README.md 12.4 KB
Newer Older
Rayyyyy's avatar
Rayyyyy committed
1
2
3
4
5
# llama3
## 论文
[llama3](https://llama.meta.com/llama3/)

## 模型结构
Rayyyyy's avatar
Rayyyyy committed
6
7
8
9
10
Llama-3中选择了一个相对标准的decoder-only的transformer架构。与Llama-2相比,做了几个关键的改进:
- 基于超过15T token训练数据,大小相当于Llama 2数据集的7倍还多,增强了推理、代码生成和指令跟随等方面的能力;
- 支持8K长文本(之前是4k),改进的tokenizer具有128K tokens的词汇量,可以更有效地对语言进行编码,从而大大提高了模型的性能;
- 采用分组查询注意力(grouped query attention,GQA)、掩码等技术,帮助开发者以最低的能耗获取绝佳的性能。
- 在8,192个tokens的序列上训练模型,使用掩码来确保self-attention不会跨越文档边界。
Rayyyyy's avatar
Rayyyyy committed
11

Rayyyyy's avatar
Rayyyyy committed
12
13
14
15
16
## 算法原理
<div align=center>
    <img src="./doc/method.png"/>
</div>

Rayyyyy's avatar
Rayyyyy committed
17
18
19

## 环境配置
-v 路径、docker_name和imageID根据实际情况修改
Rayyyyy's avatar
Rayyyyy committed
20
**注意**:bitsandbytes库功能不全,暂不支持量化相关
Rayyyyy's avatar
Rayyyyy committed
21
22
23

### Docker(方法一)
```bash
Rayyyyy's avatar
Rayyyyy committed
24
25
docker pull image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.1.0-centos7.6-dtk24.04-py310
docker run -it -v /path/your_code_data/:/path/your_code_data/ -v /opt/hyhal/:/opt/hyhal/:ro --shm-size=80G --privileged=true --device=/dev/kfd --device=/dev/dri/ --group-add video --name docker_name imageID bash
Rayyyyy's avatar
Rayyyyy committed
26
27
28
29
30
31
32

cd /your_code_path/llama3_pytorch
pip install -e .
```

### Dockerfile(方法二)
```bash
Rayyyyy's avatar
Rayyyyy committed
33
cd docker
Rayyyyy's avatar
Rayyyyy committed
34
docker build --no-cache -t llama3:latest .
Rayyyyy's avatar
Rayyyyy committed
35
docker run -it -v /path/your_code_data/:/path/your_code_data/ -v /opt/hyhal/:/opt/hyhal/:ro --shm-size=80G --privileged=true --device=/dev/kfd --device=/dev/dri/ --group-add video --name docker_name imageID bash
Rayyyyy's avatar
Rayyyyy committed
36
37
38
39
40
41
42
43

cd /your_code_path/llama3_pytorch
pip install -e .
```

### Anaconda(方法三)
关于本项目DCU显卡所需的特殊深度学习库可从[光合](https://developer.hpccube.com/tool/)开发者社区下载安装。
```bash
Rayyyyy's avatar
Rayyyyy committed
44
DTK驱动: dtk24.04
Rayyyyy's avatar
Rayyyyy committed
45
python: python3.10
Rayyyyy's avatar
Rayyyyy committed
46
torch: 2.1.0
Rayyyyy's avatar
Rayyyyy committed
47
xtuner: 0.1.18
Rayyyyy's avatar
Rayyyyy committed
48
```
Rayyyyy's avatar
Rayyyyy committed
49
`Tips:以上dtk驱动、python、torch等DCU相关工具版本需要严格一一对应`
Rayyyyy's avatar
Rayyyyy committed
50
51
52
53
54
55
56

其它非深度学习库安装方式如下:
```bash
pip install -e .
```

## 数据集
Rayyyyy's avatar
Rayyyyy committed
57
58
59
60
61
62
```
├── llama3_pytorch
│   ├── datasets
│       ├── alpaca_data.json
│       └── multi_turn_dataset_2.json
```
Rayyyyy's avatar
Rayyyyy committed
63
64

## 训练
Rayyyyy's avatar
Rayyyyy committed
65
### xtuner微调方法
Rayyyyy's avatar
Rayyyyy committed
66
1. 训练库安装(非llama3_pytorch目录下),请注意所需库版本
Rayyyyy's avatar
Rayyyyy committed
67
```bash
Rayyyyy's avatar
Rayyyyy committed
68
69
pip uninstall flash-attn # 2.0.4+82379d7.abi0.dtk2404.torch2.1
# docker环境含有deepspeed的可不进行安装, 需要对照版本是否一致即可
Rayyyyy's avatar
Rayyyyy committed
70
pip install deepspeed-0.12.3+das1.0+gita724046.abi0.dtk2404.torch2.1.0-cp310-cp310-manylinux2014_x86_64.whl
Rayyyyy's avatar
Rayyyyy committed
71
72
73
git clone -b v0.1.18 https://github.com/InternLM/xtuner.git
cd xtuner
pip install -e '.[all]'
Rayyyyy's avatar
Rayyyyy committed
74
pip install mmengine==0.10.3
Rayyyyy's avatar
Rayyyyy committed
75
76
# 注意bitsandbytes库版本,如果环境中一致可不安装,否则需要重新安装
pip install bitsandbytes-0.37.0+das1.0+gitd3d888f.abi0.dtk2404.torch2.1-py3-none-any.whl
Rayyyyy's avatar
Rayyyyy committed
77
78
```
2. 下载预训练模型,具体模型请修改`download_models.py`
Rayyyyy's avatar
Rayyyyy committed
79
80
81
82
83
```bash
cd /your_code_path/llama3_pytorch
pip install modelscope
python download_models.py
```
Rayyyyy's avatar
Rayyyyy committed
84
85
86
87
3. 修改[llama3_8b_instruct_qlora_alpaca_e3_M.py](./llama3_8b_instruct_qlora_alpaca_e3_M.py)代码中的`pretrained_model_name_or_path``data_path`为本地模型、数据地址;
4. 根据硬件环境和自身训练需求来调整`max_length``batch_size``accumulative_counts``max_epochs``lr``save_steps``evaluation_freq`、model.lora中的`r``lora_alpha`参数,默认参数支持4*32G;
5. ${DCU_NUM}参数修改为要使用的DCU卡数量,不同数据集需要修改llama3_8b_instruct_qlora_alpaca_e3_M.py中`SYSTEM``evaluation_inputs``dataset_map_fn``train_dataloader.sampler``train_cfg`参数设置,详情请参考代码注释项,当前默认alpaca数据集,**`--work-dir`设定保存模型路径**
6. 执行
Rayyyyy's avatar
Rayyyyy committed
88
```bash
Rayyyyy's avatar
Rayyyyy committed
89
90
bash finetune.sh
or
Rayyyyy's avatar
Rayyyyy committed
91
92
NPROC_PER_NODE=${DCU_NUM} xtuner train ./llama3_8b_instruct_qlora_alpaca_e3_M.py --deepspeed deepspeed_zero2
```
Rayyyyy's avatar
Rayyyyy committed
93
94

## 推理
Rayyyyy's avatar
Rayyyyy committed
95
预训练模型下载方法请参考下面的[预训练权重](#预训练权重)章节,不同的模型需要不同的模型并行(MP)值,如下表所示:
Rayyyyy's avatar
Rayyyyy committed
96

Rayyyyy's avatar
Rayyyyy committed
97
98
99
|  Model | MP |
|--------|----|
| 8B     | 1  |
Rayyyyy's avatar
Rayyyyy committed
100
| 70B    | 8  |
Rayyyyy's avatar
Rayyyyy committed
101
102
103
104

所有模型都支持序列长度高达8192个tokens,但我们根据max_seq_len和max_batch_size值预先分配缓存。根据你的硬件设置。

**Tips:**
Rayyyyy's avatar
Rayyyyy committed
105
- `–nproc_per_node`需要根据模型的MP值进行设置(参考上表)。
Rayyyyy's avatar
Rayyyyy committed
106
107
108
- `max_seq_len``max_batch_size`参数按需设置。

### Pretrained模型
Rayyyyy's avatar
Rayyyyy committed
109
这些模型都没有针对聊天或者Q&A进行微调。可以参考`example_text_completion.py`里的用例。
Rayyyyy's avatar
Rayyyyy committed
110

Rayyyyy's avatar
Rayyyyy committed
111
- Meta-Llama-3-8B 模型示例,Meta-Llama-3-70B模型仅需替换–-nproc_per_node、--ckpt_dir、--tokenizer_path对应模型地址即可。
Rayyyyy's avatar
Rayyyyy committed
112
```bash
Rayyyyy's avatar
Rayyyyy committed
113
torchrun --nproc_per_node 1 example_text_completion.py \
Rayyyyy's avatar
Rayyyyy committed
114
    --ckpt_dir Meta-Llama-3-8B/original/ \
Rayyyyy's avatar
Rayyyyy committed
115
116
117
118
    --tokenizer_path Meta-Llama-3-8B/original/tokenizer.model \
    --max_seq_len 128 --max_batch_size 4
```

Rayyyyy's avatar
Rayyyyy committed
119
### Instruction-tuned模型
Rayyyyy's avatar
Rayyyyy committed
120
经过微调的模型被训练用于对话应用程序。为了获得模型的预期特性和性能,需要遵循 [`ChatFormat`](llama/tokenizer.py#L202)中定义的特定格式:
Rayyyyy's avatar
Rayyyyy committed
121
122
- 提示以特殊令牌`<|begin_of_text|>`开始,之后跟随一个或多个消息。
- 每条消息以标签`<|start_header_id|>`开始,角色为`system``user`或者`assistant`、并以标签`<|end_header_id|>`结束。
Rayyyyy's avatar
Rayyyyy committed
123
124
- 在双换行符`\n\n`之后,消息的内容随之而来。
- 每条消息的结尾由`<|eot_id|>`令牌标记。
Rayyyyy's avatar
Rayyyyy committed
125
126
127

您还可以部署额外的分类器来过滤被认为不安全的输入和输出。有关如何向推理代码的输入和输出添加安全检查器,请参阅[llama-recipes repo](https://github.com/meta-llama/llama-recipes/blob/main/recipes/inference/local_inference/inference.py)

Rayyyyy's avatar
Rayyyyy committed
128
- Meta-Llama-3-8B-Instruct 模型示例,Meta-Llama-3-70B-Instruct模型仅需替换–-nproc_per_node、--ckpt_dir、--tokenizer_path对应模型地址即可。
Rayyyyy's avatar
Rayyyyy committed
129
130
131
132
133
134
```bash
torchrun --nproc_per_node 1 example_chat_completion.py \
    --ckpt_dir Meta-Llama-3-8B-Instruct/original/ \
    --tokenizer_path Meta-Llama-3-8B-Instruct/original/tokenizer.model \
    --max_seq_len 512 --max_batch_size 6
```
Rayyyyy's avatar
Rayyyyy committed
135

Rayyyyy's avatar
Rayyyyy committed
136
137
## 多轮对话
1. 确认环境安装及模型下载完毕;
Rayyyyy's avatar
Rayyyyy committed
138
2. 修改[chat.sh](./chat.sh)文件中的`--ckpt_dir``--tokenizer_path`参数为本地模型地址,`--max_seq_len`根据自身需求进行修改,调整该值可以增加多轮对话模型的记忆长度,不过需要注意的是这可能会增加模型运算的时间和内存需求;
Rayyyyy's avatar
Rayyyyy committed
139
140
141
142
3. 执行:
```bash
bash chat.sh
```
Rayyyyy's avatar
Rayyyyy committed
143

Rayyyyy's avatar
Rayyyyy committed
144
## Evaluation
Rayyyyy's avatar
Rayyyyy committed
145
1. 安装`llama-recipes``lm-eval`
Rayyyyy's avatar
Rayyyyy committed
146
147
148
149
150
151
152
153
```bash
# llama-recipes 下载
git clone https://github.com/meta-llama/llama-recipes.git
cd ./llama-recipes/recipes/evaluation/
# 修改eval.py第15行代码,将from lm_eval.utils import make_table 改为
from lm_eval.evaluator import make_table
# 修改eval.py第121行代码,num_fewshot参数的默认值改为0
default=0
Rayyyyy's avatar
Rayyyyy committed
154
155
# 修改eval.py第215行代码,use_cache=args.use_cache 修改为
no_cache=args.use_cache
Rayyyyy's avatar
Rayyyyy committed
156
157
158

# 返回根目录
cd ~
Rayyyyy's avatar
Rayyyyy committed
159

Rayyyyy's avatar
Rayyyyy committed
160
161
162
163
164
165
166
# lm-eval 下载
git clone http://developer.hpccube.com/codes/chenych/lm-evaluation-harness.git
cd ./lm-evaluation-harness/
pip install -e .
cd ../
```

Rayyyyy's avatar
Rayyyyy committed
167
2. 修改待测模型**pretrained**参数地址,例如`/home/Meta-Llama-3-8B-Instruct`,特别地,当前仅支持`hellaswag`数据集进行测试验证。执行以下命令:
Rayyyyy's avatar
Rayyyyy committed
168
```bash
Rayyyyy's avatar
Rayyyyy committed
169
cd /path_of/llama-recipes/recipes/evaluation
Rayyyyy's avatar
Rayyyyy committed
170
171
python eval.py --model hf --model_args pretrained=/home/llama3/Meta-Llama-3-8B-Instruct,dtype="float" --tasks hellaswag --device cuda --batch_size 8
```
Rayyyyy's avatar
Rayyyyy committed
172
173
174
<div align=center>
    <img src="./doc/evaluation.png"/>
</div>
Rayyyyy's avatar
Rayyyyy committed
175

Rayyyyy's avatar
Rayyyyy committed
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
## result
- Meta-Llama-3-8B-Instruct
<div align=center>
    <img src="./doc/Meta-Llama-3-8B-Instruct.png"/>
</div>

- Meta-Llama-3-8B
<div align=center>
    <img src="./doc/Meta-Llama-3-8B.png"/>
</div>

### 精度
暂无

## 应用场景
### 算法类别
对话问答

### 热点应用行业
制造,广媒,家居,教育

## 预训练权重
1. 环境安装
```bash
pip install -U huggingface_hub hf_transfer
export HF_ENDPOINT=https://hf-mirror.com
```

2. 预训练模型下载,**token**参数通过huggingface账号获取
Rayyyyy's avatar
Rayyyyy committed
205
206
207
208
209
210
211

- Meta-Llama-3-8B 模型
```bash
mkdir Meta-Llama-3-8B
huggingface-cli download meta-llama/Meta-Llama-3-8B --include "original/*" --local-dir Meta-Llama-3-8B --token hf_*
```

Rayyyyy's avatar
Rayyyyy committed
212
213
214
- Meta-Llama-3-8B-Instruct 模型
```bash
mkdir Meta-Llama-3-8B-Instruct
Rayyyyy's avatar
Rayyyyy committed
215
huggingface-cli download meta-llama/Meta-Llama-3-8B-Instruct --include "original/*" --local-dir Meta-Llama-3-8B-Instruct --token hf_*
Rayyyyy's avatar
Rayyyyy committed
216
```
Rayyyyy's avatar
Rayyyyy committed
217

Rayyyyy's avatar
Rayyyyy committed
218
219
220
221
222
223
224
225
226
227
228
229
- Meta-Llama-3-70B 模型
```bash
mkdir Meta-Llama-3-70B
huggingface-cli download meta-llama/Meta-Llama-3-70B --include "original/*" --local-dir Meta-Llama-3-70B --token hf_*
```

- Meta-Llama-3-70B-Instruct 模型
```bash
mkdir Meta-Llama-3-70B-Instruct
huggingface-cli download meta-llama/Meta-Llama-3-70B-Instruct --include "original/*" --local-dir Meta-Llama-3-70B-Instruct --token hf_*
```

Rayyyyy's avatar
Rayyyyy committed
230
231
模型目录结构如下:
```bash
Rayyyyy's avatar
Rayyyyy committed
232
├── model_save_path
Rayyyyy's avatar
Rayyyyy committed
233
│   ├── Meta-Llama-3-8B
Rayyyyy's avatar
Rayyyyy committed
234
│       ├── original
Rayyyyy's avatar
Rayyyyy committed
235
236
237
│           ├── consolidated.00.pth
│           ├── params.json
│           └── tokenizer.model
Rayyyyy's avatar
Rayyyyy committed
238
239
240
241
242
243
244
245
246
247
248
249
250
251
│       ├── config.json
│       ├── configuration.json
│       ├── generation_config.json
│       ├── LICENSE
│       ├── model-00001-of-00004.safetensors
│       ├── model-00002-of-00004.safetensors
│       ├── model-00003-of-00004.safetensors
│       ├── model-00004-of-00004.safetensors
│       ├── model.safetensors.index.json
│       ├── README.md
│       ├── special_tokens_map.json
│       ├── tokenizer_config.json
│       ├── tokenizer.json
│       └── USE_POLICY.md
Rayyyyy's avatar
Rayyyyy committed
252
│   ├── Meta-Llama-3-8B-Instruct
Rayyyyy's avatar
Rayyyyy committed
253
│       ├── original
Rayyyyy's avatar
Rayyyyy committed
254
255
256
│           ├── consolidated.00.pth
│           ├── params.json
│           └── tokenizer.model
Rayyyyy's avatar
Rayyyyy committed
257
258
259
260
261
262
263
264
265
266
267
268
269
270
│       ├── config.json
│       ├── configuration.json
│       ├── generation_config.json
│       ├── LICENSE
│       ├── model-00001-of-00004.safetensors
│       ├── model-00002-of-00004.safetensors
│       ├── model-00003-of-00004.safetensors
│       ├── model-00004-of-00004.safetensors
│       ├── model.safetensors.index.json
│       ├── README.md
│       ├── special_tokens_map.json
│       ├── tokenizer_config.json
│       ├── tokenizer.json
│       └── USE_POLICY.md
Rayyyyy's avatar
Rayyyyy committed
271
│   ├── Meta-Llama-3-70B
Rayyyyy's avatar
Rayyyyy committed
272
│       ├── original
Rayyyyy's avatar
Rayyyyy committed
273
│           ├── consolidated.00.pth
Rayyyyy's avatar
Rayyyyy committed
274
│           ...
Rayyyyy's avatar
Rayyyyy committed
275
276
277
│           ├── consolidated.07.pth
│           ├── params.json
│           └── tokenizer.model
Rayyyyy's avatar
Rayyyyy committed
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
│       ├── config.json
│       ├── generation_config.json
│       ├── LICENSE
│       ├── README.md
│       ├── model-00001-of-00030.safetensors
│       ├── model-00002-of-00030.safetensors
│       ...
│       ├── model-000029-of-00030.safetensors
│       ├── model-000030-of-00030.safetensors
│       ├── model.safetensors.index.json
│       ├── tokenizer_config.json
│       ├── tokenizer.json
│       ├── tokenizer_config.json
│       ├── special_tokens_map.json
│       └── USE_POLICY.md
Rayyyyy's avatar
Rayyyyy committed
293
│   └── Meta-Llama-3-70B-Instruct
Rayyyyy's avatar
Rayyyyy committed
294
│       ├── original
Rayyyyy's avatar
Rayyyyy committed
295
│           ├── consolidated.00.pth
Rayyyyy's avatar
Rayyyyy committed
296
│           ...
Rayyyyy's avatar
Rayyyyy committed
297
│           ├── consolidated.07.pth
Rayyyyy's avatar
Rayyyyy committed
298
299
│           ├── params.json
│           └── tokenizer.model
Rayyyyy's avatar
Rayyyyy committed
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
│       ├── config.json
│       ├── generation_config.json
│       ├── LICENSE
│       ├── README.md
│       ├── model-00001-of-00030.safetensors
│       ├── model-00002-of-00030.safetensors
│       ...
│       ├── model-000029-of-00030.safetensors
│       ├── model-000030-of-00030.safetensors
│       ├── model.safetensors.index.json
│       ├── tokenizer_config.json
│       ├── tokenizer.json
│       ├── tokenizer_config.json
│       ├── special_tokens_map.json
│       └── USE_POLICY.md
Rayyyyy's avatar
Rayyyyy committed
315
316
```

Rayyyyy's avatar
Rayyyyy committed
317
318
319
320
321
## 源码仓库及问题反馈
- https://developer.hpccube.com/codes/modelzoo/llama3_pytorch

## 参考资料
- https://github.com/meta-llama/llama3
Rayyyyy's avatar
Rayyyyy committed
322
- https://github.com/InternLM/xtuner
Rayyyyy's avatar
Rayyyyy committed
323
- https://github.com/meta-llama/llama-recipes