refactor: change sampling to torch.inference_mode

main.py

@@ -202,7 +202,7 @@ def generate_strokes(
                                   char_seq_lengths=priming_text_len_tensor)
         
     
-    with torch.no_grad():
+    with torch.inference_mode():
         try:
             stroke_tensors = SCRIPTED_MODEL.sample(
                 char_seq,