Support for openai/transcription

[flebarbe]

Description

Hi,

I'm looking at implementing openai's transcription API in EcoLogits.

It seems doable: We do have the tokens in the returned data (see segments in the transcription object), so we have a metric to base our calculation on.

Is anyone already working on this ? Do you forsee any obstacle ?

Thanks !

[samuelrince]

Hello @flebarbe,

As we discussed IRL, we first need to have energy benchmarks on speech-to-text models before implementing it!