MaziyarPanahi commited on
Commit
4f839a7
1 Parent(s): af46a4b

Upload folder using huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +248 -0
README.md ADDED
@@ -0,0 +1,248 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - quantized
4
+ - 2-bit
5
+ - 3-bit
6
+ - 4-bit
7
+ - 5-bit
8
+ - 6-bit
9
+ - 8-bit
10
+ - GGUF
11
+ - transformers
12
+ - safetensors
13
+ - gguf
14
+ - gemma
15
+ - text-generation
16
+ - conversational
17
+ - arxiv:2312.11805
18
+ - arxiv:2009.03300
19
+ - arxiv:1905.07830
20
+ - arxiv:1911.11641
21
+ - arxiv:1904.09728
22
+ - arxiv:1905.10044
23
+ - arxiv:1907.10641
24
+ - arxiv:1811.00937
25
+ - arxiv:1809.02789
26
+ - arxiv:1911.01547
27
+ - arxiv:1705.03551
28
+ - arxiv:2107.03374
29
+ - arxiv:2108.07732
30
+ - arxiv:2110.14168
31
+ - arxiv:2304.06364
32
+ - arxiv:2206.04615
33
+ - arxiv:1804.06876
34
+ - arxiv:2110.08193
35
+ - arxiv:2009.11462
36
+ - arxiv:2101.11718
37
+ - arxiv:1804.09301
38
+ - arxiv:2109.07958
39
+ - arxiv:2203.09509
40
+ - license:other
41
+ - autotrain_compatible
42
+ - endpoints_compatible
43
+ - text-generation-inference
44
+ - region:us
45
+ - text-generation
46
+ model_name: gemma-2b-it-GGUF
47
+ base_model: google/gemma-2b-it
48
+ inference: false
49
+ model_creator: google
50
+ pipeline_tag: text-generation
51
+ quantized_by: MaziyarPanahi
52
+ ---
53
+ # [MaziyarPanahi/gemma-2b-it-GGUF](https://huggingface.co/MaziyarPanahi/gemma-2b-it-GGUF)
54
+ - Model creator: [google](https://huggingface.co/google)
55
+ - Original model: [google/gemma-2b-it](https://huggingface.co/google/gemma-2b-it)
56
+
57
+ ## Description
58
+ [MaziyarPanahi/gemma-2b-it-GGUF](https://huggingface.co/MaziyarPanahi/gemma-2b-it-GGUF) contains GGUF format model files for [google/gemma-2b-it](https://huggingface.co/google/gemma-2b-it).
59
+
60
+ ## How to use
61
+ Thanks to [TheBloke](https://huggingface.co/TheBloke) for preparing an amazing README on how to use GGUF models:
62
+
63
+ ### About GGUF
64
+
65
+ 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.
66
+
67
+ Here is an incomplete list of clients and libraries that are known to support GGUF:
68
+
69
+ * [llama.cpp](https://github.com/ggerganov/llama.cpp). The source project for GGUF. Offers a CLI and a server option.
70
+ * [text-generation-webui](https://github.com/oobabooga/text-generation-webui), the most widely used web UI, with many features and powerful extensions. Supports GPU acceleration.
71
+ * [KoboldCpp](https://github.com/LostRuins/koboldcpp), a fully featured web UI, with GPU accel across all platforms and GPU architectures. Especially good for story telling.
72
+ * [GPT4All](https://gpt4all.io/index.html), a free and open source local running GUI, supporting Windows, Linux and macOS with full GPU accel.
73
+ * [LM Studio](https://lmstudio.ai/), an easy-to-use and powerful local GUI for Windows and macOS (Silicon), with GPU acceleration. Linux available, in beta as of 27/11/2023.
74
+ * [LoLLMS Web UI](https://github.com/ParisNeo/lollms-webui), a great web UI with many interesting and unique features, including a full model library for easy model selection.
75
+ * [Faraday.dev](https://faraday.dev/), an attractive and easy to use character-based chat GUI for Windows and macOS (both Silicon and Intel), with GPU acceleration.
76
+ * [llama-cpp-python](https://github.com/abetlen/llama-cpp-python), a Python library with GPU accel, LangChain support, and OpenAI-compatible API server.
77
+ * [candle](https://github.com/huggingface/candle), a Rust ML framework with a focus on performance, including GPU support, and ease of use.
78
+ * [ctransformers](https://github.com/marella/ctransformers), a Python library with GPU accel, LangChain support, and OpenAI-compatible AI server. Note, as of time of writing (November 27th 2023), ctransformers has not been updated in a long time and does not support many recent models.
79
+
80
+ ### Explanation of quantisation methods
81
+
82
+ <details>
83
+ <summary>Click to see details</summary>
84
+
85
+ The new methods available are:
86
+
87
+ * GGML_TYPE_Q2_K - "type-1" 2-bit quantization in super-blocks containing 16 blocks, each block having 16 weight. Block scales and mins are quantized with 4 bits. This ends up effectively using 2.5625 bits per weight (bpw)
88
+ * GGML_TYPE_Q3_K - "type-0" 3-bit quantization in super-blocks containing 16 blocks, each block having 16 weights. Scales are quantized with 6 bits. This end up using 3.4375 bpw.
89
+ * GGML_TYPE_Q4_K - "type-1" 4-bit quantization in super-blocks containing 8 blocks, each block having 32 weights. Scales and mins are quantized with 6 bits. This ends up using 4.5 bpw.
90
+ * GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw
91
+ * GGML_TYPE_Q6_K - "type-0" 6-bit quantization. Super-blocks with 16 blocks, each block having 16 weights. Scales are quantized with 8 bits. This ends up using 6.5625 bpw
92
+
93
+ ## How to download GGUF files
94
+
95
+ **Note for manual downloaders:** You almost never want to clone the entire repo! Multiple different quantisation formats are provided, and most users only want to pick and download a single file.
96
+
97
+ The following clients/libraries will automatically download models for you, providing a list of available models to choose from:
98
+
99
+ * LM Studio
100
+ * LoLLMS Web UI
101
+ * Faraday.dev
102
+
103
+ ### In `text-generation-webui`
104
+
105
+ Under Download Model, you can enter the model repo: [MaziyarPanahi/gemma-2b-it-GGUF](https://huggingface.co/MaziyarPanahi/gemma-2b-it-GGUF) and below it, a specific filename to download, such as: gemma-2b-it-GGUF.Q4_K_M.gguf.
106
+
107
+ Then click Download.
108
+
109
+ ### On the command line, including multiple files at once
110
+
111
+ I recommend using the `huggingface-hub` Python library:
112
+
113
+ ```shell
114
+ pip3 install huggingface-hub
115
+ ```
116
+
117
+ Then you can download any individual model file to the current directory, at high speed, with a command like this:
118
+
119
+ ```shell
120
+ huggingface-cli download MaziyarPanahi/gemma-2b-it-GGUF gemma-2b-it-GGUF.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
121
+ ```
122
+ </details>
123
+ <details>
124
+ <summary>More advanced huggingface-cli download usage (click to read)</summary>
125
+
126
+ You can also download multiple files at once with a pattern:
127
+
128
+ ```shell
129
+ huggingface-cli download [MaziyarPanahi/gemma-2b-it-GGUF](https://huggingface.co/MaziyarPanahi/gemma-2b-it-GGUF) --local-dir . --local-dir-use-symlinks False --include='*Q4_K*gguf'
130
+ ```
131
+
132
+ For more documentation on downloading with `huggingface-cli`, please see: [HF -> Hub Python Library -> Download files -> Download from the CLI](https://huggingface.co/docs/huggingface_hub/guides/download#download-from-the-cli).
133
+
134
+ To accelerate downloads on fast connections (1Gbit/s or higher), install `hf_transfer`:
135
+
136
+ ```shell
137
+ pip3 install hf_transfer
138
+ ```
139
+
140
+ And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:
141
+
142
+ ```shell
143
+ HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download MaziyarPanahi/gemma-2b-it-GGUF gemma-2b-it-GGUF.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
144
+ ```
145
+
146
+ Windows Command Line users: You can set the environment variable by running `set HF_HUB_ENABLE_HF_TRANSFER=1` before the download command.
147
+ </details>
148
+
149
+ ## Example `llama.cpp` command
150
+
151
+ Make sure you are using `llama.cpp` from commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221) or later.
152
+
153
+ ```shell
154
+ ./main -ngl 35 -m gemma-2b-it-GGUF.Q4_K_M.gguf --color -c 32768 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "<|im_start|>system
155
+ {system_message}<|im_end|>
156
+ <|im_start|>user
157
+ {prompt}<|im_end|>
158
+ <|im_start|>assistant"
159
+ ```
160
+
161
+ Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
162
+
163
+ Change `-c 32768` 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. Note that longer sequence lengths require much more resources, so you may need to reduce this value.
164
+
165
+ If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`
166
+
167
+ For other parameters and how to use them, please refer to [the llama.cpp documentation](https://github.com/ggerganov/llama.cpp/blob/master/examples/main/README.md)
168
+
169
+ ## How to run in `text-generation-webui`
170
+
171
+ Further instructions can be found in the text-generation-webui documentation, here: [text-generation-webui/docs/04 ‐ Model Tab.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/04%20%E2%80%90%20Model%20Tab.md#llamacpp).
172
+
173
+ ## How to run from Python code
174
+
175
+ 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. Note that at the time of writing (Nov 27th 2023), ctransformers has not been updated for some time and is not compatible with some recent models. Therefore I recommend you use llama-cpp-python.
176
+
177
+ ### How to load this model in Python code, using llama-cpp-python
178
+
179
+ For full documentation, please see: [llama-cpp-python docs](https://abetlen.github.io/llama-cpp-python/).
180
+
181
+ #### First install the package
182
+
183
+ Run one of the following commands, according to your system:
184
+
185
+ ```shell
186
+ # Base ctransformers with no GPU acceleration
187
+ pip install llama-cpp-python
188
+ # With NVidia CUDA acceleration
189
+ CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python
190
+ # Or with OpenBLAS acceleration
191
+ CMAKE_ARGS="-DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS" pip install llama-cpp-python
192
+ # Or with CLBLast acceleration
193
+ CMAKE_ARGS="-DLLAMA_CLBLAST=on" pip install llama-cpp-python
194
+ # Or with AMD ROCm GPU acceleration (Linux only)
195
+ CMAKE_ARGS="-DLLAMA_HIPBLAS=on" pip install llama-cpp-python
196
+ # Or with Metal GPU acceleration for macOS systems only
197
+ CMAKE_ARGS="-DLLAMA_METAL=on" pip install llama-cpp-python
198
+
199
+ # In windows, to set the variables CMAKE_ARGS in PowerShell, follow this format; eg for NVidia CUDA:
200
+ $env:CMAKE_ARGS = "-DLLAMA_OPENBLAS=on"
201
+ pip install llama-cpp-python
202
+ ```
203
+
204
+ #### Simple llama-cpp-python example code
205
+
206
+ ```python
207
+ from llama_cpp import Llama
208
+
209
+ # Set gpu_layers to the number of layers to offload to GPU. Set to 0 if no GPU acceleration is available on your system.
210
+ llm = Llama(
211
+ model_path="./gemma-2b-it-GGUF.Q4_K_M.gguf", # Download the model file first
212
+ n_ctx=32768, # The max sequence length to use - note that longer sequence lengths require much more resources
213
+ n_threads=8, # The number of CPU threads to use, tailor to your system and the resulting performance
214
+ n_gpu_layers=35 # The number of layers to offload to GPU, if you have GPU acceleration available
215
+ )
216
+
217
+ # Simple inference example
218
+ output = llm(
219
+ "<|im_start|>system
220
+ {system_message}<|im_end|>
221
+ <|im_start|>user
222
+ {prompt}<|im_end|>
223
+ <|im_start|>assistant", # Prompt
224
+ max_tokens=512, # Generate up to 512 tokens
225
+ stop=["</s>"], # Example stop token - not necessarily correct for this specific model! Please check before using.
226
+ echo=True # Whether to echo the prompt
227
+ )
228
+
229
+ # Chat Completion API
230
+
231
+ llm = Llama(model_path="./gemma-2b-it-GGUF.Q4_K_M.gguf", chat_format="llama-2") # Set chat_format according to the model you are using
232
+ llm.create_chat_completion(
233
+ messages = [
234
+ {"role": "system", "content": "You are a story writing assistant."},
235
+ {
236
+ "role": "user",
237
+ "content": "Write a story about llamas."
238
+ }
239
+ ]
240
+ )
241
+ ```
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)