From 4d473894fddf4778a5cbf8449de1b617c5f44983 Mon Sep 17 00:00:00 2001
From: Mark Mueller <mark.mueller@inf.ethz.ch>
Date: Thu, 8 Feb 2024 17:56:18 +0100
Subject: [PATCH] Slim Orca data parsing

Former-commit-id: ca57d27c39d4e7bc3dd7c3207a23d23d2cbd446b
---
 src/llmtuner/data/aligner.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/src/llmtuner/data/aligner.py b/src/llmtuner/data/aligner.py
index d4f281bc..cd3a7ea4 100644
--- a/src/llmtuner/data/aligner.py
+++ b/src/llmtuner/data/aligner.py
@@ -60,6 +60,7 @@ def convert_sharegpt(examples: Dict[str, List[Any]], dataset_attr: "DatasetAttr"
             if dataset_attr.system_tag and message[dataset_attr.role_tag] == dataset_attr.system_tag:
                 outputs["system"].append(message[dataset_attr.content_tag])
                 n_sys = 1
+                continue
 
             if (turn_idx - n_sys) % 2 == 0:
                 accept_tags = [dataset_attr.user_tag, dataset_attr.observation_tag]