From a825f7dedcaa2069d0d014ed31914a031b5bd658 Mon Sep 17 00:00:00 2001
From: MaoXun <41497489+mao-code@users.noreply.github.com>
Date: Fri, 22 Mar 2024 20:48:10 +0800
Subject: [PATCH] Update rewards.py

add comment in batch_size variable in llava reward to make it clear
---
 ddpo_pytorch/rewards.py | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/ddpo_pytorch/rewards.py b/ddpo_pytorch/rewards.py
index 9409a2f..b3aca11 100644
--- a/ddpo_pytorch/rewards.py
+++ b/ddpo_pytorch/rewards.py
@@ -56,7 +56,10 @@ def llava_strict_satisfaction():
     from io import BytesIO
     import pickle
 
+    # this batch size is related to sample_batch_size
+    # set it with the sample_batch_size in your comfig   
     batch_size = 4
+    
     url = "http://127.0.0.1:8085"
     sess = requests.Session()
     retries = Retry(
@@ -124,7 +127,10 @@ def llava_bertscore():
     from io import BytesIO
     import pickle
 
+    # this batch size is related to sample_batch_size
+    # set it with the sample_batch_size in your comfig
     batch_size = 16
+    
     url = "http://127.0.0.1:8085"
     sess = requests.Session()
     retries = Retry(