join20252
Results
1
issues of
join20252
内存爆了
16
Starting training..., iters: 1000 Iter 1: Val loss 9.100, Val took 5657.756s Iter 10: Train loss 9.754, Learning Rate 1.000e-05, It/sec 0.005, Tokens/sec 6.189, Trained Tokens 12868, Peak mem 81.864...