how long will the training process takes?

[guchengxi1994]

i have trained the model on my own dataset for 27000 rounds, but the d_oss is still between 1.5 to 4.0 and the g_loss is between 9.0 to 40, how many iterations will it takes ?

[knicholes]

I calculated how many it would take for me to train 1M epocs, and it ended up to be ~7 days with my RTX 3090. If your model isn't converging, maybe you should add more data or look through your dataset to make sure there isn't anything in there that is wrong and screwing up the model's attempt to converge.

[jdunkelheit]

For me who's training on the CPU, 1 million iterations would take 254 days, but the model is already showing some results after 12,000 iterations.
As far as I know, the time it takes to converge depends on how much data you have and the complexity of the problem.