Spaces:
Running
Expanding AI Energy Score Research to other GPUs
Hi AI Energy team, love the wonderful work being done to quantify energy ratings of these various models! I'm currently working on the AI emissions problem from a hardware perspective at Georgia Tech, where we are building software to match efficient compute resources to LLM training jobs.
Since the true emissions of AI usage can change dramatically based on the efficiency of newer GPUs with different architectures, I was wondering if there were any plans to expand training and evaluating energy scores beyond the standard NVIDIA H100 GPUs? This could potentially help developers make informed decisions about both model selection and infrastructure choice simultaneously. Appreciate the feedback!
Hi @PranavViswanath , we will consider other AI hardware accelerators in the future. Is there any specific GPU you have in mind?