Update app.py

app.py

@@ -13,7 +13,7 @@ from dl_hf_model import dl_hf_model
 from loguru import logger
 
 
-URL = "https://huggingface.co/s3nh/LLaMA-2-7B-32K-GGML/resolve/main/LLaMA-2-7B-32K.ggmlv3.q4_1.bin"  # 4.05G
+URL = "https://huggingface.co/s3nh/LLaMA-2-7B-32K-GGML/resolve/main/LLaMA-2-7B-32K.ggmlv3.q8_0.bin"  # 4.05G
 
 _ = (
     "golay" in platform.node()
@@ -24,7 +24,7 @@ _ = (
 )
 
 if _:
-    url = "https://huggingface.co/s3nh/LLaMA-2-7B-32K-GGML/resolve/main/LLaMA-2-7B-32K.ggmlv3.q4_1.bin"  # 2.87G
+    url = "https://huggingface.co/s3nh/LLaMA-2-7B-32K-GGML/resolve/main/LLaMA-2-7B-32K.ggmlv3.q8_0.bin"  # 2.87G
 
 
 prompt_template = """Below is an instruction that describes a task. Write a response that appropriately completes the request.
@@ -84,6 +84,9 @@ USER: {question}
 ASSISTANT: """
 
 
+prompt_template = """<|prompt|>:{question}</s>
+<|answer|>:"""
+
 _ = [elm for elm in prompt_template.splitlines() if elm.strip()]
 stop_string = [elm.split(":")[0] + ":" for elm in _][-2]