When running a pipeline in AI Center, you have an option to enable GPU. This speeds up the rate at which the pipeline is ran, but it is also more expensive per-hour.
Say you are training a machine learning extractor with a lot of fields, and a lot of training documents. Would choosing GPU ever save you AI units due to cutting processing time? Or will the increased per-hour cost always lead to it ultimately being more expensive?
Thanks.
Hi, @edevries
If training a ML extractor with many fields and lots of documents choosing GPU can save AI units because it trains way faster and cutting total time..even though GPUs cost more per hour the big time saving usually make the total AI units lower. But for smaller training CPUs cheaper hourly cost might still be better overall
It totally depends on the time with gpu and without..
generally if the dataset is as large as 100k or so it would make difference for smaller sets I dont think there would be much
cheers
This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.