When will the training stop?

Elsaam2y / DINet_optimized

An optimized pipeline for DINet reducing inference latency for up to 60% 🚀. Kudos for the authors of the original repo for this amazing work.

93 stars 15 forks source link

Closed einsqing closed 4 months ago

einsqing commented 5 months ago

Loss_DI: 0.2508 Loss_GI: 0.2469 Loss_DV: 0.2495 Loss_GV: 0.2887 Loss_perception: 1.3895 Loss_sync: 0.0031 lr_g = 0.0000353 elapsed_time: 2.4532

When will the training stop? Is the learning rate below 500, or is it based on other indicators?