DonImages commited on
Commit
fc20c03
·
verified ·
1 Parent(s): 90933cd

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +11 -11
app.py CHANGED
@@ -13,23 +13,23 @@ else:
13
 
14
  # Load the Stable Diffusion 3.5 model
15
  model_id = "stabilityai/stable-diffusion-3.5-medium"
16
- pipe = StableDiffusion3Pipeline.from_pretrained(model_id, torch_dtype=torch.float16)
17
- pipe.to("cpu")
 
 
 
18
 
19
  # Custom method to load and apply LoRA weights to the Stable Diffusion pipeline
20
  def load_lora_model(pipe, lora_model_path):
21
- # Load the LoRA weights (assuming it's a PyTorch .pth file)
22
  lora_weights = torch.load(lora_model_path, map_location="cpu")
23
-
24
- # Apply LoRA weights to the parameters of the model
25
- for name, param in pipe.named_parameters():
26
  if name in lora_weights:
27
- param.data += lora_weights[name] # Apply LoRA weights to the parameters
28
-
29
- return pipe # Return the updated model
30
 
31
- # Define the path to the LoRA model (Local path in your Hugging Face Space)
32
- lora_model_path = "./lora_model.pth" # Local path to LoRA model
33
 
34
  # Load and apply the LoRA model weights
35
  pipe = load_lora_model(pipe, lora_model_path)
 
13
 
14
  # Load the Stable Diffusion 3.5 model
15
  model_id = "stabilityai/stable-diffusion-3.5-medium"
16
+ pipe = StableDiffusion3Pipeline.from_pretrained(model_id) # Removed torch_dtype argument
17
+ pipe.to("cpu") # Ensuring it runs on CPU
18
+
19
+ # Define the path to the LoRA model
20
+ lora_model_path = "./lora_model.pth" # Assuming the file is saved locally
21
 
22
  # Custom method to load and apply LoRA weights to the Stable Diffusion pipeline
23
  def load_lora_model(pipe, lora_model_path):
24
+ # Load the LoRA weights
25
  lora_weights = torch.load(lora_model_path, map_location="cpu")
26
+
27
+ # Apply weights to the UNet submodule
28
+ for name, param in pipe.unet.named_parameters(): # Accessing unet parameters
29
  if name in lora_weights:
30
+ param.data += lora_weights[name]
 
 
31
 
32
+ return pipe
 
33
 
34
  # Load and apply the LoRA model weights
35
  pipe = load_lora_model(pipe, lora_model_path)