bash
展开代码
>>  [2026-01-13 09:37:10] iteration      961/    1637 | consumed samples:       123008 | elapsed time per iteration (ms): 52266.9 | memory(GiB): 43.51 | elapsed time: 13h 29m 14s | remaining time: 9h 29m 14s | learning rate: 4.583093E-06 | global batch size:   128 | lm loss: 7.373063E-05 | load_balancing_loss: 0.000000E+00 | loss scale: 1.0 | grad norm: 0.034 | number of skipped iterations:   0 | number of nan iterations:   0 |