Step... (208001/250000 | Loss: 1.6939743757247925, Acc: 0.6590701937675476): 59%|█████████████ | 148622/250000 [76:01:19<63:54:00, 2.27s/it]

Step... (200001/250000 | Loss: 1.6965291500091553, Acc: 0.6575480699539185): 56%|████████████▎ | 140093/250000 [71:33:45<52:41:15, 1.73s/it]

Model at 191k steps, acc 0.6562. Dataset stats

Step... (190001/250000 | Loss: 1.705819845199585, Acc: 0.6563901901245117): 52%|████████████ | 130628/250000 [66:37:26<59:10:21, 1.78s/it]

Dump 187k steps, acc 0.6574