Failed to offload sd3.5L to cpu

sd3 branch
when I tried to lora sd3.5-Large，lode_safetensors failed where I found that the modes are set to be offloaded to cpu
codes:
```
def load_target_model(self, args, weight_dtype, accelerator):
# currently offload to cpu for some models

# if the file is fp8 and we are using fp8_base, we can load it as is (fp8)
loading_dtype = None if args.fp8_base else weight_dtype

# if we load to cpu, flux.to(fp8) takes a long time, so we should load to gpu in future
state_dict = utils.load_safetensors(
args.pretrained_model_name_or_path, "cpu", disable_mmap=args.disable_mmap_load_safetensors, dtype=loading_dtype
)
mmdit = sd3_utils.load_mmdit(state_dict, loading_dtype, "cpu")
```
```
Traceback (most recent call last):
File "/home/bml/workspace/jinyibo/workspace/llm-factory/src/sd-scripts-sd3/library/utils.py", line 367, in load_safetensors
state_dict = load_file(path, device=device)
File "/opt/conda/envs/codelab/lib/python3.10/site-packages/safetensors/torch.py", line 313, in load_file
with safe_open(filename, framework="pt", device=device) as f:
OSError: No such device (os error 19)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/home/bml/workspace/jinyibo/workspace/llm-factory/src/sd-scripts-sd3/sd3_train_network.py", line 480, in
trainer.train(args)
File "/home/bml/workspace/jinyibo/workspace/llm-factory/src/sd-scripts-sd3/train_network.py", line 413, in train
model_version, text_encoder, vae, unet = self.load_target_model(args, weight_dtype, accelerator)
File "/home/bml/workspace/jinyibo/workspace/llm-factory/src/sd-scripts-sd3/sd3_train_network.py", line 70, in load_target_model
state_dict = utils.load_safetensors(
File "/home/bml/workspace/jinyibo/workspace/llm-factory/src/sd-scripts-sd3/library/utils.py", line 369, in load_safetensors
state_dict = load_file(path) # prevent device invalid Error
File "/opt/conda/envs/codelab/lib/python3.10/site-packages/safetensors/torch.py", line 313, in load_file
with safe_open(filename, framework="pt", device=device) as f:
OSError: No such device (os error 19)
```

Start by reading library/utils.py around load_safetensors and sd3_train_network.py around load_target_model, then inspect how the safetensors loader handles the requested device and its fallback. Reproduce the SD3.5-Large LoRA loading path and confirm that the model can be loaded without the reported No such device error.

Written by the indexing model from the issue text.