ModelArts的推理平台提供了一系列的基础镜像,用户可以基于这些基础镜像构建自定义镜像,用于部署推理服务。
X86架构(CPU/GPU)的推理基础镜像
表1 TensorFlow
AI引擎版本 |
支持的运行环境 |
镜像名称 |
URI |
2.1.0 |
CPU
GPU(cuda10.1) |
tensorflow_2.1.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64 |
swr.{region_id}.myhuaweicloud.com/atelier/tensorflow_2_1:tensorflow_2.1.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64-20221121111529-d65d817 |
1.15.5 |
CPU
GPU(cuda11.4) |
tensorflow_1.15.5-cuda_11.4-py_3.8-ubuntu_20.04-x86_64 |
swr.{region_id}.myhuaweicloud.com/aip/tensorflow_1_15:tensorflow_1.15.5-cuda_11.4-py_3.8-ubuntu_20.04-x86_64-20220524162601-50d6a18 |
2.6.0 |
CPU
GPU(cuda11.2) |
tensorflow_2.6.0-cuda_11.2-py_3.7-ubuntu_18.04-x86_64 |
swr.{region_id}.myhuaweicloud.com/aip/tensorflow_2_6:tensorflow_2.6.0-cuda_11.2-py_3.7-ubuntu_18.04-x86_64-20220524162601-50d6a18 |
表2 PyTorch
AI引擎版本 |
支持的运行环境 |
镜像名称 |
URI |
1.8.0 |
CPU
GPU(cuda10.2) |
pytorch_1.8.0-cuda_10.2-py_3.7-ubuntu_18.04-x86_64 |
swr.{region_id}.myhuaweicloud.com/atelier/pytorch_1_8:pytorch_1.8.0-cuda_10.2-py_3.7-ubuntu_18.04-x86_64-20221118143845-d65d817 |
1.8.2 |
CPU
GPU(cuda11.1) |
pytorch_1.8.2-cuda_11.1-py_3.7-ubuntu_18.04-x86_64 |
swr.{region_id}.myhuaweicloud.com/aip/pytorch_1_8:pytorch_1.8.2-cuda_11.1-py_3.7-ubuntu_18.04-x86_64-20220524162601-50d6a18 |
表3 MindSpore
AI引擎版本 |
支持的运行环境 |
镜像名称 |
URI |
1.7.0 |
CPU |
mindspore_1.7.0-cpu-py_3.7-ubuntu_18.04-x86_64 |
swr.{region_id}.myhuaweicloud.com/atelier/mindspore_1_7_0:mindspore_1.7.0-cpu-py_3.7-ubuntu_18.04-x86_64-20220702120711-8590b76 |
1.7.0 |
GPU(cuda10.1) |
mindspore_1.7.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64 |
swr.{region_id}.myhuaweicloud.com/atelier/mindspore_1_7_0:mindspore_1.7.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64-20220702120711-8590b76 |
1.7.0 |
GPU(cuda11.1) |
mindspore_1.7.0-cuda_11.1-py_3.7-ubuntu_18.04-x86_64 |
swr.{region_id}.myhuaweicloud.com/atelier/mindspore_1_7_0:mindspore_1.7.0-cuda_11.1-py_3.7-ubuntu_18.04-x86_64-20220702120711-8590b76 |
推理基础镜像详情TensorFlow(CPU/GPU)
ModelArts提供了以下TensorFlow(CPU/GPU)推理基础镜像:
引擎版本一:tensorflow_2.1.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64
- 镜像地址:swr.{region_id}.myhuaweicloud.com/atelier/tensorflow_2_1:tensorflow_2.1.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64-20221121111529-d65d817
- 镜像构建时间:20220713110657(yyyy-mm-dd-hh-mm-ss)
- 镜像系统版本:Ubuntu 18.04.4 LTS
- cuda:10.1.243
- cudnn:7.6.5.32
- Python解释器路径及版本:/home/ma-user/anaconda3/envs/TensorFlow-2.1/bin/python, python 3.7.10
- 三方包安装路径:/home/ma-user/anaconda3/envs/TensorFlow-2.1/lib/python3.7/site-packages
引擎版本二: tensorflow_1.15.5-cuda_11.4-py_3.8-ubuntu_20.04-x86_64
- 镜像地址:swr.{region_id}.myhuaweicloud.com/aip/tensorflow_1_15:tensorflow_1.15.5-cuda_11.4-py_3.8-ubuntu_20.04-x86_64-20220524162601-50d6a18
- 镜像构建时间:20220524162601(yyyy-mm-dd-hh-mm-ss)
- 镜像系统版本:Ubuntu 20.04.4 LTS
- cuda:11.4.3
- cudnn:8.2.4.15
- Python解释器路径及版本:/home/ma-user/anaconda3/envs/TensorFlow-1.15.5/bin/python, python 3.8.13
- 三方包安装路径:/home/ma-user/anaconda3/envs/TensorFlow-1.15.5/lib/python3.8/site-packages
引擎版本三:tensorflow_2.6.0-cuda_11.2-py_3.7-ubuntu_18.04-x86_64
- 镜像地址:swr.{region_id}.myhuaweicloud.com/aip/tensorflow_2_6:tensorflow_2.6.0-cuda_11.2-py_3.7-ubuntu_18.04-x86_64-20220524162601-50d6a18
- 镜像构建时间:20220524162601(yyyy-mm-dd-hh-mm-ss)
- 镜像系统版本:Ubuntu 18.04.4 LTS
- cuda:11.2.0
- cudnn:8.1.1.33
- Python解释器路径及版本:/home/ma-user/anaconda3/envs/TensorFlow-2.6.0/bin/python, python 3.7.10
- 三方包安装路径:/home/ma-user/anaconda3/envs/TensorFlow-2.6.0/lib/python3.7/site-packages
推理基础镜像详情PyTorch(CPU/GPU)
ModelArts提供了以下PyTorch(CPU/GPU)推理基础镜像:
引擎版本一:pytorch_1.8.0-cuda_10.2-py_3.7-ubuntu_18.04-x86_64
- 镜像地址:swr.{region_id}.myhuaweicloud.com/atelier/pytorch_1_8:pytorch_1.8.0-cuda_10.2-py_3.7-ubuntu_18.04-x86_64-20221118143845-d65d817
- 镜像构建时间:20220713110657 (yyyy-mm-dd-hh-mm-ss)
- 镜像系统版本:Ubuntu 18.04.4 LTS
- cuda:10.2.89
- cudnn:7.6.5.32
- Python解释器路径及版本:/home/ma-user/anaconda3/envs/PyTorch-1.8/bin/python, python 3.7.10
- 三方包安装路径:/home/ma-user/anaconda3/envs/PyTorch-1.8/lib/python3.7/site-packages
引擎版本二:pytorch_1.8.2-cuda_11.1-py_3.7-ubuntu_18.04-x86_64
- 镜像地址:swr.{region_id}.myhuaweicloud.com/aip/pytorch_1_8:pytorch_1.8.2-cuda_11.1-py_3.7-ubuntu_18.04-x86_64-20220524162601-50d6a18
- 镜像构建时间:20220524162601(yyyy-mm-dd-hh-mm-ss)
- 镜像系统版本:Ubuntu 18.04.4 LTS
- cuda:11.1.1
- cudnn:8.0.5.39
- Python解释器路径及版本:/home/ma-user/anaconda3/envs/PyTorch-1.8.2/bin/python, python 3.7.10
- 三方包安装路径:/home/ma-user/anaconda3/envs/PyTorch-1.8.2/lib/python3.7/site-packages
推理基础镜像详情MindSpore(CPU/GPU)
ModelArts提供了以下MindSpore(CPU/GPU)推理基础镜像:
引擎版本一:mindspore_1.7.0-cpu-py_3.7-ubuntu_18.04-x86_64
- 镜像地址:swr.{region_id}.myhuaweicloud.com/atelier/mindspore_1_7_0:mindspore_1.7.0-cpu-py_3.7-ubuntu_18.04-x86_64-20220702120711-8590b76
- 镜像构建时间:20220702120711(yyyy-mm-dd-hh-mm-ss)
- 镜像系统版本:Ubuntu 18.04.4 LTS
- Python解释器路径及版本:/home/ma-user/anaconda3/envs/MindSpore/bin/python, python 3.7.10
- 三方包安装路径:/home/ma-user/anaconda3/envs/MindSpore/lib/python3.7/site-packages
引擎版本二:mindspore_1.7.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64
- 镜像地址:swr.{region_id}.myhuaweicloud.com/atelier/mindspore_1_7_0:mindspore_1.7.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64-20220702120711-8590b76
- 镜像构建时间:20220702120711(yyyy-mm-dd-hh-mm-ss)
- 镜像系统版本:Ubuntu 18.04.4 LTS
- cuda:10.1.243
- cudnn:7.6.5.32
- Python解释器路径及版本:/home/ma-user/anaconda3/envs/MindSpore/bin/python, python 3.7.10
- 三方包安装路径:/home/ma-user/anaconda3/envs/MindSpore/lib/python3.7/site-packages
引擎版本三:mindspore_1.7.0-cuda_11.1-py_3.7-ubuntu_18.04-x86_64
- 镜像地址:swr.{region_id}.myhuaweicloud.com/atelier/mindspore_1_7_0:mindspore_1.7.0-cuda_11.1-py_3.7-ubuntu_18.04-x86_64-20220702120711-8590b76
- 镜像构建时间:20220702120711(yyyy-mm-dd-hh-mm-ss)
- 镜像系统版本:Ubuntu 18.04.4 LTS
- cuda:11.1.1
- cudnn:8.0.5.39
- Python解释器路径及版本:/home/ma-user/anaconda3/envs/MindSpore/bin/python, python 3.7.10
- 三方包安装路径:/home/ma-user/anaconda3/envs/MindSpore/lib/python3.7/site-packages