general
Is it expected for the second response to be different when running a model with parallelism and sending two completion requests at the same time?
Er
Erfan Safari
Asked on Nov 20, 2023
Unexpected randomness could be introduced at various layers, which could potentially cause the second response to be different. It would be helpful to have the environment information and a script to reproduce the behavior in order to investigate further.
Dec 01, 2023Edited by