Sign inSign up

movahub/pytorch

By movahub

•Updated 7 months ago

Pytorch + YOLO model

Image
Networking
Security
Machine learning & AI
0

496

movahub/pytorch repository overview

⁠PyTorch + YOLO model

PyTorch Docker image with CUDA 12.6 and cuDNN 9 (runtime). Suitable for GPU training and inference.

⁠Images

TagDescription
2.9.1-cuda12.6-cudnn9-runtimePyTorch + CUDA + system libs (OpenCV). Generic base.
2.9.1-cuda12.6-cudnn9-yolo26mSame as above + YOLO weights (yolo26m.pt, yolo26n.pt) in /app/weights for Ultralytics offline (e.g. ECS training).

⁠Build (base images)

cd pytorch-docker
# 1. Generic GPU base (no YOLO) — phải build + push trước vì gpu.yolo26m kế thừa từ đây
docker build --platform=linux/amd64 -f Dockerfile.gpu -t movahub/pytorch:2.9.1-cuda12.6-cudnn9-runtime .
docker push movahub/pytorch:2.9.1-cuda12.6-cudnn9-runtime
# 2. GPU + YOLO weights (for sphere-ai training) — extends từ image trên
docker build --platform=linux/amd64 -f Dockerfile.gpu.yolo26m -t movahub/pytorch:2.9.1-cuda12.6-cudnn9-yolo26m .
docker push movahub/pytorch:2.9.1-cuda12.6-cudnn9-yolo26m

⁠Quick start

docker pull movahub/pytorch:2.9.1-cuda12.6-cudnn9-runtime
docker run --gpus all -it movahub/pytorch:2.9.1-cuda12.6-cudnn9-runtime python -c "import torch; print(torch.cuda.is_available())"

⁠Use as base image

FROM movahub/pytorch:2.9.1-cuda12.6-cudnn9-runtime
WORKDIR /app
# Install your dependencies and copy app code

⁠Requirements

  • Docker with NVIDIA Container Toolkit (for GPU support)
  • NVIDIA driver compatible with CUDA 12.6

⁠License

See PyTorch license⁠.

Tag summary

Content type

Image

Digest

sha256:8c4c8a298…

Size

383.6 MB

Last updated

7 months ago

docker pull movahub/pytorch:cpu-python3.11-yolo26m