| {"loss": 1.51931369, "token_acc": 0.65133395, "grad_norm": 0.5639717, "learning_rate": 8.11e-06, "memory(GiB)": 11.56, "train_speed(iter/s)": 0.103188, "epoch": 0.00813008, "global_step/max_steps": "1/369", "percentage": "0.27%", "elapsed_time": "3s", "remaining_time": "24m 14s"} | |
| {"loss": 1.25695169, "token_acc": 0.70556827, "grad_norm": 0.51285839, "learning_rate": 1.622e-05, "memory(GiB)": 11.56, "train_speed(iter/s)": 0.150757, "epoch": 0.01626016, "global_step/max_steps": "2/369", "percentage": "0.54%", "elapsed_time": "7s", "remaining_time": "23m 1s"} | |
| {"loss": 1.51376784, "token_acc": 0.65096953, "grad_norm": 0.50419378, "learning_rate": 2.432e-05, "memory(GiB)": 11.56, "train_speed(iter/s)": 0.178341, "epoch": 0.02439024, "global_step/max_steps": "3/369", "percentage": "0.81%", "elapsed_time": "11s", "remaining_time": "22m 31s"} | |
| {"loss": 1.44488692, "token_acc": 0.65977444, "grad_norm": 0.32598454, "learning_rate": 3.243e-05, "memory(GiB)": 11.56, "train_speed(iter/s)": 0.196312, "epoch": 0.03252033, "global_step/max_steps": "4/369", "percentage": "1.08%", "elapsed_time": "14s", "remaining_time": "22m 15s"} | |
| {"loss": 1.1349889, "token_acc": 0.72309028, "grad_norm": 0.54342312, "learning_rate": 4.054e-05, "memory(GiB)": 11.56, "train_speed(iter/s)": 0.208955, "epoch": 0.04065041, "global_step/max_steps": "5/369", "percentage": "1.36%", "elapsed_time": "18s", "remaining_time": "22m 4s"} | |
| {"loss": 1.64801562, "token_acc": 0.6032849, "grad_norm": 0.45636648, "learning_rate": 4.865e-05, "memory(GiB)": 11.56, "train_speed(iter/s)": 0.216078, "epoch": 0.04878049, "global_step/max_steps": "6/369", "percentage": "1.63%", "elapsed_time": "22s", "remaining_time": "22m 12s"} | |