From 334f02cd017ba5ac30aef33feb93b5a4eedb8f29 Mon Sep 17 00:00:00 2001
From: Wing Lian <wing.lian@gmail.com>
Date: Fri, 19 Jan 2024 23:32:24 -0500
Subject: [PATCH] warn about not pre-processing

---
 src/axolotl/utils/trainer.py | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/src/axolotl/utils/trainer.py b/src/axolotl/utils/trainer.py
index 9a9eeab4b9..6d2f08c8ec 100644
--- a/src/axolotl/utils/trainer.py
+++ b/src/axolotl/utils/trainer.py
@@ -107,6 +107,10 @@ def drop_long_seq(sample, sequence_len=2048):
 
 
 def process_datasets_for_packing(cfg, train_dataset, eval_dataset, tokenizer):
+    if cfg.is_preprocess:
+        LOG.warning(
+            "Processing datasets during training can lead to VRAM instability. Please pre-process your dataset"
+        )
     drop_long = partial(drop_long_seq, sequence_len=cfg.sequence_len)
     with zero_first(is_main_process()):
         if cfg.group_by_length: