TheBloke commited on
Commit
d509cb9
1 Parent(s): 88a4e64

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +20 -18
README.md CHANGED
@@ -49,7 +49,7 @@ This repo contains GGUF format model files for [Eric Hartford's Wizard-Vicuna-30
49
  <!-- README_GGUF.md-about-gguf start -->
50
  ### About GGUF
51
 
52
- GGUF is a new format introduced by the llama.cpp team on August 21st 2023. It is a replacement for GGML, which is no longer supported by llama.cpp. GGUF offers numerous advantages over GGML, such as better tokenisation, and support for special tokens. It is also supports metadata, and is designed to be extensible.
53
 
54
  Here is an incomplate list of clients and libraries that are known to support GGUF:
55
 
@@ -87,7 +87,7 @@ A chat between a curious user and an artificial intelligence assistant. The assi
87
  <!-- compatibility_gguf start -->
88
  ## Compatibility
89
 
90
- These quantised GGUFv2 files are compatible with llama.cpp from August 27th onwards, as of commit [d0cee0d36d5be95a0d9088b674dbb27354107221](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221)
91
 
92
  They are also compatible with many third party UIs and libraries - please see the list at the top of this README.
93
 
@@ -151,7 +151,7 @@ Then click Download.
151
  I recommend using the `huggingface-hub` Python library:
152
 
153
  ```shell
154
- pip3 install huggingface-hub>=0.17.1
155
  ```
156
 
157
  Then you can download any individual model file to the current directory, at high speed, with a command like this:
@@ -180,25 +180,25 @@ pip3 install hf_transfer
180
  And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:
181
 
182
  ```shell
183
- HUGGINGFACE_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/Wizard-Vicuna-30B-Uncensored-GGUF Wizard-Vicuna-30B-Uncensored.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
184
  ```
185
 
186
- Windows CLI users: Use `set HUGGINGFACE_HUB_ENABLE_HF_TRANSFER=1` before running the download command.
187
  </details>
188
  <!-- README_GGUF.md-how-to-download end -->
189
 
190
  <!-- README_GGUF.md-how-to-run start -->
191
  ## Example `llama.cpp` command
192
 
193
- Make sure you are using `llama.cpp` from commit [d0cee0d36d5be95a0d9088b674dbb27354107221](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221) or later.
194
 
195
  ```shell
196
- ./main -ngl 32 -m Wizard-Vicuna-30B-Uncensored.Q4_K_M.gguf --color -c 4096 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions. USER: {prompt} ASSISTANT:"
197
  ```
198
 
199
  Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
200
 
201
- Change `-c 4096` to the desired sequence length. For extended sequence models - eg 8K, 16K, 32K - the necessary RoPE scaling parameters are read from the GGUF file and set by llama.cpp automatically.
202
 
203
  If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`
204
 
@@ -212,22 +212,24 @@ Further instructions here: [text-generation-webui/docs/llama.cpp.md](https://git
212
 
213
  You can use GGUF models from Python using the [llama-cpp-python](https://github.com/abetlen/llama-cpp-python) or [ctransformers](https://github.com/marella/ctransformers) libraries.
214
 
215
- ### How to load this model from Python using ctransformers
216
 
217
  #### First install the package
218
 
219
- ```bash
 
 
220
  # Base ctransformers with no GPU acceleration
221
- pip install ctransformers>=0.2.24
222
  # Or with CUDA GPU acceleration
223
- pip install ctransformers[cuda]>=0.2.24
224
- # Or with ROCm GPU acceleration
225
- CT_HIPBLAS=1 pip install ctransformers>=0.2.24 --no-binary ctransformers
226
- # Or with Metal GPU acceleration for macOS systems
227
- CT_METAL=1 pip install ctransformers>=0.2.24 --no-binary ctransformers
228
  ```
229
 
230
- #### Simple example code to load one of these GGUF models
231
 
232
  ```python
233
  from ctransformers import AutoModelForCausalLM
@@ -240,7 +242,7 @@ print(llm("AI is going to"))
240
 
241
  ## How to use with LangChain
242
 
243
- Here's guides on using llama-cpp-python or ctransformers with LangChain:
244
 
245
  * [LangChain + llama-cpp-python](https://python.langchain.com/docs/integrations/llms/llamacpp)
246
  * [LangChain + ctransformers](https://python.langchain.com/docs/integrations/providers/ctransformers)
 
49
  <!-- README_GGUF.md-about-gguf start -->
50
  ### About GGUF
51
 
52
+ GGUF is a new format introduced by the llama.cpp team on August 21st 2023. It is a replacement for GGML, which is no longer supported by llama.cpp.
53
 
54
  Here is an incomplate list of clients and libraries that are known to support GGUF:
55
 
 
87
  <!-- compatibility_gguf start -->
88
  ## Compatibility
89
 
90
+ These quantised GGUFv2 files are compatible with llama.cpp from August 27th onwards, as of commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221)
91
 
92
  They are also compatible with many third party UIs and libraries - please see the list at the top of this README.
93
 
 
151
  I recommend using the `huggingface-hub` Python library:
152
 
153
  ```shell
154
+ pip3 install huggingface-hub
155
  ```
156
 
157
  Then you can download any individual model file to the current directory, at high speed, with a command like this:
 
180
  And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:
181
 
182
  ```shell
183
+ HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/Wizard-Vicuna-30B-Uncensored-GGUF Wizard-Vicuna-30B-Uncensored.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
184
  ```
185
 
186
+ Windows Command Line users: You can set the environment variable by running `set HF_HUB_ENABLE_HF_TRANSFER=1` before the download command.
187
  </details>
188
  <!-- README_GGUF.md-how-to-download end -->
189
 
190
  <!-- README_GGUF.md-how-to-run start -->
191
  ## Example `llama.cpp` command
192
 
193
+ Make sure you are using `llama.cpp` from commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221) or later.
194
 
195
  ```shell
196
+ ./main -ngl 32 -m Wizard-Vicuna-30B-Uncensored.Q4_K_M.gguf --color -c 2048 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions. USER: {prompt} ASSISTANT:"
197
  ```
198
 
199
  Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
200
 
201
+ Change `-c 2048` to the desired sequence length. For extended sequence models - eg 8K, 16K, 32K - the necessary RoPE scaling parameters are read from the GGUF file and set by llama.cpp automatically.
202
 
203
  If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`
204
 
 
212
 
213
  You can use GGUF models from Python using the [llama-cpp-python](https://github.com/abetlen/llama-cpp-python) or [ctransformers](https://github.com/marella/ctransformers) libraries.
214
 
215
+ ### How to load this model in Python code, using ctransformers
216
 
217
  #### First install the package
218
 
219
+ Run one of the following commands, according to your system:
220
+
221
+ ```shell
222
  # Base ctransformers with no GPU acceleration
223
+ pip install ctransformers
224
  # Or with CUDA GPU acceleration
225
+ pip install ctransformers[cuda]
226
+ # Or with AMD ROCm GPU acceleration (Linux only)
227
+ CT_HIPBLAS=1 pip install ctransformers --no-binary ctransformers
228
+ # Or with Metal GPU acceleration for macOS systems only
229
+ CT_METAL=1 pip install ctransformers --no-binary ctransformers
230
  ```
231
 
232
+ #### Simple ctransformers example code
233
 
234
  ```python
235
  from ctransformers import AutoModelForCausalLM
 
242
 
243
  ## How to use with LangChain
244
 
245
+ Here are guides on using llama-cpp-python and ctransformers with LangChain:
246
 
247
  * [LangChain + llama-cpp-python](https://python.langchain.com/docs/integrations/llms/llamacpp)
248
  * [LangChain + ctransformers](https://python.langchain.com/docs/integrations/providers/ctransformers)