init

README.md

@@ -0,0 +1,28 @@
+# MVDream-hf
+
+modified from https://github.com/KokeCacao/mvdream-hf.
+
+### convert weights
+```bash
+# download original ckpt
+wget https://huggingface.co/MVDream/MVDream/resolve/main/sd-v2.1-base-4view.pt
+wget https://raw.githubusercontent.com/bytedance/MVDream/main/mvdream/configs/sd-v2-base.yaml
+
+# convert
+python convert_mvdream_to_diffusers.py --checkpoint_path ./sd-v2.1-base-4view.pt --dump_path ./weights --original_config_file ./sd-v2-base.yaml --half --to_safetensors --test
+```
+
+### run pipeline
+```python
+import torch
+import kiui
+from mvdream.pipeline_mvdream import MVDreamStableDiffusionPipeline
+
+pipe = MVDreamStableDiffusionPipeline.from_pretrained('./weights', torch_dtype=torch.float16)
+pipe = pipe.to("cuda")
+
+prompt = "a photo of an astronaut riding a horse on mars"
+image = pipe(prompt) # np.ndarray [4, 256, 256, 3]
+
+kiui.vis.plot_image(image)
+```

convert_mvdream_to_diffusers.py

@@ -0,0 +1,408 @@
+# Modified from https://github.com/huggingface/diffusers/blob/bc691231360a4cbc7d19a58742ebb8ed0f05e027/scripts/convert_original_stable_diffusion_to_diffusers.py
+
+import argparse
+import torch
+import sys
+
+sys.path.insert(0, '.')
+
+from diffusers.models import (
+    AutoencoderKL,
+)
+from omegaconf import OmegaConf
+from diffusers.schedulers import DDIMScheduler
+from diffusers.utils import logging
+from typing import Any
+from accelerate import init_empty_weights
+from accelerate.utils import set_module_tensor_to_device
+from mvdream.models import MultiViewUNetWrapperModel
+from mvdream.pipeline_mvdream import MVDreamStableDiffusionPipeline
+from transformers import CLIPTokenizer, CLIPTextModel
+
+logger = logging.get_logger(__name__)
+
+
+def assign_to_checkpoint(paths, checkpoint, old_checkpoint, attention_paths_to_split=None, additional_replacements=None, config=None):
+    """
+    This does the final conversion step: take locally converted weights and apply a global renaming to them. It splits
+    attention layers, and takes into account additional replacements that may arise.
+    Assigns the weights to the new checkpoint.
+    """
+    assert isinstance(paths, list), "Paths should be a list of dicts containing 'old' and 'new' keys."
+
+    # Splits the attention layers into three variables.
+    if attention_paths_to_split is not None:
+        for path, path_map in attention_paths_to_split.items():
+            old_tensor = old_checkpoint[path]
+            channels = old_tensor.shape[0] // 3
+
+            target_shape = (-1, channels) if len(old_tensor.shape) == 3 else (-1)
+
+            assert config is not None
+            num_heads = old_tensor.shape[0] // config["num_head_channels"] // 3
+
+            old_tensor = old_tensor.reshape((num_heads, 3 * channels // num_heads) + old_tensor.shape[1:])
+            query, key, value = old_tensor.split(channels // num_heads, dim=1)
+
+            checkpoint[path_map["query"]] = query.reshape(target_shape)
+            checkpoint[path_map["key"]] = key.reshape(target_shape)
+            checkpoint[path_map["value"]] = value.reshape(target_shape)
+
+    for path in paths:
+        new_path = path["new"]
+
+        # These have already been assigned
+        if attention_paths_to_split is not None and new_path in attention_paths_to_split:
+            continue
+
+        # Global renaming happens here
+        new_path = new_path.replace("middle_block.0", "mid_block.resnets.0")
+        new_path = new_path.replace("middle_block.1", "mid_block.attentions.0")
+        new_path = new_path.replace("middle_block.2", "mid_block.resnets.1")
+
+        if additional_replacements is not None:
+            for replacement in additional_replacements:
+                new_path = new_path.replace(replacement["old"], replacement["new"])
+
+        # proj_attn.weight has to be converted from conv 1D to linear
+        is_attn_weight = "proj_attn.weight" in new_path or ("attentions" in new_path and "to_" in new_path)
+        shape = old_checkpoint[path["old"]].shape
+        if is_attn_weight and len(shape) == 3:
+            checkpoint[new_path] = old_checkpoint[path["old"]][:, :, 0]
+        elif is_attn_weight and len(shape) == 4:
+            checkpoint[new_path] = old_checkpoint[path["old"]][:, :, 0, 0]
+        else:
+            checkpoint[new_path] = old_checkpoint[path["old"]]
+
+
+def shave_segments(path, n_shave_prefix_segments=1):
+    """
+    Removes segments. Positive values shave the first segments, negative shave the last segments.
+    """
+    if n_shave_prefix_segments >= 0:
+        return ".".join(path.split(".")[n_shave_prefix_segments:])
+    else:
+        return ".".join(path.split(".")[:n_shave_prefix_segments])
+
+
+def create_vae_diffusers_config(original_config, image_size: int):
+    """
+    Creates a config for the diffusers based on the config of the LDM model.
+    """
+    vae_params = original_config.model.params.first_stage_config.params.ddconfig
+    _ = original_config.model.params.first_stage_config.params.embed_dim
+
+    block_out_channels = [vae_params.ch * mult for mult in vae_params.ch_mult]
+    down_block_types = ["DownEncoderBlock2D"] * len(block_out_channels)
+    up_block_types = ["UpDecoderBlock2D"] * len(block_out_channels)
+
+    config = {
+        "sample_size": image_size,
+        "in_channels": vae_params.in_channels,
+        "out_channels": vae_params.out_ch,
+        "down_block_types": tuple(down_block_types),
+        "up_block_types": tuple(up_block_types),
+        "block_out_channels": tuple(block_out_channels),
+        "latent_channels": vae_params.z_channels,
+        "layers_per_block": vae_params.num_res_blocks,
+    }
+    return config
+
+
+def convert_ldm_vae_checkpoint(checkpoint, config):
+    # extract state dict for VAE
+    vae_state_dict = {}
+    vae_key = "first_stage_model."
+    keys = list(checkpoint.keys())
+    for key in keys:
+        if key.startswith(vae_key):
+            vae_state_dict[key.replace(vae_key, "")] = checkpoint.get(key)
+
+    new_checkpoint = {}
+
+    new_checkpoint["encoder.conv_in.weight"] = vae_state_dict["encoder.conv_in.weight"]
+    new_checkpoint["encoder.conv_in.bias"] = vae_state_dict["encoder.conv_in.bias"]
+    new_checkpoint["encoder.conv_out.weight"] = vae_state_dict["encoder.conv_out.weight"]
+    new_checkpoint["encoder.conv_out.bias"] = vae_state_dict["encoder.conv_out.bias"]
+    new_checkpoint["encoder.conv_norm_out.weight"] = vae_state_dict["encoder.norm_out.weight"]
+    new_checkpoint["encoder.conv_norm_out.bias"] = vae_state_dict["encoder.norm_out.bias"]
+
+    new_checkpoint["decoder.conv_in.weight"] = vae_state_dict["decoder.conv_in.weight"]
+    new_checkpoint["decoder.conv_in.bias"] = vae_state_dict["decoder.conv_in.bias"]
+    new_checkpoint["decoder.conv_out.weight"] = vae_state_dict["decoder.conv_out.weight"]
+    new_checkpoint["decoder.conv_out.bias"] = vae_state_dict["decoder.conv_out.bias"]
+    new_checkpoint["decoder.conv_norm_out.weight"] = vae_state_dict["decoder.norm_out.weight"]
+    new_checkpoint["decoder.conv_norm_out.bias"] = vae_state_dict["decoder.norm_out.bias"]
+
+    new_checkpoint["quant_conv.weight"] = vae_state_dict["quant_conv.weight"]
+    new_checkpoint["quant_conv.bias"] = vae_state_dict["quant_conv.bias"]
+    new_checkpoint["post_quant_conv.weight"] = vae_state_dict["post_quant_conv.weight"]
+    new_checkpoint["post_quant_conv.bias"] = vae_state_dict["post_quant_conv.bias"]
+
+    # Retrieves the keys for the encoder down blocks only
+    num_down_blocks = len({".".join(layer.split(".")[:3]) for layer in vae_state_dict if "encoder.down" in layer})
+    down_blocks = {layer_id: [key for key in vae_state_dict if f"down.{layer_id}" in key] for layer_id in range(num_down_blocks)}
+
+    # Retrieves the keys for the decoder up blocks only
+    num_up_blocks = len({".".join(layer.split(".")[:3]) for layer in vae_state_dict if "decoder.up" in layer})
+    up_blocks = {layer_id: [key for key in vae_state_dict if f"up.{layer_id}" in key] for layer_id in range(num_up_blocks)}
+
+    for i in range(num_down_blocks):
+        resnets = [key for key in down_blocks[i] if f"down.{i}" in key and f"down.{i}.downsample" not in key]
+
+        if f"encoder.down.{i}.downsample.conv.weight" in vae_state_dict:
+            new_checkpoint[f"encoder.down_blocks.{i}.downsamplers.0.conv.weight"] = vae_state_dict.pop(f"encoder.down.{i}.downsample.conv.weight")
+            new_checkpoint[f"encoder.down_blocks.{i}.downsamplers.0.conv.bias"] = vae_state_dict.pop(f"encoder.down.{i}.downsample.conv.bias")
+
+        paths = renew_vae_resnet_paths(resnets)
+        meta_path = {"old": f"down.{i}.block", "new": f"down_blocks.{i}.resnets"}
+        assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
+
+    mid_resnets = [key for key in vae_state_dict if "encoder.mid.block" in key]
+    num_mid_res_blocks = 2
+    for i in range(1, num_mid_res_blocks + 1):
+        resnets = [key for key in mid_resnets if f"encoder.mid.block_{i}" in key]
+
+        paths = renew_vae_resnet_paths(resnets)
+        meta_path = {"old": f"mid.block_{i}", "new": f"mid_block.resnets.{i - 1}"}
+        assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
+
+    mid_attentions = [key for key in vae_state_dict if "encoder.mid.attn" in key]
+    paths = renew_vae_attention_paths(mid_attentions)
+    meta_path = {"old": "mid.attn_1", "new": "mid_block.attentions.0"}
+    assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
+    conv_attn_to_linear(new_checkpoint)
+
+    for i in range(num_up_blocks):
+        block_id = num_up_blocks - 1 - i
+        resnets = [key for key in up_blocks[block_id] if f"up.{block_id}" in key and f"up.{block_id}.upsample" not in key]
+
+        if f"decoder.up.{block_id}.upsample.conv.weight" in vae_state_dict:
+            new_checkpoint[f"decoder.up_blocks.{i}.upsamplers.0.conv.weight"] = vae_state_dict[f"decoder.up.{block_id}.upsample.conv.weight"]
+            new_checkpoint[f"decoder.up_blocks.{i}.upsamplers.0.conv.bias"] = vae_state_dict[f"decoder.up.{block_id}.upsample.conv.bias"]
+
+        paths = renew_vae_resnet_paths(resnets)
+        meta_path = {"old": f"up.{block_id}.block", "new": f"up_blocks.{i}.resnets"}
+        assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
+
+    mid_resnets = [key for key in vae_state_dict if "decoder.mid.block" in key]
+    num_mid_res_blocks = 2
+    for i in range(1, num_mid_res_blocks + 1):
+        resnets = [key for key in mid_resnets if f"decoder.mid.block_{i}" in key]
+
+        paths = renew_vae_resnet_paths(resnets)
+        meta_path = {"old": f"mid.block_{i}", "new": f"mid_block.resnets.{i - 1}"}
+        assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
+
+    mid_attentions = [key for key in vae_state_dict if "decoder.mid.attn" in key]
+    paths = renew_vae_attention_paths(mid_attentions)
+    meta_path = {"old": "mid.attn_1", "new": "mid_block.attentions.0"}
+    assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
+    conv_attn_to_linear(new_checkpoint)
+    return new_checkpoint
+
+
+def renew_vae_resnet_paths(old_list, n_shave_prefix_segments=0):
+    """
+    Updates paths inside resnets to the new naming scheme (local renaming)
+    """
+    mapping = []
+    for old_item in old_list:
+        new_item = old_item
+
+        new_item = new_item.replace("nin_shortcut", "conv_shortcut")
+        new_item = shave_segments(new_item, n_shave_prefix_segments=n_shave_prefix_segments)
+
+        mapping.append({"old": old_item, "new": new_item})
+
+    return mapping
+
+
+def renew_vae_attention_paths(old_list, n_shave_prefix_segments=0):
+    """
+    Updates paths inside attentions to the new naming scheme (local renaming)
+    """
+    mapping = []
+    for old_item in old_list:
+        new_item = old_item
+
+        new_item = new_item.replace("norm.weight", "group_norm.weight")
+        new_item = new_item.replace("norm.bias", "group_norm.bias")
+
+        new_item = new_item.replace("q.weight", "to_q.weight")
+        new_item = new_item.replace("q.bias", "to_q.bias")
+
+        new_item = new_item.replace("k.weight", "to_k.weight")
+        new_item = new_item.replace("k.bias", "to_k.bias")
+
+        new_item = new_item.replace("v.weight", "to_v.weight")
+        new_item = new_item.replace("v.bias", "to_v.bias")
+
+        new_item = new_item.replace("proj_out.weight", "to_out.0.weight")
+        new_item = new_item.replace("proj_out.bias", "to_out.0.bias")
+
+        new_item = shave_segments(new_item, n_shave_prefix_segments=n_shave_prefix_segments)
+
+        mapping.append({"old": old_item, "new": new_item})
+
+    return mapping
+
+
+def conv_attn_to_linear(checkpoint):
+    keys = list(checkpoint.keys())
+    attn_keys = ["query.weight", "key.weight", "value.weight"]
+    for key in keys:
+        if ".".join(key.split(".")[-2:]) in attn_keys:
+            if checkpoint[key].ndim > 2:
+                checkpoint[key] = checkpoint[key][:, :, 0, 0]
+        elif "proj_attn.weight" in key:
+            if checkpoint[key].ndim > 2:
+                checkpoint[key] = checkpoint[key][:, :, 0]
+
+def create_unet_config(original_config) -> Any:
+    return OmegaConf.to_container(original_config.model.params.unet_config.params, resolve=True)
+
+def convert_from_original_mvdream_ckpt(checkpoint_path, original_config_file, device):
+    checkpoint = torch.load(checkpoint_path, map_location=device)
+    # print(f"Checkpoint: {checkpoint.keys()}")
+    torch.cuda.empty_cache()
+
+    original_config = OmegaConf.load(original_config_file)
+    # print(f"Original Config: {original_config}")
+    prediction_type = "epsilon"
+    image_size = 256
+    num_train_timesteps = getattr(original_config.model.params, "timesteps", None) or 1000
+    beta_start = getattr(original_config.model.params, "linear_start", None) or 0.02
+    beta_end = getattr(original_config.model.params, "linear_end", None) or 0.085
+    scheduler = DDIMScheduler(
+        beta_end=beta_end,
+        beta_schedule="scaled_linear",
+        beta_start=beta_start,
+        num_train_timesteps=num_train_timesteps,
+        steps_offset=1,
+        clip_sample=False,
+        set_alpha_to_one=False,
+        prediction_type=prediction_type,
+    )
+    scheduler.register_to_config(clip_sample=False)
+
+    # Convert the UNet2DConditionModel model.
+    # upcast_attention = None
+    # unet_config = create_unet_diffusers_config(original_config, image_size=image_size)
+    # unet_config["upcast_attention"] = upcast_attention
+    # with init_empty_weights():
+    #     unet = UNet2DConditionModel(**unet_config)
+    # converted_unet_checkpoint = convert_ldm_unet_checkpoint(
+    #     checkpoint, unet_config, path=None, extract_ema=extract_ema
+    # )
+    # print(f"Unet Config: {original_config.model.params.unet_config.params}")
+    unet_config = create_unet_config(original_config)
+    unet: MultiViewUNetWrapperModel = MultiViewUNetWrapperModel(**unet_config)
+    unet.register_to_config(**unet_config)
+    # print(f"Unet State Dict: {unet.state_dict().keys()}")
+    unet.load_state_dict({key.replace("model.diffusion_model.", "unet."): value for key, value in checkpoint.items() if key.replace("model.diffusion_model.", "unet.") in unet.state_dict()})
+    for param_name, param in unet.state_dict().items():
+        set_module_tensor_to_device(unet, param_name, device=device, value=param)
+
+    # Convert the VAE model.
+    vae_config = create_vae_diffusers_config(original_config, image_size=image_size)
+    converted_vae_checkpoint = convert_ldm_vae_checkpoint(checkpoint, vae_config)
+
+    if ("model" in original_config and "params" in original_config.model and "scale_factor" in original_config.model.params):
+        vae_scaling_factor = original_config.model.params.scale_factor
+    else:
+        vae_scaling_factor = 0.18215 # default SD scaling factor
+
+    vae_config["scaling_factor"] = vae_scaling_factor
+
+    with init_empty_weights():
+        vae = AutoencoderKL(**vae_config)
+
+    for param_name, param in converted_vae_checkpoint.items():
+        set_module_tensor_to_device(vae, param_name, device=device, value=param)
+
+    if original_config.model.params.unet_config.params.context_dim == 768:
+        tokenizer: CLIPTokenizer = CLIPTokenizer.from_pretrained("openai/clip-vit-large-patch14")
+        text_encoder: CLIPTextModel = CLIPTextModel.from_pretrained("openai/clip-vit-large-patch14").to(device=device) # type: ignore
+    elif original_config.model.params.unet_config.params.context_dim == 1024:
+        tokenizer: CLIPTokenizer = CLIPTokenizer.from_pretrained("stabilityai/stable-diffusion-2-1", subfolder="tokenizer")
+        text_encoder: CLIPTextModel = CLIPTextModel.from_pretrained("stabilityai/stable-diffusion-2-1", subfolder="text_encoder").to(device=device) # type: ignore
+    else:
+        raise ValueError(f"Unknown context_dim: {original_config.model.paams.unet_config.params.context_dim}")
+
+    pipe = MVDreamStableDiffusionPipeline(
+        vae=vae,
+        unet=unet,
+        tokenizer=tokenizer,
+        text_encoder=text_encoder,
+        scheduler=scheduler,
+    )
+
+    return pipe
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument("--checkpoint_path", default=None, type=str, required=True, help="Path to the checkpoint to convert.")
+    parser.add_argument(
+        "--original_config_file",
+        default=None,
+        type=str,
+        help="The YAML config file corresponding to the original architecture.",
+    )
+    parser.add_argument(
+        "--to_safetensors",
+        action="store_true",
+        help="Whether to store pipeline in safetensors format or not.",
+    )
+    parser.add_argument("--half", action="store_true", help="Save weights in half precision.")
+    parser.add_argument("--test", action="store_true", help="Whether to test inference after convertion.")
+    parser.add_argument("--dump_path", default=None, type=str, required=True, help="Path to the output model.")
+    parser.add_argument("--device", type=str, help="Device to use (e.g. cpu, cuda:0, cuda:1, etc.)")
+    args = parser.parse_args()
+    
+    args.device = torch.device(args.device if args.device is not None else "cuda" if torch.cuda.is_available() else "cpu")
+
+    pipe = convert_from_original_mvdream_ckpt(
+        checkpoint_path=args.checkpoint_path,
+        original_config_file=args.original_config_file,
+        device=args.device,
+    )
+
+    if args.half:
+        pipe.to(torch_dtype=torch.float16)
+
+    print(f"Saving pipeline to {args.dump_path}...")
+    pipe.save_pretrained(args.dump_path, safe_serialization=args.to_safetensors)
+    
+    if args.test:
+        try:
+            print(f"Testing each subcomponent of the pipeline...")
+            images = pipe(
+                prompt="Head of Hatsune Miku",
+                negative_prompt="painting, bad quality, flat",
+                output_type="pil",
+                guidance_scale=7.5,
+                num_inference_steps=50,
+                device=args.device,
+            )
+            for i, image in enumerate(images):
+                image.save(f"image_{i}.png") # type: ignore
+
+            print(f"Testing entire pipeline...")
+            loaded_pipe: MVDreamStableDiffusionPipeline = MVDreamStableDiffusionPipeline.from_pretrained(args.dump_path, safe_serialization=args.to_safetensors) # type: ignore
+            images = loaded_pipe(
+                prompt="Head of Hatsune Miku",
+                negative_prompt="painting, bad quality, flat",
+                output_type="pil",
+                guidance_scale=7.5,
+                num_inference_steps=50,
+                device=args.device,
+            )
+            for i, image in enumerate(images):
+                image.save(f"image_{i}.png") # type: ignore
+        except Exception as e:
+            print(f"Failed to test inference: {e}")
+            raise e from e
+        print("Inference test passed!")

main.py

@@ -0,0 +1,11 @@
+import torch
+import kiui
+from mvdream.pipeline_mvdream import MVDreamStableDiffusionPipeline
+
+pipe = MVDreamStableDiffusionPipeline.from_pretrained('./weights', torch_dtype=torch.float16)
+pipe = pipe.to("cuda")
+
+prompt = "a photo of an astronaut riding a horse on mars"
+image = pipe(prompt)
+
+kiui.vis.plot_image(image)

mvdream/attention.py

@@ -0,0 +1,352 @@
+# obtained and modified from https://github.com/bytedance/MVDream
+
+import math
+import torch
+import torch.nn.functional as F
+
+from inspect import isfunction
+from torch import nn, einsum
+from torch.amp.autocast_mode import autocast
+from einops import rearrange, repeat
+from typing import Optional, Any
+from .util import checkpoint
+
+try:
+    import xformers # type: ignore
+    import xformers.ops # type: ignore
+    XFORMERS_IS_AVAILBLE = True
+except:
+    XFORMERS_IS_AVAILBLE = False
+
+# CrossAttn precision handling
+import os
+
+_ATTN_PRECISION = os.environ.get("ATTN_PRECISION", "fp32")
+
+
+def uniq(arr):
+    return {el: True for el in arr}.keys()
+
+
+def default(val, d):
+    if val is not None:
+        return val
+    return d() if isfunction(d) else d
+
+
+def max_neg_value(t):
+    return -torch.finfo(t.dtype).max
+
+
+def init_(tensor):
+    dim = tensor.shape[-1]
+    std = 1 / math.sqrt(dim)
+    tensor.uniform_(-std, std)
+    return tensor
+
+
+# feedforward
+class GEGLU(nn.Module):
+
+    def __init__(self, dim_in, dim_out):
+        super().__init__()
+        self.proj = nn.Linear(dim_in, dim_out * 2)
+
+    def forward(self, x):
+        x, gate = self.proj(x).chunk(2, dim=-1)
+        return x * F.gelu(gate)
+
+
+class FeedForward(nn.Module):
+
+    def __init__(self, dim, dim_out=None, mult=4, glu=False, dropout=0.):
+        super().__init__()
+        inner_dim = int(dim * mult)
+        dim_out = default(dim_out, dim)
+        project_in = nn.Sequential(nn.Linear(dim, inner_dim), nn.GELU()) if not glu else GEGLU(dim, inner_dim)
+
+        self.net = nn.Sequential(project_in, nn.Dropout(dropout), nn.Linear(inner_dim, dim_out))
+
+    def forward(self, x):
+        return self.net(x)
+
+
+def zero_module(module):
+    """
+    Zero out the parameters of a module and return it.
+    """
+    for p in module.parameters():
+        p.detach().zero_()
+    return module
+
+
+def Normalize(in_channels):
+    return torch.nn.GroupNorm(num_groups=32, num_channels=in_channels, eps=1e-6, affine=True)
+
+
+class SpatialSelfAttention(nn.Module):
+
+    def __init__(self, in_channels):
+        super().__init__()
+        self.in_channels = in_channels
+
+        self.norm = Normalize(in_channels)
+        self.q = torch.nn.Conv2d(in_channels, in_channels, kernel_size=1, stride=1, padding=0)
+        self.k = torch.nn.Conv2d(in_channels, in_channels, kernel_size=1, stride=1, padding=0)
+        self.v = torch.nn.Conv2d(in_channels, in_channels, kernel_size=1, stride=1, padding=0)
+        self.proj_out = torch.nn.Conv2d(in_channels, in_channels, kernel_size=1, stride=1, padding=0)
+
+    def forward(self, x):
+        h_ = x
+        h_ = self.norm(h_)
+        q = self.q(h_)
+        k = self.k(h_)
+        v = self.v(h_)
+
+        # compute attention
+        b, c, h, w = q.shape
+        q = rearrange(q, 'b c h w -> b (h w) c')
+        k = rearrange(k, 'b c h w -> b c (h w)')
+        w_ = torch.einsum('bij,bjk->bik', q, k)
+
+        w_ = w_ * (int(c)**(-0.5))
+        w_ = torch.nn.functional.softmax(w_, dim=2)
+
+        # attend to values
+        v = rearrange(v, 'b c h w -> b c (h w)')
+        w_ = rearrange(w_, 'b i j -> b j i')
+        h_ = torch.einsum('bij,bjk->bik', v, w_)
+        h_ = rearrange(h_, 'b c (h w) -> b c h w', h=h)
+        h_ = self.proj_out(h_)
+
+        return x + h_
+
+
+class CrossAttention(nn.Module):
+
+    def __init__(self, query_dim, context_dim=None, heads=8, dim_head=64, dropout=0.):
+        super().__init__()
+        inner_dim = dim_head * heads
+        context_dim = default(context_dim, query_dim)
+
+        self.scale = dim_head**-0.5
+        self.heads = heads
+
+        self.to_q = nn.Linear(query_dim, inner_dim, bias=False)
+        self.to_k = nn.Linear(context_dim, inner_dim, bias=False)
+        self.to_v = nn.Linear(context_dim, inner_dim, bias=False)
+
+        self.to_out = nn.Sequential(nn.Linear(inner_dim, query_dim), nn.Dropout(dropout))
+
+    def forward(self, x, context=None, mask=None):
+        h = self.heads
+
+        q = self.to_q(x)
+        context = default(context, x)
+        k = self.to_k(context)
+        v = self.to_v(context)
+
+        q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q, k, v))
+
+        # force cast to fp32 to avoid overflowing
+        if _ATTN_PRECISION == "fp32":
+            with autocast(enabled=False, device_type='cuda'):
+                q, k = q.float(), k.float()
+                sim = einsum('b i d, b j d -> b i j', q, k) * self.scale
+        else:
+            sim = einsum('b i d, b j d -> b i j', q, k) * self.scale
+
+        del q, k
+
+        if mask is not None:
+            mask = rearrange(mask, 'b ... -> b (...)')
+            max_neg_value = -torch.finfo(sim.dtype).max
+            mask = repeat(mask, 'b j -> (b h) () j', h=h)
+            sim.masked_fill_(~mask, max_neg_value)
+
+        # attention, what we cannot get enough of
+        sim = sim.softmax(dim=-1)
+
+        out = einsum('b i j, b j d -> b i d', sim, v)
+        out = rearrange(out, '(b h) n d -> b n (h d)', h=h)
+        return self.to_out(out)
+
+
+class MemoryEfficientCrossAttention(nn.Module):
+    # https://github.com/MatthieuTPHR/diffusers/blob/d80b531ff8060ec1ea982b65a1b8df70f73aa67c/src/diffusers/models/attention.py#L223
+    def __init__(self, query_dim, context_dim=None, heads=8, dim_head=64, dropout=0.0):
+        super().__init__()
+        # print(f"Setting up {self.__class__.__name__}. Query dim is {query_dim}, context_dim is {context_dim} and using {heads} heads.")
+        inner_dim = dim_head * heads
+        context_dim = default(context_dim, query_dim)
+
+        self.heads = heads
+        self.dim_head = dim_head
+
+        self.to_q = nn.Linear(query_dim, inner_dim, bias=False)
+        self.to_k = nn.Linear(context_dim, inner_dim, bias=False)
+        self.to_v = nn.Linear(context_dim, inner_dim, bias=False)
+
+        self.to_out = nn.Sequential(nn.Linear(inner_dim, query_dim), nn.Dropout(dropout))
+        self.attention_op: Optional[Any] = None
+
+    def forward(self, x, context=None, mask=None):
+        q = self.to_q(x)
+        context = default(context, x)
+        k = self.to_k(context)
+        v = self.to_v(context)
+
+        b, _, _ = q.shape
+        q, k, v = map(
+            lambda t: t.unsqueeze(3).reshape(b, t.shape[1], self.heads, self.dim_head).permute(0, 2, 1, 3).reshape(b * self.heads, t.shape[1], self.dim_head).contiguous(),
+            (q, k, v),
+        )
+
+        # actually compute the attention, what we cannot get enough of
+        out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=self.attention_op)
+
+        if mask is not None:
+            raise NotImplementedError
+        out = (out.unsqueeze(0).reshape(b, self.heads, out.shape[1], self.dim_head).permute(0, 2, 1, 3).reshape(b, out.shape[1], self.heads * self.dim_head))
+        return self.to_out(out)
+
+
+class BasicTransformerBlock(nn.Module):
+    ATTENTION_MODES = {
+        "softmax": CrossAttention, # vanilla attention
+        "softmax-xformers": MemoryEfficientCrossAttention
+    }
+
+    def __init__(self, dim, n_heads, d_head, dropout=0., context_dim=None, gated_ff=True, checkpoint=True, disable_self_attn=False):
+        super().__init__()
+        attn_mode = "softmax-xformers" if XFORMERS_IS_AVAILBLE else "softmax"
+        assert attn_mode in self.ATTENTION_MODES
+        attn_cls = self.ATTENTION_MODES[attn_mode]
+        self.disable_self_attn = disable_self_attn
+        self.attn1 = attn_cls(query_dim=dim, heads=n_heads, dim_head=d_head, dropout=dropout, context_dim=context_dim if self.disable_self_attn else None) # is a self-attention if not self.disable_self_attn
+        self.ff = FeedForward(dim, dropout=dropout, glu=gated_ff)
+        self.attn2 = attn_cls(query_dim=dim, context_dim=context_dim, heads=n_heads, dim_head=d_head, dropout=dropout) # is self-attn if context is none
+        self.norm1 = nn.LayerNorm(dim)
+        self.norm2 = nn.LayerNorm(dim)
+        self.norm3 = nn.LayerNorm(dim)
+        self.checkpoint = checkpoint
+
+    def forward(self, x, context=None):
+        return checkpoint(self._forward, (x, context), self.parameters(), self.checkpoint)
+
+    def _forward(self, x, context=None):
+        x = self.attn1(self.norm1(x), context=context if self.disable_self_attn else None) + x
+        x = self.attn2(self.norm2(x), context=context) + x
+        x = self.ff(self.norm3(x)) + x
+        return x
+
+
+class SpatialTransformer(nn.Module):
+    """
+    Transformer block for image-like data.
+    First, project the input (aka embedding)
+    and reshape to b, t, d.
+    Then apply standard transformer action.
+    Finally, reshape to image
+    NEW: use_linear for more efficiency instead of the 1x1 convs
+    """
+
+    def __init__(self, in_channels, n_heads, d_head, depth=1, dropout=0., context_dim=None, disable_self_attn=False, use_linear=False, use_checkpoint=True):
+        super().__init__()
+        assert context_dim is not None
+        if not isinstance(context_dim, list):
+            context_dim = [context_dim]
+        self.in_channels = in_channels
+        inner_dim = n_heads * d_head
+        self.norm = Normalize(in_channels)
+        if not use_linear:
+            self.proj_in = nn.Conv2d(in_channels, inner_dim, kernel_size=1, stride=1, padding=0)
+        else:
+            self.proj_in = nn.Linear(in_channels, inner_dim)
+
+        self.transformer_blocks = nn.ModuleList([BasicTransformerBlock(inner_dim, n_heads, d_head, dropout=dropout, context_dim=context_dim[d], disable_self_attn=disable_self_attn, checkpoint=use_checkpoint) for d in range(depth)])
+        if not use_linear:
+            self.proj_out = zero_module(nn.Conv2d(inner_dim, in_channels, kernel_size=1, stride=1, padding=0))
+        else:
+            self.proj_out = zero_module(nn.Linear(in_channels, inner_dim))
+        self.use_linear = use_linear
+
+    def forward(self, x, context=None):
+        # note: if no context is given, cross-attention defaults to self-attention
+        if not isinstance(context, list):
+            context = [context]
+        b, c, h, w = x.shape
+        x_in = x
+        x = self.norm(x)
+        if not self.use_linear:
+            x = self.proj_in(x)
+        x = rearrange(x, 'b c h w -> b (h w) c').contiguous()
+        if self.use_linear:
+            x = self.proj_in(x)
+        for i, block in enumerate(self.transformer_blocks):
+            x = block(x, context=context[i])
+        if self.use_linear:
+            x = self.proj_out(x)
+        x = rearrange(x, 'b (h w) c -> b c h w', h=h, w=w).contiguous()
+        if not self.use_linear:
+            x = self.proj_out(x)
+        return x + x_in
+
+
+class BasicTransformerBlock3D(BasicTransformerBlock):
+
+    def forward(self, x, context=None, num_frames=1):
+        return checkpoint(self._forward, (x, context, num_frames), self.parameters(), self.checkpoint)
+
+    def _forward(self, x, context=None, num_frames=1):
+        x = rearrange(x, "(b f) l c -> b (f l) c", f=num_frames).contiguous()
+        x = self.attn1(self.norm1(x), context=context if self.disable_self_attn else None) + x
+        x = rearrange(x, "b (f l) c -> (b f) l c", f=num_frames).contiguous()
+        x = self.attn2(self.norm2(x), context=context) + x
+        x = self.ff(self.norm3(x)) + x
+        return x
+
+
+class SpatialTransformer3D(nn.Module):
+    ''' 3D self-attention '''
+
+    def __init__(self, in_channels, n_heads, d_head, depth=1, dropout=0., context_dim=None, disable_self_attn=False, use_linear=False, use_checkpoint=True):
+        super().__init__()
+        assert context_dim is not None
+        if not isinstance(context_dim, list):
+            context_dim = [context_dim]
+        self.in_channels = in_channels
+        inner_dim = n_heads * d_head
+        self.norm = Normalize(in_channels)
+        if not use_linear:
+            self.proj_in = nn.Conv2d(in_channels, inner_dim, kernel_size=1, stride=1, padding=0)
+        else:
+            self.proj_in = nn.Linear(in_channels, inner_dim)
+
+        self.transformer_blocks = nn.ModuleList([BasicTransformerBlock3D(inner_dim, n_heads, d_head, dropout=dropout, context_dim=context_dim[d], disable_self_attn=disable_self_attn, checkpoint=use_checkpoint) for d in range(depth)])
+        if not use_linear:
+            self.proj_out = zero_module(nn.Conv2d(inner_dim, in_channels, kernel_size=1, stride=1, padding=0))
+        else:
+            self.proj_out = zero_module(nn.Linear(in_channels, inner_dim))
+        self.use_linear = use_linear
+
+    def forward(self, x, context=None, num_frames=1):
+        # note: if no context is given, cross-attention defaults to self-attention
+        if not isinstance(context, list):
+            context = [context]
+        b, c, h, w = x.shape
+        x_in = x
+        x = self.norm(x)
+        if not self.use_linear:
+            x = self.proj_in(x)
+        x = rearrange(x, 'b c h w -> b (h w) c').contiguous()
+        if self.use_linear:
+            x = self.proj_in(x)
+        for i, block in enumerate(self.transformer_blocks):
+            x = block(x, context=context[i], num_frames=num_frames)
+        if self.use_linear:
+            x = self.proj_out(x)
+        x = rearrange(x, 'b (h w) c -> b c h w', h=h, w=w).contiguous()
+        if not self.use_linear:
+            x = self.proj_out(x)
+        return x + x_in

mvdream/models.py

@@ -0,0 +1,775 @@
+# obtained and modified from https://github.com/bytedance/MVDream
+
+import math
+import numpy as np
+import torch as th
+import torch.nn as nn
+import torch.nn.functional as F
+
+from abc import abstractmethod
+from .util import (
+    checkpoint,
+    conv_nd,
+    linear,
+    avg_pool_nd,
+    zero_module,
+    normalization,
+    timestep_embedding,
+)
+from .attention import SpatialTransformer, SpatialTransformer3D
+from diffusers.configuration_utils import ConfigMixin
+from diffusers.models.modeling_utils import ModelMixin
+from typing import Any, List, Optional
+from torch import Tensor
+
+
+class MultiViewUNetWrapperModel(ModelMixin, ConfigMixin):
+
+    def __init__(self, 
+            image_size,
+            in_channels,
+            model_channels,
+            out_channels,
+            num_res_blocks,
+            attention_resolutions,
+            dropout=0,
+            channel_mult=(1, 2, 4, 8),
+            conv_resample=True,
+            dims=2,
+            num_classes=None,
+            use_checkpoint=False,
+            num_heads=-1,
+            num_head_channels=-1,
+            num_heads_upsample=-1,
+            use_scale_shift_norm=False,
+            resblock_updown=False,
+            use_new_attention_order=False,
+            use_spatial_transformer=False, # custom transformer support
+            transformer_depth=1, # custom transformer support
+            context_dim=None, # custom transformer support
+            n_embed=None, # custom support for prediction of discrete ids into codebook of first stage vq model
+            legacy=True,
+            disable_self_attentions=None,
+            num_attention_blocks=None,
+            disable_middle_self_attn=False,
+            use_linear_in_transformer=False,
+            adm_in_channels=None,
+            camera_dim=None,):
+        super().__init__()
+        self.unet: MultiViewUNetModel = MultiViewUNetModel(
+            image_size=image_size,
+            in_channels=in_channels,
+            model_channels=model_channels,
+            out_channels=out_channels,
+            num_res_blocks=num_res_blocks,
+            attention_resolutions=attention_resolutions,
+            dropout=dropout,
+            channel_mult=channel_mult,
+            conv_resample=conv_resample,
+            dims=dims,
+            num_classes=num_classes,
+            use_checkpoint=use_checkpoint,
+            num_heads=num_heads,
+            num_head_channels=num_head_channels,
+            num_heads_upsample=num_heads_upsample,
+            use_scale_shift_norm=use_scale_shift_norm,
+            resblock_updown=resblock_updown,
+            use_new_attention_order=use_new_attention_order,
+            use_spatial_transformer=use_spatial_transformer,
+            transformer_depth=transformer_depth,
+            context_dim=context_dim,
+            n_embed=n_embed,
+            legacy=legacy,
+            disable_self_attentions=disable_self_attentions,
+            num_attention_blocks=num_attention_blocks,
+            disable_middle_self_attn=disable_middle_self_attn,
+            use_linear_in_transformer=use_linear_in_transformer,
+            adm_in_channels=adm_in_channels,
+            camera_dim=camera_dim,
+        )
+
+    def forward(self, *args, **kwargs):
+        return self.unet(*args, **kwargs)
+
+
+class TimestepBlock(nn.Module):
+    """
+    Any module where forward() takes timestep embeddings as a second argument.
+    """
+
+    @abstractmethod
+    def forward(self, x, emb):
+        """
+        Apply the module to `x` given `emb` timestep embeddings.
+        """
+
+
+class TimestepEmbedSequential(nn.Sequential, TimestepBlock):
+    """
+    A sequential module that passes timestep embeddings to the children that
+    support it as an extra input.
+    """
+
+    def forward(self, x, emb, context=None, num_frames=1):
+        for layer in self:
+            if isinstance(layer, TimestepBlock):
+                x = layer(x, emb)
+            elif isinstance(layer, SpatialTransformer3D):
+                x = layer(x, context, num_frames=num_frames)
+            elif isinstance(layer, SpatialTransformer):
+                x = layer(x, context)
+            else:
+                x = layer(x)
+        return x
+
+
+class Upsample(nn.Module):
+    """
+    An upsampling layer with an optional convolution.
+    :param channels: channels in the inputs and outputs.
+    :param use_conv: a bool determining if a convolution is applied.
+    :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
+                 upsampling occurs in the inner-two dimensions.
+    """
+
+    def __init__(self, channels, use_conv, dims=2, out_channels=None, padding=1):
+        super().__init__()
+        self.channels = channels
+        self.out_channels = out_channels or channels
+        self.use_conv = use_conv
+        self.dims = dims
+        if use_conv:
+            self.conv = conv_nd(dims, self.channels, self.out_channels, 3, padding=padding)
+
+    def forward(self, x):
+        assert x.shape[1] == self.channels
+        if self.dims == 3:
+            x = F.interpolate(x, (x.shape[2], x.shape[3] * 2, x.shape[4] * 2), mode="nearest")
+        else:
+            x = F.interpolate(x, scale_factor=2, mode="nearest")
+        if self.use_conv:
+            x = self.conv(x)
+        return x
+
+
+class Downsample(nn.Module):
+    """
+    A downsampling layer with an optional convolution.
+    :param channels: channels in the inputs and outputs.
+    :param use_conv: a bool determining if a convolution is applied.
+    :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
+                 downsampling occurs in the inner-two dimensions.
+    """
+
+    def __init__(self, channels, use_conv, dims=2, out_channels=None, padding=1):
+        super().__init__()
+        self.channels = channels
+        self.out_channels = out_channels or channels
+        self.use_conv = use_conv
+        self.dims = dims
+        stride = 2 if dims != 3 else (1, 2, 2)
+        if use_conv:
+            self.op = conv_nd(dims, self.channels, self.out_channels, 3, stride=stride, padding=padding)
+        else:
+            assert self.channels == self.out_channels
+            self.op = avg_pool_nd(dims, kernel_size=stride, stride=stride)
+
+    def forward(self, x):
+        assert x.shape[1] == self.channels
+        return self.op(x)
+
+
+class ResBlock(TimestepBlock):
+    """
+    A residual block that can optionally change the number of channels.
+    :param channels: the number of input channels.
+    :param emb_channels: the number of timestep embedding channels.
+    :param dropout: the rate of dropout.
+    :param out_channels: if specified, the number of out channels.
+    :param use_conv: if True and out_channels is specified, use a spatial
+        convolution instead of a smaller 1x1 convolution to change the
+        channels in the skip connection.
+    :param dims: determines if the signal is 1D, 2D, or 3D.
+    :param use_checkpoint: if True, use gradient checkpointing on this module.
+    :param up: if True, use this block for upsampling.
+    :param down: if True, use this block for downsampling.
+    """
+
+    def __init__(
+        self,
+        channels,
+        emb_channels,
+        dropout,
+        out_channels=None,
+        use_conv=False,
+        use_scale_shift_norm=False,
+        dims=2,
+        use_checkpoint=False,
+        up=False,
+        down=False,
+    ):
+        super().__init__()
+        self.channels = channels
+        self.emb_channels = emb_channels
+        self.dropout = dropout
+        self.out_channels = out_channels or channels
+        self.use_conv = use_conv
+        self.use_checkpoint = use_checkpoint
+        self.use_scale_shift_norm = use_scale_shift_norm
+
+        self.in_layers = nn.Sequential(
+            normalization(channels),
+            nn.SiLU(),
+            conv_nd(dims, channels, self.out_channels, 3, padding=1),
+        )
+
+        self.updown = up or down
+
+        if up:
+            self.h_upd = Upsample(channels, False, dims)
+            self.x_upd = Upsample(channels, False, dims)
+        elif down:
+            self.h_upd = Downsample(channels, False, dims)
+            self.x_upd = Downsample(channels, False, dims)
+        else:
+            self.h_upd = self.x_upd = nn.Identity()
+
+        self.emb_layers = nn.Sequential(
+            nn.SiLU(),
+            linear(
+                emb_channels,
+                2 * self.out_channels if use_scale_shift_norm else self.out_channels,
+            ),
+        )
+        self.out_layers = nn.Sequential(
+            normalization(self.out_channels),
+            nn.SiLU(),
+            nn.Dropout(p=dropout),
+            zero_module(conv_nd(dims, self.out_channels, self.out_channels, 3, padding=1)),
+        )
+
+        if self.out_channels == channels:
+            self.skip_connection = nn.Identity()
+        elif use_conv:
+            self.skip_connection = conv_nd(dims, channels, self.out_channels, 3, padding=1)
+        else:
+            self.skip_connection = conv_nd(dims, channels, self.out_channels, 1)
+
+    def forward(self, x, emb):
+        """
+        Apply the block to a Tensor, conditioned on a timestep embedding.
+        :param x: an [N x C x ...] Tensor of features.
+        :param emb: an [N x emb_channels] Tensor of timestep embeddings.
+        :return: an [N x C x ...] Tensor of outputs.
+        """
+        return checkpoint(self._forward, (x, emb), self.parameters(), self.use_checkpoint)
+
+    def _forward(self, x, emb):
+        if self.updown:
+            in_rest, in_conv = self.in_layers[:-1], self.in_layers[-1]
+            h = in_rest(x)
+            h = self.h_upd(h)
+            x = self.x_upd(x)
+            h = in_conv(h)
+        else:
+            h = self.in_layers(x)
+        emb_out = self.emb_layers(emb).type(h.dtype)
+        while len(emb_out.shape) < len(h.shape):
+            emb_out = emb_out[..., None]
+        if self.use_scale_shift_norm:
+            out_norm, out_rest = self.out_layers[0], self.out_layers[1:]
+            scale, shift = th.chunk(emb_out, 2, dim=1)
+            h = out_norm(h) * (1 + scale) + shift
+            h = out_rest(h)
+        else:
+            h = h + emb_out
+            h = self.out_layers(h)
+        return self.skip_connection(x) + h
+
+
+class AttentionBlock(nn.Module):
+    """
+    An attention block that allows spatial positions to attend to each other.
+    Originally ported from here, but adapted to the N-d case.
+    https://github.com/hojonathanho/diffusion/blob/1e0dceb3b3495bbe19116a5e1b3596cd0706c543/diffusion_tf/models/unet.py#L66.
+    """
+
+    def __init__(
+        self,
+        channels,
+        num_heads=1,
+        num_head_channels=-1,
+        use_checkpoint=False,
+        use_new_attention_order=False,
+    ):
+        super().__init__()
+        self.channels = channels
+        if num_head_channels == -1:
+            self.num_heads = num_heads
+        else:
+            assert (channels % num_head_channels == 0), f"q,k,v channels {channels} is not divisible by num_head_channels {num_head_channels}"
+            self.num_heads = channels // num_head_channels
+        self.use_checkpoint = use_checkpoint
+        self.norm = normalization(channels)
+        self.qkv = conv_nd(1, channels, channels * 3, 1)
+        if use_new_attention_order:
+            # split qkv before split heads
+            self.attention = QKVAttention(self.num_heads)
+        else:
+            # split heads before split qkv
+            self.attention = QKVAttentionLegacy(self.num_heads)
+
+        self.proj_out = zero_module(conv_nd(1, channels, channels, 1))
+
+    def forward(self, x):
+        return checkpoint(self._forward, (x,), self.parameters(), True) # TODO: check checkpoint usage, is True # TODO: fix the .half call!!!
+        #return pt_checkpoint(self._forward, x)  # pytorch
+
+    def _forward(self, x):
+        b, c, *spatial = x.shape
+        x = x.reshape(b, c, -1)
+        qkv = self.qkv(self.norm(x))
+        h = self.attention(qkv)
+        h = self.proj_out(h)
+        return (x + h).reshape(b, c, *spatial)
+
+
+def count_flops_attn(model, _x, y):
+    """
+    A counter for the `thop` package to count the operations in an
+    attention operation.
+    Meant to be used like:
+        macs, params = thop.profile(
+            model,
+            inputs=(inputs, timestamps),
+            custom_ops={QKVAttention: QKVAttention.count_flops},
+        )
+    """
+    b, c, *spatial = y[0].shape
+    num_spatial = int(np.prod(spatial))
+    # We perform two matmuls with the same number of ops.
+    # The first computes the weight matrix, the second computes
+    # the combination of the value vectors.
+    matmul_ops = 2 * b * (num_spatial**2) * c
+    model.total_ops += th.DoubleTensor([matmul_ops])
+
+
+class QKVAttentionLegacy(nn.Module):
+    """
+    A module which performs QKV attention. Matches legacy QKVAttention + input/ouput heads shaping
+    """
+
+    def __init__(self, n_heads):
+        super().__init__()
+        self.n_heads = n_heads
+
+    def forward(self, qkv):
+        """
+        Apply QKV attention.
+        :param qkv: an [N x (H * 3 * C) x T] tensor of Qs, Ks, and Vs.
+        :return: an [N x (H * C) x T] tensor after attention.
+        """
+        bs, width, length = qkv.shape
+        assert width % (3 * self.n_heads) == 0
+        ch = width // (3 * self.n_heads)
+        q, k, v = qkv.reshape(bs * self.n_heads, ch * 3, length).split(ch, dim=1)
+        scale = 1 / math.sqrt(math.sqrt(ch))
+        weight = th.einsum("bct,bcs->bts", q * scale, k * scale) # More stable with f16 than dividing afterwards
+        weight = th.softmax(weight.float(), dim=-1).type(weight.dtype)
+        a = th.einsum("bts,bcs->bct", weight, v)
+        return a.reshape(bs, -1, length)
+
+    @staticmethod
+    def count_flops(model, _x, y):
+        return count_flops_attn(model, _x, y)
+
+
+class QKVAttention(nn.Module):
+    """
+    A module which performs QKV attention and splits in a different order.
+    """
+
+    def __init__(self, n_heads):
+        super().__init__()
+        self.n_heads = n_heads
+
+    def forward(self, qkv):
+        """
+        Apply QKV attention.
+        :param qkv: an [N x (3 * H * C) x T] tensor of Qs, Ks, and Vs.
+        :return: an [N x (H * C) x T] tensor after attention.
+        """
+        bs, width, length = qkv.shape
+        assert width % (3 * self.n_heads) == 0
+        ch = width // (3 * self.n_heads)
+        q, k, v = qkv.chunk(3, dim=1)
+        scale = 1 / math.sqrt(math.sqrt(ch))
+        weight = th.einsum(
+            "bct,bcs->bts",
+            (q * scale).view(bs * self.n_heads, ch, length),
+            (k * scale).view(bs * self.n_heads, ch, length),
+        ) # More stable with f16 than dividing afterwards
+        weight = th.softmax(weight.float(), dim=-1).type(weight.dtype)
+        a = th.einsum("bts,bcs->bct", weight, v.reshape(bs * self.n_heads, ch, length))
+        return a.reshape(bs, -1, length)
+
+    @staticmethod
+    def count_flops(model, _x, y):
+        return count_flops_attn(model, _x, y)
+
+
+class Timestep(nn.Module):
+
+    def __init__(self, dim):
+        super().__init__()
+        self.dim = dim
+
+    def forward(self, t):
+        return timestep_embedding(t, self.dim)
+
+
+class MultiViewUNetModel(nn.Module):
+    """
+    The full multi-view UNet model with attention, timestep embedding and camera embedding.
+    :param in_channels: channels in the input Tensor.
+    :param model_channels: base channel count for the model.
+    :param out_channels: channels in the output Tensor.
+    :param num_res_blocks: number of residual blocks per downsample.
+    :param attention_resolutions: a collection of downsample rates at which
+        attention will take place. May be a set, list, or tuple.
+        For example, if this contains 4, then at 4x downsampling, attention
+        will be used.
+    :param dropout: the dropout probability.
+    :param channel_mult: channel multiplier for each level of the UNet.
+    :param conv_resample: if True, use learned convolutions for upsampling and
+        downsampling.
+    :param dims: determines if the signal is 1D, 2D, or 3D.
+    :param num_classes: if specified (as an int), then this model will be
+        class-conditional with `num_classes` classes.
+    :param use_checkpoint: use gradient checkpointing to reduce memory usage.
+    :param num_heads: the number of attention heads in each attention layer.
+    :param num_heads_channels: if specified, ignore num_heads and instead use
+                               a fixed channel width per attention head.
+    :param num_heads_upsample: works with num_heads to set a different number
+                               of heads for upsampling. Deprecated.
+    :param use_scale_shift_norm: use a FiLM-like conditioning mechanism.
+    :param resblock_updown: use residual blocks for up/downsampling.
+    :param use_new_attention_order: use a different attention pattern for potentially
+                                    increased efficiency.
+    :param camera_dim: dimensionality of camera input.
+    """
+
+    def __init__(
+            self,
+            image_size,
+            in_channels,
+            model_channels,
+            out_channels,
+            num_res_blocks,
+            attention_resolutions,
+            dropout=0,
+            channel_mult=(1, 2, 4, 8),
+            conv_resample=True,
+            dims=2,
+            num_classes=None,
+            use_checkpoint=False,
+            num_heads=-1,
+            num_head_channels=-1,
+            num_heads_upsample=-1,
+            use_scale_shift_norm=False,
+            resblock_updown=False,
+            use_new_attention_order=False,
+            use_spatial_transformer=False, # custom transformer support
+            transformer_depth=1, # custom transformer support
+            context_dim=None, # custom transformer support
+            n_embed=None, # custom support for prediction of discrete ids into codebook of first stage vq model
+            legacy=True,
+            disable_self_attentions=None,
+            num_attention_blocks=None,
+            disable_middle_self_attn=False,
+            use_linear_in_transformer=False,
+            adm_in_channels=None,
+            camera_dim=None,
+    ):
+        super().__init__()
+        if use_spatial_transformer:
+            assert context_dim is not None, 'Fool!! You forgot to include the dimension of your cross-attention conditioning...'
+
+        if context_dim is not None:
+            assert use_spatial_transformer, 'Fool!! You forgot to use the spatial transformer for your cross-attention conditioning...'
+            from omegaconf.listconfig import ListConfig
+            if type(context_dim) == ListConfig:
+                context_dim = list(context_dim)
+
+        if num_heads_upsample == -1:
+            num_heads_upsample = num_heads
+
+        if num_heads == -1:
+            assert num_head_channels != -1, 'Either num_heads or num_head_channels has to be set'
+
+        if num_head_channels == -1:
+            assert num_heads != -1, 'Either num_heads or num_head_channels has to be set'
+
+        self.image_size = image_size
+        self.in_channels = in_channels
+        self.model_channels = model_channels
+        self.out_channels = out_channels
+        if isinstance(num_res_blocks, int):
+            self.num_res_blocks = len(channel_mult) * [num_res_blocks]
+        else:
+            if len(num_res_blocks) != len(channel_mult):
+                raise ValueError("provide num_res_blocks either as an int (globally constant) or "
+                                 "as a list/tuple (per-level) with the same length as channel_mult")
+            self.num_res_blocks = num_res_blocks
+        if disable_self_attentions is not None:
+            # should be a list of booleans, indicating whether to disable self-attention in TransformerBlocks or not
+            assert len(disable_self_attentions) == len(channel_mult)
+        if num_attention_blocks is not None:
+            assert len(num_attention_blocks) == len(self.num_res_blocks)
+            assert all(map(lambda i: self.num_res_blocks[i] >= num_attention_blocks[i], range(len(num_attention_blocks))))
+            print(f"Constructor of UNetModel received num_attention_blocks={num_attention_blocks}. "
+                  f"This option has LESS priority than attention_resolutions {attention_resolutions}, "
+                  f"i.e., in cases where num_attention_blocks[i] > 0 but 2**i not in attention_resolutions, "
+                  f"attention will still not be set.")
+
+        self.attention_resolutions = attention_resolutions
+        self.dropout = dropout
+        self.channel_mult = channel_mult
+        self.conv_resample = conv_resample
+        self.num_classes = num_classes
+        self.use_checkpoint = use_checkpoint
+        self.num_heads = num_heads
+        self.num_head_channels = num_head_channels
+        self.num_heads_upsample = num_heads_upsample
+        self.predict_codebook_ids = n_embed is not None
+
+        time_embed_dim = model_channels * 4
+        self.time_embed = nn.Sequential(
+            linear(model_channels, time_embed_dim),
+            nn.SiLU(),
+            linear(time_embed_dim, time_embed_dim),
+        )
+
+        if camera_dim is not None:
+            time_embed_dim = model_channels * 4
+            self.camera_embed = nn.Sequential(
+                linear(camera_dim, time_embed_dim),
+                nn.SiLU(),
+                linear(time_embed_dim, time_embed_dim),
+            )
+
+        if self.num_classes is not None:
+            if isinstance(self.num_classes, int):
+                self.label_emb = nn.Embedding(self.num_classes, time_embed_dim)
+            elif self.num_classes == "continuous":
+                # print("setting up linear c_adm embedding layer")
+                self.label_emb = nn.Linear(1, time_embed_dim)
+            elif self.num_classes == "sequential":
+                assert adm_in_channels is not None
+                self.label_emb = nn.Sequential(nn.Sequential(
+                    linear(adm_in_channels, time_embed_dim),
+                    nn.SiLU(),
+                    linear(time_embed_dim, time_embed_dim),
+                ))
+            else:
+                raise ValueError()
+
+        self.input_blocks = nn.ModuleList([TimestepEmbedSequential(conv_nd(dims, in_channels, model_channels, 3, padding=1))])
+        self._feature_size = model_channels
+        input_block_chans = [model_channels]
+        ch = model_channels
+        ds = 1
+        for level, mult in enumerate(channel_mult):
+            for nr in range(self.num_res_blocks[level]):
+                layers: List[Any] = [ResBlock(
+                    ch,
+                    time_embed_dim,
+                    dropout,
+                    out_channels=mult * model_channels,
+                    dims=dims,
+                    use_checkpoint=use_checkpoint,
+                    use_scale_shift_norm=use_scale_shift_norm,
+                )]
+                ch = mult * model_channels
+                if ds in attention_resolutions:
+                    if num_head_channels == -1:
+                        dim_head = ch // num_heads
+                    else:
+                        num_heads = ch // num_head_channels
+                        dim_head = num_head_channels
+                    if legacy:
+                        #num_heads = 1
+                        dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+                    if disable_self_attentions is not None:
+                        disabled_sa = disable_self_attentions[level]
+                    else:
+                        disabled_sa = False
+
+                    if num_attention_blocks is None or nr < num_attention_blocks[level]:
+                        layers.append(AttentionBlock(
+                            ch,
+                            use_checkpoint=use_checkpoint,
+                            num_heads=num_heads,
+                            num_head_channels=dim_head,
+                            use_new_attention_order=use_new_attention_order,
+                        ) if not use_spatial_transformer else SpatialTransformer3D(ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim, disable_self_attn=disabled_sa, use_linear=use_linear_in_transformer, use_checkpoint=use_checkpoint))
+                self.input_blocks.append(TimestepEmbedSequential(*layers))
+                self._feature_size += ch
+                input_block_chans.append(ch)
+            if level != len(channel_mult) - 1:
+                out_ch = ch
+                self.input_blocks.append(TimestepEmbedSequential(ResBlock(
+                    ch,
+                    time_embed_dim,
+                    dropout,
+                    out_channels=out_ch,
+                    dims=dims,
+                    use_checkpoint=use_checkpoint,
+                    use_scale_shift_norm=use_scale_shift_norm,
+                    down=True,
+                ) if resblock_updown else Downsample(ch, conv_resample, dims=dims, out_channels=out_ch)))
+                ch = out_ch
+                input_block_chans.append(ch)
+                ds *= 2
+                self._feature_size += ch
+
+        if num_head_channels == -1:
+            dim_head = ch // num_heads
+        else:
+            num_heads = ch // num_head_channels
+            dim_head = num_head_channels
+        if legacy:
+            #num_heads = 1
+            dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+        self.middle_block = TimestepEmbedSequential(
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+            AttentionBlock(
+                ch,
+                use_checkpoint=use_checkpoint,
+                num_heads=num_heads,
+                num_head_channels=dim_head,
+                use_new_attention_order=use_new_attention_order,
+            ) if not use_spatial_transformer else SpatialTransformer3D( # always uses a self-attn
+                ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim, disable_self_attn=disable_middle_self_attn, use_linear=use_linear_in_transformer, use_checkpoint=use_checkpoint),
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+        )
+        self._feature_size += ch
+
+        self.output_blocks = nn.ModuleList([])
+        for level, mult in list(enumerate(channel_mult))[::-1]:
+            for i in range(self.num_res_blocks[level] + 1):
+                ich = input_block_chans.pop()
+                layers = [ResBlock(
+                    ch + ich,
+                    time_embed_dim,
+                    dropout,
+                    out_channels=model_channels * mult,
+                    dims=dims,
+                    use_checkpoint=use_checkpoint,
+                    use_scale_shift_norm=use_scale_shift_norm,
+                )]
+                ch = model_channels * mult
+                if ds in attention_resolutions:
+                    if num_head_channels == -1:
+                        dim_head = ch // num_heads
+                    else:
+                        num_heads = ch // num_head_channels
+                        dim_head = num_head_channels
+                    if legacy:
+                        #num_heads = 1
+                        dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+                    if disable_self_attentions is not None:
+                        disabled_sa = disable_self_attentions[level]
+                    else:
+                        disabled_sa = False
+
+                    if num_attention_blocks is None or i < num_attention_blocks[level]:
+                        layers.append(AttentionBlock(
+                            ch,
+                            use_checkpoint=use_checkpoint,
+                            num_heads=num_heads_upsample,
+                            num_head_channels=dim_head,
+                            use_new_attention_order=use_new_attention_order,
+                        ) if not use_spatial_transformer else SpatialTransformer3D(ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim, disable_self_attn=disabled_sa, use_linear=use_linear_in_transformer, use_checkpoint=use_checkpoint))
+                if level and i == self.num_res_blocks[level]:
+                    out_ch = ch
+                    layers.append(ResBlock(
+                        ch,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=out_ch,
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                        up=True,
+                    ) if resblock_updown else Upsample(ch, conv_resample, dims=dims, out_channels=out_ch))
+                    ds //= 2
+                self.output_blocks.append(TimestepEmbedSequential(*layers))
+                self._feature_size += ch
+
+        self.out = nn.Sequential(
+            normalization(ch),
+            nn.SiLU(),
+            zero_module(conv_nd(dims, model_channels, out_channels, 3, padding=1)),
+        )
+        if self.predict_codebook_ids:
+            self.id_predictor = nn.Sequential(
+                normalization(ch),
+                conv_nd(dims, model_channels, n_embed, 1),
+                #nn.LogSoftmax(dim=1)  # change to cross_entropy and produce non-normalized logits
+            )
+
+    def forward(self, x, timesteps=None, context=None, y: Optional[Tensor] = None, camera=None, num_frames=1, **kwargs):
+        """
+        Apply the model to an input batch.
+        :param x: an [(N x F) x C x ...] Tensor of inputs. F is the number of frames (views).
+        :param timesteps: a 1-D batch of timesteps.
+        :param context: conditioning plugged in via crossattn
+        :param y: an [N] Tensor of labels, if class-conditional.
+        :param num_frames: a integer indicating number of frames for tensor reshaping.
+        :return: an [(N x F) x C x ...] Tensor of outputs. F is the number of frames (views).
+        """
+        assert x.shape[0] % num_frames == 0, "[UNet] input batch size must be dividable by num_frames!"
+        assert (y is not None) == (self.num_classes is not None), "must specify y if and only if the model is class-conditional"
+        hs = []
+        t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False).to(x.dtype)
+        
+        emb = self.time_embed(t_emb)
+
+        if self.num_classes is not None:
+            assert y is not None
+            assert y.shape[0] == x.shape[0]
+            emb = emb + self.label_emb(y)
+
+        # Add camera embeddings
+        if camera is not None:
+            assert camera.shape[0] == emb.shape[0]
+            emb = emb + self.camera_embed(camera)
+
+        h = x
+        for module in self.input_blocks:
+            h = module(h, emb, context, num_frames=num_frames)
+            hs.append(h)
+        h = self.middle_block(h, emb, context, num_frames=num_frames)
+        for module in self.output_blocks:
+            h = th.cat([h, hs.pop()], dim=1)
+            h = module(h, emb, context, num_frames=num_frames)
+        h = h.type(x.dtype)
+        if self.predict_codebook_ids:
+            return self.id_predictor(h)
+        else:
+            return self.out(h)

mvdream/pipeline_mvdream.py

@@ -0,0 +1,484 @@
+import torch
+import numpy as np
+import inspect
+from typing import Callable, List, Optional, Union
+from transformers import CLIPTextModel, CLIPTokenizer
+from diffusers import AutoencoderKL, DiffusionPipeline
+from diffusers.utils import (
+    deprecate,
+    is_accelerate_available,
+    is_accelerate_version,
+    logging,
+)
+from diffusers.configuration_utils import FrozenDict
+from diffusers.schedulers import DDIMScheduler
+try:
+    from diffusers import randn_tensor # old import # type: ignore
+except ImportError:
+    from diffusers.utils.torch_utils import randn_tensor # new import # type: ignore
+
+from .models import MultiViewUNetWrapperModel
+from accelerate.utils import set_module_tensor_to_device
+
+logger = logging.get_logger(__name__) # pylint: disable=invalid-name
+
+def create_camera_to_world_matrix(elevation, azimuth):
+    elevation = np.radians(elevation)
+    azimuth = np.radians(azimuth)
+    # Convert elevation and azimuth angles to Cartesian coordinates on a unit sphere
+    x = np.cos(elevation) * np.sin(azimuth)
+    y = np.sin(elevation)
+    z = np.cos(elevation) * np.cos(azimuth)
+
+    # Calculate camera position, target, and up vectors
+    camera_pos = np.array([x, y, z])
+    target = np.array([0, 0, 0])
+    up = np.array([0, 1, 0])
+
+    # Construct view matrix
+    forward = target - camera_pos
+    forward /= np.linalg.norm(forward)
+    right = np.cross(forward, up)
+    right /= np.linalg.norm(right)
+    new_up = np.cross(right, forward)
+    new_up /= np.linalg.norm(new_up)
+    cam2world = np.eye(4)
+    cam2world[:3, :3] = np.array([right, new_up, -forward]).T
+    cam2world[:3, 3] = camera_pos
+    return cam2world
+
+
+def convert_opengl_to_blender(camera_matrix):
+    if isinstance(camera_matrix, np.ndarray):
+        # Construct transformation matrix to convert from OpenGL space to Blender space
+        flip_yz = np.array([[1, 0, 0, 0], [0, 0, -1, 0], [0, 1, 0, 0], [0, 0, 0, 1]])
+        camera_matrix_blender = np.dot(flip_yz, camera_matrix)
+    else:
+        # Construct transformation matrix to convert from OpenGL space to Blender space
+        flip_yz = torch.tensor([[1, 0, 0, 0], [0, 0, -1, 0], [0, 1, 0, 0], [0, 0, 0, 1]])
+        if camera_matrix.ndim == 3:
+            flip_yz = flip_yz.unsqueeze(0)
+        camera_matrix_blender = torch.matmul(flip_yz.to(camera_matrix), camera_matrix)
+    return camera_matrix_blender
+
+
+def get_camera(num_frames, elevation=15, azimuth_start=0, azimuth_span=360, blender_coord=True):
+    angle_gap = azimuth_span / num_frames
+    cameras = []
+    for azimuth in np.arange(azimuth_start, azimuth_span + azimuth_start, angle_gap):
+        camera_matrix = create_camera_to_world_matrix(elevation, azimuth)
+        if blender_coord:
+            camera_matrix = convert_opengl_to_blender(camera_matrix)
+        cameras.append(camera_matrix.flatten())
+    return torch.tensor(np.stack(cameras, 0)).float()
+
+
+class MVDreamStableDiffusionPipeline(DiffusionPipeline):
+
+    def __init__(
+        self,
+        vae: AutoencoderKL,
+        unet: MultiViewUNetWrapperModel,
+        tokenizer: CLIPTokenizer,
+        text_encoder: CLIPTextModel,
+        scheduler: DDIMScheduler,
+        requires_safety_checker: bool = False,
+    ):
+        super().__init__()
+
+        if hasattr(scheduler.config, "steps_offset") and scheduler.config.steps_offset != 1: # type: ignore
+            deprecation_message = (f"The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`"
+                                   f" should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure " # type: ignore
+                                   "to update the config accordingly as leaving `steps_offset` might led to incorrect results"
+                                   " in future versions. If you have downloaded this checkpoint from the Hugging Face Hub,"
+                                   " it would be very nice if you could open a Pull request for the `scheduler/scheduler_config.json`"
+                                   " file")
+            deprecate("steps_offset!=1", "1.0.0", deprecation_message, standard_warn=False)
+            new_config = dict(scheduler.config)
+            new_config["steps_offset"] = 1
+            scheduler._internal_dict = FrozenDict(new_config)
+
+        if hasattr(scheduler.config, "clip_sample") and scheduler.config.clip_sample is True: # type: ignore
+            deprecation_message = (f"The configuration file of this scheduler: {scheduler} has not set the configuration `clip_sample`."
+                                   " `clip_sample` should be set to False in the configuration file. Please make sure to update the"
+                                   " config accordingly as not setting `clip_sample` in the config might lead to incorrect results in"
+                                   " future versions. If you have downloaded this checkpoint from the Hugging Face Hub, it would be very"
+                                   " nice if you could open a Pull request for the `scheduler/scheduler_config.json` file")
+            deprecate("clip_sample not set", "1.0.0", deprecation_message, standard_warn=False)
+            new_config = dict(scheduler.config)
+            new_config["clip_sample"] = False
+            scheduler._internal_dict = FrozenDict(new_config)
+
+        self.register_modules(
+            vae=vae,
+            unet=unet,
+            scheduler=scheduler,
+            tokenizer=tokenizer,
+            text_encoder=text_encoder,
+        )
+        self.vae_scale_factor = 2**(len(self.vae.config.block_out_channels) - 1)
+        self.register_to_config(requires_safety_checker=requires_safety_checker)
+
+    def enable_vae_slicing(self):
+        r"""
+        Enable sliced VAE decoding.
+
+        When this option is enabled, the VAE will split the input tensor in slices to compute decoding in several
+        steps. This is useful to save some memory and allow larger batch sizes.
+        """
+        self.vae.enable_slicing()
+
+    def disable_vae_slicing(self):
+        r"""
+        Disable sliced VAE decoding. If `enable_vae_slicing` was previously invoked, this method will go back to
+        computing decoding in one step.
+        """
+        self.vae.disable_slicing()
+
+    def enable_vae_tiling(self):
+        r"""
+        Enable tiled VAE decoding.
+
+        When this option is enabled, the VAE will split the input tensor into tiles to compute decoding and encoding in
+        several steps. This is useful to save a large amount of memory and to allow the processing of larger images.
+        """
+        self.vae.enable_tiling()
+
+    def disable_vae_tiling(self):
+        r"""
+        Disable tiled VAE decoding. If `enable_vae_tiling` was previously invoked, this method will go back to
+        computing decoding in one step.
+        """
+        self.vae.disable_tiling()
+
+    def enable_sequential_cpu_offload(self, gpu_id=0):
+        r"""
+        Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
+        text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
+        `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called.
+        Note that offloading happens on a submodule basis. Memory savings are higher than with
+        `enable_model_cpu_offload`, but performance is lower.
+        """
+        if is_accelerate_available() and is_accelerate_version(">=", "0.14.0"):
+            from accelerate import cpu_offload
+        else:
+            raise ImportError("`enable_sequential_cpu_offload` requires `accelerate v0.14.0` or higher")
+
+        device = torch.device(f"cuda:{gpu_id}")
+
+        if self.device.type != "cpu":
+            self.to("cpu", silence_dtype_warnings=True)
+            torch.cuda.empty_cache() # otherwise we don't see the memory savings (but they probably exist)
+
+        for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
+            cpu_offload(cpu_offloaded_model, device)
+
+    def enable_model_cpu_offload(self, gpu_id=0):
+        r"""
+        Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared
+        to `enable_sequential_cpu_offload`, this method moves one whole model at a time to the GPU when its `forward`
+        method is called, and the model remains in GPU until the next model runs. Memory savings are lower than with
+        `enable_sequential_cpu_offload`, but performance is much better due to the iterative execution of the `unet`.
+        """
+        if is_accelerate_available() and is_accelerate_version(">=", "0.17.0.dev0"):
+            from accelerate import cpu_offload_with_hook
+        else:
+            raise ImportError("`enable_model_offload` requires `accelerate v0.17.0` or higher.")
+
+        device = torch.device(f"cuda:{gpu_id}")
+
+        if self.device.type != "cpu":
+            self.to("cpu", silence_dtype_warnings=True)
+            torch.cuda.empty_cache() # otherwise we don't see the memory savings (but they probably exist)
+
+        hook = None
+        for cpu_offloaded_model in [self.text_encoder, self.unet, self.vae]:
+            _, hook = cpu_offload_with_hook(cpu_offloaded_model, device, prev_module_hook=hook)
+
+        # We'll offload the last model manually.
+        self.final_offload_hook = hook
+
+    @property
+    def _execution_device(self):
+        r"""
+        Returns the device on which the pipeline's models will be executed. After calling
+        `pipeline.enable_sequential_cpu_offload()` the execution device can only be inferred from Accelerate's module
+        hooks.
+        """
+        if not hasattr(self.unet, "_hf_hook"):
+            return self.device
+        for module in self.unet.modules():
+            if (hasattr(module, "_hf_hook") and hasattr(module._hf_hook, "execution_device") and module._hf_hook.execution_device is not None):
+                return torch.device(module._hf_hook.execution_device)
+        return self.device
+
+    def _encode_prompt(
+        self,
+        prompt,
+        device,
+        num_images_per_prompt,
+        do_classifier_free_guidance: bool,
+        negative_prompt=None,
+    ):
+        r"""
+        Encodes the prompt into text encoder hidden states.
+
+        Args:
+             prompt (`str` or `List[str]`, *optional*):
+                prompt to be encoded
+            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*):
+                The prompt or prompts not to guide the image generation. If not defined, one has to pass
+                `negative_prompt_embeds`. instead. If not defined, one has to pass `negative_prompt_embeds`. instead.
+                Ignored when not using guidance (i.e., ignored if `guidance_scale` is less than `1`).
+            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.
+            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
+                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
+                argument.
+        """
+        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:
+            raise ValueError(f"`prompt` should be either a string or a list of strings, but got {type(prompt)}.")
+
+        text_inputs = self.tokenizer(
+            prompt,
+            padding="max_length",
+            max_length=self.tokenizer.model_max_length,
+            truncation=True,
+            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.model_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.model_max_length} tokens: {removed_text}")
+
+        if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+            attention_mask = text_inputs.attention_mask.to(device)
+        else:
+            attention_mask = None
+
+        prompt_embeds = self.text_encoder(
+            text_input_ids.to(device),
+            attention_mask=attention_mask,
+        )
+        prompt_embeds = prompt_embeds[0]
+
+        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
+
+        bs_embed, seq_len, _ = prompt_embeds.shape
+        # 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(bs_embed * num_images_per_prompt, seq_len, -1)
+
+        # get unconditional embeddings for classifier free guidance
+        if do_classifier_free_guidance:
+            uncond_tokens: List[str]
+            if negative_prompt is None:
+                uncond_tokens = [""] * batch_size
+            elif type(prompt) is not type(negative_prompt):
+                raise TypeError(f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
+                                f" {type(prompt)}.")
+            elif isinstance(negative_prompt, str):
+                uncond_tokens = [negative_prompt]
+            elif batch_size != len(negative_prompt):
+                raise ValueError(f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
+                                 f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
+                                 " the batch size of `prompt`.")
+            else:
+                uncond_tokens = negative_prompt
+
+            max_length = prompt_embeds.shape[1]
+            uncond_input = self.tokenizer(
+                uncond_tokens,
+                padding="max_length",
+                max_length=max_length,
+                truncation=True,
+                return_tensors="pt",
+            )
+
+            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+                attention_mask = uncond_input.attention_mask.to(device)
+            else:
+                attention_mask = None
+
+            negative_prompt_embeds = self.text_encoder(
+                uncond_input.input_ids.to(device),
+                attention_mask=attention_mask,
+            )
+            negative_prompt_embeds = negative_prompt_embeds[0]
+
+            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
+            seq_len = negative_prompt_embeds.shape[1]
+
+            negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
+
+            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
+            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+
+            # For classifier free guidance, we need to do two forward passes.
+            # Here we concatenate the unconditional and text embeddings into a single batch
+            # to avoid doing two forward passes
+            prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds])
+
+        return prompt_embeds
+
+    def decode_latents(self, latents):
+        latents = 1 / self.vae.config.scaling_factor * latents
+        image = self.vae.decode(latents).sample
+        image = (image / 2 + 0.5).clamp(0, 1)
+        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
+        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
+        return image
+
+    def prepare_extra_step_kwargs(self, generator, eta):
+        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
+        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
+        # eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502
+        # and should be between [0, 1]
+
+        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        extra_step_kwargs = {}
+        if accepts_eta:
+            extra_step_kwargs["eta"] = eta
+
+        # check if the scheduler accepts generator
+        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        if accepts_generator:
+            extra_step_kwargs["generator"] = generator
+        return extra_step_kwargs
+
+    def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
+        shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
+        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.")
+
+        if latents is None:
+            latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
+        else:
+            latents = latents.to(device)
+
+        # scale the initial noise by the standard deviation required by the scheduler
+        latents = latents * self.scheduler.init_noise_sigma
+        return latents
+
+    @torch.no_grad()
+    def __call__(
+        self,
+        prompt: str = "a car",
+        height: int = 256,
+        width: int = 256,
+        num_inference_steps: int = 50,
+        guidance_scale: float = 7.0,
+        negative_prompt: str = "bad quality",
+        num_images_per_prompt: int = 1,
+        eta: float = 0.0,
+        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
+        output_type: Optional[str] = "image",
+        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback_steps: int = 1,
+        batch_size: int = 4,
+        device = torch.device("cuda:0"),
+    ):
+        self.unet = self.unet.to(device=device)
+        self.vae = self.vae.to(device=device)
+
+        self.text_encoder = self.text_encoder.to(device=device)
+
+
+        # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
+        # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
+        # corresponds to doing no classifier free guidance.
+        do_classifier_free_guidance = guidance_scale > 1.0
+
+        # Prepare timesteps
+        self.scheduler.set_timesteps(num_inference_steps, device=device)
+        timesteps = self.scheduler.timesteps
+
+        _prompt_embeds: torch.Tensor = self._encode_prompt(
+            prompt=prompt,
+            device=device,
+            num_images_per_prompt=num_images_per_prompt,
+            do_classifier_free_guidance=do_classifier_free_guidance,
+            negative_prompt=negative_prompt,
+        ) # type: ignore
+        prompt_embeds_neg, prompt_embeds_pos = _prompt_embeds.chunk(2)
+
+        # Prepare latent variables
+        latents: torch.Tensor = self.prepare_latents(
+            batch_size * num_images_per_prompt,
+            4,
+            height,
+            width,
+            prompt_embeds_pos.dtype,
+            device,
+            generator,
+            None,
+        )
+        
+        camera = get_camera(batch_size).to(dtype=latents.dtype, device=device)
+
+        # Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
+        extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)
+
+        # Denoising loop
+        num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
+        with self.progress_bar(total=num_inference_steps) as progress_bar:
+            for i, t in enumerate(timesteps):
+                # expand the latents if we are doing classifier free guidance
+                multiplier = 2 if do_classifier_free_guidance else 1
+                latent_model_input = torch.cat([latents] * multiplier)
+                latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
+
+                # predict the noise residual
+                noise_pred = self.unet.forward(
+                    x=latent_model_input,
+                    timesteps=torch.tensor([t] * 4 * multiplier, dtype=latent_model_input.dtype, device=device),
+                    context=torch.cat([prompt_embeds_neg] * 4 + [prompt_embeds_pos] * 4),
+                    num_frames=4,
+                    camera=torch.cat([camera] * multiplier),
+                )
+
+                # perform guidance
+                if do_classifier_free_guidance:
+                    noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
+                    noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
+
+                # compute the previous noisy sample x_t -> x_t-1
+                # latents = self.scheduler.step(noise_pred.to(dtype=torch.float32), t, latents.to(dtype=torch.float32)).prev_sample.to(prompt_embeds.dtype)
+                latents: torch.Tensor = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
+
+                # 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 callback is not None and i % callback_steps == 0:
+                        callback(i, t, latents) # type: ignore
+
+        # Post-processing
+        if output_type == "latent":
+            image = latents
+        elif output_type == "pil":
+            image = self.decode_latents(latents)
+            image = self.numpy_to_pil(image)
+        else:
+            image = self.decode_latents(latents)
+
+        # Offload last model to CPU
+        if hasattr(self, "final_offload_hook") and self.final_offload_hook is not None:
+            self.final_offload_hook.offload()
+
+        return image

mvdream/util.py

@@ -0,0 +1,320 @@
+# adopted from
+# https://github.com/openai/improved-diffusion/blob/main/improved_diffusion/gaussian_diffusion.py
+# and
+# https://github.com/lucidrains/denoising-diffusion-pytorch/blob/7706bdfc6f527f58d33f84b7b522e61e6e3164b3/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py
+# and
+# https://github.com/openai/guided-diffusion/blob/0ba878e517b276c45d1195eb29f6f5f72659a05b/guided_diffusion/nn.py
+#
+# thanks!
+
+import math
+import torch
+import torch.nn as nn
+import numpy as np
+import importlib
+from einops import repeat
+from typing import Any
+
+
+def instantiate_from_config(config):
+    if not "target" in config:
+        if config == '__is_first_stage__':
+            return None
+        elif config == "__is_unconditional__":
+            return None
+        raise KeyError("Expected key `target` to instantiate.")
+    return get_obj_from_str(config["target"])(**config.get("params", dict()))
+
+
+def get_obj_from_str(string, reload=False):
+    module, cls = string.rsplit(".", 1)
+    if reload:
+        module_imp = importlib.import_module(module)
+        importlib.reload(module_imp)
+    return getattr(importlib.import_module(module, package=None), cls)
+
+
+def make_beta_schedule(schedule,
+                       n_timestep,
+                       linear_start=1e-4,
+                       linear_end=2e-2,
+                       cosine_s=8e-3):
+    if schedule == "linear":
+        betas = (torch.linspace(linear_start**0.5,
+                                linear_end**0.5,
+                                n_timestep,
+                                dtype=torch.float64)**2)
+
+    elif schedule == "cosine":
+        timesteps = (
+            torch.arange(n_timestep + 1, dtype=torch.float64) / n_timestep +
+            cosine_s)
+        alphas = timesteps / (1 + cosine_s) * np.pi / 2
+        alphas = torch.cos(alphas).pow(2)
+        alphas = alphas / alphas[0]
+        betas = 1 - alphas[1:] / alphas[:-1]
+        betas = np.clip(betas, a_min=0, a_max=0.999)
+
+    elif schedule == "sqrt_linear":
+        betas = torch.linspace(linear_start,
+                               linear_end,
+                               n_timestep,
+                               dtype=torch.float64)
+    elif schedule == "sqrt":
+        betas = torch.linspace(linear_start,
+                               linear_end,
+                               n_timestep,
+                               dtype=torch.float64)**0.5
+    else:
+        raise ValueError(f"schedule '{schedule}' unknown.")
+    return betas.numpy()  # type: ignore
+
+
+def make_ddim_timesteps(ddim_discr_method,
+                        num_ddim_timesteps,
+                        num_ddpm_timesteps,
+                        verbose=True):
+    if ddim_discr_method == 'uniform':
+        c = num_ddpm_timesteps // num_ddim_timesteps
+        ddim_timesteps = np.asarray(list(range(0, num_ddpm_timesteps, c)))
+    elif ddim_discr_method == 'quad':
+        ddim_timesteps = ((np.linspace(0, np.sqrt(num_ddpm_timesteps * .8),
+                                       num_ddim_timesteps))**2).astype(int)
+    else:
+        raise NotImplementedError(
+            f'There is no ddim discretization method called "{ddim_discr_method}"'
+        )
+
+    # assert ddim_timesteps.shape[0] == num_ddim_timesteps
+    # add one to get the final alpha values right (the ones from first scale to data during sampling)
+    steps_out = ddim_timesteps + 1
+    if verbose:
+        print(f'Selected timesteps for ddim sampler: {steps_out}')
+    return steps_out
+
+
+def make_ddim_sampling_parameters(alphacums,
+                                  ddim_timesteps,
+                                  eta,
+                                  verbose=True):
+    # select alphas for computing the variance schedule
+    alphas = alphacums[ddim_timesteps]
+    alphas_prev = np.asarray([alphacums[0]] +
+                             alphacums[ddim_timesteps[:-1]].tolist())
+
+    # according the the formula provided in https://arxiv.org/abs/2010.02502
+    sigmas = eta * np.sqrt(
+        (1 - alphas_prev) / (1 - alphas) * (1 - alphas / alphas_prev))
+    if verbose:
+        print(
+            f'Selected alphas for ddim sampler: a_t: {alphas}; a_(t-1): {alphas_prev}'
+        )
+        print(
+            f'For the chosen value of eta, which is {eta}, '
+            f'this results in the following sigma_t schedule for ddim sampler {sigmas}'
+        )
+    return sigmas, alphas, alphas_prev
+
+
+def betas_for_alpha_bar(num_diffusion_timesteps, alpha_bar, max_beta=0.999):
+    """
+    Create a beta schedule that discretizes the given alpha_t_bar function,
+    which defines the cumulative product of (1-beta) over time from t = [0,1].
+    :param num_diffusion_timesteps: the number of betas to produce.
+    :param alpha_bar: a lambda that takes an argument t from 0 to 1 and
+                      produces the cumulative product of (1-beta) up to that
+                      part of the diffusion process.
+    :param max_beta: the maximum beta to use; use values lower than 1 to
+                     prevent singularities.
+    """
+    betas = []
+    for i in range(num_diffusion_timesteps):
+        t1 = i / num_diffusion_timesteps
+        t2 = (i + 1) / num_diffusion_timesteps
+        betas.append(min(1 - alpha_bar(t2) / alpha_bar(t1), max_beta))
+    return np.array(betas)
+
+
+def extract_into_tensor(a, t, x_shape):
+    b, *_ = t.shape
+    out = a.gather(-1, t)
+    return out.reshape(b, *((1, ) * (len(x_shape) - 1)))
+
+
+def checkpoint(func, inputs, params, flag):
+    """
+    Evaluate a function without caching intermediate activations, allowing for
+    reduced memory at the expense of extra compute in the backward pass.
+    :param func: the function to evaluate.
+    :param inputs: the argument sequence to pass to `func`.
+    :param params: a sequence of parameters `func` depends on but does not
+                   explicitly take as arguments.
+    :param flag: if False, disable gradient checkpointing.
+    """
+    if flag:
+        args = tuple(inputs) + tuple(params)
+        return CheckpointFunction.apply(func, len(inputs), *args)
+    else:
+        return func(*inputs)
+
+
+class CheckpointFunction(torch.autograd.Function):
+
+    @staticmethod
+    def forward(ctx, run_function, length, *args):
+        ctx.run_function = run_function
+        ctx.input_tensors = list(args[:length])
+        ctx.input_params = list(args[length:])
+
+        with torch.no_grad():
+            output_tensors = ctx.run_function(*ctx.input_tensors)
+        return output_tensors
+
+    @staticmethod
+    def backward(ctx, *output_grads):
+        ctx.input_tensors = [
+            x.detach().requires_grad_(True) for x in ctx.input_tensors
+        ]
+        with torch.enable_grad():
+            # Fixes a bug where the first op in run_function modifies the
+            # Tensor storage in place, which is not allowed for detach()'d
+            # Tensors.
+            shallow_copies = [x.view_as(x) for x in ctx.input_tensors]
+            output_tensors = ctx.run_function(*shallow_copies)
+        input_grads = torch.autograd.grad(
+            output_tensors,
+            ctx.input_tensors + ctx.input_params,
+            output_grads,
+            allow_unused=True,
+        )
+        del ctx.input_tensors
+        del ctx.input_params
+        del output_tensors
+        return (None, None) + input_grads
+
+
+def timestep_embedding(timesteps, dim, max_period=10000, repeat_only=False):
+    """
+    Create sinusoidal timestep embeddings.
+    :param timesteps: a 1-D Tensor of N indices, one per batch element.
+                      These may be fractional.
+    :param dim: the dimension of the output.
+    :param max_period: controls the minimum frequency of the embeddings.
+    :return: an [N x dim] Tensor of positional embeddings.
+    """
+    if not repeat_only:
+        half = dim // 2
+        freqs = torch.exp(
+            -math.log(max_period) *
+            torch.arange(start=0, end=half, dtype=torch.float32) /
+            half).to(device=timesteps.device)
+        args = timesteps[:, None] * freqs[None]
+        embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
+        if dim % 2:
+            embedding = torch.cat(
+                [embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
+    else:
+        embedding = repeat(timesteps, 'b -> b d', d=dim)
+    # import pdb; pdb.set_trace()
+    return embedding
+
+
+def zero_module(module):
+    """
+    Zero out the parameters of a module and return it.
+    """
+    for p in module.parameters():
+        p.detach().zero_()
+    return module
+
+
+def scale_module(module, scale):
+    """
+    Scale the parameters of a module and return it.
+    """
+    for p in module.parameters():
+        p.detach().mul_(scale)
+    return module
+
+
+def mean_flat(tensor):
+    """
+    Take the mean over all non-batch dimensions.
+    """
+    return tensor.mean(dim=list(range(1, len(tensor.shape))))
+
+
+def normalization(channels):
+    """
+    Make a standard normalization layer.
+    :param channels: number of input channels.
+    :return: an nn.Module for normalization.
+    """
+    return GroupNorm32(32, channels)
+
+
+# PyTorch 1.7 has SiLU, but we support PyTorch 1.5.
+class SiLU(nn.Module):
+
+    def forward(self, x):
+        return x * torch.sigmoid(x)
+
+
+class GroupNorm32(nn.GroupNorm):
+
+    def forward(self, x):
+        return super().forward(x)
+
+
+def conv_nd(dims, *args, **kwargs):
+    """
+    Create a 1D, 2D, or 3D convolution module.
+    """
+    if dims == 1:
+        return nn.Conv1d(*args, **kwargs)
+    elif dims == 2:
+        return nn.Conv2d(*args, **kwargs)
+    elif dims == 3:
+        return nn.Conv3d(*args, **kwargs)
+    raise ValueError(f"unsupported dimensions: {dims}")
+
+
+def linear(*args, **kwargs):
+    """
+    Create a linear module.
+    """
+    return nn.Linear(*args, **kwargs)
+
+
+def avg_pool_nd(dims, *args, **kwargs):
+    """
+    Create a 1D, 2D, or 3D average pooling module.
+    """
+    if dims == 1:
+        return nn.AvgPool1d(*args, **kwargs)
+    elif dims == 2:
+        return nn.AvgPool2d(*args, **kwargs)
+    elif dims == 3:
+        return nn.AvgPool3d(*args, **kwargs)
+    raise ValueError(f"unsupported dimensions: {dims}")
+
+
+class HybridConditioner(nn.Module):
+
+    def __init__(self, c_concat_config, c_crossattn_config):
+        super().__init__()
+        self.concat_conditioner: Any = instantiate_from_config(c_concat_config)
+        self.crossattn_conditioner: Any = instantiate_from_config(
+            c_crossattn_config)
+
+    def forward(self, c_concat, c_crossattn):
+        c_concat = self.concat_conditioner(c_concat)
+        c_crossattn = self.crossattn_conditioner(c_crossattn)
+        return {'c_concat': [c_concat], 'c_crossattn': [c_crossattn]}
+
+
+def noise_like(shape, device, repeat=False):
+    repeat_noise = lambda: torch.randn((1, *shape[1:]), device=device).repeat(
+        shape[0], *((1, ) * (len(shape) - 1)))
+    noise = lambda: torch.randn(shape, device=device)
+    return repeat_noise() if repeat else noise()