README.md

This is the Course docker file folder.

Here has 3 kinds of courses, CV, DL, LLM. CV and DL are recommended to run on dGPUs, and LLM is suggested to run on 395.

Upon buiding, the scripts inside each folder will copy related notebooks from projects, download neccessary files, and build then push the image.

Course images remain notebook and course focused. Browser coding environments are built from the separate generic code image line in dockerfiles/Code/ as auplc-code-cpu and auplc-code-gpu; do not create per-course VS Code image variants such as course-specific code-server images.

Use the existing course target for course notebook images:

make -C dockerfiles courses GPU_TARGET=gfx1151

Course resources should land on their course content. Set each official course resource's custom.resources.metadata.<resource>.defaultPath to the same path as the image WORKDIR, such as /opt/workspace/CV. The official image contract verifier checks that metadata and image contract:

make -C dockerfiles verify-resource-contracts

The verifier is for official images. At runtime, the Hub doesn't check whether a configured path exists in arbitrary or custom images. Custom course images must create their configured landing path themselves. For non-official images that already declare the desired Docker WORKDIR, omit defaultPath to preserve the image's initial folder.

Use the code targets only for generic code-server environments:

make -C dockerfiles code-cpu
make -C dockerfiles code-gpu GPU_TARGET=gfx1151
make -C dockerfiles code