asap-zzhou

Asap-zzhou's group workspace

Group: sft-reproduction

552

496

531

1-3

of 3

Timestamps visible

2025-10-21 21:38:14

{'loss': 0.2341, 'grad_norm': 1.2375109195709229, 'learning_rate': 4.627503662947841e-09, 'epoch': 4.94}

2025-10-21 21:39:57

{'loss': 0.2819, 'grad_norm': 0.7424988746643066, 'learning_rate': 3.331606645023411e-09, 'epoch': 4.95}

2025-10-21 21:41:41

{'loss': 0.2693, 'grad_norm': 1.384178876876831, 'learning_rate': 2.248065796588983e-09, 'epoch': 4.96}

2025-10-21 21:43:24

{'loss': 0.2151, 'grad_norm': 1.7919690608978271, 'learning_rate': 1.376927167646147e-09, 'epoch': 4.97}

2025-10-21 21:45:07

{'loss': 0.213, 'grad_norm': 1.06334388256073, 'learning_rate': 7.182277811962346e-10, 'epoch': 4.98}

2025-10-21 21:46:50

{'loss': 0.2914, 'grad_norm': 1.1211055517196655, 'learning_rate': 2.719956316682426e-10, 'epoch': 4.99}

2025-10-21 21:48:33

{'loss': 0.2077, 'grad_norm': 0.9333163499832153, 'learning_rate': 3.82496837292301e-11, 'epoch': 5.0}

2025-10-21 21:50:56

{'train_runtime': 60254.9897, 'train_samples_per_second': 68.24, 'train_steps_per_second': 0.089, 'train_loss': 0.2889569140949837, 'epoch': 5.0}