File size: 20,767 Bytes
04efded
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
cc356f9
4f20837
f3823ee
04efded
 
 
4118d23
 
1fee0a1
04efded
 
 
 
 
 
 
 
 
 
93fb5cd
04efded
 
 
9d02ba9
04efded
93fb5cd
04efded
 
 
9d02ba9
04efded
41641a2
b65f261
04efded
 
 
1122542
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b2c79c3
1122542
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
04efded
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
---
language:
- en
- ko
pipeline_tag: text-generation
inference: false
tags:
- facebook
- meta
- pytorch
- llama
- llama-2
- kollama
- llama-2-ko
- llama-2-ko-chat
---
> 🚧 Note: this repo is under construction 🚧

# **Llama-2-Ko-Chat** πŸ¦™πŸ‡°πŸ‡·
<img src=https://github.com/boostcampaitech5/level2_klue-nlp-08/assets/96530685/9dc3a02a-f554-42c7-9aa7-e7bd8c8870b7 style="max-width: 700px; width: 100%" />

Llama-2-Ko-7b-Chat은 [beomi/llama-2-ko-7b 40B](https://huggingface.co/beomi/llama-2-ko-7b)λ₯Ό ν† λŒ€λ‘œ λ§Œλ“€μ–΄μ‘ŒμŠ΅λ‹ˆλ‹€. ν•™μŠ΅ λ°μ΄ν„°λŠ” [nlpai-lab/kullm-v2](https://huggingface.co/datasets/nlpai-lab/kullm-v2)λ₯Ό 톡해 ν•™μŠ΅ν•˜μ˜€μŠ΅λ‹ˆλ‹€. 아직 ν•™μŠ΅μ΄ 진행 쀑이며 μΆ”ν›„ [beomi/llama-2-ko-7b](https://huggingface.co/beomi/llama-2-ko-7b)의 μ—…λ°μ΄νŠΈμ— 따라 μΆ”κ°€λ‘œ ν›ˆλ ¨μ„ 진행할 κ³„νšμž…λ‹ˆλ‹€.

## Model Details

**Backbone Model** beomi/llama-2-ko-7b

**Model Developers** Taemin Kim (kfkas), Kim Juwon (uomnf97)

**data** nlpai-lab/kullm-v2

**prompts** kullm

**Input** Models input text only.

**Output** Models generate text only.

---
| Model              | 질문 : μΊ ν•‘ 여행에 ν•„μš”ν•œ 10가지 ν’ˆλͺ©μ˜ λͺ©λ‘μ„ μƒμ„±ν•©λ‹ˆλ‹€.                                                                                                                                                                                                                                                                                                                                                                                                |
|--------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| Llama-2-7b-chat-hf | `[Hello! As a helpful and respectful assistant, I'd be happy to help you with your camping trip. To ensure a safe and enjoyable experience, here is a list of 10 essential items you may need for your camping trip:Tent: A sturdy, waterproof tent to provide shelter and protection from the elements.Sleeping Bag: A warm, insulated sleeping bag to keep you cozy during the night.Sleeping Pad: A comfortable sleeping pad to provide extra cushioning and insulation.Backpack: A sturdy, comfortable backpack to carry all your gear.Clothing: Pack layers of breathable, moisture-wicking clothing to stay warm and dry.Footwear: Sturdy, waterproof hiking boots or shoes to protect your feet and provide traction.Headlamp: A reliable, battery-powered headlamp to light your way during the night.First Aid Kit: A well-stocked first aid kit to treat any injuries or illnesses.Water Bottle: A refillable, BPA-free water bottle to stay hydrated.Personal Hygiene Items: Toothbrush, toothpaste, biodegradable soap, and toilet paper to maintain personal hygiene.]`                                                                                                                      |
| Llama-2-ko         | `[3일용 μΊ ν•‘μž₯ μš©ν’ˆ - 의λ₯˜ (5일) : μ†μ˜· , 잠옷, κΈ΄νŒ” 및 λ°˜νŒ”μ΄ ν•„μš”ν•˜λ―€λ‘œ μ€€λΉ„ν•˜κΈ°!\u200b- μ‹λŸ‰ (9일 이상)- μ—°λ£Œ (6일 μ‚¬μš© μ˜ˆμ •μ΄μ§€λ§Œ μ΅œμ†Œ 7 일뢄 λ³΄μœ ν•΄μ•Ό ν•©λ‹ˆλ‹€.)\u200b- ν…νŠΈ/거주지역 : 20 ~ 40 평방 λ―Έν„°κ°€ ν•„μš”ν•©λ‹ˆλ‹€.- λ¬Ό 탱크 (물탱컀 / μˆ˜μ‘±κ΄€ )\u200b캠핑에 μ‚¬μš©ν•  8 λͺ… (λ‚¨μž 4λͺ… / μ—¬μž 4λͺ…이 ν•„μš”ν•©λ‹ˆλ‹€!).이 μž‘μ—…μ—λŠ” 두 개의 쑰건 λΆ€κ°€κ°€ μžˆμŠ΅λ‹ˆλ‹€.(쑰건 1) λͺ¨λ“  μ•„μ΄ν…œ λͺ©λ‘μ—μ„œ 쀑볡 제거λ₯Όμœ„ν•œ 1 번 반볡.이 μ˜΅μ…˜μ€ 쀑볡을 μ‹λ³„ν•˜μ—¬ μ œκ±°ν• λΏλ§Œ μ•„λ‹ˆλΌ μΆ”κ°€ ν•­λͺ©μ΄ 포함 된 카운트λ₯Ό μž¬μ„€μ • ν•  λ•Œ μœ μš©ν•˜κ΅°μš”.(쑰건 = "λͺ¨λ“  μ•„μ΄ν…œ λ¦¬μŠ€νŠΈμ—μ„œ ν•œ 개 μ΄μƒμ˜ 동일 μ•„μ΄ν…œ" μ΄κ±°λ‚˜ "(λͺ¨λ“  ν•­λͺ©)"μ—μ„œ λ™μΌν•œ 숫자 κ°’μœΌλ‘œ 식별 된)\u200b(쑰건 2) μž‘μ—…μ‹œ 각 행이 ν•˜λ‚˜μ˜ μ•„μ΄ν…œ 만 포함 ν•  ν•„μš”κ°€ μ—†μœΌλ―€λ‘œ μ•„μ΄ν…œμ΄ μ„œλ‘œ λ‹€λ₯Έ μ„ ν–‰μž λ‘œλΆ€ν„° μ—¬λŸ¬ 번]` |
| Llama-2-ko-Chat    | `[1) ν…νŠΈ - κ°€μž₯ κΈ°λ³Έμ μ΄λ©΄μ„œ μœ μš©ν•œ μž₯λΉ„ 쀑 ν•˜λ‚˜μ΄λ©° μΊ ν•‘ 여행을 νŽΈμ•ˆν•˜κ²Œ ν•΄μ€λ‹ˆλ‹€.2) 베개 및 μΉ¨λŒ€ μ‹œνŠΈμ™€ 같이 ν‘Ήμ‹ ν•œ 수면 μš©ν’ˆμ„ μ±™κΈ°λŠ” 것도 μžŠμ§€ λ§ˆμ„Έμš”!3) 맀트리슀 - νŽΈμ•ˆν•¨κ³Ό 지지λ ₯을 μ œκ³΅ν•˜μ—¬ μž μ„ 잘 자게 ν•©λ‹ˆλ‹€.4) 의λ₯˜ - 땀을 흘리고 νœ΄μ‹ν•˜λŠ” μ‹œκ°„μ„ 늘리기 μœ„ν•΄ 방수재질이 ν¬ν•¨λœ κΈ΄νŒ”μ˜·κ³Ό λ°˜λ°”μ§€λ₯Ό μ±™κ²¨μš”5) 식기 μ„ΈνŠΈ - μ•Όμ™Έ μ‘°λ¦¬μ—μ„œ μŒμ‹μ„ 먹으렀면 λ°˜λ“œμ‹œ 챙겨야 ν•  λ„κ΅¬μ§€μš”.6) μŠ€ν† λΈŒμ™€ μ—°λ£Œ λ˜λŠ” νœ΄λŒ€μš© κ°€μŠ€ λ Œν„΄ λ“± μ—΄ 기ꡬλ₯Ό κ°€μ Έμ˜΅λ‹ˆλ‹€; μŒμ‹ 쑰리에 맀우 νŽΈλ¦¬ν•˜κ²Œ μ‚¬μš©ν•  수 μžˆμŠ΅λ‹ˆλ‹€.7) λžœν„΄, ν”Œλž˜μ‹œ λ˜λŠ” LED μŠ€νƒ λ“œλ₯Ό μ€€λΉ„ν•˜μ„Έμš”! μ–΄λ‘μš΄ 밀에 더 μ•ˆμ „ν•˜λ„λ‘ 돕고, μ•Όμ™Έμ—μ„œ μš”λ¦¬ν•˜κ³  놀 λ•Œ ν•„μˆ˜μ μΈ ν•„μˆ˜ μ†Œν’ˆμ΄ 될 κ²ƒμž…λ‹ˆλ‹€.8) 손전등 - 밀늦게 이동할 μˆ˜λ„ 있고 μˆ²μ† μ˜€μ†”κΈΈμ„ 걸을 λ•Œλ„ μ΄μš©ν•  κ²ƒμ΄λ―€λ‘œ λ°˜λ“œμ‹œ 가지고 μžˆμ–΄μ•Ό ν•΄μš”.9) κ΅¬κΈ‰μ•½μƒμž - λ‹€μ–‘ν•œ 상해에 λŒ€λΉ„ν•΄μ„œ μ•ŒμΉΌλ¦¬ 미슀트, μ†Œλ…μ œ, ν•­μƒμ œ, μ—°κ³  λ“± 일반적이고 인기 μžˆλŠ” ꡬ급 약물을 항상 λΉ„μΆ•ν•΄μ•Ό ν•΄μš”. μΊ ν”„μ—μ„œλŠ” 응급 상황에 μ²˜ν•˜λ”λΌλ„ μ‰½κ²Œ 찾을 수 μžˆλ„λ‘ μƒμžμ— λ‹΄μŠ΅λ‹ˆλ‹€.10) νœ΄λŒ€ν° 보쑰 배터리/μΆ©μ „κΈ° λ˜λŠ” μΆ©μ „ κ°€λŠ₯ν•œ μ™ΈλΆ€ μ½˜μ„ΌνŠΈλ‚˜ μ—°κ²°λœ μ™ΈλΆ€ 전원μž₯치 - 이것은 μŠ€λ§ˆνŠΈν°μ΄λ‚˜ 슀마트 μ›ŒμΉ˜λ₯Ό μœ„ν•œ ν•„μˆ˜ν’ˆμΌ 뿐만 μ•„λ‹ˆλΌ λ‹€λ₯Έ μ‚¬λžŒλ“€μ΄ ν•„μš”λ‘œ ν•˜λŠ” μƒν™©μ—μ„œ κΈ΄κΈ‰ 연락이 λ˜λ„λ‘ 도와 쀄 μˆ˜λ„ μžˆλ‹€λŠ” 점 κΈ°μ–΅ν•˜μ„Έμš”.]`                                                                                                                                                                                                                                                                                        |
---
| Model              | 질문 : 삼원색에 λŒ€ν•΄ μ•Œλ €μ€˜                                                                                                                                                                                                                                                                                                        |
|--------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| Llama-2-7b-chat-hf | `[Regarding your question, "삼원색에 λŒ€ν•΄ μ•Œλ €μ€˜" (which means "Tell me about trichromacy" in Korean), I can provide some information. Trichromacy is a concept in vision science that refers to the ability of the human visual system to perceive and distinguish between three primary colors: red, green, and blue...]` |
| Llama-2-ko         | `[C1, R0.625 G8 B275\u200b3-4, μ‚Όκ°ν˜• 그리기.\u200bμ•„λž˜μ—λŠ” μ‚Όκ°ν˜•μ„ κ·Έλ¦Όλ‹ˆλ‹€ (그림을 κ·Έλ €μ•Ό ν•  μ˜μ—­μ˜ μ’Œν‘œλ₯Ό 지정) - λͺ…령어와 ν•¨κ»˜ μ‚¬μš©λ˜λŠ” λ¬Έμžμ—΄λ‘œ μž…λ ₯ν•  μ’Œν‘œ κ°’κ³Ό λ™μΌν•œ κ²°κ³Όλ₯Ό μ–»μŒ - 예λ₯Ό λ“€μ–΄ "X=-19. Y=0" 으둜 μž…λ ₯을 ν•˜λ©΄ μ’Œν‘œκ°’μ„ μ§€μ •λœ κ°’κ³Ό κ°™κ²Œ 좜λ ₯을 ν•œλ‹€. ##########################\u200b##3_0-6-1, μ’Œν‘œ μž…λ ₯으둜 ν…μŠ€νŠΈ νŒŒμΌμ„ μƒμ„±ν•˜λΌ(ex C4 D6 E10 A10 )-λͺ…λ Ήλ¬Έμ—μ„œ 각 문자λ₯Ό μž…λ ₯κ³Ό λ™μΌν•˜λ‹€. -이 λͺ…령은 ν˜„μž¬ μˆ˜ν–‰μ€‘μΈ ν”„λ‘œκ·Έλž¨μ„ μ’…λ£Œν•œλ‹€.-νŒŒμΌμ€ μ‹€ν–‰ κ°€λŠ₯ν•œ 파일과 μΌμΉ˜ν• ]`                 |
| Llama-2-ko-Chat    | `[물체λ₯Ό μΈμ‹ν•˜κ³  색상에 λŒ€ν•œ μ‹œκ°μ μΈ 이미지λ₯Ό μƒμ„±ν•˜κΈ° μœ„ν•΄ 두 개의 λˆˆμ€ μ„œλ‘œ λ°˜λŒ€μͺ½ λ°©ν–₯μ—μ„œ 물체의 색상 정보와 λͺ…μ•” 정보λ₯Ό 받아듀이기 λ•Œλ¬Έμ—, 인간은 일반적으둜 색상을 삼원색이라고 ν•˜λŠ” μ„Έ 가지 원좔 ν˜•νƒœ(λΉ¨κ°•, 녹색, νŒŒλž‘μƒ‰)둜 λΆ„λ₯˜ν•©λ‹ˆλ‹€. 이 쀑 ν•œκ°€μ§€ 색상이 λ‹€λ₯Έ 색상보닀 λˆˆμ— 훨씬 더 빨리 λ“€μ–΄μ˜€λŠ” κ²½ν–₯이 μžˆλ‹€κ³  ν•©λ‹ˆλ‹€. ν•˜μ§€λ§Œ λͺ¨λ“  μ‚¬λžŒμ΄ κ·Έλ ‡μ§€λŠ” μ•ŠμœΌλ―€λ‘œ 항상 μ‚Όκ°ν˜• λͺ¨μ–‘μœΌλ‘œ 색상을 λΆ„λ₯˜ν•˜μ§€λŠ” μ•ŠμŠ΅λ‹ˆλ‹€. ν•˜μ§€λ§Œ 삼원색이 우리 λˆˆμ— 잘 μ „λ‹¬λ˜λ©° 색상 ꡬ별에 μ€‘μš”ν•˜λ‹€λŠ” 것은 λΆ€μ •ν•  수 μ—†μŠ΅λ‹ˆλ‹€.]`                                                                                                                                                                                     |
---

## ν›ˆλ ¨ 진행 ν˜„ν™©
<img src=https://github.com/taemin6697/Paper_Review/assets/96530685/b9a697a2-ef06-4b1c-97e1-e72b20d9a8b5 style="max-width: 700px; width: 100%" />
---

### Inference

```python
def gen(x, model, tokenizer, device):
    prompt = (
        f"μ•„λž˜λŠ” μž‘μ—…μ„ μ„€λͺ…ν•˜λŠ” λͺ…λ Ήμ–΄μž…λ‹ˆλ‹€. μš”μ²­μ„ 적절히 μ™„λ£Œν•˜λŠ” 응닡을 μž‘μ„±ν•˜μ„Έμš”.\n\n### λͺ…λ Ήμ–΄:\n{x}\n\n### 응닡:"
    )
    len_prompt = len(prompt)
    gened = model.generate(
        **tokenizer(prompt, return_tensors="pt", return_token_type_ids=False).to(
            device
        ),
        max_new_tokens=1024,
        early_stopping=True,
        do_sample=True,
        top_k=10,
        top_p=0.92,
        no_repeat_ngram_size=3,
        eos_token_id=2,
        repetition_penalty=1.2,
        num_beams=3
    )
    return tokenizer.decode(gened[0])[len_prompt:]

def LLM_infer(input):
    device = (
        torch.device("cuda:0") if torch.cuda.is_available() else torch.device("cpu")
    )
    model_id = "kfkas/Llama-2-ko-7b-Chat"
    model = AutoModelForCausalLM.from_pretrained(
        model_id, device_map={"": 0},torch_dtype=torch.float16, low_cpu_mem_usage=True
    )
    tokenizer = AutoTokenizer.from_pretrained(model_id)
    model.eval()
    model.config.use_cache = (True)
    tokenizer.pad_token = tokenizer.eos_token
    output = gen(input, model=model, tokenizer=tokenizer, device=device)

    return output


if __name__ == "__main__":
    text = LLM_infer("살인죄λ₯Ό μ•Œλ €μ€˜")
    print(text)
```

## Note for oobabooga/text-generation-webui

Remove `ValueError` at `load_tokenizer` function(line 109 or near), in `modules/models.py`.

```python
diff --git a/modules/models.py b/modules/models.py
index 232d5fa..de5b7a0 100644
--- a/modules/models.py
+++ b/modules/models.py
@@ -106,7 +106,7 @@ def load_tokenizer(model_name, model):
                 trust_remote_code=shared.args.trust_remote_code,
                 use_fast=False
             )
-        except ValueError:
+        except:
             tokenizer = AutoTokenizer.from_pretrained(
                 path_to_model,
                 trust_remote_code=shared.args.trust_remote_code,
```

Since Llama-2-Ko uses FastTokenizer provided by HF tokenizers NOT sentencepiece package, 
it is required to use `use_fast=True` option when initialize tokenizer.

Apple Sillicon does not support BF16 computing, use CPU instead. (BF16 is supported when using NVIDIA GPU)


---

> Below is the original model card of the Llama-2 model.

# **Llama 2**
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 7B pretrained model, converted for the Hugging Face Transformers format. Links to other models can be found in the index at the bottom.

## Model Details
*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.*

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.

**Model Developers** Meta

**Variations** Llama 2 comes in a range of parameter sizes β€” 7B, 13B, and 70B β€” as well as pretrained and fine-tuned variations.

**Input** Models input text only.

**Output** Models generate text only.

**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.


||Training Data|Params|Content Length|GQA|Tokens|LR|
|---|---|---|---|---|---|---|
|Llama 2|*A new mix of publicly available online data*|7B|4k|&#10007;|2.0T|3.0 x 10<sup>-4</sup>|
|Llama 2|*A new mix of publicly available online data*|13B|4k|&#10007;|2.0T|3.0 x 10<sup>-4</sup>|
|Llama 2|*A new mix of publicly available online data*|70B|4k|&#10004;|2.0T|1.5 x 10<sup>-4</sup>|

*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.

**Model Dates** Llama 2 was trained between January 2023 and July 2023.

**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.

**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/)

**Research Paper** ["Llama-2: Open Foundation and Fine-tuned Chat Models"](arxiv.org/abs/2307.09288)

## Intended Use
**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.

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).

**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.

## Hardware and Software
**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.

**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.

||Time (GPU hours)|Power Consumption (W)|Carbon Emitted(tCO<sub>2</sub>eq)|
|---|---|---|---|
|Llama 2 7B|184320|400|31.22|
|Llama 2 13B|368640|400|62.44|
|Llama 2 70B|1720320|400|291.42|
|Total|3311616||539.00|

**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.

## Training Data
**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.

**Data Freshness** The pretraining data has a cutoff of September 2022, but some tuning data is more recent, up to July 2023.

## Evaluation Results

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.

|Model|Size|Code|Commonsense Reasoning|World Knowledge|Reading Comprehension|Math|MMLU|BBH|AGI Eval|
|---|---|---|---|---|---|---|---|---|---|
|Llama 1|7B|14.1|60.8|46.2|58.5|6.95|35.1|30.3|23.9|
|Llama 1|13B|18.9|66.1|52.6|62.3|10.9|46.9|37.0|33.9|
|Llama 1|33B|26.0|70.0|58.4|67.6|21.4|57.8|39.8|41.7|
|Llama 1|65B|30.7|70.7|60.5|68.6|30.8|63.4|43.5|47.6|
|Llama 2|7B|16.8|63.9|48.9|61.3|14.6|45.3|32.6|29.3|
|Llama 2|13B|24.5|66.9|55.4|65.8|28.7|54.8|39.4|39.1|
|Llama 2|70B|**37.5**|**71.9**|**63.6**|**69.4**|**35.2**|**68.9**|**51.2**|**54.2**|

**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.

|||TruthfulQA|Toxigen|
|---|---|---|---|
|Llama 1|7B|27.42|23.00|
|Llama 1|13B|41.74|23.08|
|Llama 1|33B|44.19|22.57|
|Llama 1|65B|48.71|21.77|
|Llama 2|7B|33.29|**21.25**|
|Llama 2|13B|41.86|26.10|
|Llama 2|70B|**50.18**|24.60|

**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).


|||TruthfulQA|Toxigen|
|---|---|---|---|
|Llama-2-Chat|7B|57.04|**0.00**|
|Llama-2-Chat|13B|62.18|**0.00**|
|Llama-2-Chat|70B|**64.14**|0.01|

**Evaluation of fine-tuned LLMs on different safety datasets.** Same metric definitions as above.

## Ethical Considerations and Limitations
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.

Please see the Responsible Use Guide available at [https://ai.meta.com/llama/responsible-use-guide/](https://ai.meta.com/llama/responsible-use-guide)

## Reporting Issues
Please report any software β€œbug,” or other problems with the models through one of the following means:
- Reporting issues with the model: [github.com/facebookresearch/llama](http://github.com/facebookresearch/llama)
- Reporting problematic content generated by the model: [developers.facebook.com/llama_output_feedback](http://developers.facebook.com/llama_output_feedback)
- Reporting bugs and security concerns: [facebook.com/whitehat/info](http://facebook.com/whitehat/info)

## Llama Model Index
|Model|Llama2|Llama2-hf|Llama2-chat|Llama2-chat-hf|
|---|---|---|---|---|
|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)|
|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)|
|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)|