`prepare_model_for_int8_training` has been deprecated for quite some time, with PEFT v0.10.0, it has been removed. Please use `prepare_model_for_kbit_training` instead.

lora_clm_with_additional_tokens.ipynb

@@ -41,8 +41,8 @@
     "    PeftConfig,\n",
     "    PeftModel,\n",
     "    get_peft_model,\n",
-    "    prepare_model_for_int8_training,\n",
     ")\n",
+    "from peft import prepare_model_for_kbit_training as prepare_model_for_int8_training\n",
     "from transformers import (\n",
     "    AutoModelForCausalLM,\n",
     "    AutoTokenizer,\n",