diff --git a/LLMs/training/sft_mistral_7b_garole/README.md b/LLMs/training/sft_mistral_7b_garole/README.md
new file mode 100644
index 0000000..a4328fc
--- /dev/null
+++ b/LLMs/training/sft_mistral_7b_garole/README.md
@@ -0,0 +1,3 @@
+# Garole Optimizer
+
+With [garole](https://huggingface.co/papers/2403.03507), you could pre-train the 7B LLM model on consumer type hardwares.