Python 3.7 / CUDA 11.7 dependency environment
A Linux x86_64 dependency environment for reproducible container builds. This repository contains an environment archive, not a trained model.
Core versions: Python 3.7.12 and PyTorch 1.13.1 with CUDA 11.7. Retained native libraries are unchanged from the source environment. Cached bytecode, editable-install links and private installation-path metadata are omitted. The optional local MultiScaleDeformableAttention and safetensors packages are omitted.
The exact public python-lzf 0.2.6 source is bundled under share/public-dependency-sources/; compile it with the included Python interpreter during image construction. Dependency installation is a build-time step.
The archive contains per-package upstream license files and an embedded content/provenance manifest. Licenses remain those of the individual dependencies. See PROVENANCE.json and SHA256SUMS for the exact archive identity.
The environment needs a clean Linux image build, import/model checks and hardware validation for each intended application. An archive content audit does not itself establish runtime compatibility.