Update app.py

app.py

@@ -382,7 +382,7 @@ def plot_confusion_matrix(y_true, y_pred, title, light_mode=False):
 
 def identical_train_test_split(output_emb, output_raw, labels, train_percentage):
 
-    torch.manual_seed(seed)
+    torch.manual_seed(42)
     
     N = output_emb.shape[0]
     indices = torch.randperm(N) 
@@ -623,7 +623,7 @@ with gr.Blocks(css="""
                   </ul>
                 </li>
                 <li><b>🔗 Tip 1</b>: Instructions for organizing your dataset are available at the bottom of the page.</li>
-                <li><b>🔗 Tip 2</b>: As the computations and inference are performed on HuggingFace CPUs, please use small datasets for faster demo experience (say <500 samples). </li>
+                <li><b>🔗 Tip 2</b>: As the computations and inference are performed on HuggingFace CPUs, please use small datasets for faster demo experience (say <400 samples). Clone the model from <a href="https://huggingface.co/wi-lab/lwm" target="_blank">here</a> and use any number of samples locally.</li>
                 <li><b>🔗 Tip 3</b>: Your dataset will be normalized automatically based on outdoor environments. </li>
                 <li><b>💼 No Downstream Model</b>: Instead of a complex downstream model, we classify each sample based on its distance to the centroid of training samples from each class (LoS/NLoS).</il>
               </ul>