TheBloke commited on
Commit
26dc0f3
1 Parent(s): 0d8254f

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +371 -0
README.md ADDED
@@ -0,0 +1,371 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: CausalLM/14B
3
+ datasets:
4
+ - JosephusCheung/GuanacoDataset
5
+ - Open-Orca/OpenOrca
6
+ - stingning/ultrachat
7
+ - meta-math/MetaMathQA
8
+ - liuhaotian/LLaVA-Instruct-150K
9
+ - jondurbin/airoboros-3.1
10
+ - WizardLM/WizardLM_evol_instruct_V2_196k
11
+ - RyokoAI/ShareGPT52K
12
+ - RyokoAI/Fandom23K
13
+ - milashkaarshif/MoeGirlPedia_wikitext_raw_archive
14
+ - wikipedia
15
+ - wiki_lingua
16
+ - fnlp/moss-003-sft-data
17
+ - garage-bAInd/Open-Platypus
18
+ - LDJnr/Puffin
19
+ - openbmb/llava_zh
20
+ - BAAI/COIG
21
+ - TigerResearch/tigerbot-zhihu-zh-10k
22
+ - liwu/MNBVC
23
+ - teknium/openhermes
24
+ inference: false
25
+ language:
26
+ - en
27
+ - zh
28
+ license: wtfpl
29
+ model_creator: CausalLM
30
+ model_name: CausalLM 14B
31
+ model_type: llama
32
+ pipeline_tag: text-generation
33
+ prompt_template: '<|im_start|>system
34
+
35
+ {system_message}<|im_end|>
36
+
37
+ <|im_start|>user
38
+
39
+ {prompt}<|im_end|>
40
+
41
+ <|im_start|>assistant
42
+
43
+ '
44
+ quantized_by: TheBloke
45
+ tags:
46
+ - llama
47
+ - llama2
48
+ ---
49
+
50
+ <!-- header start -->
51
+ <!-- 200823 -->
52
+ <div style="width: auto; margin-left: auto; margin-right: auto">
53
+ <img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
54
+ </div>
55
+ <div style="display: flex; justify-content: space-between; width: 100%;">
56
+ <div style="display: flex; flex-direction: column; align-items: flex-start;">
57
+ <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://discord.gg/theblokeai">Chat & support: TheBloke's Discord server</a></p>
58
+ </div>
59
+ <div style="display: flex; flex-direction: column; align-items: flex-end;">
60
+ <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>
61
+ </div>
62
+ </div>
63
+ <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>
64
+ <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
65
+ <!-- header end -->
66
+
67
+ # CausalLM 14B - GGUF
68
+ - Model creator: [CausalLM](https://huggingface.co/CausalLM)
69
+ - Original model: [CausalLM 14B](https://huggingface.co/CausalLM/14B)
70
+
71
+ <!-- description start -->
72
+ ## Description
73
+
74
+ This repo contains GGUF format model files for [CausalLM's CausalLM 14B](https://huggingface.co/CausalLM/14B).
75
+
76
+ <!-- description end -->
77
+ <!-- README_GGUF.md-about-gguf start -->
78
+ ### About GGUF
79
+
80
+ 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.
81
+
82
+ Here is an incomplate list of clients and libraries that are known to support GGUF:
83
+
84
+ * [llama.cpp](https://github.com/ggerganov/llama.cpp). The source project for GGUF. Offers a CLI and a server option.
85
+ * [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.
86
+ * [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.
87
+ * [LM Studio](https://lmstudio.ai/), an easy-to-use and powerful local GUI for Windows and macOS (Silicon), with GPU acceleration.
88
+ * [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.
89
+ * [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.
90
+ * [ctransformers](https://github.com/marella/ctransformers), a Python library with GPU accel, LangChain support, and OpenAI-compatible AI server.
91
+ * [llama-cpp-python](https://github.com/abetlen/llama-cpp-python), a Python library with GPU accel, LangChain support, and OpenAI-compatible API server.
92
+ * [candle](https://github.com/huggingface/candle), a Rust ML framework with a focus on performance, including GPU support, and ease of use.
93
+
94
+ <!-- README_GGUF.md-about-gguf end -->
95
+ <!-- repositories-available start -->
96
+ ## Repositories available
97
+
98
+ * [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/CausalLM-14B-AWQ)
99
+ * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/CausalLM-14B-GPTQ)
100
+ * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/CausalLM-14B-GGUF)
101
+ * [CausalLM's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/CausalLM/14B)
102
+ <!-- repositories-available end -->
103
+
104
+ <!-- prompt-template start -->
105
+ ## Prompt template: ChatML
106
+
107
+ ```
108
+ <|im_start|>system
109
+ {system_message}<|im_end|>
110
+ <|im_start|>user
111
+ {prompt}<|im_end|>
112
+ <|im_start|>assistant
113
+
114
+ ```
115
+
116
+ <!-- prompt-template end -->
117
+ <!-- licensing start -->
118
+ ## Licensing
119
+
120
+ The creator of the source model has listed its license as `wtfpl`, and this quantization has therefore used that same license.
121
+
122
+ As this model is based on Llama 2, it is also subject to the Meta Llama 2 license terms, and the license files for that are additionally included. It should therefore be considered as being claimed to be licensed under both licenses. I contacted Hugging Face for clarification on dual licensing but they do not yet have an official position. Should this change, or should Meta provide any feedback on this situation, I will update this section accordingly.
123
+
124
+ In the meantime, any questions regarding licensing, and in particular how these two licenses might interact, should be directed to the original model repository: [CausalLM's CausalLM 14B](https://huggingface.co/CausalLM/14B).
125
+ <!-- licensing end -->
126
+ <!-- compatibility_gguf start -->
127
+ ## Compatibility
128
+
129
+ 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)
130
+
131
+ They are also compatible with many third party UIs and libraries - please see the list at the top of this README.
132
+
133
+ ## Explanation of quantisation methods
134
+ <details>
135
+ <summary>Click to see details</summary>
136
+
137
+ The new methods available are:
138
+ * 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)
139
+ * 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.
140
+ * 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.
141
+ * GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw
142
+ * 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
143
+
144
+ Refer to the Provided Files table below to see what files use which methods, and how.
145
+ </details>
146
+ <!-- compatibility_gguf end -->
147
+
148
+ <!-- README_GGUF.md-provided-files start -->
149
+ ## Provided files
150
+
151
+ | Name | Quant method | Bits | Size | Max RAM required | Use case |
152
+ | ---- | ---- | ---- | ---- | ---- | ----- |
153
+ | [causallm_14b.Q4_0.gguf](https://huggingface.co/TheBloke/CausalLM-14B-GGUF/blob/main/causallm_14b.Q4_0.gguf) | Q4_0 | 4 | 8.17 GB| 10.67 GB | legacy; small, very high quality loss - prefer using Q3_K_M |
154
+ | [causallm_14b.Q4_1.gguf](https://huggingface.co/TheBloke/CausalLM-14B-GGUF/blob/main/causallm_14b.Q4_1.gguf) | Q4_1 | 4 | 9.01 GB| 11.51 GB | legacy; small, substantial quality loss - lprefer using Q3_K_L |
155
+ | [causallm_14b.Q5_0.gguf](https://huggingface.co/TheBloke/CausalLM-14B-GGUF/blob/main/causallm_14b.Q5_0.gguf) | Q5_0 | 5 | 9.85 GB| 12.35 GB | legacy; medium, balanced quality - prefer using Q4_K_M |
156
+ | [causallm_14b.Q5_1.gguf](https://huggingface.co/TheBloke/CausalLM-14B-GGUF/blob/main/causallm_14b.Q5_1.gguf) | Q5_1 | 5 | 10.68 GB| 13.18 GB | legacy; medium, low quality loss - prefer using Q5_K_M |
157
+ | [causallm_14b.Q8_0.gguf](https://huggingface.co/TheBloke/CausalLM-14B-GGUF/blob/main/causallm_14b.Q8_0.gguf) | Q8_0 | 8 | 15.06 GB| 17.56 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
+ - LM Studio
172
+ - LoLLMS Web UI
173
+ - Faraday.dev
174
+
175
+ ### In `text-generation-webui`
176
+
177
+ Under Download Model, you can enter the model repo: TheBloke/CausalLM-14B-GGUF and below it, a specific filename to download, such as: causallm_14b.Q4_K_M.gguf.
178
+
179
+ Then click Download.
180
+
181
+ ### On the command line, including multiple files at once
182
+
183
+ I recommend using the `huggingface-hub` Python library:
184
+
185
+ ```shell
186
+ pip3 install huggingface-hub
187
+ ```
188
+
189
+ Then you can download any individual model file to the current directory, at high speed, with a command like this:
190
+
191
+ ```shell
192
+ huggingface-cli download TheBloke/CausalLM-14B-GGUF causallm_14b.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
193
+ ```
194
+
195
+ <details>
196
+ <summary>More advanced huggingface-cli download usage</summary>
197
+
198
+ You can also download multiple files at once with a pattern:
199
+
200
+ ```shell
201
+ huggingface-cli download TheBloke/CausalLM-14B-GGUF --local-dir . --local-dir-use-symlinks False --include='*Q4_K*gguf'
202
+ ```
203
+
204
+ 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).
205
+
206
+ To accelerate downloads on fast connections (1Gbit/s or higher), install `hf_transfer`:
207
+
208
+ ```shell
209
+ pip3 install hf_transfer
210
+ ```
211
+
212
+ And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:
213
+
214
+ ```shell
215
+ HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/CausalLM-14B-GGUF causallm_14b.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
216
+ ```
217
+
218
+ Windows Command Line users: You can set the environment variable by running `set HF_HUB_ENABLE_HF_TRANSFER=1` before the download command.
219
+ </details>
220
+ <!-- README_GGUF.md-how-to-download end -->
221
+
222
+ <!-- README_GGUF.md-how-to-run start -->
223
+ ## Example `llama.cpp` command
224
+
225
+ Make sure you are using `llama.cpp` from commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221) or later.
226
+
227
+ ```shell
228
+ ./main -ngl 32 -m causallm_14b.Q4_K_M.gguf --color -c 4096 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "<|im_start|>system\n{system_message}<|im_end|>\n<|im_start|>user\n{prompt}<|im_end|>\n<|im_start|>assistant"
229
+ ```
230
+
231
+ Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
232
+
233
+ 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.
234
+
235
+ If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`
236
+
237
+ 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)
238
+
239
+ ## How to run in `text-generation-webui`
240
+
241
+ Further instructions here: [text-generation-webui/docs/llama.cpp.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/llama.cpp.md).
242
+
243
+ ## How to run from Python code
244
+
245
+ 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.
246
+
247
+ ### How to load this model in Python code, using ctransformers
248
+
249
+ #### First install the package
250
+
251
+ Run one of the following commands, according to your system:
252
+
253
+ ```shell
254
+ # Base ctransformers with no GPU acceleration
255
+ pip install ctransformers
256
+ # Or with CUDA GPU acceleration
257
+ pip install ctransformers[cuda]
258
+ # Or with AMD ROCm GPU acceleration (Linux only)
259
+ CT_HIPBLAS=1 pip install ctransformers --no-binary ctransformers
260
+ # Or with Metal GPU acceleration for macOS systems only
261
+ CT_METAL=1 pip install ctransformers --no-binary ctransformers
262
+ ```
263
+
264
+ #### Simple ctransformers example code
265
+
266
+ ```python
267
+ from ctransformers import AutoModelForCausalLM
268
+
269
+ # Set gpu_layers to the number of layers to offload to GPU. Set to 0 if no GPU acceleration is available on your system.
270
+ llm = AutoModelForCausalLM.from_pretrained("TheBloke/CausalLM-14B-GGUF", model_file="causallm_14b.Q4_K_M.gguf", model_type="llama", gpu_layers=50)
271
+
272
+ print(llm("AI is going to"))
273
+ ```
274
+
275
+ ## How to use with LangChain
276
+
277
+ Here are guides on using llama-cpp-python and ctransformers with LangChain:
278
+
279
+ * [LangChain + llama-cpp-python](https://python.langchain.com/docs/integrations/llms/llamacpp)
280
+ * [LangChain + ctransformers](https://python.langchain.com/docs/integrations/providers/ctransformers)
281
+
282
+ <!-- README_GGUF.md-how-to-run end -->
283
+
284
+ <!-- footer start -->
285
+ <!-- 200823 -->
286
+ ## Discord
287
+
288
+ For further support, and discussions on these models and AI in general, join us at:
289
+
290
+ [TheBloke AI's Discord server](https://discord.gg/theblokeai)
291
+
292
+ ## Thanks, and how to contribute
293
+
294
+ Thanks to the [chirper.ai](https://chirper.ai) team!
295
+
296
+ Thanks to Clay from [gpus.llm-utils.org](llm-utils)!
297
+
298
+ 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.
299
+
300
+ 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.
301
+
302
+ Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
303
+
304
+ * Patreon: https://patreon.com/TheBlokeAI
305
+ * Ko-Fi: https://ko-fi.com/TheBlokeAI
306
+
307
+ **Special thanks to**: Aemon Algiz.
308
+
309
+ **Patreon special mentions**: Pierre Kircher, Stanislav Ovsiannikov, Michael Levine, Eugene Pentland, Andrey, 준교 김, Randy H, Fred von Graf, Artur Olbinski, Caitlyn Gatomon, terasurfer, Jeff Scroggin, James Bentley, Vadim, Gabriel Puliatti, Harry Royden McLaughlin, Sean Connelly, Dan Guido, Edmond Seymore, Alicia Loh, subjectnull, AzureBlack, Manuel Alberto Morcote, Thomas Belote, Lone Striker, Chris Smitley, Vitor Caleffi, Johann-Peter Hartmann, Clay Pascal, biorpg, Brandon Frisco, sidney chen, transmissions 11, Pedro Madruga, jinyuan sun, Ajan Kanaga, Emad Mostaque, Trenton Dambrowitz, Jonathan Leane, Iucharbius, usrbinkat, vamX, George Stoitzev, Luke Pendergrass, theTransient, Olakabola, Swaroop Kallakuri, Cap'n Zoog, Brandon Phillips, Michael Dempsey, Nikolai Manek, danny, Matthew Berman, Gabriel Tamborski, alfie_i, Raymond Fosdick, Tom X Nguyen, Raven Klaugh, LangChain4j, Magnesian, Illia Dulskyi, David Ziegler, Mano Prime, Luis Javier Navarrete Lozano, Erik Bjäreholt, 阿明, Nathan Dryer, Alex, Rainer Wilmers, zynix, TL, Joseph William Delisle, John Villwock, Nathan LeClaire, Willem Michiel, Joguhyik, GodLy, OG, Alps Aficionado, Jeffrey Morgan, ReadyPlayerEmma, Tiffany J. Kim, Sebastain Graf, Spencer Kim, Michael Davis, webtim, Talal Aujan, knownsqashed, John Detwiler, Imad Khwaja, Deo Leter, Jerry Meng, Elijah Stavena, Rooh Singh, Pieter, SuperWojo, Alexandros Triantafyllidis, Stephen Murray, Ai Maven, ya boyyy, Enrico Ros, Ken Nordquist, Deep Realms, Nicholas, Spiking Neurons AB, Elle, Will Dee, Jack West, RoA, Luke @flexchar, Viktor Bowallius, Derek Yates, Subspace Studios, jjj, Toran Billups, Asp the Wyvern, Fen Risland, Ilya, NimbleBox.ai, Chadd, Nitin Borwankar, Emre, Mandus, Leonard Tan, Kalila, K, Trailburnt, S_X, Cory Kujawski
310
+
311
+
312
+ Thank you to all my generous patrons and donaters!
313
+
314
+ And thank you again to a16z for their generous grant.
315
+
316
+ <!-- footer end -->
317
+
318
+ <!-- original-model-card start -->
319
+ # Original model card: CausalLM's CausalLM 14B
320
+
321
+ ![](https://huggingface.co/JosephusCheung/tmp/resolve/main/14.17b.png)
322
+
323
+ # Read Me:
324
+
325
+ Also see [7B Version](https://huggingface.co/CausalLM/7B)
326
+
327
+ This model was trained based on the model weights of Qwen and LLaMA2. The training process utilized a model structure that was identical to LLaMA2, using the same attention calculation method as the original MHA LLaMA2 models, and no additional scaling applied to the Relative Positional Encoding (RoPE).
328
+
329
+ We manually curated a SFT dataset of 1.3B tokens for training, utilizing open source datasets from Hugging Face. For most of these sentences, we performed manual or synthetic rewrites and generated alternate language versions using larger language models. Additionally, we conducted augmented text training using carefully selected entries from Wikipedia, as well as featured entries from Fandom and filtered entries from Moegirlpedia. In order to strike a balance between efficiency and quality, 100% of the data used for training was synthetic data, no direct use of text from the internet or original texts from publicly available datasets was employed for fine-tuning.
330
+
331
+ The 7B version of the model is a distilled version of the 14B model, specifically designed for speculative sampling. Therefore, it is important to exercise caution when directly using the model, as it may produce hallucinations or unreliable outputs.
332
+
333
+ Please note that the model was trained on unfiltered internet data. Since we do not have the capacity to vet all of it, there may be a substantial amount of objectionable content, pornography, violence, and offensive language present that we are unable to remove. Therefore, you will still need to complete your own checks on the model's safety and filter keywords in the output. Due to computational resource constraints, we are presently unable to implement RLHF for the model's ethics and safety, nor training on SFT samples that refuse to answer certain questions for restrictive fine-tuning.
334
+
335
+ Bonus: The model underwent some fine-tuning on the prompt format introduced in LLaVA1.5 that is unrelated to image attention calculation. Therefore, aligning the ViT Projection module with frozen LM under visual instructions would enable rapid implementation of effective multimodal capabilities.
336
+
337
+ ## PROMPT FORMAT:
338
+ [chatml](https://github.com/openai/openai-python/blob/main/chatml.md)
339
+
340
+ **System Prompt must not be empty!**
341
+
342
+ ## MMLU:
343
+ stem ACC: 64.19
344
+
345
+ Humanities ACC: 61.40
346
+
347
+ other ACC: 71.64
348
+
349
+ social ACC: 75.37
350
+
351
+ **AVERAGE ACC:67.36**
352
+
353
+
354
+ ## CEval (Val):
355
+ STEM ACC: 66.71
356
+
357
+ Social Science ACC: 85.10
358
+
359
+ Humanities ACC: 76.68
360
+
361
+ Other ACC: 70.23
362
+
363
+ Hard ACC:54.71
364
+
365
+ **AVERAGE ACC:73.10**
366
+
367
+ ## GSM8K
368
+
369
+ **Zero-shot ACC 0.7012888551933283**
370
+
371
+ <!-- original-model-card end -->