2023-01-16 08:31:13 -05:00
# Text generation web UI
2022-12-21 14:49:30 -05:00
2023-08-15 14:40:37 -04:00
A Gradio web UI for Large Language Models.
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
2023-08-15 16:16:24 -04:00
|![Image1](https://github.com/oobabooga/screenshots/raw/main/print_instruct.png) | ![Image2 ](https://github.com/oobabooga/screenshots/raw/main/print_chat.png ) |
2023-01-14 21:25:05 -05:00
|:---:|:---:|
2023-08-15 16:16:24 -04:00
|![Image1](https://github.com/oobabooga/screenshots/raw/main/print_default.png) | ![Image2 ](https://github.com/oobabooga/screenshots/raw/main/print_parameters.png ) |
2022-12-21 12:04:51 -05:00
2023-01-07 16:40:51 -05:00
## Features
2023-08-15 14:40:37 -04:00
* 3 interface modes: default (two columns), notebook, and chat
2023-12-08 11:39:20 -05:00
* Multiple model backends: [Transformers ](https://github.com/huggingface/transformers ), [llama.cpp ](https://github.com/ggerganov/llama.cpp ) (through [llama-cpp-python ](https://github.com/abetlen/llama-cpp-python )), [ExLlama ](https://github.com/turboderp/exllama ), [ExLlamaV2 ](https://github.com/turboderp/exllamav2 ), [AutoGPTQ ](https://github.com/PanQiWei/AutoGPTQ ), [AutoAWQ ](https://github.com/casper-hansen/AutoAWQ ), [GPTQ-for-LLaMa ](https://github.com/qwopqwop200/GPTQ-for-LLaMa ), [CTransformers ](https://github.com/marella/ctransformers ), [QuIP# ](https://github.com/Cornell-RelaxML/quip-sharp )
2023-06-05 14:16:06 -04:00
* Dropdown menu for quickly switching between different models
2023-08-15 14:40:37 -04:00
* LoRA: load and unload LoRAs on the fly, train a new LoRA using QLoRA
* Precise instruction templates for chat mode, including Llama-2-chat, Alpaca, Vicuna, WizardLM, StableLM, and many others
* 4-bit, 8-bit, and CPU inference through the transformers library
* Use llama.cpp models with transformers samplers (`llamacpp_HF` loader)
2023-05-10 11:09:47 -04:00
* [Multimodal pipelines, including LLaVA and MiniGPT-4 ](https://github.com/oobabooga/text-generation-webui/tree/main/extensions/multimodal )
2023-10-23 13:09:57 -04:00
* [Extensions framework ](https://github.com/oobabooga/text-generation-webui/wiki/07-%E2%80%90-Extensions )
* [Custom chat characters ](https://github.com/oobabooga/text-generation-webui/wiki/03-%E2%80%90-Parameters-Tab#character )
2023-06-05 14:16:06 -04:00
* Markdown output with LaTeX rendering, to use for instance with [GALACTICA ](https://github.com/paperswithcode/galai )
2023-11-07 11:05:06 -05:00
* OpenAI-compatible API server with Chat and Completions endpoints -- see the [examples ](https://github.com/oobabooga/text-generation-webui/wiki/12-%E2%80%90-OpenAI-API#examples )
2023-06-05 14:16:06 -04:00
2023-10-21 19:03:43 -04:00
## Documentation
To learn how to use the various features, check out the Documentation:
https://github.com/oobabooga/text-generation-webui/wiki
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
2023-10-21 02:58:05 -04:00
1) Clone or [download ](https://github.com/oobabooga/text-generation-webui/archive/refs/heads/main.zip ) the repository.
2023-09-22 18:20:09 -04:00
2) Run the `start_linux.sh` , `start_windows.bat` , `start_macos.sh` , or `start_wsl.bat` script depending on your OS.
3) Select your GPU vendor when asked.
2023-09-22 22:23:16 -04:00
4) Have fun!
2023-04-18 01:40:36 -04:00
2023-09-22 22:31:32 -04:00
#### How it works
2023-03-29 01:11:51 -04:00
2023-09-22 22:23:16 -04:00
The script creates a folder called `installer_files` where it sets up a Conda environment using Miniconda. The installation is self-contained: if you want to reinstall, just delete `installer_files` and run the start script again.
2023-09-22 14:52:20 -04:00
2023-10-20 20:13:33 -04:00
To launch the webui in the future after it is already installed, run the same `start` script.
2023-09-22 14:52:20 -04:00
2023-09-22 22:31:32 -04:00
#### Getting updates
2023-09-22 14:52:20 -04:00
2023-09-22 22:31:32 -04:00
Run `update_linux.sh` , `update_windows.bat` , `update_macos.sh` , or `update_wsl.bat` .
#### Running commands
If you ever need to install something manually in the `installer_files` environment, you can launch an interactive shell using the cmd script: `cmd_linux.sh` , `cmd_windows.bat` , `cmd_macos.sh` , or `cmd_wsl.bat` .
#### Defining command-line flags
2023-09-22 14:52:20 -04:00
To define persistent command-line flags like `--listen` or `--api` , edit the `CMD_FLAGS.txt` file with a text editor and add them there. Flags can also be provided directly to the start scripts, for instance, `./start-linux.sh --listen` .
2023-09-22 22:31:32 -04:00
#### Other info
2023-09-22 14:52:20 -04:00
* There is no need to run any of those scripts as admin/root.
2023-12-13 15:55:41 -05:00
* For additional instructions about AMD setup and WSL setup, consult [the documentation ](https://github.com/oobabooga/text-generation-webui/wiki ).
2023-09-22 15:00:06 -04:00
* The installer has been tested mostly on NVIDIA GPUs. If you can find a way to improve it for your AMD/Intel Arc/Mac Metal GPU, you are highly encouraged to submit a PR to this repository. The main file to be edited is `one_click.py` .
2023-10-21 02:46:23 -04:00
* For automated installation, you can use the `GPU_CHOICE` , `USE_CUDA118` , `LAUNCH_AFTER_INSTALL` , and `INSTALL_EXTENSIONS` environment variables. For instance: `GPU_CHOICE=A USE_CUDA118=FALSE LAUNCH_AFTER_INSTALL=FALSE INSTALL_EXTENSIONS=FALSE ./start_linux.sh` .
2023-03-29 01:11:51 -04:00
### Manual installation using Conda
2023-08-15 14:40:37 -04:00
Recommended if you have some experience with the command-line.
2023-03-17 17:51:07 -04:00
2023-04-05 13:42:58 -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
2023-08-15 14:40:37 -04:00
On Linux or WSL, it can be automatically installed with these two commands ([source](https://educe-ubc.github.io/conda.html)):
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
```
2023-04-05 13:42:58 -04:00
#### 1. Create a new conda environment
2023-01-05 22:05:11 -05:00
2023-01-12 23:29:36 -05:00
```
2023-10-21 02:46:23 -04:00
conda create -n textgen python=3.11
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
2023-04-05 13:42:58 -04:00
#### 2. Install Pytorch
2023-01-21 20:54:35 -05:00
2023-03-17 17:51:07 -04:00
| System | GPU | Command |
|--------|---------|---------|
2023-10-21 02:46:23 -04:00
| Linux/WSL | NVIDIA | `pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121` |
2023-07-25 18:13:07 -04:00
| Linux/WSL | CPU only | `pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu` |
2023-10-08 09:12:13 -04:00
| Linux | AMD | `pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.6` |
2023-07-25 18:13:07 -04:00
| MacOS + MPS | Any | `pip3 install torch torchvision torchaudio` |
2023-10-21 02:46:23 -04:00
| Windows | NVIDIA | `pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121` |
2023-07-25 17:58:13 -04:00
| Windows | CPU only | `pip3 install torch torchvision torchaudio` |
2023-02-11 05:56:12 -05:00
2023-10-20 20:13:33 -04:00
The up-to-date commands can be found here: https://pytorch.org/get-started/locally/.
2023-03-17 21:57:46 -04:00
2023-12-13 15:55:41 -05:00
For NVIDIA, you also need to install the CUDA runtime libraries:
2023-10-13 00:02:44 -04:00
```
2023-12-13 15:55:41 -05:00
conda install -y -c "nvidia/label/cuda-12.1.1" cuda-runtime
```
If you need `nvcc` to compile some library manually, replace the command above with
```
conda install -y -c "nvidia/label/cuda-12.1.1" cuda
2023-10-13 00:02:44 -04:00
```
2023-04-05 13:42:58 -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
2023-10-20 20:13:33 -04:00
pip install -r < requirements file according to table below >
2023-01-12 23:29:36 -05:00
```
2022-12-21 11:17:06 -05:00
2023-10-20 20:13:33 -04:00
Requirements file to use:
2023-08-18 12:37:20 -04:00
2023-10-20 20:13:33 -04:00
| GPU | CPU | requirements file to use |
|--------|---------|---------|
| NVIDIA | has AVX2 | `requirements.txt` |
| NVIDIA | no AVX2 | `requirements_noavx2.txt` |
| AMD | has AVX2 | `requirements_amd.txt` |
| AMD | no AVX2 | `requirements_amd_noavx2.txt` |
| CPU only | has AVX2 | `requirements_cpu_only.txt` |
| CPU only | no AVX2 | `requirements_cpu_only_noavx2.txt` |
| Apple | Intel | `requirements_apple_intel.txt` |
| Apple | Apple Silicon | `requirements_apple_silicon.txt` |
2023-08-18 12:37:20 -04:00
2023-10-20 20:13:33 -04:00
##### AMD GPU on Windows
2023-08-18 12:37:20 -04:00
2023-10-20 20:13:33 -04:00
1) Use `requirements_cpu_only.txt` or `requirements_cpu_only_noavx2.txt` in the command above.
2023-09-24 08:58:29 -04:00
2023-10-20 20:13:33 -04:00
2) Manually install llama-cpp-python using the appropriate command for your hardware: [Installation from PyPI ](https://github.com/abetlen/llama-cpp-python#installation-with-hardware-acceleration ).
* Use the `LLAMA_HIPBLAS=on` toggle.
* Note the [Windows remarks ](https://github.com/abetlen/llama-cpp-python#windows-remarks ).
2023-08-30 12:37:53 -04:00
2023-10-20 20:13:33 -04:00
3) Manually install AutoGPTQ: [Installation ](https://github.com/PanQiWei/AutoGPTQ#install-from-source ).
* Perform the from-source installation - there are no prebuilt ROCm packages for Windows.
2023-08-30 12:37:53 -04:00
2023-10-20 20:13:33 -04:00
4) Manually install [ExLlama ](https://github.com/turboderp/exllama ) by simply cloning it into the `repositories` folder (it will be automatically compiled at runtime after that):
2023-08-30 12:37:53 -04:00
2023-10-20 20:13:33 -04:00
```sh
2023-08-30 12:37:53 -04:00
cd text-generation-webui
2023-09-17 18:26:05 -04:00
git clone https://github.com/turboderp/exllama repositories/exllama
2023-08-30 12:37:53 -04:00
```
2023-08-18 12:37:20 -04:00
2023-10-21 02:46:23 -04:00
##### Older NVIDIA GPUs
1) For Kepler GPUs and older, you will need to install CUDA 11.8 instead of 12:
2023-05-25 10:21:52 -04:00
2023-10-21 02:46:23 -04:00
```
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
conda install -y -c "nvidia/label/cuda-11.8.0" cuda-runtime
```
2023-05-25 10:21:52 -04:00
2023-10-21 02:46:23 -04:00
2) bitsandbytes >= 0.39 may not work. In that case, to use `--load-in-8bit` , you may have to downgrade like this:
* Linux: `pip install bitsandbytes==0.38.1`
* Windows: `pip install https://github.com/jllllll/bitsandbytes-windows-webui/raw/main/bitsandbytes-0.38.1-py3-none-any.whl`
2023-05-25 10:21:52 -04:00
2023-10-20 20:13:33 -04:00
##### Manual install
2023-12-03 20:46:34 -05:00
The requirements*.txt above contain various precompiled wheels. If you wish to compile things manually, or if you need to because no suitable wheels are available for your hardware, you can use `requirements_nowheels.txt` and then install your desired loaders manually.
2023-10-20 20:13:33 -04:00
2023-03-17 18:59:37 -04:00
### Alternative: Docker
2023-04-06 21:50:58 -04:00
```
2023-11-30 00:27:23 -05:00
ln -s docker/{nvidia/Dockerfile,docker-compose.yml,.dockerignore} .
2023-04-18 11:47:43 -04:00
cp docker/.env.example .env
2023-11-30 00:20:23 -05:00
# Edit .env and set:
# TORCH_CUDA_ARCH_LIST based on your GPU model
# APP_RUNTIME_GID your host user's group id (run `id -g` in a terminal)
# BUILD_EXTENIONS optionally add comma separated list of extensions to build
2023-04-07 09:48:47 -04:00
docker compose up --build
2023-04-06 21:50:58 -04:00
```
2023-11-17 22:52:30 -05:00
* You need to have Docker Compose v2.17 or higher installed. See [this guide ](https://github.com/oobabooga/text-generation-webui/wiki/09-%E2%80%90-Docker ) for instructions.
2023-05-30 21:14:49 -04:00
* For additional docker files, check out [this repository ](https://github.com/Atinoda/text-generation-webui-docker ).
2023-03-17 17:51:07 -04:00
2023-04-06 12:24:01 -04:00
### Updating the requirements
2023-10-20 20:13:33 -04:00
From time to time, the `requirements*.txt` changes. To update, use these commands:
2023-04-06 12:24:01 -04:00
```
conda activate textgen
cd text-generation-webui
2023-10-20 20:13:33 -04:00
pip install -r < requirements file that you ' ve used > --upgrade
2023-04-06 12:24:01 -04:00
```
2022-12-21 12:37:50 -05:00
2023-08-15 14:40:37 -04:00
## Downloading models
2023-01-05 22:13:26 -05:00
2023-08-15 14:40:37 -04:00
Models should be placed in the `text-generation-webui/models` folder. They are usually downloaded from [Hugging Face ](https://huggingface.co/models?pipeline_tag=text-generation&sort=downloads ).
2023-01-06 18:00:58 -05:00
2023-08-15 14:40:37 -04:00
* Transformers or GPTQ models are made of several files and must be placed in a subfolder. Example:
2023-01-06 18:00:58 -05:00
2023-08-15 14:40:37 -04:00
```
2023-08-15 15:16:21 -04:00
text-generation-webui
2023-08-15 14:40:37 -04:00
├── models
│ ├── lmsys_vicuna-33b-v1.3
│ │ ├── config.json
│ │ ├── generation_config.json
│ │ ├── pytorch_model-00001-of-00007.bin
│ │ ├── pytorch_model-00002-of-00007.bin
│ │ ├── pytorch_model-00003-of-00007.bin
│ │ ├── pytorch_model-00004-of-00007.bin
│ │ ├── pytorch_model-00005-of-00007.bin
│ │ ├── pytorch_model-00006-of-00007.bin
│ │ ├── pytorch_model-00007-of-00007.bin
│ │ ├── pytorch_model.bin.index.json
│ │ ├── special_tokens_map.json
│ │ ├── tokenizer_config.json
│ │ └── tokenizer.model
```
2023-01-06 18:00:58 -05:00
2023-09-11 10:30:56 -04:00
* GGUF models are a single file and should be placed directly into `models` . Example:
2023-01-23 08:02:35 -05:00
2023-08-15 14:40:37 -04:00
```
2023-08-15 15:21:50 -04:00
text-generation-webui
2023-08-15 14:40:37 -04:00
├── models
2023-09-15 16:00:46 -04:00
│ ├── llama-2-13b-chat.Q4_K_M.gguf
2023-08-15 14:40:37 -04:00
```
2023-07-25 23:49:00 -04:00
2023-08-30 17:19:04 -04:00
In both cases, you can use the "Model" tab of the UI to download the model from Hugging Face automatically. It is also possible to download via the command-line with `python download-model.py organization/model` (use `--help` to see all the options).
2023-07-25 23:49:00 -04:00
#### GPT-4chan
< details >
< summary >
Instructions
< / summary >
[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:
* 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/ )
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.
After downloading the model, follow these steps:
1. Place the files under `models/gpt4chan_model_float16` or `models/gpt4chan_model` .
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's tokenizer files (they will be automatically detected when you attempt to load GPT-4chan):
```
python download-model.py EleutherAI/gpt-j-6B --text-only
```
2023-08-15 16:16:24 -04:00
When you load this model in default or notebook modes, the "HTML" tab will show the generated text in 4chan format:
![Image3 ](https://github.com/oobabooga/screenshots/raw/main/gpt4chan.png )
2023-07-25 23:49:00 -04:00
< / details >
2023-06-05 14:16:06 -04:00
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
2023-10-20 20:13:33 -04:00
Then browse to
2023-01-06 18:00:58 -05:00
`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
2023-04-07 09:47:10 -04:00
| Flag | Description |
|--------------------------------------------|-------------|
2023-10-21 21:52:59 -04:00
| `-h` , `--help` | show this help message and exit |
| `--multi-user` | Multi-user mode. Chat histories are not saved or automatically loaded. WARNING: this is likely not safe for sharing publicly. |
2023-04-24 12:19:42 -04:00
| `--character CHARACTER` | The name of the character to load in chat mode by default. |
2023-04-07 09:47:10 -04:00
| `--model MODEL` | Name of the model to load by default. |
2023-04-25 21:58:48 -04:00
| `--lora LORA [LORA ...]` | The list of LoRAs to load. If you want to load more than one LoRA, write the names separated by spaces. |
2023-04-07 09:47:10 -04:00
| `--model-dir MODEL_DIR` | Path to directory with all the models. |
| `--lora-dir LORA_DIR` | Path to directory with all the loras. |
2023-04-12 20:24:26 -04:00
| `--model-menu` | Show a model menu in the terminal when the web UI is first launched. |
2023-05-28 21:34:12 -04:00
| `--settings SETTINGS_FILE` | Load the default interface settings from this yaml file. See `settings-template.yaml` for an example. If you create a file called `settings.yaml` , this file will be loaded by default without the need to use the `--settings` flag. |
2023-04-07 09:47:10 -04:00
| `--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. |
2023-10-21 21:52:59 -04:00
| `--chat-buttons` | Show buttons on the chat tab instead of a hover menu. |
2023-04-01 12:55:15 -04:00
2023-06-16 18:00:37 -04:00
#### Model loader
| Flag | Description |
|--------------------------------------------|-------------|
2023-12-08 11:39:20 -05:00
| `--loader LOADER` | Choose the model loader manually, otherwise, it will get autodetected. Valid options: Transformers, llama.cpp, llamacpp_HF, ExLlama_HF, ExLlamav2_HF, AutoGPTQ, AutoAWQ, GPTQ-for-LLaMa, ExLlama, ExLlamav2, ctransformers, QuIP#. |
2023-06-16 18:00:37 -04:00
2023-04-01 12:55:15 -04:00
#### Accelerate/transformers
2023-04-07 09:47:10 -04:00
| Flag | Description |
|---------------------------------------------|-------------|
2023-10-21 21:52:59 -04:00
| `--cpu` | Use the CPU to generate text. Warning: Training on CPU is extremely slow. |
2023-04-07 09:47:10 -04:00
| `--auto-devices` | Automatically split the model across the available GPU(s) and CPU. |
2023-10-21 21:52:59 -04:00
| `--gpu-memory GPU_MEMORY [GPU_MEMORY ...]` | Maximum 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. |
| `--cpu-memory CPU_MEMORY` | Maximum CPU memory in GiB to allocate for offloaded weights. Same as above. |
2023-04-07 09:47:10 -04:00
| `--disk` | If the model is too large for your GPU(s) and CPU combined, send the remaining layers to the disk. |
2023-10-21 21:52:59 -04:00
| `--disk-cache-dir DISK_CACHE_DIR` | Directory to save the disk cache to. Defaults to "cache". |
| `--load-in-8bit` | Load the model with 8-bit precision (using bitsandbytes). |
2023-04-07 09:47:10 -04:00
| `--bf16` | Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU. |
2023-10-21 21:52:59 -04:00
| `--no-cache` | Set `use_cache` to `False` while generating text. This reduces VRAM usage slightly, but it comes at a performance cost. |
| `--xformers` | Use xformer's memory efficient attention. This is really old and probably doesn't do anything. |
| `--sdp-attention` | Use PyTorch 2.0's SDP attention. Same as above. |
| `--trust-remote-code` | Set `trust_remote_code=True` while loading the model. Necessary for some models. |
2023-11-16 22:57:55 -05:00
| `--no_use_fast` | Set use_fast=False while loading the tokenizer (it's True by default). Use this if you have any problems related to use_fast. |
2023-11-04 12:59:33 -04:00
| `--use_flash_attention_2` | Set use_flash_attention_2=True while loading the model. |
2023-04-01 12:55:15 -04:00
2023-05-25 00:14:13 -04:00
#### Accelerate 4-bit
2023-10-21 21:52:59 -04:00
⚠️ Requires minimum compute of 7.0 on Windows at the moment.
2023-05-25 00:23:46 -04:00
2023-05-25 00:14:13 -04:00
| Flag | Description |
|---------------------------------------------|-------------|
| `--load-in-4bit` | Load the model with 4-bit precision (using bitsandbytes). |
2023-10-21 21:52:59 -04:00
| `--use_double_quant` | use_double_quant for 4-bit. |
2023-05-25 00:14:13 -04:00
| `--compute_dtype COMPUTE_DTYPE` | compute dtype for 4-bit. Valid options: bfloat16, float16, float32. |
| `--quant_type QUANT_TYPE` | quant_type for 4-bit. Valid options: nf4, fp4. |
2023-10-21 21:52:59 -04:00
#### llama.cpp
2023-04-01 12:55:15 -04:00
2023-04-07 09:47:10 -04:00
| Flag | Description |
|-------------|-------------|
2023-10-21 21:52:59 -04:00
| `--n_ctx N_CTX` | Size of the prompt context. |
2023-05-02 17:25:28 -04:00
| `--threads` | Number of threads to use. |
2023-10-02 00:27:04 -04:00
| `--threads-batch THREADS_BATCH` | Number of threads to use for batches/prompt processing. |
2023-10-22 15:22:06 -04:00
| `--no_mul_mat_q` | Disable the mulmat kernels. |
2023-10-21 21:52:59 -04:00
| `--n_batch` | Maximum number of prompt tokens to batch together when calling llama_eval. |
2023-09-26 21:05:00 -04:00
| `--no-mmap` | Prevent mmap from being used. |
| `--mlock` | Force the system to keep the model in RAM. |
2023-10-21 21:52:59 -04:00
| `--n-gpu-layers N_GPU_LAYERS` | Number of layers to offload to the GPU. |
| `--tensor_split TENSOR_SPLIT` | Split the model across multiple GPUs. Comma-separated list of proportions. Example: 18,17. |
| `--numa` | Activate NUMA task allocation for llama.cpp. |
2023-11-07 17:45:11 -05:00
| `--logits_all` | Needs to be set for perplexity evaluation to work. Otherwise, ignore it, as it makes prompt processing slower. |
2023-10-21 21:52:59 -04:00
| `--cache-capacity CACHE_CAPACITY` | Maximum cache capacity (llama-cpp-python). Examples: 2000MiB, 2GiB. When provided without units, bytes will be assumed. |
2023-04-01 12:55:15 -04:00
2023-10-21 21:52:59 -04:00
#### ExLlama
2023-08-13 22:09:03 -04:00
2023-10-21 21:52:59 -04:00
| Flag | Description |
|------------------|-------------|
|`--gpu-split` | Comma-separated list of VRAM (in GB) to use per GPU device for model layers. Example: 20,7,7. |
|`--max_seq_len MAX_SEQ_LEN` | Maximum sequence length. |
|`--cfg-cache` | ExLlama_HF: Create an additional cache for CFG negative prompts. Necessary to use CFG with that loader, but not necessary for CFG with base ExLlama. |
2023-11-02 11:19:42 -04:00
|`--no_flash_attn` | Force flash-attention to not be used. |
2023-11-02 14:23:04 -04:00
|`--cache_8bit` | Use 8-bit cache to save VRAM. |
2023-08-13 22:09:03 -04:00
2023-06-05 14:41:48 -04:00
#### AutoGPTQ
| Flag | Description |
|------------------|-------------|
2023-06-15 22:59:54 -04:00
| `--triton` | Use triton. |
| `--no_inject_fused_attention` | Disable the use of fused attention, which will use less VRAM at the cost of slower inference. |
| `--no_inject_fused_mlp` | Triton mode only: disable the use of fused MLP, which will use less VRAM at the cost of slower inference. |
2023-06-23 11:24:43 -04:00
| `--no_use_cuda_fp16` | This can make models faster on some systems. |
2023-06-15 22:59:54 -04:00
| `--desc_act` | For models that don't have a quantize_config.json, this parameter is used to define whether to set desc_act or not in BaseQuantizeConfig. |
2023-08-12 02:26:58 -04:00
| `--disable_exllama` | Disable ExLlama kernel, which can improve inference speed on some systems. |
2023-06-05 14:41:48 -04:00
#### GPTQ-for-LLaMa
2023-04-01 12:55:15 -04:00
2023-04-07 09:47:10 -04:00
| Flag | Description |
|---------------------------|-------------|
2023-04-17 09:55:35 -04:00
| `--wbits WBITS` | Load a pre-quantized model with specified precision in bits. 2, 3, 4 and 8 are supported. |
| `--model_type MODEL_TYPE` | Model type of pre-quantized model. Currently LLaMA, OPT, and GPT-J are supported. |
| `--groupsize GROUPSIZE` | Group size. |
2023-05-17 09:41:09 -04:00
| `--pre_layer PRE_LAYER [PRE_LAYER ...]` | The number of layers to allocate to the GPU. Setting this parameter enables CPU offloading for 4-bit models. For multi-gpu, write the numbers separated by spaces, eg `--pre_layer 30 60` . |
2023-05-04 14:17:20 -04:00
| `--checkpoint CHECKPOINT` | The path to the quantized checkpoint file. If not specified, it will be automatically detected. |
2023-10-21 21:52:59 -04:00
| `--monkey-patch` | Apply the monkey patch for using LoRAs with quantized models. |
#### ctransformers
| Flag | Description |
|-------------|-------------|
| `--model_type MODEL_TYPE` | Model type of pre-quantized model. Currently gpt2, gptj, gptneox, falcon, llama, mpt, starcoder (gptbigcode), dollyv2, and replit are supported. |
2023-04-01 12:55:15 -04:00
#### DeepSpeed
2023-04-07 09:47:10 -04:00
| Flag | Description |
|---------------------------------------|-------------|
| `--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. |
2023-04-07 09:47:10 -04:00
| `--local_rank LOCAL_RANK` | DeepSpeed: Optional argument for distributed setups. |
2023-04-01 12:55:15 -04:00
#### RWKV
2023-04-07 09:47:10 -04:00
| Flag | Description |
|---------------------------------|-------------|
| `--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
2023-09-12 18:05:21 -04:00
#### RoPE (for llama.cpp, ExLlama, ExLlamaV2, and transformers)
2023-07-25 17:58:13 -04:00
| Flag | Description |
|------------------|-------------|
2023-10-21 21:52:59 -04:00
| `--alpha_value ALPHA_VALUE` | Positional embeddings alpha factor for NTK RoPE scaling. Use either this or `compress_pos_emb` , not both. |
| `--rope_freq_base ROPE_FREQ_BASE` | If greater than 0, will be used instead of alpha_value. Those two are related by `rope_freq_base = 10000 * alpha_value ^ (64 / 63)` . |
| `--compress_pos_emb COMPRESS_POS_EMB` | Positional embeddings compression factor. Should be set to `(context length) / (model's original context length)` . Equal to `1/rope_freq_scale` . |
2023-07-25 17:58:13 -04:00
2023-04-01 12:55:15 -04:00
#### Gradio
2023-04-07 09:47:10 -04:00
| Flag | Description |
|---------------------------------------|-------------|
| `--listen` | Make the web UI reachable from your local network. |
| `--listen-port LISTEN_PORT` | The listening port that the server will use. |
2023-10-21 21:52:59 -04:00
| `--listen-host LISTEN_HOST` | The hostname that the server will use. |
2023-04-07 09:47:10 -04:00
| `--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. |
2023-10-21 21:52:59 -04:00
| `--gradio-auth USER:PWD` | Set Gradio authentication password in the format "username:password". Multiple credentials can also be supplied with "u1:p1,u2:p2,u3:p3". |
| `--gradio-auth-path GRADIO_AUTH_PATH` | Set the Gradio authentication file path. The file should contain one or more user:password pairs in the same format as above. |
2023-08-04 12:57:31 -04:00
| `--ssl-keyfile SSL_KEYFILE` | The path to the SSL certificate key file. |
| `--ssl-certfile SSL_CERTFILE` | The path to the SSL certificate cert file. |
2023-01-09 08:58:46 -05:00
2023-04-23 14:52:43 -04:00
#### API
| Flag | Description |
|---------------------------------------|-------------|
| `--api` | Enable the API extension. |
| `--public-api` | Create a public URL for the API using Cloudfare. |
2023-08-08 21:20:27 -04:00
| `--public-api-id PUBLIC_API_ID` | Tunnel ID for named Cloudflare Tunnel. Use together with public-api option. |
2023-11-06 00:38:29 -05:00
| `--api-port API_PORT` | The listening port for the API. |
| `--api-key API_KEY` | API authentication key. |
2023-11-18 20:33:27 -05:00
| `--admin-key ADMIN_KEY` | API authentication key for admin tasks like loading and unloading models. If not set, will be the same as --api-key. |
2023-11-18 21:38:39 -05:00
| `--nowebui` | Do not launch the Gradio UI. Useful for launching the API in standalone mode. |
2023-04-23 14:52:43 -04:00
2023-05-09 19:18:02 -04:00
#### Multimodal
| Flag | Description |
|---------------------------------------|-------------|
| `--multimodal-pipeline PIPELINE` | The multimodal pipeline to use. Examples: `llava-7b` , `llava-13b` . |
2023-10-22 00:22:17 -04:00
## Google Colab notebook
2023-01-05 23:33:21 -05:00
2023-10-22 00:22:17 -04:00
https://colab.research.google.com/github/oobabooga/text-generation-webui/blob/main/Colab-TextGen-GPU.ipynb
2023-01-09 16:12:41 -05:00
2022-12-21 14:52:23 -05:00
## Contributing
2023-08-03 17:36:35 -04:00
If you would like to contribute to the project, check out the [Contributing guidelines ](https://github.com/oobabooga/text-generation-webui/wiki/Contributing-guidelines ).
2023-02-07 23:17:50 -05:00
2023-06-27 12:56:14 -04:00
## Community
2023-08-15 14:40:37 -04:00
* Subreddit: https://www.reddit.com/r/oobabooga/
2023-06-27 12:56:14 -04:00
* Discord: https://discord.gg/jwZCF2dPQN
2023-08-21 14:57:51 -04:00
## Acknowledgment
In August 2023, [Andreessen Horowitz ](https://a16z.com/ ) (a16z) provided a generous grant to encourage and support my independent work on this project. I am **extremely** grateful for their trust and recognition, which will allow me to dedicate more time towards realizing the full potential of text-generation-webui.