Upload hy3dgen/shapegen/models/hunyuan3ddit.py with huggingface_hub

hy3dgen/shapegen/models/hunyuan3ddit.py

@@ -0,0 +1,390 @@
+# Open Source Model Licensed under the Apache License Version 2.0
+# and Other Licenses of the Third-Party Components therein:
+# The below Model in this distribution may have been modified by THL A29 Limited
+# ("Tencent Modifications"). All Tencent Modifications are Copyright (C) 2024 THL A29 Limited.
+
+# Copyright (C) 2024 THL A29 Limited, a Tencent company.  All rights reserved.
+# The below software and/or models in this distribution may have been
+# modified by THL A29 Limited ("Tencent Modifications").
+# All Tencent Modifications are Copyright (C) THL A29 Limited.
+
+# Hunyuan 3D is licensed under the TENCENT HUNYUAN NON-COMMERCIAL LICENSE AGREEMENT
+# except for the third-party components listed below.
+# Hunyuan 3D does not impose any additional limitations beyond what is outlined
+# in the repsective licenses of these third-party components.
+# Users must comply with all terms and conditions of original licenses of these third-party
+# components and must ensure that the usage of the third party components adheres to
+# all relevant laws and regulations.
+
+# For avoidance of doubts, Hunyuan 3D means the large language models and
+# their software and algorithms, including trained model weights, parameters (including
+# optimizer states), machine-learning model code, inference-enabling code, training-enabling code,
+# fine-tuning enabling code and other elements of the foregoing made publicly available
+# by Tencent in accordance with TENCENT HUNYUAN COMMUNITY LICENSE AGREEMENT.
+
+import math
+from dataclasses import dataclass
+from typing import List, Tuple, Optional
+
+import torch
+from einops import rearrange
+from torch import Tensor, nn
+
+
+def attention(q: Tensor, k: Tensor, v: Tensor, **kwargs) -> Tensor:
+    x = torch.nn.functional.scaled_dot_product_attention(q, k, v)
+    x = rearrange(x, "B H L D -> B L (H D)")
+    return x
+
+
+def timestep_embedding(t: Tensor, dim, max_period=10000, time_factor: float = 1000.0):
+    """
+    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.
+    """
+    t = time_factor * t
+    half = dim // 2
+    freqs = torch.exp(-math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half).to(
+        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(t)
+    return embedding
+
+
+class MLPEmbedder(nn.Module):
+    def __init__(self, in_dim: int, hidden_dim: int):
+        super().__init__()
+        self.in_layer = nn.Linear(in_dim, hidden_dim, bias=True)
+        self.silu = nn.SiLU()
+        self.out_layer = nn.Linear(hidden_dim, hidden_dim, bias=True)
+
+    def forward(self, x: Tensor) -> Tensor:
+        return self.out_layer(self.silu(self.in_layer(x)))
+
+
+class RMSNorm(torch.nn.Module):
+    def __init__(self, dim: int):
+        super().__init__()
+        self.scale = nn.Parameter(torch.ones(dim))
+
+    def forward(self, x: Tensor):
+        x_dtype = x.dtype
+        x = x.float()
+        rrms = torch.rsqrt(torch.mean(x ** 2, dim=-1, keepdim=True) + 1e-6)
+        return (x * rrms).to(dtype=x_dtype) * self.scale
+
+
+class QKNorm(torch.nn.Module):
+    def __init__(self, dim: int):
+        super().__init__()
+        self.query_norm = RMSNorm(dim)
+        self.key_norm = RMSNorm(dim)
+
+    def forward(self, q: Tensor, k: Tensor, v: Tensor) -> Tuple[Tensor, Tensor]:
+        q = self.query_norm(q)
+        k = self.key_norm(k)
+        return q.to(v), k.to(v)
+
+
+class SelfAttention(nn.Module):
+    def __init__(
+        self,
+        dim: int,
+        num_heads: int = 8,
+        qkv_bias: bool = False,
+    ):
+        super().__init__()
+        self.num_heads = num_heads
+        head_dim = dim // num_heads
+
+        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
+        self.norm = QKNorm(head_dim)
+        self.proj = nn.Linear(dim, dim)
+
+    def forward(self, x: Tensor, pe: Tensor) -> Tensor:
+        qkv = self.qkv(x)
+        q, k, v = rearrange(qkv, "B L (K H D) -> K B H L D", K=3, H=self.num_heads)
+        q, k = self.norm(q, k, v)
+        x = attention(q, k, v, pe=pe)
+        x = self.proj(x)
+        return x
+
+
+@dataclass
+class ModulationOut:
+    shift: Tensor
+    scale: Tensor
+    gate: Tensor
+
+
+class Modulation(nn.Module):
+    def __init__(self, dim: int, double: bool):
+        super().__init__()
+        self.is_double = double
+        self.multiplier = 6 if double else 3
+        self.lin = nn.Linear(dim, self.multiplier * dim, bias=True)
+
+    def forward(self, vec: Tensor) -> Tuple[ModulationOut, Optional[ModulationOut]]:
+        out = self.lin(nn.functional.silu(vec))[:, None, :]
+        out = out.chunk(self.multiplier, dim=-1)
+
+        return (
+            ModulationOut(*out[:3]),
+            ModulationOut(*out[3:]) if self.is_double else None,
+        )
+
+
+class DoubleStreamBlock(nn.Module):
+    def __init__(
+        self,
+        hidden_size: int,
+        num_heads: int,
+        mlp_ratio: float,
+        qkv_bias: bool = False,
+    ):
+        super().__init__()
+        mlp_hidden_dim = int(hidden_size * mlp_ratio)
+        self.num_heads = num_heads
+        self.hidden_size = hidden_size
+        self.img_mod = Modulation(hidden_size, double=True)
+        self.img_norm1 = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
+        self.img_attn = SelfAttention(dim=hidden_size, num_heads=num_heads, qkv_bias=qkv_bias)
+
+        self.img_norm2 = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
+        self.img_mlp = nn.Sequential(
+            nn.Linear(hidden_size, mlp_hidden_dim, bias=True),
+            nn.GELU(approximate="tanh"),
+            nn.Linear(mlp_hidden_dim, hidden_size, bias=True),
+        )
+
+        self.txt_mod = Modulation(hidden_size, double=True)
+        self.txt_norm1 = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
+        self.txt_attn = SelfAttention(dim=hidden_size, num_heads=num_heads, qkv_bias=qkv_bias)
+
+        self.txt_norm2 = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
+        self.txt_mlp = nn.Sequential(
+            nn.Linear(hidden_size, mlp_hidden_dim, bias=True),
+            nn.GELU(approximate="tanh"),
+            nn.Linear(mlp_hidden_dim, hidden_size, bias=True),
+        )
+
+    def forward(self, img: Tensor, txt: Tensor, vec: Tensor, pe: Tensor) -> Tuple[Tensor, Tensor]:
+        img_mod1, img_mod2 = self.img_mod(vec)
+        txt_mod1, txt_mod2 = self.txt_mod(vec)
+
+        img_modulated = self.img_norm1(img)
+        img_modulated = (1 + img_mod1.scale) * img_modulated + img_mod1.shift
+        img_qkv = self.img_attn.qkv(img_modulated)
+        img_q, img_k, img_v = rearrange(img_qkv, "B L (K H D) -> K B H L D", K=3, H=self.num_heads)
+        img_q, img_k = self.img_attn.norm(img_q, img_k, img_v)
+
+        txt_modulated = self.txt_norm1(txt)
+        txt_modulated = (1 + txt_mod1.scale) * txt_modulated + txt_mod1.shift
+        txt_qkv = self.txt_attn.qkv(txt_modulated)
+        txt_q, txt_k, txt_v = rearrange(txt_qkv, "B L (K H D) -> K B H L D", K=3, H=self.num_heads)
+        txt_q, txt_k = self.txt_attn.norm(txt_q, txt_k, txt_v)
+
+        q = torch.cat((txt_q, img_q), dim=2)
+        k = torch.cat((txt_k, img_k), dim=2)
+        v = torch.cat((txt_v, img_v), dim=2)
+
+        attn = attention(q, k, v, pe=pe)
+        txt_attn, img_attn = attn[:, : txt.shape[1]], attn[:, txt.shape[1]:]
+
+        img = img + img_mod1.gate * self.img_attn.proj(img_attn)
+        img = img + img_mod2.gate * self.img_mlp((1 + img_mod2.scale) * self.img_norm2(img) + img_mod2.shift)
+
+        txt = txt + txt_mod1.gate * self.txt_attn.proj(txt_attn)
+        txt = txt + txt_mod2.gate * self.txt_mlp((1 + txt_mod2.scale) * self.txt_norm2(txt) + txt_mod2.shift)
+        return img, txt
+
+
+class SingleStreamBlock(nn.Module):
+    """
+    A DiT block with parallel linear layers as described in
+    https://arxiv.org/abs/2302.05442 and adapted modulation interface.
+    """
+
+    def __init__(
+        self,
+        hidden_size: int,
+        num_heads: int,
+        mlp_ratio: float = 4.0,
+        qk_scale: Optional[float] = None,
+    ):
+        super().__init__()
+
+        self.hidden_dim = hidden_size
+        self.num_heads = num_heads
+        head_dim = hidden_size // num_heads
+        self.scale = qk_scale or head_dim ** -0.5
+
+        self.mlp_hidden_dim = int(hidden_size * mlp_ratio)
+        # qkv and mlp_in
+        self.linear1 = nn.Linear(hidden_size, hidden_size * 3 + self.mlp_hidden_dim)
+        # proj and mlp_out
+        self.linear2 = nn.Linear(hidden_size + self.mlp_hidden_dim, hidden_size)
+
+        self.norm = QKNorm(head_dim)
+
+        self.hidden_size = hidden_size
+        self.pre_norm = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
+
+        self.mlp_act = nn.GELU(approximate="tanh")
+        self.modulation = Modulation(hidden_size, double=False)
+
+    def forward(self, x: Tensor, vec: Tensor, pe: Tensor) -> Tensor:
+        mod, _ = self.modulation(vec)
+
+        x_mod = (1 + mod.scale) * self.pre_norm(x) + mod.shift
+        qkv, mlp = torch.split(self.linear1(x_mod), [3 * self.hidden_size, self.mlp_hidden_dim], dim=-1)
+
+        q, k, v = rearrange(qkv, "B L (K H D) -> K B H L D", K=3, H=self.num_heads)
+        q, k = self.norm(q, k, v)
+
+        # compute attention
+        attn = attention(q, k, v, pe=pe)
+        # compute activation in mlp stream, cat again and run second linear layer
+        output = self.linear2(torch.cat((attn, self.mlp_act(mlp)), 2))
+        return x + mod.gate * output
+
+
+class LastLayer(nn.Module):
+    def __init__(self, hidden_size: int, patch_size: int, out_channels: int):
+        super().__init__()
+        self.norm_final = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
+        self.linear = nn.Linear(hidden_size, patch_size * patch_size * out_channels, bias=True)
+        self.adaLN_modulation = nn.Sequential(nn.SiLU(), nn.Linear(hidden_size, 2 * hidden_size, bias=True))
+
+    def forward(self, x: Tensor, vec: Tensor) -> Tensor:
+        shift, scale = self.adaLN_modulation(vec).chunk(2, dim=1)
+        x = (1 + scale[:, None, :]) * self.norm_final(x) + shift[:, None, :]
+        x = self.linear(x)
+        return x
+
+
+class Hunyuan3DDiT(nn.Module):
+    def __init__(
+        self,
+        in_channels: int = 64,
+        context_in_dim: int = 1536,
+        hidden_size: int = 1024,
+        mlp_ratio: float = 4.0,
+        num_heads: int = 16,
+        depth: int = 16,
+        depth_single_blocks: int = 32,
+        axes_dim: List[int] = [64],
+        theta: int = 10_000,
+        qkv_bias: bool = True,
+        time_factor: float = 1000,
+        ckpt_path: Optional[str] = None,
+        **kwargs,
+    ):
+        super().__init__()
+        self.in_channels = in_channels
+        self.context_in_dim = context_in_dim
+        self.hidden_size = hidden_size
+        self.mlp_ratio = mlp_ratio
+        self.num_heads = num_heads
+        self.depth = depth
+        self.depth_single_blocks = depth_single_blocks
+        self.axes_dim = axes_dim
+        self.theta = theta
+        self.qkv_bias = qkv_bias
+        self.time_factor = time_factor
+        self.out_channels = self.in_channels
+
+        if hidden_size % num_heads != 0:
+            raise ValueError(
+                f"Hidden size {hidden_size} must be divisible by num_heads {num_heads}"
+            )
+        pe_dim = hidden_size // num_heads
+        if sum(axes_dim) != pe_dim:
+            raise ValueError(f"Got {axes_dim} but expected positional dim {pe_dim}")
+        self.hidden_size = hidden_size
+        self.num_heads = num_heads
+        self.latent_in = nn.Linear(self.in_channels, self.hidden_size, bias=True)
+        self.time_in = MLPEmbedder(in_dim=256, hidden_dim=self.hidden_size)
+        self.cond_in = nn.Linear(context_in_dim, self.hidden_size)
+
+        self.double_blocks = nn.ModuleList(
+            [
+                DoubleStreamBlock(
+                    self.hidden_size,
+                    self.num_heads,
+                    mlp_ratio=mlp_ratio,
+                    qkv_bias=qkv_bias,
+                )
+                for _ in range(depth)
+            ]
+        )
+
+        self.single_blocks = nn.ModuleList(
+            [
+                SingleStreamBlock(
+                    self.hidden_size,
+                    self.num_heads,
+                    mlp_ratio=mlp_ratio,
+                )
+                for _ in range(depth_single_blocks)
+            ]
+        )
+
+        self.final_layer = LastLayer(self.hidden_size, 1, self.out_channels)
+
+        if ckpt_path is not None:
+            print('restored denoiser ckpt', ckpt_path)
+
+            ckpt = torch.load(ckpt_path, map_location="cpu")
+            if 'state_dict' not in ckpt:
+                # deepspeed ckpt
+                state_dict = {}
+                for k in ckpt.keys():
+                    new_k = k.replace('_forward_module.', '')
+                    state_dict[new_k] = ckpt[k]
+            else:
+                state_dict = ckpt["state_dict"]
+
+            final_state_dict = {}
+            for k, v in state_dict.items():
+                if k.startswith('model.'):
+                    final_state_dict[k.replace('model.', '')] = v
+                else:
+                    final_state_dict[k] = v
+            missing, unexpected = self.load_state_dict(final_state_dict, strict=False)
+            print('unexpected keys:', unexpected)
+            print('missing keys:', missing)
+
+    def forward(
+        self,
+        x,
+        t,
+        contexts,
+        **kwargs,
+    ) -> Tensor:
+        cond = contexts['main']
+        latent = self.latent_in(x)
+        vec = self.time_in(timestep_embedding(t, 256, self.time_factor).to(dtype=latent.dtype))
+        cond = self.cond_in(cond)
+        pe = None
+
+        for block in self.double_blocks:
+            latent, cond = block(img=latent, txt=cond, vec=vec, pe=pe)
+
+        latent = torch.cat((cond, latent), 1)
+        for block in self.single_blocks:
+            latent = block(latent, vec=vec, pe=pe)
+
+        latent = latent[:, cond.shape[1]:, ...]
+        latent = self.final_layer(latent, vec)
+        return latent