From 0ecad4b178f12edd0c174136ab2f81715e0c154a Mon Sep 17 00:00:00 2001
From: Victor Nogueira <felladrin@gmail.com>
Date: Thu, 13 Mar 2025 13:20:18 +0100
Subject: [PATCH] [dataset] fix ultrachat_200k dataset (#7259)

The `HuggingFaceH4/ultrachat_200k` dataset doesn't contain the default "train" split. The correct split is "train_sft".
---
 data/dataset_info.json | 1 +
 1 file changed, 1 insertion(+)

diff --git a/data/dataset_info.json b/data/dataset_info.json
index 7aaee14f..aba2c6c5 100644
--- a/data/dataset_info.json
+++ b/data/dataset_info.json
@@ -232,6 +232,7 @@
   "ultrachat_200k": {
     "hf_hub_url": "HuggingFaceH4/ultrachat_200k",
     "ms_hub_url": "AI-ModelScope/ultrachat_200k",
+    "split": "train_sft",
     "formatting": "sharegpt",
     "columns": {
       "messages": "messages"