From 263c7a5b059d3b0dd8aa528889f3ce5281b0000b Mon Sep 17 00:00:00 2001 From: Amil Dravid <46203730+avdravid@users.noreply.github.com> Date: Thu, 12 Sep 2024 19:15:16 -0700 Subject: [PATCH] Update README.md --- README.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/README.md b/README.md index e04cb34..c802c64 100644 --- a/README.md +++ b/README.md @@ -54,6 +54,8 @@ $ bash train.sh ``` which conducts Dreambooth LoRA fine-tuning by running `train_dreambooth.py` given a folder of identity images. This is based on [PEFT](https://github.com/huggingface/peft/tree/main/examples/lora_dreambooth). Download the folders of identity images from this [link](https://huggingface.co/datasets/wangkua1/w2w-celeba-generated/tree/main). All you need to do is change ``--instance_data_dir="celeba_generated0/0"`` to the identity folder and ``--output_dir="output0"`` to the desired output directory. +After conducting Dreambooth fine-tuning, you can see how we flatten the weights and conduct PCA in ``other/creating_weights_dataset.ipynb``. + ## Acknowledgments Our code is based on implementations from the following repos: