'Why does the loss/alpha of my SAC algorithm not converge after changing the graphics card?
My graphics card changed from 970 to 3090 TI, the code has not changed, but, from the tensorboard, the loss/alpha has not converged. The red line is the result of training with 970 before, and the sky blue line is the result of training with 3090 Ti. , what could be the cause?
I have tried modifying the seed, retraining from 0,and reinstall display card driver, cuda, cudnn, and so on, and the results are the same.
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|