Update app.py

app.py

@@ -1,34 +1,61 @@
 import gradio as gr
 import numpy as np
 import torch
-import dnnlib
-import legacy
+import pickle  #0.7) # Using a truncation_psi < 1 often gives better faces
+    
+    # Apply the gender transformation
+    w = w + gender_direction * strength
+    
+    # Synthesize the image from the modified w <--- CHANGE 1: Import pickle instead of legacy
 
 # Load pretrained StyleGAN2 model
-with open("stylegan2-ffhq-config-f.pkl", "rb") as f:
-    G = legacy.load_network_pkl(f)["G_ema"].to(torch.device("cpu"))  # 👈 RIGHT HERE
+print("Loading StyleGAN2 model...")
+device = torch.device("cpu")
+with open("stylegan2-ffhq-config-
+    img = G.synthesis(w, noise_mode="const")
+    
+    # Post-process the imagef.pkl", "rb") as f:
+    # CHANGE 2: Use pickle.load() instead of legacy. for display
+    img = (img.clamp(-1, 1) + 1) / 2.load_network_pkl()
+    G = pickle.load(f)["G_ema"].to(device)0
+    img = (img * 255).to(torch.uint8)
+    img = img.perm
+print("Model loaded successfully.")
 
 # Load gender direction vector
-gender_direction = np.load("stylegan2directions/gender.npy")
-gender_direction = torch.from_numpy(gender_direction).to(torch.float32).to(torch.device("cpu"))  # 👈 AND HERE
-
-def edit_gender(seed, strength):
-    rnd = np.random.RandomState(seed)
-    z = torch.from_numpy(rnd.randn(1, G.z_dim)).to(torch.float32).to(torch.device("cpu"))
-    w = G.mapping(z, None, truncation_psi=0.5)
-    w = w + gender_direction * strength
-    img = G.synthesis(w, noise_mode="const")
-    img = (img.clamp(-1, 1) + 1) * 127.5
-    img = img.permute(0, 2, 3, 1)[0].cpu().numpy().astype(np.uint8)
+print("Loading gender direction vector...")
+gender_direction = npute(0, 2, 3, 1)[0].cpu().numpy()
+    
     return img
 
+#.load("stylegan2directions/gender.npy")
+gender_direction = torch.from_numpy(gender Create and launch the Gradio interface
 gr.Interface(
     fn=edit_gender,
     inputs=[
-        gr.Slider(0, 10000, step=1, label="Random Seed"),
-        gr.Slider(-5, 5, step=0.1, label="Gender Strength (← Feminine | Masculine →)")
+        _direction).to(torch.float32).to(device)
+print("Vector loaded successfully.")
+
+
+def edit_gendergr.Slider(0, 10000, step=1, value=1234, label="Random Seed"),
+        gr.Slider(-5, 5, step=0.1, value=0, label(seed, strength):
+    # Ensure seed is an integer for the random state
+    seed = int(seed)
+="Gender Strength (← Feminine | Masculine →)")
     ],
-    outputs=gr.Image(label="Generated Face"),
+    outputs=gr.Image(label    
+    # Generate latent code from the seed
+    rnd = np.random.RandomState(seed)
+    z = torch.from_numpy(rnd.randn(1, G.z_dim)).to(torch="Generated Face"),
     title="StyleGAN2 Gender Editor",
-    description="Use the sliders to masculinize or feminize a generated face."
-).launch()
+    description="Move the slider to change the gender expression.float32).to(device)
+    
+    # Map the latent code to the intermediate space (W)
+    w = G.mapping(z, None, truncation_psi=0.7)
+    
+    # Apply the gender direction transformation
+    w = w + gender_direction * strength
+    
+    # Synt of the generated face. Change the seed to get a new face.",
+    allow_flagging="never"
+).launch()