2023-01-16 08:31:13 -05:00
# Text generation web UI
2022-12-21 14:49:30 -05:00
2023-03-31 16:49:11 -04:00
A gradio web UI for running Large Language Models like LLaMA, llama.cpp, GPT-J, OPT, and GALACTICA.
2022-12-21 14:49:30 -05:00
Its goal is to become the [AUTOMATIC1111/stable-diffusion-webui ](https://github.com/AUTOMATIC1111/stable-diffusion-webui ) of text generation.
2022-12-21 11:17:06 -05:00
2023-02-05 21:14:25 -05:00
[[Try it on Google Colab]](https://colab.research.google.com/github/oobabooga/AI-Notebooks/blob/main/Colab-TextGen-GPU.ipynb)
2023-01-22 17:29:24 -05:00
2023-01-22 15:28:51 -05:00
| |  |
2023-01-14 21:25:05 -05:00
|:---:|:---:|
2023-01-15 02:27:22 -05:00
| |  |
2022-12-21 12:04:51 -05:00
2023-01-07 16:40:51 -05:00
## Features
2023-04-01 13:15:19 -04:00
* Dropdown menu for switching between models
* Notebook mode that resembles OpenAI's playground
* Chat mode for conversation and role playing
* Nice HTML output for GPT-4chan
* Markdown output for [GALACTICA ](https://github.com/paperswithcode/galai ), including LaTeX rendering
2023-04-02 19:50:06 -04:00
* [Custom chat characters ](https://github.com/oobabooga/text-generation-webui/wiki/Custom-chat-characters )
2023-04-01 13:15:19 -04:00
* Advanced chat features (send images, get audio responses with TTS)
* Very efficient text streaming
* Parameter presets
* 8-bit mode
* Layers splitting across GPU(s), CPU, and disk
* CPU mode
* [FlexGen ](https://github.com/oobabooga/text-generation-webui/wiki/FlexGen )
* [DeepSpeed ZeRO-3 ](https://github.com/oobabooga/text-generation-webui/wiki/DeepSpeed )
* API [with ](https://github.com/oobabooga/text-generation-webui/blob/main/api-example-streaming.py ) streaming and [without ](https://github.com/oobabooga/text-generation-webui/blob/main/api-example.py ) streaming
* [LLaMA model, including 4-bit GPTQ ](https://github.com/oobabooga/text-generation-webui/wiki/LLaMA-model )
* [llama.cpp ](https://github.com/oobabooga/text-generation-webui/wiki/llama.cpp-models ) ** \*NEW!\***
* [RWKV model ](https://github.com/oobabooga/text-generation-webui/wiki/RWKV-model )
* [LoRa (loading and training) ](https://github.com/oobabooga/text-generation-webui/wiki/Using-LoRAs )
* Softprompts
* [Extensions ](https://github.com/oobabooga/text-generation-webui/wiki/Extensions )
* [Google Colab ](https://github.com/oobabooga/text-generation-webui/wiki/Running-on-Colab )
2023-01-07 16:40:51 -05:00
2023-03-17 17:51:07 -04:00
## Installation
2022-12-21 11:17:06 -05:00
2023-03-29 01:11:51 -04:00
### One-click installers
[oobabooga-windows.zip ](https://github.com/oobabooga/text-generation-webui/releases/download/installers/oobabooga-windows.zip )
Just download the zip above, extract it, and double click on "install". The web UI and all its dependencies will be installed in the same folder.
* To download a model, double click on "download-model"
* To start the web UI, double click on "start-webui"
Source codes: https://github.com/oobabooga/one-click-installers
> **Note**
>
> Thanks to [@jllllll](https://github.com/jllllll) and [@ClayShoaf](https://github.com/ClayShoaf), the Windows 1-click installer now sets up 8-bit and 4-bit requirements out of the box. No additional installation steps are necessary.
> **Note**
>
> There is no need to run the installer as admin.
### Manual installation using Conda
2023-03-29 01:22:19 -04:00
Recommended if you have some experience with the command-line.
2023-03-17 17:51:07 -04:00
2023-04-02 20:12:41 -04:00
On Windows, I additionally recommend carrying out the installation on WSL instead of the base system: [WSL installation guide ](https://github.com/oobabooga/text-generation-webui/wiki/WSL-installation-guide ).
2023-03-17 17:51:07 -04:00
2023-04-01 13:31:10 -04:00
0. Install Conda
2023-03-29 01:24:11 -04:00
2023-04-01 13:15:19 -04:00
https://docs.conda.io/en/latest/miniconda.html
2023-03-17 17:51:07 -04:00
2023-03-17 19:25:42 -04:00
On Linux or WSL, it can be automatically installed with these two commands:
2023-03-17 18:52:22 -04:00
```
curl -sL "https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh" > "Miniconda3.sh"
bash Miniconda3.sh
```
Source: https://educe-ubc.github.io/conda.html
2023-04-01 13:31:10 -04:00
1. Create a new conda environment
2023-01-05 22:05:11 -05:00
2023-01-12 23:29:36 -05:00
```
2023-03-17 17:51:07 -04:00
conda create -n textgen python=3.10.9
2023-01-12 23:29:36 -05:00
conda activate textgen
2023-01-19 22:20:35 -05:00
```
2023-01-12 23:27:29 -05:00
2023-04-01 13:31:10 -04:00
2. Install Pytorch
2023-01-21 20:54:35 -05:00
2023-03-17 17:51:07 -04:00
| System | GPU | Command |
|--------|---------|---------|
2023-03-18 10:56:04 -04:00
| Linux/WSL | NVIDIA | `pip3 install torch torchvision torchaudio` |
2023-03-17 17:51:07 -04:00
| Linux | AMD | `pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.4.2` |
2023-03-18 10:56:04 -04:00
| MacOS + MPS (untested) | Any | `pip3 install torch torchvision torchaudio` |
2023-02-11 05:56:12 -05:00
2023-03-17 21:57:46 -04:00
The up to date commands can be found here: https://pytorch.org/get-started/locally/.
MacOS users, refer to the comments here: https://github.com/oobabooga/text-generation-webui/pull/393
2023-03-17 17:51:07 -04:00
2023-03-18 11:24:54 -04:00
2023-04-01 13:31:10 -04:00
3. Install the web UI
2023-01-05 22:05:11 -05:00
2023-01-12 23:29:36 -05:00
```
2023-03-17 17:51:07 -04:00
git clone https://github.com/oobabooga/text-generation-webui
cd text-generation-webui
pip install -r requirements.txt
2023-01-12 23:29:36 -05:00
```
2022-12-21 11:17:06 -05:00
2023-03-18 11:24:54 -04:00
> **Note**
>
> For bitsandbytes and `--load-in-8bit` to work on Linux/WSL, this dirty fix is currently necessary: https://github.com/oobabooga/text-generation-webui/issues/400#issuecomment-1474876859
2023-03-17 18:59:37 -04:00
2023-03-23 22:28:50 -04:00
2023-03-29 01:11:51 -04:00
### Alternative: manual Windows installation
2023-03-24 00:40:43 -04:00
2023-03-29 10:02:29 -04:00
As an alternative to the recommended WSL method, you can install the web UI natively on Windows using this guide. It will be a lot harder and the performance may be slower: [Windows installation guide ](https://github.com/oobabooga/text-generation-webui/wiki/Windows-installation-guide ).
2023-03-17 18:59:37 -04:00
### Alternative: Docker
https://github.com/oobabooga/text-generation-webui/issues/174, https://github.com/oobabooga/text-generation-webui/issues/87
2023-03-17 17:51:07 -04:00
2022-12-21 12:37:50 -05:00
## Downloading models
2023-03-17 17:55:01 -04:00
Models should be placed inside the `models` folder.
2022-12-21 12:37:50 -05:00
2023-04-01 13:38:04 -04:00
[Hugging Face ](https://huggingface.co/models?pipeline_tag=text-generation&sort=downloads ) is the main place to download models. These are some examples:
2022-12-21 12:37:50 -05:00
2023-04-01 13:49:03 -04:00
* [Pythia ](https://huggingface.co/models?sort=downloads&search=eleutherai%2Fpythia+deduped )
2023-01-07 17:19:55 -05:00
* [OPT ](https://huggingface.co/models?search=facebook/opt )
* [GALACTICA ](https://huggingface.co/models?search=facebook/galactica )
2023-03-17 17:55:01 -04:00
* [GPT-J 6B ](https://huggingface.co/EleutherAI/gpt-j-6B/tree/main )
2023-01-05 22:13:26 -05:00
2023-02-13 19:49:19 -05:00
You can automatically download a model from HF using the script `download-model.py` :
2023-01-06 18:00:58 -05:00
python download-model.py organization/model
2023-04-01 13:15:19 -04:00
For example:
2023-01-06 18:00:58 -05:00
python download-model.py facebook/opt-1.3b
2023-02-14 11:06:26 -05:00
If you want to download a model manually, note that all you need are the json, txt, and pytorch\*.bin (or model*.safetensors) files. The remaining files are not necessary.
2023-01-23 08:02:35 -05:00
2023-04-01 13:15:19 -04:00
#### GPT-4chan
2022-12-21 12:37:50 -05:00
2023-01-09 14:19:57 -05:00
[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 12:37:50 -05:00
* 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/ )
2023-01-12 23:39:09 -05:00
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.
2023-01-11 21:00:31 -05:00
2023-01-12 23:37:12 -05:00
After downloading the model, follow these steps:
1. Place the files under `models/gpt4chan_model_float16` or `models/gpt4chan_model` .
2023-01-16 09:19:23 -05:00
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 ).
2023-02-11 22:42:56 -05:00
3. Download GPT-J 6B's tokenizer files (they will be automatically detected when you attempt to load GPT-4chan):
2023-01-11 21:01:50 -05:00
2023-01-11 21:04:06 -05:00
```
2023-02-11 22:42:56 -05:00
python download-model.py EleutherAI/gpt-j-6B --text-only
2023-01-11 21:04:06 -05:00
```
2023-01-07 21:13:43 -05:00
2023-01-16 08:07:00 -05:00
## Starting the web UI
2022-12-21 11:17:06 -05:00
conda activate textgen
2023-03-17 20:05:17 -04:00
cd text-generation-webui
2022-12-21 11:17:06 -05:00
python server.py
2023-01-06 18:00:58 -05:00
Then browse to
`http://localhost:7860/?__theme=dark`
2023-01-06 18:07:58 -05:00
Optionally, you can use the following command-line flags:
2023-01-06 18:00:58 -05:00
2023-04-01 12:55:15 -04:00
#### Basic settings
2023-03-17 10:58:45 -04:00
| Flag | Description |
|------------------|-------------|
| `-h` , `--help` | show this help message and exit |
| `--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.|
2023-04-04 21:28:49 -04:00
| `--cai-chat` | Launch the web UI in chat mode with a style similar to the Character.AI website. |
2023-03-31 20:18:05 -04:00
| `--model MODEL` | Name of the model to load by default. |
| `--lora LORA` | Name of the LoRA to apply to the model by default. |
| `--model-dir MODEL_DIR` | Path to directory with all the models |
| `--lora-dir LORA_DIR` | Path to directory with all the loras |
| `--no-stream` | Don't stream the text output in real time. |
| `--settings SETTINGS_FILE` | Load the default interface settings from this json file. See `settings-template.json` for an example. If you create a file called `settings.json` , this file will be loaded by default without the need to use the `--settings` flag.|
| `--extensions EXTENSIONS [EXTENSIONS ...]` | The list of extensions to load. If you want to load more than one extension, write the names separated by spaces. |
| `--verbose` | Print the prompts to the terminal. |
2023-04-01 12:55:15 -04:00
#### Accelerate/transformers
| Flag | Description |
|------------------|-------------|
2023-03-17 10:58:45 -04:00
| `--cpu` | Use the CPU to generate text.|
2023-03-31 20:18:05 -04:00
| `--auto-devices` | Automatically split the model across the available GPU(s) and CPU.|
| `--gpu-memory GPU_MEMORY [GPU_MEMORY ...]` | Maxmimum GPU memory in GiB to be allocated per GPU. Example: `--gpu-memory 10` for a single GPU, `--gpu-memory 10 5` for two GPUs. You can also set values in MiB like `--gpu-memory 3500MiB` . |
2023-04-01 12:55:15 -04:00
| `--cpu-memory CPU_MEMORY` | Maximum CPU memory in GiB to allocate for offloaded weights. Same as above.|
2023-03-31 20:18:05 -04:00
| `--disk` | If the model is too large for your GPU(s) and CPU combined, send the remaining layers to the disk. |
| `--disk-cache-dir DISK_CACHE_DIR` | Directory to save the disk cache to. Defaults to `cache/` . |
2023-03-17 10:58:45 -04:00
| `--load-in-8bit` | Load the model with 8-bit precision.|
2023-03-31 20:18:05 -04:00
| `--bf16` | Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU. |
| `--no-cache` | Set `use_cache` to False while generating text. This reduces the VRAM usage a bit with a performance cost. |
2023-04-01 12:55:15 -04:00
#### llama.cpp
| Flag | Description |
|------------------|-------------|
2023-03-31 20:18:05 -04:00
| `--threads` | Number of threads to use in llama.cpp. |
2023-04-01 12:55:15 -04:00
#### GPTQ
| Flag | Description |
|------------------|-------------|
2023-03-25 23:11:33 -04:00
| `--wbits WBITS` | GPTQ: Load a pre-quantized model with specified precision in bits. 2, 3, 4 and 8 are supported. |
2023-03-28 16:46:00 -04:00
| `--model_type MODEL_TYPE` | GPTQ: Model type of pre-quantized model. Currently LLaMA, OPT, and GPT-J are supported. |
2023-03-25 23:11:33 -04:00
| `--groupsize GROUPSIZE` | GPTQ: Group size. |
2023-04-03 16:34:25 -04:00
| `--pre_layer PRE_LAYER` | GPTQ: The number of layers to allocate to the GPU. Setting this parameter enables CPU offloading for 4-bit models. |
2023-04-01 12:55:15 -04:00
#### FlexGen
| Flag | Description |
|------------------|-------------|
2023-03-19 18:21:41 -04:00
| `--flexgen` | Enable the use of FlexGen offloading. |
2023-03-17 10:58:45 -04:00
| `--percent PERCENT [PERCENT ...]` | FlexGen: allocation percentages. Must be 6 numbers separated by spaces (default: 0, 100, 100, 0, 100, 0). |
| `--compress-weight` | FlexGen: Whether to compress weight (default: False).|
| `--pin-weight [PIN_WEIGHT]` | FlexGen: whether to pin weights (setting this to False reduces CPU memory by 20%). |
2023-04-01 12:55:15 -04:00
#### DeepSpeed
| Flag | Description |
|------------------|-------------|
2023-02-16 15:29:48 -05:00
| `--deepspeed` | Enable the use of DeepSpeed ZeRO-3 for inference via the Transformers integration. |
2023-03-17 10:58:45 -04:00
| `--nvme-offload-dir NVME_OFFLOAD_DIR` | DeepSpeed: Directory to use for ZeRO-3 NVME offloading. |
| `--local_rank LOCAL_RANK` | DeepSpeed: Optional argument for distributed setups. |
2023-04-01 12:55:15 -04:00
#### RWKV
| Flag | Description |
|------------------|-------------|
2023-03-17 10:58:45 -04:00
| `--rwkv-strategy RWKV_STRATEGY` | RWKV: The strategy to use while loading the model. Examples: "cpu fp32", "cuda fp16", "cuda fp16i8". |
| `--rwkv-cuda-on` | RWKV: Compile the CUDA kernel for better performance. |
2023-04-01 12:55:15 -04:00
#### Gradio
| Flag | Description |
|------------------|-------------|
2023-03-27 22:35:01 -04:00
| `--listen` | Make the web UI reachable from your local network. |
| `--listen-port LISTEN_PORT` | The listening port that the server will use. |
| `--share` | Create a public URL. This is useful for running the web UI on Google Colab or similar. |
| `--auto-launch` | Open the web UI in the default browser upon launch. |
2023-03-27 22:39:26 -04:00
| `--gradio-auth-path GRADIO_AUTH_PATH` | Set the gradio authentication file path. The file should contain one or more user:password pairs in this format: "u1:p1,u2:p2,u3:p3" |
2023-01-09 08:58:46 -05:00
2023-03-19 18:25:52 -04:00
Out of memory errors? [Check the low VRAM guide ](https://github.com/oobabooga/text-generation-webui/wiki/Low-VRAM-guide ).
2023-01-19 09:25:17 -05:00
2023-01-05 23:33:21 -05:00
## Presets
Inference settings presets can be created under `presets/` as text files. These files are detected automatically at startup.
2023-01-23 18:44:27 -05:00
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.
2023-04-01 13:15:19 -04:00
[Visualization ](https://user-images.githubusercontent.com/112222186/228956352-1addbdb9-2456-465a-b51d-089f462cd385.png )
2023-01-09 16:12:41 -05:00
## System requirements
2023-01-09 16:18:35 -05:00
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 16:12:41 -05:00
2022-12-21 14:52:23 -05:00
## Contributing
2023-01-16 08:07:00 -05:00
Pull requests, suggestions, and issue reports are welcome.
2023-01-09 14:15:54 -05:00
2023-03-12 11:46:53 -04:00
Before reporting a bug, make sure that you have:
2023-01-29 12:37:05 -05:00
2023-03-12 11:46:53 -04:00
1. Created a conda environment and installed the dependencies exactly as in the *Installation* section above.
2. [Searched ](https://github.com/oobabooga/text-generation-webui/issues ) to see if an issue already exists for the issue you encountered.
2023-02-07 23:17:50 -05:00
2023-01-15 01:00:59 -05:00
## Credits
2023-01-09 14:15:54 -05:00
2023-03-16 11:45:27 -04:00
- Gradio dropdown menu refresh button, code for reloading the interface: https://github.com/AUTOMATIC1111/stable-diffusion-webui
2023-03-09 09:13:03 -05:00
- Verbose preset: Anonymous 4chan user.
2023-01-15 02:29:19 -05:00
- NovelAI and KoboldAI presets: https://github.com/KoboldAI/KoboldAI-Client/wiki/Settings-Presets
2023-03-30 16:34:44 -04:00
- Code for early stopping in chat mode, code for some of the sliders: https://github.com/PygmalionAI/gradio-ui/
