zeroGPU try

app.py

@@ -1,6 +1,7 @@
 import os
 from time import time_ns
 
+import spaces
 import gradio as gr
 import torch
 from huggingface_hub import Repository
@@ -14,6 +15,7 @@ from kgen.metainfo import SPECIAL, TARGET
 MODEL_PATH = "KBlueLeaf/DanTagGen"
 
 
+@spaces.GPU
 @torch.no_grad()
 def get_result(
     text_model: LlamaForCausalLM,
@@ -30,6 +32,7 @@ def get_result(
     escape_bracket: bool = False,
     temperature: float = 1.35,
 ):
+    text_model.eval().half().cuda()
     start = time_ns()
     print("=" * 50, "\n")
     # Use LLM to predict possible summary

requirements.txt

@@ -3,4 +3,5 @@ transformers
 llama-cpp-python
 gradio
 requests
-sentencepiece
+sentencepiece
+spaces