Update app.py (#2)

- Update app.py (a91007977f3f198b4c04e58e1f27fe3456eb5e38)


Co-authored-by: Peter Lin <[email protected]>

app.py

@@ -2,6 +2,7 @@ import gradio as gr
 import torch
 from diffusers import StableDiffusionXLPipeline, EulerDiscreteScheduler
 from huggingface_hub import hf_hub_download
+from safetensors.torch import load_file
 import spaces
 
 
@@ -9,10 +10,10 @@ import spaces
 base = "stabilityai/stable-diffusion-xl-base-1.0"
 repo = "ByteDance/SDXL-Lightning"
 checkpoints = {
-    "1-Step" : ["sdxl_lightning_1step_unet_x0.pth", 1],
-    "2-Step" : ["sdxl_lightning_2step_unet.pth", 2],
-    "4-Step" : ["sdxl_lightning_4step_unet.pth", 4],
-    "8-Step" : ["sdxl_lightning_8step_unet.pth", 8],
+    "1-Step" : ["sdxl_lightning_1step_unet_x0.safetensors", 1],
+    "2-Step" : ["sdxl_lightning_2step_unet.safetensors", 2],
+    "4-Step" : ["sdxl_lightning_4step_unet.safetensors", 4],
+    "8-Step" : ["sdxl_lightning_8step_unet.safetensors", 8],
 }
 
 
@@ -35,7 +36,7 @@ def generate_image(prompt, ckpt):
         # Ensure sampler uses "trailing" timesteps.
         pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
         
-    pipe.unet.load_state_dict(torch.load(hf_hub_download(repo, checkpoint), map_location="cuda"))
+    pipe.unet.load_state_dict(load_file(hf_hub_download(repo, checkpoint), device="cuda"))
     image = pipe(prompt, num_inference_steps=num_inference_steps, guidance_scale=0).images[0]
     return image