Update chat interface logic

app.py

@@ -134,16 +134,24 @@ def answer_query(
 EMB_KEY_PATH       = "staria_keys_embed.json"
 VALUE_TEXT_PATH    = "staria_values.json"
 MODEL_ID           = "JLee0/staria-pdf-chatbot-lora"
+BASE_ID = "MLP-KTLim/llama-3-Korean-Bllossom-8B"
 
-tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, padding_side="right")
+from peft import PeftModel
+
+tokenizer = AutoTokenizer.from_pretrained(BASE_ID)
 tokenizer.pad_token = tokenizer.eos_token
 
-import torch
-model = AutoModelForCausalLM.from_pretrained(
-    MODEL_ID,
+base_model = AutoModelForCausalLM.from_pretrained(
+    BASE_ID,
     load_in_8bit=True,
     device_map="auto"
 )
+model = PeftModel.from_pretrained(
+    base_model,
+    MODEL_ID,
+    device_map="auto"
+)
+
 
 rag_embedder = SentenceTransformer("JLee0/rag-embedder-staria-10epochs")
 SYSTEM_PROMPT = PROMPTS["staria_after"]