Update README.md

README.md

@@ -71,7 +71,7 @@ You can run the smashed model on cards with less than 12 GB of memory with these
     ```python 
     
     import torch
-    
+
     from optimum.quanto import freeze, qfloat8, quantize
     
     from diffusers import FlowMatchEulerDiscreteScheduler, AutoencoderKL
@@ -104,17 +104,17 @@ You can run the smashed model on cards with less than 12 GB of memory with these
     )
     pipe.text_encoder_2 = text_encoder_2
     pipe.transformer = transformer
-    pipe.enable_model_cpu_offload()
-    
+    # pipe.enable_model_cpu_offload()
+    pipe.to('cuda')
+    print('done')
     generator = torch.Generator().manual_seed(12345)
-    image = pipe(
+    pipe(
     "a cute apple smiling",
     guidance_scale=0.0,
     num_inference_steps=4,
     max_sequence_length=256,
     generator=torch.Generator("cpu").manual_seed(0)
       ).images[0]
-    image.save("flux-schnell.png")
     ```
 
 ## Configurations