When training SDXL with fsdp, the error was raised when calling
self.model.feedforward_model, self.model.guidance_model = accelerator.prepare(
self.model.feedforward_model, self.model.guidance_model
)
I think it is because in self.model.guidance_model, real_unet is fp16 but fake_unet is float32, but most related to the package version
Here is what I installed. Any ideas to fix this
accelerate 0.23.0
aiofiles 23.2.1
aiohappyeyeballs 2.4.4
aiohttp 3.10.11
aiosignal 1.3.1
annotated-doc 0.0.4
annotated-types 0.7.0
anyio 4.5.2
async-timeout 5.0.1
attrs 25.3.0
certifi 2025.11.12
charset-normalizer 3.4.4
clean-fid 0.1.35
click 8.1.8
clip 1.0
cmake 4.2.0
contourpy 1.1.1
cycler 0.12.1
datasets 3.0.1
diffusers 0.28.2
dill 0.3.8
DMD2 0.0.1 /home/DMD2
eval_type_backport 0.3.1
evaluate 0.4.6
exceptiongroup 1.3.1
fairscale 0.4.13
fastapi 0.124.4
ffmpy 0.5.0
filelock 3.16.1
fonttools 4.57.0
frozenlist 1.5.0
fsspec 2024.6.1
ftfy 6.2.3
gitdb 4.0.12
GitPython 3.1.45
gradio 4.44.1
gradio_client 1.3.0
h11 0.16.0
hf-xet 1.2.0
httpcore 1.0.9
httpx 0.28.1
huggingface-hub 0.22.0
idna 3.11
image-reward 1.5
imageio 2.35.1
importlib_metadata 8.5.0
importlib_resources 6.4.5
Jinja2 3.1.6
kiwisolver 1.4.7
lit 18.1.8
lmdb 1.7.5
markdown-it-py 3.0.0
MarkupSafe 2.1.5
matplotlib 3.7.5
mdurl 0.1.2
mpmath 1.3.0
multidict 6.1.0
multiprocess 0.70.16
networkx 3.1
numpy 1.24.4
nvidia-cublas-cu11 11.10.3.66
nvidia-cuda-cupti-cu11 11.7.101
nvidia-cuda-nvrtc-cu11 11.7.99
nvidia-cuda-runtime-cu11 11.7.99
nvidia-cudnn-cu11 8.5.0.96
nvidia-cufft-cu11 10.9.0.58
nvidia-curand-cu11 10.2.10.91
nvidia-cusolver-cu11 11.4.0.1
nvidia-cusparse-cu11 11.7.4.91
nvidia-nccl-cu11 2.14.3
nvidia-nvtx-cu11 11.7.91
open_clip_torch 2.32.0
opencv-python 4.12.0.88
orjson 3.10.15
packaging 25.0
pandas 2.0.3
peft 0.12.0
pillow 10.4.0
pip 24.2
piq 0.7.0
platformdirs 4.3.6
propcache 0.2.0
protobuf 5.29.5
psutil 7.1.3
pyarrow 17.0.0
pydantic 2.10.6
pydantic_core 2.27.2
pydub 0.25.1
Pygments 2.19.2
pyparsing 3.1.4
python-dateutil 2.9.0.post0
python-multipart 0.0.20
pytz 2025.2
PyYAML 6.0.3
regex 2024.11.6
requests 2.32.4
rich 14.2.0
ruff 0.14.10
safetensors 0.5.3
scipy 1.10.1
semantic-version 2.10.0
sentry-sdk 2.48.0
setuptools 75.1.0
shellingham 1.5.4
six 1.17.0
smmap 5.0.2
sniffio 1.3.1
starlette 0.44.0
sympy 1.13.3
timm 0.6.13
tokenizers 0.19.1
tomlkit 0.12.0
torch 2.0.1
torchvision 0.15.2
tqdm 4.67.1
transformers 4.40.2
triton 2.0.0
typer 0.20.0
typing_extensions 4.13.2
tzdata 2025.3
urllib3 2.2.3
uvicorn 0.33.0
wandb 0.23.1
wcwidth 0.2.14
websockets 12.0
wheel 0.44.0
xxhash 3.6.0
yarl 1.15.2
zipp 3.20.2
When training SDXL with fsdp, the error was raised when calling
self.model.feedforward_model, self.model.guidance_model = accelerator.prepare(
self.model.feedforward_model, self.model.guidance_model
)
I think it is because in self.model.guidance_model, real_unet is fp16 but fake_unet is float32, but most related to the package version
Here is what I installed. Any ideas to fix this