Added compile

app.py

@@ -3,6 +3,7 @@ import tiktoken
 import gradio as gr
 import torch.nn.functional as F
 from model import GPT, GPTConfig
+torch._dynamo.reset()
 
 device = 'cpu'
 if torch.cuda.is_available():
@@ -14,6 +15,7 @@ model = GPT(GPTConfig())
 model.load_state_dict(torch.load("gpt2.pt", map_location=torch.device(device)), strict=False)
 
 model.to(device)
+model = torch.compile(model, fullgraph=True, backend="cudagraphs")
 
 enc = tiktoken.get_encoding('gpt2')
 
@@ -53,7 +55,7 @@ def inference(input_text, num_return_sequences, max_length):
 
 title = "GPT-2 trained on Shakespeare Plays dataset"
 description = "A simple Gradio interface to generate text from GPT-2 model trained on Shakespeare Plays"
-examples = [["Please put on these earmuffs because I can't you hear.", 2, 20], 
+examples = [["Please put on these earmuffs because I can't hear you.", 2, 20], 
            ["Twin 4-month-olds slept in the shade of the palm tree while the mother tanned in the sun.", 2, 20], 
            ["Happiness can be found in the depths of chocolate pudding.", 2, 20], 
            ["Seek success, but always be prepared for random cats.", 2, 20],