I see a lot of requests for the groq team to deploy models that aren’t properly evaluated by the requesters. Some look poorly thought-out.
What if Groq developed a tool to help requesters evaluate a model’s performance and suitability for deployment?
Groq must have an internal evaluation process that checks if a model is suitable. Why not share that process as a public tool?
Or formalize model requests as a tool. (submit the huggingface link, at least - even if the groq team manually reviews models, there should be a checklist that says, hey, this tool has been requested before, and has been rejected for its limitations. Please try suggesting another model.)
1 Like