AI开发平台MODELARTS-推理场景介绍:支持的模型列表和权重文件
支持的模型列表和权重文件
本方案支持vLLM的v0.6.0版本。不同vLLM版本支持的模型列表有差异,具体如表3所示。
序号 |
模型名称 |
是否支持fp16/bf16推理 |
是否支持W4A16量化 |
是否支持W8A8量化 |
是否支持W8A16量化 |
是否支持 kv-cache-int8量化 |
开源权重获取地址 |
---|---|---|---|---|---|---|---|
1 |
llama-7b |
√ |
√ |
√ |
√ |
√ |
|
2 |
llama-13b |
√ |
√ |
√ |
√ |
√ |
|
3 |
llama-65b |
√ |
√ |
√ |
√ |
√ |
|
4 |
llama2-7b |
√ |
√ |
√ |
√ |
√ |
|
5 |
llama2-13b |
√ |
√ |
√ |
√ |
√ |
|
6 |
llama2-70b |
√ |
√ |
√ |
√ |
√ |
|
7 |
llama3-8b |
√ |
√ |
√ |
√ |
√ |
|
8 |
llama3-70b |
√ |
√ |
√ |
√ |
√ |
|
9 |
yi-6b |
√ |
√ |
√ |
√ |
√ |
|
10 |
yi-9b |
√ |
√ |
√ |
√ |
√ |
|
11 |
yi-34b |
√ |
√ |
√ |
√ |
√ |
|
12 |
deepseek-llm-7b |
√ |
x |
x |
x |
x |
|
13 |
deepseek-coder-33b-instruct |
√ |
x |
x |
x |
x |
https://huggingface.co/deepseek-ai/deepseek-coder-33b-instruct |
14 |
deepseek-llm-67b |
√ |
x |
x |
x |
x |
|
15 |
qwen-7b |
√ |
√ |
√ |
√ |
x |
|
16 |
qwen-14b |
√ |
√ |
√ |
√ |
x |
|
17 |
qwen-72b |
√ |
√ |
√ |
√ |
x |
|
18 |
qwen1.5-0.5b |
√ |
√ |
√ |
√ |
x |
|
19 |
qwen1.5-7b |
√ |
√ |
√ |
√ |
x |
|
20 |
qwen1.5-1.8b |
√ |
√ |
√ |
√ |
x |
|
21 |
qwen1.5-14b |
√ |
√ |
√ |
√ |
x |
|
22 |
qwen1.5-32b |
√ |
√ |
√ |
√ |
x |
|
23 |
qwen1.5-72b |
√ |
√ |
√ |
√ |
x |
|
24 |
qwen1.5-110b |
√ |
√ |
√ |
√ |
x |
|
25 |
qwen2-0.5b |
√ |
√ |
√ |
√ |
x |
|
26 |
qwen2-1.5b |
√ |
√ |
√ |
√ |
x |
|
27 |
qwen2-7b |
√ |
√ |
x |
√ |
x |
|
28 |
qwen2-72b |
√ |
√ |
√ |
√ |
x |
|
29 |
baichuan2-7b |
√ |
x |
x |
√ |
x |
|
30 |
baichuan2-13b |
√ |
x |
x |
√ |
x |
|
31 |
gemma-2b |
√ |
x |
x |
x |
x |
|
32 |
gemma-7b |
√ |
x |
x |
x |
x |
|
33 |
chatglm2-6b |
√ |
x |
x |
x |
x |
|
34 |
chatglm3-6b |
√ |
x |
x |
x |
x |
|
35 |
glm-4-9b |
√ |
x |
x |
x |
x |
|
36 |
mistral-7b |
√ |
x |
x |
x |
x |
|
37 |
mixtral-8x7b |
√ |
x |
x |
x |
x |
|
38 |
falcon-11b |
√ |
x |
x |
x |
x |
|
39 |
qwen2-57b-a14b |
√ |
x |
x |
x |
x |
|
40 |
llama3.1-8b |
√ |
√ |
√ |
√ |
x |
https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct |
41 |
llama3.1-70b |
√ |
√ |
√ |
√ |
x |
https://huggingface.co/meta-llama/Meta-Llama-3.1-70B-Instruct |
42 |
llama-3.1-405B |
√ |
√ |
x |
x |
x |
https://huggingface.co/hugging-quants/Meta-Llama-3.1-405B-Instruct-AWQ-INT4 |
43 |
llava-1.5-7b |
√ |
x |
x |
x |
x |
|
44 |
llava-1.5-13b |
√ |
x |
x |
x |
x |
|
45 |
llava-v1.6-7b |
√ |
x |
x |
x |
x |
https://huggingface.co/llava-hf/llava-v1.6-vicuna-7b-hf/tree/main |
46 |
llava-v1.6-13b |
√ |
x |
x |
x |
x |
https://huggingface.co/llava-hf/llava-v1.6-vicuna-13b-hf/tree/main |
47 |
llava-v1.6-34b |
√ |
x |
x |
x |
x |
|
48 |
internvl2-26B |
√ |
x |
x |
x |
x |
|
49 |
MiniCPM-v2.6 |
√ |
x |
x |
x |
x |
|
50 |
deepseek-v2-236b |
x |
x |
√ |
x |
x |
|
51 |
deepseek-v2-lite-16b |
√ |
x |
√ |
x |
x |
各模型支持的卡数请参见附录:基于vLLM不同模型推理支持最小卡数和最大序列说明章节。
- ModelArts推理部署_OBS导入_模型包规范-华为云
- ModelArts计费说明_计费简介_ModelArts怎么计费
- ModelArts模型训练_创建训练作业_如何创建训练作业
- ModelArts推理部署_纳管Atlas 500_边缘服务-华为云
- ModelArts推理部署_模型_AI应用来源-华为云
- 华为云内容审核服务_内容审核有什么作用_华为云内容审核的优势
- ModelArts是什么_AI开发平台_ModelArts功能
- 弹性文件存储介绍_好用的文件存储_文件存储免费试用-华为云
- 华为云CodeArts API_API设计支持公共模型
- ModelArts模型训练_模型训练简介_如何训练模型