
-                ckpt_kwargs: Dict[str, Any] = (
-                    {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
-                )
-                hidden_states = torch.utils.checkpoint.checkpoint(
-                    create_custom_forward(block),
-                    hidden_states,
-                    temb,
-                    image_rotary_emb,
-                    **ckpt_kwargs,
-                )
-
-            else:
-                hidden_states = block(
-                    hidden_states=hidden_states,
-                    temb=temb,
-                    image_rotary_emb=image_rotary_emb,
-                )
-
-            # controlnet residual
-            if controlnet_single_block_samples is not None:
-                interval_control = len(self.single_transformer_blocks) / len(
-                    controlnet_single_block_samples
-                )
-                interval_control = int(np.ceil(interval_control))
-                hidden_states[:, encoder_hidden_states.shape[1] :, ...] = (
-                    hidden_states[:, encoder_hidden_states.shape[1] :, ...]
-                    + controlnet_single_block_samples[index_block // interval_control]
-                )
-
-        hidden_states = hidden_states[:, encoder_hidden_states.shape[1] :, ...]
-
-        hidden_states = self.norm_out(hidden_states, temb)
-        output = self.proj_out(hidden_states)
-
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
-        if not return_dict:
-            return (output,)
-
-        return Transformer2DModelOutput(sample=output)