Environnemental informations for Llama 3.3 70B

[davidson-canada-devops]

In the Llama 3.3 Model Card in Github and Hugging Face, the text presentign GPU hours and carbon emissions does not match the table bellow. It seems like the text from model card of llama 3.1 have been reused.

"Training Energy Use Training utilized a cumulative of 39.3M GPU hours of computation on H100-80GB (TDP of 700W) type hardware, per the table below. Training time is the total GPU time required for training each model and power consumption is the peak power capacity per GPU device used, adjusted for power usage efficiency.

Training Greenhouse Gas Emissions Estimated total location-based greenhouse gas emissions were 11,390 tons CO2eq for training."

39.3M GPU hours => 7M GPU hours
11,390 tons CO2eq for training => 2,040 tons CO2eq for training

Also, would you consider to add the following informations in the model card :

• The location of the datacenter used to train the model (or the carbon intensity of the energy mix in this aera)
• Le number of H100 GPU used to train the model

Would be higly appreciated for AI environnemental studies.

Source :

• https://github.com/meta-llama/llama-models/blob/main/models/llama3_3/MODEL_CARD.md#hardware-and-software
• https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct