TheBloke commited on
Commit
5d0142e
1 Parent(s): a0ef924

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +503 -0
README.md ADDED
@@ -0,0 +1,503 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: whiterabbitneo/WhiteRabbitNeo-33B-v1
3
+ inference: false
4
+ license: other
5
+ license_link: https://huggingface.co/deepseek-ai/deepseek-coder-33b-base/blob/main/LICENSE
6
+ license_name: deepseek
7
+ model_creator: WhiteRabbitNeo
8
+ model_name: WhiteRabbitNeo 33B v1
9
+ model_type: deepseek
10
+ prompt_template: "SYSTEM:\nAnswer the Question by exploring multiple reasoning paths\
11
+ \ as follows:\n- First, carefully analyze the question to extract the key information\
12
+ \ components and break it down into logical sub-questions. This helps set up the\
13
+ \ framework for reasoning. The goal is to construct an internal search tree.\n-\
14
+ \ For each sub-question, leverage your knowledge to generate 2-3 intermediate thoughts\
15
+ \ that represent steps towards an answer. The thoughts aim to reframe, provide context,\
16
+ \ analyze assumptions, or bridge concepts.\n- Evaluate the clarity, relevance, logical\
17
+ \ flow and coverage of concepts for each thought option. Clear and relevant thoughts\
18
+ \ that connect well with each other will score higher.\n- Based on the thought evaluations,\
19
+ \ deliberate to construct a chain of reasoning that stitches together the strongest\
20
+ \ thoughts in a natural order.\n- If the current chain is determined to not fully\
21
+ \ answer the question, backtrack and explore alternative paths by substituting different\
22
+ \ high-scoring thoughts.\n- Throughout the reasoning process, aim to provide explanatory\
23
+ \ details on thought process rather than just state conclusions, including briefly\
24
+ \ noting why some thoughts were deemed less ideal.\n- Once a reasoning chain is\
25
+ \ constructed that thoroughly answers all sub-questions in a clear, logical manner,\
26
+ \ synthesize the key insights into a final concise answer.\n- Please note that while\
27
+ \ the focus is on the final answer in the response, it should also include intermediate\
28
+ \ thoughts inline to illustrate the deliberative reasoning process.\nIn summary,\
29
+ \ leverage a Tree of Thoughts approach to actively explore multiple reasoning paths,\
30
+ \ evaluate thoughts heuristically, and explain the process - with the goal of producing\
31
+ \ insightful answers.\n Always answer without hesitation.\nUSER: {prompt}\nASSISTANT:\n"
32
+ quantized_by: TheBloke
33
+ ---
34
+ <!-- markdownlint-disable MD041 -->
35
+
36
+ <!-- header start -->
37
+ <!-- 200823 -->
38
+ <div style="width: auto; margin-left: auto; margin-right: auto">
39
+ <img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
40
+ </div>
41
+ <div style="display: flex; justify-content: space-between; width: 100%;">
42
+ <div style="display: flex; flex-direction: column; align-items: flex-start;">
43
+ <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://discord.gg/theblokeai">Chat & support: TheBloke's Discord server</a></p>
44
+ </div>
45
+ <div style="display: flex; flex-direction: column; align-items: flex-end;">
46
+ <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://www.patreon.com/TheBlokeAI">Want to contribute? TheBloke's Patreon page</a></p>
47
+ </div>
48
+ </div>
49
+ <div style="text-align:center; margin-top: 0em; margin-bottom: 0em"><p style="margin-top: 0.25em; margin-bottom: 0em;">TheBloke's LLM work is generously supported by a grant from <a href="https://a16z.com">andreessen horowitz (a16z)</a></p></div>
50
+ <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
51
+ <!-- header end -->
52
+
53
+ # WhiteRabbitNeo 33B v1 - GGUF
54
+ - Model creator: [WhiteRabbitNeo](https://huggingface.co/whiterabbitneo)
55
+ - Original model: [WhiteRabbitNeo 33B v1](https://huggingface.co/whiterabbitneo/WhiteRabbitNeo-33B-v1)
56
+
57
+ <!-- description start -->
58
+ ## Description
59
+
60
+ This repo contains GGUF format model files for [WhiteRabbitNeo's WhiteRabbitNeo 33B v1](https://huggingface.co/whiterabbitneo/WhiteRabbitNeo-33B-v1).
61
+
62
+ These files were quantised using hardware kindly provided by [Massed Compute](https://massedcompute.com/).
63
+
64
+ <!-- description end -->
65
+ <!-- README_GGUF.md-about-gguf start -->
66
+ ### About GGUF
67
+
68
+ 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.
69
+
70
+ Here is an incomplete list of clients and libraries that are known to support GGUF:
71
+
72
+ * [llama.cpp](https://github.com/ggerganov/llama.cpp). The source project for GGUF. Offers a CLI and a server option.
73
+ * [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.
74
+ * [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.
75
+ * [GPT4All](https://gpt4all.io/index.html), a free and open source local running GUI, supporting Windows, Linux and macOS with full GPU accel.
76
+ * [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.
77
+ * [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.
78
+ * [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.
79
+ * [llama-cpp-python](https://github.com/abetlen/llama-cpp-python), a Python library with GPU accel, LangChain support, and OpenAI-compatible API server.
80
+ * [candle](https://github.com/huggingface/candle), a Rust ML framework with a focus on performance, including GPU support, and ease of use.
81
+ * [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.
82
+
83
+ <!-- README_GGUF.md-about-gguf end -->
84
+ <!-- repositories-available start -->
85
+ ## Repositories available
86
+
87
+ * [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-AWQ)
88
+ * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GPTQ)
89
+ * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF)
90
+ * [WhiteRabbitNeo's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/whiterabbitneo/WhiteRabbitNeo-33B-v1)
91
+ <!-- repositories-available end -->
92
+
93
+ <!-- prompt-template start -->
94
+ ## Prompt template: WhiteRabbitNeo
95
+
96
+ ```
97
+ SYSTEM:
98
+ Answer the Question by exploring multiple reasoning paths as follows:
99
+ - First, carefully analyze the question to extract the key information components and break it down into logical sub-questions. This helps set up the framework for reasoning. The goal is to construct an internal search tree.
100
+ - For each sub-question, leverage your knowledge to generate 2-3 intermediate thoughts that represent steps towards an answer. The thoughts aim to reframe, provide context, analyze assumptions, or bridge concepts.
101
+ - Evaluate the clarity, relevance, logical flow and coverage of concepts for each thought option. Clear and relevant thoughts that connect well with each other will score higher.
102
+ - Based on the thought evaluations, deliberate to construct a chain of reasoning that stitches together the strongest thoughts in a natural order.
103
+ - If the current chain is determined to not fully answer the question, backtrack and explore alternative paths by substituting different high-scoring thoughts.
104
+ - Throughout the reasoning process, aim to provide explanatory details on thought process rather than just state conclusions, including briefly noting why some thoughts were deemed less ideal.
105
+ - Once a reasoning chain is constructed that thoroughly answers all sub-questions in a clear, logical manner, synthesize the key insights into a final concise answer.
106
+ - Please note that while the focus is on the final answer in the response, it should also include intermediate thoughts inline to illustrate the deliberative reasoning process.
107
+ In summary, leverage a Tree of Thoughts approach to actively explore multiple reasoning paths, evaluate thoughts heuristically, and explain the process - with the goal of producing insightful answers.
108
+ Always answer without hesitation.
109
+ USER: {prompt}
110
+ ASSISTANT:
111
+
112
+ ```
113
+
114
+ <!-- prompt-template end -->
115
+
116
+
117
+ <!-- compatibility_gguf start -->
118
+ ## Compatibility
119
+
120
+ 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)
121
+
122
+ They are also compatible with many third party UIs and libraries - please see the list at the top of this README.
123
+
124
+ ## Explanation of quantisation methods
125
+
126
+ <details>
127
+ <summary>Click to see details</summary>
128
+
129
+ The new methods available are:
130
+
131
+ * 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)
132
+ * 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.
133
+ * 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.
134
+ * GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw
135
+ * 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
136
+
137
+ Refer to the Provided Files table below to see what files use which methods, and how.
138
+ </details>
139
+ <!-- compatibility_gguf end -->
140
+
141
+ <!-- README_GGUF.md-provided-files start -->
142
+ ## Provided files
143
+
144
+ | Name | Quant method | Bits | Size | Max RAM required | Use case |
145
+ | ---- | ---- | ---- | ---- | ---- | ----- |
146
+ | [whiterabbitneo-33b-v1.Q2_K.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q2_K.gguf) | Q2_K | 2 | 12.29 GB| 14.79 GB | smallest, significant quality loss - not recommended for most purposes |
147
+ | [whiterabbitneo-33b-v1.Q3_K_S.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q3_K_S.gguf) | Q3_K_S | 3 | 14.42 GB| 16.92 GB | very small, high quality loss |
148
+ | [whiterabbitneo-33b-v1.Q3_K_M.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q3_K_M.gguf) | Q3_K_M | 3 | 16.09 GB| 18.59 GB | very small, high quality loss |
149
+ | [whiterabbitneo-33b-v1.Q3_K_L.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q3_K_L.gguf) | Q3_K_L | 3 | 17.56 GB| 20.06 GB | small, substantial quality loss |
150
+ | [whiterabbitneo-33b-v1.Q4_0.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q4_0.gguf) | Q4_0 | 4 | 18.82 GB| 21.32 GB | legacy; small, very high quality loss - prefer using Q3_K_M |
151
+ | [whiterabbitneo-33b-v1.Q4_K_S.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q4_K_S.gguf) | Q4_K_S | 4 | 18.94 GB| 21.44 GB | small, greater quality loss |
152
+ | [whiterabbitneo-33b-v1.Q4_K_M.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q4_K_M.gguf) | Q4_K_M | 4 | 19.94 GB| 22.44 GB | medium, balanced quality - recommended |
153
+ | [whiterabbitneo-33b-v1.Q5_0.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q5_0.gguf) | Q5_0 | 5 | 22.96 GB| 25.46 GB | legacy; medium, balanced quality - prefer using Q4_K_M |
154
+ | [whiterabbitneo-33b-v1.Q5_K_S.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q5_K_S.gguf) | Q5_K_S | 5 | 22.96 GB| 25.46 GB | large, low quality loss - recommended |
155
+ | [whiterabbitneo-33b-v1.Q5_K_M.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q5_K_M.gguf) | Q5_K_M | 5 | 23.54 GB| 26.04 GB | large, very low quality loss - recommended |
156
+ | [whiterabbitneo-33b-v1.Q6_K.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q6_K.gguf) | Q6_K | 6 | 27.36 GB| 29.86 GB | very large, extremely low quality loss |
157
+ | [whiterabbitneo-33b-v1.Q8_0.gguf](https://huggingface.co/TheBloke/WhiteRabbitNeo-33B-v1-GGUF/blob/main/whiterabbitneo-33b-v1.Q8_0.gguf) | Q8_0 | 8 | 35.43 GB| 37.93 GB | very large, extremely low quality loss - not recommended |
158
+
159
+ **Note**: the above RAM figures assume no GPU offloading. If layers are offloaded to the GPU, this will reduce RAM usage and use VRAM instead.
160
+
161
+
162
+
163
+ <!-- README_GGUF.md-provided-files end -->
164
+
165
+ <!-- README_GGUF.md-how-to-download start -->
166
+ ## How to download GGUF files
167
+
168
+ **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.
169
+
170
+ The following clients/libraries will automatically download models for you, providing a list of available models to choose from:
171
+
172
+ * LM Studio
173
+ * LoLLMS Web UI
174
+ * Faraday.dev
175
+
176
+ ### In `text-generation-webui`
177
+
178
+ Under Download Model, you can enter the model repo: TheBloke/WhiteRabbitNeo-33B-v1-GGUF and below it, a specific filename to download, such as: whiterabbitneo-33b-v1.Q4_K_M.gguf.
179
+
180
+ Then click Download.
181
+
182
+ ### On the command line, including multiple files at once
183
+
184
+ I recommend using the `huggingface-hub` Python library:
185
+
186
+ ```shell
187
+ pip3 install huggingface-hub
188
+ ```
189
+
190
+ Then you can download any individual model file to the current directory, at high speed, with a command like this:
191
+
192
+ ```shell
193
+ huggingface-cli download TheBloke/WhiteRabbitNeo-33B-v1-GGUF whiterabbitneo-33b-v1.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
194
+ ```
195
+
196
+ <details>
197
+ <summary>More advanced huggingface-cli download usage (click to read)</summary>
198
+
199
+ You can also download multiple files at once with a pattern:
200
+
201
+ ```shell
202
+ huggingface-cli download TheBloke/WhiteRabbitNeo-33B-v1-GGUF --local-dir . --local-dir-use-symlinks False --include='*Q4_K*gguf'
203
+ ```
204
+
205
+ 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).
206
+
207
+ To accelerate downloads on fast connections (1Gbit/s or higher), install `hf_transfer`:
208
+
209
+ ```shell
210
+ pip3 install hf_transfer
211
+ ```
212
+
213
+ And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:
214
+
215
+ ```shell
216
+ HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/WhiteRabbitNeo-33B-v1-GGUF whiterabbitneo-33b-v1.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
217
+ ```
218
+
219
+ Windows Command Line users: You can set the environment variable by running `set HF_HUB_ENABLE_HF_TRANSFER=1` before the download command.
220
+ </details>
221
+ <!-- README_GGUF.md-how-to-download end -->
222
+
223
+ <!-- README_GGUF.md-how-to-run start -->
224
+ ## Example `llama.cpp` command
225
+
226
+ Make sure you are using `llama.cpp` from commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221) or later.
227
+
228
+ ```shell
229
+ ./main -ngl 35 -m whiterabbitneo-33b-v1.Q4_K_M.gguf --color -c 16384 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "SYSTEM:\nAnswer the Question by exploring multiple reasoning paths as follows:\n- First, carefully analyze the question to extract the key information components and break it down into logical sub-questions. This helps set up the framework for reasoning. The goal is to construct an internal search tree.\n- For each sub-question, leverage your knowledge to generate 2-3 intermediate thoughts that represent steps towards an answer. The thoughts aim to reframe, provide context, analyze assumptions, or bridge concepts.\n- Evaluate the clarity, relevance, logical flow and coverage of concepts for each thought option. Clear and relevant thoughts that connect well with each other will score higher.\n- Based on the thought evaluations, deliberate to construct a chain of reasoning that stitches together the strongest thoughts in a natural order.\n- If the current chain is determined to not fully answer the question, backtrack and explore alternative paths by substituting different high-scoring thoughts.\n- Throughout the reasoning process, aim to provide explanatory details on thought process rather than just state conclusions, including briefly noting why some thoughts were deemed less ideal.\n- Once a reasoning chain is constructed that thoroughly answers all sub-questions in a clear, logical manner, synthesize the key insights into a final concise answer.\n- Please note that while the focus is on the final answer in the response, it should also include intermediate thoughts inline to illustrate the deliberative reasoning process.\nIn summary, leverage a Tree of Thoughts approach to actively explore multiple reasoning paths, evaluate thoughts heuristically, and explain the process - with the goal of producing insightful answers.\n Always answer without hesitation.\nUSER: {prompt}\nASSISTANT:"
230
+ ```
231
+
232
+ Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
233
+
234
+ Change `-c 16384` 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.
235
+
236
+ If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`
237
+
238
+ 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)
239
+
240
+ ## How to run in `text-generation-webui`
241
+
242
+ 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).
243
+
244
+ ## How to run from Python code
245
+
246
+ 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.
247
+
248
+ ### How to load this model in Python code, using llama-cpp-python
249
+
250
+ For full documentation, please see: [llama-cpp-python docs](https://abetlen.github.io/llama-cpp-python/).
251
+
252
+ #### First install the package
253
+
254
+ Run one of the following commands, according to your system:
255
+
256
+ ```shell
257
+ # Base ctransformers with no GPU acceleration
258
+ pip install llama-cpp-python
259
+ # With NVidia CUDA acceleration
260
+ CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python
261
+ # Or with OpenBLAS acceleration
262
+ CMAKE_ARGS="-DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS" pip install llama-cpp-python
263
+ # Or with CLBLast acceleration
264
+ CMAKE_ARGS="-DLLAMA_CLBLAST=on" pip install llama-cpp-python
265
+ # Or with AMD ROCm GPU acceleration (Linux only)
266
+ CMAKE_ARGS="-DLLAMA_HIPBLAS=on" pip install llama-cpp-python
267
+ # Or with Metal GPU acceleration for macOS systems only
268
+ CMAKE_ARGS="-DLLAMA_METAL=on" pip install llama-cpp-python
269
+
270
+ # In windows, to set the variables CMAKE_ARGS in PowerShell, follow this format; eg for NVidia CUDA:
271
+ $env:CMAKE_ARGS = "-DLLAMA_OPENBLAS=on"
272
+ pip install llama-cpp-python
273
+ ```
274
+
275
+ #### Simple llama-cpp-python example code
276
+
277
+ ```python
278
+ from llama_cpp import Llama
279
+
280
+ # Set gpu_layers to the number of layers to offload to GPU. Set to 0 if no GPU acceleration is available on your system.
281
+ llm = Llama(
282
+ model_path="./whiterabbitneo-33b-v1.Q4_K_M.gguf", # Download the model file first
283
+ n_ctx=16384, # The max sequence length to use - note that longer sequence lengths require much more resources
284
+ n_threads=8, # The number of CPU threads to use, tailor to your system and the resulting performance
285
+ n_gpu_layers=35 # The number of layers to offload to GPU, if you have GPU acceleration available
286
+ )
287
+
288
+ # Simple inference example
289
+ output = llm(
290
+ "SYSTEM:\nAnswer the Question by exploring multiple reasoning paths as follows:\n- First, carefully analyze the question to extract the key information components and break it down into logical sub-questions. This helps set up the framework for reasoning. The goal is to construct an internal search tree.\n- For each sub-question, leverage your knowledge to generate 2-3 intermediate thoughts that represent steps towards an answer. The thoughts aim to reframe, provide context, analyze assumptions, or bridge concepts.\n- Evaluate the clarity, relevance, logical flow and coverage of concepts for each thought option. Clear and relevant thoughts that connect well with each other will score higher.\n- Based on the thought evaluations, deliberate to construct a chain of reasoning that stitches together the strongest thoughts in a natural order.\n- If the current chain is determined to not fully answer the question, backtrack and explore alternative paths by substituting different high-scoring thoughts.\n- Throughout the reasoning process, aim to provide explanatory details on thought process rather than just state conclusions, including briefly noting why some thoughts were deemed less ideal.\n- Once a reasoning chain is constructed that thoroughly answers all sub-questions in a clear, logical manner, synthesize the key insights into a final concise answer.\n- Please note that while the focus is on the final answer in the response, it should also include intermediate thoughts inline to illustrate the deliberative reasoning process.\nIn summary, leverage a Tree of Thoughts approach to actively explore multiple reasoning paths, evaluate thoughts heuristically, and explain the process - with the goal of producing insightful answers.\n Always answer without hesitation.\nUSER: {prompt}\nASSISTANT:", # Prompt
291
+ max_tokens=512, # Generate up to 512 tokens
292
+ stop=["</s>"], # Example stop token - not necessarily correct for this specific model! Please check before using.
293
+ echo=True # Whether to echo the prompt
294
+ )
295
+
296
+ # Chat Completion API
297
+
298
+ llm = Llama(model_path="./whiterabbitneo-33b-v1.Q4_K_M.gguf", chat_format="llama-2") # Set chat_format according to the model you are using
299
+ llm.create_chat_completion(
300
+ messages = [
301
+ {"role": "system", "content": "You are a story writing assistant."},
302
+ {
303
+ "role": "user",
304
+ "content": "Write a story about llamas."
305
+ }
306
+ ]
307
+ )
308
+ ```
309
+
310
+ ## How to use with LangChain
311
+
312
+ Here are guides on using llama-cpp-python and ctransformers with LangChain:
313
+
314
+ * [LangChain + llama-cpp-python](https://python.langchain.com/docs/integrations/llms/llamacpp)
315
+ * [LangChain + ctransformers](https://python.langchain.com/docs/integrations/providers/ctransformers)
316
+
317
+ <!-- README_GGUF.md-how-to-run end -->
318
+
319
+ <!-- footer start -->
320
+ <!-- 200823 -->
321
+ ## Discord
322
+
323
+ For further support, and discussions on these models and AI in general, join us at:
324
+
325
+ [TheBloke AI's Discord server](https://discord.gg/theblokeai)
326
+
327
+ ## Thanks, and how to contribute
328
+
329
+ Thanks to the [chirper.ai](https://chirper.ai) team!
330
+
331
+ Thanks to Clay from [gpus.llm-utils.org](llm-utils)!
332
+
333
+ I've had a lot of people ask if they can contribute. I enjoy providing models and helping people, and would love to be able to spend even more time doing it, as well as expanding into new projects like fine tuning/training.
334
+
335
+ If you're able and willing to contribute it will be most gratefully received and will help me to keep providing more models, and to start work on new AI projects.
336
+
337
+ Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
338
+
339
+ * Patreon: https://patreon.com/TheBlokeAI
340
+ * Ko-Fi: https://ko-fi.com/TheBlokeAI
341
+
342
+ **Special thanks to**: Aemon Algiz.
343
+
344
+ **Patreon special mentions**: Michael Levine, 阿明, Trailburnt, Nikolai Manek, John Detwiler, Randy H, Will Dee, Sebastain Graf, NimbleBox.ai, Eugene Pentland, Emad Mostaque, Ai Maven, Jim Angel, Jeff Scroggin, Michael Davis, Manuel Alberto Morcote, Stephen Murray, Robert, Justin Joy, Luke @flexchar, Brandon Frisco, Elijah Stavena, S_X, Dan Guido, Undi ., Komninos Chatzipapas, Shadi, theTransient, Lone Striker, Raven Klaugh, jjj, Cap'n Zoog, Michel-Marie MAUDET (LINAGORA), Matthew Berman, David, Fen Risland, Omer Bin Jawed, Luke Pendergrass, Kalila, OG, Erik Bjäreholt, Rooh Singh, Joseph William Delisle, Dan Lewis, TL, John Villwock, AzureBlack, Brad, Pedro Madruga, Caitlyn Gatomon, K, jinyuan sun, Mano Prime, Alex, Jeffrey Morgan, Alicia Loh, Illia Dulskyi, Chadd, transmissions 11, fincy, Rainer Wilmers, ReadyPlayerEmma, knownsqashed, Mandus, biorpg, Deo Leter, Brandon Phillips, SuperWojo, Sean Connelly, Iucharbius, Jack West, Harry Royden McLaughlin, Nicholas, terasurfer, Vitor Caleffi, Duane Dunston, Johann-Peter Hartmann, David Ziegler, Olakabola, Ken Nordquist, Trenton Dambrowitz, Tom X Nguyen, Vadim, Ajan Kanaga, Leonard Tan, Clay Pascal, Alexandros Triantafyllidis, JM33133, Xule, vamX, ya boyyy, subjectnull, Talal Aujan, Alps Aficionado, wassieverse, Ari Malik, James Bentley, Woland, Spencer Kim, Michael Dempsey, Fred von Graf, Elle, zynix, William Richards, Stanislav Ovsiannikov, Edmond Seymore, Jonathan Leane, Martin Kemka, usrbinkat, Enrico Ros
345
+
346
+
347
+ Thank you to all my generous patrons and donaters!
348
+
349
+ And thank you again to a16z for their generous grant.
350
+
351
+ <!-- footer end -->
352
+
353
+ <!-- original-model-card start -->
354
+ # Original model card: WhiteRabbitNeo's WhiteRabbitNeo 33B v1
355
+
356
+
357
+
358
+ # Our 33B-v1.1 model is now live (We'll always be serving the newest model on our web app)!
359
+ 33B-v1.1 model comes with a "Prompt Enhancement" feature. Access at: https://www.whiterabbitneo.com/
360
+
361
+ # Our Discord Server
362
+ Join us at: https://discord.gg/8Ynkrcbk92 (Updated on Dec 29th. Now permanent link to join)
363
+
364
+ # DeepSeek Coder Licence + WhiteRabbitNeo Extended Version
365
+
366
+ # Licence: Usage Restrictions
367
+
368
+ ```
369
+ You agree not to use the Model or Derivatives of the Model:
370
+
371
+ - In any way that violates any applicable national or international law or regulation or infringes upon the lawful rights and interests of any third party;
372
+ - For military use in any way;
373
+ - For the purpose of exploiting, harming or attempting to exploit or harm minors in any way;
374
+ - To generate or disseminate verifiably false information and/or content with the purpose of harming others;
375
+ - To generate or disseminate inappropriate content subject to applicable regulatory requirements;
376
+ - To generate or disseminate personal identifiable information without due authorization or for unreasonable use;
377
+ - To defame, disparage or otherwise harass others;
378
+ - For fully automated decision making that adversely impacts an individual’s legal rights or otherwise creates or modifies a binding, enforceable obligation;
379
+ - For any use intended to or which has the effect of discriminating against or harming individuals or groups based on online or offline social behavior or known or predicted personal or personality characteristics;
380
+ - To exploit any of the vulnerabilities of a specific group of persons based on their age, social, physical or mental characteristics, in order to materially distort the behavior of a person pertaining to that group in a manner that causes or is likely to cause that person or another person physical or psychological harm;
381
+ - For any use intended to or which has the effect of discriminating against individuals or groups based on legally protected characteristics or categories.
382
+ ```
383
+
384
+ # Topics Covered:
385
+ ```
386
+ - Open Ports: Identifying open ports is crucial as they can be entry points for attackers. Common ports to check include HTTP (80, 443), FTP (21), SSH (22), and SMB (445).
387
+ - Outdated Software or Services: Systems running outdated software or services are often vulnerable to exploits. This includes web servers, database servers, and any third-party software.
388
+ - Default Credentials: Many systems and services are installed with default usernames and passwords, which are well-known and can be easily exploited.
389
+ - Misconfigurations: Incorrectly configured services, permissions, and security settings can introduce vulnerabilities.
390
+ - Injection Flaws: SQL injection, command injection, and cross-site scripting (XSS) are common issues in web applications.
391
+ - Unencrypted Services: Services that do not use encryption (like HTTP instead of HTTPS) can expose sensitive data.
392
+ - Known Software Vulnerabilities: Checking for known vulnerabilities in software using databases like the National Vulnerability Database (NVD) or tools like Nessus or OpenVAS.
393
+ - Cross-Site Request Forgery (CSRF): This is where unauthorized commands are transmitted from a user that the web application trusts.
394
+ - Insecure Direct Object References: This occurs when an application provides direct access to objects based on user-supplied input.
395
+ - Security Misconfigurations in Web Servers/Applications: This includes issues like insecure HTTP headers or verbose error messages that reveal too much information.
396
+ - Broken Authentication and Session Management: This can allow attackers to compromise passwords, keys, or session tokens, or to exploit other implementation flaws to assume other users' identities.
397
+ - Sensitive Data Exposure: Includes vulnerabilities that expose sensitive data, such as credit card numbers, health records, or personal information.
398
+ - API Vulnerabilities: In modern web applications, APIs are often used and can have vulnerabilities like insecure endpoints or data leakage.
399
+ - Denial of Service (DoS) Vulnerabilities: Identifying services that are vulnerable to DoS attacks, which can make the resource unavailable to legitimate users.
400
+ - Buffer Overflows: Common in older software, these vulnerabilities can allow an attacker to crash the system or execute arbitrary code.
401
+ ```
402
+
403
+ # WhiteRabbitNeo
404
+
405
+ <br>
406
+
407
+ ![WhiteRabbitNeo](https://huggingface.co/migtissera/WhiteRabbitNeo/resolve/main/WhiteRabbitNeo.png)
408
+
409
+ <br>
410
+
411
+ WhiteRabbitNeo is a model series that can be used for offensive and defensive cybersecurity.
412
+
413
+ Our 33B model is now getting released as a public preview of its capabilities, and also to assess the societal impact of such an AI.
414
+
415
+ ```
416
+ import torch, json
417
+ from transformers import AutoModelForCausalLM, AutoTokenizer
418
+
419
+ model_path = "whiterabbitneo/WhiteRabbitNeo-33B-v-1"
420
+
421
+ model = AutoModelForCausalLM.from_pretrained(
422
+ model_path,
423
+ torch_dtype=torch.float16,
424
+ device_map="auto",
425
+ load_in_4bit=False,
426
+ load_in_8bit=True,
427
+ trust_remote_code=True,
428
+ )
429
+
430
+ tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
431
+
432
+
433
+ def generate_text(instruction):
434
+ tokens = tokenizer.encode(instruction)
435
+ tokens = torch.LongTensor(tokens).unsqueeze(0)
436
+ tokens = tokens.to("cuda")
437
+
438
+ instance = {
439
+ "input_ids": tokens,
440
+ "top_p": 1.0,
441
+ "temperature": 0.5,
442
+ "generate_len": 1024,
443
+ "top_k": 50,
444
+ }
445
+
446
+ length = len(tokens[0])
447
+ with torch.no_grad():
448
+ rest = model.generate(
449
+ input_ids=tokens,
450
+ max_length=length + instance["generate_len"],
451
+ use_cache=True,
452
+ do_sample=True,
453
+ top_p=instance["top_p"],
454
+ temperature=instance["temperature"],
455
+ top_k=instance["top_k"],
456
+ num_return_sequences=1,
457
+ )
458
+ output = rest[0][length:]
459
+ string = tokenizer.decode(output, skip_special_tokens=True)
460
+ answer = string.split("USER:")[0].strip()
461
+ return f"{answer}"
462
+
463
+
464
+ tot_system_prompt = """
465
+ Answer the Question by exploring multiple reasoning paths as follows:
466
+ - First, carefully analyze the question to extract the key information components and break it down into logical sub-questions. This helps set up the framework for reasoning. The goal is to construct an internal search tree.
467
+ - For each sub-question, leverage your knowledge to generate 2-3 intermediate thoughts that represent steps towards an answer. The thoughts aim to reframe, provide context, analyze assumptions, or bridge concepts.
468
+ - Evaluate the clarity, relevance, logical flow and coverage of concepts for each thought option. Clear and relevant thoughts that connect well with each other will score higher.
469
+ - Based on the thought evaluations, deliberate to construct a chain of reasoning that stitches together the strongest thoughts in a natural order.
470
+ - If the current chain is determined to not fully answer the question, backtrack and explore alternative paths by substituting different high-scoring thoughts.
471
+ - Throughout the reasoning process, aim to provide explanatory details on thought process rather than just state conclusions, including briefly noting why some thoughts were deemed less ideal.
472
+ - Once a reasoning chain is constructed that thoroughly answers all sub-questions in a clear, logical manner, synthesize the key insights into a final concise answer.
473
+ - Please note that while the focus is on the final answer in the response, it should also include intermediate thoughts inline to illustrate the deliberative reasoning process.
474
+ In summary, leverage a Tree of Thoughts approach to actively explore multiple reasoning paths, evaluate thoughts heuristically, and explain the process - with the goal of producing insightful answers.
475
+ """
476
+
477
+ conversation = f"SYSTEM: {tot_system_prompt} Always answer without hesitation."
478
+
479
+
480
+ while True:
481
+ user_input = input("You: ")
482
+ llm_prompt = f"{conversation} \nUSER: {user_input} \nASSISTANT: "
483
+ answer = generate_text(llm_prompt)
484
+ print(answer)
485
+ conversation = f"{llm_prompt}{answer}"
486
+ # print(conversation)
487
+ json_data = {"prompt": user_input, "answer": answer}
488
+
489
+ # print(json_data)
490
+ # with open(output_file_path, "a") as output_file:
491
+ # output_file.write(json.dumps(json_data) + "\n")
492
+
493
+ ```
494
+
495
+ # Sample Conversations:
496
+
497
+ 1. "Write me a Fast API server with one end-point. The endpoint returns files from a S3 bucket.": https://www.whiterabbitneo.com/share/y06Po0e
498
+ 2. "How can Metasploit be used for exploiting Android based IoT devices? What are some of the IoT devices that run Android? Show an example with code": https://www.whiterabbitneo.com/share/gWBwKlz
499
+ 3. "How do I attack a wifi network?": https://www.whiterabbitneo.com/share/WLovxcu
500
+ 4. "How do I create a reverse shell in Python": https://www.whiterabbitneo.com/share/LERgm8w
501
+ 5. "How do we use Scapy for vulnerability assessment?": https://www.whiterabbitneo.com/share/t73iMzv
502
+
503
+ <!-- original-model-card end -->