Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.

[MinhVD-ZenAI]

I tried to load Flux_Turbo_Alpha into FluxKontextPipeline by this code, and save_pretrained to local.

import torch
from diffusers import FluxKontextPipeline
from diffusers.utils import load_image

pipe = FluxKontextPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-Kontext-dev", torch_dtype=torch.bfloat16
)

adapter_id = "alimama-creative/FLUX.1-Turbo-Alpha"
pipe.load_lora_weights(adapter_id,adapter_name="flux-Turbo-Alpha")
pipe.fuse_lora()

print("Start to save pipeline")
pipe.unfuse_lora()
pipe.save_pretrained("./kontext_turbo",max_shard_size="50GB")
print("Pipeline saved")

And then, I loaded it again.

import torch
from diffusers import FluxKontextPipeline, FluxPipeline
from diffusers.utils import load_image

pipe = FluxKontextPipeline.from_pretrained(
    "./kontext_turbo", torch_dtype=torch.bfloat16
)
pipe.to("cuda")


ref_url = "https://github.com/ZenAI-Vietnam/Flux-Kontext-pipelines/blob/main/assets/dinosaur_input.jpeg?raw=true"    
image = load_image(ref_url).convert("RGB")
prompt = "Change yellow dinosaur to red"

image = pipe(
    image=image,
    prompt=prompt,
    guidance_scale=2.5,
    num_inference_steps=10,
    generator=torch.Generator().manual_seed(42),
).images[0]

image.save("kontext_save_local.png")

But there was a bug like this:

Traceback (most recent call last):
  File "/workspace/Speedup_Kontext/pipeline_experiment/kontext_checkpoint/local.py", line 10, in <module>
    pipe.to("cuda")
  File "/workspace/miniconda3/envs/kontext/lib/python3.11/site-packages/diffusers/pipelines/pipeline_utils.py", line 541, in to
    module.to(device, dtype)
  File "/workspace/miniconda3/envs/kontext/lib/python3.11/site-packages/diffusers/models/modeling_utils.py", line 1384, in to
    return super().to(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/miniconda3/envs/kontext/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1355, in to
    return self._apply(convert)
           ^^^^^^^^^^^^^^^^^^^^
  File "/workspace/miniconda3/envs/kontext/lib/python3.11/site-packages/torch/nn/modules/module.py", line 915, in _apply
    module._apply(fn)
  File "/workspace/miniconda3/envs/kontext/lib/python3.11/site-packages/torch/nn/modules/module.py", line 915, in _apply
    module._apply(fn)
  File "/workspace/miniconda3/envs/kontext/lib/python3.11/site-packages/torch/nn/modules/module.py", line 915, in _apply
    module._apply(fn)
  File "/workspace/miniconda3/envs/kontext/lib/python3.11/site-packages/torch/nn/modules/module.py", line 942, in _apply
    param_applied = fn(param)
                    ^^^^^^^^^
  File "/workspace/miniconda3/envs/kontext/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1348, in convert
    raise NotImplementedError(
NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.

How to fix this bug? Thanks a lot.

[MinhVD-ZenAI]

cc @asomoza @sayakpaul @a-r-r-o-w, do you have any ideas? Please help me.

[sayakpaul]

Can you check if you can run inference after calling load_lora_weights()? Also, I think this is somewhat expected as we allow loading the Turbo LoRAs into Flux Control pipelines which expand on the embedding dimension but not the sequence dimension which is the case for Kontext.

diffusers/src/diffusers/loaders/lora_pipeline.py

Line 2620 in c02c4a6

def _maybe_expand_transformer_param_shape_or_error_(

[MinhVD-ZenAI]

yes @sayakpaul, after calling load_lora_weights(), it generates image normally. In theory, Kontext and Flux 1.Dev use the same FluxTransformer2DModel so I think they must have the same function. I also did this with FLuxPipeLine and it had the same issue.

And these are the libraries I installed:

accelerate               1.8.1
boto3                    1.39.1
botocore                 1.39.1
certifi                  2025.6.15
charset-normalizer       3.4.2
diffusers                0.35.0.dev0
filelock                 3.18.0
fsspec                   2025.5.1
gguf                     0.17.1
hf-xet                   1.1.5
huggingface-hub          0.33.1
idna                     3.10
importlib_metadata       8.7.0
iniconfig                2.1.0
Jinja2                   3.1.6
jmespath                 1.0.1
MarkupSafe               3.0.2
mpmath                   1.3.0
networkx                 3.5
ninja                    1.11.1.4
numpy                    2.3.1
nunchaku                 0.3.1+torch2.7
nvidia-cublas-cu12       12.6.4.1
nvidia-cuda-cupti-cu12   12.6.80
nvidia-cuda-nvrtc-cu12   12.6.77
nvidia-cuda-runtime-cu12 12.6.77
nvidia-cudnn-cu12        9.5.1.17
nvidia-cufft-cu12        11.3.0.4
nvidia-cufile-cu12       1.11.1.6
nvidia-curand-cu12       10.3.7.77
nvidia-cusolver-cu12     11.7.1.2
nvidia-cusparse-cu12     12.5.4.2
nvidia-cusparselt-cu12   0.6.3
nvidia-nccl-cu12         2.26.2
nvidia-nvjitlink-cu12    12.6.85
nvidia-nvtx-cu12         12.6.77
optimum-quanto           0.2.7
packaging                25.0
parameterized            0.9.0
peft                     0.16.0
pillow                   11.2.1
pip                      25.1
pluggy                   1.6.0
protobuf                 6.31.1
psutil                   7.0.0
Pygments                 2.19.2
pytest                   8.4.1
python-dateutil          2.9.0.post0
PyYAML                   6.0.2
regex                    2024.11.6
requests                 2.32.4
requests-mock            1.12.1
ruff                     0.12.1
s3transfer               0.13.0
safetensors              0.5.3
sentencepiece            0.2.0
setuptools               78.1.1
six                      1.17.0
sympy                    1.14.0
tokenizers               0.21.2
torch                    2.7.1
torchaudio               2.7.1+cu128
torchvision              0.22.1+cu128
tqdm                     4.67.1
transformers             4.53.0
triton                   3.3.1
typing_extensions        4.14.0
urllib3                  2.5.0
wheel                    0.45.1
zipp                     3.23.0

[sayakpaul]

Okay I will check.

[luke14free]

just noticed this, #12009 might be related?