empirischtech Inference VM — Third-Party Software Licenses
Last updated: May 2026
This virtual machine image bundles the following open-source software and proprietary components. Each retains its own license; redistribution and use of this VM image and its components are governed by the terms of those licenses listed below.
The empirischtech image itself (the build scripts, systemd units, helper commands, and configuration files we authored) is © 2026 empirischtech and distributed for use as part of this VM offering.
Operating System
| Component | License | Source |
|---|---|---|
| Ubuntu 24.04 LTS | Mix of GPL, LGPL, BSD, MIT (per-package) | ubuntu.com |
| Linux kernel | GPL-2.0 | kernel.org |
| systemd | LGPL-2.1+ | systemd/systemd |
| OpenSSH | BSD | openssh.com |
| openssl | Apache-2.0 / OpenSSL License | openssl.org |
| unattended-upgrades | GPL-2.0+ | unattended-upgrades |
| fail2ban | GPL-2.0+ | fail2ban |
| logrotate | GPL-2.0 | logrotate |
| dnsutils (BIND tools) | MPL-2.0 | isc.org |
Per-package source code is available via:
apt-get source <package-name>Full per-package copyright and license text is available at:
/usr/share/doc/<package-name>/copyrightNVIDIA Components (PROPRIETARY — see notice)
The NVIDIA driver and CUDA toolkit included in this image are governed by NVIDIA's Software License Agreement
and the CUDA EULA, NOT by the open-source licenses above. Use, redistribution, and modification are subject
to those agreements. Full notices are in /opt/empirischtech/NVIDIA-EULA-NOTICE.txt.
| Component | License | Source |
|---|---|---|
| nvidia-driver-580-open (kernel modules) | MIT/GPL-2.0 dual-license | open-gpu-kernel-modules |
| NVIDIA proprietary userspace libraries | NVIDIA Software License Agreement | NVIDIA license |
| CUDA Toolkit 13.0 | NVIDIA CUDA EULA | CUDA EULA |
| NVIDIA Container Toolkit | Apache-2.0 | nvidia-container-toolkit |
Containerization & Web Server
| Component | License | Source |
|---|---|---|
| Docker CE | Apache-2.0 | moby/moby |
| containerd | Apache-2.0 | containerd |
| Caddy (web server / reverse proxy) | Apache-2.0 | caddyserver/caddy |
Python Runtime & Tooling
| Component | License | Source |
|---|---|---|
| Python 3.12 | PSF License Agreement | python.org |
| uv (Python package manager) | Apache-2.0 OR MIT | astral-sh/uv |
| pip, setuptools, wheel | MIT / PSF | pypi.org |
Inference Stack (/opt/inference/.venv)
| Component | License | Source |
|---|---|---|
| vLLM | Apache-2.0 | vllm-project/vllm |
| PyTorch (cu130 build) | BSD-3-Clause | pytorch |
| huggingface_hub, hf_transfer | Apache-2.0 | huggingface_hub |
| openai (Python SDK) | Apache-2.0 | openai-python |
Pre-trained Model (Downloaded at First Boot)
This model is NOT baked into this image. It is downloaded from Hugging Face on first boot. By using it you accept its license.
Inference Model: empirischtech/DeepSeek-R1-Distill-Qwen-32B-gptq-4bit
- License: Creative Commons Attribution 4.0 International (CC-BY-4.0)
- Source: Hugging Face model card
- License text: CC-BY-4.0 legal code
- Attribution: 4-bit GPTQ quantization by empirischtech of DeepSeek-R1-Distill-Qwen-32B (DeepSeek AI, MIT-licensed), itself a distillation of Qwen2.5-32B (Alibaba Cloud, Apache-2.0).
A copy of this attribution is also placed at /opt/models/MODEL-LICENSES.txt on the deployed VM after first boot completes.
Source Code Availability
For GPL-licensed components on this VM, source code is available from the upstream project websites listed above and via:
apt-get source <package-name>
For LGPL-licensed components, source code is available on request. NVIDIA provides per-component source
for its open-source contributions at oss-requests@nvidia.com.
Where to find full license texts on the VM
| Path | Contents |
|---|---|
/opt/empirischtech/LICENSE.md | VM license bundle (mirrors this notice) |
/opt/empirischtech/NVIDIA-EULA-NOTICE.txt | NVIDIA-specific notices |
/usr/share/doc/<package>/copyright | Per-Debian-package copyrights |
/opt/inference/.venv/lib/python3.12/site-packages/<package>/LICENSE* | Per-Python-package licenses |
/opt/models/MODEL-LICENSES.txt | Pre-trained model attribution (after first boot) |
Disclaimer of Warranty
This VM image and its bundled components are provided "AS IS" without warranty of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose, and noninfringement. In no event shall the authors or copyright holders of any component be liable for any claim, damages, or other liability arising from the use of this image or its software.
Contact
For questions about this VM image or its licensing, contact Empirisch Tech:
+43-1-2396082