update delpoy

app.py

@@ -1,14 +1,13 @@
 # Author: Huzheng Yang
 # %%
-USE_HUGGINGFACE_SPACE = True
-BATCH_SIZE = 4
-
-if USE_HUGGINGFACE_SPACE:  # huggingface ZeroGPU, dynamic GPU allocation 
-    try:
-        import spaces
-    except ImportError:
-        USE_HUGGINGFACE_SPACE = False  # run on local machine
-        BATCH_SIZE = 1
+import os
+USE_HUGGINGFACE_ZEROGPU = os.getenv("USE_HUGGINGFACE_ZEROGPU", "False").lower() in ["true", "1", "yes"]
+#%%
+if USE_HUGGINGFACE_ZEROGPU:  # huggingface ZeroGPU, dynamic GPU allocation 
+    import spaces
+    BATCH_SIZE = 4
+else:  # run on local machine
+    BATCH_SIZE = 1
 
 import gradio as gr
 
@@ -18,7 +17,6 @@ from PIL import Image
 import numpy as np
 import time
 import threading
-import os
 
 from ncut_pytorch.backbone import extract_features, load_model
 from ncut_pytorch.backbone import MODEL_DICT, LAYER_DICT, RES_DICT
@@ -257,6 +255,7 @@ def ncut_run(
         logging_str += "Running NCut for each image independently\n"
         rgb = []
         for i_image in range(features.shape[0]):
+            logging_str += f"Image #{i_image+1}\n"
             feature = features[i_image]
             _rgb, _logging_str, _ = compute_ncut(
                 feature[None],
@@ -315,7 +314,7 @@ def _ncut_run(*args, **kwargs):
             torch.cuda.empty_cache()
         return [], "Error: " + str(e)
 
-if USE_HUGGINGFACE_SPACE:
+if USE_HUGGINGFACE_ZEROGPU:
     @spaces.GPU(duration=20)
     def quick_run(*args, **kwargs):
         return _ncut_run(*args, **kwargs)
@@ -332,7 +331,7 @@ if USE_HUGGINGFACE_SPACE:
     def super_duper_long_run(*args, **kwargs):
         return _ncut_run(*args, **kwargs)
 
-if not USE_HUGGINGFACE_SPACE:
+if not USE_HUGGINGFACE_ZEROGPU:
     def quick_run(*args, **kwargs):
         return _ncut_run(*args, **kwargs)
 
@@ -739,7 +738,7 @@ with demo:
                 with gr.Accordion("➡️ Recursion config", open=True):
                     l1_num_eig_slider = gr.Slider(1, 1000, step=1, label="Recursion #1: N eigenvectors", value=100, elem_id="l1_num_eig")
                     l2_num_eig_slider = gr.Slider(1, 1000, step=1, label="Recursion #2: N eigenvectors", value=50, elem_id="l2_num_eig")
-                    l3_num_eig_slider = gr.Slider(1, 1000, step=1, label="Recursion #3: N eigenvectors", value=25, elem_id="l3_num_eig")
+                    l3_num_eig_slider = gr.Slider(1, 1000, step=1, label="Recursion #3: N eigenvectors", value=50, elem_id="l3_num_eig")
                     metric_dropdown = gr.Dropdown(["euclidean", "cosine"], label="Recursion distance metric", value="cosine", elem_id="recursion_metric")
                     
                 [
@@ -826,7 +825,7 @@ with demo:
         )    
     
     with gr.Tab('Text'): 
-        if USE_HUGGINGFACE_SPACE:
+        if USE_HUGGINGFACE_ZEROGPU:
             from app_text import make_demo
         else:
             from draft_gradio_app_text import make_demo
@@ -901,7 +900,7 @@ with demo:
         
 
 
-if USE_HUGGINGFACE_SPACE:
+if USE_HUGGINGFACE_ZEROGPU:
     from ncut_pytorch.backbone import download_all_models
     threading.Thread(target=download_all_models).start()
     from ncut_pytorch.backbone_text import download_all_models