TheBloke commited on
Commit
7f81ada
1 Parent(s): e9d37d9

Initial GGUF model commit

Browse files
Files changed (1) hide show
  1. README.md +227 -0
README.md ADDED
@@ -0,0 +1,227 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets:
3
+ - pg19
4
+ inference: false
5
+ library_name: transformers
6
+ license: llama2
7
+ metrics:
8
+ - perplexity
9
+ model_creator: NousResearch
10
+ model_link: https://huggingface.co/NousResearch/Yarn-Llama-2-7b-64k
11
+ model_name: Yarn Llama 2 7B 64K
12
+ model_type: llama
13
+ quantized_by: TheBloke
14
+ ---
15
+
16
+ <!-- header start -->
17
+ <!-- 200823 -->
18
+ <div style="width: auto; margin-left: auto; margin-right: auto">
19
+ <img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
20
+ </div>
21
+ <div style="display: flex; justify-content: space-between; width: 100%;">
22
+ <div style="display: flex; flex-direction: column; align-items: flex-start;">
23
+ <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://discord.gg/theblokeai">Chat & support: TheBloke's Discord server</a></p>
24
+ </div>
25
+ <div style="display: flex; flex-direction: column; align-items: flex-end;">
26
+ <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>
27
+ </div>
28
+ </div>
29
+ <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>
30
+ <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
31
+ <!-- header end -->
32
+
33
+ # Yarn Llama 2 7B 64K - GGUF
34
+ - Model creator: [NousResearch](https://huggingface.co/NousResearch)
35
+ - Original model: [Yarn Llama 2 7B 64K](https://huggingface.co/NousResearch/Yarn-Llama-2-7b-64k)
36
+
37
+ ## Description
38
+
39
+ This repo contains GGUF format model files for [NousResearch's Yarn Llama 2 7B 64K](https://huggingface.co/NousResearch/Yarn-Llama-2-7b-64k).
40
+
41
+ <!-- README_GGUF.md-about-gguf start -->
42
+ ### About GGUF
43
+
44
+ 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.
45
+
46
+ 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.
47
+
48
+ As of August 25th, here is a list of clients and libraries that are known to support GGUF:
49
+ * [llama.cpp](https://github.com/ggerganov/llama.cpp).
50
+ * [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.
51
+ * [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.
52
+ * [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.
53
+ * [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.
54
+ * [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.
55
+ * [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.
56
+ * [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.
57
+
58
+ The clients and libraries below are expecting to add GGUF support shortly:
59
+ <!-- README_GGUF.md-about-gguf end -->
60
+
61
+ <!-- repositories-available start -->
62
+ ## Repositories available
63
+
64
+ * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GPTQ)
65
+ * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF)
66
+ * [2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference (deprecated)](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGML)
67
+ * [NousResearch's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/NousResearch/Yarn-Llama-2-7b-64k)
68
+ <!-- repositories-available end -->
69
+
70
+ <!-- prompt-template start -->
71
+ ## Prompt template: None
72
+
73
+ ```
74
+ {prompt}
75
+
76
+ ```
77
+
78
+ <!-- prompt-template end -->
79
+ <!-- compatibility_gguf start -->
80
+ ## Compatibility
81
+
82
+ 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)
83
+
84
+ They are now also compatible with many third party UIs and libraries - please see the list at the top of the README.
85
+
86
+ ## Explanation of quantisation methods
87
+ <details>
88
+ <summary>Click to see details</summary>
89
+
90
+ The new methods available are:
91
+ * 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)
92
+ * 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.
93
+ * 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.
94
+ * GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw
95
+ * 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
96
+
97
+ Refer to the Provided Files table below to see what files use which methods, and how.
98
+ </details>
99
+ <!-- compatibility_gguf end -->
100
+
101
+ <!-- README_GGUF.md-provided-files start -->
102
+ ## Provided files
103
+
104
+ | Name | Quant method | Bits | Size | Max RAM required | Use case |
105
+ | ---- | ---- | ---- | ---- | ---- | ----- |
106
+ | [yarn-llama-2-7b-64k.Q2_K.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q2_K.gguf) | Q2_K | 2 | 2.83 GB| 5.33 GB | smallest, significant quality loss - not recommended for most purposes |
107
+ | [yarn-llama-2-7b-64k.Q3_K_S.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q3_K_S.gguf) | Q3_K_S | 3 | 2.95 GB| 5.45 GB | very small, high quality loss |
108
+ | [yarn-llama-2-7b-64k.Q3_K_M.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q3_K_M.gguf) | Q3_K_M | 3 | 3.30 GB| 5.80 GB | very small, high quality loss |
109
+ | [yarn-llama-2-7b-64k.Q3_K_L.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q3_K_L.gguf) | Q3_K_L | 3 | 3.60 GB| 6.10 GB | small, substantial quality loss |
110
+ | [yarn-llama-2-7b-64k.Q4_0.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q4_0.gguf) | Q4_0 | 4 | 3.83 GB| 6.33 GB | legacy; small, very high quality loss - prefer using Q3_K_M |
111
+ | [yarn-llama-2-7b-64k.Q4_K_S.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q4_K_S.gguf) | Q4_K_S | 4 | 3.86 GB| 6.36 GB | small, greater quality loss |
112
+ | [yarn-llama-2-7b-64k.Q4_K_M.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q4_K_M.gguf) | Q4_K_M | 4 | 4.08 GB| 6.58 GB | medium, balanced quality - recommended |
113
+ | [yarn-llama-2-7b-64k.Q5_0.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q5_0.gguf) | Q5_0 | 5 | 4.65 GB| 7.15 GB | legacy; medium, balanced quality - prefer using Q4_K_M |
114
+ | [yarn-llama-2-7b-64k.Q5_K_S.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q5_K_S.gguf) | Q5_K_S | 5 | 4.65 GB| 7.15 GB | large, low quality loss - recommended |
115
+ | [yarn-llama-2-7b-64k.Q5_K_M.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q5_K_M.gguf) | Q5_K_M | 5 | 4.78 GB| 7.28 GB | large, very low quality loss - recommended |
116
+ | [yarn-llama-2-7b-64k.Q6_K.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q6_K.gguf) | Q6_K | 6 | 5.53 GB| 8.03 GB | very large, extremely low quality loss |
117
+ | [yarn-llama-2-7b-64k.Q8_0.gguf](https://huggingface.co/TheBloke/Yarn-Llama-2-7B-64K-GGUF/blob/main/yarn-llama-2-7b-64k.Q8_0.gguf) | Q8_0 | 8 | 7.16 GB| 9.66 GB | very large, extremely low quality loss - not recommended |
118
+
119
+ **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.
120
+ <!-- README_GGUF.md-provided-files end -->
121
+
122
+ <!-- README_GGUF.md-how-to-run start -->
123
+ ## How to run in `llama.cpp`
124
+
125
+ Make sure you are using `llama.cpp` from commit [6381d4e110bd0ec02843a60bbeb8b6fc37a9ace9](https://github.com/ggerganov/llama.cpp/commit/6381d4e110bd0ec02843a60bbeb8b6fc37a9ace9) or later.
126
+
127
+ For compatibility with older versions of llama.cpp, or for use with third-party clients and libaries, please use GGML files instead.
128
+
129
+ ```
130
+ ./main -t 10 -ngl 32 -m yarn-llama-2-7b-64k.q4_K_M.gguf --color -c 4096 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "Write a story about llamas"
131
+ ```
132
+ 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`.
133
+
134
+ Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
135
+
136
+ 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.
137
+
138
+ If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`
139
+
140
+ 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)
141
+
142
+ ## How to run in `text-generation-webui`
143
+
144
+ Further instructions here: [text-generation-webui/docs/llama.cpp.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/llama.cpp.md).
145
+ <!-- README_GGUF.md-how-to-run end -->
146
+
147
+ <!-- footer start -->
148
+ <!-- 200823 -->
149
+ ## Discord
150
+
151
+ For further support, and discussions on these models and AI in general, join us at:
152
+
153
+ [TheBloke AI's Discord server](https://discord.gg/theblokeai)
154
+
155
+ ## Thanks, and how to contribute.
156
+
157
+ Thanks to the [chirper.ai](https://chirper.ai) team!
158
+
159
+ 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.
160
+
161
+ 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.
162
+
163
+ Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
164
+
165
+ * Patreon: https://patreon.com/TheBlokeAI
166
+ * Ko-Fi: https://ko-fi.com/TheBlokeAI
167
+
168
+ **Special thanks to**: Aemon Algiz.
169
+
170
+ **Patreon special mentions**: Kacper Wikieł, knownsqashed, Leonard Tan, Asp the Wyvern, Daniel P. Andersen, Luke Pendergrass, Stanislav Ovsiannikov, RoA, Dave, Ai Maven, Kalila, Will Dee, Imad Khwaja, Nitin Borwankar, Joseph William Delisle, Tony Hughes, Cory Kujawski, Rishabh Srivastava, Russ Johnson, Stephen Murray, Lone Striker, Johann-Peter Hartmann, Elle, J, Deep Realms, SuperWojo, Raven Klaugh, Sebastain Graf, ReadyPlayerEmma, Alps Aficionado, Mano Prime, Derek Yates, Gabriel Puliatti, Mesiah Bishop, Magnesian, Sean Connelly, biorpg, Iucharbius, Olakabola, Fen Risland, Space Cruiser, theTransient, Illia Dulskyi, Thomas Belote, Spencer Kim, Pieter, John Detwiler, Fred von Graf, Michael Davis, Swaroop Kallakuri, subjectnull, Clay Pascal, Subspace Studios, Chris Smitley, Enrico Ros, usrbinkat, Steven Wood, alfie_i, David Ziegler, Willem Michiel, Matthew Berman, Andrey, Pyrater, Jeffrey Morgan, vamX, LangChain4j, Luke @flexchar, Trenton Dambrowitz, Pierre Kircher, Alex, Sam, James Bentley, Edmond Seymore, Eugene Pentland, Pedro Madruga, Rainer Wilmers, Dan Guido, Nathan LeClaire, Spiking Neurons AB, Talal Aujan, zynix, Artur Olbinski, Michael Levine, 阿明, K, John Villwock, Nikolai Manek, Femi Adebogun, senxiiz, Deo Leter, NimbleBox.ai, Viktor Bowallius, Geoffrey Montalvo, Mandus, Ajan Kanaga, ya boyyy, Jonathan Leane, webtim, Brandon Frisco, danny, Alexandros Triantafyllidis, Gabriel Tamborski, Randy H, terasurfer, Vadim, Junyu Yang, Vitor Caleffi, Chadd, transmissions 11
171
+
172
+
173
+ Thank you to all my generous patrons and donaters!
174
+
175
+ And thank you again to a16z for their generous grant.
176
+
177
+ <!-- footer end -->
178
+
179
+ <!-- original-model-card start -->
180
+ # Original model card: NousResearch's Yarn Llama 2 7B 64K
181
+
182
+ # Model Card: Nous-Yarn-Llama-2-7b-64k
183
+
184
+
185
+
186
+ ## Model Description
187
+
188
+ Nous-Yarn-Llama-2-7b-64k is a state-of-the-art language model for long context, further pretrained on long context data for 400 steps.
189
+ This model is the Flash Attention 2 patched version of the original model: https://huggingface.co/conceptofmind/Yarn-Llama-2-7b-64k
190
+
191
+ Note that this model **requires** the [Flash Attention library](https://pypi.org/project/flash-attn/) in order to function correctly, see the Model Usage section for installation instructions.
192
+
193
+ ## Model Training
194
+
195
+ Starting from the base Llama 2 models, this model was further pretrained on a subset of the PG19 dataset, allowing it to effectively utilize up to 64k tokens of context.
196
+
197
+ ## Collaborators
198
+
199
+ - [bloc97](https://github.com/bloc97): Methods, Paper and evals
200
+ - [@theemozilla](https://twitter.com/theemozilla): Methods, Paper and evals
201
+ - [@EnricoShippole](https://twitter.com/EnricoShippole): Model Training
202
+ - [honglu2875](https://github.com/honglu2875): Paper and evals
203
+
204
+ The authors would like to thank Stability AI, Carper AI, and Eleuther AI for their generous support of significant computing resources that enabled the training of these models and the completion of this research. We would also like to thank Jonathan Tow and Dakota Mahan directly for their help in advising on the use of the Stability AI compute cluster. Additionally, we would like to thank a16z, and PygmalionAI, for providing resources to run evaluations and experiments on the models.
205
+
206
+ ## Usage and Prompt Format
207
+
208
+ Install FA2 and Rotary Extensions:
209
+ ```
210
+ pip install flash-attn --no-build-isolation
211
+ pip install git+https://github.com/HazyResearch/flash-attention.git#subdirectory=csrc/rotary
212
+ ```
213
+
214
+ There are no specific prompt formats as this is a pretrained base model.
215
+
216
+ ## Benchmark Results
217
+
218
+ TODO
219
+
220
+ ## Future Plans
221
+ We plan to continue training when we have more compute and to improve the dataset and/or instruct tune the models in order to improve the long context performance even further.
222
+
223
+ ## Model Usage
224
+
225
+ The model is available for download on HuggingFace.
226
+
227
+ <!-- original-model-card end -->