From b37ddf97783da826454aec2662329ef0ea38dc4c Mon Sep 17 00:00:00 2001
From: Wing Lian <wing@axolotl.ai>
Date: Sun, 6 Jul 2025 21:55:09 -0400
Subject: [PATCH] don't use tokenizer parallelism when using packing (#2862)
 [skip ci]

---
 src/axolotl/utils/trainer.py | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/src/axolotl/utils/trainer.py b/src/axolotl/utils/trainer.py
index 278fbed5b..06853451c 100644
--- a/src/axolotl/utils/trainer.py
+++ b/src/axolotl/utils/trainer.py
@@ -609,6 +609,9 @@ def prepare_opinionated_env(cfg):
     if cfg.qlora_sharded_model_loading:
         # model loading is forked after the tokenizer
         os.environ["TOKENIZERS_PARALLELISM"] = "false"
+    if cfg.sample_packing:
+        # multipack parallel packing sampler defaults to using fork
+        os.environ["TOKENIZERS_PARALLELISM"] = "false"
 
 
 def setup_trainer(