remote llm

app.py

@@ -1,9 +1,9 @@
 import streamlit as st
 from evaluations.repo_evaluations import evaluate
-from evaluations.models import LocalLLM
+from evaluations.models import RemoteLLM
 import requests
 
-# model = LocalLLM("codellama/CodeLlama-7b-Instruct-hf")
+model = RemoteLLM("codellama/CodeLlama-7b-Instruct-hf")
 
 st.write("\n")
 st.write("Welcome to the online reproducibility evaluation tool!")
@@ -14,7 +14,7 @@ repo_link = st.text_input("Github repository link:", value="", type="default", h
 
 if (repo_link):
     verbose = 4 if checkbox else 3
-    evaluate(llm=None, verbose=verbose, repo_url=repo_link)
+    evaluate(llm=model, verbose=verbose, repo_url=repo_link)
 
 with st.form("my_form"):
     st.write("Notice something wrong? Please tell us so we can improve.")

evaluations/models.py

@@ -23,13 +23,10 @@ class LocalLLM():
     return res
 
 class RemoteLLM():
-  def __init__(self):
+  def __init__(self, model_name):
     token = os.getenv("hfToken")
-    API_URL = "https://api-inference.huggingface.co/models/openlm-research/open_llama_3b_v2"
-    headers = {"Authorization": f"Bearer {token}", "x-wait-for-model": "true"}
-
     self.client = InferenceClient(
-        "meta-llama/Llama-3.1-8B-Instruct",
+        model_name,
         token=token,
     )