各种 Python 的 pip 包和 conda 包的名称都是什么含义?
标签
开发/Python
命令行/pip
命令行/conda
开发/AI/CUDA
字数
264 字
阅读时间
2 分钟
TL;DR
cp38
- CPython 3.8py3
- Python 3cu12
- 适用于 CUDA 12cpu_generic
- 适用于通用 CPUmanylinux1_x86_64
- pypa/manylinux: Python wheels that work on any linux (almost)
通过 pip
和 conda
安装的时候我们会遇到这样的输出:
shell
Collecting torch
Downloading torch-2.2.1-cp38-cp38-manylinux1_x86_64.whl.metadata (25 kB)
Collecting nvidia-cuda-nvrtc-cu12==12.1.105 (from torch)
Using cached nvidia_cuda_nvrtc_cu12-12.1.105-py3-none-manylinux1_x86_64.whl.metadata (1.5 kB)
# ...
manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.0 kB)
Downloading torch-2.2.1-cp38-cp38-manylinux1_x86_64.whl (755.5 MB)
━━━╺━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 65.6/755.5 MB 14.5 MB/s eta 0:00:48
ERROR: Operation cancelled by user
(mlcommons-training) jovyan@neko-python-test-0:~/Git/NVIDIA/apex$ pip install torch==1.12
Collecting torch==1.12
Downloading torch-1.12.0-cp38-cp38-manylinux1_x86_64.whl.metadata (22 kB)
Collecting typing-extensions (from torch==1.12)
Using cached typing_extensions-4.10.0-py3-none-any.whl.metadata (3.0 kB)
Downloading torch-1.12.0-cp38-cp38-manylinux1_x86_64.whl (776.3 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 776.3/776.3 MB 2.1 MB/s eta 0:00:00
Using cached typing_extensions-4.10.0-py3-none-any.whl (33 kB)
Installing collected packages: typing-extensions, torch
Successfully installed torch-1.12.0 typing-extensions-4.10.0
shell
python_abi conda-forge/linux-64::python_abi-3.8-4_cp38
pytorch conda-forge/linux-64::pytorch-2.1.2-cpu_generic_py38h1a28abb_1
pytorch-mutex pytorch/noarch::pytorch-mutex-1.0-cpu
torchaudio pytorch/linux-64::torchaudio-2.1.2-py38_cpu
torchvision pytorch/linux-64::torchvision-0.16.2-py38_cpu
shell
package | build
---------------------------|-----------------
pytorch-2.1.2 |cpu_generic_py38h1a28abb_1 25.2 MB conda-forge
pytorch-mutex-1.0 | cpu 3 KB pytorch
torchaudio-2.1.2 | py38_cpu 4.8 MB pytorch
torchvision-0.16.2 | py38_cpu 11.2 MB pytorch
------------------------------------------------------------
Total: 206.2 MB
这些输出虽然各自有各自的标签习惯,但是整体的规律还是相似的。
写了 cpu
的就是 CPU 可以用的,架构也会一并包含在内。