jaty54 commited on
Commit
5084ce4
1 Parent(s): 722a6ef

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -393
README.md CHANGED
@@ -1,393 +1,7 @@
1
- # Text generation web UI
2
-
3
- A Gradio web UI for Large Language Models.
4
-
5
- Its goal is to become the [AUTOMATIC1111/stable-diffusion-webui](https://github.com/AUTOMATIC1111/stable-diffusion-webui) of text generation.
6
-
7
- |![Image1](https://github.com/oobabooga/screenshots/raw/main/print_instruct.png) | ![Image2](https://github.com/oobabooga/screenshots/raw/main/print_chat.png) |
8
- |:---:|:---:|
9
- |![Image1](https://github.com/oobabooga/screenshots/raw/main/print_default.png) | ![Image2](https://github.com/oobabooga/screenshots/raw/main/print_parameters.png) |
10
-
11
- ## Features
12
-
13
- * 3 interface modes: default (two columns), notebook, and chat
14
- * Multiple model backends: [transformers](https://github.com/huggingface/transformers), [llama.cpp](https://github.com/ggerganov/llama.cpp), [ExLlama](https://github.com/turboderp/exllama), [AutoGPTQ](https://github.com/PanQiWei/AutoGPTQ), [GPTQ-for-LLaMa](https://github.com/qwopqwop200/GPTQ-for-LLaMa), [ctransformers](https://github.com/marella/ctransformers)
15
- * Dropdown menu for quickly switching between different models
16
- * LoRA: load and unload LoRAs on the fly, train a new LoRA using QLoRA
17
- * Precise instruction templates for chat mode, including Llama-2-chat, Alpaca, Vicuna, WizardLM, StableLM, and many others
18
- * 4-bit, 8-bit, and CPU inference through the transformers library
19
- * Use llama.cpp models with transformers samplers (`llamacpp_HF` loader)
20
- * [Multimodal pipelines, including LLaVA and MiniGPT-4](https://github.com/oobabooga/text-generation-webui/tree/main/extensions/multimodal)
21
- * [Extensions framework](docs/Extensions.md)
22
- * [Custom chat characters](docs/Chat-mode.md)
23
- * Very efficient text streaming
24
- * Markdown output with LaTeX rendering, to use for instance with [GALACTICA](https://github.com/paperswithcode/galai)
25
- * API, including endpoints for websocket streaming ([see the examples](https://github.com/oobabooga/text-generation-webui/blob/main/api-examples))
26
-
27
- To learn how to use the various features, check out the Documentation: https://github.com/oobabooga/text-generation-webui/tree/main/docs
28
-
29
- ## Installation
30
-
31
- ### One-click installers
32
-
33
- | Windows | Linux | macOS | WSL |
34
- |--------|--------|--------|--------|
35
- | [oobabooga-windows.zip](https://github.com/oobabooga/text-generation-webui/releases/download/installers/oobabooga_windows.zip) | [oobabooga-linux.zip](https://github.com/oobabooga/text-generation-webui/releases/download/installers/oobabooga_linux.zip) |[oobabooga-macos.zip](https://github.com/oobabooga/text-generation-webui/releases/download/installers/oobabooga_macos.zip) | [oobabooga-wsl.zip](https://github.com/oobabooga/text-generation-webui/releases/download/installers/oobabooga_wsl.zip) |
36
-
37
- Just download the zip above, extract it, and double-click on "start". The web UI and all its dependencies will be installed in the same folder.
38
-
39
- * The source codes and more information can be found here: https://github.com/oobabooga/one-click-installers
40
- * There is no need to run the installers as admin.
41
- * Huge thanks to [@jllllll](https://github.com/jllllll), [@ClayShoaf](https://github.com/ClayShoaf), and [@xNul](https://github.com/xNul) for their contributions to these installers.
42
-
43
- ### Manual installation using Conda
44
-
45
- Recommended if you have some experience with the command-line.
46
-
47
- #### 0. Install Conda
48
-
49
- https://docs.conda.io/en/latest/miniconda.html
50
-
51
- On Linux or WSL, it can be automatically installed with these two commands ([source](https://educe-ubc.github.io/conda.html)):
52
-
53
- ```
54
- curl -sL "https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh" > "Miniconda3.sh"
55
- bash Miniconda3.sh
56
- ```
57
-
58
- #### 1. Create a new conda environment
59
-
60
- ```
61
- conda create -n textgen python=3.10.9
62
- conda activate textgen
63
- ```
64
-
65
- #### 2. Install Pytorch
66
-
67
- | System | GPU | Command |
68
- |--------|---------|---------|
69
- | Linux/WSL | NVIDIA | `pip3 install torch torchvision torchaudio` |
70
- | Linux/WSL | CPU only | `pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu` |
71
- | Linux | AMD | `pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.4.2` |
72
- | MacOS + MPS | Any | `pip3 install torch torchvision torchaudio` |
73
- | Windows | NVIDIA | `pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117` |
74
- | Windows | CPU only | `pip3 install torch torchvision torchaudio` |
75
-
76
- The up-to-date commands can be found here: https://pytorch.org/get-started/locally/.
77
-
78
- #### 3. Install the web UI
79
-
80
- ```
81
- git clone https://github.com/oobabooga/text-generation-webui
82
- cd text-generation-webui
83
- pip install -r requirements.txt
84
- ```
85
-
86
- #### AMD, Metal, Intel Arc, and CPUs without AVCX2
87
-
88
- 1) Replace the last command above with
89
-
90
- ```
91
- pip install -r requirements_nocuda.txt
92
- ```
93
-
94
- 2) Manually install llama-cpp-python using the appropriate command for your hardware: [Installation from PyPI](https://github.com/abetlen/llama-cpp-python#installation-from-pypi).
95
-
96
- 3) AMD: Manually install AutoGPTQ: [Installation](https://github.com/PanQiWei/AutoGPTQ#installation).
97
-
98
- 4) AMD: 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):
99
-
100
- ```
101
- cd text-generation-webui
102
- mkdir repositories
103
- cd repositories
104
- git clone https://github.com/turboderp/exllama
105
- ```
106
-
107
- #### bitsandbytes on older NVIDIA GPUs
108
-
109
- bitsandbytes >= 0.39 may not work. In that case, to use `--load-in-8bit`, you may have to downgrade like this:
110
-
111
- * Linux: `pip install bitsandbytes==0.38.1`
112
- * Windows: `pip install https://github.com/jllllll/bitsandbytes-windows-webui/raw/main/bitsandbytes-0.38.1-py3-none-any.whl`
113
-
114
- ### Alternative: Docker
115
-
116
- ```
117
- ln -s docker/{Dockerfile,docker-compose.yml,.dockerignore} .
118
- cp docker/.env.example .env
119
- # Edit .env and set TORCH_CUDA_ARCH_LIST based on your GPU model
120
- docker compose up --build
121
- ```
122
-
123
- * You need to have docker compose v2.17 or higher installed. See [this guide](https://github.com/oobabooga/text-generation-webui/blob/main/docs/Docker.md) for instructions.
124
- * For additional docker files, check out [this repository](https://github.com/Atinoda/text-generation-webui-docker).
125
-
126
- ### Updating the requirements
127
-
128
- From time to time, the `requirements.txt` changes. To update, use these commands:
129
-
130
- ```
131
- conda activate textgen
132
- cd text-generation-webui
133
- pip install -r requirements.txt --upgrade
134
- ```
135
-
136
- ## Downloading models
137
-
138
- 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).
139
-
140
- * Transformers or GPTQ models are made of several files and must be placed in a subfolder. Example:
141
-
142
- ```
143
- text-generation-webui
144
- ├── models
145
- │   ├── lmsys_vicuna-33b-v1.3
146
- │   │   ├── config.json
147
- │   │   ├── generation_config.json
148
- │   │   ├── pytorch_model-00001-of-00007.bin
149
- │   │   ├── pytorch_model-00002-of-00007.bin
150
- │   │   ├── pytorch_model-00003-of-00007.bin
151
- │   │   ├── pytorch_model-00004-of-00007.bin
152
- │   │   ├── pytorch_model-00005-of-00007.bin
153
- │   │   ├── pytorch_model-00006-of-00007.bin
154
- │   │   ├── pytorch_model-00007-of-00007.bin
155
- │   │   ├── pytorch_model.bin.index.json
156
- │   │   ├── special_tokens_map.json
157
- │   │   ├── tokenizer_config.json
158
- │   │   └── tokenizer.model
159
- ```
160
-
161
- * GGML/GGUF models are a single file and should be placed directly into `models`. Example:
162
-
163
- ```
164
- text-generation-webui
165
- ├── models
166
- │   ├── llama-13b.ggmlv3.q4_K_M.bin
167
- ```
168
-
169
- 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).
170
-
171
- #### GPT-4chan
172
-
173
- <details>
174
- <summary>
175
- Instructions
176
- </summary>
177
-
178
- [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:
179
-
180
- * Torrent: [16-bit](https://archive.org/details/gpt4chan_model_float16) / [32-bit](https://archive.org/details/gpt4chan_model)
181
- * Direct download: [16-bit](https://theswissbay.ch/pdf/_notpdf_/gpt4chan_model_float16/) / [32-bit](https://theswissbay.ch/pdf/_notpdf_/gpt4chan_model/)
182
-
183
- 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.
184
-
185
- After downloading the model, follow these steps:
186
-
187
- 1. Place the files under `models/gpt4chan_model_float16` or `models/gpt4chan_model`.
188
- 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).
189
- 3. Download GPT-J 6B's tokenizer files (they will be automatically detected when you attempt to load GPT-4chan):
190
-
191
- ```
192
- python download-model.py EleutherAI/gpt-j-6B --text-only
193
- ```
194
-
195
- When you load this model in default or notebook modes, the "HTML" tab will show the generated text in 4chan format:
196
-
197
- ![Image3](https://github.com/oobabooga/screenshots/raw/main/gpt4chan.png)
198
-
199
- </details>
200
-
201
- ## Starting the web UI
202
-
203
- conda activate textgen
204
- cd text-generation-webui
205
- python server.py
206
-
207
- Then browse to
208
-
209
- `http://localhost:7860/?__theme=dark`
210
-
211
- Optionally, you can use the following command-line flags:
212
-
213
- #### Basic settings
214
-
215
- | Flag | Description |
216
- |--------------------------------------------|-------------|
217
- | `-h`, `--help` | Show this help message and exit. |
218
- | `--multi-user` | Multi-user mode. Chat histories are not saved or automatically loaded. WARNING: this is highly experimental. |
219
- | `--character CHARACTER` | The name of the character to load in chat mode by default. |
220
- | `--model MODEL` | Name of the model to load by default. |
221
- | `--lora LORA [LORA ...]` | The list of LoRAs to load. If you want to load more than one LoRA, write the names separated by spaces. |
222
- | `--model-dir MODEL_DIR` | Path to directory with all the models. |
223
- | `--lora-dir LORA_DIR` | Path to directory with all the loras. |
224
- | `--model-menu` | Show a model menu in the terminal when the web UI is first launched. |
225
- | `--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. |
226
- | `--extensions EXTENSIONS [EXTENSIONS ...]` | The list of extensions to load. If you want to load more than one extension, write the names separated by spaces. |
227
- | `--verbose` | Print the prompts to the terminal. |
228
-
229
- #### Model loader
230
-
231
- | Flag | Description |
232
- |--------------------------------------------|-------------|
233
- | `--loader LOADER` | Choose the model loader manually, otherwise, it will get autodetected. Valid options: transformers, autogptq, gptq-for-llama, exllama, exllama_hf, llamacpp, rwkv, ctransformers |
234
-
235
- #### Accelerate/transformers
236
-
237
- | Flag | Description |
238
- |---------------------------------------------|-------------|
239
- | `--cpu` | Use the CPU to generate text. Warning: Training on CPU is extremely slow.|
240
- | `--auto-devices` | Automatically split the model across the available GPU(s) and CPU. |
241
- | `--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`. |
242
- | `--cpu-memory CPU_MEMORY` | Maximum CPU memory in GiB to allocate for offloaded weights. Same as above.|
243
- | `--disk` | If the model is too large for your GPU(s) and CPU combined, send the remaining layers to the disk. |
244
- | `--disk-cache-dir DISK_CACHE_DIR` | Directory to save the disk cache to. Defaults to `cache/`. |
245
- | `--load-in-8bit` | Load the model with 8-bit precision (using bitsandbytes).|
246
- | `--bf16` | Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU. |
247
- | `--no-cache` | Set `use_cache` to False while generating text. This reduces the VRAM usage a bit with a performance cost. |
248
- | `--xformers` | Use xformer's memory efficient attention. This should increase your tokens/s. |
249
- | `--sdp-attention` | Use torch 2.0's sdp attention. |
250
- | `--trust-remote-code` | Set trust_remote_code=True while loading a model. Necessary for ChatGLM and Falcon. |
251
-
252
- #### Accelerate 4-bit
253
-
254
- ⚠️ Requires minimum compute of 7.0 on Windows at the moment.
255
-
256
- | Flag | Description |
257
- |---------------------------------------------|-------------|
258
- | `--load-in-4bit` | Load the model with 4-bit precision (using bitsandbytes). |
259
- | `--compute_dtype COMPUTE_DTYPE` | compute dtype for 4-bit. Valid options: bfloat16, float16, float32. |
260
- | `--quant_type QUANT_TYPE` | quant_type for 4-bit. Valid options: nf4, fp4. |
261
- | `--use_double_quant` | use_double_quant for 4-bit. |
262
-
263
- #### GGML/GGUF (for llama.cpp and ctransformers)
264
-
265
- | Flag | Description |
266
- |-------------|-------------|
267
- | `--threads` | Number of threads to use. |
268
- | `--n_batch` | Maximum number of prompt tokens to batch together when calling llama_eval. |
269
- | `--n-gpu-layers N_GPU_LAYERS` | Number of layers to offload to the GPU. Only works if llama-cpp-python was compiled with BLAS. Set this to 1000000000 to offload all layers to the GPU. |
270
- | `--n_ctx N_CTX` | Size of the prompt context. |
271
-
272
- #### llama.cpp
273
-
274
- | Flag | Description |
275
- |---------------|---------------|
276
- | `--no-mmap` | Prevent mmap from being used. |
277
- | `--mlock` | Force the system to keep the model in RAM. |
278
- | `--mul_mat_q` | Activate new mulmat kernels. |
279
- | `--cache-capacity CACHE_CAPACITY` | Maximum cache capacity. Examples: 2000MiB, 2GiB. When provided without units, bytes will be assumed. |
280
- | `--tensor_split TENSOR_SPLIT` | Split the model across multiple GPUs, comma-separated list of proportions, e.g. 18,17 |
281
- | `--llama_cpp_seed SEED` | Seed for llama-cpp models. Default 0 (random). |
282
- | `--n_gqa N_GQA` | GGML only (not used by GGUF): Grouped-Query Attention. Must be 8 for llama-2 70b. |
283
- | `--rms_norm_eps RMS_NORM_EPS` | GGML only (not used by GGUF): 5e-6 is a good value for llama-2 models. |
284
- | `--cpu` | Use the CPU version of llama-cpp-python instead of the GPU-accelerated version. |
285
- |`--cfg-cache` | llamacpp_HF: Create an additional cache for CFG negative prompts. |
286
-
287
- #### ctransformers
288
-
289
- | Flag | Description |
290
- |-------------|-------------|
291
- | `--model_type MODEL_TYPE` | Model type of pre-quantized model. Currently gpt2, gptj, gptneox, falcon, llama, mpt, starcoder (gptbigcode), dollyv2, and replit are supported. |
292
-
293
- #### AutoGPTQ
294
-
295
- | Flag | Description |
296
- |------------------|-------------|
297
- | `--triton` | Use triton. |
298
- | `--no_inject_fused_attention` | Disable the use of fused attention, which will use less VRAM at the cost of slower inference. |
299
- | `--no_inject_fused_mlp` | Triton mode only: disable the use of fused MLP, which will use less VRAM at the cost of slower inference. |
300
- | `--no_use_cuda_fp16` | This can make models faster on some systems. |
301
- | `--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. |
302
- | `--disable_exllama` | Disable ExLlama kernel, which can improve inference speed on some systems. |
303
-
304
- #### ExLlama
305
-
306
- | Flag | Description |
307
- |------------------|-------------|
308
- |`--gpu-split` | Comma-separated list of VRAM (in GB) to use per GPU device for model layers, e.g. `20,7,7` |
309
- |`--max_seq_len MAX_SEQ_LEN` | Maximum sequence length. |
310
- |`--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. |
311
-
312
- #### GPTQ-for-LLaMa
313
-
314
- | Flag | Description |
315
- |---------------------------|-------------|
316
- | `--wbits WBITS` | Load a pre-quantized model with specified precision in bits. 2, 3, 4 and 8 are supported. |
317
- | `--model_type MODEL_TYPE` | Model type of pre-quantized model. Currently LLaMA, OPT, and GPT-J are supported. |
318
- | `--groupsize GROUPSIZE` | Group size. |
319
- | `--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`. |
320
- | `--checkpoint CHECKPOINT` | The path to the quantized checkpoint file. If not specified, it will be automatically detected. |
321
- | `--monkey-patch` | Apply the monkey patch for using LoRAs with quantized models.
322
-
323
- #### DeepSpeed
324
-
325
- | Flag | Description |
326
- |---------------------------------------|-------------|
327
- | `--deepspeed` | Enable the use of DeepSpeed ZeRO-3 for inference via the Transformers integration. |
328
- | `--nvme-offload-dir NVME_OFFLOAD_DIR` | DeepSpeed: Directory to use for ZeRO-3 NVME offloading. |
329
- | `--local_rank LOCAL_RANK` | DeepSpeed: Optional argument for distributed setups. |
330
-
331
- #### RWKV
332
-
333
- | Flag | Description |
334
- |---------------------------------|-------------|
335
- | `--rwkv-strategy RWKV_STRATEGY` | RWKV: The strategy to use while loading the model. Examples: "cpu fp32", "cuda fp16", "cuda fp16i8". |
336
- | `--rwkv-cuda-on` | RWKV: Compile the CUDA kernel for better performance. |
337
-
338
- #### RoPE (for llama.cpp, ExLlama, and transformers)
339
-
340
- | Flag | Description |
341
- |------------------|-------------|
342
- | `--alpha_value ALPHA_VALUE` | Positional embeddings alpha factor for NTK RoPE scaling. Use either this or compress_pos_emb, not both. |
343
- | `--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). |
344
- | `--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. |
345
-
346
- #### Gradio
347
-
348
- | Flag | Description |
349
- |---------------------------------------|-------------|
350
- | `--listen` | Make the web UI reachable from your local network. |
351
- | `--listen-host LISTEN_HOST` | The hostname that the server will use. |
352
- | `--listen-port LISTEN_PORT` | The listening port that the server will use. |
353
- | `--share` | Create a public URL. This is useful for running the web UI on Google Colab or similar. |
354
- | `--auto-launch` | Open the web UI in the default browser upon launch. |
355
- | `--gradio-auth USER:PWD` | set gradio authentication like "username:password"; or comma-delimit multiple like "u1:p1,u2:p2,u3:p3" |
356
- | `--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" |
357
- | `--ssl-keyfile SSL_KEYFILE` | The path to the SSL certificate key file. |
358
- | `--ssl-certfile SSL_CERTFILE` | The path to the SSL certificate cert file. |
359
-
360
- #### API
361
-
362
- | Flag | Description |
363
- |---------------------------------------|-------------|
364
- | `--api` | Enable the API extension. |
365
- | `--public-api` | Create a public URL for the API using Cloudfare. |
366
- | `--public-api-id PUBLIC_API_ID` | Tunnel ID for named Cloudflare Tunnel. Use together with public-api option. |
367
- | `--api-blocking-port BLOCKING_PORT` | The listening port for the blocking API. |
368
- | `--api-streaming-port STREAMING_PORT` | The listening port for the streaming API. |
369
-
370
- #### Multimodal
371
-
372
- | Flag | Description |
373
- |---------------------------------------|-------------|
374
- | `--multimodal-pipeline PIPELINE` | The multimodal pipeline to use. Examples: `llava-7b`, `llava-13b`. |
375
-
376
- ## Presets
377
-
378
- Inference settings presets can be created under `presets/` as yaml files. These files are detected automatically at startup.
379
-
380
- The presets that are included by default are the result of a contest that received 7215 votes. More details can be found [here](https://github.com/oobabooga/oobabooga.github.io/blob/main/arena/results.md).
381
-
382
- ## Contributing
383
-
384
- If you would like to contribute to the project, check out the [Contributing guidelines](https://github.com/oobabooga/text-generation-webui/wiki/Contributing-guidelines).
385
-
386
- ## Community
387
-
388
- * Subreddit: https://www.reddit.com/r/oobabooga/
389
- * Discord: https://discord.gg/jwZCF2dPQN
390
-
391
- ## Acknowledgment
392
-
393
- 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.
 
1
+ title: {{title}}
2
+ emoji: {{emoji}}
3
+ colorFrom: {{colorFrom}}
4
+ colorTo: {{colorTo}}
5
+ sdk: {{sdk}}
6
+ sdk_version: {{sdkVersion}}
7
+ app_file: app.py