Fixed some stuff

app.py

@@ -1,9 +1,3 @@
-from diffusers import DDIMScheduler
-from pipeline_stable_diffusion_xl_opt import StableDiffusionXLPipeline
-from injection_utils import regiter_attention_editor_diffusers
-from bounded_attention import BoundedAttention
-from pytorch_lightning import seed_everything
-
 import spaces
 import gradio as gr
 import torch
@@ -11,6 +5,12 @@ import nltk
 import numpy as np
 from PIL import Image, ImageDraw
 
+from diffusers import DDIMScheduler
+from pipeline_stable_diffusion_xl_opt import StableDiffusionXLPipeline
+from injection_utils import regiter_attention_editor_diffusers
+from bounded_attention import BoundedAttention
+from pytorch_lightning import seed_everything
+
 from functools import partial
 
 RESOLUTION = 256
@@ -113,14 +113,15 @@ def convert_token_indices(token_indices, nested=False):
 def draw(sketchpad):
     boxes = []
     for i, layer in enumerate(sketchpad['layers']):
-        mask = (layer != 0)
-        if mask.sum() < 0:
-            raise gr.Error(f'Box in layer {i} is too small')
-
-        x1x2 = np.where(mask.max(0) != 0)[0] / RESOLUTION
-        y1y2 = np.where(mask.max(1) != 0)[0] / RESOLUTION
-        y1, y2 = y1y2.min(), y1y2.max()
-        x1, x2 = x1x2.min(), x1x2.max()
+        non_zeros = layer.nonzero()
+        x1 = x2 = y1 = y2 = 0
+        if len(non_zeros[0]) > 0:
+            x1x2 = non_zeros[1] / layer.shape[1]
+            y1y2 = non_zeros[0] / layer.shape[0]
+            x1 = x1x2.min()
+            x2 = x1x2.max()
+            y1 = y1y2.min()
+            y2 = y1y2.max()
 
         if (x2 - x1 < MIN_SIZE) or (y2 - y1 < MIN_SIZE):
             raise gr.Error(f'Box in layer {i} is too small')
@@ -220,20 +221,20 @@ def main():
             )
 
             subject_token_indices = gr.Textbox(
-                label="The token indices of each subject (separate indices for the same subject with commas, and between different subjects with semicolons)",
+                label="The token indices of each subject (separate indices for the same subject with commas, and for different subjects with semicolons)",
             )
 
             filter_token_indices = gr.Textbox(
-                label="The token indices to filter, i.e. conjunctions, number, postional relations, etc. (if left empty, this will be automatically inferred)",
+                label="Optional: The token indices to filter, i.e. conjunctions, numbers, postional relations, etc. (if left empty, this will be automatically inferred)",
             )
 
             num_tokens = gr.Textbox(
-                label="The number of tokens in the prompt (can be left empty, but we recommend filling this, so we can verify your input, as sometimes rare words are split into more than one token)",
+                label="Optional: The number of tokens in the prompt (We use this to verify your input, as sometimes rare words are split into more than one token)",
             )
 
             with gr.Row():
                 sketchpad = gr.Sketchpad(label="Sketch Pad", width=RESOLUTION, height=RESOLUTION)
-                layout_image = gr.Image(type="pil", label="Bounding Boxes", interactive=False, width=RESOLUTION, height=RESOLUTION)
+                layout_image = gr.Image(type="pil", label="Bounding Boxes", interactive=False, width=RESOLUTION, height=RESOLUTION, scale=1)
 
             with gr.Row():
                 clear_button = gr.Button(value='Clear')
@@ -278,7 +279,7 @@ def main():
                         </div>
                     """
                     gr.HTML(description)
-                    batch_size = gr.Slider(minimum=1, maximum=5, step=1, value=1, label="Number of samples")
+                    batch_size = gr.Slider(minimum=1, maximum=5, step=1, value=1, label="Number of samples (currently limited to one sample)")
                     init_step_size = gr.Slider(minimum=0, maximum=50, step=0.5, value=25, label="Initial step size")
                     final_step_size = gr.Slider(minimum=0, maximum=20, step=0.5, value=10, label="Final step size")
                     num_clusters_per_subject = gr.Slider(minimum=0, maximum=5, step=0.5, value=3, label="Number of clusters per subject")