how to use the trained lora safetensors in pipeline?

I used your great work to train a model, and the generated safetensors files works in WebUI;
however it could not be used in pure code with pipeline, such as
```
# in main
pipeline = StableDiffusionPipeline.from_ckpt('e:\\xxx\\trained\\stable-diffusion-v1-5-lxq-3.safetensors')
pipeline.to('cuda')
drawn = pipeline(
prompt='a girl with a cat',
height=1024,
width=1024,
num_inference_steps=20,
guidance_scale=5,
negative_prompt='lowres'
)
image = drawn.images[0]
image.show()
```

would result in error, like

```
global_step key not found in model
Traceback (most recent call last):
File "E:\xxx\taiyi\drawer\SafetensorsDrawTest.py", line 4, in
pipeline = StableDiffusionPipeline.from_ckpt(
File "E:\xxxr\venv\lib\site-packages\diffusers\loaders.py", line 1284, in from_ckpt
pipe = download_from_original_stable_diffusion_ckpt(
File "E:\xxx\venv\lib\site-packages\diffusers\pipelines\stable_diffusion\convert_from_ckpt.py", line 1177, in download_from_original_stable_diffusion_ckpt
converted_unet_checkpoint = convert_ldm_unet_checkpoint(
File "E:\xxx\venv\lib\site-packages\diffusers\pipelines\stable_diffusion\convert_from_ckpt.py", line 376, in convert_ldm_unet_checkpoint
new_checkpoint["time_embedding.linear_1.weight"] = unet_state_dict["time_embed.0.weight"]
KeyError: 'time_embed.0.weight'
```

Could you please give some advice?
Thx.

Start with the reported StableDiffusionPipeline.from_ckpt call and the traceback through download_from_original_stable_diffusion_ckpt and convert_ldm_unet_checkpoint. Determine whether the trained safetensors format is supported by this pipeline, then document the supported loading procedure or the incompatibility and its workaround.

Written by the indexing model from the issue text.