From 1f831fb61faadeb02a8a618fd52c16420bca060a Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 6 Feb 2025 02:24:13 -0800
Subject: [PATCH] Update rl.py

---
 unsloth/models/rl.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/unsloth/models/rl.py b/unsloth/models/rl.py
index fe1587f56b..c785870306 100644
--- a/unsloth/models/rl.py
+++ b/unsloth/models/rl.py
@@ -308,7 +308,7 @@ def _patch_trl_rl_trainers(trainer_file = "grpo_trainer"):
     sampling_params = sampling_params[0]
     # Replace with our vLLM engine
     sampling_params = \
-        " "*8 + "self.llm = model.vllm_engine; " + \
+        " "*8 + "self.llm = model.vllm_engine; self._last_loaded_step = 0; " + \
         sampling_params # Add spaces
     new_vllm_part = f"\n    if {args}.use_vllm:\n{sampling_params}\n    else:\n"
     __init__ = __init__.replace(vllm_part, new_vllm_part)