From d76960d1e562e42236cce728e1c980fad7f17ee3 Mon Sep 17 00:00:00 2001 From: wy-luke Date: Thu, 21 Sep 2023 21:59:02 +0800 Subject: [PATCH] docs: add re-run, troubleshooting, models --- README.md | 77 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 77 insertions(+) diff --git a/README.md b/README.md index c8fbf84..3fee9aa 100644 --- a/README.md +++ b/README.md @@ -32,8 +32,81 @@ Theoretically, it should work on both Apple Silicon and Intel CPU with any versi 5. Open your web browser and enter `http://127.0.0.1:7860` (the underlined part in the image above) to access the Stable Diffusion web UI. + > Note 1: Normally, it will automatically open the web browser. + + > **Note 2: Don't close the terminal app while using Stable Diffusion web UI, just keep it running background. If you accidentally close the terminal, see step 7 below.** + 6. If the installation fails, you will be prompted whether to retry. Type `y` or just simply **press the Enter key** to try again. Type `n` to exit. +## Re-run + +Use the command below to **re-run** Stable Diffusion web UI after the installation sucessfully completes. + +```bash +/bin/bash $HOME/stable-diffusion-webui/webui.sh +``` + +If you move the `stable-diffusion-webui` folder to another path, you need to change the command above accordingly. Besides, you need always append `/webui.sh` in the end. + +```bash +/bin/bash /your/path/of/sd/webui.sh +``` + +## Troubleshooting + +### \*\*\*\* not implemented for 'Half' + +Most likely, this is due to poor compatibility with AMD graphics cards. + +The solution is simple. Run the command below, then execute the **re-run** command above again to start and use SD. + +```bash +echo 'export COMMANDLINE_ARGS="--upcast-sampling --no-half-vae --use-cpu interrogate --precision full --no-half --skip-torch-cuda-test"' > $HOME/stable-diffusion-webui/webui-user.sh +``` + +### MPS backend out of memory + +This issue is caused by insufficient memory. You can reduce the memory requirement by adding `PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.7` before the **re-run** command to make SD run correctly. + +However, this will result in longer processing time to generate a image. + +Replace the **re-run** command with: + +```bash +PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.7 /bin/bash $HOME/stable-diffusion-webui/webui.sh +``` + +If you still encounter this issue, you can try further reducing the value, such as: + +```bash +PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.5 /bin/bash $HOME/stable-diffusion-webui/webui.sh +``` + +## Download models + +Place the models downloaded to `stable-diffusion-webui/models/Stable-diffusion` folder. Then refresh the Stable Diffusion web UI page in the browser and you will see it in the list. + +1. Some popular official Stable Diffusion 1.x models: + + - [Stable DIffusion 1.4](https://huggingface.co/CompVis/stable-diffusion-v-1-4-original) ([sd-v1-4.ckpt](https://huggingface.co/CompVis/stable-diffusion-v-1-4-original/resolve/main/sd-v1-4.ckpt)) + - [Stable Diffusion 1.5](https://huggingface.co/runwayml/stable-diffusion-v1-5) ([v1-5-pruned-emaonly.ckpt](https://huggingface.co/runwayml/stable-diffusion-v1-5/resolve/main/v1-5-pruned-emaonly.ckpt)) + - [Stable Diffusion 1.5 Inpainting](https://huggingface.co/runwayml/stable-diffusion-inpainting) ([sd-v1-5-inpainting.ckpt](https://huggingface.co/runwayml/stable-diffusion-inpainting/resolve/main/sd-v1-5-inpainting.ckpt)) + +2. Stable Diffusion 2.0 and 2.1 models: + + - [Stable Diffusion 2.0](https://huggingface.co/stabilityai/stable-diffusion-2) ([768-v-ema.ckpt](https://huggingface.co/stabilityai/stable-diffusion-2/resolve/main/768-v-ema.ckpt)) + - [Stable Diffusion 2.1](https://huggingface.co/stabilityai/stable-diffusion-2-1) ([v2-1_768-ema-pruned.ckpt](https://huggingface.co/stabilityai/stable-diffusion-2-1/resolve/main/v2-1_768-ema-pruned.ckpt)) + + These models require both a model and a configuration file, and image width & height will need to be set to 768 or higher when generating images. + + For the configuration file, hold down the option key on the keyboard and click [here](https://github.com/Stability-AI/stablediffusion/raw/main/configs/stable-diffusion/v2-inference-v.yaml) to download `v2-inference-v.yaml` (it may download as `v2-inference-v.yaml.yml`). In Finder select that file then go to the menu and select `File` \> `Get Info`. In the window that appears select the filename and change it to the filename of the model, except with the file extension `.yaml` instead of `.ckpt`, press return on the keyboard (confirm changing the file extension if prompted), and place it in the same folder as the model (e.g. if you downloaded the `768-v-ema.ckpt` model, rename it to `768-v-ema.yaml` and put it in `stable-diffusion-webui/models/Stable-diffusion` along with the model). + + - [Stable Diffusion 2.0 depth model](https://huggingface.co/stabilityai/stable-diffusion-2-depth) ([512-depth-ema.ckpt](https://huggingface.co/stabilityai/stable-diffusion-2-depth/resolve/main/512-depth-ema.ckpt)). + + Download the `v2-midas-inference.yaml` configuration file by holding down option on the keyboard and clicking [here](https://github.com/Stability-AI/stablediffusion/raw/main/configs/stable-diffusion/v2-midas-inference.yaml), then rename it with the `.yaml` extension in the same way as mentioned above and put it in `stable-diffusion-webui/models/Stable-diffusion` along with the model. Note that this model works at image dimensions of 512 width/height or higher instead of 768. + +3. You could also download all kinds of models in [civitai](https://civitai.com/), a great site! + ## Features 1. Use [Homebrew](https://brew.sh/) to install required dependencies. If you already have it installed, it will be used automatically. If not, it will be installed for you, but without modifying your system environment – your system will remain unchanged. @@ -69,3 +142,7 @@ Theoretically, it should work on both Apple Silicon and Intel CPU with any versi /opt/homebrew/bin/micromamba shell init -s bash -p ~/micromamba source ~/.bashrc ``` + +## Credit + +Thanks to [AUTOMATIC1111](https://github.com/AUTOMATIC1111) and the [stable-diffusion-webui](https://github.com/AUTOMATIC1111/stable-diffusion-webui).