This project provides over 30 images for the following combinations:
- OS Versions:
jammy,focal - CUDA Versions:
12.4.1,11.8.0 - Container Types:
base,pytorch,tensorflow=2.15.0 - Python Versions:
3.9,3.10,3.11
The resultant images are significantly smaller than the NVIDIA base images, with the smallest image being 399MB, which is 398% smaller than the NVIDIA base image. The images are built using Micromamba, a fast, reliable, and secure package manager for data science and machine learning.
Some of the highlights of the images are:
| Name | Description | Image Size | Nvidia Image Size | Size Reduction |
|---|---|---|---|---|
civo_jammy_python_3.11_cuda_12.4.1_base |
Python 3.11, CUDA 12.4, base image | 399.22MB | 1.99GB | -398.4% |
civo-python-cuda12-pytorch |
Base image with PyTorch, PyTorch-audio, etc. | 7.93GB | 8.68GB | -9.45% |
civo_jammy_python_3.11_cuda_12.4.1_pytorch |
Base image with PyTorch 2.3.0 | 1.99GB | 8.68GB | -336% |
civo_jammy_python_3.11_cuda_12.4.1_tensorflow |
Base image with TensorFlow 2.15.0 | 2.3GB | 6.62GB | -187.83% |
The following table lists the images and their compressed sizes:
| Image Name | Size (in MB) |
|---|---|
| civo_jammy_python_3.10_cuda_12.4.1_pytorch | 709M |
| civo_jammy_python_3.9_cuda_12.4.1_pytorch | 704M |
| civo_jammy_python_3.11_cuda_12.4.1_pytorch | 737M |
| civo_jammy_python_3.9_cuda_12.4.1_tensorflow | 650M |
| civo_focal_python_3.10_cuda_11.8.0_tensorflow | 661M |
| civo_jammy_python_3.10_cuda_12.4.1_base | 170M |
| civo_focal_python_3.9_cuda_11.8.0_pytorch | 704M |
| civo_focal_python_3.11_cuda_11.8.0_pytorch | 737M |
| civo_jammy_python_3.10_cuda_11.8.0_pytorch | 708M |
| civo_jammy_python_3.9_cuda_11.8.0_pytorch | 706M |
| civo_jammy_python_3.11_cuda_12.4.1_base | 186M |
| civo_jammy_python_3.11_cuda_11.8.0_pytorch | 735M |
| civo_jammy_python_3.9_cuda_12.4.1_base | 166M |
| civo_jammy_python_3.9_cuda_11.8.0_tensorflow | 649M |
| civo_jammy_python_3.11_cuda_11.8.0_tensorflow | 677M |
| civo_jammy_python_3.10_cuda_11.8.0_tensorflow | 653M |
| civo_jammy_python_3.10_cuda_12.4.1_tensorflow | 655M |
| civo_jammy_python_3.11_cuda_12.4.1_tensorflow | 678M |
| civo_focal_python_3.9_cuda_11.8.0_tensorflow | 650M |
| civo_focal_python_3.11_cuda_12.4.1_tensorflow | 680M |
| civo_focal_python_3.10_cuda_12.4.1_pytorch | 710M |
| civo_focal_python_3.9_cuda_12.4.1_tensorflow | 651M |
| civo_focal_python_3.10_cuda_11.8.0_pytorch | 709M |
| civo_focal_python_3.11_cuda_11.8.0_tensorflow | 678M |
| civo_focal_python_3.10_cuda_12.4.1_tensorflow | 656M |
| civo_focal_python_3.9_cuda_12.4.1_pytorch | 706M |
| civo_focal_python_3.11_cuda_12.4.1_pytorch | 738M |
| civo_jammy_python_3.9_cuda_11.8.0_base | 165M |
| civo_focal_python_3.10_cuda_11.8.0_base | 170M |
| civo_focal_python_3.11_cuda_11.8.0_base | 186M |
| civo_focal_python_3.9_cuda_11.8.0_base | 166M |
| civo_jammy_python_3.10_cuda_11.8.0_base | 169M |
| civo_jammy_python_3.11_cuda_11.8.0_base | 184M |
| civo_focal_python_3.10_cuda_12.4.1_base | 172M |
| civo_focal_python_3.11_cuda_12.4.1_base | 187M |
| civo_focal_python_3.9_cuda_12.4.1_base | 168M |
| civo_focal_python_3.11_cuda_12.4.1_pytorch | 738M |
| civo_jammy_python_3.9_cuda_11.8.0_base | 165M |
| civo_focal_python_3.10_cuda_11.8.0_base | 170M |
| civo_focal_python_3.11_cuda_11.8.0_base | 186M |
| civo_focal_python_3.9_cuda_11.8.0_base | 166M |
| civo_jammy_python_3.10_cuda_11.8.0_base | 169M |
| civo_jammy_python_3.11_cuda_11.8.0_base | 184M |
| civo_focal_python_3.10_cuda_12.4.1_base | 172M |
| civo_focal_python_3.11_cuda_12.4.1_base | 187M |
| civo_focal_python_3.9_cuda_12.4.1_base | 168M |
Check out civo-vllm-docker for an example implementation.