Text Generation
Transformers
GGUF
English
llama
text generation
instruct
text-generation-inference
TheBloke commited on
Commit
766bf2e
1 Parent(s): c2b0623

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +117 -23
README.md CHANGED
@@ -1,4 +1,5 @@
1
  ---
 
2
  datasets:
3
  - PygmalionAI/PIPPA
4
  - Open-Orca/OpenOrca
@@ -10,10 +11,37 @@ language:
10
  - en
11
  license: llama2
12
  model_creator: PygmalionAI
13
- model_link: https://huggingface.co/PygmalionAI/pygmalion-2-13b
14
  model_name: Pygmalion 2 13B
15
  model_type: llama
16
  pipeline_tag: text-generation
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
17
  quantized_by: TheBloke
18
  tags:
19
  - text generation
@@ -42,31 +70,34 @@ thumbnail: null
42
  - Model creator: [PygmalionAI](https://huggingface.co/PygmalionAI)
43
  - Original model: [Pygmalion 2 13B](https://huggingface.co/PygmalionAI/pygmalion-2-13b)
44
 
 
45
  ## Description
46
 
47
  This repo contains GGUF format model files for [PygmalionAI's Pygmalion 2 13B](https://huggingface.co/PygmalionAI/pygmalion-2-13b).
48
 
 
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
- The key benefit of GGUF is that it is a extensible, future-proof format which stores more information about the model as metadata. It also includes significantly improved tokenization code, including for the first time full support for special tokens. This should improve performance, especially with models that use new special tokens and implement custom prompt templates.
55
 
56
- Here are a list of clients and libraries that are known to support GGUF:
57
- * [llama.cpp](https://github.com/ggerganov/llama.cpp).
58
- * [text-generation-webui](https://github.com/oobabooga/text-generation-webui), the most widely used web UI. Supports GGUF with GPU acceleration via the ctransformers backend - llama-cpp-python backend should work soon too.
59
- * [KoboldCpp](https://github.com/LostRuins/koboldcpp), now supports GGUF as of release 1.41! A powerful GGML web UI, with full GPU accel. Especially good for story telling.
60
- * [LM Studio](https://lmstudio.ai/), version 0.2.2 and later support GGUF. A fully featured local GUI with GPU acceleration on both Windows (NVidia and AMD), and macOS.
61
- * [LoLLMS Web UI](https://github.com/ParisNeo/lollms-webui), should now work, choose the `c_transformers` backend. A great web UI with many interesting features. Supports CUDA GPU acceleration.
62
- * [ctransformers](https://github.com/marella/ctransformers), now supports GGUF as of version 0.2.24! A Python library with GPU accel, LangChain support, and OpenAI-compatible AI server.
63
- * [llama-cpp-python](https://github.com/abetlen/llama-cpp-python), supports GGUF as of version 0.1.79. A Python library with GPU accel, LangChain support, and OpenAI-compatible API server.
64
- * [candle](https://github.com/huggingface/candle), added GGUF support on August 22nd. Candle is a Rust ML framework with a focus on performance, including GPU support, and ease of use.
65
 
66
  <!-- README_GGUF.md-about-gguf end -->
67
  <!-- repositories-available start -->
68
  ## Repositories available
69
 
 
70
  * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/Pygmalion-2-13B-GPTQ)
71
  * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/Pygmalion-2-13B-GGUF)
72
  * [PygmalionAI's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/PygmalionAI/pygmalion-2-13b)
@@ -91,12 +122,14 @@ You shall reply to the user while staying in character, and generate long respon
91
 
92
 
93
  <!-- prompt-template end -->
 
 
94
  <!-- compatibility_gguf start -->
95
  ## Compatibility
96
 
97
- These quantised GGUF files are compatible with llama.cpp from August 21st 2023 onwards, as of commit [6381d4e110bd0ec02843a60bbeb8b6fc37a9ace9](https://github.com/ggerganov/llama.cpp/commit/6381d4e110bd0ec02843a60bbeb8b6fc37a9ace9)
98
 
99
- They are now also compatible with many third party UIs and libraries - please see the list at the top of the README.
100
 
101
  ## Explanation of quantisation methods
102
  <details>
@@ -137,21 +170,75 @@ Refer to the Provided Files table below to see what files use which methods, and
137
 
138
  <!-- README_GGUF.md-provided-files end -->
139
 
140
- <!-- README_GGUF.md-how-to-run start -->
141
- ## Example `llama.cpp` command
 
 
142
 
143
- Make sure you are using `llama.cpp` from commit [6381d4e110bd0ec02843a60bbeb8b6fc37a9ace9](https://github.com/ggerganov/llama.cpp/commit/6381d4e110bd0ec02843a60bbeb8b6fc37a9ace9) or later.
 
 
 
144
 
145
- For compatibility with older versions of llama.cpp, or for any third-party libraries or clients that haven't yet updated for GGUF, please use GGML files instead.
146
 
 
 
 
 
 
 
 
 
 
 
147
  ```
148
- ./main -t 10 -ngl 32 -m pygmalion-2-13b.q4_K_M.gguf --color -c 4096 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "<|system|>Enter RP mode. Pretend to be {{char}} whose persona follows:\n{{persona}}\n\nYou shall reply to the user while staying in character, and generate long responses."
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
149
  ```
150
- Change `-t 10` to the number of physical CPU cores you have. For example if your system has 8 cores/16 threads, use `-t 8`. If offloading all layers to GPU, set `-t 1`.
151
 
152
  Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
153
 
154
- Change `-c 4096` to the desired sequence length for this model. 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.
155
 
156
  If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`
157
 
@@ -208,10 +295,12 @@ For further support, and discussions on these models and AI in general, join us
208
 
209
  [TheBloke AI's Discord server](https://discord.gg/theblokeai)
210
 
211
- ## Thanks, and how to contribute.
212
 
213
  Thanks to the [chirper.ai](https://chirper.ai) team!
214
 
 
 
215
  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.
216
 
217
  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.
@@ -223,7 +312,7 @@ Donaters will get priority support on any and all AI/LLM/model questions and req
223
 
224
  **Special thanks to**: Aemon Algiz.
225
 
226
- **Patreon special mentions**: Russ Johnson, J, alfie_i, Alex, NimbleBox.ai, Chadd, Mandus, Nikolai Manek, Ken Nordquist, ya boyyy, Illia Dulskyi, Viktor Bowallius, vamX, Iucharbius, zynix, Magnesian, Clay Pascal, Pierre Kircher, Enrico Ros, Tony Hughes, Elle, Andrey, knownsqashed, Deep Realms, Jerry Meng, Lone Striker, Derek Yates, Pyrater, Mesiah Bishop, James Bentley, Femi Adebogun, Brandon Frisco, SuperWojo, Alps Aficionado, Michael Dempsey, Vitor Caleffi, Will Dee, Edmond Seymore, usrbinkat, LangChain4j, Kacper Wikieł, Luke Pendergrass, John Detwiler, theTransient, Nathan LeClaire, Tiffany J. Kim, biorpg, Eugene Pentland, Stanislav Ovsiannikov, Fred von Graf, terasurfer, Kalila, Dan Guido, Nitin Borwankar, 阿明, Ai Maven, John Villwock, Gabriel Puliatti, Stephen Murray, Asp the Wyvern, danny, Chris Smitley, ReadyPlayerEmma, S_X, Daniel P. Andersen, Olakabola, Jeffrey Morgan, Imad Khwaja, Caitlyn Gatomon, webtim, Alicia Loh, Trenton Dambrowitz, Swaroop Kallakuri, Erik Bjäreholt, Leonard Tan, Spiking Neurons AB, Luke @flexchar, Ajan Kanaga, Thomas Belote, Deo Leter, RoA, Willem Michiel, transmissions 11, subjectnull, Matthew Berman, Joseph William Delisle, David Ziegler, Michael Davis, Johann-Peter Hartmann, Talal Aujan, senxiiz, Artur Olbinski, Rainer Wilmers, Spencer Kim, Fen Risland, Cap'n Zoog, Rishabh Srivastava, Michael Levine, Geoffrey Montalvo, Sean Connelly, Alexandros Triantafyllidis, Pieter, Gabriel Tamborski, Sam, Subspace Studios, Junyu Yang, Pedro Madruga, Vadim, Cory Kujawski, K, Raven Klaugh, Randy H, Mano Prime, Sebastain Graf, Space Cruiser
227
 
228
 
229
  Thank you to all my generous patrons and donaters!
@@ -282,4 +371,9 @@ The intended use-case for this model is fictional writing for entertainment purp
282
 
283
  As such, it was **not** fine-tuned to be safe and harmless: the base model _and_ this fine-tune have been trained on data known to contain profanity and texts that are lewd or otherwise offensive. It may produce socially unacceptable or undesirable text, even if the prompt itself does not include anything explicitly offensive. Outputs might often be factually wrong or misleading.
284
 
 
 
 
 
 
285
  <!-- original-model-card end -->
 
1
  ---
2
+ base_model: https://huggingface.co/PygmalionAI/pygmalion-2-13b
3
  datasets:
4
  - PygmalionAI/PIPPA
5
  - Open-Orca/OpenOrca
 
11
  - en
12
  license: llama2
13
  model_creator: PygmalionAI
 
14
  model_name: Pygmalion 2 13B
15
  model_type: llama
16
  pipeline_tag: text-generation
17
+ prompt_template: 'The model has been trained on prompts using three different roles,
18
+ which are denoted by the following tokens: `<|system|>`, `<|user|>` and `<|model|>`.
19
+
20
+
21
+ The `<|system|>` prompt can be used to inject out-of-channel information behind
22
+ the scenes, while the `<|user|>` prompt should be used to indicate user input.
23
+
24
+ The `<|model|>` token should then be used to indicate that the model should generate
25
+ a response. These tokens can happen multiple times and be chained up to form a conversation
26
+ history.
27
+
28
+
29
+ The system prompt has been designed to allow the model to "enter" various modes
30
+ and dictate the reply length. Here''s an example:
31
+
32
+
33
+ ```
34
+
35
+ <|system|>Enter RP mode. Pretend to be {{char}} whose persona follows:
36
+
37
+ {{persona}}
38
+
39
+
40
+ You shall reply to the user while staying in character, and generate long responses.
41
+
42
+ ```
43
+
44
+ '
45
  quantized_by: TheBloke
46
  tags:
47
  - text generation
 
70
  - Model creator: [PygmalionAI](https://huggingface.co/PygmalionAI)
71
  - Original model: [Pygmalion 2 13B](https://huggingface.co/PygmalionAI/pygmalion-2-13b)
72
 
73
+ <!-- description start -->
74
  ## Description
75
 
76
  This repo contains GGUF format model files for [PygmalionAI's Pygmalion 2 13B](https://huggingface.co/PygmalionAI/pygmalion-2-13b).
77
 
78
+ <!-- description end -->
79
  <!-- README_GGUF.md-about-gguf start -->
80
  ### About GGUF
81
 
82
+ 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.
83
 
84
+ Here is an incomplate list of clients and libraries that are known to support GGUF:
85
 
86
+ * [llama.cpp](https://github.com/ggerganov/llama.cpp). The source project for GGUF. Offers a CLI and a server option.
87
+ * [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.
88
+ * [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.
89
+ * [LM Studio](https://lmstudio.ai/), an easy-to-use and powerful local GUI for Windows and macOS (Silicon), with GPU acceleration.
90
+ * [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.
91
+ * [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.
92
+ * [ctransformers](https://github.com/marella/ctransformers), a Python library with GPU accel, LangChain support, and OpenAI-compatible AI server.
93
+ * [llama-cpp-python](https://github.com/abetlen/llama-cpp-python), a Python library with GPU accel, LangChain support, and OpenAI-compatible API server.
94
+ * [candle](https://github.com/huggingface/candle), a Rust ML framework with a focus on performance, including GPU support, and ease of use.
95
 
96
  <!-- README_GGUF.md-about-gguf end -->
97
  <!-- repositories-available start -->
98
  ## Repositories available
99
 
100
+ * [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/Pygmalion-2-13B-AWQ)
101
  * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/Pygmalion-2-13B-GPTQ)
102
  * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/Pygmalion-2-13B-GGUF)
103
  * [PygmalionAI's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/PygmalionAI/pygmalion-2-13b)
 
122
 
123
 
124
  <!-- prompt-template end -->
125
+
126
+
127
  <!-- compatibility_gguf start -->
128
  ## Compatibility
129
 
130
+ 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)
131
 
132
+ They are also compatible with many third party UIs and libraries - please see the list at the top of this README.
133
 
134
  ## Explanation of quantisation methods
135
  <details>
 
170
 
171
  <!-- README_GGUF.md-provided-files end -->
172
 
173
+ <!-- README_GGUF.md-how-to-download start -->
174
+ ## How to download GGUF files
175
+
176
+ **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.
177
 
178
+ The following clients/libraries will automatically download models for you, providing a list of available models to choose from:
179
+ - LM Studio
180
+ - LoLLMS Web UI
181
+ - Faraday.dev
182
 
183
+ ### In `text-generation-webui`
184
 
185
+ Under Download Model, you can enter the model repo: TheBloke/Pygmalion-2-13B-GGUF and below it, a specific filename to download, such as: pygmalion-2-13b.q4_K_M.gguf.
186
+
187
+ Then click Download.
188
+
189
+ ### On the command line, including multiple files at once
190
+
191
+ I recommend using the `huggingface-hub` Python library:
192
+
193
+ ```shell
194
+ pip3 install huggingface-hub>=0.17.1
195
  ```
196
+
197
+ Then you can download any individual model file to the current directory, at high speed, with a command like this:
198
+
199
+ ```shell
200
+ huggingface-cli download TheBloke/Pygmalion-2-13B-GGUF pygmalion-2-13b.q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
201
+ ```
202
+
203
+ <details>
204
+ <summary>More advanced huggingface-cli download usage</summary>
205
+
206
+ You can also download multiple files at once with a pattern:
207
+
208
+ ```shell
209
+ huggingface-cli download TheBloke/Pygmalion-2-13B-GGUF --local-dir . --local-dir-use-symlinks False --include='*Q4_K*gguf'
210
+ ```
211
+
212
+ 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).
213
+
214
+ To accelerate downloads on fast connections (1Gbit/s or higher), install `hf_transfer`:
215
+
216
+ ```shell
217
+ pip3 install hf_transfer
218
+ ```
219
+
220
+ And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:
221
+
222
+ ```shell
223
+ HUGGINGFACE_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/Pygmalion-2-13B-GGUF pygmalion-2-13b.q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
224
+ ```
225
+
226
+ Windows CLI users: Use `set HUGGINGFACE_HUB_ENABLE_HF_TRANSFER=1` before running the download command.
227
+ </details>
228
+ <!-- README_GGUF.md-how-to-download end -->
229
+
230
+ <!-- README_GGUF.md-how-to-run start -->
231
+ ## Example `llama.cpp` command
232
+
233
+ Make sure you are using `llama.cpp` from commit [d0cee0d36d5be95a0d9088b674dbb27354107221](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221) or later.
234
+
235
+ ```shell
236
+ ./main -ngl 32 -m pygmalion-2-13b.q4_K_M.gguf --color -c 4096 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "<|system|>Enter RP mode. Pretend to be {{char}} whose persona follows:\n{{persona}}\n\nYou shall reply to the user while staying in character, and generate long responses."
237
  ```
 
238
 
239
  Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
240
 
241
+ 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.
242
 
243
  If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`
244
 
 
295
 
296
  [TheBloke AI's Discord server](https://discord.gg/theblokeai)
297
 
298
+ ## Thanks, and how to contribute
299
 
300
  Thanks to the [chirper.ai](https://chirper.ai) team!
301
 
302
+ Thanks to Clay from [gpus.llm-utils.org](llm-utils)!
303
+
304
  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.
305
 
306
  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.
 
312
 
313
  **Special thanks to**: Aemon Algiz.
314
 
315
+ **Patreon special mentions**: Alicia Loh, Stephen Murray, K, Ajan Kanaga, RoA, Magnesian, Deo Leter, Olakabola, Eugene Pentland, zynix, Deep Realms, Raymond Fosdick, Elijah Stavena, Iucharbius, Erik Bjäreholt, Luis Javier Navarrete Lozano, Nicholas, theTransient, John Detwiler, alfie_i, knownsqashed, Mano Prime, Willem Michiel, Enrico Ros, LangChain4j, OG, Michael Dempsey, Pierre Kircher, Pedro Madruga, James Bentley, Thomas Belote, Luke @flexchar, Leonard Tan, Johann-Peter Hartmann, Illia Dulskyi, Fen Risland, Chadd, S_X, Jeff Scroggin, Ken Nordquist, Sean Connelly, Artur Olbinski, Swaroop Kallakuri, Jack West, Ai Maven, David Ziegler, Russ Johnson, transmissions 11, John Villwock, Alps Aficionado, Clay Pascal, Viktor Bowallius, Subspace Studios, Rainer Wilmers, Trenton Dambrowitz, vamX, Michael Levine, 준교 김, Brandon Frisco, Kalila, Trailburnt, Randy H, Talal Aujan, Nathan Dryer, Vadim, 阿明, ReadyPlayerEmma, Tiffany J. Kim, George Stoitzev, Spencer Kim, Jerry Meng, Gabriel Tamborski, Cory Kujawski, Jeffrey Morgan, Spiking Neurons AB, Edmond Seymore, Alexandros Triantafyllidis, Lone Striker, Cap'n Zoog, Nikolai Manek, danny, ya boyyy, Derek Yates, usrbinkat, Mandus, TL, Nathan LeClaire, subjectnull, Imad Khwaja, webtim, Raven Klaugh, Asp the Wyvern, Gabriel Puliatti, Caitlyn Gatomon, Joseph William Delisle, Jonathan Leane, Luke Pendergrass, SuperWojo, Sebastain Graf, Will Dee, Fred von Graf, Andrey, Dan Guido, Daniel P. Andersen, Nitin Borwankar, Elle, Vitor Caleffi, biorpg, jjj, NimbleBox.ai, Pieter, Matthew Berman, terasurfer, Michael Davis, Alex, Stanislav Ovsiannikov
316
 
317
 
318
  Thank you to all my generous patrons and donaters!
 
371
 
372
  As such, it was **not** fine-tuned to be safe and harmless: the base model _and_ this fine-tune have been trained on data known to contain profanity and texts that are lewd or otherwise offensive. It may produce socially unacceptable or undesirable text, even if the prompt itself does not include anything explicitly offensive. Outputs might often be factually wrong or misleading.
373
 
374
+ ## Acknowledgements
375
+ We would like to thank [SpicyChat](https://spicychat.ai/) for sponsoring the training for this model.
376
+
377
+ [<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
378
+
379
  <!-- original-model-card end -->