Update to qwen:1.8b

app/rag.py

@@ -15,14 +15,12 @@ class ChatPDF:
     chain = None
 
     def __init__(self):
-        self.model = ChatOllama(model="phi3:mini")
-        self.text_splitter = RecursiveCharacterTextSplitter(chunk_size=1536, chunk_overlap=152)
-        # self.text_splitter = RecursiveCharacterTextSplitter(chunk_size=1024, chunk_overlap=100)
+        #8k context window
+        self.model = ChatOllama(model="qwen:1.8b")
+        self.text_splitter = RecursiveCharacterTextSplitter(chunk_size=128, chunk_overlap=8)
         self.prompt = PromptTemplate.from_template(
             """
-            <s> [INST] You are an assistant for question-answering tasks. Use the following pieces of retrieved context 
-            to answer the question. If you don't know the answer, just say that you don't know. Use three sentences
-             maximum and keep the answer concise. [/INST] </s> 
+            <s> [INST] Use the pieces of context to answer the question concisely.[/INST] </s> 
             [INST] Question: {question} 
             Context: {context} 
             Answer: [/INST]
@@ -38,7 +36,7 @@ class ChatPDF:
         self.retriever = vector_store.as_retriever(
             search_type="similarity_score_threshold",
             search_kwargs={
-                "k": 2,
+                "k": 56,
                 "score_threshold": 0.5,
             },
         )

start_service.sh

@@ -7,7 +7,7 @@ ollama serve &
 sleep 5
 
 # Pull and run <YOUR_MODEL_NAME>
-ollama pull phi3:mini
+ollama pull qwen:1.8b
 
 #
 fastapi run /code/app/main.py --port 7860