From ad81b5d838cf593a741230bb9ed8e93637c32702 Mon Sep 17 00:00:00 2001
From: Ziteng Zhang <87107332+Jasonzzt@users.noreply.github.com>
Date: Fri, 10 Nov 2023 15:19:25 +0800
Subject: [PATCH] Update qlora README.md (#9422)

---
 python/llm/example/CPU/QLoRA-FineTuning/README.md | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/python/llm/example/CPU/QLoRA-FineTuning/README.md b/python/llm/example/CPU/QLoRA-FineTuning/README.md
index 5acd255b..caa81a70 100644
--- a/python/llm/example/CPU/QLoRA-FineTuning/README.md
+++ b/python/llm/example/CPU/QLoRA-FineTuning/README.md
@@ -20,7 +20,12 @@ pip install datasets
 
 ### 2. Finetune model
 
+If the machine memory is not enough, you can try to set `use_gradient_checkpointing=True` in [here](https://github.com/intel-analytics/BigDL/blob/1747ffe60019567482b6976a24b05079274e7fc8/python/llm/example/CPU/QLoRA-FineTuning/qlora_finetuning_cpu.py#L53C6-L53C6).
+
+And remember to use `bigdl-llm-init` before you start finetuning, which can accelerate the job.
+
 ```
+source bigdl-llm-init -t
 python ./qlora_finetuning_cpu.py --repo-id-or-model-path REPO_ID_OR_MODEL_PATH --dataset DATASET
 ```
 
@@ -73,4 +78,4 @@ Inference time: 2.864234209060669 s
 “QLoRA fine-tuning using BigDL-LLM 4bit optimizations on Intel CPU is Efficient and convenient” ->:
 -------------------- Output --------------------
 “QLoRA fine-tuning using BigDL-LLM 4bit optimizations on Intel CPU is Efficient and convenient” ->: ['bigdl'] ['deep-learning'] ['distributed-computing'] ['intel'] ['optimization'] ['training'] ['training-speed']
-```
\ No newline at end of file
+```