wip

invokeai/app/invocations/fields.py

@@ -133,6 +133,7 @@ class FieldDescriptions:
     clip_embed_model = "CLIP Embed loader"
     unet = "UNet (scheduler, LoRAs)"
     transformer = "Transformer"
+    mmditx = "MMDiTX"
     vae = "VAE"
     cond = "Conditioning tensor"
     controlnet_model = "ControlNet model to load"
@@ -140,6 +141,7 @@ class FieldDescriptions:
     lora_model = "LoRA model to load"
     main_model = "Main model (UNet, VAE, CLIP) to load"
     flux_model = "Flux model (Transformer) to load"
+    sd3_model = "SD3 model (MMDiTX) to load"
     sdxl_main_model = "SDXL Main model (UNet, VAE, CLIP1, CLIP2) to load"
     sdxl_refiner_model = "SDXL Refiner Main Modde (UNet, VAE, CLIP2) to load"
     onnx_main_model = "ONNX Main model (UNet, VAE, CLIP) to load"

invokeai/app/invocations/flux_model_loader.py

@@ -0,0 +1,86 @@
+from typing import Literal
+
+from invokeai.app.invocations.baseinvocation import (
+    BaseInvocation,
+    BaseInvocationOutput,
+    Classification,
+    invocation,
+    invocation_output,
+)
+from invokeai.app.invocations.fields import FieldDescriptions, Input, InputField, OutputField, UIType
+from invokeai.app.invocations.model import CLIPField, ModelIdentifierField, T5EncoderField, TransformerField, VAEField
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.flux.util import max_seq_lengths
+from invokeai.backend.model_manager.config import CheckpointConfigBase, SubModelType
+
+
+@invocation_output("flux_model_loader_output")
+class FluxModelLoaderOutput(BaseInvocationOutput):
+    """Flux base model loader output"""
+
+    transformer: TransformerField = OutputField(description=FieldDescriptions.transformer, title="Transformer")
+    clip: CLIPField = OutputField(description=FieldDescriptions.clip, title="CLIP")
+    t5_encoder: T5EncoderField = OutputField(description=FieldDescriptions.t5_encoder, title="T5 Encoder")
+    vae: VAEField = OutputField(description=FieldDescriptions.vae, title="VAE")
+    max_seq_len: Literal[256, 512] = OutputField(
+        description="The max sequence length to used for the T5 encoder. (256 for schnell transformer, 512 for dev transformer)",
+        title="Max Seq Length",
+    )
+
+
+@invocation(
+    "flux_model_loader",
+    title="Flux Main Model",
+    tags=["model", "flux"],
+    category="model",
+    version="1.0.4",
+    classification=Classification.Prototype,
+)
+class FluxModelLoaderInvocation(BaseInvocation):
+    """Loads a flux base model, outputting its submodels."""
+
+    model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.flux_model,
+        ui_type=UIType.FluxMainModel,
+        input=Input.Direct,
+    )
+
+    t5_encoder_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.t5_encoder, ui_type=UIType.T5EncoderModel, input=Input.Direct, title="T5 Encoder"
+    )
+
+    clip_embed_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.clip_embed_model,
+        ui_type=UIType.CLIPEmbedModel,
+        input=Input.Direct,
+        title="CLIP Embed",
+    )
+
+    vae_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.vae_model, ui_type=UIType.FluxVAEModel, title="VAE"
+    )
+
+    def invoke(self, context: InvocationContext) -> FluxModelLoaderOutput:
+        for key in [self.model.key, self.t5_encoder_model.key, self.clip_embed_model.key, self.vae_model.key]:
+            if not context.models.exists(key):
+                raise ValueError(f"Unknown model: {key}")
+
+        transformer = self.model.model_copy(update={"submodel_type": SubModelType.Transformer})
+        vae = self.vae_model.model_copy(update={"submodel_type": SubModelType.VAE})
+
+        tokenizer = self.clip_embed_model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
+        clip_encoder = self.clip_embed_model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
+
+        tokenizer2 = self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.Tokenizer2})
+        t5_encoder = self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.TextEncoder2})
+
+        transformer_config = context.models.get_config(transformer)
+        assert isinstance(transformer_config, CheckpointConfigBase)
+
+        return FluxModelLoaderOutput(
+            transformer=TransformerField(transformer=transformer, loras=[]),
+            clip=CLIPField(tokenizer=tokenizer, text_encoder=clip_encoder, loras=[], skipped_layers=0),
+            t5_encoder=T5EncoderField(tokenizer=tokenizer2, text_encoder=t5_encoder),
+            vae=VAEField(vae=vae),
+            max_seq_len=max_seq_lengths[transformer_config.config_path],
+        )

invokeai/app/invocations/model.py

@@ -1,5 +1,5 @@
 import copy
-from typing import List, Literal, Optional
+from typing import List, Optional
 
 from pydantic import BaseModel, Field
 
@@ -13,11 +13,9 @@ from invokeai.app.invocations.baseinvocation import (
 from invokeai.app.invocations.fields import FieldDescriptions, Input, InputField, OutputField, UIType
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.app.shared.models import FreeUConfig
-from invokeai.backend.flux.util import max_seq_lengths
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
     BaseModelType,
-    CheckpointConfigBase,
     ModelType,
     SubModelType,
 )
@@ -139,78 +137,6 @@ class ModelIdentifierInvocation(BaseInvocation):
         return ModelIdentifierOutput(model=self.model)
 
 
-@invocation_output("flux_model_loader_output")
-class FluxModelLoaderOutput(BaseInvocationOutput):
-    """Flux base model loader output"""
-
-    transformer: TransformerField = OutputField(description=FieldDescriptions.transformer, title="Transformer")
-    clip: CLIPField = OutputField(description=FieldDescriptions.clip, title="CLIP")
-    t5_encoder: T5EncoderField = OutputField(description=FieldDescriptions.t5_encoder, title="T5 Encoder")
-    vae: VAEField = OutputField(description=FieldDescriptions.vae, title="VAE")
-    max_seq_len: Literal[256, 512] = OutputField(
-        description="The max sequence length to used for the T5 encoder. (256 for schnell transformer, 512 for dev transformer)",
-        title="Max Seq Length",
-    )
-
-
-@invocation(
-    "flux_model_loader",
-    title="Flux Main Model",
-    tags=["model", "flux"],
-    category="model",
-    version="1.0.4",
-    classification=Classification.Prototype,
-)
-class FluxModelLoaderInvocation(BaseInvocation):
-    """Loads a flux base model, outputting its submodels."""
-
-    model: ModelIdentifierField = InputField(
-        description=FieldDescriptions.flux_model,
-        ui_type=UIType.FluxMainModel,
-        input=Input.Direct,
-    )
-
-    t5_encoder_model: ModelIdentifierField = InputField(
-        description=FieldDescriptions.t5_encoder, ui_type=UIType.T5EncoderModel, input=Input.Direct, title="T5 Encoder"
-    )
-
-    clip_embed_model: ModelIdentifierField = InputField(
-        description=FieldDescriptions.clip_embed_model,
-        ui_type=UIType.CLIPEmbedModel,
-        input=Input.Direct,
-        title="CLIP Embed",
-    )
-
-    vae_model: ModelIdentifierField = InputField(
-        description=FieldDescriptions.vae_model, ui_type=UIType.FluxVAEModel, title="VAE"
-    )
-
-    def invoke(self, context: InvocationContext) -> FluxModelLoaderOutput:
-        for key in [self.model.key, self.t5_encoder_model.key, self.clip_embed_model.key, self.vae_model.key]:
-            if not context.models.exists(key):
-                raise ValueError(f"Unknown model: {key}")
-
-        transformer = self.model.model_copy(update={"submodel_type": SubModelType.Transformer})
-        vae = self.vae_model.model_copy(update={"submodel_type": SubModelType.VAE})
-
-        tokenizer = self.clip_embed_model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
-        clip_encoder = self.clip_embed_model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
-
-        tokenizer2 = self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.Tokenizer2})
-        t5_encoder = self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.TextEncoder2})
-
-        transformer_config = context.models.get_config(transformer)
-        assert isinstance(transformer_config, CheckpointConfigBase)
-
-        return FluxModelLoaderOutput(
-            transformer=TransformerField(transformer=transformer, loras=[]),
-            clip=CLIPField(tokenizer=tokenizer, text_encoder=clip_encoder, loras=[], skipped_layers=0),
-            t5_encoder=T5EncoderField(tokenizer=tokenizer2, text_encoder=t5_encoder),
-            vae=VAEField(vae=vae),
-            max_seq_len=max_seq_lengths[transformer_config.config_path],
-        )
-
-
 @invocation(
     "main_model_loader",
     title="Main Model",

invokeai/app/invocations/sd3_model_loader.py

@@ -0,0 +1,102 @@
+from invokeai.app.invocations.baseinvocation import (
+    BaseInvocation,
+    BaseInvocationOutput,
+    Classification,
+    invocation,
+    invocation_output,
+)
+from invokeai.app.invocations.fields import FieldDescriptions, Input, InputField, OutputField, UIType
+from invokeai.app.invocations.model import CLIPField, ModelIdentifierField, T5EncoderField, TransformerField, VAEField
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.model_manager.config import CheckpointConfigBase, SubModelType
+
+
+@invocation_output("sd3_model_loader_output")
+class Sd3ModelLoaderOutput(BaseInvocationOutput):
+    """SD3 base model loader output."""
+
+    mmditx: TransformerField = OutputField(description=FieldDescriptions.mmditx, title="MMDiTX")
+    clip_l: CLIPField = OutputField(description=FieldDescriptions.clip, title="CLIP L")
+    clip_g: CLIPField = OutputField(description=FieldDescriptions.clip, title="CLIP G")
+    t5_encoder: T5EncoderField = OutputField(description=FieldDescriptions.t5_encoder, title="T5 Encoder")
+    vae: VAEField = OutputField(description=FieldDescriptions.vae, title="VAE")
+
+
+@invocation(
+    "sd3_model_loader",
+    title="SD3 Main Model",
+    tags=["model", "sd3"],
+    category="model",
+    version="1.0.0",
+    classification=Classification.Prototype,
+)
+class Sd3ModelLoaderInvocation(BaseInvocation):
+    """Loads a SD3 base model, outputting its submodels."""
+
+    # TODO(ryand): Create a UIType.Sd3MainModelField to use here.
+    model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.sd3_model,
+        ui_type=UIType.MainModel,
+        input=Input.Direct,
+    )
+
+    # TODO(ryand): Make the text encoders optional.
+    # Note: The text encoders are optional for SD3. The model was trained with dropout, so any can be left out at
+    # inference time. Typically, only the T5 encoder is omitted, since it is the largest by far.
+    t5_encoder_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.t5_encoder, ui_type=UIType.T5EncoderModel, input=Input.Direct, title="T5 Encoder"
+    )
+
+    clip_l_embed_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.clip_embed_model,
+        ui_type=UIType.CLIPEmbedModel,
+        input=Input.Direct,
+        title="CLIP L Embed",
+    )
+
+    clip_g_embed_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.clip_embed_model,
+        ui_type=UIType.CLIPEmbedModel,
+        input=Input.Direct,
+        title="CLIP G Embed",
+    )
+
+    # TODO(ryand): Create a UIType.Sd3VaModelField to use here.
+    vae_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.vae_model, ui_type=UIType.VAEModel, title="VAE"
+    )
+
+    def invoke(self, context: InvocationContext) -> Sd3ModelLoaderOutput:
+        for key in [
+            self.model.key,
+            self.t5_encoder_model.key,
+            self.clip_l_embed_model.key,
+            self.clip_g_embed_model.key,
+            self.vae_model.key,
+        ]:
+            if not context.models.exists(key):
+                raise ValueError(f"Unknown model: {key}")
+
+        # TODO(ryand): Figure out the sub-model types for SD3.
+        mmditx = self.model.model_copy(update={"submodel_type": SubModelType.Transformer})
+        vae = self.vae_model.model_copy(update={"submodel_type": SubModelType.VAE})
+
+        tokenizer_l = self.clip_l_embed_model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
+        clip_encoder_l = self.clip_l_embed_model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
+
+        tokenizer_g = self.clip_g_embed_model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
+        clip_encoder_g = self.clip_g_embed_model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
+
+        tokenizer_t5 = self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.Tokenizer2})
+        t5_encoder = self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.TextEncoder2})
+
+        transformer_config = context.models.get_config(mmditx)
+        assert isinstance(transformer_config, CheckpointConfigBase)
+
+        return Sd3ModelLoaderOutput(
+            mmditx=TransformerField(transformer=mmditx, loras=[]),
+            clip_l=CLIPField(tokenizer=tokenizer_l, text_encoder=clip_encoder_l, loras=[], skipped_layers=0),
+            clip_g=CLIPField(tokenizer=tokenizer_g, text_encoder=clip_encoder_g, loras=[], skipped_layers=0),
+            t5_encoder=T5EncoderField(tokenizer=tokenizer_t5, text_encoder=t5_encoder),
+            vae=VAEField(vae=vae),
+        )

invokeai/backend/model_manager/config.py

@@ -53,6 +53,8 @@ class BaseModelType(str, Enum):
     Any = "any"
     StableDiffusion1 = "sd-1"
     StableDiffusion2 = "sd-2"
+    # TODO(ryand): Should this just be StableDiffusion3?
+    StableDiffusion35 = "sd-3.5"
     StableDiffusionXL = "sdxl"
     StableDiffusionXLRefiner = "sdxl-refiner"
     Flux = "flux"

invokeai/backend/model_manager/load/model_loaders/sd3.py

@@ -0,0 +1,55 @@
+from pathlib import Path
+from typing import Optional
+
+from invokeai.backend.model_manager.config import (
+    AnyModel,
+    AnyModelConfig,
+    BaseModelType,
+    CheckpointConfigBase,
+    MainCheckpointConfig,
+    ModelFormat,
+    ModelType,
+    SubModelType,
+)
+from invokeai.backend.model_manager.load.load_default import ModelLoader
+from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+
+
+@ModelLoaderRegistry.register(base=BaseModelType.StableDiffusion35, type=ModelType.Main, format=ModelFormat.Checkpoint)
+class FluxCheckpointModel(ModelLoader):
+    """Class to load main models."""
+
+    def _load_model(
+        self,
+        config: AnyModelConfig,
+        submodel_type: Optional[SubModelType] = None,
+    ) -> AnyModel:
+        if not isinstance(config, CheckpointConfigBase):
+            raise ValueError("Only CheckpointConfigBase models are currently supported here.")
+
+        match submodel_type:
+            case SubModelType.Transformer:
+                return self._load_from_singlefile(config)
+
+        raise ValueError(
+            f"Only Transformer submodels are currently supported. Received: {submodel_type.value if submodel_type else 'None'}"
+        )
+
+    def _load_from_singlefile(
+        self,
+        config: AnyModelConfig,
+    ) -> AnyModel:
+        assert isinstance(config, MainCheckpointConfig)
+        model_path = Path(config.path)
+
+        # model = Flux(params[config.config_path])
+        # sd = load_file(model_path)
+        # if "model.diffusion_model.double_blocks.0.img_attn.norm.key_norm.scale" in sd:
+        #     sd = convert_bundle_to_flux_transformer_checkpoint(sd)
+        # new_sd_size = sum([ten.nelement() * torch.bfloat16.itemsize for ten in sd.values()])
+        # self._ram_cache.make_room(new_sd_size)
+        # for k in sd.keys():
+        #     # We need to cast to bfloat16 due to it being the only currently supported dtype for inference
+        #     sd[k] = sd[k].to(torch.bfloat16)
+        # model.load_state_dict(sd, assign=True)
+        return model

invokeai/backend/model_manager/probe.py

@@ -37,6 +37,7 @@ from invokeai.backend.model_manager.config import (
 from invokeai.backend.model_manager.util.model_util import lora_token_vector_length, read_checkpoint_meta
 from invokeai.backend.quantization.gguf.ggml_tensor import GGMLTensor
 from invokeai.backend.quantization.gguf.loaders import gguf_sd_loader
+from invokeai.backend.sd3.sd3_state_dict_utils import is_sd3_checkpoint
 from invokeai.backend.spandrel_image_to_image_model import SpandrelImageToImageModel
 from invokeai.backend.util.silence_warnings import SilenceWarnings
 
@@ -120,6 +121,7 @@ class ModelProbe(object):
         "T2IAdapter": ModelType.T2IAdapter,
         "CLIPModel": ModelType.CLIPEmbed,
         "CLIPTextModel": ModelType.CLIPEmbed,
+        "CLIPTextModelWithProjection": ModelType.CLIPEmbed,
         "T5EncoderModel": ModelType.T5Encoder,
         "FluxControlNetModel": ModelType.ControlNet,
     }
@@ -241,6 +243,11 @@ class ModelProbe(object):
         for key in [str(k) for k in ckpt.keys()]:
             if key.startswith(
                 (
+                    # The following prefixes appear when multiple models have been bundled together in a single file (I
+                    # believe the format originated in ComfyUI).
+                    # first_stage_model = VAE
+                    # cond_stage_model = Text Encoder
+                    # model.diffusion_model = UNet / Transformer
                     "cond_stage_model.",
                     "first_stage_model.",
                     "model.diffusion_model.",
@@ -397,6 +404,9 @@ class ModelProbe(object):
                     #   is used rather than attempting to support flux with separate model types and format
                     #   If changed in the future, please fix me
                     config_file = "flux-schnell"
+            elif base_type == BaseModelType.StableDiffusion35:
+                # TODO(ryand): Think about what to do here.
+                config_file = "sd3.5-large"
             else:
                 config_file = LEGACY_CONFIGS[base_type][variant_type]
                 if isinstance(config_file, dict):  # need another tier for sd-2.x models
@@ -516,7 +526,7 @@ class CheckpointProbeBase(ProbeBase):
     def get_variant_type(self) -> ModelVariantType:
         model_type = ModelProbe.get_model_type_from_checkpoint(self.model_path, self.checkpoint)
         base_type = self.get_base_type()
-        if model_type != ModelType.Main or base_type == BaseModelType.Flux:
+        if model_type != ModelType.Main or base_type in (BaseModelType.Flux, BaseModelType.StableDiffusion35):
             return ModelVariantType.Normal
         state_dict = self.checkpoint.get("state_dict") or self.checkpoint
         in_channels = state_dict["model.diffusion_model.input_blocks.0.0.weight"].shape[1]
@@ -541,6 +551,10 @@ class PipelineCheckpointProbe(CheckpointProbeBase):
             or "model.diffusion_model.double_blocks.0.img_attn.norm.key_norm.scale" in state_dict
         ):
             return BaseModelType.Flux
+
+        if is_sd3_checkpoint(state_dict):
+            return BaseModelType.StableDiffusion35
+
         key_name = "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_k.weight"
         if key_name in state_dict and state_dict[key_name].shape[-1] == 768:
             return BaseModelType.StableDiffusion1

invokeai/backend/sd3/mmditx.py

@@ -0,0 +1,891 @@
+# This file was originally copied from:
+# https://github.com/Stability-AI/sd3.5/blob/19bf11c4e1e37324c5aa5a61f010d4127848a09c/mmditx.py
+
+
+### This file contains impls for MM-DiT, the core model component of SD3
+
+import math
+from typing import Dict, List, Optional
+
+import numpy as np
+import torch
+from einops import rearrange, repeat
+
+from invokeai.backend.sd3.other_impls import Mlp, attention
+
+
+class PatchEmbed(torch.nn.Module):
+    """2D Image to Patch Embedding"""
+
+    def __init__(
+        self,
+        img_size: Optional[int] = 224,
+        patch_size: int = 16,
+        in_chans: int = 3,
+        embed_dim: int = 768,
+        flatten: bool = True,
+        bias: bool = True,
+        strict_img_size: bool = True,
+        dynamic_img_pad: bool = False,
+        dtype: torch.dtype | None = None,
+        device: torch.device | None = None,
+    ):
+        super().__init__()
+        self.patch_size = (patch_size, patch_size)
+        if img_size is not None:
+            self.img_size = (img_size, img_size)
+            self.grid_size = tuple([s // p for s, p in zip(self.img_size, self.patch_size, strict=False)])
+            self.num_patches = self.grid_size[0] * self.grid_size[1]
+        else:
+            self.img_size = None
+            self.grid_size = None
+            self.num_patches = None
+
+        # flatten spatial dim and transpose to channels last, kept for bwd compat
+        self.flatten = flatten
+        self.strict_img_size = strict_img_size
+        self.dynamic_img_pad = dynamic_img_pad
+
+        self.proj = torch.nn.Conv2d(
+            in_chans,
+            embed_dim,
+            kernel_size=patch_size,
+            stride=patch_size,
+            bias=bias,
+            dtype=dtype,
+            device=device,
+        )
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        x = self.proj(x)
+        if self.flatten:
+            x = x.flatten(2).transpose(1, 2)  # NCHW -> NLC
+        return x
+
+
+def modulate(x: torch.Tensor, shift: torch.Tensor | None, scale: torch.Tensor) -> torch.Tensor:
+    if shift is None:
+        shift = torch.zeros_like(scale)
+    return x * (1 + scale.unsqueeze(1)) + shift.unsqueeze(1)
+
+
+#################################################################################
+#                   Sine/Cosine Positional Embedding Functions                  #
+#################################################################################
+
+
+def get_2d_sincos_pos_embed(
+    embed_dim: int,
+    grid_size: int,
+    cls_token: bool = False,
+    extra_tokens: int = 0,
+    scaling_factor: Optional[float] = None,
+    offset: Optional[float] = None,
+):
+    """
+    grid_size: int of the grid height and width
+    return:
+    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)
+    """
+    grid_h = np.arange(grid_size, dtype=np.float32)
+    grid_w = np.arange(grid_size, dtype=np.float32)
+    grid = np.meshgrid(grid_w, grid_h)  # here w goes first
+    grid = np.stack(grid, axis=0)
+    if scaling_factor is not None:
+        grid = grid / scaling_factor
+    if offset is not None:
+        grid = grid - offset
+    grid = grid.reshape([2, 1, grid_size, grid_size])
+    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)
+    if cls_token and extra_tokens > 0:
+        pos_embed = np.concatenate([np.zeros([extra_tokens, embed_dim]), pos_embed], axis=0)
+    return pos_embed
+
+
+def get_2d_sincos_pos_embed_from_grid(embed_dim: int, grid):
+    assert embed_dim % 2 == 0
+    # use half of dimensions to encode grid_h
+    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)
+    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)
+    emb = np.concatenate([emb_h, emb_w], axis=1)  # (H*W, D)
+    return emb
+
+
+def get_1d_sincos_pos_embed_from_grid(embed_dim: int, pos):
+    """
+    embed_dim: output dimension for each position
+    pos: a list of positions to be encoded: size (M,)
+    out: (M, D)
+    """
+    assert embed_dim % 2 == 0
+    omega = np.arange(embed_dim // 2, dtype=np.float64)
+    omega /= embed_dim / 2.0
+    omega = 1.0 / 10000**omega  # (D/2,)
+    pos = pos.reshape(-1)  # (M,)
+    out = np.einsum("m,d->md", pos, omega)  # (M, D/2), outer product
+    emb_sin = np.sin(out)  # (M, D/2)
+    emb_cos = np.cos(out)  # (M, D/2)
+    return np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)
+
+
+#################################################################################
+#               Embedding Layers for Timesteps and Class Labels                 #
+#################################################################################
+
+
+class TimestepEmbedder(torch.nn.Module):
+    """Embeds scalar timesteps into vector representations."""
+
+    def __init__(self, hidden_size, frequency_embedding_size=256, dtype=None, device=None):
+        super().__init__()
+        self.mlp = torch.nn.Sequential(
+            torch.nn.Linear(
+                frequency_embedding_size,
+                hidden_size,
+                bias=True,
+                dtype=dtype,
+                device=device,
+            ),
+            torch.nn.SiLU(),
+            torch.nn.Linear(hidden_size, hidden_size, bias=True, dtype=dtype, device=device),
+        )
+        self.frequency_embedding_size = frequency_embedding_size
+
+    @staticmethod
+    def timestep_embedding(t, dim, max_period=10000):
+        """
+        Create sinusoidal timestep embeddings.
+        :param t: 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, D) Tensor of positional embeddings.
+        """
+        half = dim // 2
+        freqs = torch.exp(-math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half).to(
+            device=t.device
+        )
+        args = t[:, None].float() * 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)
+        if torch.is_floating_point(t):
+            embedding = embedding.to(dtype=t.dtype)
+        return embedding
+
+    def forward(self, t, dtype, **kwargs):
+        t_freq = self.timestep_embedding(t, self.frequency_embedding_size).to(dtype)
+        t_emb = self.mlp(t_freq)
+        return t_emb
+
+
+class VectorEmbedder(torch.nn.Module):
+    """Embeds a flat vector of dimension input_dim"""
+
+    def __init__(self, input_dim: int, hidden_size: int, dtype=None, device=None):
+        super().__init__()
+        self.mlp = torch.nn.Sequential(
+            torch.nn.Linear(input_dim, hidden_size, bias=True, dtype=dtype, device=device),
+            torch.nn.SiLU(),
+            torch.nn.Linear(hidden_size, hidden_size, bias=True, dtype=dtype, device=device),
+        )
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        return self.mlp(x)
+
+
+#################################################################################
+#                                 Core DiT Model                                #
+#################################################################################
+
+
+def split_qkv(qkv, head_dim):
+    qkv = qkv.reshape(qkv.shape[0], qkv.shape[1], 3, -1, head_dim).movedim(2, 0)
+    return qkv[0], qkv[1], qkv[2]
+
+
+def optimized_attention(qkv, num_heads):
+    return attention(qkv[0], qkv[1], qkv[2], num_heads)
+
+
+class SelfAttention(torch.nn.Module):
+    ATTENTION_MODES = ("xformers", "torch", "torch-hb", "math", "debug")
+
+    def __init__(
+        self,
+        dim: int,
+        num_heads: int = 8,
+        qkv_bias: bool = False,
+        qk_scale: Optional[float] = None,
+        attn_mode: str = "xformers",
+        pre_only: bool = False,
+        qk_norm: Optional[str] = None,
+        rmsnorm: bool = False,
+        dtype=None,
+        device=None,
+    ):
+        super().__init__()
+        self.num_heads = num_heads
+        self.head_dim = dim // num_heads
+
+        self.qkv = torch.nn.Linear(dim, dim * 3, bias=qkv_bias, dtype=dtype, device=device)
+        if not pre_only:
+            self.proj = torch.nn.Linear(dim, dim, dtype=dtype, device=device)
+        assert attn_mode in self.ATTENTION_MODES
+        self.attn_mode = attn_mode
+        self.pre_only = pre_only
+
+        if qk_norm == "rms":
+            self.ln_q = RMSNorm(
+                self.head_dim,
+                elementwise_affine=True,
+                eps=1.0e-6,
+                dtype=dtype,
+                device=device,
+            )
+            self.ln_k = RMSNorm(
+                self.head_dim,
+                elementwise_affine=True,
+                eps=1.0e-6,
+                dtype=dtype,
+                device=device,
+            )
+        elif qk_norm == "ln":
+            self.ln_q = torch.nn.LayerNorm(
+                self.head_dim,
+                elementwise_affine=True,
+                eps=1.0e-6,
+                dtype=dtype,
+                device=device,
+            )
+            self.ln_k = torch.nn.LayerNorm(
+                self.head_dim,
+                elementwise_affine=True,
+                eps=1.0e-6,
+                dtype=dtype,
+                device=device,
+            )
+        elif qk_norm is None:
+            self.ln_q = torch.nn.Identity()
+            self.ln_k = torch.nn.Identity()
+        else:
+            raise ValueError(qk_norm)
+
+    def pre_attention(self, x: torch.Tensor):
+        B, L, C = x.shape
+        qkv = self.qkv(x)
+        q, k, v = split_qkv(qkv, self.head_dim)
+        q = self.ln_q(q).reshape(q.shape[0], q.shape[1], -1)
+        k = self.ln_k(k).reshape(q.shape[0], q.shape[1], -1)
+        return (q, k, v)
+
+    def post_attention(self, x: torch.Tensor) -> torch.Tensor:
+        assert not self.pre_only
+        x = self.proj(x)
+        return x
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        (q, k, v) = self.pre_attention(x)
+        x = attention(q, k, v, self.num_heads)
+        x = self.post_attention(x)
+        return x
+
+
+class RMSNorm(torch.nn.Module):
+    def __init__(
+        self,
+        dim: int,
+        elementwise_affine: bool = False,
+        eps: float = 1e-6,
+        device=None,
+        dtype=None,
+    ):
+        """
+        Initialize the RMSNorm normalization layer.
+        Args:
+            dim (int): The dimension of the input tensor.
+            eps (float, optional): A small value added to the denominator for numerical stability. Default is 1e-6.
+        Attributes:
+            eps (float): A small value added to the denominator for numerical stability.
+            weight (torch.nn.Parameter): Learnable scaling parameter.
+        """
+        super().__init__()
+        self.eps = eps
+        self.learnable_scale = elementwise_affine
+        if self.learnable_scale:
+            self.weight = torch.nn.Parameter(torch.empty(dim, device=device, dtype=dtype))
+        else:
+            self.register_parameter("weight", None)
+
+    def _norm(self, x):
+        """
+        Apply the RMSNorm normalization to the input tensor.
+        Args:
+            x (torch.Tensor): The input tensor.
+        Returns:
+            torch.Tensor: The normalized tensor.
+        """
+        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)
+
+    def forward(self, x):
+        """
+        Forward pass through the RMSNorm layer.
+        Args:
+            x (torch.Tensor): The input tensor.
+        Returns:
+            torch.Tensor: The output tensor after applying RMSNorm.
+        """
+        x = self._norm(x)
+        if self.learnable_scale:
+            return x * self.weight.to(device=x.device, dtype=x.dtype)
+        else:
+            return x
+
+
+class SwiGLUFeedForward(torch.nn.Module):
+    def __init__(
+        self,
+        dim: int,
+        hidden_dim: int,
+        multiple_of: int,
+        ffn_dim_multiplier: Optional[float] = None,
+    ):
+        """
+        Initialize the FeedForward module.
+
+        Args:
+            dim (int): Input dimension.
+            hidden_dim (int): Hidden dimension of the feedforward layer.
+            multiple_of (int): Value to ensure hidden dimension is a multiple of this value.
+            ffn_dim_multiplier (float, optional): Custom multiplier for hidden dimension. Defaults to None.
+
+        Attributes:
+            w1 (ColumnParallelLinear): Linear transformation for the first layer.
+            w2 (RowParallelLinear): Linear transformation for the second layer.
+            w3 (ColumnParallelLinear): Linear transformation for the third layer.
+
+        """
+        super().__init__()
+        hidden_dim = int(2 * hidden_dim / 3)
+        # custom dim factor multiplier
+        if ffn_dim_multiplier is not None:
+            hidden_dim = int(ffn_dim_multiplier * hidden_dim)
+        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)
+
+        self.w1 = torch.nn.Linear(dim, hidden_dim, bias=False)
+        self.w2 = torch.nn.Linear(hidden_dim, dim, bias=False)
+        self.w3 = torch.nn.Linear(dim, hidden_dim, bias=False)
+
+    def forward(self, x):
+        return self.w2(torch.nn.functional.silu(self.w1(x)) * self.w3(x))
+
+
+class DismantledBlock(torch.nn.Module):
+    """A DiT block with gated adaptive layer norm (adaLN) conditioning."""
+
+    ATTENTION_MODES = ("xformers", "torch", "torch-hb", "math", "debug")
+
+    def __init__(
+        self,
+        hidden_size: int,
+        num_heads: int,
+        mlp_ratio: float = 4.0,
+        attn_mode: str = "xformers",
+        qkv_bias: bool = False,
+        pre_only: bool = False,
+        rmsnorm: bool = False,
+        scale_mod_only: bool = False,
+        swiglu: bool = False,
+        qk_norm: Optional[str] = None,
+        x_block_self_attn: bool = False,
+        dtype=None,
+        device=None,
+        **block_kwargs,
+    ):
+        super().__init__()
+        assert attn_mode in self.ATTENTION_MODES
+        if not rmsnorm:
+            self.norm1 = torch.nn.LayerNorm(
+                hidden_size,
+                elementwise_affine=False,
+                eps=1e-6,
+                dtype=dtype,
+                device=device,
+            )
+        else:
+            self.norm1 = RMSNorm(hidden_size, elementwise_affine=False, eps=1e-6)
+        self.attn = SelfAttention(
+            dim=hidden_size,
+            num_heads=num_heads,
+            qkv_bias=qkv_bias,
+            attn_mode=attn_mode,
+            pre_only=pre_only,
+            qk_norm=qk_norm,
+            rmsnorm=rmsnorm,
+            dtype=dtype,
+            device=device,
+        )
+        if x_block_self_attn:
+            assert not pre_only
+            assert not scale_mod_only
+            self.x_block_self_attn = True
+            self.attn2 = SelfAttention(
+                dim=hidden_size,
+                num_heads=num_heads,
+                qkv_bias=qkv_bias,
+                attn_mode=attn_mode,
+                pre_only=False,
+                qk_norm=qk_norm,
+                rmsnorm=rmsnorm,
+                dtype=dtype,
+                device=device,
+            )
+        else:
+            self.x_block_self_attn = False
+        if not pre_only:
+            if not rmsnorm:
+                self.norm2 = torch.nn.LayerNorm(
+                    hidden_size,
+                    elementwise_affine=False,
+                    eps=1e-6,
+                    dtype=dtype,
+                    device=device,
+                )
+            else:
+                self.norm2 = RMSNorm(hidden_size, elementwise_affine=False, eps=1e-6)
+        mlp_hidden_dim = int(hidden_size * mlp_ratio)
+        if not pre_only:
+            if not swiglu:
+                self.mlp = Mlp(
+                    in_features=hidden_size,
+                    hidden_features=mlp_hidden_dim,
+                    act_layer=torch.nn.GELU(approximate="tanh"),
+                    dtype=dtype,
+                    device=device,
+                )
+            else:
+                self.mlp = SwiGLUFeedForward(dim=hidden_size, hidden_dim=mlp_hidden_dim, multiple_of=256)
+        self.scale_mod_only = scale_mod_only
+        if x_block_self_attn:
+            assert not pre_only
+            assert not scale_mod_only
+            n_mods = 9
+        elif not scale_mod_only:
+            n_mods = 6 if not pre_only else 2
+        else:
+            n_mods = 4 if not pre_only else 1
+        self.adaLN_modulation = torch.nn.Sequential(
+            torch.nn.SiLU(),
+            torch.nn.Linear(hidden_size, n_mods * hidden_size, bias=True, dtype=dtype, device=device),
+        )
+        self.pre_only = pre_only
+
+    def pre_attention(self, x: torch.Tensor, c: torch.Tensor):
+        assert x is not None, "pre_attention called with None input"
+        if not self.pre_only:
+            if not self.scale_mod_only:
+                shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = self.adaLN_modulation(c).chunk(
+                    6, dim=1
+                )
+            else:
+                shift_msa = None
+                shift_mlp = None
+                scale_msa, gate_msa, scale_mlp, gate_mlp = self.adaLN_modulation(c).chunk(4, dim=1)
+            qkv = self.attn.pre_attention(modulate(self.norm1(x), shift_msa, scale_msa))
+            return qkv, (x, gate_msa, shift_mlp, scale_mlp, gate_mlp)
+        else:
+            if not self.scale_mod_only:
+                shift_msa, scale_msa = self.adaLN_modulation(c).chunk(2, dim=1)
+            else:
+                shift_msa = None
+                scale_msa = self.adaLN_modulation(c)
+            qkv = self.attn.pre_attention(modulate(self.norm1(x), shift_msa, scale_msa))
+            return qkv, None
+
+    def post_attention(
+        self,
+        attn: torch.Tensor,
+        x: torch.Tensor,
+        gate_msa: torch.Tensor,
+        shift_mlp: torch.Tensor,
+        scale_mlp: torch.Tensor,
+        gate_mlp: torch.Tensor,
+    ) -> torch.Tensor:
+        assert not self.pre_only
+        x = x + gate_msa.unsqueeze(1) * self.attn.post_attention(attn)
+        x = x + gate_mlp.unsqueeze(1) * self.mlp(modulate(self.norm2(x), shift_mlp, scale_mlp))
+        return x
+
+    def pre_attention_x(
+        self, x: torch.Tensor, c: torch.Tensor
+    ) -> tuple[
+        tuple[torch.Tensor, torch.Tensor, torch.Tensor],
+        tuple[torch.Tensor, torch.Tensor, torch.Tensor],
+        tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor],
+    ]:
+        assert self.x_block_self_attn
+        (
+            shift_msa,
+            scale_msa,
+            gate_msa,
+            shift_mlp,
+            scale_mlp,
+            gate_mlp,
+            shift_msa2,
+            scale_msa2,
+            gate_msa2,
+        ) = self.adaLN_modulation(c).chunk(9, dim=1)
+        x_norm = self.norm1(x)
+        qkv = self.attn.pre_attention(modulate(x_norm, shift_msa, scale_msa))
+        qkv2 = self.attn2.pre_attention(modulate(x_norm, shift_msa2, scale_msa2))
+        return (
+            qkv,
+            qkv2,
+            (
+                x,
+                gate_msa,
+                shift_mlp,
+                scale_mlp,
+                gate_mlp,
+                gate_msa2,
+            ),
+        )
+
+    def post_attention_x(
+        self,
+        attn: torch.Tensor,
+        attn2: torch.Tensor,
+        x: torch.Tensor,
+        gate_msa: torch.Tensor,
+        shift_mlp: torch.Tensor,
+        scale_mlp: torch.Tensor,
+        gate_mlp: torch.Tensor,
+        gate_msa2: torch.Tensor,
+        attn1_dropout: float = 0.0,
+    ):
+        assert not self.pre_only
+        if attn1_dropout > 0.0:
+            # Use torch.bernoulli to implement dropout, only dropout the batch dimension
+            attn1_dropout = torch.bernoulli(torch.full((attn.size(0), 1, 1), 1 - attn1_dropout, device=attn.device))
+            attn_ = gate_msa.unsqueeze(1) * self.attn.post_attention(attn) * attn1_dropout
+        else:
+            attn_ = gate_msa.unsqueeze(1) * self.attn.post_attention(attn)
+        x = x + attn_
+        attn2_ = gate_msa2.unsqueeze(1) * self.attn2.post_attention(attn2)
+        x = x + attn2_
+        mlp_ = gate_mlp.unsqueeze(1) * self.mlp(modulate(self.norm2(x), shift_mlp, scale_mlp))
+        x = x + mlp_
+        return x, (gate_msa, gate_msa2, gate_mlp, attn_, attn2_)
+
+    def forward(self, x: torch.Tensor, c: torch.Tensor):
+        assert not self.pre_only
+        if self.x_block_self_attn:
+            (q, k, v), (q2, k2, v2), intermediates = self.pre_attention_x(x, c)
+            attn = attention(q, k, v, self.attn.num_heads)
+            attn2 = attention(q2, k2, v2, self.attn2.num_heads)
+            return self.post_attention_x(attn, attn2, *intermediates)
+        else:
+            (q, k, v), intermediates = self.pre_attention(x, c)
+            attn = attention(q, k, v, self.attn.num_heads)
+            return self.post_attention(attn, *intermediates)
+
+
+def block_mixing(
+    context: torch.Tensor, x: torch.Tensor, context_block: DismantledBlock, x_block: DismantledBlock, c: torch.Tensor
+):
+    assert context is not None, "block_mixing called with None context"
+    context_qkv, context_intermediates = context_block.pre_attention(context, c)
+
+    if x_block.x_block_self_attn:
+        x_qkv, x_qkv2, x_intermediates = x_block.pre_attention_x(x, c)
+    else:
+        x_qkv, x_intermediates = x_block.pre_attention(x, c)
+
+    o: list[torch.Tensor] = []
+    for t in range(3):
+        o.append(torch.cat((context_qkv[t], x_qkv[t]), dim=1))
+    q, k, v = tuple(o)
+
+    attn = attention(q, k, v, x_block.attn.num_heads)
+    context_attn, x_attn = (
+        attn[:, : context_qkv[0].shape[1]],
+        attn[:, context_qkv[0].shape[1] :],
+    )
+
+    if not context_block.pre_only:
+        context = context_block.post_attention(context_attn, *context_intermediates)
+    else:
+        context = None
+
+    if x_block.x_block_self_attn:
+        x_q2, x_k2, x_v2 = x_qkv2
+        attn2 = attention(x_q2, x_k2, x_v2, x_block.attn2.num_heads)
+    else:
+        x = x_block.post_attention(x_attn, *x_intermediates)
+
+    return context, x
+
+
+class JointBlock(torch.nn.Module):
+    """just a small wrapper to serve as a fsdp unit"""
+
+    def __init__(self, *args, **kwargs):
+        super().__init__()
+        pre_only = kwargs.pop("pre_only")
+        qk_norm = kwargs.pop("qk_norm", None)
+        x_block_self_attn = kwargs.pop("x_block_self_attn", False)
+        self.context_block = DismantledBlock(*args, pre_only=pre_only, qk_norm=qk_norm, **kwargs)
+        self.x_block = DismantledBlock(
+            *args,
+            pre_only=False,
+            qk_norm=qk_norm,
+            x_block_self_attn=x_block_self_attn,
+            **kwargs,
+        )
+
+    def forward(self, *args, **kwargs):
+        return block_mixing(*args, context_block=self.context_block, x_block=self.x_block, **kwargs)
+
+
+class FinalLayer(torch.nn.Module):
+    """
+    The final layer of DiT.
+    """
+
+    def __init__(
+        self,
+        hidden_size: int,
+        patch_size: int,
+        out_channels: int,
+        total_out_channels: Optional[int] = None,
+        dtype: Optional[torch.dtype] = None,
+        device: Optional[torch.device] = None,
+    ):
+        super().__init__()
+        self.norm_final = torch.nn.LayerNorm(
+            hidden_size, elementwise_affine=False, eps=1e-6, dtype=dtype, device=device
+        )
+        self.linear = (
+            torch.nn.Linear(
+                hidden_size,
+                patch_size * patch_size * out_channels,
+                bias=True,
+                dtype=dtype,
+                device=device,
+            )
+            if (total_out_channels is None)
+            else torch.nn.Linear(hidden_size, total_out_channels, bias=True, dtype=dtype, device=device)
+        )
+        self.adaLN_modulation = torch.nn.Sequential(
+            torch.nn.SiLU(),
+            torch.nn.Linear(hidden_size, 2 * hidden_size, bias=True, dtype=dtype, device=device),
+        )
+
+    def forward(self, x: torch.Tensor, c: torch.Tensor) -> torch.Tensor:
+        shift, scale = self.adaLN_modulation(c).chunk(2, dim=1)
+        x = modulate(self.norm_final(x), shift, scale)
+        x = self.linear(x)
+        return x
+
+
+class MMDiTX(torch.nn.Module):
+    """Diffusion model with a Transformer backbone."""
+
+    def __init__(
+        self,
+        input_size: int | None = 32,
+        patch_size: int = 2,
+        in_channels: int = 4,
+        depth: int = 28,
+        mlp_ratio: float = 4.0,
+        learn_sigma: bool = False,
+        adm_in_channels: Optional[int] = None,
+        context_embedder_config: Optional[Dict] = None,
+        register_length: int = 0,
+        attn_mode: str = "torch",
+        rmsnorm: bool = False,
+        scale_mod_only: bool = False,
+        swiglu: bool = False,
+        out_channels: Optional[int] = None,
+        pos_embed_scaling_factor: Optional[float] = None,
+        pos_embed_offset: Optional[float] = None,
+        pos_embed_max_size: Optional[int] = None,
+        num_patches: Optional[int] = None,
+        qk_norm: Optional[str] = None,
+        x_block_self_attn_layers: Optional[List[int]] = None,
+        qkv_bias: bool = True,
+        dtype: Optional[torch.dtype] = None,
+        device: Optional[torch.device] = None,
+        verbose: bool = False,
+    ):
+        super().__init__()
+        if verbose:
+            print(
+                f"mmdit initializing with: {input_size=}, {patch_size=}, {in_channels=}, {depth=}, {mlp_ratio=}, {learn_sigma=}, {adm_in_channels=}, {context_embedder_config=}, {register_length=}, {attn_mode=}, {rmsnorm=}, {scale_mod_only=}, {swiglu=}, {out_channels=}, {pos_embed_scaling_factor=}, {pos_embed_offset=}, {pos_embed_max_size=}, {num_patches=}, {qk_norm=}, {qkv_bias=}, {dtype=}, {device=}"
+            )
+        self.dtype = dtype
+        self.learn_sigma = learn_sigma
+        self.in_channels = in_channels
+        default_out_channels = in_channels * 2 if learn_sigma else in_channels
+        self.out_channels = out_channels if out_channels is not None else default_out_channels
+        self.patch_size = patch_size
+        self.pos_embed_scaling_factor = pos_embed_scaling_factor
+        self.pos_embed_offset = pos_embed_offset
+        self.pos_embed_max_size = pos_embed_max_size
+        self.x_block_self_attn_layers = x_block_self_attn_layers or []
+
+        # apply magic --> this defines a head_size of 64
+        hidden_size = 64 * depth
+        num_heads = depth
+
+        self.num_heads = num_heads
+
+        self.x_embedder = PatchEmbed(
+            input_size,
+            patch_size,
+            in_channels,
+            hidden_size,
+            bias=True,
+            strict_img_size=self.pos_embed_max_size is None,
+            dtype=dtype,
+            device=device,
+        )
+        self.t_embedder = TimestepEmbedder(hidden_size, dtype=dtype, device=device)
+
+        if adm_in_channels is not None:
+            assert isinstance(adm_in_channels, int)
+            self.y_embedder = VectorEmbedder(adm_in_channels, hidden_size, dtype=dtype, device=device)
+
+        self.context_embedder = torch.nn.Identity()
+        if context_embedder_config is not None:
+            if context_embedder_config["target"] == "torch.nn.Linear":
+                self.context_embedder = torch.nn.Linear(**context_embedder_config["params"], dtype=dtype, device=device)
+
+        self.register_length = register_length
+        if self.register_length > 0:
+            self.register = torch.nn.Parameter(torch.randn(1, register_length, hidden_size, dtype=dtype, device=device))
+
+        # num_patches = self.x_embedder.num_patches
+        # Will use fixed sin-cos embedding:
+        # just use a buffer already
+        if num_patches is not None:
+            self.register_buffer(
+                "pos_embed",
+                torch.zeros(1, num_patches, hidden_size, dtype=dtype, device=device),
+            )
+        else:
+            self.pos_embed = None
+
+        self.joint_blocks = torch.nn.ModuleList(
+            [
+                JointBlock(
+                    hidden_size,
+                    num_heads,
+                    mlp_ratio=mlp_ratio,
+                    qkv_bias=qkv_bias,
+                    attn_mode=attn_mode,
+                    pre_only=i == depth - 1,
+                    rmsnorm=rmsnorm,
+                    scale_mod_only=scale_mod_only,
+                    swiglu=swiglu,
+                    qk_norm=qk_norm,
+                    x_block_self_attn=(i in self.x_block_self_attn_layers),
+                    dtype=dtype,
+                    device=device,
+                )
+                for i in range(depth)
+            ]
+        )
+
+        self.final_layer = FinalLayer(hidden_size, patch_size, self.out_channels, dtype=dtype, device=device)
+
+    def cropped_pos_embed(self, hw: torch.Size) -> torch.Tensor:
+        assert self.pos_embed_max_size is not None
+        p = self.x_embedder.patch_size[0]
+        h, w = hw
+        # patched size
+        h = h // p
+        w = w // p
+        assert h <= self.pos_embed_max_size, (h, self.pos_embed_max_size)
+        assert w <= self.pos_embed_max_size, (w, self.pos_embed_max_size)
+        top = (self.pos_embed_max_size - h) // 2
+        left = (self.pos_embed_max_size - w) // 2
+        spatial_pos_embed: torch.Tensor = rearrange(
+            self.pos_embed,
+            "1 (h w) c -> 1 h w c",
+            h=self.pos_embed_max_size,
+            w=self.pos_embed_max_size,
+        )  # type: ignore Type checking does not correctly infer the type of the self.pos_embed buffer.
+        spatial_pos_embed = spatial_pos_embed[:, top : top + h, left : left + w, :]
+        spatial_pos_embed = rearrange(spatial_pos_embed, "1 h w c -> 1 (h w) c")
+        return spatial_pos_embed
+
+    def unpatchify(self, x: torch.Tensor, hw: Optional[torch.Size] = None) -> torch.Tensor:
+        """
+        x: (N, T, patch_size**2 * C)
+        imgs: (N, H, W, C)
+        """
+        c = self.out_channels
+        p = self.x_embedder.patch_size[0]
+        if hw is None:
+            h = w = int(x.shape[1] ** 0.5)
+        else:
+            h, w = hw
+            h = h // p
+            w = w // p
+        assert h * w == x.shape[1]
+
+        x = x.reshape(shape=(x.shape[0], h, w, p, p, c))
+        x = torch.einsum("nhwpqc->nchpwq", x)
+        imgs = x.reshape(shape=(x.shape[0], c, h * p, w * p))
+        return imgs
+
+    def forward_core_with_concat(
+        self,
+        x: torch.Tensor,
+        c_mod: torch.Tensor,
+        context: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        if self.register_length > 0:
+            context = torch.cat(
+                (
+                    repeat(self.register, "1 ... -> b ...", b=x.shape[0]),
+                    context if context is not None else torch.Tensor([]).type_as(x),
+                ),
+                1,
+            )
+
+        # context is B, L', D
+        # x is B, L, D
+        for block in self.joint_blocks:
+            context, x = block(context, x, c=c_mod)
+
+        x = self.final_layer(x, c_mod)  # (N, T, patch_size ** 2 * out_channels)
+        return x
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        t: torch.Tensor,
+        y: Optional[torch.Tensor] = None,
+        context: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        """
+        Forward pass of DiT.
+        x: (N, C, H, W) tensor of spatial inputs (images or latent representations of images)
+        t: (N,) tensor of diffusion timesteps
+        y: (N,) tensor of class labels
+        """
+        hw = x.shape[-2:]
+        x = self.x_embedder(x) + self.cropped_pos_embed(hw)
+        c = self.t_embedder(t, dtype=x.dtype)  # (N, D)
+        if y is not None:
+            y = self.y_embedder(y)  # (N, D)
+            c = c + y  # (N, D)
+
+        context = self.context_embedder(context)
+
+        x = self.forward_core_with_concat(x, c, context)
+
+        x = self.unpatchify(x, hw=hw)  # (N, out_channels, H, W)
+        return x

invokeai/backend/sd3/other_impls.py

@@ -0,0 +1,795 @@
+# This file was originally copied from:
+# https://github.com/Stability-AI/sd3.5/blob/19bf11c4e1e37324c5aa5a61f010d4127848a09c/other_impls.py
+
+### This file contains impls for underlying related models (CLIP, T5, etc)
+
+import math
+from typing import Callable, Optional
+
+import torch
+from transformers import CLIPTokenizer, T5TokenizerFast
+
+#################################################################################################
+### Core/Utility
+#################################################################################################
+
+
+def attention(
+    q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, heads: int, mask: Optional[torch.Tensor] = None
+) -> torch.Tensor:
+    """Convenience wrapper around a basic attention operation"""
+    b, _, dim_head = q.shape
+    dim_head //= heads
+    q, k, v = map(lambda t: t.view(b, -1, heads, dim_head).transpose(1, 2), (q, k, v))
+    out = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=mask, dropout_p=0.0, is_causal=False)
+    return out.transpose(1, 2).reshape(b, -1, heads * dim_head)
+
+
+class Mlp(torch.nn.Module):
+    """MLP as used in Vision Transformer, MLP-Mixer and related networks"""
+
+    def __init__(
+        self,
+        in_features: int,
+        hidden_features: Optional[int] = None,
+        out_features: Optional[int] = None,
+        act_layer: Callable[[torch.Tensor], torch.Tensor] | None = None,
+        bias: bool = True,
+        dtype: Optional[torch.dtype] = None,
+        device: Optional[torch.device] = None,
+    ):
+        super().__init__()
+        out_features = out_features or in_features
+        hidden_features = hidden_features or in_features
+        if act_layer is None:
+            act_layer = torch.nn.functional.gelu
+
+        self.fc1 = torch.nn.Linear(in_features, hidden_features, bias=bias, dtype=dtype, device=device)
+        self.act = act_layer
+        self.fc2 = torch.nn.Linear(hidden_features, out_features, bias=bias, dtype=dtype, device=device)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        x = self.fc1(x)
+        x = self.act(x)
+        x = self.fc2(x)
+        return x
+
+
+#################################################################################################
+### CLIP
+#################################################################################################
+
+
+class CLIPAttention(torch.nn.Module):
+    def __init__(self, embed_dim, heads, dtype, device):
+        super().__init__()
+        self.heads = heads
+        self.q_proj = torch.nn.Linear(embed_dim, embed_dim, bias=True, dtype=dtype, device=device)
+        self.k_proj = torch.nn.Linear(embed_dim, embed_dim, bias=True, dtype=dtype, device=device)
+        self.v_proj = torch.nn.Linear(embed_dim, embed_dim, bias=True, dtype=dtype, device=device)
+        self.out_proj = torch.nn.Linear(embed_dim, embed_dim, bias=True, dtype=dtype, device=device)
+
+    def forward(self, x, mask=None):
+        q = self.q_proj(x)
+        k = self.k_proj(x)
+        v = self.v_proj(x)
+        out = attention(q, k, v, self.heads, mask)
+        return self.out_proj(out)
+
+
+ACTIVATIONS = {
+    "quick_gelu": lambda a: a * torch.sigmoid(1.702 * a),
+    "gelu": torch.nn.functional.gelu,
+}
+
+
+class CLIPLayer(torch.nn.Module):
+    def __init__(
+        self,
+        embed_dim,
+        heads,
+        intermediate_size,
+        intermediate_activation,
+        dtype,
+        device,
+    ):
+        super().__init__()
+        self.layer_norm1 = torch.nn.LayerNorm(embed_dim, dtype=dtype, device=device)
+        self.self_attn = CLIPAttention(embed_dim, heads, dtype, device)
+        self.layer_norm2 = torch.nn.LayerNorm(embed_dim, dtype=dtype, device=device)
+        # self.mlp = CLIPMLP(embed_dim, intermediate_size, intermediate_activation, dtype, device)
+        self.mlp = Mlp(
+            embed_dim,
+            intermediate_size,
+            embed_dim,
+            act_layer=ACTIVATIONS[intermediate_activation],
+            dtype=dtype,
+            device=device,
+        )
+
+    def forward(self, x, mask=None):
+        x += self.self_attn(self.layer_norm1(x), mask)
+        x += self.mlp(self.layer_norm2(x))
+        return x
+
+
+class CLIPEncoder(torch.nn.Module):
+    def __init__(
+        self,
+        num_layers,
+        embed_dim,
+        heads,
+        intermediate_size,
+        intermediate_activation,
+        dtype,
+        device,
+    ):
+        super().__init__()
+        self.layers = torch.nn.ModuleList(
+            [
+                CLIPLayer(
+                    embed_dim,
+                    heads,
+                    intermediate_size,
+                    intermediate_activation,
+                    dtype,
+                    device,
+                )
+                for i in range(num_layers)
+            ]
+        )
+
+    def forward(self, x, mask=None, intermediate_output=None):
+        if intermediate_output is not None:
+            if intermediate_output < 0:
+                intermediate_output = len(self.layers) + intermediate_output
+        intermediate = None
+        for i, l in enumerate(self.layers):
+            x = l(x, mask)
+            if i == intermediate_output:
+                intermediate = x.clone()
+        return x, intermediate
+
+
+class CLIPEmbeddings(torch.nn.Module):
+    def __init__(self, embed_dim, vocab_size=49408, num_positions=77, dtype=None, device=None):
+        super().__init__()
+        self.token_embedding = torch.nn.Embedding(vocab_size, embed_dim, dtype=dtype, device=device)
+        self.position_embedding = torch.nn.Embedding(num_positions, embed_dim, dtype=dtype, device=device)
+
+    def forward(self, input_tokens):
+        return self.token_embedding(input_tokens) + self.position_embedding.weight
+
+
+class CLIPTextModel_(torch.nn.Module):
+    def __init__(self, config_dict, dtype, device):
+        num_layers = config_dict["num_hidden_layers"]
+        embed_dim = config_dict["hidden_size"]
+        heads = config_dict["num_attention_heads"]
+        intermediate_size = config_dict["intermediate_size"]
+        intermediate_activation = config_dict["hidden_act"]
+        super().__init__()
+        self.embeddings = CLIPEmbeddings(embed_dim, dtype=torch.float32, device=device)
+        self.encoder = CLIPEncoder(
+            num_layers,
+            embed_dim,
+            heads,
+            intermediate_size,
+            intermediate_activation,
+            dtype,
+            device,
+        )
+        self.final_layer_norm = torch.nn.LayerNorm(embed_dim, dtype=dtype, device=device)
+
+    def forward(self, input_tokens, intermediate_output=None, final_layer_norm_intermediate=True):
+        x = self.embeddings(input_tokens)
+        causal_mask = torch.empty(x.shape[1], x.shape[1], dtype=x.dtype, device=x.device).fill_(float("-inf")).triu_(1)
+        x, i = self.encoder(x, mask=causal_mask, intermediate_output=intermediate_output)
+        x = self.final_layer_norm(x)
+        if i is not None and final_layer_norm_intermediate:
+            i = self.final_layer_norm(i)
+        pooled_output = x[
+            torch.arange(x.shape[0], device=x.device),
+            input_tokens.to(dtype=torch.int, device=x.device).argmax(dim=-1),
+        ]
+        return x, i, pooled_output
+
+
+class CLIPTextModel(torch.nn.Module):
+    def __init__(self, config_dict, dtype, device):
+        super().__init__()
+        self.num_layers = config_dict["num_hidden_layers"]
+        self.text_model = CLIPTextModel_(config_dict, dtype, device)
+        embed_dim = config_dict["hidden_size"]
+        self.text_projection = torch.nn.Linear(embed_dim, embed_dim, bias=False, dtype=dtype, device=device)
+        self.text_projection.weight.copy_(torch.eye(embed_dim))
+        self.dtype = dtype
+
+    def get_input_embeddings(self):
+        return self.text_model.embeddings.token_embedding
+
+    def set_input_embeddings(self, embeddings):
+        self.text_model.embeddings.token_embedding = embeddings
+
+    def forward(self, *args, **kwargs):
+        x = self.text_model(*args, **kwargs)
+        out = self.text_projection(x[2])
+        return (x[0], x[1], out, x[2])
+
+
+def parse_parentheses(string):
+    result = []
+    current_item = ""
+    nesting_level = 0
+    for char in string:
+        if char == "(":
+            if nesting_level == 0:
+                if current_item:
+                    result.append(current_item)
+                    current_item = "("
+                else:
+                    current_item = "("
+            else:
+                current_item += char
+            nesting_level += 1
+        elif char == ")":
+            nesting_level -= 1
+            if nesting_level == 0:
+                result.append(current_item + ")")
+                current_item = ""
+            else:
+                current_item += char
+        else:
+            current_item += char
+    if current_item:
+        result.append(current_item)
+    return result
+
+
+def token_weights(string, current_weight):
+    a = parse_parentheses(string)
+    out = []
+    for x in a:
+        weight = current_weight
+        if len(x) >= 2 and x[-1] == ")" and x[0] == "(":
+            x = x[1:-1]
+            xx = x.rfind(":")
+            weight *= 1.1
+            if xx > 0:
+                try:
+                    weight = float(x[xx + 1 :])
+                    x = x[:xx]
+                except:
+                    pass
+            out += token_weights(x, weight)
+        else:
+            out += [(x, current_weight)]
+    return out
+
+
+def escape_important(text):
+    text = text.replace("\\)", "\0\1")
+    text = text.replace("\\(", "\0\2")
+    return text
+
+
+def unescape_important(text):
+    text = text.replace("\0\1", ")")
+    text = text.replace("\0\2", "(")
+    return text
+
+
+class SDTokenizer:
+    def __init__(
+        self,
+        max_length=77,
+        pad_with_end=True,
+        tokenizer=None,
+        has_start_token=True,
+        pad_to_max_length=True,
+        min_length=None,
+        extra_padding_token=None,
+    ):
+        self.tokenizer = tokenizer
+        self.max_length = max_length
+        self.min_length = min_length
+
+        empty = self.tokenizer("")["input_ids"]
+        if has_start_token:
+            self.tokens_start = 1
+            self.start_token = empty[0]
+            self.end_token = empty[1]
+        else:
+            self.tokens_start = 0
+            self.start_token = None
+            self.end_token = empty[0]
+        self.pad_with_end = pad_with_end
+        self.pad_to_max_length = pad_to_max_length
+        self.extra_padding_token = extra_padding_token
+
+        vocab = self.tokenizer.get_vocab()
+        self.inv_vocab = {v: k for k, v in vocab.items()}
+        self.max_word_length = 8
+
+    def tokenize_with_weights(self, text: str, return_word_ids=False):
+        """
+        Tokenize the text, with weight values - presume 1.0 for all and ignore other features here.
+        The details aren't relevant for a reference impl, and weights themselves has weak effect on SD3.
+        """
+        if self.pad_with_end:
+            pad_token = self.end_token
+        else:
+            pad_token = 0
+
+        text = escape_important(text)
+        parsed_weights = token_weights(text, 1.0)
+
+        # tokenize words
+        tokens = []
+        for weighted_segment, weight in parsed_weights:
+            to_tokenize = unescape_important(weighted_segment).replace("\n", " ").split(" ")
+            to_tokenize = [x for x in to_tokenize if x != ""]
+            for word in to_tokenize:
+                # parse word
+                tokens.append([(t, weight) for t in self.tokenizer(word)["input_ids"][self.tokens_start : -1]])
+
+        # reshape token array to CLIP input size
+        batched_tokens = []
+        batch = []
+        if self.start_token is not None:
+            batch.append((self.start_token, 1.0, 0))
+        batched_tokens.append(batch)
+        for i, t_group in enumerate(tokens):
+            # determine if we're going to try and keep the tokens in a single batch
+            is_large = len(t_group) >= self.max_word_length
+
+            while len(t_group) > 0:
+                if len(t_group) + len(batch) > self.max_length - 1:
+                    remaining_length = self.max_length - len(batch) - 1
+                    # break word in two and add end token
+                    if is_large:
+                        batch.extend([(t, w, i + 1) for t, w in t_group[:remaining_length]])
+                        batch.append((self.end_token, 1.0, 0))
+                        t_group = t_group[remaining_length:]
+                    # add end token and pad
+                    else:
+                        batch.append((self.end_token, 1.0, 0))
+                        if self.pad_to_max_length:
+                            batch.extend([(pad_token, 1.0, 0)] * (remaining_length))
+                    # start new batch
+                    batch = []
+                    if self.start_token is not None:
+                        batch.append((self.start_token, 1.0, 0))
+                    batched_tokens.append(batch)
+                else:
+                    batch.extend([(t, w, i + 1) for t, w in t_group])
+                    t_group = []
+
+        # pad extra padding token first befor getting to the end token
+        if self.extra_padding_token is not None:
+            batch.extend([(self.extra_padding_token, 1.0, 0)] * (self.min_length - len(batch) - 1))
+        # fill last batch
+        batch.append((self.end_token, 1.0, 0))
+        if self.pad_to_max_length:
+            batch.extend([(pad_token, 1.0, 0)] * (self.max_length - len(batch)))
+        if self.min_length is not None and len(batch) < self.min_length:
+            batch.extend([(pad_token, 1.0, 0)] * (self.min_length - len(batch)))
+
+        if not return_word_ids:
+            batched_tokens = [[(t, w) for t, w, _ in x] for x in batched_tokens]
+
+        return batched_tokens
+
+    def untokenize(self, token_weight_pair):
+        return list(map(lambda a: (a, self.inv_vocab[a[0]]), token_weight_pair))
+
+
+class SDXLClipGTokenizer(SDTokenizer):
+    def __init__(self, tokenizer):
+        super().__init__(pad_with_end=False, tokenizer=tokenizer)
+
+
+class SD3Tokenizer:
+    def __init__(self):
+        clip_tokenizer = CLIPTokenizer.from_pretrained("openai/clip-vit-large-patch14")
+        self.clip_l = SDTokenizer(tokenizer=clip_tokenizer)
+        self.clip_g = SDXLClipGTokenizer(clip_tokenizer)
+        self.t5xxl = T5XXLTokenizer()
+
+    def tokenize_with_weights(self, text: str):
+        out = {}
+        out["l"] = self.clip_l.tokenize_with_weights(text)
+        out["g"] = self.clip_g.tokenize_with_weights(text)
+        out["t5xxl"] = self.t5xxl.tokenize_with_weights(text[:226])
+        return out
+
+
+class ClipTokenWeightEncoder:
+    def encode_token_weights(self, token_weight_pairs):
+        tokens = list(map(lambda a: a[0], token_weight_pairs[0]))
+        out, pooled = self([tokens])
+        if pooled is not None:
+            first_pooled = pooled[0:1].cpu()
+        else:
+            first_pooled = pooled
+        output = [out[0:1]]
+        return torch.cat(output, dim=-2).cpu(), first_pooled
+
+
+class SDClipModel(torch.nn.Module, ClipTokenWeightEncoder):
+    """Uses the CLIP transformer encoder for text (from huggingface)"""
+
+    LAYERS = ["last", "pooled", "hidden"]
+
+    def __init__(
+        self,
+        device="cpu",
+        max_length=77,
+        layer="last",
+        layer_idx=None,
+        textmodel_json_config=None,
+        dtype=None,
+        model_class=CLIPTextModel,
+        special_tokens={"start": 49406, "end": 49407, "pad": 49407},
+        layer_norm_hidden_state=True,
+        return_projected_pooled=True,
+    ):
+        super().__init__()
+        assert layer in self.LAYERS
+        self.transformer = model_class(textmodel_json_config, dtype, device)
+        self.num_layers = self.transformer.num_layers
+        self.max_length = max_length
+        self.transformer = self.transformer.eval()
+        for param in self.parameters():
+            param.requires_grad = False
+        self.layer = layer
+        self.layer_idx = None
+        self.special_tokens = special_tokens
+        self.logit_scale = torch.nn.Parameter(torch.tensor(4.6055))
+        self.layer_norm_hidden_state = layer_norm_hidden_state
+        self.return_projected_pooled = return_projected_pooled
+        if layer == "hidden":
+            assert layer_idx is not None
+            assert abs(layer_idx) < self.num_layers
+            self.set_clip_options({"layer": layer_idx})
+        self.options_default = (
+            self.layer,
+            self.layer_idx,
+            self.return_projected_pooled,
+        )
+
+    def set_clip_options(self, options):
+        layer_idx = options.get("layer", self.layer_idx)
+        self.return_projected_pooled = options.get("projected_pooled", self.return_projected_pooled)
+        if layer_idx is None or abs(layer_idx) > self.num_layers:
+            self.layer = "last"
+        else:
+            self.layer = "hidden"
+            self.layer_idx = layer_idx
+
+    def forward(self, tokens):
+        backup_embeds = self.transformer.get_input_embeddings()
+        device = backup_embeds.weight.device
+        tokens = torch.LongTensor(tokens).to(device)
+        outputs = self.transformer(
+            tokens,
+            intermediate_output=self.layer_idx,
+            final_layer_norm_intermediate=self.layer_norm_hidden_state,
+        )
+        self.transformer.set_input_embeddings(backup_embeds)
+        if self.layer == "last":
+            z = outputs[0]
+        else:
+            z = outputs[1]
+        pooled_output = None
+        if len(outputs) >= 3:
+            if not self.return_projected_pooled and len(outputs) >= 4 and outputs[3] is not None:
+                pooled_output = outputs[3].float()
+            elif outputs[2] is not None:
+                pooled_output = outputs[2].float()
+        return z.float(), pooled_output
+
+
+class SDXLClipG(SDClipModel):
+    """Wraps the CLIP-G model into the SD-CLIP-Model interface"""
+
+    def __init__(self, config, device="cpu", layer="penultimate", layer_idx=None, dtype=None):
+        if layer == "penultimate":
+            layer = "hidden"
+            layer_idx = -2
+        super().__init__(
+            device=device,
+            layer=layer,
+            layer_idx=layer_idx,
+            textmodel_json_config=config,
+            dtype=dtype,
+            special_tokens={"start": 49406, "end": 49407, "pad": 0},
+            layer_norm_hidden_state=False,
+        )
+
+
+class T5XXLModel(SDClipModel):
+    """Wraps the T5-XXL model into the SD-CLIP-Model interface for convenience"""
+
+    def __init__(self, config, device="cpu", layer="last", layer_idx=None, dtype=None):
+        super().__init__(
+            device=device,
+            layer=layer,
+            layer_idx=layer_idx,
+            textmodel_json_config=config,
+            dtype=dtype,
+            special_tokens={"end": 1, "pad": 0},
+            model_class=T5,
+        )
+
+
+#################################################################################################
+### T5 implementation, for the T5-XXL text encoder portion, largely pulled from upstream impl
+#################################################################################################
+
+
+class T5XXLTokenizer(SDTokenizer):
+    """Wraps the T5 Tokenizer from HF into the SDTokenizer interface"""
+
+    def __init__(self):
+        super().__init__(
+            pad_with_end=False,
+            tokenizer=T5TokenizerFast.from_pretrained("google/t5-v1_1-xxl"),
+            has_start_token=False,
+            pad_to_max_length=False,
+            max_length=99999999,
+            min_length=77,
+        )
+
+
+class T5LayerNorm(torch.nn.Module):
+    def __init__(self, hidden_size, eps=1e-6, dtype=None, device=None):
+        super().__init__()
+        self.weight = torch.nn.Parameter(torch.ones(hidden_size, dtype=dtype, device=device))
+        self.variance_epsilon = eps
+
+    def forward(self, x):
+        variance = x.pow(2).mean(-1, keepdim=True)
+        x = x * torch.rsqrt(variance + self.variance_epsilon)
+        return self.weight.to(device=x.device, dtype=x.dtype) * x
+
+
+class T5DenseGatedActDense(torch.nn.Module):
+    def __init__(self, model_dim, ff_dim, dtype, device):
+        super().__init__()
+        self.wi_0 = torch.nn.Linear(model_dim, ff_dim, bias=False, dtype=dtype, device=device)
+        self.wi_1 = torch.nn.Linear(model_dim, ff_dim, bias=False, dtype=dtype, device=device)
+        self.wo = torch.nn.Linear(ff_dim, model_dim, bias=False, dtype=dtype, device=device)
+
+    def forward(self, x):
+        hidden_gelu = torch.nn.functional.gelu(self.wi_0(x), approximate="tanh")
+        hidden_linear = self.wi_1(x)
+        x = hidden_gelu * hidden_linear
+        x = self.wo(x)
+        return x
+
+
+class T5LayerFF(torch.nn.Module):
+    def __init__(self, model_dim, ff_dim, dtype, device):
+        super().__init__()
+        self.DenseReluDense = T5DenseGatedActDense(model_dim, ff_dim, dtype, device)
+        self.layer_norm = T5LayerNorm(model_dim, dtype=dtype, device=device)
+
+    def forward(self, x):
+        forwarded_states = self.layer_norm(x)
+        forwarded_states = self.DenseReluDense(forwarded_states)
+        x += forwarded_states
+        return x
+
+
+class T5Attention(torch.nn.Module):
+    def __init__(self, model_dim, inner_dim, num_heads, relative_attention_bias, dtype, device):
+        super().__init__()
+        # Mesh TensorFlow initialization to avoid scaling before softmax
+        self.q = torch.nn.Linear(model_dim, inner_dim, bias=False, dtype=dtype, device=device)
+        self.k = torch.nn.Linear(model_dim, inner_dim, bias=False, dtype=dtype, device=device)
+        self.v = torch.nn.Linear(model_dim, inner_dim, bias=False, dtype=dtype, device=device)
+        self.o = torch.nn.Linear(inner_dim, model_dim, bias=False, dtype=dtype, device=device)
+        self.num_heads = num_heads
+        self.relative_attention_bias = None
+        if relative_attention_bias:
+            self.relative_attention_num_buckets = 32
+            self.relative_attention_max_distance = 128
+            self.relative_attention_bias = torch.nn.Embedding(
+                self.relative_attention_num_buckets, self.num_heads, device=device
+            )
+
+    @staticmethod
+    def _relative_position_bucket(relative_position, bidirectional=True, num_buckets=32, max_distance=128):
+        """
+        Adapted from Mesh Tensorflow:
+        https://github.com/tensorflow/mesh/blob/0cb87fe07da627bf0b7e60475d59f95ed6b5be3d/mesh_tensorflow/transformer/transformer_layers.py#L593
+
+        Translate relative position to a bucket number for relative attention. The relative position is defined as
+        memory_position - query_position, i.e. the distance in tokens from the attending position to the attended-to
+        position. If bidirectional=False, then positive relative positions are invalid. We use smaller buckets for
+        small absolute relative_position and larger buckets for larger absolute relative_positions. All relative
+        positions >=max_distance map to the same bucket. All relative positions <=-max_distance map to the same bucket.
+        This should allow for more graceful generalization to longer sequences than the model has been trained on
+
+        Args:
+            relative_position: an int32 Tensor
+            bidirectional: a boolean - whether the attention is bidirectional
+            num_buckets: an integer
+            max_distance: an integer
+
+        Returns:
+            a Tensor with the same shape as relative_position, containing int32 values in the range [0, num_buckets)
+        """
+        relative_buckets = 0
+        if bidirectional:
+            num_buckets //= 2
+            relative_buckets += (relative_position > 0).to(torch.long) * num_buckets
+            relative_position = torch.abs(relative_position)
+        else:
+            relative_position = -torch.min(relative_position, torch.zeros_like(relative_position))
+        # now relative_position is in the range [0, inf)
+        # half of the buckets are for exact increments in positions
+        max_exact = num_buckets // 2
+        is_small = relative_position < max_exact
+        # The other half of the buckets are for logarithmically bigger bins in positions up to max_distance
+        relative_position_if_large = max_exact + (
+            torch.log(relative_position.float() / max_exact)
+            / math.log(max_distance / max_exact)
+            * (num_buckets - max_exact)
+        ).to(torch.long)
+        relative_position_if_large = torch.min(
+            relative_position_if_large,
+            torch.full_like(relative_position_if_large, num_buckets - 1),
+        )
+        relative_buckets += torch.where(is_small, relative_position, relative_position_if_large)
+        return relative_buckets
+
+    def compute_bias(self, query_length, key_length, device):
+        """Compute binned relative position bias"""
+        context_position = torch.arange(query_length, dtype=torch.long, device=device)[:, None]
+        memory_position = torch.arange(key_length, dtype=torch.long, device=device)[None, :]
+        relative_position = memory_position - context_position  # shape (query_length, key_length)
+        relative_position_bucket = self._relative_position_bucket(
+            relative_position,  # shape (query_length, key_length)
+            bidirectional=True,
+            num_buckets=self.relative_attention_num_buckets,
+            max_distance=self.relative_attention_max_distance,
+        )
+        values = self.relative_attention_bias(relative_position_bucket)  # shape (query_length, key_length, num_heads)
+        values = values.permute([2, 0, 1]).unsqueeze(0)  # shape (1, num_heads, query_length, key_length)
+        return values
+
+    def forward(self, x, past_bias=None):
+        q = self.q(x)
+        k = self.k(x)
+        v = self.v(x)
+        if self.relative_attention_bias is not None:
+            past_bias = self.compute_bias(x.shape[1], x.shape[1], x.device)
+        if past_bias is not None:
+            mask = past_bias
+        out = attention(q, k * ((k.shape[-1] / self.num_heads) ** 0.5), v, self.num_heads, mask)
+        return self.o(out), past_bias
+
+
+class T5LayerSelfAttention(torch.nn.Module):
+    def __init__(
+        self,
+        model_dim,
+        inner_dim,
+        ff_dim,
+        num_heads,
+        relative_attention_bias,
+        dtype,
+        device,
+    ):
+        super().__init__()
+        self.SelfAttention = T5Attention(model_dim, inner_dim, num_heads, relative_attention_bias, dtype, device)
+        self.layer_norm = T5LayerNorm(model_dim, dtype=dtype, device=device)
+
+    def forward(self, x, past_bias=None):
+        output, past_bias = self.SelfAttention(self.layer_norm(x), past_bias=past_bias)
+        x += output
+        return x, past_bias
+
+
+class T5Block(torch.nn.Module):
+    def __init__(
+        self,
+        model_dim,
+        inner_dim,
+        ff_dim,
+        num_heads,
+        relative_attention_bias,
+        dtype,
+        device,
+    ):
+        super().__init__()
+        self.layer = torch.nn.ModuleList()
+        self.layer.append(
+            T5LayerSelfAttention(
+                model_dim,
+                inner_dim,
+                ff_dim,
+                num_heads,
+                relative_attention_bias,
+                dtype,
+                device,
+            )
+        )
+        self.layer.append(T5LayerFF(model_dim, ff_dim, dtype, device))
+
+    def forward(self, x, past_bias=None):
+        x, past_bias = self.layer[0](x, past_bias)
+        x = self.layer[-1](x)
+        return x, past_bias
+
+
+class T5Stack(torch.nn.Module):
+    def __init__(
+        self,
+        num_layers,
+        model_dim,
+        inner_dim,
+        ff_dim,
+        num_heads,
+        vocab_size,
+        dtype,
+        device,
+    ):
+        super().__init__()
+        self.embed_tokens = torch.nn.Embedding(vocab_size, model_dim, device=device)
+        self.block = torch.nn.ModuleList(
+            [
+                T5Block(
+                    model_dim,
+                    inner_dim,
+                    ff_dim,
+                    num_heads,
+                    relative_attention_bias=(i == 0),
+                    dtype=dtype,
+                    device=device,
+                )
+                for i in range(num_layers)
+            ]
+        )
+        self.final_layer_norm = T5LayerNorm(model_dim, dtype=dtype, device=device)
+
+    def forward(self, input_ids, intermediate_output=None, final_layer_norm_intermediate=True):
+        intermediate = None
+        x = self.embed_tokens(input_ids)
+        past_bias = None
+        for i, l in enumerate(self.block):
+            x, past_bias = l(x, past_bias)
+            if i == intermediate_output:
+                intermediate = x.clone()
+        x = self.final_layer_norm(x)
+        if intermediate is not None and final_layer_norm_intermediate:
+            intermediate = self.final_layer_norm(intermediate)
+        return x, intermediate
+
+
+class T5(torch.nn.Module):
+    def __init__(self, config_dict, dtype, device):
+        super().__init__()
+        self.num_layers = config_dict["num_layers"]
+        self.encoder = T5Stack(
+            self.num_layers,
+            config_dict["d_model"],
+            config_dict["d_model"],
+            config_dict["d_ff"],
+            config_dict["num_heads"],
+            config_dict["vocab_size"],
+            dtype,
+            device,
+        )
+        self.dtype = dtype
+
+    def get_input_embeddings(self):
+        return self.encoder.embed_tokens
+
+    def set_input_embeddings(self, embeddings):
+        self.encoder.embed_tokens = embeddings
+
+    def forward(self, *args, **kwargs):
+        return self.encoder(*args, **kwargs)

invokeai/backend/sd3/sd3_impls.py

@@ -0,0 +1,609 @@
+# This file was originally copied from:
+# https://github.com/Stability-AI/sd3.5/blob/19bf11c4e1e37324c5aa5a61f010d4127848a09c/sd3_impls.py
+
+
+### Impls of the SD3 core diffusion model and VAE
+
+import math
+import re
+
+import einops
+import torch
+from PIL import Image
+from tqdm import tqdm
+
+from invokeai.backend.sd3.mmditx import MMDiTX
+
+#################################################################################################
+### MMDiT Model Wrapping
+#################################################################################################
+
+
+class ModelSamplingDiscreteFlow(torch.nn.Module):
+    """Helper for sampler scheduling (ie timestep/sigma calculations) for Discrete Flow models"""
+
+    def __init__(self, shift: float = 1.0):
+        super().__init__()
+        self.shift = shift
+        timesteps = 1000
+        ts = self.sigma(torch.arange(1, timesteps + 1, 1))
+        self.register_buffer("sigmas", ts)
+
+    @property
+    def sigma_min(self):
+        return self.sigmas[0]
+
+    @property
+    def sigma_max(self):
+        return self.sigmas[-1]
+
+    def timestep(self, sigma: torch.Tensor) -> torch.Tensor:
+        return sigma * 1000
+
+    def sigma(self, timestep: torch.Tensor):
+        timestep = timestep / 1000.0
+        if self.shift == 1.0:
+            return timestep
+        return self.shift * timestep / (1 + (self.shift - 1) * timestep)
+
+    def calculate_denoised(
+        self, sigma: torch.Tensor, model_output: torch.Tensor, model_input: torch.Tensor
+    ) -> torch.Tensor:
+        sigma = sigma.view(sigma.shape[:1] + (1,) * (model_output.ndim - 1))
+        return model_input - model_output * sigma
+
+    def noise_scaling(self, sigma, noise, latent_image, max_denoise=False):
+        return sigma * noise + (1.0 - sigma) * latent_image
+
+
+class BaseModel(torch.nn.Module):
+    """Wrapper around the core MM-DiT model"""
+
+    def __init__(
+        self,
+        shift=1.0,
+        device=None,
+        dtype=torch.float32,
+        file=None,
+        prefix="",
+        verbose=False,
+    ):
+        super().__init__()
+        # Important configuration values can be quickly determined by checking shapes in the source file
+        # Some of these will vary between models (eg 2B vs 8B primarily differ in their depth, but also other details change)
+        patch_size = file.get_tensor(f"{prefix}x_embedder.proj.weight").shape[2]
+        depth = file.get_tensor(f"{prefix}x_embedder.proj.weight").shape[0] // 64
+        num_patches = file.get_tensor(f"{prefix}pos_embed").shape[1]
+        pos_embed_max_size = round(math.sqrt(num_patches))
+        adm_in_channels = file.get_tensor(f"{prefix}y_embedder.mlp.0.weight").shape[1]
+        context_shape = file.get_tensor(f"{prefix}context_embedder.weight").shape
+        qk_norm = "rms" if f"{prefix}joint_blocks.0.context_block.attn.ln_k.weight" in file.keys() else None
+        x_block_self_attn_layers = sorted(
+            [
+                int(key.split(".x_block.attn2.ln_k.weight")[0].split(".")[-1])
+                for key in list(filter(re.compile(".*.x_block.attn2.ln_k.weight").match, file.keys()))
+            ]
+        )
+
+        context_embedder_config = {
+            "target": "torch.nn.Linear",
+            "params": {
+                "in_features": context_shape[1],
+                "out_features": context_shape[0],
+            },
+        }
+        self.diffusion_model = MMDiTX(
+            input_size=None,
+            pos_embed_scaling_factor=None,
+            pos_embed_offset=None,
+            pos_embed_max_size=pos_embed_max_size,
+            patch_size=patch_size,
+            in_channels=16,
+            depth=depth,
+            num_patches=num_patches,
+            adm_in_channels=adm_in_channels,
+            context_embedder_config=context_embedder_config,
+            qk_norm=qk_norm,
+            x_block_self_attn_layers=x_block_self_attn_layers,
+            device=device,
+            dtype=dtype,
+            verbose=verbose,
+        )
+        self.model_sampling = ModelSamplingDiscreteFlow(shift=shift)
+
+    def apply_model(
+        self, x: torch.Tensor, sigma: float, c_crossattn: torch.Tensor | None = None, y: torch.Tensor | None = None
+    ):
+        dtype = self.get_dtype()
+        timestep = self.model_sampling.timestep(sigma).float()
+        model_output = self.diffusion_model(x.to(dtype), timestep, context=c_crossattn.to(dtype), y=y.to(dtype)).float()
+        return self.model_sampling.calculate_denoised(sigma, model_output, x)
+
+    def forward(self, *args, **kwargs):
+        return self.apply_model(*args, **kwargs)
+
+    def get_dtype(self):
+        return self.diffusion_model.dtype
+
+
+class CFGDenoiser(torch.nn.Module):
+    """Helper for applying CFG Scaling to diffusion outputs"""
+
+    def __init__(self, model):
+        super().__init__()
+        self.model = model
+
+    def forward(self, x, timestep, cond, uncond, cond_scale):
+        # Run cond and uncond in a batch together
+        batched = self.model.apply_model(
+            torch.cat([x, x]),
+            torch.cat([timestep, timestep]),
+            c_crossattn=torch.cat([cond["c_crossattn"], uncond["c_crossattn"]]),
+            y=torch.cat([cond["y"], uncond["y"]]),
+        )
+        # Then split and apply CFG Scaling
+        pos_out, neg_out = batched.chunk(2)
+        scaled = neg_out + (pos_out - neg_out) * cond_scale
+        return scaled
+
+
+class SD3LatentFormat:
+    """Latents are slightly shifted from center - this class must be called after VAE Decode to correct for the shift"""
+
+    def __init__(self):
+        self.scale_factor = 1.5305
+        self.shift_factor = 0.0609
+
+    def process_in(self, latent):
+        return (latent - self.shift_factor) * self.scale_factor
+
+    def process_out(self, latent):
+        return (latent / self.scale_factor) + self.shift_factor
+
+    def decode_latent_to_preview(self, x0):
+        """Quick RGB approximate preview of sd3 latents"""
+        factors = torch.tensor(
+            [
+                [-0.0645, 0.0177, 0.1052],
+                [0.0028, 0.0312, 0.0650],
+                [0.1848, 0.0762, 0.0360],
+                [0.0944, 0.0360, 0.0889],
+                [0.0897, 0.0506, -0.0364],
+                [-0.0020, 0.1203, 0.0284],
+                [0.0855, 0.0118, 0.0283],
+                [-0.0539, 0.0658, 0.1047],
+                [-0.0057, 0.0116, 0.0700],
+                [-0.0412, 0.0281, -0.0039],
+                [0.1106, 0.1171, 0.1220],
+                [-0.0248, 0.0682, -0.0481],
+                [0.0815, 0.0846, 0.1207],
+                [-0.0120, -0.0055, -0.0867],
+                [-0.0749, -0.0634, -0.0456],
+                [-0.1418, -0.1457, -0.1259],
+            ],
+            device="cpu",
+        )
+        latent_image = x0[0].permute(1, 2, 0).cpu() @ factors
+
+        latents_ubyte = (
+            ((latent_image + 1) / 2)
+            .clamp(0, 1)  # change scale from -1..1 to 0..1
+            .mul(0xFF)  # to 0..255
+            .byte()
+        ).cpu()
+
+        return Image.fromarray(latents_ubyte.numpy())
+
+
+#################################################################################################
+### Samplers
+#################################################################################################
+
+
+def append_dims(x, target_dims):
+    """Appends dimensions to the end of a tensor until it has target_dims dimensions."""
+    dims_to_append = target_dims - x.ndim
+    return x[(...,) + (None,) * dims_to_append]
+
+
+def to_d(x, sigma, denoised):
+    """Converts a denoiser output to a Karras ODE derivative."""
+    return (x - denoised) / append_dims(sigma, x.ndim)
+
+
+@torch.no_grad()
+@torch.autocast("cuda", dtype=torch.float16)
+def sample_euler(model, x, sigmas, extra_args=None):
+    """Implements Algorithm 2 (Euler steps) from Karras et al. (2022)."""
+    extra_args = {} if extra_args is None else extra_args
+    s_in = x.new_ones([x.shape[0]])
+    for i in tqdm(range(len(sigmas) - 1)):
+        sigma_hat = sigmas[i]
+        denoised = model(x, sigma_hat * s_in, **extra_args)
+        d = to_d(x, sigma_hat, denoised)
+        dt = sigmas[i + 1] - sigma_hat
+        # Euler method
+        x = x + d * dt
+    return x
+
+
+@torch.no_grad()
+@torch.autocast("cuda", dtype=torch.float16)
+def sample_dpmpp_2m(model, x, sigmas, extra_args=None):
+    """DPM-Solver++(2M)."""
+    extra_args = {} if extra_args is None else extra_args
+    s_in = x.new_ones([x.shape[0]])
+    sigma_fn = lambda t: t.neg().exp()
+    t_fn = lambda sigma: sigma.log().neg()
+    old_denoised = None
+    for i in tqdm(range(len(sigmas) - 1)):
+        denoised = model(x, sigmas[i] * s_in, **extra_args)
+        t, t_next = t_fn(sigmas[i]), t_fn(sigmas[i + 1])
+        h = t_next - t
+        if old_denoised is None or sigmas[i + 1] == 0:
+            x = (sigma_fn(t_next) / sigma_fn(t)) * x - (-h).expm1() * denoised
+        else:
+            h_last = t - t_fn(sigmas[i - 1])
+            r = h_last / h
+            denoised_d = (1 + 1 / (2 * r)) * denoised - (1 / (2 * r)) * old_denoised
+            x = (sigma_fn(t_next) / sigma_fn(t)) * x - (-h).expm1() * denoised_d
+        old_denoised = denoised
+    return x
+
+
+#################################################################################################
+### VAE
+#################################################################################################
+
+
+def Normalize(in_channels, num_groups=32, dtype=torch.float32, device=None):
+    return torch.nn.GroupNorm(
+        num_groups=num_groups,
+        num_channels=in_channels,
+        eps=1e-6,
+        affine=True,
+        dtype=dtype,
+        device=device,
+    )
+
+
+class ResnetBlock(torch.nn.Module):
+    def __init__(self, *, in_channels, out_channels=None, dtype=torch.float32, device=None):
+        super().__init__()
+        self.in_channels = in_channels
+        out_channels = in_channels if out_channels is None else out_channels
+        self.out_channels = out_channels
+
+        self.norm1 = Normalize(in_channels, dtype=dtype, device=device)
+        self.conv1 = torch.nn.Conv2d(
+            in_channels,
+            out_channels,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            dtype=dtype,
+            device=device,
+        )
+        self.norm2 = Normalize(out_channels, dtype=dtype, device=device)
+        self.conv2 = torch.nn.Conv2d(
+            out_channels,
+            out_channels,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            dtype=dtype,
+            device=device,
+        )
+        if self.in_channels != self.out_channels:
+            self.nin_shortcut = torch.nn.Conv2d(
+                in_channels,
+                out_channels,
+                kernel_size=1,
+                stride=1,
+                padding=0,
+                dtype=dtype,
+                device=device,
+            )
+        else:
+            self.nin_shortcut = None
+        self.swish = torch.nn.SiLU(inplace=True)
+
+    def forward(self, x):
+        hidden = x
+        hidden = self.norm1(hidden)
+        hidden = self.swish(hidden)
+        hidden = self.conv1(hidden)
+        hidden = self.norm2(hidden)
+        hidden = self.swish(hidden)
+        hidden = self.conv2(hidden)
+        if self.in_channels != self.out_channels:
+            x = self.nin_shortcut(x)
+        return x + hidden
+
+
+class AttnBlock(torch.nn.Module):
+    def __init__(self, in_channels, dtype=torch.float32, device=None):
+        super().__init__()
+        self.norm = Normalize(in_channels, dtype=dtype, device=device)
+        self.q = torch.nn.Conv2d(
+            in_channels,
+            in_channels,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            dtype=dtype,
+            device=device,
+        )
+        self.k = torch.nn.Conv2d(
+            in_channels,
+            in_channels,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            dtype=dtype,
+            device=device,
+        )
+        self.v = torch.nn.Conv2d(
+            in_channels,
+            in_channels,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            dtype=dtype,
+            device=device,
+        )
+        self.proj_out = torch.nn.Conv2d(
+            in_channels,
+            in_channels,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            dtype=dtype,
+            device=device,
+        )
+
+    def forward(self, x):
+        hidden = self.norm(x)
+        q = self.q(hidden)
+        k = self.k(hidden)
+        v = self.v(hidden)
+        b, c, h, w = q.shape
+        q, k, v = map(
+            lambda x: einops.rearrange(x, "b c h w -> b 1 (h w) c").contiguous(),
+            (q, k, v),
+        )
+        hidden = torch.nn.functional.scaled_dot_product_attention(q, k, v)  # scale is dim ** -0.5 per default
+        hidden = einops.rearrange(hidden, "b 1 (h w) c -> b c h w", h=h, w=w, c=c, b=b)
+        hidden = self.proj_out(hidden)
+        return x + hidden
+
+
+class Downsample(torch.nn.Module):
+    def __init__(self, in_channels, dtype=torch.float32, device=None):
+        super().__init__()
+        self.conv = torch.nn.Conv2d(
+            in_channels,
+            in_channels,
+            kernel_size=3,
+            stride=2,
+            padding=0,
+            dtype=dtype,
+            device=device,
+        )
+
+    def forward(self, x):
+        pad = (0, 1, 0, 1)
+        x = torch.nn.functional.pad(x, pad, mode="constant", value=0)
+        x = self.conv(x)
+        return x
+
+
+class Upsample(torch.nn.Module):
+    def __init__(self, in_channels, dtype=torch.float32, device=None):
+        super().__init__()
+        self.conv = torch.nn.Conv2d(
+            in_channels,
+            in_channels,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            dtype=dtype,
+            device=device,
+        )
+
+    def forward(self, x):
+        x = torch.nn.functional.interpolate(x, scale_factor=2.0, mode="nearest")
+        x = self.conv(x)
+        return x
+
+
+class VAEEncoder(torch.nn.Module):
+    def __init__(
+        self,
+        ch=128,
+        ch_mult=(1, 2, 4, 4),
+        num_res_blocks=2,
+        in_channels=3,
+        z_channels=16,
+        dtype=torch.float32,
+        device=None,
+    ):
+        super().__init__()
+        self.num_resolutions = len(ch_mult)
+        self.num_res_blocks = num_res_blocks
+        # downsampling
+        self.conv_in = torch.nn.Conv2d(
+            in_channels,
+            ch,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            dtype=dtype,
+            device=device,
+        )
+        in_ch_mult = (1,) + tuple(ch_mult)
+        self.in_ch_mult = in_ch_mult
+        self.down = torch.nn.ModuleList()
+        for i_level in range(self.num_resolutions):
+            block = torch.nn.ModuleList()
+            attn = torch.nn.ModuleList()
+            block_in = ch * in_ch_mult[i_level]
+            block_out = ch * ch_mult[i_level]
+            for i_block in range(num_res_blocks):
+                block.append(
+                    ResnetBlock(
+                        in_channels=block_in,
+                        out_channels=block_out,
+                        dtype=dtype,
+                        device=device,
+                    )
+                )
+                block_in = block_out
+            down = torch.nn.Module()
+            down.block = block
+            down.attn = attn
+            if i_level != self.num_resolutions - 1:
+                down.downsample = Downsample(block_in, dtype=dtype, device=device)
+            self.down.append(down)
+        # middle
+        self.mid = torch.nn.Module()
+        self.mid.block_1 = ResnetBlock(in_channels=block_in, out_channels=block_in, dtype=dtype, device=device)
+        self.mid.attn_1 = AttnBlock(block_in, dtype=dtype, device=device)
+        self.mid.block_2 = ResnetBlock(in_channels=block_in, out_channels=block_in, dtype=dtype, device=device)
+        # end
+        self.norm_out = Normalize(block_in, dtype=dtype, device=device)
+        self.conv_out = torch.nn.Conv2d(
+            block_in,
+            2 * z_channels,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            dtype=dtype,
+            device=device,
+        )
+        self.swish = torch.nn.SiLU(inplace=True)
+
+    def forward(self, x):
+        # downsampling
+        hs = [self.conv_in(x)]
+        for i_level in range(self.num_resolutions):
+            for i_block in range(self.num_res_blocks):
+                h = self.down[i_level].block[i_block](hs[-1])
+                hs.append(h)
+            if i_level != self.num_resolutions - 1:
+                hs.append(self.down[i_level].downsample(hs[-1]))
+        # middle
+        h = hs[-1]
+        h = self.mid.block_1(h)
+        h = self.mid.attn_1(h)
+        h = self.mid.block_2(h)
+        # end
+        h = self.norm_out(h)
+        h = self.swish(h)
+        h = self.conv_out(h)
+        return h
+
+
+class VAEDecoder(torch.nn.Module):
+    def __init__(
+        self,
+        ch=128,
+        out_ch=3,
+        ch_mult=(1, 2, 4, 4),
+        num_res_blocks=2,
+        resolution=256,
+        z_channels=16,
+        dtype=torch.float32,
+        device=None,
+    ):
+        super().__init__()
+        self.num_resolutions = len(ch_mult)
+        self.num_res_blocks = num_res_blocks
+        block_in = ch * ch_mult[self.num_resolutions - 1]
+        curr_res = resolution // 2 ** (self.num_resolutions - 1)
+        # z to block_in
+        self.conv_in = torch.nn.Conv2d(
+            z_channels,
+            block_in,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            dtype=dtype,
+            device=device,
+        )
+        # middle
+        self.mid = torch.nn.Module()
+        self.mid.block_1 = ResnetBlock(in_channels=block_in, out_channels=block_in, dtype=dtype, device=device)
+        self.mid.attn_1 = AttnBlock(block_in, dtype=dtype, device=device)
+        self.mid.block_2 = ResnetBlock(in_channels=block_in, out_channels=block_in, dtype=dtype, device=device)
+        # upsampling
+        self.up = torch.nn.ModuleList()
+        for i_level in reversed(range(self.num_resolutions)):
+            block = torch.nn.ModuleList()
+            block_out = ch * ch_mult[i_level]
+            for i_block in range(self.num_res_blocks + 1):
+                block.append(
+                    ResnetBlock(
+                        in_channels=block_in,
+                        out_channels=block_out,
+                        dtype=dtype,
+                        device=device,
+                    )
+                )
+                block_in = block_out
+            up = torch.nn.Module()
+            up.block = block
+            if i_level != 0:
+                up.upsample = Upsample(block_in, dtype=dtype, device=device)
+                curr_res = curr_res * 2
+            self.up.insert(0, up)  # prepend to get consistent order
+        # end
+        self.norm_out = Normalize(block_in, dtype=dtype, device=device)
+        self.conv_out = torch.nn.Conv2d(
+            block_in,
+            out_ch,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            dtype=dtype,
+            device=device,
+        )
+        self.swish = torch.nn.SiLU(inplace=True)
+
+    def forward(self, z):
+        # z to block_in
+        hidden = self.conv_in(z)
+        # middle
+        hidden = self.mid.block_1(hidden)
+        hidden = self.mid.attn_1(hidden)
+        hidden = self.mid.block_2(hidden)
+        # upsampling
+        for i_level in reversed(range(self.num_resolutions)):
+            for i_block in range(self.num_res_blocks + 1):
+                hidden = self.up[i_level].block[i_block](hidden)
+            if i_level != 0:
+                hidden = self.up[i_level].upsample(hidden)
+        # end
+        hidden = self.norm_out(hidden)
+        hidden = self.swish(hidden)
+        hidden = self.conv_out(hidden)
+        return hidden
+
+
+class SDVAE(torch.nn.Module):
+    def __init__(self, dtype=torch.float32, device=None):
+        super().__init__()
+        self.encoder = VAEEncoder(dtype=dtype, device=device)
+        self.decoder = VAEDecoder(dtype=dtype, device=device)
+
+    @torch.autocast("cuda", dtype=torch.float16)
+    def decode(self, latent):
+        return self.decoder(latent)
+
+    @torch.autocast("cuda", dtype=torch.float16)
+    def encode(self, image):
+        hidden = self.encoder(image)
+        mean, logvar = torch.chunk(hidden, 2, dim=1)
+        logvar = torch.clamp(logvar, -30.0, 20.0)
+        std = torch.exp(0.5 * logvar)
+        return mean + std * torch.randn_like(mean)

invokeai/backend/sd3/sd3_infer.py

@@ -0,0 +1,426 @@
+# This file was originally copied from:
+# https://github.com/Stability-AI/sd3.5/blob/19bf11c4e1e37324c5aa5a61f010d4127848a09c/sd3_infer.py
+
+# NOTE: Must have folder `models` with the following files:
+# - `clip_g.safetensors` (openclip bigG, same as SDXL)
+# - `clip_l.safetensors` (OpenAI CLIP-L, same as SDXL)
+# - `t5xxl.safetensors` (google T5-v1.1-XXL)
+# - `sd3_medium.safetensors` (or whichever main MMDiT model file)
+# Also can have
+# - `sd3_vae.safetensors` (holds the VAE separately if needed)
+
+import datetime
+import math
+import os
+
+import fire
+import numpy as np
+import sd3_impls
+import torch
+from other_impls import SD3Tokenizer, SDClipModel, SDXLClipG, T5XXLModel
+from PIL import Image
+from safetensors import safe_open
+from sd3_impls import SDVAE, BaseModel, CFGDenoiser, SD3LatentFormat
+from tqdm import tqdm
+
+#################################################################################################
+### Wrappers for model parts
+#################################################################################################
+
+
+def load_into(f, model, prefix, device, dtype=None):
+    """Just a debugging-friendly hack to apply the weights in a safetensors file to the pytorch module."""
+    for key in f.keys():
+        if key.startswith(prefix) and not key.startswith("loss."):
+            path = key[len(prefix) :].split(".")
+            obj = model
+            for p in path:
+                if obj is list:
+                    obj = obj[int(p)]
+                else:
+                    obj = getattr(obj, p, None)
+                    if obj is None:
+                        print(f"Skipping key '{key}' in safetensors file as '{p}' does not exist in python model")
+                        break
+            if obj is None:
+                continue
+            try:
+                tensor = f.get_tensor(key).to(device=device)
+                if dtype is not None:
+                    tensor = tensor.to(dtype=dtype)
+                obj.requires_grad_(False)
+                obj.set_(tensor)
+            except Exception as e:
+                print(f"Failed to load key '{key}' in safetensors file: {e}")
+                raise e
+
+
+CLIPG_CONFIG = {
+    "hidden_act": "gelu",
+    "hidden_size": 1280,
+    "intermediate_size": 5120,
+    "num_attention_heads": 20,
+    "num_hidden_layers": 32,
+}
+
+
+class ClipG:
+    def __init__(self):
+        with safe_open("models/clip_g.safetensors", framework="pt", device="cpu") as f:
+            self.model = SDXLClipG(CLIPG_CONFIG, device="cpu", dtype=torch.float32)
+            load_into(f, self.model.transformer, "", "cpu", torch.float32)
+
+
+CLIPL_CONFIG = {
+    "hidden_act": "quick_gelu",
+    "hidden_size": 768,
+    "intermediate_size": 3072,
+    "num_attention_heads": 12,
+    "num_hidden_layers": 12,
+}
+
+
+class ClipL:
+    def __init__(self):
+        with safe_open("models/clip_l.safetensors", framework="pt", device="cpu") as f:
+            self.model = SDClipModel(
+                layer="hidden",
+                layer_idx=-2,
+                device="cpu",
+                dtype=torch.float32,
+                layer_norm_hidden_state=False,
+                return_projected_pooled=False,
+                textmodel_json_config=CLIPL_CONFIG,
+            )
+            load_into(f, self.model.transformer, "", "cpu", torch.float32)
+
+
+T5_CONFIG = {
+    "d_ff": 10240,
+    "d_model": 4096,
+    "num_heads": 64,
+    "num_layers": 24,
+    "vocab_size": 32128,
+}
+
+
+class T5XXL:
+    def __init__(self):
+        with safe_open("models/t5xxl.safetensors", framework="pt", device="cpu") as f:
+            self.model = T5XXLModel(T5_CONFIG, device="cpu", dtype=torch.float32)
+            load_into(f, self.model.transformer, "", "cpu", torch.float32)
+
+
+class SD3:
+    def __init__(self, model, shift, verbose=False):
+        with safe_open(model, framework="pt", device="cpu") as f:
+            self.model = BaseModel(
+                shift=shift,
+                file=f,
+                prefix="model.diffusion_model.",
+                device="cpu",
+                dtype=torch.float16,
+                verbose=verbose,
+            ).eval()
+            load_into(f, self.model, "model.", "cpu", torch.float16)
+
+
+class VAE:
+    def __init__(self, model):
+        with safe_open(model, framework="pt", device="cpu") as f:
+            self.model = SDVAE(device="cpu", dtype=torch.float16).eval().cpu()
+            prefix = ""
+            if any(k.startswith("first_stage_model.") for k in f.keys()):
+                prefix = "first_stage_model."
+            load_into(f, self.model, prefix, "cpu", torch.float16)
+
+
+#################################################################################################
+### Main inference logic
+#################################################################################################
+
+
+# Note: Sigma shift value, publicly released models use 3.0
+SHIFT = 3.0
+# Naturally, adjust to the width/height of the model you have
+WIDTH = 1024
+HEIGHT = 1024
+# Pick your prompt
+PROMPT = "a photo of a cat"
+# Most models prefer the range of 4-5, but still work well around 7
+CFG_SCALE = 4.5
+# Different models want different step counts but most will be good at 50, albeit that's slow to run
+# sd3_medium is quite decent at 28 steps
+STEPS = 40
+# Seed
+SEED = 23
+# SEEDTYPE = "fixed"
+SEEDTYPE = "rand"
+# SEEDTYPE = "roll"
+# Actual model file path
+# MODEL = "models/sd3_medium.safetensors"
+# MODEL = "models/sd3.5_large_turbo.safetensors"
+MODEL = "models/sd3.5_large.safetensors"
+# VAE model file path, or set None to use the same model file
+VAEFile = None  # "models/sd3_vae.safetensors"
+# Optional init image file path
+INIT_IMAGE = None
+# If init_image is given, this is the percentage of denoising steps to run (1.0 = full denoise, 0.0 = no denoise at all)
+DENOISE = 0.6
+# Output file path
+OUTDIR = "outputs"
+# SAMPLER
+# SAMPLER = "euler"
+SAMPLER = "dpmpp_2m"
+
+
+class SD3Inferencer:
+    def print(self, txt):
+        if self.verbose:
+            print(txt)
+
+    def load(self, model=MODEL, vae=VAEFile, shift=SHIFT, verbose=False):
+        self.verbose = verbose
+        print("Loading tokenizers...")
+        # NOTE: if you need a reference impl for a high performance CLIP tokenizer instead of just using the HF transformers one,
+        # check https://github.com/Stability-AI/StableSwarmUI/blob/master/src/Utils/CliplikeTokenizer.cs
+        # (T5 tokenizer is different though)
+        self.tokenizer = SD3Tokenizer()
+        print("Loading OpenAI CLIP L...")
+        self.clip_l = ClipL()
+        print("Loading OpenCLIP bigG...")
+        self.clip_g = ClipG()
+        print("Loading Google T5-v1-XXL...")
+        self.t5xxl = T5XXL()
+        print(f"Loading SD3 model {os.path.basename(model)}...")
+        self.sd3 = SD3(model, shift, verbose)
+        print("Loading VAE model...")
+        self.vae = VAE(vae or model)
+        print("Models loaded.")
+
+    def get_empty_latent(self, width, height):
+        self.print("Prep an empty latent...")
+        return torch.ones(1, 16, height // 8, width // 8, device="cpu") * 0.0609
+
+    def get_sigmas(self, sampling, steps):
+        start = sampling.timestep(sampling.sigma_max)
+        end = sampling.timestep(sampling.sigma_min)
+        timesteps = torch.linspace(start, end, steps)
+        sigs = []
+        for x in range(len(timesteps)):
+            ts = timesteps[x]
+            sigs.append(sampling.sigma(ts))
+        sigs += [0.0]
+        return torch.FloatTensor(sigs)
+
+    def get_noise(self, seed, latent):
+        generator = torch.manual_seed(seed)
+        self.print(f"dtype = {latent.dtype}, layout = {latent.layout}, device = {latent.device}")
+        return torch.randn(
+            latent.size(),
+            dtype=torch.float32,
+            layout=latent.layout,
+            generator=generator,
+            device="cpu",
+        ).to(latent.dtype)
+
+    def get_cond(self, prompt):
+        self.print("Encode prompt...")
+        tokens = self.tokenizer.tokenize_with_weights(prompt)
+        l_out, l_pooled = self.clip_l.model.encode_token_weights(tokens["l"])
+        g_out, g_pooled = self.clip_g.model.encode_token_weights(tokens["g"])
+        t5_out, t5_pooled = self.t5xxl.model.encode_token_weights(tokens["t5xxl"])
+        lg_out = torch.cat([l_out, g_out], dim=-1)
+        lg_out = torch.nn.functional.pad(lg_out, (0, 4096 - lg_out.shape[-1]))
+        return torch.cat([lg_out, t5_out], dim=-2), torch.cat((l_pooled, g_pooled), dim=-1)
+
+    def max_denoise(self, sigmas):
+        max_sigma = float(self.sd3.model.model_sampling.sigma_max)
+        sigma = float(sigmas[0])
+        return math.isclose(max_sigma, sigma, rel_tol=1e-05) or sigma > max_sigma
+
+    def fix_cond(self, cond):
+        cond, pooled = (cond[0].half().cuda(), cond[1].half().cuda())
+        return {"c_crossattn": cond, "y": pooled}
+
+    def do_sampling(
+        self,
+        latent,
+        seed,
+        conditioning,
+        neg_cond,
+        steps,
+        cfg_scale,
+        sampler="dpmpp_2m",
+        denoise=1.0,
+    ) -> torch.Tensor:
+        self.print("Sampling...")
+        latent = latent.half().cuda()
+        self.sd3.model = self.sd3.model.cuda()
+        noise = self.get_noise(seed, latent).cuda()
+        sigmas = self.get_sigmas(self.sd3.model.model_sampling, steps).cuda()
+        sigmas = sigmas[int(steps * (1 - denoise)) :]
+        conditioning = self.fix_cond(conditioning)
+        neg_cond = self.fix_cond(neg_cond)
+        extra_args = {"cond": conditioning, "uncond": neg_cond, "cond_scale": cfg_scale}
+        noise_scaled = self.sd3.model.model_sampling.noise_scaling(sigmas[0], noise, latent, self.max_denoise(sigmas))
+        sample_fn = getattr(sd3_impls, f"sample_{sampler}")
+        latent = sample_fn(CFGDenoiser(self.sd3.model), noise_scaled, sigmas, extra_args=extra_args)
+        latent = SD3LatentFormat().process_out(latent)
+        self.sd3.model = self.sd3.model.cpu()
+        self.print("Sampling done")
+        return latent
+
+    def vae_encode(self, image) -> torch.Tensor:
+        self.print("Encoding image to latent...")
+        image = image.convert("RGB")
+        image_np = np.array(image).astype(np.float32) / 255.0
+        image_np = np.moveaxis(image_np, 2, 0)
+        batch_images = np.expand_dims(image_np, axis=0).repeat(1, axis=0)
+        image_torch = torch.from_numpy(batch_images)
+        image_torch = 2.0 * image_torch - 1.0
+        image_torch = image_torch.cuda()
+        self.vae.model = self.vae.model.cuda()
+        latent = self.vae.model.encode(image_torch).cpu()
+        self.vae.model = self.vae.model.cpu()
+        self.print("Encoded")
+        return latent
+
+    def vae_decode(self, latent) -> Image.Image:
+        self.print("Decoding latent to image...")
+        latent = latent.cuda()
+        self.vae.model = self.vae.model.cuda()
+        image = self.vae.model.decode(latent)
+        image = image.float()
+        self.vae.model = self.vae.model.cpu()
+        image = torch.clamp((image + 1.0) / 2.0, min=0.0, max=1.0)[0]
+        decoded_np = 255.0 * np.moveaxis(image.cpu().numpy(), 0, 2)
+        decoded_np = decoded_np.astype(np.uint8)
+        out_image = Image.fromarray(decoded_np)
+        self.print("Decoded")
+        return out_image
+
+    def gen_image(
+        self,
+        prompts=[PROMPT],
+        width=WIDTH,
+        height=HEIGHT,
+        steps=STEPS,
+        cfg_scale=CFG_SCALE,
+        sampler=SAMPLER,
+        seed=SEED,
+        seed_type=SEEDTYPE,
+        out_dir=OUTDIR,
+        init_image=INIT_IMAGE,
+        denoise=DENOISE,
+    ):
+        latent = self.get_empty_latent(width, height)
+        if init_image:
+            image_data = Image.open(init_image)
+            image_data = image_data.resize((width, height), Image.LANCZOS)
+            latent = self.vae_encode(image_data)
+            latent = SD3LatentFormat().process_in(latent)
+        neg_cond = self.get_cond("")
+        seed_num = None
+        pbar = tqdm(enumerate(prompts), total=len(prompts), position=0, leave=True)
+        for i, prompt in pbar:
+            if seed_type == "roll":
+                seed_num = seed if seed_num is None else seed_num + 1
+            elif seed_type == "rand":
+                seed_num = torch.randint(0, 100000, (1,)).item()
+            else:  # fixed
+                seed_num = seed
+            conditioning = self.get_cond(prompt)
+            sampled_latent = self.do_sampling(
+                latent,
+                seed_num,
+                conditioning,
+                neg_cond,
+                steps,
+                cfg_scale,
+                sampler,
+                denoise if init_image else 1.0,
+            )
+            image = self.vae_decode(sampled_latent)
+            save_path = os.path.join(out_dir, f"{i:06d}.png")
+            self.print(f"Will save to {save_path}")
+            image.save(save_path)
+            self.print("Done")
+
+
+CONFIGS = {
+    "sd3_medium": {
+        "shift": 1.0,
+        "cfg": 5.0,
+        "steps": 50,
+        "sampler": "dpmpp_2m",
+    },
+    "sd3.5_large": {
+        "shift": 3.0,
+        "cfg": 4.5,
+        "steps": 40,
+        "sampler": "dpmpp_2m",
+    },
+    "sd3.5_large_turbo": {"shift": 3.0, "cfg": 1.0, "steps": 4, "sampler": "euler"},
+}
+
+
+@torch.no_grad()
+def main(
+    prompt=PROMPT,
+    model=MODEL,
+    out_dir=OUTDIR,
+    postfix=None,
+    seed=SEED,
+    seed_type=SEEDTYPE,
+    sampler=None,
+    steps=None,
+    cfg=None,
+    shift=None,
+    width=WIDTH,
+    height=HEIGHT,
+    vae=VAEFile,
+    init_image=INIT_IMAGE,
+    denoise=DENOISE,
+    verbose=False,
+):
+    steps = steps or CONFIGS[os.path.splitext(os.path.basename(model))[0]]["steps"]
+    cfg = cfg or CONFIGS[os.path.splitext(os.path.basename(model))[0]]["cfg"]
+    shift = shift or CONFIGS[os.path.splitext(os.path.basename(model))[0]]["shift"]
+    sampler = sampler or CONFIGS[os.path.splitext(os.path.basename(model))[0]]["sampler"]
+
+    inferencer = SD3Inferencer()
+    inferencer.load(model, vae, shift, verbose)
+
+    if isinstance(prompt, str):
+        if os.path.splitext(prompt)[-1] == ".txt":
+            with open(prompt, "r") as f:
+                prompts = [l.strip() for l in f.readlines()]
+        else:
+            prompts = [prompt]
+
+    out_dir = os.path.join(
+        out_dir,
+        os.path.splitext(os.path.basename(model))[0],
+        os.path.splitext(os.path.basename(prompt))[0][:50]
+        + (postfix or datetime.datetime.now().strftime("_%Y-%m-%dT%H-%M-%S")),
+    )
+    print(f"Saving images to {out_dir}")
+    os.makedirs(out_dir, exist_ok=False)
+
+    inferencer.gen_image(
+        prompts,
+        width,
+        height,
+        steps,
+        cfg,
+        sampler,
+        seed,
+        seed_type,
+        out_dir,
+        init_image,
+        denoise,
+    )
+
+
+fire.Fire(main)

invokeai/backend/sd3/sd3_mmditx.py

@@ -0,0 +1,72 @@
+from dataclasses import dataclass
+from typing import Literal, TypedDict
+
+import torch
+
+from invokeai.backend.sd3.mmditx import MMDiTX
+from invokeai.backend.sd3.sd3_impls import ModelSamplingDiscreteFlow
+
+
+class ContextEmbedderConfig(TypedDict):
+    target: Literal["torch.nn.Linear"]
+    params: dict[str, int]
+
+
+@dataclass
+class Sd3MMDiTXParams:
+    patch_size: int
+    depth: int
+    num_patches: int
+    pos_embed_max_size: int
+    adm_in_channels: int
+    context_shape: tuple[int, int]
+    qk_norm: Literal["rms", None]
+    x_block_self_attn_layers: list[int]
+    context_embedder_config: ContextEmbedderConfig
+
+
+class Sd3MMDiTX(torch.nn.Module):
+    """This class is based closely on
+    https://github.com/Stability-AI/sd3.5/blob/19bf11c4e1e37324c5aa5a61f010d4127848a09c/sd3_impls.py#L53
+    but has more standard model loading semantics.
+    """
+
+    def __init__(
+        self,
+        params: Sd3MMDiTXParams,
+        shift: float = 1.0,
+        device: torch.device | None = None,
+        dtype: torch.dtype | None = None,
+        verbose: bool = False,
+    ):
+        super().__init__()
+        self.diffusion_model = MMDiTX(
+            input_size=None,
+            pos_embed_scaling_factor=None,
+            pos_embed_offset=None,
+            pos_embed_max_size=params.pos_embed_max_size,
+            patch_size=params.patch_size,
+            in_channels=16,
+            depth=params.depth,
+            num_patches=params.num_patches,
+            adm_in_channels=params.adm_in_channels,
+            context_embedder_config=params.context_embedder_config,
+            qk_norm=params.qk_norm,
+            x_block_self_attn_layers=params.x_block_self_attn_layers,
+            device=device,
+            dtype=dtype,
+            verbose=verbose,
+        )
+        self.model_sampling = ModelSamplingDiscreteFlow(shift=shift)
+
+    def apply_model(self, x: torch.Tensor, sigma: torch.Tensor, c_crossattn: torch.Tensor, y: torch.Tensor):
+        dtype = self.get_dtype()
+        timestep = self.model_sampling.timestep(sigma).float()
+        model_output = self.diffusion_model(x.to(dtype), timestep, context=c_crossattn.to(dtype), y=y.to(dtype)).float()
+        return self.model_sampling.calculate_denoised(sigma, model_output, x)
+
+    def forward(self, x: torch.Tensor, sigma: float, c_crossattn: torch.Tensor, y: torch.Tensor):
+        return self.apply_model(x=x, sigma=sigma, c_crossattn=c_crossattn, y=y)
+
+    def get_dtype(self):
+        return self.diffusion_model.dtype

invokeai/backend/sd3/sd3_state_dict_utils.py

@@ -0,0 +1,70 @@
+import math
+import re
+from typing import Any, Dict
+
+from invokeai.backend.sd3.sd3_mmditx import ContextEmbedderConfig, Sd3MMDiTXParams
+
+
+def is_sd3_checkpoint(sd: Dict[str, Any]) -> bool:
+    """Is the state dict for an SD3 checkpoint like this one?:
+    https://huggingface.co/stabilityai/stable-diffusion-3.5-large/blob/main/sd3.5_large.safetensors
+
+    Note that this checkpoint format contains both the VAE and the MMDiTX model.
+
+    This is intended to be a reasonably high-precision detector, but it is not guaranteed to have perfect precision.
+    """
+    # If all of the expected keys are present, then this is very likely a SD3 checkpoint.
+    expected_keys = {
+        # VAE decoder and encoder keys.
+        "first_stage_model.decoder.conv_in.bias",
+        "first_stage_model.decoder.conv_in.weight",
+        "first_stage_model.encoder.conv_in.bias",
+        "first_stage_model.encoder.conv_in.weight",
+        # MMDiTX keys.
+        "model.diffusion_model.final_layer.linear.bias",
+        "model.diffusion_model.final_layer.linear.weight",
+        "model.diffusion_model.joint_blocks.0.context_block.attn.ln_k.weight",
+        "model.diffusion_model.joint_blocks.0.context_block.attn.ln_q.weight",
+    }
+
+    return expected_keys.issubset(sd.keys())
+
+
+def infer_sd3_mmditx_params(sd: Dict[str, Any], prefix: str = "model.diffusion_model.") -> Sd3MMDiTXParams:
+    """Infer the MMDiTX model parameters from the state dict.
+
+    This logic is based on:
+    https://github.com/Stability-AI/sd3.5/blob/19bf11c4e1e37324c5aa5a61f010d4127848a09c/sd3_impls.py#L68-L88
+    """
+    patch_size = sd[f"{prefix}x_embedder.proj.weight"].shape[2]
+    depth = sd[f"{prefix}x_embedder.proj.weight"].shape[0] // 64
+    num_patches = sd[f"{prefix}pos_embed"].shape[1]
+    pos_embed_max_size = round(math.sqrt(num_patches))
+    adm_in_channels = sd[f"{prefix}y_embedder.mlp.0.weight"].shape[1]
+    context_shape = sd[f"{prefix}context_embedder.weight"].shape
+    qk_norm = "rms" if f"{prefix}joint_blocks.0.context_block.attn.ln_k.weight" in sd else None
+    x_block_self_attn_layers = sorted(
+        [
+            int(key.split(".x_block.attn2.ln_k.weight")[0].split(".")[-1])
+            for key in list(filter(re.compile(".*.x_block.attn2.ln_k.weight").match, sd.keys()))
+        ]
+    )
+
+    context_embedder_config: ContextEmbedderConfig = {
+        "target": "torch.nn.Linear",
+        "params": {
+            "in_features": context_shape[1],
+            "out_features": context_shape[0],
+        },
+    }
+    return Sd3MMDiTXParams(
+        patch_size=patch_size,
+        depth=depth,
+        num_patches=num_patches,
+        pos_embed_max_size=pos_embed_max_size,
+        adm_in_channels=adm_in_channels,
+        context_shape=context_shape,
+        qk_norm=qk_norm,
+        x_block_self_attn_layers=x_block_self_attn_layers,
+        context_embedder_config=context_embedder_config,
+    )

tests/backend/sd3/test_sd3_state_dict_utils.py

@@ -0,0 +1,43 @@
+import pytest
+import torch
+
+from invokeai.backend.sd3.sd3_mmditx import Sd3MMDiTX
+from invokeai.backend.sd3.sd3_state_dict_utils import infer_sd3_mmditx_params, is_sd3_checkpoint
+from tests.backend.sd3.sd3_5_mmditx_state_dict import sd3_sd_shapes
+
+
+@pytest.mark.parametrize(
+    ["sd_shapes", "expected"],
+    [
+        (sd3_sd_shapes, True),
+        ({}, False),
+        ({"foo": [1]}, False),
+    ],
+)
+def test_is_sd3_checkpoint(sd_shapes: dict[str, list[int]], expected: bool):
+    # Build mock state dict from the provided shape dict.
+    sd = {k: None for k in sd_shapes}
+    assert is_sd3_checkpoint(sd) == expected
+
+
+def test_infer_sd3_mmditx_params():
+    # Build mock state dict on the meta device.
+    with torch.device("meta"):
+        sd = {k: torch.zeros(shape) for k, shape in sd3_sd_shapes.items()}
+
+    # Filter the MMDiTX parameters from the state dict.
+    sd = {k: v for k, v in sd.items() if k.startswith("model.diffusion_model.")}
+
+    params = infer_sd3_mmditx_params(sd)
+
+    # Construct model from params.
+    with torch.device("meta"):
+        model = Sd3MMDiTX(params=params)
+
+    model_sd = model.state_dict()
+
+    # Assert that the model state dict is compatible with the original state dict.
+    sd_without_prefix = {k.split("model.diffusion_model.")[-1]: v for k, v in model_sd.items()}
+    assert set(model_sd.keys()) == set(sd_without_prefix.keys())
+    for k in model_sd:
+        assert model_sd[k].shape == sd_without_prefix[k].shape