Datasets:
Could you clarify how much time is needed for the evaluation process to be completed?
same as the title.
Could you please inform me about the type of machine being used for the evaluation, and how long it is expected to take?
Because the evaluation of my fine-tuned model is taking longer than expected.
I'm wondering how much time it take to evaluate those model
Thanks
Really depends on your models and completion tokens. For reference : Causal/7B took around 30 minutes to finish while llama 7B would took around half a day.
The inference method also has a huge impact which is why we use tgi since they have all the optimization built in (flash-attn, quantization .. ).
Thank you for your detailed explanation and the efforts of your team.
May I inquire about the specific machine specifications used for the run times you mentioned?
We use workstation with 8 A6000
We are still in discussion whether or not to release a smaller subset for faster evaluation. If you have any ideas, feel free to reach out.