deep-learning-methods

Deep Learning Methods for University

Requirements

• uv
• FFmpeg (required for audio loading in project2; on Windows install the full-shared build via winget install ffmpeg)

Setup

git clone https://github.com/MaciasMucias/Deep-Learning-Methods.git
cd Deep-Learning-Methods
uv sync --all-packages

Running an experiment from the workspace root

uv run project1-cinic10/srd/project1-cinic10/experiments/train.py

Extending the project

Adding an external package — run from inside the project folder, not the root:

cd project1_cinic10
uv add torchvision

This ensures it goes into that project's pyproject.toml, not the workspace root.

Adding a dependency to the shared library — run from dl_base/:

cd dl_base
uv add numpy

All projects that depend on dl_base will inherit it automatically.

After any pyproject.toml change, run uv sync from the workspace root to update the lockfile and .venv.

Dev-only dependencies (pytest, black, etc.) — use --dev so they don't leak into production installs:

uv add --dev pytest

Importing from the shared library in your project code just works as a normal import — no path hacks needed:

from dl_base.runner import Trainer

If uv sync fails after pulling someone else's changes, it usually means the uv.lock was updated — just re-run uv sync and it'll reconcile.