You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm a researcher using CodeCarbon extensively for measuring energy consumption on various hardware platforms. I'm looking into expanding experiments to include Apple hardware, and I want to understand the implications of the current state of CodeCarbon's support for Apple's own chips.
When I run CodeCarbon on my MacBook with the M4 chip, I get the following warnings at startup:
codecarbon WARNING @ 10:25:00] Error while trying to count physical CPUs: [Errno 2] No such file or directory: 'lscpu'. Defaulting to 1.
[codecarbon INFO @ 10:25:00] [setup] RAM Tracking...
[codecarbon INFO @ 10:25:00] [setup] CPU Tracking...
[codecarbon WARNING @ 10:25:00] We saw that you have a Apple M4 but we don't know it. Please contact us.
[codecarbon INFO @ 10:25:01] Tracking Apple CPU and GPU via PowerMetrics
[codecarbon INFO @ 10:25:01] [setup] GPU Tracking...
[codecarbon INFO @ 10:25:01] No GPU found.
[codecarbon INFO @ 10:25:01] The below tracking methods have been set up:
RAM Tracking Method: RAM power estimation model
CPU Tracking Method: PowerMetrics
GPU Tracking Method: PowerMetrics
[codecarbon INFO @ 10:25:01] >>> Tracker's metadata:
[codecarbon INFO @ 10:25:01] Platform system: macOS-15.3.2-arm64-arm-64bit
[codecarbon INFO @ 10:25:01] Python version: 3.12.12
[codecarbon INFO @ 10:25:01] CodeCarbon version: 3.2.0
[codecarbon INFO @ 10:25:01] Available RAM : 24.000 GB
[codecarbon INFO @ 10:25:01] CPU count: 10 thread(s) in 1 physical CPU(s)
[codecarbon INFO @ 10:25:01] CPU model: Apple M4
[codecarbon INFO @ 10:25:01] GPU count: 1
[codecarbon INFO @ 10:25:01] GPU model: Apple M4
As far as I understand, the message We saw that you have a Apple M4 but we don't know it means that there is currently no default power values for the M4; is that correct?
When looking at the RAM measurements in emissions.csv, CodeCarbon assigns a default value of 6W for the ram_power. In my initial experiments on the MacBook, the ram_energy is an order of magnitude higher than the cpu_energy, meaning that the estimate of the RAM consumption is the main contributor to the overall consumption. Because of this, I'm unsure whether it's currently accurate enough for our purposes. Do you know what the 6W value is based on? Will this be updated if/when the M4 is added to your database?
I'll take the opportunity to thank you for amazing work on CodeCarbon! It has been extremely useful for measuring and understanding energy consumption of various AI models, and has been used in several of our publications. Keep up the good work :)
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
Hi,
I'm a researcher using CodeCarbon extensively for measuring energy consumption on various hardware platforms. I'm looking into expanding experiments to include Apple hardware, and I want to understand the implications of the current state of CodeCarbon's support for Apple's own chips.
When I run CodeCarbon on my MacBook with the M4 chip, I get the following warnings at startup:
As far as I understand, the message
We saw that you have a Apple M4 but we don't know itmeans that there is currently no default power values for the M4; is that correct?When looking at the RAM measurements in
emissions.csv, CodeCarbon assigns a default value of 6W for theram_power. In my initial experiments on the MacBook, theram_energyis an order of magnitude higher than thecpu_energy, meaning that the estimate of the RAM consumption is the main contributor to the overall consumption. Because of this, I'm unsure whether it's currently accurate enough for our purposes. Do you know what the 6W value is based on? Will this be updated if/when the M4 is added to your database?I'll take the opportunity to thank you for amazing work on CodeCarbon! It has been extremely useful for measuring and understanding energy consumption of various AI models, and has been used in several of our publications. Keep up the good work :)
Beta Was this translation helpful? Give feedback.
All reactions