Upload 2 files

app.py

@@ -1,9 +1,10 @@
 import gradio as gr
 from llama_cpp import Llama
 
-llm = Llama(model_path="vicuna-AlekseyKorshuk-7B-GPTQ-4bit-128g.GGML.bin", n_ctx=2048)
+
 
 def generate_text(prompt):
+    llm = Llama(model_path="vicuna-AlekseyKorshuk-7B-GPTQ-4bit-128g.GGML.bin", n_ctx=2048)
     output = llm(prompt, max_tokens=468, temperature=0.1, top_p=0.5, echo=False, stop=["#"])
     text = output['choices'][0]['text']
     return text

requirements.txt

@@ -1 +1 @@
-llama-cpp-python
+llama-cpp-python