diff --git a/docs/source/en/api/models/glm_image_transformer2d.md b/docs/source/en/api/models/glm_image_transformer2d.md new file mode 100644 index 0000000000..d31557c9f7 --- /dev/null +++ b/docs/source/en/api/models/glm_image_transformer2d.md @@ -0,0 +1,18 @@ + + +# GlmImageDecoderTransformer2DModel + +A Diffusion Transformer model for 2D data from [GlmImageDecoderTransformer2DModel]() + +## GlmImageDecoderTransformer2DModel + +[[autodoc]] GlmImageDecoderTransformer2DModel diff --git a/docs/source/en/api/pipelines/glm_image.md b/docs/source/en/api/pipelines/glm_image.md new file mode 100644 index 0000000000..24b5e14a1a --- /dev/null +++ b/docs/source/en/api/pipelines/glm_image.md @@ -0,0 +1,31 @@ + + +# GLM-Image + +> [!TIP] +> Make sure to check out the Schedulers [guide](../../using-diffusers/schedulers) to learn how to explore the tradeoff between scheduler speed and quality, and see the [reuse components across pipelines](../../using-diffusers/loading#reuse-a-pipeline) section to learn how to efficiently load the same components into multiple pipelines. + +This pipeline was contributed by [zRzRzRzRzRzRzR](https://github.com/zRzRzRzRzRzRzR). The original codebase can be found [here](https://huggingface.co/zai-org). The original weights can be found under [hf.co/zai-org](https://huggingface.co/zai-org). + +## GlmImageDecoderPipeline + +[[autodoc]] GlmImageDecoderPipeline + - all + - __call__ + +## GlmImageDecoderPipelineOutput + +[[autodoc]] pipelines.cogview4.pipeline_output.GlmImageDecoderPipelineOutput diff --git a/src/diffusers/models/__init__.py b/src/diffusers/models/__init__.py index c4664f00ca..29bf6016de 100755 --- a/src/diffusers/models/__init__.py +++ b/src/diffusers/models/__init__.py @@ -96,6 +96,7 @@ if is_torch_available(): _import_structure["transformers.transformer_easyanimate"] = ["EasyAnimateTransformer3DModel"] _import_structure["transformers.transformer_flux"] = ["FluxTransformer2DModel"] _import_structure["transformers.transformer_flux2"] = ["Flux2Transformer2DModel"] + _import_structure["transformers.transformer_glm_image"] = ["GlmImageDecoderTransformer2DModel"] _import_structure["transformers.transformer_hidream_image"] = ["HiDreamImageTransformer2DModel"] _import_structure["transformers.transformer_hunyuan_video"] = ["HunyuanVideoTransformer3DModel"] _import_structure["transformers.transformer_hunyuan_video15"] = ["HunyuanVideo15Transformer3DModel"] @@ -203,6 +204,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT: EasyAnimateTransformer3DModel, Flux2Transformer2DModel, FluxTransformer2DModel, + GlmImageDecoderTransformer2DModel, HiDreamImageTransformer2DModel, HunyuanDiT2DModel, HunyuanImageTransformer2DModel, diff --git a/src/diffusers/models/embeddings.py b/src/diffusers/models/embeddings.py index 37fc412adc..14fd9e2de9 100644 --- a/src/diffusers/models/embeddings.py +++ b/src/diffusers/models/embeddings.py @@ -1658,6 +1658,37 @@ class CogView3CombinedTimestepSizeEmbeddings(nn.Module): return conditioning +class GlmImageDecoderCombinedTimestepSizeEmbeddings(nn.Module): + def __init__(self, embedding_dim: int, condition_dim: int, pooled_projection_dim: int, timesteps_dim: int = 256): + super().__init__() + + self.time_proj = Timesteps(num_channels=timesteps_dim, flip_sin_to_cos=True, downscale_freq_shift=0) + self.condition_proj = Timesteps(num_channels=condition_dim, flip_sin_to_cos=True, downscale_freq_shift=0) + self.timestep_embedder = TimestepEmbedding(in_channels=timesteps_dim, time_embed_dim=embedding_dim) + self.condition_embedder = PixArtAlphaTextProjection(pooled_projection_dim, embedding_dim, act_fn="silu") + + def forward( + self, + timestep: torch.Tensor, + target_size: torch.Tensor, + crop_coords: torch.Tensor, + hidden_dtype: torch.dtype, + ) -> torch.Tensor: + timesteps_proj = self.time_proj(timestep) + + crop_coords_proj = self.condition_proj(crop_coords.flatten()).view(crop_coords.size(0), -1) + target_size_proj = self.condition_proj(target_size.flatten()).view(target_size.size(0), -1) + + # (B, 2 * condition_dim) + condition_proj = torch.cat([crop_coords_proj, target_size_proj], dim=1) + + timesteps_emb = self.timestep_embedder(timesteps_proj.to(dtype=hidden_dtype)) # (B, embedding_dim) + condition_emb = self.condition_embedder(condition_proj.to(dtype=hidden_dtype)) # (B, embedding_dim) + + conditioning = timesteps_emb + condition_emb + return conditioning + + class HunyuanDiTAttentionPool(nn.Module): # Copied from https://github.com/Tencent/HunyuanDiT/blob/cb709308d92e6c7e8d59d0dff41b74d35088db6a/hydit/modules/poolers.py#L6 diff --git a/src/diffusers/models/transformers/__init__.py b/src/diffusers/models/transformers/__init__.py index 40b5d4a0df..ea051624f2 100755 --- a/src/diffusers/models/transformers/__init__.py +++ b/src/diffusers/models/transformers/__init__.py @@ -27,6 +27,7 @@ if is_torch_available(): from .transformer_easyanimate import EasyAnimateTransformer3DModel from .transformer_flux import FluxTransformer2DModel from .transformer_flux2 import Flux2Transformer2DModel + from .transformer_glm_image import GlmImageDecoderTransformer2DModel from .transformer_hidream_image import HiDreamImageTransformer2DModel from .transformer_hunyuan_video import HunyuanVideoTransformer3DModel from .transformer_hunyuan_video15 import HunyuanVideo15Transformer3DModel diff --git a/src/diffusers/models/transformers/transformer_glm_image.py b/src/diffusers/models/transformers/transformer_glm_image.py new file mode 100644 index 0000000000..54a1c7cbdb --- /dev/null +++ b/src/diffusers/models/transformers/transformer_glm_image.py @@ -0,0 +1,753 @@ +# Copyright 2025 The CogView team, Tsinghua University & ZhipuAI and The HuggingFace Team. All rights reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from enum import Enum +from typing import Any, Dict, List, Optional, Tuple, Union + +import torch +import torch.nn as nn +import torch.nn.functional as F +from einops import rearrange + +from ...configuration_utils import ConfigMixin, register_to_config +from ...loaders import PeftAdapterMixin +from ...utils import logging +from ...utils.torch_utils import maybe_allow_in_graph +from ..attention import FeedForward +from ..attention_processor import Attention +from ..cache_utils import CacheMixin +from ..embeddings import GlmImageDecoderCombinedTimestepSizeEmbeddings +from ..modeling_outputs import Transformer2DModelOutput +from ..modeling_utils import ModelMixin +from ..normalization import LayerNorm, RMSNorm + + +logger = logging.get_logger(__name__) # pylint: disable=invalid-name + + +class GlmImageDecoderImageProjector(nn.Module): + def __init__( + self, + in_channels: int = 16, + hidden_size: int = 2560, + patch_size: int = 2, + ): + super().__init__() + self.patch_size = patch_size + + self.proj = nn.Linear(in_channels * patch_size**2, hidden_size) + + def forward(self, hidden_states: torch.Tensor) -> torch.Tensor: + batch_size, channel, height, width = hidden_states.shape + post_patch_height = height // self.patch_size + post_patch_width = width // self.patch_size + + hidden_states = hidden_states.reshape( + batch_size, channel, post_patch_height, self.patch_size, post_patch_width, self.patch_size + ) + hidden_states = hidden_states.permute(0, 2, 4, 1, 3, 5).flatten(3, 5).flatten(1, 2) + hidden_states = self.proj(hidden_states) + + return hidden_states + + +class GlmImageDecoderAdaLayerNormZero(nn.Module): + def __init__(self, embedding_dim: int, dim: int) -> None: + super().__init__() + + self.norm = nn.LayerNorm(dim, elementwise_affine=False, eps=1e-5) + self.norm_context = nn.LayerNorm(dim, elementwise_affine=False, eps=1e-5) + self.linear = nn.Linear(embedding_dim, 12 * dim, bias=True) + + def forward( + self, hidden_states: torch.Tensor, glyph_hidden_states: torch.Tensor, temb: torch.Tensor + ) -> Tuple[torch.Tensor, torch.Tensor]: + dtype = hidden_states.dtype + norm_hidden_states = self.norm(hidden_states).to(dtype=dtype) + norm_glyph_hidden_states = self.norm_context(glyph_hidden_states).to(dtype=dtype) + + emb = self.linear(temb) + ( + shift_msa, + c_shift_msa, + scale_msa, + c_scale_msa, + gate_msa, + c_gate_msa, + shift_mlp, + c_shift_mlp, + scale_mlp, + c_scale_mlp, + gate_mlp, + c_gate_mlp, + ) = emb.chunk(12, dim=1) + + hidden_states = norm_hidden_states * (1 + scale_msa.unsqueeze(1)) + shift_msa.unsqueeze(1) + glyph_hidden_states = norm_glyph_hidden_states * (1 + c_scale_msa.unsqueeze(1)) + c_shift_msa.unsqueeze(1) + + return ( + hidden_states, + gate_msa, + shift_mlp, + scale_mlp, + gate_mlp, + glyph_hidden_states, + c_gate_msa, + c_shift_mlp, + c_scale_mlp, + c_gate_mlp, + ) + + +class GlmImageDecoderAttenProcessorState(Enum): + ImageGen = "ImageGen" + ImageEditWriteKV = "ImageEditWriteKV" + ImageEditReadKV = "ImageEditReadKV" + ImageEditDontReadKV = "ImageEditNoReadKV" + + +class GlmImageDecoderAttnProcessor: + """ + Processor for implementing scaled dot-product attention for the GlmImageDecoder model. It applies a rotary + embedding on query and key vectors, but does not include spatial normalization. + + The processor supports passing an attention mask for text tokens. The attention mask should have shape (batch_size, + text_seq_length) where 1 indicates a non-padded token and 0 indicates a padded token. + """ + + def __init__(self): + if not hasattr(F, "scaled_dot_product_attention"): + raise ImportError( + "GlmImageDecoderAttnProcessor requires PyTorch 2.0. To use it, please upgrade PyTorch to 2.0." + ) + self.processor_state = GlmImageDecoderAttenProcessorState.ImageGen + self.k_cache = None + self.v_cache = None + + def __call__( + self, + attn: Attention, + hidden_states: torch.Tensor, + encoder_hidden_states: torch.Tensor, + attention_mask: Optional[torch.Tensor] = None, + image_rotary_emb: Optional[Tuple[torch.Tensor, torch.Tensor]] = None, + ) -> Tuple[torch.Tensor, torch.Tensor]: + dtype = encoder_hidden_states.dtype + + batch_size, text_seq_length, embed_dim = encoder_hidden_states.shape + batch_size, image_seq_length, embed_dim = hidden_states.shape + hidden_states = torch.cat([encoder_hidden_states, hidden_states], dim=1) + + # 1. QKV projections + query = attn.to_q(hidden_states) + key = attn.to_k(hidden_states) + value = attn.to_v(hidden_states) + + query = query.unflatten(2, (attn.heads, -1)).transpose(1, 2) + key = key.unflatten(2, (attn.heads, -1)).transpose(1, 2) + value = value.unflatten(2, (attn.heads, -1)).transpose(1, 2) + + # 2. QK normalization + if attn.norm_q is not None: + query = attn.norm_q(query).to(dtype=dtype) + if attn.norm_k is not None: + key = attn.norm_k(key).to(dtype=dtype) + + # 3. Rotational positional embeddings applied to latent stream + if image_rotary_emb is not None: + from ..embeddings import apply_rotary_emb + + query[:, :, text_seq_length:, :] = apply_rotary_emb( + query[:, :, text_seq_length:, :], image_rotary_emb, use_real_unbind_dim=-2 + ) + key[:, :, text_seq_length:, :] = apply_rotary_emb( + key[:, :, text_seq_length:, :], image_rotary_emb, use_real_unbind_dim=-2 + ) + + if self.processor_state == GlmImageDecoderAttenProcessorState.ImageEditWriteKV: + self.k_cache = key if self.k_cache is None else torch.cat([self.k_cache, key], dim=2) + self.v_cache = value if self.v_cache is None else torch.cat([self.v_cache, value], dim=2) + elif self.processor_state == GlmImageDecoderAttenProcessorState.ImageEditReadKV: + key = torch.cat([self.k_cache, key], dim=2) if self.k_cache is not None else key + value = torch.cat([self.v_cache, value], dim=2) if self.v_cache is not None else value + + # 4. Attention + if attention_mask is not None: + text_attn_mask = attention_mask + assert text_attn_mask.dim() == 2, "the shape of text_attn_mask should be (batch_size, text_seq_length)" + text_attn_mask = text_attn_mask.float().to(query.device) + mix_attn_mask = torch.ones((batch_size, text_seq_length + image_seq_length), device=query.device) + mix_attn_mask[:, :text_seq_length] = text_attn_mask + mix_attn_mask = mix_attn_mask.unsqueeze(2) + attn_mask_matrix = mix_attn_mask @ mix_attn_mask.transpose(1, 2) + attention_mask = (attn_mask_matrix > 0).unsqueeze(1).to(query.dtype) + + hidden_states = F.scaled_dot_product_attention( + query, key, value, attn_mask=attention_mask, dropout_p=0.0, is_causal=False + ) + hidden_states = hidden_states.transpose(1, 2).flatten(2, 3) + hidden_states = hidden_states.type_as(query) + + # 5. Output projection + hidden_states = attn.to_out[0](hidden_states) + hidden_states = attn.to_out[1](hidden_states) + + encoder_hidden_states, hidden_states = hidden_states.split( + [text_seq_length, hidden_states.size(1) - text_seq_length], dim=1 + ) + return hidden_states, encoder_hidden_states + + +@maybe_allow_in_graph +class GlmImageDecoderTransformerBlock(nn.Module): + def __init__( + self, + dim: int = 2560, + num_attention_heads: int = 64, + attention_head_dim: int = 40, + time_embed_dim: int = 512, + ) -> None: + super().__init__() + + # 1. Attention + self.norm1 = GlmImageDecoderAdaLayerNormZero(time_embed_dim, dim) + self.attn1 = Attention( + query_dim=dim, + heads=num_attention_heads, + dim_head=attention_head_dim, + out_dim=dim, + bias=True, + qk_norm="layer_norm", + elementwise_affine=False, + eps=1e-5, + processor=GlmImageDecoderAttnProcessor(), + ) + + # 2. Feedforward + self.norm2 = nn.LayerNorm(dim, elementwise_affine=False, eps=1e-5) + self.norm2_context = nn.LayerNorm(dim, elementwise_affine=False, eps=1e-5) + self.ff = FeedForward(dim=dim, dim_out=dim, activation_fn="gelu-approximate") + + def forward( + self, + hidden_states: torch.Tensor, + glyph_hidden_states: torch.Tensor, + temb: Optional[torch.Tensor] = None, + image_rotary_emb: Optional[ + Union[Tuple[torch.Tensor, torch.Tensor], List[Tuple[torch.Tensor, torch.Tensor]]] + ] = None, + attention_mask: Optional[Dict[str, torch.Tensor]] = None, + attention_kwargs: Optional[Dict[str, Any]] = None, + ) -> Tuple[torch.Tensor, torch.Tensor]: + # 1. Timestep conditioning + ( + norm_hidden_states, + gate_msa, + shift_mlp, + scale_mlp, + gate_mlp, + norm_glyph_hidden_states, + c_gate_msa, + c_shift_mlp, + c_scale_mlp, + c_gate_mlp, + ) = self.norm1(hidden_states, glyph_hidden_states, temb) + + # 2. Attention + if attention_kwargs is None: + attention_kwargs = {} + + attn_hidden_states, attn_glyph_hidden_states = self.attn1( + hidden_states=norm_hidden_states, + encoder_hidden_states=norm_glyph_hidden_states, + image_rotary_emb=image_rotary_emb, + attention_mask=attention_mask, + **attention_kwargs, + ) + hidden_states = hidden_states + attn_hidden_states * gate_msa.unsqueeze(1) + glyph_hidden_states = glyph_hidden_states + attn_glyph_hidden_states * c_gate_msa.unsqueeze(1) + + # 3. Feedforward + norm_hidden_states = self.norm2(hidden_states) * (1 + scale_mlp.unsqueeze(1)) + shift_mlp.unsqueeze(1) + norm_glyph_hidden_states = self.norm2_context(glyph_hidden_states) * ( + 1 + c_scale_mlp.unsqueeze(1) + ) + c_shift_mlp.unsqueeze(1) + + ff_output = self.ff(norm_hidden_states) + ff_output_context = self.ff(norm_glyph_hidden_states) + hidden_states = hidden_states + ff_output * gate_mlp.unsqueeze(1) + glyph_hidden_states = glyph_hidden_states + ff_output_context * c_gate_mlp.unsqueeze(1) + + return hidden_states, glyph_hidden_states + + +class GlmImageDecoderRotaryPosEmbed(nn.Module): + def __init__(self, dim: int, patch_size: int, theta: float = 10000.0) -> None: + super().__init__() + + self.dim = dim + self.patch_size = patch_size + self.theta = theta + + def forward(self, hidden_states: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]: + batch_size, num_channels, height, width = hidden_states.shape + height, width = height // self.patch_size, width // self.patch_size + + dim_h, dim_w = self.dim // 2, self.dim // 2 + h_inv_freq = 1.0 / ( + self.theta ** (torch.arange(0, dim_h, 2, dtype=torch.float32)[: (dim_h // 2)].float() / dim_h) + ) + w_inv_freq = 1.0 / ( + self.theta ** (torch.arange(0, dim_w, 2, dtype=torch.float32)[: (dim_w // 2)].float() / dim_w) + ) + h_seq = torch.arange(height) + w_seq = torch.arange(width) + freqs_h = torch.outer(h_seq, h_inv_freq) + freqs_w = torch.outer(w_seq, w_inv_freq) + + # Create position matrices for height and width + # [height, 1, dim//4] and [1, width, dim//4] + freqs_h = freqs_h.unsqueeze(1) + freqs_w = freqs_w.unsqueeze(0) + # Broadcast freqs_h and freqs_w to [height, width, dim//4] + freqs_h = freqs_h.expand(height, width, -1) + freqs_w = freqs_w.expand(height, width, -1) + + # Concatenate along last dimension to get [height, width, dim//2] + freqs = torch.cat([freqs_h, freqs_w], dim=-1) + freqs = torch.cat([freqs, freqs], dim=-1) # [height, width, dim] + freqs = freqs.reshape(height * width, -1) + return (freqs.cos(), freqs.sin()) + + +class GlmImageDecoderAdaLayerNormContinuous(nn.Module): + """ + GlmImageDecoder-only final AdaLN: LN(x) -> Linear(cond) -> chunk -> affine. Matches Megatron: **no activation** + before the Linear on conditioning embedding. + """ + + def __init__( + self, + embedding_dim: int, + conditioning_embedding_dim: int, + elementwise_affine: bool = True, + eps: float = 1e-5, + bias: bool = True, + norm_type: str = "layer_norm", + ): + super().__init__() + self.linear = nn.Linear(conditioning_embedding_dim, embedding_dim * 2, bias=bias) + if norm_type == "layer_norm": + self.norm = LayerNorm(embedding_dim, eps, elementwise_affine, bias) + elif norm_type == "rms_norm": + self.norm = RMSNorm(embedding_dim, eps, elementwise_affine) + else: + raise ValueError(f"unknown norm_type {norm_type}") + + def forward(self, x: torch.Tensor, conditioning_embedding: torch.Tensor) -> torch.Tensor: + # *** NO SiLU here *** + emb = self.linear(conditioning_embedding.to(x.dtype)) + scale, shift = torch.chunk(emb, 2, dim=1) + x = self.norm(x) * (1 + scale)[:, None, :] + shift[:, None, :] + return x + + +class GlmImageDecoderTransformer2DModel(ModelMixin, ConfigMixin, PeftAdapterMixin, CacheMixin): + r""" + Args: + patch_size (`int`, defaults to `2`): + The size of the patches to use in the patch embedding layer. + in_channels (`int`, defaults to `16`): + The number of channels in the input. + num_layers (`int`, defaults to `30`): + The number of layers of Transformer blocks to use. + attention_head_dim (`int`, defaults to `40`): + The number of channels in each head. + num_attention_heads (`int`, defaults to `64`): + The number of heads to use for multi-head attention. + out_channels (`int`, defaults to `16`): + The number of channels in the output. + text_embed_dim (`int`, defaults to `4096`): + Input dimension of text embeddings from the text encoder. + time_embed_dim (`int`, defaults to `512`): + Output dimension of timestep embeddings. + condition_dim (`int`, defaults to `256`): + The embedding dimension of the input SDXL-style resolution conditions (original_size, target_size, + crop_coords). + pos_embed_max_size (`int`, defaults to `128`): + The maximum resolution of the positional embeddings, from which slices of shape `H x W` are taken and added + to input patched latents, where `H` and `W` are the latent height and width respectively. A value of 128 + means that the maximum supported height and width for image generation is `128 * vae_scale_factor * + patch_size => 128 * 8 * 2 => 2048`. + sample_size (`int`, defaults to `128`): + The base resolution of input latents. If height/width is not provided during generation, this value is used + to determine the resolution as `sample_size * vae_scale_factor => 128 * 8 => 1024` + """ + + _supports_gradient_checkpointing = True + _no_split_modules = [ + "GlmImageDecoderTransformerBlock", + "GlmImageDecoderImageProjector", + "GlmImageDecoderImageProjector", + ] + _skip_layerwise_casting_patterns = ["patch_embed", "norm", "proj_out"] + + @register_to_config + def __init__( + self, + patch_size: int = 2, + in_channels: int = 16, + out_channels: int = 16, + num_layers: int = 30, + attention_head_dim: int = 40, + num_attention_heads: int = 64, + text_embed_dim: int = 4096, + glyph_embed_dim: int = 1472, + time_embed_dim: int = 512, + condition_dim: int = 256, + pos_embed_max_size: int = 128, + sample_size: int = 128, + prior_vq_quantizer_codebook_size: int = 16384, + ): + super().__init__() + + # GlmImageDecoder uses 2 additional SDXL-like conditions - target_size, crop_coords + # Each of these are sincos embeddings of shape 2 * condition_dim + pooled_projection_dim = 2 * 2 * condition_dim + inner_dim = num_attention_heads * attention_head_dim + out_channels = out_channels + + # 1. RoPE + self.rope = GlmImageDecoderRotaryPosEmbed(attention_head_dim, patch_size, theta=10000.0) + + # 2. Patch & Text-timestep embedding + self.image_projector = GlmImageDecoderImageProjector(in_channels, inner_dim, patch_size) + # 这次没有，未来可能有text_projector + # self.text_projector = FeedForward(text_embed_dim, inner_dim, activation_fn="gelu") + self.glyph_projector = FeedForward(glyph_embed_dim, inner_dim, inner_dim=inner_dim, activation_fn="gelu") + + self.prior_token_embedding = nn.Embedding(prior_vq_quantizer_codebook_size, inner_dim) + self.prior_projector = FeedForward(inner_dim, inner_dim, inner_dim=inner_dim, activation_fn="linear-silu") + + self.time_condition_embed = GlmImageDecoderCombinedTimestepSizeEmbeddings( + embedding_dim=time_embed_dim, + condition_dim=condition_dim, + pooled_projection_dim=pooled_projection_dim, + timesteps_dim=time_embed_dim, + ) + + # 3. Transformer blocks + self.transformer_blocks = nn.ModuleList( + [ + GlmImageDecoderTransformerBlock(inner_dim, num_attention_heads, attention_head_dim, time_embed_dim) + for _ in range(num_layers) + ] + ) + + # 4. Output projection + self.norm_out = GlmImageDecoderAdaLayerNormContinuous(inner_dim, time_embed_dim, elementwise_affine=False) + self.proj_out = nn.Linear(inner_dim, patch_size * patch_size * out_channels, bias=True) + + self.gradient_checkpointing = False + + def forward( + self, + hidden_states: torch.Tensor, + glyph_hidden_states: torch.Tensor, + prior_token_id: torch.Tensor, + prior_token_drop: torch.Tensor, + timestep: torch.LongTensor, + original_size: torch.Tensor, + target_size: torch.Tensor, + crop_coords: torch.Tensor, + attention_kwargs: Optional[Dict[str, Any]] = None, + return_dict: bool = True, + attention_mask: Optional[torch.Tensor] = None, + image_rotary_emb: Optional[ + Union[Tuple[torch.Tensor, torch.Tensor], List[Tuple[torch.Tensor, torch.Tensor]]] + ] = None, + ) -> Union[Tuple[torch.Tensor], Transformer2DModelOutput]: + batch_size, num_channels, height, width = hidden_states.shape + + # 1. RoPE + if image_rotary_emb is None: + image_rotary_emb = self.rope(hidden_states) + + # 2. Patch & Timestep embeddings + p = self.config.patch_size + post_patch_height = height // p + post_patch_width = width // p + + hidden_states = self.image_projector(hidden_states) + glyph_hidden_states = self.glyph_projector(glyph_hidden_states) + prior_embedding = self.prior_token_embedding(prior_token_id) + prior_embedding[prior_token_drop] *= 0.0 + prior_hidden_states = self.prior_projector(prior_embedding) + + hidden_states = hidden_states + prior_hidden_states + + temb = self.time_condition_embed(timestep, target_size, crop_coords, hidden_states.dtype) + temb = F.silu(temb) + + # 3. Transformer blocks + for block in self.transformer_blocks: + if torch.is_grad_enabled() and self.gradient_checkpointing: + hidden_states, glyph_hidden_states = self._gradient_checkpointing_func( + block, + hidden_states, + glyph_hidden_states, + temb, + image_rotary_emb, + attention_mask, + attention_kwargs, + ) + else: + hidden_states, glyph_hidden_states = block( + hidden_states, + glyph_hidden_states, + temb, + image_rotary_emb, + attention_mask, + attention_kwargs, + ) + + # 4. Output norm & projection + hidden_states = self.norm_out(hidden_states, temb) + hidden_states = self.proj_out(hidden_states) + + # 5. Unpatchify + hidden_states = hidden_states.reshape(batch_size, post_patch_height, post_patch_width, -1, p, p) + output = hidden_states.permute(0, 3, 1, 4, 2, 5).flatten(4, 5).flatten(2, 3) + + if not return_dict: + return (output,) + return Transformer2DModelOutput(sample=output) + + def set_attention_processors_state(self, state: GlmImageDecoderAttenProcessorState): + for block in self.transformer_blocks: + block.attn1.processor.processor_state = state + + def clear_attention_processors_cache(self): + for block in self.transformer_blocks: + block.attn1.processor.k_cache = None + block.attn1.processor.v_cache = None + + def repeat_attention_processors_cache(self, repeats: int): + for block in self.transformer_blocks: + if block.attn1.processor.k_cache is None or block.attn1.processor.v_cache is None: + continue + block.attn1.processor.k_cache = torch.repeat_interleave(block.attn1.processor.k_cache, repeats, dim=2) + block.attn1.processor.v_cache = torch.repeat_interleave(block.attn1.processor.v_cache, repeats, dim=2) + + +if __name__ == "__main__": + + def swap_scale_shift(weight, dim): + """ + Swap the scale and shift components in the weight tensor. + + Args: + weight (torch.Tensor): The original weight tensor. + dim (int): The dimension along which to split. + + Returns: + torch.Tensor: The modified weight tensor with scale and shift swapped. + """ + shift, scale = weight.chunk(2, dim=dim) + new_weight = torch.cat([scale, shift], dim=dim) + return new_weight + + def convert_megatron_transformer_checkpoint_to_diffusers( + ckpt_path: str, + num_layers: int, + num_heads: int, + hidden_size: int, + ): + """ + Convert a Megatron Transformer checkpoint to Diffusers format. + + Args: + ckpt_path (str): Path to the Megatron Transformer checkpoint. + num_layers (int): Number of Transformer layers. + num_heads (int): Number of attention heads. + hidden_size (int): Hidden size of the Transformer. + + Returns: + dict: The converted state dictionary compatible with Diffusers. + """ + ckpt = torch.load(ckpt_path, map_location="cpu", weights_only=False) + mega = ckpt["model"] + used_keys = set() + + def get_mega(key): + used_keys.add(key) + return mega[key] + + new_state_dict = {} + + # Patch Embedding + new_state_dict["image_projector.proj.weight"] = get_mega("encoder_expand_linear.weight").reshape( + hidden_size, 64 + ) + new_state_dict["image_projector.proj.bias"] = get_mega("encoder_expand_linear.bias") + + new_state_dict["glyph_projector.net.0.proj.weight"] = get_mega("glyph_projector.linear_fc1.weight") + new_state_dict["glyph_projector.net.0.proj.bias"] = get_mega("glyph_projector.linear_fc1.bias") + new_state_dict["glyph_projector.net.2.weight"] = get_mega("glyph_projector.linear_fc2.weight") + new_state_dict["glyph_projector.net.2.bias"] = get_mega("glyph_projector.linear_fc2.bias") + + new_state_dict["prior_token_embedding.weight"] = get_mega("xomni_token_id_embedding.weight") + new_state_dict["prior_projector.net.0.proj.weight"] = get_mega("prior_condition_embedding.0.weight") + new_state_dict["prior_projector.net.0.proj.bias"] = get_mega("prior_condition_embedding.0.bias") + new_state_dict["prior_projector.net.2.weight"] = get_mega("prior_condition_embedding.2.weight") + new_state_dict["prior_projector.net.2.bias"] = get_mega("prior_condition_embedding.2.bias") + + # Time Condition Embedding + new_state_dict["time_condition_embed.timestep_embedder.linear_1.weight"] = get_mega( + "time_embedding.time_embed.0.weight" + ) + new_state_dict["time_condition_embed.timestep_embedder.linear_1.bias"] = get_mega( + "time_embedding.time_embed.0.bias" + ) + new_state_dict["time_condition_embed.timestep_embedder.linear_2.weight"] = get_mega( + "time_embedding.time_embed.2.weight" + ) + new_state_dict["time_condition_embed.timestep_embedder.linear_2.bias"] = get_mega( + "time_embedding.time_embed.2.bias" + ) + + new_state_dict["time_condition_embed.condition_embedder.linear_1.weight"] = get_mega( + "label_embedding.label_embed.0.weight" + ) + new_state_dict["time_condition_embed.condition_embedder.linear_1.bias"] = get_mega( + "label_embedding.label_embed.0.bias" + ) + new_state_dict["time_condition_embed.condition_embedder.linear_2.weight"] = get_mega( + "label_embedding.label_embed.2.weight" + ) + new_state_dict["time_condition_embed.condition_embedder.linear_2.bias"] = get_mega( + "label_embedding.label_embed.2.bias" + ) + + # Convert each Transformer layer + from tqdm import tqdm + + for i in tqdm(range(num_layers), desc="Converting layers (Megatron->Diffusers)"): + block_prefix = f"transformer_blocks.{i}." + + # AdaLayerNorm + new_state_dict[block_prefix + "norm1.linear.weight"] = get_mega(f"decoder.layers.{i}.adaln.weight") + new_state_dict[block_prefix + "norm1.linear.bias"] = get_mega(f"decoder.layers.{i}.adaln.bias") + qkv_weight = get_mega(f"decoder.layers.{i}.self_attention.linear_qkv.weight") + qkv_bias = get_mega(f"decoder.layers.{i}.self_attention.linear_qkv.bias") + + # Reshape to match SAT logic + qkv_weight = qkv_weight.view(num_heads, 3, hidden_size // num_heads, hidden_size) + qkv_weight = qkv_weight.permute(1, 0, 2, 3).reshape(3 * hidden_size, hidden_size) + + qkv_bias = qkv_bias.view(num_heads, 3, hidden_size // num_heads) + qkv_bias = qkv_bias.permute(1, 0, 2).reshape(3 * hidden_size) + + # Assign to Diffusers keys + q, k, v = torch.chunk(qkv_weight, 3, dim=0) + qb, kb, vb = torch.chunk(qkv_bias, 3, dim=0) + + new_state_dict[block_prefix + "attn1.to_q.weight"] = q + new_state_dict[block_prefix + "attn1.to_q.bias"] = qb + new_state_dict[block_prefix + "attn1.to_k.weight"] = k + new_state_dict[block_prefix + "attn1.to_k.bias"] = kb + new_state_dict[block_prefix + "attn1.to_v.weight"] = v + new_state_dict[block_prefix + "attn1.to_v.bias"] = vb + + # Attention Output + new_state_dict[block_prefix + "attn1.to_out.0.weight"] = get_mega( + f"decoder.layers.{i}.self_attention.linear_proj.weight" + ) + new_state_dict[block_prefix + "attn1.to_out.0.bias"] = get_mega( + f"decoder.layers.{i}.self_attention.linear_proj.bias" + ) + + # MLP + new_state_dict[block_prefix + "ff.net.0.proj.weight"] = get_mega( + f"decoder.layers.{i}.mlp.linear_fc1.weight" + ) + new_state_dict[block_prefix + "ff.net.0.proj.bias"] = get_mega(f"decoder.layers.{i}.mlp.linear_fc1.bias") + new_state_dict[block_prefix + "ff.net.2.weight"] = get_mega(f"decoder.layers.{i}.mlp.linear_fc2.weight") + new_state_dict[block_prefix + "ff.net.2.bias"] = get_mega(f"decoder.layers.{i}.mlp.linear_fc2.bias") + + # Final Layers + new_state_dict["norm_out.linear.weight"] = swap_scale_shift(get_mega("adaln_final.weight"), dim=0) + new_state_dict["norm_out.linear.bias"] = swap_scale_shift(get_mega("adaln_final.bias"), dim=0) + new_state_dict["proj_out.weight"] = get_mega("output_projector.weight") + new_state_dict["proj_out.bias"] = get_mega("output_projector.bias") + + # Check for unused keys + all_keys = set(mega.keys()) + unused_keys = all_keys - used_keys + if unused_keys: + print(f"\n[WARNING] The following {len(unused_keys)} keys in mega were NOT used:") + for key in sorted(unused_keys): + print(f" - {key}") + raise ValueError( + f"Found {len(unused_keys)} unused keys in Megatron checkpoint. Please update the conversion script to handle these keys." + ) + else: + print(f"\n[INFO] All {len(all_keys)} keys in mega were successfully used.") + + return new_state_dict + + transformer = GlmImageDecoderTransformer2DModel( + patch_size=2, + in_channels=16, + num_layers=30, + attention_head_dim=128, + num_attention_heads=32, + out_channels=16, + text_embed_dim=4096, + time_embed_dim=512, + glyph_embed_dim=1472, + condition_dim=256, + pos_embed_max_size=128, + ).to(torch.bfloat16) + converted_transformer_state_dict = convert_megatron_transformer_checkpoint_to_diffusers( + ckpt_path="/workspace/ckpt/tjy/Glm-train-dev/examples/cogview/ckpts/merge/1+6_0.5+0.5/iter_0000000/mp_rank_00/model_optim_rng.pt", + num_layers=30, + num_heads=32, + hidden_size=4096, + ) + transformer.load_state_dict(converted_transformer_state_dict) + transformer.cuda() + + latent = torch.load("/workspace/ckpt/tjy/glm-train-dev/examples/cogview/latent.pt").to(torch.bfloat16) + latent = rearrange(latent, "(b h w) (c p q) -> b c (h p) (w q)", b=8, h=72, w=54, p=2, q=2) + glyph_hidden_states = torch.load( + "/workspace/ckpt/tjy/glm-train-dev/examples/cogview/glyph_condition_embedding.pt" + ).to(torch.bfloat16) + glyph_hidden_states = rearrange(glyph_hidden_states, "(b n) c -> b n c", b=8, n=2) + prior_token_id = torch.load("/workspace/ckpt/tjy/glm-train-dev/examples/cogview/xomni_token_id.pt") + prior_token_drop = torch.load("/workspace/ckpt/tjy/glm-train-dev/examples/cogview/xomni_drop.pt") + prior_token_id = rearrange(prior_token_id, "(b n) -> b n", b=8) + prior_token_drop = rearrange(prior_token_drop, "(b n)-> b n", b=8) + + with torch.no_grad(): + output = transformer( + hidden_states=latent, + glyph_hidden_states=glyph_hidden_states, + prior_token_id=prior_token_id, + prior_token_drop=prior_token_drop, + timestep=torch.tensor([999.0] * 8).cuda(), + original_size=torch.tensor([[144, 108]] * 8).cuda(), + target_size=torch.tensor([[144, 108]] * 8).cuda(), + crop_coords=torch.tensor([[0, 0]] * 8).cuda(), + ) diff --git a/src/diffusers/pipelines/glm_image/__init__.py b/src/diffusers/pipelines/glm_image/__init__.py new file mode 100644 index 0000000000..24c20c4ef5 --- /dev/null +++ b/src/diffusers/pipelines/glm_image/__init__.py @@ -0,0 +1,47 @@ +from typing import TYPE_CHECKING + +from ...utils import ( + DIFFUSERS_SLOW_IMPORT, + OptionalDependencyNotAvailable, + _LazyModule, + get_objects_from_module, + is_torch_available, + is_transformers_available, +) + + +_dummy_objects = {} +_additional_imports = {} +_import_structure = {"pipeline_output": ["GlmImageDecoderPipelineOutput"]} + +try: + if not (is_transformers_available() and is_torch_available()): + raise OptionalDependencyNotAvailable() +except OptionalDependencyNotAvailable: + from ...utils import dummy_torch_and_transformers_objects # noqa F403 + + _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects)) +else: + _import_structure["pipeline_glm_image"] = ["GlmImageDecoderPipeline"] +if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT: + try: + if not (is_transformers_available() and is_torch_available()): + raise OptionalDependencyNotAvailable() + except OptionalDependencyNotAvailable: + from ...utils.dummy_torch_and_transformers_objects import * # noqa F403 + else: + from .pipeline_glm_image import GlmImageDecoderPipeline +else: + import sys + + sys.modules[__name__] = _LazyModule( + __name__, + globals()["__file__"], + _import_structure, + module_spec=__spec__, + ) + + for name, value in _dummy_objects.items(): + setattr(sys.modules[__name__], name, value) + for name, value in _additional_imports.items(): + setattr(sys.modules[__name__], name, value) diff --git a/src/diffusers/pipelines/glm_image/pipeline_glm_image.py b/src/diffusers/pipelines/glm_image/pipeline_glm_image.py new file mode 100644 index 0000000000..825952be4b --- /dev/null +++ b/src/diffusers/pipelines/glm_image/pipeline_glm_image.py @@ -0,0 +1,786 @@ +# Copyright 2025 The CogVideoX team, Tsinghua University & ZhipuAI and The HuggingFace Team. +# All rights reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import inspect +import re +from typing import Any, Callable, Dict, List, Optional, Tuple, Union + +import numpy as np +import PIL +import torch +from transformers import AutoTokenizer, T5EncoderModel + +from ...callbacks import MultiPipelineCallbacks, PipelineCallback +from ...image_processor import VaeImageProcessor +from ...loaders import CogView4LoraLoaderMixin +from ...models import AutoencoderKL, GlmImageDecoderTransformer2DModel +from ...models.transformers.transformer_glm_image import GlmImageDecoderAttenProcessorState +from ...pipelines.pipeline_utils import DiffusionPipeline +from ...schedulers import FlowMatchEulerDiscreteScheduler +from ...utils import is_torch_xla_available, logging, replace_example_docstring +from ...utils.torch_utils import randn_tensor +from .pipeline_output import GlmImageDecoderPipelineOutput + + +if is_torch_xla_available(): + import torch_xla.core.xla_model as xm + + XLA_AVAILABLE = True +else: + XLA_AVAILABLE = False + +logger = logging.get_logger(__name__) # pylint: disable=invalid-name + +EXAMPLE_DOC_STRING = """ + Examples: + ```python + >>> import torch + >>> from diffusers import GlmImageDecoderPipeline + + >>> pipe = GlmImageDecoderPipeline.from_pretrained("THUDM/CogView4-6B", torch_dtype=torch.bfloat16) + >>> pipe.to("cuda") + + >>> prompt = "A photo of an astronaut riding a horse on mars" + >>> image = pipe(prompt).images[0] + >>> image.save("output.png") + ``` +""" + + +def calculate_shift( + image_seq_len, + base_seq_len: int = 256, + base_shift: float = 0.25, + max_shift: float = 0.75, +) -> float: + m = (image_seq_len / base_seq_len) ** 0.5 + mu = m * max_shift + base_shift + return mu + + +def retrieve_timesteps( + scheduler, + num_inference_steps: Optional[int] = None, + device: Optional[Union[str, torch.device]] = None, + timesteps: Optional[List[int]] = None, + sigmas: Optional[List[float]] = None, + **kwargs, +): + r""" + Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles + custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. + + Args: + scheduler (`SchedulerMixin`): + The scheduler to get timesteps from. + num_inference_steps (`int`): + The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` + must be `None`. + device (`str` or `torch.device`, *optional*): + The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. + timesteps (`List[int]`, *optional*): + Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, + `num_inference_steps` and `sigmas` must be `None`. + sigmas (`List[float]`, *optional*): + Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, + `num_inference_steps` and `timesteps` must be `None`. + + Returns: + `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the + second element is the number of inference steps. + """ + accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) + accepts_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys()) + + if timesteps is not None and sigmas is not None: + if not accepts_timesteps and not accepts_sigmas: + raise ValueError( + f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" + f" timestep or sigma schedules. Please check whether you are using the correct scheduler." + ) + scheduler.set_timesteps(timesteps=timesteps, sigmas=sigmas, device=device, **kwargs) + timesteps = scheduler.timesteps + num_inference_steps = len(timesteps) + elif timesteps is not None and sigmas is None: + if not accepts_timesteps: + raise ValueError( + f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" + f" timestep schedules. Please check whether you are using the correct scheduler." + ) + scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs) + timesteps = scheduler.timesteps + num_inference_steps = len(timesteps) + elif timesteps is None and sigmas is not None: + if not accepts_sigmas: + raise ValueError( + f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" + f" sigmas schedules. Please check whether you are using the correct scheduler." + ) + scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs) + timesteps = scheduler.timesteps + num_inference_steps = len(timesteps) + else: + scheduler.set_timesteps(num_inference_steps, device=device, **kwargs) + timesteps = scheduler.timesteps + return timesteps, num_inference_steps + + +# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents +def retrieve_latents( + encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample" +): + if hasattr(encoder_output, "latent_dist") and sample_mode == "sample": + return encoder_output.latent_dist.sample(generator) + elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax": + return encoder_output.latent_dist.mode() + elif hasattr(encoder_output, "latents"): + return encoder_output.latents + else: + raise AttributeError("Could not access latents of provided encoder_output") + + +class GlmImageDecoderPipeline(DiffusionPipeline, CogView4LoraLoaderMixin): + r""" + Pipeline for text-to-image generation using CogView4. + + This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods the + library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.) + + Args: + vae ([`AutoencoderKL`]): + Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations. + text_encoder ([`GlmModel`]): + Frozen text-encoder. CogView4 uses [Glm-4-9b-hf](https://huggingface.co/THUDM/Glm-4-9b-hf). + tokenizer (`PreTrainedTokenizer`): + Tokenizer of class + [PreTrainedTokenizer](https://huggingface.co/docs/transformers/main/en/main_classes/tokenizer#transformers.PreTrainedTokenizer). + transformer ([`CogView4Transformer2DModel`]): + A text conditioned `CogView4Transformer2DModel` to denoise the encoded image latents. + scheduler ([`SchedulerMixin`]): + A scheduler to be used in combination with `transformer` to denoise the encoded image latents. + """ + + _optional_components = [] + model_cpu_offload_seq = "transformer->vae" + _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"] + + def __init__( + self, + tokenizer: AutoTokenizer, + glyph_encoder: T5EncoderModel, + vae: AutoencoderKL, + transformer: GlmImageDecoderTransformer2DModel, + scheduler: FlowMatchEulerDiscreteScheduler, + ): + super().__init__() + + self.register_modules( + tokenizer=tokenizer, glyph_encoder=glyph_encoder, vae=vae, transformer=transformer, scheduler=scheduler + ) + self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8 + self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor) + + self.default_sample_size = ( + self.transformer.config.sample_size + if hasattr(self, "transformer") + and self.transformer is not None + and hasattr(self.transformer.config, "sample_size") + else 128 + ) + + def get_glyph_texts( + self, + prompt, + ): + prompt = prompt[0] if isinstance(prompt, list) else prompt + ocr_texts = ( + re.findall(r"'([^']*)'", prompt) + + re.findall(r"“([^“”]*)”", prompt) + + re.findall(r'"([^"]*)"', prompt) + + re.findall(r"「([^「」]*)」", prompt) + ) + return ocr_texts + + def _get_glyph_embeds( + self, + prompt: Union[str, List[str]] = None, + max_sequence_length: int = 2048, + device: Optional[torch.device] = None, + dtype: Optional[torch.dtype] = None, + ): + device = device or self._execution_device + dtype = dtype or self.glyph_encoder.dtype + + glyph_texts = self.get_glyph_texts(prompt) + input_ids = self.tokenizer( + glyph_texts if len(glyph_texts) > 0 else [""], + max_length=max_sequence_length, + truncation=True, + ).input_ids + input_ids = [ + [self.tokenizer.pad_token_id] * ((len(input_ids) + 1) % 2) + input_ids_ for input_ids_ in input_ids + ] + max_length = max(len(input_ids_) for input_ids_ in input_ids) + attention_mask = torch.tensor( + [[1] * len(input_ids_) + [0] * (max_length - len(input_ids_)) for input_ids_ in input_ids], device=device + ) + input_ids = torch.tensor( + [input_ids_ + [self.tokenizer.pad_token_id] * (max_length - len(input_ids_)) for input_ids_ in input_ids], + device=device, + ) + outputs = self.glyph_encoder(input_ids, attention_mask=attention_mask) + glyph_embeds = outputs.last_hidden_state[attention_mask.bool()].unsqueeze(0) + + return glyph_embeds.to(device=device, dtype=dtype) + + def encode_prompt( + self, + prompt: Union[str, List[str]], + negative_prompt: Optional[Union[str, List[str]]] = None, + do_classifier_free_guidance: bool = True, + num_images_per_prompt: int = 1, + prompt_embeds: Optional[torch.Tensor] = None, + negative_prompt_embeds: Optional[torch.Tensor] = None, + device: Optional[torch.device] = None, + dtype: Optional[torch.dtype] = None, + max_sequence_length: int = 2048, + ): + r""" + Encodes the prompt into text encoder hidden states. + + Args: + prompt (`str` or `List[str]`, *optional*): + prompt to be encoded + negative_prompt (`str` or `List[str]`, *optional*): + The prompt or prompts not to guide the image generation. If not defined, one has to pass + `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is + less than `1`). + do_classifier_free_guidance (`bool`, *optional*, defaults to `True`): + Whether to use classifier free guidance or not. + num_images_per_prompt (`int`, *optional*, defaults to 1): + Number of images that should be generated per prompt. torch device to place the resulting embeddings on + prompt_embeds (`torch.Tensor`, *optional*): + Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not + provided, text embeddings will be generated from `prompt` input argument. + negative_prompt_embeds (`torch.Tensor`, *optional*): + Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt + weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input + argument. + device: (`torch.device`, *optional*): + torch device + dtype: (`torch.dtype`, *optional*): + torch dtype + max_sequence_length (`int`, defaults to `2048`): + Maximum sequence length in encoded prompt. Can be set to other values but may lead to poorer results. + """ + device = device or self._execution_device + + prompt = [prompt] if isinstance(prompt, str) else prompt + if prompt is not None: + batch_size = len(prompt) + else: + batch_size = prompt_embeds.shape[0] + + if prompt_embeds is None: + prompt_embeds = self._get_glyph_embeds(prompt, max_sequence_length, device, dtype) + + seq_len = prompt_embeds.size(1) + prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1) + prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1) + + if do_classifier_free_guidance and negative_prompt_embeds is None: + negative_prompt = negative_prompt or "" + negative_prompt = batch_size * [negative_prompt] if isinstance(negative_prompt, str) else negative_prompt + + if prompt is not None and type(prompt) is not type(negative_prompt): + raise TypeError( + f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !=" + f" {type(prompt)}." + ) + elif batch_size != len(negative_prompt): + raise ValueError( + f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:" + f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches" + " the batch size of `prompt`." + ) + + negative_prompt_embeds = self._get_glyph_embeds(negative_prompt, max_sequence_length, device, dtype) + + seq_len = negative_prompt_embeds.size(1) + negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1) + negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1) + + return prompt_embeds, negative_prompt_embeds + + def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None): + if latents is not None: + return latents.to(device) + + shape = ( + batch_size, + num_channels_latents, + int(height) // self.vae_scale_factor, + int(width) // self.vae_scale_factor, + ) + if isinstance(generator, list) and len(generator) != batch_size: + raise ValueError( + f"You have passed a list of generators of length {len(generator)}, but requested an effective batch" + f" size of {batch_size}. Make sure the batch size matches the length of the generators." + ) + latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype) + return latents + + def check_inputs( + self, + prompt, + height, + width, + negative_prompt, + callback_on_step_end_tensor_inputs, + prompt_embeds=None, + negative_prompt_embeds=None, + ): + if ( + height is not None + and height % (self.vae_scale_factor * self.transformer.config.patch_size) != 0 + or width is not None + and width % (self.transformer.config.patch_size) != 0 + ): + logger.warning( + f"`height` and `width` have to be divisible by {self.vae_scale_factor * 2} but are {height} and {width}. Dimensions will be resized accordingly" + ) + + if callback_on_step_end_tensor_inputs is not None and not all( + k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs + ): + raise ValueError( + f"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}" + ) + + if prompt is not None and prompt_embeds is not None: + raise ValueError( + f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to" + " only forward one of the two." + ) + elif prompt is None and prompt_embeds is None: + raise ValueError( + "Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined." + ) + elif prompt is not None and (not isinstance(prompt, str) and not isinstance(prompt, list)): + raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}") + + @property + def guidance_scale(self): + return self._guidance_scale + + # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2) + # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1` + # corresponds to doing no classifier free guidance. + @property + def do_classifier_free_guidance(self): + return self._guidance_scale > 1 + + @property + def num_timesteps(self): + return self._num_timesteps + + @property + def attention_kwargs(self): + return self._attention_kwargs + + @property + def current_timestep(self): + return self._current_timestep + + @property + def interrupt(self): + return self._interrupt + + @torch.no_grad() + @replace_example_docstring(EXAMPLE_DOC_STRING) + def __call__( + self, + prior_token_id: Optional[torch.LongTensor] = None, + prompt: Optional[Union[str, List[str]]] = None, + negative_prompt: Optional[Union[str, List[str]]] = None, + condition_images: Optional[ + Union[ + torch.Tensor, PIL.Image.Image, np.ndarray, List[torch.Tensor], List[PIL.Image.Image], List[np.ndarray] + ] + ] = None, + condition_images_prior_token_id: Optional[Union[torch.LongTensor, List[torch.LongTensor]]] = None, + height: Optional[int] = None, + width: Optional[int] = None, + num_inference_steps: int = 50, + timesteps: Optional[List[int]] = None, + sigmas: Optional[List[float]] = None, + guidance_scale: float = 1.5, + num_images_per_prompt: int = 1, + generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None, + latents: Optional[torch.FloatTensor] = None, + prompt_embeds: Optional[torch.FloatTensor] = None, + negative_prompt_embeds: Optional[torch.FloatTensor] = None, + original_size: Optional[Tuple[int, int]] = None, + crops_coords_top_left: Tuple[int, int] = (0, 0), + output_type: str = "pil", + return_dict: bool = True, + attention_kwargs: Optional[Dict[str, Any]] = None, + callback_on_step_end: Optional[ + Union[Callable[[int, int, Dict], None], PipelineCallback, MultiPipelineCallbacks] + ] = None, + callback_on_step_end_tensor_inputs: List[str] = ["latents"], + max_sequence_length: int = 2048, + ) -> Union[GlmImageDecoderPipelineOutput, Tuple]: + """ + Function invoked when calling the pipeline for generation. + + Args: + image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, or `List[np.ndarray]`): + `Image`, numpy array or tensor representing an image batch to be used as the starting point. For both + numpy array and pytorch tensor, the expected value range is between `[0, 1]` If it's a tensor or a list + or tensors, the expected shape should be `(B, C, H, W)` or `(C, H, W)`. If it is a numpy array or a + list of arrays, the expected shape should be `(B, H, W, C)` or `(H, W, C)` It can also accept image + latents as `image`, but if passing latents directly it is not encoded again. + prompt (`str` or `List[str]`, *optional*): + The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`. + negative_prompt (`str` or `List[str]`, *optional*): + The prompt or prompts not to guide the image generation. If not defined, one has to pass + `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is + less than `1`). + height (`int`, *optional*, defaults to self.transformer.config.sample_size * self.vae_scale_factor): + The height in pixels of the generated image. If not provided, it is set to 2048. + width (`int`, *optional*, defaults to self.transformer.config.sample_size * self.vae_scale_factor): + The width in pixels of the generated image. If not provided it is set to 2048. + num_inference_steps (`int`, *optional*, defaults to `50`): + The number of denoising steps. More denoising steps usually lead to a higher quality image at the + expense of slower inference. + timesteps (`List[int]`, *optional*): + Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument + in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is + passed will be used. Must be in descending order. + sigmas (`List[float]`, *optional*): + Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in + their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed + will be used. + guidance_scale (`float`, *optional*, defaults to `5.0`): + Guidance scale as defined in [Classifier-Free Diffusion + Guidance](https://huggingface.co/papers/2207.12598). `guidance_scale` is defined as `w` of equation 2. + of [Imagen Paper](https://huggingface.co/papers/2205.11487). Guidance scale is enabled by setting + `guidance_scale > 1`. Higher guidance scale encourages to generate images that are closely linked to + the text `prompt`, usually at the expense of lower image quality. + num_images_per_prompt (`int`, *optional*, defaults to `1`): + The number of images to generate per prompt. + generator (`torch.Generator` or `List[torch.Generator]`, *optional*): + One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html) + to make generation deterministic. + latents (`torch.FloatTensor`, *optional*): + Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image + generation. Can be used to tweak the same generation with different prompts. If not provided, a latents + tensor will be generated by sampling using the supplied random `generator`. + prompt_embeds (`torch.FloatTensor`, *optional*): + Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not + provided, text embeddings will be generated from `prompt` input argument. + negative_prompt_embeds (`torch.FloatTensor`, *optional*): + Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt + weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input + argument. + original_size (`Tuple[int]`, *optional*, defaults to (2048, 2048)): + If `original_size` is not the same as `target_size` the image will appear to be down- or upsampled. + `original_size` defaults to `(height, width)` if not specified. Part of SDXL's micro-conditioning as + explained in section 2.2 of + [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952). + crops_coords_top_left (`Tuple[int]`, *optional*, defaults to (0, 0)): + `crops_coords_top_left` can be used to generate an image that appears to be "cropped" from the position + `crops_coords_top_left` downwards. Favorable, well-centered images are usually achieved by setting + `crops_coords_top_left` to (0, 0). Part of SDXL's micro-conditioning as explained in section 2.2 of + [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952). + output_type (`str`, *optional*, defaults to `"pil"`): + The output format of the generate image. Choose between + [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`. + return_dict (`bool`, *optional*, defaults to `True`): + Whether or not to return a [`~pipelines.stable_diffusion_xl.StableDiffusionXLPipelineOutput`] instead + of a plain tuple. + attention_kwargs (`dict`, *optional*): + A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under + `self.processor` in + [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py). + callback_on_step_end (`Callable`, *optional*): + A function that calls at the end of each denoising steps during the inference. The function is called + with the following arguments: `callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int, + callback_kwargs: Dict)`. `callback_kwargs` will include a list of all tensors as specified by + `callback_on_step_end_tensor_inputs`. + callback_on_step_end_tensor_inputs (`List`, *optional*): + The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list + will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the + `._callback_tensor_inputs` attribute of your pipeline class. + max_sequence_length (`int`, defaults to `224`): + Maximum sequence length in encoded prompt. Can be set to other values but may lead to poorer results. + + Examples: + + Returns: + [`~pipelines.cogview4.pipeline_CogView4.CogView4PipelineOutput`] or `tuple`: + [`~pipelines.cogview4.pipeline_CogView4.CogView4PipelineOutput`] if `return_dict` is True, otherwise a + `tuple`. When returning a tuple, the first element is a list with the generated images. + """ + + if isinstance(callback_on_step_end, (PipelineCallback, MultiPipelineCallbacks)): + callback_on_step_end_tensor_inputs = callback_on_step_end.tensor_inputs + + # 1. Check inputs. Raise error if not correct + self.check_inputs( + prompt, + height, + width, + negative_prompt, + callback_on_step_end_tensor_inputs, + prompt_embeds, + negative_prompt_embeds, + ) + self._guidance_scale = guidance_scale + self._attention_kwargs = attention_kwargs + self._current_timestep = None + self._interrupt = False + + # 2. Default call parameters + if prompt is not None and isinstance(prompt, str): + batch_size = 1 + elif prompt is not None and isinstance(prompt, list): + batch_size = len(prompt) + else: + batch_size = prompt_embeds.shape[0] + assert batch_size == 1, "batch_size must be 1" + + device = self._execution_device + + # 3. Encode input prompt + prompt_embeds, negative_prompt_embeds = self.encode_prompt( + prompt, + negative_prompt, + self.do_classifier_free_guidance, + num_images_per_prompt=num_images_per_prompt, + prompt_embeds=prompt_embeds, + negative_prompt_embeds=negative_prompt_embeds, + max_sequence_length=max_sequence_length, + device=device, + dtype=self.dtype, + ) + + # 4. process images + if condition_images is not None and not isinstance(condition_images, list): + condition_images = [condition_images] + condition_images_prior_token_id = [condition_images_prior_token_id] + assert condition_images is None or (len(condition_images) == len(condition_images_prior_token_id)), ( + "image and image_prior_token_id must be the same length" + ) + + if condition_images is not None: + preprocessed_condition_images = [] + for img in condition_images: + image_height, image_width = img.size[::-1] if isinstance(img, PIL.Image.Image) else img.shape[:2] + multiple_of = self.vae_scale_factor * self.transformer.config.patch_size + image_height = (image_height // multiple_of) * multiple_of + image_width = (image_width // multiple_of) * multiple_of + img = self.image_processor.preprocess(img, height=image_height, width=image_width) + preprocessed_condition_images.append(img) + height = height or image_height + width = width or image_width + condition_images = preprocessed_condition_images + height = height or self.default_sample_size * self.vae_scale_factor + width = width or self.default_sample_size * self.vae_scale_factor + + # 5. Prepare latents and (optional) condition_images kv cache + latent_channels = self.transformer.config.in_channels + latents = self.prepare_latents( + batch_size=batch_size * num_images_per_prompt, + num_channels_latents=latent_channels, + height=height, + width=width, + dtype=torch.float32, + device=device, + generator=generator, + latents=latents, + ) + + if condition_images is not None: + self.transformer.set_attention_processors_state(GlmImageDecoderAttenProcessorState.ImageEditWriteKV) + latents_mean = ( + torch.tensor(self.vae.config.latents_mean) + .view(1, self.vae.config.latent_channels, 1, 1) + .to(self.vae.device, self.vae.dtype) + ) + latents_std = ( + torch.tensor(self.vae.config.latents_std) + .view(1, self.vae.config.latent_channels, 1, 1) + .to(self.vae.device, self.vae.dtype) + ) + empty_glyph_hiddens = torch.zeros_like(prompt_embeds)[:1, :0, ...] + for condition_image, condition_image_prior_token_id in zip( + condition_images, condition_images_prior_token_id + ): + condition_image = condition_image.to(device=device, dtype=self.vae.dtype) + condition_latent = retrieve_latents( + self.vae.encode(condition_image), generator=generator, sample_mode="argmax" + ) + condition_latent = (condition_latent - latents_mean) / latents_std + _ = self.transformer( + hidden_states=condition_latent, + glyph_hidden_states=empty_glyph_hiddens, + prior_token_id=condition_image_prior_token_id, + prior_token_drop=torch.full_like(condition_image_prior_token_id, False, dtype=torch.bool), + timestep=torch.zeros((1,), device=device), + original_size=torch.tensor([condition_image.shape[-2:]], device=device), + target_size=torch.tensor([condition_image.shape[-2:]], device=device), + crop_coords=torch.zeros((1, 2), device=device), + attention_kwargs=attention_kwargs, + ) + + # 6. Prepare additional timestep conditions + original_size = original_size or (height, width) + target_size = (height, width) + original_size = torch.tensor([original_size], dtype=prompt_embeds.dtype, device=device) + target_size = torch.tensor([target_size], dtype=prompt_embeds.dtype, device=device) + crops_coords_top_left = torch.tensor([crops_coords_top_left], dtype=prompt_embeds.dtype, device=device) + + original_size = original_size.repeat(batch_size * num_images_per_prompt, 1) + target_size = target_size.repeat(batch_size * num_images_per_prompt, 1) + crops_coords_top_left = crops_coords_top_left.repeat(batch_size * num_images_per_prompt, 1) + + # Prepare timesteps + image_seq_len = ((height // self.vae_scale_factor) * (width // self.vae_scale_factor)) // ( + self.transformer.config.patch_size**2 + ) + timesteps = ( + np.linspace(self.scheduler.config.num_train_timesteps, 1.0, num_inference_steps + 1)[:-1] + if timesteps is None + else np.array(timesteps) + ) + timesteps = timesteps.astype(np.int64).astype(np.float32) + sigmas = timesteps / self.scheduler.config.num_train_timesteps if sigmas is None else sigmas + mu = calculate_shift( + image_seq_len, + self.scheduler.config.get("base_image_seq_len", 256), + self.scheduler.config.get("base_shift", 0.25), + self.scheduler.config.get("max_shift", 0.75), + ) + timesteps, num_inference_steps = retrieve_timesteps( + self.scheduler, num_inference_steps, device, timesteps, sigmas, mu=mu + ) + self._num_timesteps = len(timesteps) + + # 7. Denoising loop + transformer_dtype = self.transformer.dtype + num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0) + + prior_token_drop_cond = torch.full_like(prior_token_id, False, dtype=torch.bool) + prior_token_drop_uncond = torch.full_like(prior_token_id, True, dtype=torch.bool) + with self.progress_bar(total=num_inference_steps) as progress_bar: + for i, t in enumerate(timesteps): + if self.interrupt: + continue + + self._current_timestep = t + latent_model_input = latents.to(transformer_dtype) + + # broadcast to batch dimension in a way that's compatible with ONNX/Core ML + timestep = t.expand(latents.shape[0]) - 1 + + if condition_images is not None: + self.transformer.set_attention_processors_state(GlmImageDecoderAttenProcessorState.ImageEditReadKV) + + noise_pred_cond = self.transformer( + hidden_states=latent_model_input, + glyph_hidden_states=prompt_embeds, + prior_token_id=prior_token_id, + prior_token_drop=prior_token_drop_cond, + timestep=timestep, + original_size=original_size, + target_size=target_size, + crop_coords=crops_coords_top_left, + attention_kwargs=attention_kwargs, + return_dict=False, + )[0].float() + + # perform guidance + if self.do_classifier_free_guidance: + if condition_images is not None: + self.transformer.set_attention_processors_state( + GlmImageDecoderAttenProcessorState.ImageEditDontReadKV + ) + noise_pred_uncond = self.transformer( + hidden_states=latent_model_input, + glyph_hidden_states=negative_prompt_embeds, + prior_token_id=prior_token_id, + prior_token_drop=prior_token_drop_uncond, + timestep=timestep, + original_size=original_size, + target_size=target_size, + crop_coords=crops_coords_top_left, + attention_kwargs=attention_kwargs, + return_dict=False, + )[0].float() + + noise_pred = noise_pred_uncond + self.guidance_scale * (noise_pred_cond - noise_pred_uncond) + else: + noise_pred = noise_pred_cond + + latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0] + + # call the callback, if provided + if callback_on_step_end is not None: + callback_kwargs = {} + for k in callback_on_step_end_tensor_inputs: + callback_kwargs[k] = locals()[k] + callback_outputs = callback_on_step_end(self, i, self.scheduler.sigmas[i], callback_kwargs) + latents = callback_outputs.pop("latents", latents) + prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds) + negative_prompt_embeds = callback_outputs.pop("negative_prompt_embeds", negative_prompt_embeds) + + if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0): + progress_bar.update() + + if XLA_AVAILABLE: + xm.mark_step() + + self._current_timestep = None + self.transformer.set_attention_processors_state(GlmImageDecoderAttenProcessorState.ImageGen) + self.transformer.clear_attention_processors_cache() + + if not output_type == "latent": + latents = latents.to(self.vae.dtype) + latents_mean = ( + torch.tensor(self.vae.config.latents_mean) + .view(1, self.vae.config.latent_channels, 1, 1) + .to(latents.device, latents.dtype) + ) + latents_std = ( + torch.tensor(self.vae.config.latents_std) + .view(1, self.vae.config.latent_channels, 1, 1) + .to(latents.device, latents.dtype) + ) + latents = latents * latents_std + latents_mean + condition_images = self.vae.decode(latents, return_dict=False, generator=generator)[0] + else: + condition_images = latents + + condition_images = self.image_processor.postprocess(condition_images, output_type=output_type) + + # Offload all models + self.maybe_free_model_hooks() + + if not return_dict: + return (condition_images,) + + return GlmImageDecoderPipelineOutput(images=condition_images) diff --git a/src/diffusers/pipelines/glm_image/pipeline_output.py b/src/diffusers/pipelines/glm_image/pipeline_output.py new file mode 100644 index 0000000000..a506e527cb --- /dev/null +++ b/src/diffusers/pipelines/glm_image/pipeline_output.py @@ -0,0 +1,21 @@ +from dataclasses import dataclass +from typing import List, Union + +import numpy as np +import PIL.Image + +from ...utils import BaseOutput + + +@dataclass +class GlmImageDecoderPipelineOutput(BaseOutput): + """ + Output class for CogView3 pipelines. + + Args: + images (`List[PIL.Image.Image]` or `np.ndarray`) + List of denoised PIL images of length `batch_size` or numpy array of shape `(batch_size, height, width, + num_channels)`. PIL images or numpy array present the denoised images of the diffusion pipeline. + """ + + images: Union[List[PIL.Image.Image], np.ndarray]