Better seeding + slight Gradio demo example schedule changes

app_ctrlx.py

@@ -157,9 +157,7 @@ def inference(
     structure_schedule, appearance_schedule, use_advanced_config,
     control_config,
 ):
-    global pipe, refiner
-
-    torch.manual_seed(seed)
+    seed_everything(seed)
 
     pipe.scheduler.set_timesteps(num_inference_steps, device=device)
     timesteps = pipe.scheduler.timesteps
@@ -310,7 +308,7 @@ def inference(
                 "a 3D mesh of a cat",
                 "",
                 "a photo of a tiger standing on snow",
-                0.6, 0.6,
+                0.5, 0.6,
             ],
             [
                 "assets/images/dog__sketch.jpg",
@@ -358,7 +356,7 @@ def inference(
                 "a segmentation map of a bear and an avocado",
                 "",
                 "a realistic photo of a bear and an avocado in a forest",
-                0.6, 0.6,
+                0.5, 0.6,
             ],
             [
                 "assets/images/cat__point_cloud.jpg",
@@ -382,7 +380,7 @@ def inference(
                 "a 3D model of a person holding a sword and shield",
                 "",
                 "a photo of a medieval soldier standing on a barren field, raining",
-                0.6, 0.6,
+                0.5, 0.6,
             ],
             [
                 "assets/images/person__mesh.jpg",

ctrl_x/utils/utils.py

@@ -1,7 +1,20 @@
+import random
+from os import environ
+
+import numpy as np
 import torch
 
 
-JPEG_QUALITY = 95
+JPEG_QUALITY = 100
+
+
+def seed_everything(seed):
+    random.seed(seed)
+    environ["PYTHONHASHSEED"] = str(seed)
+    np.random.seed(seed)
+    torch.manual_seed(seed)
+    torch.backends.cudnn.deterministic = True
+    torch.backends.cudnn.benchmark = False
 
 
 def exists(x):

run_ctrlx.py

@@ -13,9 +13,6 @@
 from ctrl_x.utils.sdxl import *
 
 
-JPEG_QUALITY = 100
-
-
 @torch.no_grad()
 def inference(
     pipe, refiner, device,
@@ -27,7 +24,7 @@ def inference(
     width, height,
     structure_schedule, appearance_schedule,
 ):  
-    torch.manual_seed(seed)
+    seed_everything(seed)
 
     pipe.scheduler.set_timesteps(num_inference_steps, device=device)
     timesteps = pipe.scheduler.timesteps