From 4c798129b04955caad1c48405de168ff63a3809c Mon Sep 17 00:00:00 2001 From: Kohya S Date: Sun, 19 May 2024 19:00:32 +0900 Subject: [PATCH] update README --- README.md | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index 04769a4cf..d0f2d65b2 100644 --- a/README.md +++ b/README.md @@ -165,11 +165,14 @@ The majority of scripts is licensed under ASL 2.0 (including codes from Diffuser - Specify the learning rate and dim (rank) for each block. - See [Block-wise learning rates in LoRA](./docs/train_network_README-ja.md#階層別学習率) for details (Japanese only). +- Negative learning rates can now be specified during SDXL model training. PR [#1277](https://github.com/kohya-ss/sd-scripts/pull/1277) Thanks to Cauldrath! + - The model is trained to move away from the training images, so the model is easily collapsed. Use with caution. A value close to 0 is recommended. + - Training scripts can now output training settings to wandb or Tensor Board logs. Specify the `--log_config` option. PR [#1285](https://github.com/kohya-ss/sd-scripts/pull/1285) Thanks to ccharest93, plucked, rockerBOO, and VelocityRa! - Some settings, such as API keys and directory specifications, are not output due to security issues. - The ControlNet training script `train_controlnet.py` for SD1.5/2.x was not working, but it has been fixed. PR [#1284](https://github.com/kohya-ss/sd-scripts/pull/1284) Thanks to sdbds! - + - An option `--disable_mmap_load_safetensors` is added to disable memory mapping when loading the model's .safetensors in SDXL. PR [#1266](https://github.com/kohya-ss/sd-scripts/pull/1266) Thanks to Zovjsra! - It seems that the model file loading is faster in the WSL environment etc. - Available in `sdxl_train.py`, `sdxl_train_network.py`, `sdxl_train_textual_inversion.py`, and `sdxl_train_control_net_lllite.py`. @@ -214,6 +217,9 @@ https://github.com/kohya-ss/sd-scripts/pull/1290) Thanks to frodo821! - ブロックごとに学習率および dim (rank) を指定することができます。 - 詳細は [LoRA の階層別学習率](./docs/train_network_README-ja.md#階層別学習率) をご覧ください。 +- `sdxl_train.py` での SDXL モデル学習時に負の学習率が指定できるようになりました。PR [#1277](https://github.com/kohya-ss/sd-scripts/pull/1277) Cauldrath 氏に感謝します。 + - 学習画像から離れるように学習するため、モデルは容易に崩壊します。注意して使用してください。0 に近い値を推奨します。 + - 各学習スクリプトで学習設定を wandb や Tensor Board などのログに出力できるようになりました。`--log_config` オプションを指定してください。PR [#1285](https://github.com/kohya-ss/sd-scripts/pull/1285) ccharest93 氏、plucked 氏、rockerBOO 氏および VelocityRa 氏に感謝します。 - API キーや各種ディレクトリ指定など、一部の設定はセキュリティ上の問題があるため出力されません。