HARDWARE Requirements
AudioMuse-AI requires specific CPU instruction sets due to the pre-compiled Python packages (NumPy, SciPy, ONNX Runtime) used for audio analysis.
Supported Architectures
| Architecture | Instruction Set Required | Status | Examples |
|---|---|---|---|
| x86_64 (Intel/AMD) | AVX2 | Required | Intel 4th gen (Haswell)+, AMD Zen+ |
| ARM64 (aarch64) | NEON | Standard on all ARM64 | Raspberry Pi 4/5, AWS Graviton, Apple Silicon |
NOT Supported
| Architecture | Reason | Examples |
|---|---|---|
| x86_64 without AVX | PyPI wheels require AVX2 instructions | Intel Celeron N3160, Atom pre-2013 |
| 32-bit x86 | Not supported by dependencies | Legacy 32-bit systems |
Why AVX2? The pre-compiled PyPI wheels for NumPy, SciPy, and ONNX Runtime are optimized with AVX2 instructions. Running on CPUs without AVX2 will cause "Illegal instruction" errors.
ARM64 Works Differently: ARM processors use NEON instructions (not AVX). ARM64 PyPI wheels are compiled for NEON, which is standard on all ARM64 CPUs.
Tested HW and Configuration WITH ONNX (from version 0.7.0-beta)
| Issue ID | HW | CONFIGURATION | Supported | Notes |
|---|---|---|---|---|
| - | CPU: Intel: i5 6th gen, i5 8th gen; ARM: VM on cloud (no detail on the specific cpu) | K3S cluster, Docker Image :0.7.0-beta |
✅ Yes | - |
| - | CPU: Intel i5-12450H | K3S, Docker Image :0.7.0-beta |
✅ Yes | - |
| #14 | CPU: Intel 12600K | Docker Compose, Jellyfin, Docker Image :0.7.0-beta |
✅ Yes | user need to set at least 4GB of ram to the container |
Tested HW and Configuration WITH TENSORFLOW (till version 0.6.10-beta)
This table collect all the HW and configuration from the different issue in this repository.
IMPORTANT: If a CPU is not present in the table only means is not tested, but is still possible that works. If you have an issue on a particular CPU please raise a ticket.
| Issue ID | HW | CONFIGURATION | Supported | Notes |
|---|---|---|---|---|
| #14 | CPU: Ryzen 5600G | Docker Compose, Jellyfin | ✅ Yes | - |
| #24 | CPU: Ryzen 5600G | TrueNAS SCALE, Docker Image v0.6.0-beta / v0.6.2-beta | ✅ Yes | - |
| #25 | CPU: i5 6th gen, i5 8th gen, ARM Raspberry PI 5 | K3S cluster, Docker Image :devel |
✅ Yes | - |
| #39 | CPU: Ryzen 5 3600 | Bazzite (Fedora Atomic), Docker Image v0.6.4-beta | ✅ Yes | - |
| #55 | CPU: Intel (32c), GPU: Nvidia RTX 3060 | Docker Swarm/single docker, Navidrome | ✅ Yes | Analysis failed for high-res FLACs. Developer added fallbacks, but the issue was ultimately a problem with Navidrome's media library indexing, fixed by the user running a full scan in Navidrome. |
| #62 | CPU: Intel Xeon W-2125 | Proxmox-VE LXC (Debian), Docker Image v0.6.5-beta | ✅ Yes | - |
| #66 | CPU: E5-2697 | Docker Compose (via Portainer), Navidrome 0.58.0 | ✅ Yes | - |
| #67 | CPU: Intel i7-10850H | Arch Linux, Docker Image latest-nvidia |
✅ Yes | - |
| #69 | CPU: Ryzen 5 PRO 4650G | Ubuntu 24.04.3 LTS, Docker Image v0.6.7-beta | ✅ Yes | - |
| #73 | CPU: Intel core i5 1035G1 | Docker Compose, Jellyfin 10.10.7 | ✅ Yes | Database showed zero tracks after analysis. A bug related to float precision on certain CPUs was fixed by casting to Float32 in the :devel branch. |
| #74 | CPU: Amd Ryzen 3600 | Docker Compose, Navidrome v0.58.0 | ✅ Yes | - |
| #65 | CPU: N100 | Docker Compose, Navidrome | ✅ Yes | - |
| #93 | CPU: AMD Ryzen AI 9 HX 370 w/ Radeon 890M 64bit | Podman with docker-compose (v5.6.1), Jellyfin v10.10.7, AudioMuse-AI v0.6.8-beta | 🚧 In Progress | A bug with CPU-specific behavior was fixed in the :devel branch by adding ENV TF_ENABLE_ONEDNN_OPTS=0. Probably a new parameter will be added in deployment/yaml file. |
| #56 | CPU: Intel Celeron CPU N3160 | Docker Compose, Unraid 7.1.4 | ❌ No | Flask app failed with an Illegal instruction error. This is a hardware limitation: AVX2 CPU support is required (NumPy, SciPy, ONNX Runtime PyPI wheels are compiled with AVX2), which the Celeron N3160 lacks. |