Unnecessary download when loading BF16 variant

[francois-rozet]

Describe the bug

Loading the BF16 variant of SANA leads to downloading and loading non-BF16 files.

Reproduction

import torch

from diffusers import SanaPipeline

pipe = SanaPipeline.from_pretrained(
    "Efficient-Large-Model/Sana_1600M_1024px_BF16_diffusers",
    torch_dtype=torch.bfloat16,
    variant="bf16",
)

A mixture of bf16 and non-bf16 filenames will be loaded.
Loaded bf16 filenames:
[transformer/diffusion_pytorch_model.bf16.safetensors, text_encoder/model.bf16-00001-of-00002.safetensors, vae/diffusion_pytorch_model.bf16.safetensors, text_encoder/model.bf16-00002-of-00002.safetensors]
Loaded non-bf16 filenames:
[transformer/diffusion_pytorch_model-00002-of-00002.safetensors, transformer/diffusion_pytorch_model-00001-of-00002.safetensors
If this behavior is not expected, please check your folder structure.
Loading pipeline components...:   0%|          | 0/5 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]

Logs

System Info

• 🤗 Diffusers version: 0.32.2
• Platform: Linux-6.1.97.1.fi-x86_64-with-glibc2.28
• Running on Google Colab?: No
• Python version: 3.10.13
• PyTorch version (GPU?): 2.4.1+cu121 (False)
• Flax version (CPU?/GPU?/TPU?): not installed (NA)
• Jax version: not installed
• JaxLib version: not installed
• Huggingface_hub version: 0.29.3
• Transformers version: 4.50.0
• Accelerate version: 1.0.0
• PEFT version: not installed
• Bitsandbytes version: not installed
• Safetensors version: 0.4.5
• xFormers version: not installed
• Accelerator: NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.
• Using GPU in script?: No
• Using distributed or parallel set-up in script?: No

Who can help?

No response

[nitinmukesh]

I thought this was fixed
#10646

[EDIT]
Can you try installing from source
pip uninstall diffusers
pip install git+https://github.com/huggingface/diffuser.git

[francois-rozet]

Hi @nitinmukesh, thanks for the answer. This is indeed fixed in the latest (still unpublished) version. I will close this issue, but it would be good to publish a new release with this fix.