[BugFix] Fix load_weights error when loading HunyuanImage3.0 (#1598)

Signed-off-by: Semmer2 <semmer@live.cn>
This commit is contained in:
Ting FU authored and GitHub committed 2026-03-02 08:40:48 +00:00
1 parent e37a89ffcb
commit 1ca198eaff
2 files changed
+13 -43

No files matched your search

@@ -1,11 +1,9 @@
# SPDX-License-Identifier: Apache-2.0 # SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import glob
import inspect import inspect
import logging import logging
import math import math
import os
from collections.abc import Callable, Iterable from collections.abc import Callable, Iterable
from dataclasses import dataclass from dataclasses import dataclass
from typing import Any, cast from typing import Any, cast
@@ -22,7 +20,6 @@ from diffusers.utils.outputs import BaseOutput
from diffusers.utils.torch_utils import randn_tensor from diffusers.utils.torch_utils import randn_tensor
from einops import rearrange from einops import rearrange
from PIL import Image from PIL import Image
from safetensors.torch import load_file
from torch import nn from torch import nn
from torchvision import transforms from torchvision import transforms
from transformers import PretrainedConfig, Siglip2ImageProcessorFast from transformers import PretrainedConfig, Siglip2ImageProcessorFast
@@ -357,14 +354,6 @@ def build_batch_2d_rope(
return stacked_cos, stacked_sin return stacked_cos, stacked_sin
def get_full_state_dict(model_path):
files = glob.glob(os.path.join(model_path, "*.safetensors"))
full_sd = {}
for f in files:
full_sd.update(load_file(f))
return full_sd
def rotate_half(x): def rotate_half(x):
"""Rotates half the hidden dims of the input.""" """Rotates half the hidden dims of the input."""
x1 = x[..., : x.shape[-1] // 2] x1 = x[..., : x.shape[-1] // 2]
@@ -1744,7 +1733,6 @@ class HunyuanImage3Model(nn.Module):
lora_vocab = (lora_config.lora_extra_vocab_size * (lora_config.max_loras or 1)) if lora_config else 0 lora_vocab = (lora_config.lora_extra_vocab_size * (lora_config.max_loras or 1)) if lora_config else 0
self.vocab_size = config.vocab_size + lora_vocab self.vocab_size = config.vocab_size + lora_vocab
self.org_vocab_size = config.vocab_size self.org_vocab_size = config.vocab_size
self.wte = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
if get_pp_group().is_first_rank or (config.tie_word_embeddings and get_pp_group().is_last_rank): if get_pp_group().is_first_rank or (config.tie_word_embeddings and get_pp_group().is_last_rank):
self.embed_tokens = VocabParallelEmbedding( self.embed_tokens = VocabParallelEmbedding(
self.vocab_size, self.vocab_size,
@@ -2047,7 +2035,7 @@ class HunyuanImage3Model(nn.Module):
return_dict = return_dict if return_dict is not None else self.config.use_return_dict return_dict = return_dict if return_dict is not None else self.config.use_return_dict
if inputs_embeds is None: if inputs_embeds is None:
inputs_embeds = self.wte(input_ids) inputs_embeds = self.embed_tokens(input_ids)
# embed positions # embed positions
hidden_states = inputs_embeds hidden_states = inputs_embeds
@@ -37,7 +37,6 @@ from .hunyuan_image_3_transformer import (
UNetDown, UNetDown,
UNetUp, UNetUp,
build_batch_2d_rope, build_batch_2d_rope,
get_full_state_dict,
real_batched_index_select, real_batched_index_select,
) )
@@ -113,10 +112,10 @@ class HunyuanImage3Pipeline(HunyuanImage3PreTrainedModel, GenerationMixin):
self.vllm_config = get_current_vllm_config() self.vllm_config = get_current_vllm_config()
self.post_init() self.post_init()
def pre_load(self): def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
tp_rank = get_tensor_model_parallel_rank() skip_prefixes = ["lm_head."] if self.hf_config.tie_word_embeddings else []
state_dict = get_full_state_dict(self.od_config.model) # List of unexpected keywords in weight names
non_layer_prefixes = [ non_model_layer_prefixes = [
"vae", "vae",
"vision_model", "vision_model",
"vision_aligner", "vision_aligner",
@@ -129,32 +128,15 @@ class HunyuanImage3Pipeline(HunyuanImage3PreTrainedModel, GenerationMixin):
"time_embed_2", "time_embed_2",
"final_layer.model", "final_layer.model",
] ]
filtered_sd = {} tp_rank = get_tensor_model_parallel_rank()
for k, v in state_dict.items(): device_str = f"{self.model.device.type}:{tp_rank}"
if any(k.startswith(prefix) for prefix in non_layer_prefixes): named_modules = dict(self.named_modules())
filtered_sd[k] = v for prefix in non_model_layer_prefixes:
mod = named_modules.get(prefix)
if mod:
mod.to(device_str)
missing, unexpected = self.load_state_dict(filtered_sd, strict=False)
for prefix in non_layer_prefixes:
if hasattr(self, prefix.split(".")[0]):
module = dict(self.named_modules()).get(prefix)
if module:
module.to(f"{self.model.device.type}:{tp_rank}")
def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
self.pre_load()
skip_prefixes = ["lm_head."] if self.hf_config.tie_word_embeddings else []
# List of unexpected keywords in weight names
unexpected_keywords = [ unexpected_keywords = [
"vae",
"vision_aligner",
"vision_model",
"final_layer",
"patch_embed",
"timestep_emb",
"time_embed",
"time_embed_2",
"guidance_emb", "guidance_emb",
"timestep_r_emb", "timestep_r_emb",
] ]
@@ -892,7 +874,7 @@ class HunyuanImage3Pipeline(HunyuanImage3PreTrainedModel, GenerationMixin):
custom_pos_emb = self.get_pos_emb(custom_pos_emb, position_ids) custom_pos_emb = self.get_pos_emb(custom_pos_emb, position_ids)
inputs_embeds = self.model.wte(input_ids) inputs_embeds = self.model.embed_tokens(input_ids)
bsz, seq_len, n_embd = inputs_embeds.shape bsz, seq_len, n_embd = inputs_embeds.shape