Update app.py

app.py

@@ -124,10 +124,13 @@ def sample_then_run(net):
     standev = torch.std(proj, 0)
     
     # sample
-    sample = torch.zeros([1, 1000]).to(device)
+    sample = torch.zeros([1, 10000]).to(device)
+
+    #only first 1000 PCs
     for i in range(1000):
         sample[0, i] = torch.normal(m[i], standev[i], (1,1))
 
+    
     net = "model_"+str(uuid.uuid4())[:4]+".pt"
     torch.save(sample, net)
     
@@ -148,7 +151,7 @@ def inference(net, prompt, negative_prompt, guidance_scale, ddim_steps, seed):
     v.to(device)
     
     weights = torch.load(net).to(device)
-    network = LoRAw2w(weights, mean, std, v[:, :1000], 
+    network = LoRAw2w(weights, mean, std, v[:, :10000], 
                     unet,
                     rank=1,
                     multiplier=1.0,
@@ -215,7 +218,7 @@ def edit_inference(net, prompt, negative_prompt, guidance_scale, ddim_steps, see
 
     
     weights = torch.load(net).to(device)
-    network = LoRAw2w(weights, mean, std, v[:, :1000], 
+    network = LoRAw2w(weights, mean, std, v[:, :10000], 
                     unet,
                     rank=1,
                     multiplier=1.0,
@@ -386,32 +389,25 @@ def run_inversion(self, dict, pcs, epochs, weight_decay,lr):
     
     
 @spaces.GPU
-def file_upload(self, file):
-    proj = torch.load(file.name).to(device)
+def file_upload(file):
+    device="cuda"
+    weights = torch.load(file.name).to(device)
+    net = "model_"+str(uuid.uuid4())[:4]+".pt"
+    torch.save(weights, net)
         
     #pad to 10000 Principal components to keep everything consistent
-    pcs = proj.shape[1]
+    pcs = net.shape[1]
     padding =  torch.zeros((1,10000-pcs)).to(device)
-    proj = torch.cat((proj, padding), 1)
-    unet, _, _, _, _ = load_models(device)
-        
+    weights = torch.cat((weights, padding), 1)
+       
         
-    network = LoRAw2w( proj, mean, std, v[:, :10000], 
-                                unet,
-                                rank=1,
-                                multiplier=1.0,
-                                alpha=27.0,
-                                train_method="xattn-strict"
-                            ).to(device, torch.bfloat16)
-                
-            
-    prompt = "sks person"
-    negative_prompt = "low quality, blurry, unfinished, nudity"
+    image = prompt = "sks person"
+    negative_prompt = "low quality, blurry, unfinished, nudity, weapon"
     seed = 5
     cfg = 3.0
     steps = 25
-    image = inference( prompt, negative_prompt, cfg, steps, seed)
-    return image
+    image = inference(net, prompt, negative_prompt, cfg, steps, seed)
+    return net,net,image
     
     
 
@@ -516,7 +512,7 @@ with gr.Blocks(css="style.css") as demo:
         submit.click(
                 fn=edit_inference, inputs=[net, prompt, negative_prompt, cfg, steps, seed, injection_step, a1, a2, a3, a4], outputs=[net, gallery]
             )
-        # file_input.change(fn=model.file_upload, inputs=file_input, outputs = gallery)
+        file_input.change(fn=file_upload, inputs=[file_input, net], outputs = [net, gallery])