updating

Author: chasecadet

dockerfiles/llm_predictor/src/model.py

@@ -46,10 +46,8 @@ def predict(self, request: dict, headers: dict) -> dict:
         instruction = data["instruction"]
         context = data["context"]
         query = data["input"]
-
         logger.info(f"Received instruction: {instruction}")
         logger.info(f"Received input: {query}")
-
         # generation parameters
         temperature = data.get("temperature", .2)
         max_tokens = data.get("max_tokens", 200)

dockerfiles/llm_transformer/dockerfile

@@ -1,10 +1,17 @@
 FROM python:3.10.6-slim
+
 COPY src/requirements.txt /src/requirements.txt
+
 RUN DEBIAN_FRONTEND=noninteractive
+
 RUN apt update && apt install build-essential -y
+
 RUN pip install --upgrade pip && pip install -r /src/requirements.txt
+
 COPY src/model.py /src/model.py
+
 WORKDIR /src
+
 RUN chmod -R 777 /src
-ENTRYPOINT ["python", "model.py"]
 
+ENTRYPOINT ["python", "model.py"]

dockerfiles/llm_transformer/src/model.py

@@ -39,8 +39,8 @@ def _build_vectorstore_url(self):
         deployment_name = self.vectorstore_name
         model_name = deployment_name
         # Build the vectorstore URL
-        svc = f'{deployment_name}-predictor-default.{namespace}.{domain_name}'
-        url = f"https://{svc}/v1/models/{model_name}:predict"
+        svc = f'{deployment_name}-predictor.{namespace}.{domain_name}'
+        url = f"http://{svc}/v1/models/{model_name}:predict"
         return url
 
     @property
@@ -51,16 +51,11 @@ def _http_client(self):
         return self._http_client_instance
 
     def preprocess(self, request: dict, headers: dict) -> dict:
-        self.authorization = headers["authorization"]
-
         data = request["instances"][0]
         query = data["input"]
-
         logger.info(f"Received question: {query}")
-
         num_docs = data.get("num_docs", 4)
         context = data.get("context", None)
-
         if context:
             logger.info(f"Received context: {context}")
             logger.info(f"Skipping retrieval step...")
@@ -71,7 +66,7 @@ def preprocess(self, request: dict, headers: dict) -> dict:
                 f"Receiving relevant docs from: {self.vectorstore_url}")
 
             response = requests.post(
-                self.vectorstore_url, json=payload, headers=headers,
+                self.vectorstore_url, json=payload,
                 verify=False)
             response = json.loads(response.text)        
             context = "\n".join(response["predictions"])