fixes

app.py

@@ -9,7 +9,8 @@ model = TokenBasedLanguageModel()
 m = model.to(device)
 
 print("Loading checkpoint from file")
-checkpoint = torch.load("improved-v5.bin")
+# second param is to work on huggingface cpu
+checkpoint = torch.load("improved-v5.bin", map_location=torch.device('cpu'))
 model.load_state_dict(checkpoint["model_state_dict"])
 print("State restored")
 
@@ -44,4 +45,5 @@ demo = gr.Interface(generate_llm,
                     inputs=[gr.TextArea(placeholder = "In the midst of chaos."), gr.Number(value = 512, maximum = 2048, minimum = 1, step = 1, label = "Max tokens"), gr.Checkbox(label = "Show probs, 10x slower")],
                     outputs=[gr.TextArea(label = "Output"), gr.Text(placeholder = "tok/s and other stats", label = "Stats"), gr.TextArea(label = "Probability stats")])
 
-demo.launch(share = True)
+if __name__ == "__main__":
+    demo.launch(share = True)