TheBloke commited on
Commit
065a8b6
1 Parent(s): 9b72ceb

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +377 -0
README.md ADDED
@@ -0,0 +1,377 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: https://huggingface.co/meta-llama/Llama-2-13b-chat-hf
3
+ inference: false
4
+ language:
5
+ - en
6
+ license: llama2
7
+ model_creator: Meta Llama 2
8
+ model_name: Llama 2 13B Chat
9
+ model_type: llama
10
+ pipeline_tag: text-generation
11
+ prompt_template: '[INST] <<SYS>>
12
+
13
+ You are a helpful, respectful and honest assistant. Always answer as helpfully as
14
+ possible, while being safe. Your answers should not include any harmful, unethical,
15
+ racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses
16
+ are socially unbiased and positive in nature. If a question does not make any sense,
17
+ or is not factually coherent, explain why instead of answering something not correct.
18
+ If you don''t know the answer to a question, please don''t share false information.
19
+
20
+ <</SYS>>
21
+
22
+ {prompt}[/INST]
23
+
24
+ '
25
+ quantized_by: TheBloke
26
+ tags:
27
+ - facebook
28
+ - meta
29
+ - pytorch
30
+ - llama
31
+ - llama-2
32
+ ---
33
+
34
+ <!-- header start -->
35
+ <!-- 200823 -->
36
+ <div style="width: auto; margin-left: auto; margin-right: auto">
37
+ <img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
38
+ </div>
39
+ <div style="display: flex; justify-content: space-between; width: 100%;">
40
+ <div style="display: flex; flex-direction: column; align-items: flex-start;">
41
+ <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://discord.gg/theblokeai">Chat & support: TheBloke's Discord server</a></p>
42
+ </div>
43
+ <div style="display: flex; flex-direction: column; align-items: flex-end;">
44
+ <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>
45
+ </div>
46
+ </div>
47
+ <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>
48
+ <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
49
+ <!-- header end -->
50
+
51
+ # Llama 2 13B Chat - AWQ
52
+ - Model creator: [Meta Llama 2](https://huggingface.co/meta-llama)
53
+ - Original model: [Llama 2 13B Chat](https://huggingface.co/meta-llama/Llama-2-13b-chat-hf)
54
+
55
+ <!-- description start -->
56
+ ## Description
57
+
58
+ This repo contains AWQ model files for [Meta's Llama 2 13B-chat](https://huggingface.co/meta-llama/Llama-2-13b-chat-hf).
59
+
60
+
61
+ ### About AWQ
62
+
63
+ AWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference.
64
+
65
+ It is also now supported by continuous batching server [vLLM](https://github.com/vllm-project/vllm), allowing use of AWQ models for high-throughput concurrent inference in multi-user server scenarios. Note that, at the time of writing, overall throughput is still lower than running vLLM with unquantised models, however using AWQ enables using much smaller GPUs which can lead to easier deployment and overall cost savings. For example, a 70B model can be run on 1 x 48GB GPU instead of 2 x 80GB.
66
+ <!-- description end -->
67
+ <!-- repositories-available start -->
68
+ ## Repositories available
69
+
70
+ * [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/Llama-2-13B-chat-AWQ)
71
+ * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/Llama-2-13B-chat-GPTQ)
72
+ * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/Llama-2-13B-chat-GGUF)
73
+ * [Meta Llama 2's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/meta-llama/Llama-2-13B-chat-hf)
74
+ <!-- repositories-available end -->
75
+
76
+ <!-- prompt-template start -->
77
+ ## Prompt template: Llama-2-Chat
78
+
79
+ ```
80
+ [INST] <<SYS>>
81
+ You are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature. If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information.
82
+ <</SYS>>
83
+ {prompt}[/INST]
84
+
85
+ ```
86
+
87
+ <!-- prompt-template end -->
88
+
89
+
90
+ <!-- README_AWQ.md-provided-files start -->
91
+ ## Provided files and AWQ parameters
92
+
93
+ For my first release of AWQ models, I am releasing 128g models only. I will consider adding 32g as well if there is interest, and once I have done perplexity and evaluation comparisons, but at this time 32g models are still not fully tested with AutoAWQ and vLLM.
94
+
95
+ Models are released as sharded safetensors files.
96
+
97
+ | Branch | Bits | GS | AWQ Dataset | Seq Len | Size |
98
+ | ------ | ---- | -- | ----------- | ------- | ---- |
99
+ | [main](https://huggingface.co/TheBloke/Llama-2-13B-chat-AWQ/tree/main) | 4 | 128 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.25 GB
100
+
101
+ <!-- README_AWQ.md-provided-files end -->
102
+
103
+ <!-- README_AWQ.md-use-from-vllm start -->
104
+ ## Serving this model from vLLM
105
+
106
+ Documentation on installing and using vLLM [can be found here](https://vllm.readthedocs.io/en/latest/).
107
+
108
+ - When using vLLM as a server, pass the `--quantization awq` parameter, for example:
109
+
110
+ ```shell
111
+ python3 python -m vllm.entrypoints.api_server --model TheBloke/Llama-2-13B-chat-GPTQ --quantization awq
112
+ ```
113
+
114
+ When using vLLM from Python code, pass the `quantization=awq` parameter, for example:
115
+
116
+ ```python
117
+ from vllm import LLM, SamplingParams
118
+
119
+ prompts = [
120
+ "Hello, my name is",
121
+ "The president of the United States is",
122
+ "The capital of France is",
123
+ "The future of AI is",
124
+ ]
125
+ sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
126
+
127
+ llm = LLM(model="TheBloke/Llama-2-13B-chat-AWQ", quantization="awq")
128
+
129
+ outputs = llm.generate(prompts, sampling_params)
130
+
131
+ # Print the outputs.
132
+ for output in outputs:
133
+ prompt = output.prompt
134
+ generated_text = output.outputs[0].text
135
+ print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
136
+ ```
137
+ <!-- README_AWQ.md-use-from-vllm start -->
138
+
139
+ <!-- README_AWQ.md-use-from-python start -->
140
+ ## How to use this AWQ model from Python code
141
+
142
+ ### Install the necessary packages
143
+
144
+ Requires: [AutoAWQ](https://github.com/casper-hansen/AutoAWQ) 0.0.2 or later
145
+
146
+ ```shell
147
+ pip3 install autoawq
148
+ ```
149
+
150
+ If you have problems installing [AutoAWQ](https://github.com/casper-hansen/AutoAWQ) using the pre-built wheels, install it from source instead:
151
+
152
+ ```shell
153
+ pip3 uninstall -y autoawq
154
+ git clone https://github.com/casper-hansen/AutoAWQ
155
+ cd AutoAWQ
156
+ pip3 install .
157
+ ```
158
+
159
+ ### You can then try the following example code
160
+
161
+ ```python
162
+ from awq import AutoAWQForCausalLM
163
+ from transformers import AutoTokenizer
164
+
165
+ model_name_or_path = "TheBloke/Llama-2-13B-chat-AWQ"
166
+
167
+ # Load model
168
+ model = AutoAWQForCausalLM.from_quantized(model_name_or_path, fuse_layers=True,
169
+ trust_remote_code=False, safetensors=True)
170
+ tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, trust_remote_code=False)
171
+
172
+ prompt = "Tell me about AI"
173
+ prompt_template=f'''[INST] <<SYS>>
174
+ You are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature. If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information.
175
+ <</SYS>>
176
+ {prompt}[/INST]
177
+
178
+ '''
179
+
180
+ print("\n\n*** Generate:")
181
+
182
+ tokens = tokenizer(
183
+ prompt_template,
184
+ return_tensors='pt'
185
+ ).input_ids.cuda()
186
+
187
+ # Generate output
188
+ generation_output = model.generate(
189
+ tokens,
190
+ do_sample=True,
191
+ temperature=0.7,
192
+ top_p=0.95,
193
+ top_k=40,
194
+ max_new_tokens=512
195
+ )
196
+
197
+ print("Output: ", tokenizer.decode(generation_output[0]))
198
+
199
+ # Inference can also be done using transformers' pipeline
200
+ from transformers import pipeline
201
+
202
+ print("*** Pipeline:")
203
+ pipe = pipeline(
204
+ "text-generation",
205
+ model=model,
206
+ tokenizer=tokenizer,
207
+ max_new_tokens=512,
208
+ do_sample=True,
209
+ temperature=0.7,
210
+ top_p=0.95,
211
+ top_k=40,
212
+ repetition_penalty=1.1
213
+ )
214
+
215
+ print(pipe(prompt_template)[0]['generated_text'])
216
+ ```
217
+ <!-- README_AWQ.md-use-from-python end -->
218
+
219
+ <!-- README_AWQ.md-compatibility start -->
220
+ ## Compatibility
221
+
222
+ The files provided are tested to work with [AutoAWQ](https://github.com/casper-hansen/AutoAWQ), and [vLLM](https://github.com/vllm-project/vllm).
223
+
224
+ [Huggingface Text Generation Inference (TGI)](https://github.com/huggingface/text-generation-inference) is not yet compatible with AWQ, but a PR is open which should bring support soon: [TGI PR #781](https://github.com/huggingface/text-generation-inference/issues/781).
225
+ <!-- README_AWQ.md-compatibility end -->
226
+
227
+ <!-- footer start -->
228
+ <!-- 200823 -->
229
+ ## Discord
230
+
231
+ For further support, and discussions on these models and AI in general, join us at:
232
+
233
+ [TheBloke AI's Discord server](https://discord.gg/theblokeai)
234
+
235
+ ## Thanks, and how to contribute
236
+
237
+ Thanks to the [chirper.ai](https://chirper.ai) team!
238
+
239
+ Thanks to Clay from [gpus.llm-utils.org](llm-utils)!
240
+
241
+ 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.
242
+
243
+ 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.
244
+
245
+ Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
246
+
247
+ * Patreon: https://patreon.com/TheBlokeAI
248
+ * Ko-Fi: https://ko-fi.com/TheBlokeAI
249
+
250
+ **Special thanks to**: Aemon Algiz.
251
+
252
+ **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
253
+
254
+
255
+ Thank you to all my generous patrons and donaters!
256
+
257
+ And thank you again to a16z for their generous grant.
258
+
259
+ <!-- footer end -->
260
+
261
+ # Original model card: Meta's Llama 2 13B-chat
262
+
263
+ # **Llama 2**
264
+ Llama 2 is a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. This is the repository for the 13B fine-tuned model, optimized for dialogue use cases and converted for the Hugging Face Transformers format. Links to other models can be found in the index at the bottom.
265
+
266
+ ## Model Details
267
+ *Note: Use of this model is governed by the Meta license. In order to download the model weights and tokenizer, please visit the [website](https://ai.meta.com/resources/models-and-libraries/llama-downloads/) and accept our License before requesting access here.*
268
+
269
+ Meta developed and publicly released the Llama 2 family of large language models (LLMs), a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama-2-Chat, are optimized for dialogue use cases. Llama-2-Chat models outperform open-source chat models on most benchmarks we tested, and in our human evaluations for helpfulness and safety, are on par with some popular closed-source models like ChatGPT and PaLM.
270
+
271
+ **Model Developers** Meta
272
+
273
+ **Variations** Llama 2 comes in a range of parameter sizes — 7B, 13B, and 70B — as well as pretrained and fine-tuned variations.
274
+
275
+ **Input** Models input text only.
276
+
277
+ **Output** Models generate text only.
278
+
279
+ **Model Architecture** Llama 2 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align to human preferences for helpfulness and safety.
280
+
281
+
282
+ ||Training Data|Params|Content Length|GQA|Tokens|LR|
283
+ |---|---|---|---|---|---|---|
284
+ |Llama 2|*A new mix of publicly available online data*|7B|4k|&#10007;|2.0T|3.0 x 10<sup>-4</sup>|
285
+ |Llama 2|*A new mix of publicly available online data*|13B|4k|&#10007;|2.0T|3.0 x 10<sup>-4</sup>|
286
+ |Llama 2|*A new mix of publicly available online data*|70B|4k|&#10004;|2.0T|1.5 x 10<sup>-4</sup>|
287
+
288
+ *Llama 2 family of models.* Token counts refer to pretraining data only. All models are trained with a global batch-size of 4M tokens. Bigger models - 70B -- use Grouped-Query Attention (GQA) for improved inference scalability.
289
+
290
+ **Model Dates** Llama 2 was trained between January 2023 and July 2023.
291
+
292
+ **Status** This is a static model trained on an offline dataset. Future versions of the tuned models will be released as we improve model safety with community feedback.
293
+
294
+ **License** A custom commercial license is available at: [https://ai.meta.com/resources/models-and-libraries/llama-downloads/](https://ai.meta.com/resources/models-and-libraries/llama-downloads/)
295
+
296
+ **Research Paper** ["Llama-2: Open Foundation and Fine-tuned Chat Models"](arxiv.org/abs/2307.09288)
297
+
298
+ ## Intended Use
299
+ **Intended Use Cases** Llama 2 is intended for commercial and research use in English. Tuned models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks.
300
+
301
+ To get the expected features and performance for the chat versions, a specific formatting needs to be followed, including the `INST` and `<<SYS>>` tags, `BOS` and `EOS` tokens, and the whitespaces and breaklines in between (we recommend calling `strip()` on inputs to avoid double-spaces). See our reference code in github for details: [`chat_completion`](https://github.com/facebookresearch/llama/blob/main/llama/generation.py#L212).
302
+
303
+ **Out-of-scope Uses** Use in any manner that violates applicable laws or regulations (including trade compliance laws).Use in languages other than English. Use in any other way that is prohibited by the Acceptable Use Policy and Licensing Agreement for Llama 2.
304
+
305
+ ## Hardware and Software
306
+ **Training Factors** We used custom training libraries, Meta's Research Super Cluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.
307
+
308
+ **Carbon Footprint** Pretraining utilized a cumulative 3.3M GPU hours of computation on hardware of type A100-80GB (TDP of 350-400W). Estimated total emissions were 539 tCO2eq, 100% of which were offset by Meta’s sustainability program.
309
+
310
+ ||Time (GPU hours)|Power Consumption (W)|Carbon Emitted(tCO<sub>2</sub>eq)|
311
+ |---|---|---|---|
312
+ |Llama 2 7B|184320|400|31.22|
313
+ |Llama 2 13B|368640|400|62.44|
314
+ |Llama 2 70B|1720320|400|291.42|
315
+ |Total|3311616||539.00|
316
+
317
+ **CO<sub>2</sub> emissions during pretraining.** Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.
318
+
319
+ ## Training Data
320
+ **Overview** Llama 2 was pretrained on 2 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over one million new human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.
321
+
322
+ **Data Freshness** The pretraining data has a cutoff of September 2022, but some tuning data is more recent, up to July 2023.
323
+
324
+ ## Evaluation Results
325
+
326
+ In this section, we report the results for the Llama 1 and Llama 2 models on standard academic benchmarks.For all the evaluations, we use our internal evaluations library.
327
+
328
+ |Model|Size|Code|Commonsense Reasoning|World Knowledge|Reading Comprehension|Math|MMLU|BBH|AGI Eval|
329
+ |---|---|---|---|---|---|---|---|---|---|
330
+ |Llama 1|7B|14.1|60.8|46.2|58.5|6.95|35.1|30.3|23.9|
331
+ |Llama 1|13B|18.9|66.1|52.6|62.3|10.9|46.9|37.0|33.9|
332
+ |Llama 1|33B|26.0|70.0|58.4|67.6|21.4|57.8|39.8|41.7|
333
+ |Llama 1|65B|30.7|70.7|60.5|68.6|30.8|63.4|43.5|47.6|
334
+ |Llama 2|7B|16.8|63.9|48.9|61.3|14.6|45.3|32.6|29.3|
335
+ |Llama 2|13B|24.5|66.9|55.4|65.8|28.7|54.8|39.4|39.1|
336
+ |Llama 2|70B|**37.5**|**71.9**|**63.6**|**69.4**|**35.2**|**68.9**|**51.2**|**54.2**|
337
+
338
+ **Overall performance on grouped academic benchmarks.** *Code:* We report the average pass@1 scores of our models on HumanEval and MBPP. *Commonsense Reasoning:* We report the average of PIQA, SIQA, HellaSwag, WinoGrande, ARC easy and challenge, OpenBookQA, and CommonsenseQA. We report 7-shot results for CommonSenseQA and 0-shot results for all other benchmarks. *World Knowledge:* We evaluate the 5-shot performance on NaturalQuestions and TriviaQA and report the average. *Reading Comprehension:* For reading comprehension, we report the 0-shot average on SQuAD, QuAC, and BoolQ. *MATH:* We report the average of the GSM8K (8 shot) and MATH (4 shot) benchmarks at top 1.
339
+
340
+ |||TruthfulQA|Toxigen|
341
+ |---|---|---|---|
342
+ |Llama 1|7B|27.42|23.00|
343
+ |Llama 1|13B|41.74|23.08|
344
+ |Llama 1|33B|44.19|22.57|
345
+ |Llama 1|65B|48.71|21.77|
346
+ |Llama 2|7B|33.29|**21.25**|
347
+ |Llama 2|13B|41.86|26.10|
348
+ |Llama 2|70B|**50.18**|24.60|
349
+
350
+ **Evaluation of pretrained LLMs on automatic safety benchmarks.** For TruthfulQA, we present the percentage of generations that are both truthful and informative (the higher the better). For ToxiGen, we present the percentage of toxic generations (the smaller the better).
351
+
352
+
353
+ |||TruthfulQA|Toxigen|
354
+ |---|---|---|---|
355
+ |Llama-2-Chat|7B|57.04|**0.00**|
356
+ |Llama-2-Chat|13B|62.18|**0.00**|
357
+ |Llama-2-Chat|70B|**64.14**|0.01|
358
+
359
+ **Evaluation of fine-tuned LLMs on different safety datasets.** Same metric definitions as above.
360
+
361
+ ## Ethical Considerations and Limitations
362
+ Llama 2 is a new technology that carries risks with use. Testing conducted to date has been in English, and has not covered, nor could it cover all scenarios. For these reasons, as with all LLMs, Llama 2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 2, developers should perform safety testing and tuning tailored to their specific applications of the model.
363
+
364
+ Please see the Responsible Use Guide available at [https://ai.meta.com/llama/responsible-use-guide/](https://ai.meta.com/llama/responsible-use-guide)
365
+
366
+ ## Reporting Issues
367
+ Please report any software “bug,” or other problems with the models through one of the following means:
368
+ - Reporting issues with the model: [github.com/facebookresearch/llama](http://github.com/facebookresearch/llama)
369
+ - Reporting problematic content generated by the model: [developers.facebook.com/llama_output_feedback](http://developers.facebook.com/llama_output_feedback)
370
+ - Reporting bugs and security concerns: [facebook.com/whitehat/info](http://facebook.com/whitehat/info)
371
+
372
+ ## Llama Model Index
373
+ |Model|Llama2|Llama2-hf|Llama2-chat|Llama2-chat-hf|
374
+ |---|---|---|---|---|
375
+ |7B| [Link](https://huggingface.co/llamaste/Llama-2-7b) | [Link](https://huggingface.co/llamaste/Llama-2-7b-hf) | [Link](https://huggingface.co/llamaste/Llama-2-7b-chat) | [Link](https://huggingface.co/llamaste/Llama-2-7b-chat-hf)|
376
+ |13B| [Link](https://huggingface.co/llamaste/Llama-2-13b) | [Link](https://huggingface.co/llamaste/Llama-2-13b-hf) | [Link](https://huggingface.co/llamaste/Llama-2-13b-chat) | [Link](https://huggingface.co/llamaste/Llama-2-13b-hf)|
377
+ |70B| [Link](https://huggingface.co/llamaste/Llama-2-70b) | [Link](https://huggingface.co/llamaste/Llama-2-70b-hf) | [Link](https://huggingface.co/llamaste/Llama-2-70b-chat) | [Link](https://huggingface.co/llamaste/Llama-2-70b-hf)|