Adding a new provider / comparing with IBM Granite

[jotak]

Hi,
Thanks for writing this tool! I'm no expert at all neither in AI nor in environmental impacts, but still very interested in what you're doing.
How doable would it be to add a new provider, IBM / Granite? I'd be very curious to see how it compares with other models. I've found this document from IBM that provides data, including carbon emission estimates, but don't know if that would be sufficient to integrate in ecologits.

[adrienbanse]

Hey @jotak, thanks for your message :)

If I'm not mistaken, IBM Granite is open source and part of the HuggingFace models (see here), so you can already use Ecologits via the huggingface-hub provider (here is a tutorial).

To compare models, you can also directly call the compute_llm_impacts function from https://github.com/genai-impact/ecologits/blob/0cd0df2b05690a73e2afa3f9353ef32c26e8f512/ecologits/impacts/llm.py#L424

For each request, you'll have to provide

• The number of active parameters of the models
• The number of total parameters of the models
• The number of output token (you can fix it to a constant value to compare)
• The request latency (that you can measure with time for example)

Let me know if you have further questions!