2023-01-16 13:31:13 +00:00
# Text generation web UI
2022-12-21 19:49:30 +00:00
2023-01-16 13:07:00 +00:00
A gradio web UI for running large language models like gpt-j-6B, gpt-neo, opt, galactica, and pygmalion on your own computer.
2022-12-21 19:49:30 +00:00
Its goal is to become the [AUTOMATIC1111/stable-diffusion-webui ](https://github.com/AUTOMATIC1111/stable-diffusion-webui ) of text generation.
2022-12-21 16:17:06 +00:00
2023-01-26 19:17:15 +00:00
[[Try it on Google Colab]](https://colab.research.google.com/github/oobabooga/AI-Notebooks/blob/main/Colab-TextGen-GPU.ipynb) -> Updated January 24th, responses are 100x better now < -
2023-01-22 22:29:24 +00:00
2023-01-22 20:28:51 +00:00
| |  |
2023-01-15 02:25:05 +00:00
|:---:|:---:|
2023-01-15 07:27:22 +00:00
| |  |
2022-12-21 17:04:51 +00:00
2023-01-07 21:40:51 +00:00
## Features
* Switch between different models using a dropdown menu.
* Notebook mode that resembles OpenAI's playground.
2023-01-22 04:30:55 +00:00
* Chat mode for conversation and role playing.
2023-01-23 13:02:35 +00:00
* Generate nice HTML output for GPT-4chan.
* Generate Markdown output for [GALACTICA ](https://github.com/paperswithcode/galai ), including LaTeX support.
2023-01-29 06:13:22 +00:00
* Support for [Pygmalion ](https://huggingface.co/models?search=pygmalionai/pygmalion ) and custom characters in JSON or TavernAI Character Card formats ([see the FAQ](https://github.com/oobabooga/text-generation-webui/wiki/Pygmalion-chat-model-FAQ)).
2023-01-23 13:02:35 +00:00
* Stream the text output in real time.
2023-01-07 21:40:51 +00:00
* Load parameter presets from text files.
2023-02-02 14:00:57 +00:00
* Load large models in 8-bit mode ([see here](https://github.com/oobabooga/text-generation-webui/issues/20#issuecomment-1411650652) if you are on Windows).
2023-01-23 13:02:35 +00:00
* Split large models across your GPU(s), CPU, and disk.
2023-01-09 14:11:05 +00:00
* CPU mode.
2023-01-11 19:51:32 +00:00
* Get responses via API.
2023-01-30 12:45:31 +00:00
* Support for extensions ([guide](https://github.com/oobabooga/text-generation-webui/wiki/Extensions)).
2023-01-22 22:29:24 +00:00
* Works on Google Colab ([see the full guide](https://github.com/oobabooga/text-generation-webui/wiki/Running-on-Colab)).
2023-01-07 21:40:51 +00:00
2022-12-21 16:17:06 +00:00
## Installation
2023-01-22 01:54:35 +00:00
Open a terminal and copy and paste these commands one at a time ([install conda](https://docs.conda.io/en/latest/miniconda.html) first if you don't have it already):
2023-01-06 03:05:11 +00:00
2023-01-13 04:29:36 +00:00
```
conda create -n textgen
conda activate textgen
2023-01-26 15:32:21 +00:00
conda install torchvision torchaudio pytorch-cuda=11.7 git -c pytorch -c nvidia
2023-01-20 03:20:35 +00:00
git clone https://github.com/oobabooga/text-generation-webui
cd text-generation-webui
2023-01-22 01:54:35 +00:00
pip install -r requirements.txt
2023-01-20 03:20:35 +00:00
```
2023-01-13 04:27:29 +00:00
2023-01-22 01:54:35 +00:00
The third line assumes that you have an NVIDIA GPU.
* If you have an AMD GPU, you should install the ROCm version of pytorch instead.
2023-01-23 16:41:50 +00:00
* If you are running in CPU mode, you just need the standard pytorch and should replace the third command with this one:
2023-01-06 03:05:11 +00:00
2023-01-13 04:29:36 +00:00
```
2023-01-22 01:54:35 +00:00
conda install pytorch torchvision torchaudio git -c pytorch
2023-01-13 04:29:36 +00:00
```
2022-12-21 16:17:06 +00:00
2023-01-20 00:08:26 +00:00
Once you have completed these steps, you should be able to start the web UI. However, you will first need to download a model.
2023-01-13 04:27:29 +00:00
2022-12-21 17:37:50 +00:00
## Downloading models
2023-01-16 14:19:23 +00:00
Models should be placed under `models/model-name` . For instance, `models/gpt-j-6B` for [GPT-J 6B ](https://huggingface.co/EleutherAI/gpt-j-6B/tree/main ).
2022-12-21 17:37:50 +00:00
2023-01-15 06:35:57 +00:00
#### Hugging Face
2022-12-21 17:37:50 +00:00
2023-01-15 06:35:57 +00:00
[Hugging Face ](https://huggingface.co/models?pipeline_tag=text-generation&sort=downloads ) is the main place to download models. These are some noteworthy examples:
2022-12-21 17:37:50 +00:00
2023-01-16 14:19:23 +00:00
* [GPT-J 6B ](https://huggingface.co/EleutherAI/gpt-j-6B/tree/main )
* [GPT-Neo ](https://huggingface.co/models?pipeline_tag=text-generation&sort=downloads&search=eleutherai+%2F+gpt-neo )
2023-01-07 22:19:55 +00:00
* [OPT ](https://huggingface.co/models?search=facebook/opt )
* [GALACTICA ](https://huggingface.co/models?search=facebook/galactica )
2023-01-06 03:13:26 +00:00
* [\*-Erebus ](https://huggingface.co/models?search=erebus )
2023-01-13 12:33:38 +00:00
* [Pygmalion ](https://huggingface.co/models?search=pygmalion )
2023-01-06 03:13:26 +00:00
2023-01-23 13:02:35 +00:00
You can automatically download a model from HF using the script `download-model.py` . Its usage is very simple:
2023-01-06 23:00:58 +00:00
python download-model.py organization/model
For instance:
python download-model.py facebook/opt-1.3b
2023-01-23 13:02:35 +00:00
If you want to download a model manually, note that all you need are the json, txt, and pytorch\*.bin files. The remaining files are not necessary.
2023-01-09 19:19:57 +00:00
#### GPT-4chan
2022-12-21 17:37:50 +00:00
2023-01-09 19:19:57 +00:00
[GPT-4chan ](https://huggingface.co/ykilcher/gpt-4chan ) has been shut down from Hugging Face, so you need to download it elsewhere. You have two options:
2022-12-21 17:37:50 +00:00
* Torrent: [16-bit ](https://archive.org/details/gpt4chan_model_float16 ) / [32-bit ](https://archive.org/details/gpt4chan_model )
* Direct download: [16-bit ](https://theswissbay.ch/pdf/_notpdf_/gpt4chan_model_float16/ ) / [32-bit ](https://theswissbay.ch/pdf/_notpdf_/gpt4chan_model/ )
2023-01-13 04:39:09 +00:00
The 32-bit version is only relevant if you intend to run the model in CPU mode. Otherwise, you should use the 16-bit version.
2023-01-12 02:00:31 +00:00
2023-01-13 04:37:12 +00:00
After downloading the model, follow these steps:
1. Place the files under `models/gpt4chan_model_float16` or `models/gpt4chan_model` .
2023-01-16 14:19:23 +00:00
2. Place GPT-J 6B's config.json file in that same folder: [config.json ](https://huggingface.co/EleutherAI/gpt-j-6B/raw/main/config.json ).
3. Download GPT-J 6B under `models/gpt-j-6B` :
2023-01-12 02:01:50 +00:00
2023-01-12 02:04:06 +00:00
```
python download-model.py EleutherAI/gpt-j-6B
```
2023-01-08 02:13:43 +00:00
2023-01-16 14:19:23 +00:00
You don't really need all of GPT-J 6B's files, just the tokenizer files, but you might as well download the whole thing. Those files will be automatically detected when you attempt to load GPT-4chan.
2023-01-13 04:27:29 +00:00
2023-01-09 21:58:24 +00:00
#### Converting to pytorch (optional)
2022-12-21 19:49:30 +00:00
2023-01-23 13:05:25 +00:00
The script `convert-to-torch.py` allows you to convert models to .pt format, which is sometimes 10x faster to load to the GPU:
2022-12-21 19:49:30 +00:00
2023-01-08 17:37:43 +00:00
python convert-to-torch.py models/model-name
2022-12-21 19:49:30 +00:00
2023-01-16 13:07:00 +00:00
The output model will be saved to `torch-dumps/model-name.pt` . When you load a new model, the web UI first looks for this .pt file; if it is not found, it loads the model as usual from `models/model-name` .
2022-12-21 19:49:30 +00:00
2023-01-16 13:07:00 +00:00
## Starting the web UI
2022-12-21 16:17:06 +00:00
conda activate textgen
python server.py
2023-01-06 23:00:58 +00:00
Then browse to
`http://localhost:7860/?__theme=dark`
2023-01-25 22:37:44 +00:00
2023-01-06 23:07:58 +00:00
Optionally, you can use the following command-line flags:
2023-01-06 23:00:58 +00:00
2023-01-15 18:40:03 +00:00
| Flag | Description |
|-------------|-------------|
2023-01-15 18:41:36 +00:00
| `-h` , `--help` | show this help message and exit |
| `--model MODEL` | Name of the model to load by default. |
2023-01-16 13:07:00 +00:00
| `--notebook` | Launch the web UI in notebook mode, where the output is written to the same text box as the input. |
| `--chat` | Launch the web UI in chat mode.|
2023-01-25 22:43:30 +00:00
| `--cai-chat` | Launch the web UI in chat mode with a style similar to Character.AI's. If the file `img_bot.png` or `img_bot.jpg` exists in the same folder as server.py, this image will be used as the bot's profile picture. Similarly, `img_me.png` or `img_me.jpg` will be used as your profile picture. |
2023-01-15 18:41:36 +00:00
| `--cpu` | Use the CPU to generate text.|
| `--load-in-8bit` | Load the model with 8-bit precision.|
2023-01-19 14:09:24 +00:00
| `--auto-devices` | Automatically split the model across the available GPU(s) and CPU.|
| `--disk` | If the model is too large for your GPU(s) and CPU combined, send the remaining layers to the disk. |
2023-01-21 05:48:06 +00:00
| `--disk-cache-dir DISK_CACHE_DIR` | Directory to save the disk cache to. Defaults to `cache/` . |
2023-01-21 03:33:41 +00:00
| `--gpu-memory GPU_MEMORY` | Maximum GPU memory in GiB to allocate. This is useful if you get out of memory errors while trying to generate text. Must be an integer number. |
2023-01-21 06:05:55 +00:00
| `--cpu-memory CPU_MEMORY` | Maximum CPU memory in GiB to allocate for offloaded weights. Must be an integer number. Defaults to 99.|
2023-01-22 19:19:11 +00:00
| `--no-stream` | Don't stream the text output in real time. This improves the text generation performance.|
2023-01-16 19:35:45 +00:00
| `--settings SETTINGS_FILE` | Load the default interface settings from this json file. See `settings-template.json` for an example.|
2023-01-27 03:40:39 +00:00
| `--extensions EXTENSIONS` | The list of extensions to load. If you want to load more than one extension, write the names separated by commas and between quotation marks, "like,this". |
2023-01-21 02:45:16 +00:00
| `--listen` | Make the web UI reachable from your local network.|
2023-01-29 05:54:36 +00:00
| `--listen-port LISTEN_PORT` | The listening port that the server will use. |
2023-01-19 20:31:29 +00:00
| `--share` | Create a public URL. This is useful for running the web UI on Google Colab or similar. |
2023-01-26 05:12:53 +00:00
| `--verbose` | Print the prompts to the terminal. |
2023-01-09 13:58:46 +00:00
2023-01-19 14:25:17 +00:00
Out of memory errors? [Check this guide ](https://github.com/oobabooga/text-generation-webui/wiki/Low-VRAM-guide ).
2023-01-06 04:33:21 +00:00
## Presets
Inference settings presets can be created under `presets/` as text files. These files are detected automatically at startup.
2023-01-23 23:44:27 +00:00
By default, 10 presets by NovelAI and KoboldAI are included. These were selected out of a sample of 43 presets after applying a K-Means clustering algorithm and selecting the elements closest to the average of each cluster.
2023-01-09 21:12:41 +00:00
## System requirements
2023-01-09 21:18:35 +00:00
Check the [wiki ](https://github.com/oobabooga/text-generation-webui/wiki/System-requirements ) for some examples of VRAM and RAM usage in both GPU and CPU mode.
2023-01-09 21:12:41 +00:00
2022-12-21 19:52:23 +00:00
## Contributing
2023-01-16 13:07:00 +00:00
Pull requests, suggestions, and issue reports are welcome.
2023-01-09 19:15:54 +00:00
2023-01-29 17:37:05 +00:00
Before reporting a bug, make sure that you have created a conda environment and installed the dependencies exactly as in the *Installation* section above.
2023-01-15 06:00:59 +00:00
## Credits
2023-01-09 19:15:54 +00:00
2023-01-15 07:29:19 +00:00
- NovelAI and KoboldAI presets: https://github.com/KoboldAI/KoboldAI-Client/wiki/Settings-Presets
2023-01-25 13:17:55 +00:00
- Pygmalion preset, code for early stopping in chat mode: https://github.com/PygmalionAI/gradio-ui/
2023-01-16 14:23:45 +00:00
- Verbose preset: Anonymous 4chan user.
2023-01-22 03:02:46 +00:00
- Gradio dropdown menu refresh button: https://github.com/AUTOMATIC1111/stable-diffusion-webui
