From 80c0957beaade2f99cb9532011ecd3094270100b Mon Sep 17 00:00:00 2001
From: Andrew Lapp <andrew@rew.la>
Date: Wed, 24 Jan 2024 17:25:07 -0600
Subject: [PATCH] Bump vLLM version

---
 docs/reference/vllm.md | 5 -----
 outlines/serve/vllm.py | 2 +-
 pyproject.toml         | 2 +-
 3 files changed, 2 insertions(+), 7 deletions(-)

diff --git a/docs/reference/vllm.md b/docs/reference/vllm.md
index 3e51f1316..063699a6d 100644
--- a/docs/reference/vllm.md
+++ b/docs/reference/vllm.md
@@ -4,13 +4,8 @@ Outlines can be deployed as an LLM service using the vLLM inference engine and a
 
 ```bash
 pip install outlines[serve]
-pip install "pydantic>=2.0"
 ```
 
-!!! Warning
-
-    Updating Pydantic to v2 after the installation is necessary.
-
 You can then start the server with:
 
 ```bash
diff --git a/outlines/serve/vllm.py b/outlines/serve/vllm.py
index bbf0a50c3..ee9c7000e 100644
--- a/outlines/serve/vllm.py
+++ b/outlines/serve/vllm.py
@@ -51,7 +51,7 @@ def __init__(self, regex_string, llm):
             An instance of `vllm.LLM`
 
         """
-        tokenizer = self.adapt_tokenizer(llm.tokenizer)
+        tokenizer = self.adapt_tokenizer(llm.tokenizer.tokenizer)
 
         fsm = RegexFSM(regex_string, tokenizer)
         self.fsm = fsm
diff --git a/pyproject.toml b/pyproject.toml
index ee6067545..1d8a04ba2 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -60,7 +60,7 @@ test = [
     "huggingface_hub"
 ]
 serve = [
-    "vllm>=0.2.6",
+    "vllm>=0.3.0",
     "ray==2.9.0",
     "uvicorn",
     "fastapi"