gpu

.gitignore

@@ -0,0 +1,2 @@
+.vscode
+.tmp/res

app.py

@@ -2,7 +2,7 @@
 import time
 import os
 import gradio as gr
-# import spaces
+import spaces
 from pnpxai.core.experiment.auto_explanation import AutoExplanationForImageClassification
 from pnpxai.core.detector.detector import extract_graph_data, symbolic_trace
 import matplotlib.pyplot as plt
@@ -343,7 +343,7 @@ class Experiment(Component):
             _plots += [gr.Image(value=None, label="Blank", visible=False)] * ((buffer_n_rows - n_rows) * PLOT_PER_LINE)
             return _plots
         
-        # @spaces.GPU
+        @spaces.GPU
         def render_plots(data_id, *metric_inputs):
             # Clear Cache Files
             # print(f"GPU Check: {torch.cuda.is_available()}")
@@ -485,7 +485,7 @@ class ExplainerCheckbox(Component):
         idx = [metric.__class__.__name__ for metric in metric_info[0]].index(metric_name)
         return metric_info[1][idx]
 
-    # @spaces.GPU
+    @spaces.GPU
     def optimize(self):
         # if self.explainer_name in ["Lime", "KernelShap", "IntegratedGradients"]:
         #     gr.Info("Lime, KernelShap and IntegratedGradients currently do not support hyperparameter optimization.")
@@ -649,8 +649,8 @@ from torch.utils.data import DataLoader
 from helpers import get_imagenet_dataset, get_torchvision_model, denormalize_image
 
 os.environ['GRADIO_TEMP_DIR'] = '.tmp'
-# device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-device = torch.device("cpu")
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+# device = torch.device("cpu")
 
 def target_visualizer(x): return dataset.dataset.idx_to_label(x.item())