Google Coral TPU M.2 PCIe Installation in Frigate LXC on Proxmox | Driver Setup | Frigate Part 2
The Google Coral TPU (Tensor Processing Unit) is a hardware accelerator designed for Machine Learning Tasks used in AI Models. It’s used in frigate due to its processing power capabilities that help in AI detection in the Frigate software. Allocating only your CPU into detection is heavy work and inefficient, usually you would want to use OpenVINO to reduce the overall usage of the CPU, but what if you wanted to allocate that CPU usage to another program? Hence, a solution is of using the Coral TPU.
Resources
Official Installation Instructions:
Commands Used in the video:
Google Coral TPU Datasheet:
Home Assistant Forum (To check if your driver is detected):
Apex driver fix:
Coral LXC Instructions:
gasket-dkms fix:
Additional Information on switching to no subscription PVE in Proxmox:
Playlists
Proxmox Course:
Basic Linux Commands:
Frigate Series:
Timestamps
1:33 Replacing the Wi-Fi chip into the Google Coral TPU
2:08 Installing the Linux Driver (Steps 1 – 4) (PVE)
3:51 Installing the Linux Driver (Steps 1 – 4) (LXC)
5:13 Stuck at step 5 Fix
5:51 Stuck at step 6 Fix
7:32 Switching to no subscription on Proxmox
9:35 Edit Proxmox LXC config file
12:04 gasket-dkms fix
14:46 Fixing USB mounts and apparmor
17:09 Setting Frigate Docker config file
17:47 Setting up Frigate Configuration file to Coral PCIe
19:08 Upgrading from frigate 0.13 to 0.14
Please take note of the requirements you need in order to run this into your server, my laptop was bought back in 2018 and is compatible with MSI-X, it’s best to do research first and know if your home server device supports MSI-X.
Requirements
1. Server with Proxmox installed (x86-64)
2. LXC Container for Frigate
3. Support for MSI-X as defined in the PCI 3.0 specification (My case was a Laptop from 2018 that has PCIe slots)
4. An available slot M.2 Slot (A+E or B+M)
5. Latest Python Installed
[ad_2]
source
