Verify the Hailo Accelerator and Runtime

Verify the Hailo device before running an AI Lab container. This separates hardware or driver failures from model and application problems.

1. Check power and PCIe discovery

bash

vcgencmd get_throttled
lspci -nn | grep -i hailo

The first command should normally report throttled=0x0. The second command must identify the installed Hailo processor. Stop here if PCIe does not see the device; reinstalling Python packages cannot fix missing PCIe enumeration.

2. Check the driver and device node

bash

lsmod | grep -E 'hailo1x_pci|hailo_pci'
ls -l /dev/hailo0

Hailo-8 commonly uses the hailo_pci driver name, while Hailo-10H software may use hailo1x_pci. The exact name depends on the supported software stack. The character device /dev/hailo0 must exist before it can be passed to Docker.

3. Check HailoRT and firmware communication

bash

hailortcli --version
sudo hailortcli fw-control identify

The identify command should report the board and device architecture without a driver-operation error. Record the output together with the kernel version:

bash

uname -r
modinfo hailo_pci 2>/dev/null | grep -E 'version|vermagic' || true
modinfo hailo1x_pci 2>/dev/null | grep -E 'version|vermagic' || true

4. Diagnose failures in order

Result	Next action
No Hailo entry in `lspci`	Check power, module seating, M.2/PCIe compatibility, and the product-specific PCIe setup.
PCIe device exists but no `/dev/hailo0`	Inspect the loaded driver and `sudo dmesg -T
Device node exists but identify fails	Check permissions, firmware, and HailoRT/driver compatibility.
Identify succeeds	Continue to the matching Hailo-8 or Hailo-10H deployment tool and model card.

For products shipped with Hailo-8, also consult Seeed's R21xx system configuration guide and the accelerator section in the exact product Wiki.