Update app.py

app.py

@@ -157,7 +157,7 @@ import gradio as gr
 import numpy as np
 from PIL import Image
 
-from accelerate import Accelerator
+#from accelerate import Accelerator
 
 #import diffusers
 from diffusers import AutoencoderKL, StableDiffusionXLPipeline
@@ -177,7 +177,7 @@ os.putenv('HF_HUB_ENABLE_HF_TRANSFER','1')
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 os.environ["SAFETENSORS_FAST_GPU"] = "1"
 
-accelerator = Accelerator(mixed_precision="bf16") # Example
+#accelerator = Accelerator(mixed_precision="bf16") # Example
 
 upscaler = UpscaleWithModel.from_pretrained("Kim2091/ClearRealityV1").to(torch.device("cuda:0"))
 
@@ -203,9 +203,9 @@ def load_and_prepare_model():
     print(f'init noise scale: {pipe.scheduler.init_noise_sigma}')
     pipe.watermark=None
     pipe.safety_checker=None  
-    #pipe.to(torch.device('cuda:0'), torch.bfloat16)
-    pipe.to(torch.bfloat16)
-    pipe.to(accelerator.device)
+    pipe.to(torch.device('cuda:0'), torch.bfloat16)
+    #pipe.to(torch.bfloat16)
+    #pipe.to(accelerator.device)
     return pipe
 
 #hidet.option.parallel_build(False)
@@ -273,8 +273,7 @@ def generate_30(
     filename = uploadNote(prompt,num_inference_steps,guidance_scale,timestamp)
     upload_to_ftp(filename) 
     batch_options = options.copy()
-    with accelerator.autocast():
-        rv_image = pipe(**batch_options).images[0]
+    rv_image = pipe(**batch_options).images[0]
     sd_image_path = f"rv_C_{timestamp}.png"
     rv_image.save(sd_image_path,optimize=False,compress_level=0)
     upload_to_ftp(sd_image_path)