Repository navigation
CPU on Raspberry Pi 5 not recognized #790
Description
Activity
Hello,
Thanks, we may support it with https://github.com/librerpi/rpi-tools/blob/master/pi5_voltage.py@prmths128 has recently done some test using CodeCarbon on a Rpi.
Can you both have a look to #791 and tell me if the TDP seems OK ?
Reacted by Freja ThoresenHello,
I've run tests with the master branch and PR #791.
From the PR #791 (
codecarbon-3.0.0rc4)[codecarbon INFO @ 13:05:17] Codecarbon is taking the configuration from global file: /home/pi/.codecarbon.config [codecarbon WARNING @ 13:05:17] Error while trying to count physical CPUs: invalid literal for int() with base 10: '-'. Defaulting to 1. [codecarbon INFO @ 13:05:17] [setup] RAM Tracking... [codecarbon INFO @ 13:05:17] [setup] CPU Tracking... [codecarbon WARNING @ 13:05:17] No CPU tracking mode found. Falling back on estimation based on TDP for CPU. Linux OS detected: Please ensure RAPL files exist at /sys/class/powercap/intel-rapl/subsystem to measure CPU [codecarbon INFO @ 13:05:17] CPU Model on constant consumption mode: Cortex-A76 [codecarbon WARNING @ 13:05:17] No CPU tracking mode found. Falling back on CPU constant mode. [codecarbon INFO @ 13:05:17] [setup] GPU Tracking... [codecarbon INFO @ 13:05:17] No GPU found. [codecarbon INFO @ 13:05:17] >>> Tracker's metadata: [codecarbon INFO @ 13:05:17] Platform system: Linux-6.6.62+rpt-rpi-2712-aarch64-with-glibc2.36 [codecarbon INFO @ 13:05:17] Python version: 3.11.2 [codecarbon INFO @ 13:05:17] CodeCarbon version: 3.0.0_rc4 [codecarbon INFO @ 13:05:17] Available RAM : 7.860 GB [codecarbon INFO @ 13:05:17] CPU count: 4 thread(s) in 1 physical CPU(s) [codecarbon INFO @ 13:05:17] CPU model: Cortex-A76 [codecarbon INFO @ 13:05:17] GPU count: None [codecarbon INFO @ 13:05:17] GPU model: None [codecarbon INFO @ 13:05:18] Saving emissions data to file /home/pi/Documents/green-ai-bench/data/raw/cpu_inference_emissions.csv [codecarbon INFO @ 13:05:18] Graceful stopping: collecting and writing information. Please wait a few seconds... [codecarbon INFO @ 13:05:18] Energy consumed for RAM : 0.000000 kWh. RAM Power : 2.9473571777343754 W [codecarbon INFO @ 13:05:18] Delta energy consumed for CPU with constant : 0.000000 kWh, power : 6.0 W [codecarbon INFO @ 13:05:18] Energy consumed for All CPU : 0.000000 kWh [codecarbon INFO @ 13:05:18] 0.000000 kWh of electricity used since the beginning. [codecarbon INFO @ 13:05:18] Done!And from the master branch (
codecarbon-2.8.3):[codecarbon INFO @ 13:17:32] Codecarbon is taking the configuration from global file: /home/pi/.codecarbon.config [codecarbon INFO @ 13:17:32] [setup] RAM Tracking... [codecarbon INFO @ 13:17:32] [setup] CPU Tracking... [codecarbon WARNING @ 13:17:32] No CPU tracking mode found. Falling back on CPU constant mode. Linux OS detected: Please ensure RAPL files exist at \sys\class\powercap\intel-rapl to measure CPU [codecarbon WARNING @ 13:17:32] We saw that you have a Cortex-A76 but we don't know it. Please contact us. [codecarbon INFO @ 13:17:32] CPU Model on constant consumption mode: Cortex-A76 [codecarbon INFO @ 13:17:32] [setup] GPU Tracking... [codecarbon INFO @ 13:17:32] No GPU found. [codecarbon INFO @ 13:17:32] >>> Tracker's metadata: [codecarbon INFO @ 13:17:32] Platform system: Linux-6.6.62+rpt-rpi-2712-aarch64-with-glibc2.36 [codecarbon INFO @ 13:17:32] Python version: 3.11.2 [codecarbon INFO @ 13:17:32] CodeCarbon version: 2.8.3 [codecarbon INFO @ 13:17:32] Available RAM : 7.860 GB [codecarbon INFO @ 13:17:32] CPU count: 4 [codecarbon INFO @ 13:17:32] CPU model: Cortex-A76 [codecarbon INFO @ 13:17:32] GPU count: None [codecarbon INFO @ 13:17:32] GPU model: None [codecarbon INFO @ 13:17:33] Saving emissions data to file /home/pi/Documents/green-ai-bench/data/raw/cpu_inference_emissions.csv [codecarbon INFO @ 13:17:33] Graceful stopping: collecting and writing information. Please wait a few seconds... [codecarbon INFO @ 13:17:33] Energy consumed for RAM : 0.000000 kWh. RAM Power : 2.9473571777343754 W [codecarbon INFO @ 13:17:33] Energy consumed for all CPUs : 0.000001 kWh. Total CPU Power : 42.5 W [codecarbon INFO @ 13:17:33] 0.000001 kWh of electricity used since the beginning. [codecarbon INFO @ 13:17:33] Done!If you managed to get a reading with a Rpi, can you please advice @prmths128 @benoit-cty?
Many thanks!
Hello,
I've pushed a new branchfeat/rpi_powerhttps://github.com/mlco2/codecarbon/tree/feat/rpi_power
to add power metrics for raspberry using their tools ; can you try it ?
It should give proper readings for CPU and RAM.Hello,
I tried out the
feat/rpi_powerbranch. It seems the CPU is now getting recognised, but I'm getting zeros for the energy usage. Did you manage to get a reading on a Raspberry Pi, which version?[codecarbon INFO @ 15:42:28] Codecarbon is taking the configuration from global file: /home/pi/.codecarbon.config [codecarbon INFO @ 15:42:28] [setup] RAM Tracking... [codecarbon INFO @ 15:42:28] [setup] CPU Tracking... [codecarbon INFO @ 15:42:28] Tracking CPU via raspberry utils [codecarbon INFO @ 15:42:28] [setup] GPU Tracking... [codecarbon INFO @ 15:42:28] No GPU found. [codecarbon INFO @ 15:42:28] >>> Tracker's metadata: [codecarbon INFO @ 15:42:28] Platform system: Linux-6.6.62+rpt-rpi-2712-aarch64-with-glibc2.36 [codecarbon INFO @ 15:42:28] Python version: 3.10.0 [codecarbon INFO @ 15:42:28] CodeCarbon version: 2.8.3 [codecarbon INFO @ 15:42:28] Available RAM : 7.860 GB [codecarbon INFO @ 15:42:28] CPU count: 4 [codecarbon INFO @ 15:42:28] CPU model: Cortex-A76 [codecarbon INFO @ 15:42:28] GPU count: None [codecarbon INFO @ 15:42:28] GPU model: None [codecarbon INFO @ 15:42:29] Emissions data (if any) will be saved to file /home/pi/Documents/green-ai-bench/data/raw/cpu_inference_emissions.csv [codecarbon INFO @ 15:42:29] Graceful stopping: collecting and writing information. Please wait a few seconds... [codecarbon INFO @ 15:42:29] Energy consumed for RAM : 0.000000 kWh. RAM Power : 2.9473571777343754 W [codecarbon INFO @ 15:42:29] Energy consumed for all CPUs : 0.000000 kWh. Total CPU Power : 5.953340186965816 W [codecarbon INFO @ 15:42:29] Energy consumed for RAMs : 0.000000 kWh. Total RAM Power : 0.047042817949260005 W [codecarbon INFO @ 15:42:29] 0.000000 kWh of electricity used since the beginning. [codecarbon INFO @ 15:42:29] Done!Cheers,
Hello,
I'm not sure the correct version is called.
With this command:export PYTHONPATH=".:$PYTHONPATH" python codecarbon/cli/main.py monitor --no-api 2 2
I'm getting:
CodeCarbon is going in an infinite loop to monitor this machine. [codecarbon INFO @ 17:40:53] [setup] RAM Tracking... [codecarbon INFO @ 17:40:53] [setup] CPU Tracking... [codecarbon INFO @ 17:40:53] Tracking CPU via raspberry utils [codecarbon INFO @ 17:40:53] [setup] GPU Tracking... [codecarbon INFO @ 17:40:53] No GPU found. [codecarbon INFO @ 17:40:53] >>> Tracker's metadata: [codecarbon INFO @ 17:40:53] Platform system: Linux-6.6.51+rpt-rpi-v8-aarch64-with-glibc2.40 [codecarbon INFO @ 17:40:53] Python version: 3.13.1 [codecarbon INFO @ 17:40:53] CodeCarbon version: 2.8.3 [codecarbon INFO @ 17:40:53] Available RAM : 7.759 GB [codecarbon INFO @ 17:40:53] CPU count: 4 [codecarbon INFO @ 17:40:53] CPU model: Cortex-A76 [codecarbon INFO @ 17:40:53] GPU count: None [codecarbon INFO @ 17:40:53] GPU model: None [codecarbon INFO @ 17:40:56] Emissions data (if any) will be saved to file /mnt/sda1/codecarbon/emissions.csv [codecarbon INFO @ 17:40:58] measuring from hardware=Raspberry (Cortex-A76 > RAM) [codecarbon INFO @ 17:40:58] Energy consumed for RAMs : 0.000000 kWh. Total RAM Power : 0.00648188698122 W [codecarbon INFO @ 17:40:58] measuring from hardware=Raspberry (Cortex-A76 > CPU) [codecarbon INFO @ 17:40:58] Energy consumed for all CPUs : 0.000001 kWh. Total CPU Power : 1.60725786863836 W [codecarbon INFO @ 17:40:58] 0.000001 kWh of electricity used since the beginning. [codecarbon INFO @ 17:41:00] measuring from hardware=Raspberry (Cortex-A76 > RAM) [codecarbon INFO @ 17:41:00] Energy consumed for RAMs : 0.000000 kWh. Total RAM Power : 0.006475451698800001 W [codecarbon INFO @ 17:41:00] measuring from hardware=Raspberry (Cortex-A76 > CPU) [codecarbon INFO @ 17:41:00] Energy consumed for all CPUs : 0.000002 kWh. Total CPU Power : 1.7625656386553101 W [codecarbon INFO @ 17:41:00] 0.000002 kWh of electricity used since the beginning.I've added a new version with some logging to the commit (branch
feat/rpi_power) to check the measure used, it showsmeasuring from hardware...and should show the raspberry hardware for CPU/RAM like above.Hello,
Thanks for the help, I think it works now.
CodeCarbon is going in an infinite loop to monitor this machine. [codecarbon INFO @ 09:20:16] Codecarbon is taking the configuration from global file: /home/pi/.codecarbon.config [codecarbon INFO @ 09:20:16] [setup] RAM Tracking... [codecarbon INFO @ 09:20:16] [setup] CPU Tracking... [codecarbon INFO @ 09:20:16] Tracking CPU via raspberry utils [codecarbon INFO @ 09:20:16] [setup] GPU Tracking... [codecarbon INFO @ 09:20:16] No GPU found. [codecarbon INFO @ 09:20:16] >>> Tracker's metadata: [codecarbon INFO @ 09:20:16] Platform system: Linux-6.6.62+rpt-rpi-2712-aarch64-with-glibc2.36 [codecarbon INFO @ 09:20:16] Python version: 3.11.2 [codecarbon INFO @ 09:20:16] CodeCarbon version: 2.8.3 [codecarbon INFO @ 09:20:16] Available RAM : 7.860 GB [codecarbon INFO @ 09:20:16] CPU count: 4 [codecarbon INFO @ 09:20:16] CPU model: Cortex-A76 [codecarbon INFO @ 09:20:16] GPU count: None [codecarbon INFO @ 09:20:16] GPU model: None [codecarbon INFO @ 09:20:17] Emissions data (if any) will be saved to file /home/pi/Documents/codecarbon/emissions.csv [codecarbon INFO @ 09:20:19] measuring from hardware=Raspberry (Cortex-A76 > RAM) [codecarbon INFO @ 09:20:19] Energy consumed for RAMs : 0.000000 kWh. Total RAM Power : 0.05351741464734 W [codecarbon INFO @ 09:20:19] measuring from hardware=Raspberry (Cortex-A76 > CPU) [codecarbon INFO @ 09:20:19] Energy consumed for all CPUs : 0.000001 kWh. Total CPU Power : 2.38590095336374 W [codecarbon INFO @ 09:20:19] 0.000001 kWh of electricity used since the beginning. [codecarbon INFO @ 09:20:21] measuring from hardware=Raspberry (Cortex-A76 > RAM) [codecarbon INFO @ 09:20:21] Energy consumed for RAMs : 0.000000 kWh. Total RAM Power : 0.13027155368436 W [codecarbon INFO @ 09:20:21] measuring from hardware=Raspberry (Cortex-A76 > CPU) [codecarbon INFO @ 09:20:21] Energy consumed for all CPUs : 0.000002 kWh. Total CPU Power : 2.05160794672903 W [codecarbon INFO @ 09:20:21] 0.000003 kWh of electricity used since the beginning. [codecarbon INFO @ 09:20:21] 0.000095 g.CO2eq/s mean an estimation of 2.9968769284364143 kg.CO2eq/year [codecarbon INFO @ 09:20:23] measuring from hardware=Raspberry (Cortex-A76 > RAM) [codecarbon INFO @ 09:20:23] Energy consumed for RAMs : 0.000000 kWh. Total RAM Power : 0.059869230269670004 W [codecarbon INFO @ 09:20:23] measuring from hardware=Raspberry (Cortex-A76 > CPU)
Cheers!
Ok, cool!
You can keep using the branch for the moment, i'll close the ticket when it's finished and merged.Reacted by Amine Saboni
Dear CodeCarbon team,
I am running codecarbon on a Raspberry Pi 5, and the CPU (and Hailo AI acceleration device) are not recognized.
Furthermore, this folder doesn't exist on the system
/sys/class/powercap.