| { | |
| "best_metric": 0.023628102615475655, | |
| "best_model_checkpoint": "./mixstral/01-04-24-Weni-WeniGPT-QA-Zephyr-7B-5.0.1-KTO_WeniGPT Experiment using KTO trainer with no collator, Mixstral model and no system prompt.-2_max_steps-262_batch_32_2024-04-01_ppid_2059/checkpoint-100", | |
| "epoch": 0.7582938388625592, | |
| "eval_steps": 50, | |
| "global_step": 100, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.15, | |
| "grad_norm": 1.2250428199768066, | |
| "learning_rate": 0.00019133858267716535, | |
| "loss": 0.3931, | |
| "step": 20, | |
| "train/kl": 1.2274655103683472, | |
| "train/logps/chosen": -192.38479614257812, | |
| "train/logps/rejected": -224.69888305664062, | |
| "train/rewards/chosen": -0.14804823696613312, | |
| "train/rewards/margins": 1.4835635870695114, | |
| "train/rewards/rejected": -1.6316118240356445 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "grad_norm": 2.3648412227630615, | |
| "learning_rate": 0.00017559055118110238, | |
| "loss": 0.1177, | |
| "step": 40, | |
| "train/kl": 0.9526737332344055, | |
| "train/logps/chosen": -162.2818603515625, | |
| "train/logps/rejected": -265.06201171875, | |
| "train/rewards/chosen": 3.2411201000213623, | |
| "train/rewards/margins": 10.128443956375122, | |
| "train/rewards/rejected": -6.88732385635376 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval/kl": 0.0, | |
| "eval/logps/chosen": -149.90289306640625, | |
| "eval/logps/rejected": -403.8857727050781, | |
| "eval/rewards/chosen": 4.500288486480713, | |
| "eval/rewards/margins": 24.8637433052063, | |
| "eval/rewards/rejected": -20.363454818725586, | |
| "eval_loss": 0.04682931676506996, | |
| "eval_runtime": 1026.2439, | |
| "eval_samples_per_second": 0.476, | |
| "eval_steps_per_second": 0.119, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "grad_norm": 1.3626991510391235, | |
| "learning_rate": 0.00015984251968503938, | |
| "loss": 0.0493, | |
| "step": 60, | |
| "train/kl": 0.10373685508966446, | |
| "train/logps/chosen": -142.3507080078125, | |
| "train/logps/rejected": -377.33282470703125, | |
| "train/rewards/chosen": 5.036634922027588, | |
| "train/rewards/margins": 22.215031147003174, | |
| "train/rewards/rejected": -17.178396224975586 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "grad_norm": 0.18376566469669342, | |
| "learning_rate": 0.0001440944881889764, | |
| "loss": 0.0223, | |
| "step": 80, | |
| "train/kl": 0.26756417751312256, | |
| "train/logps/chosen": -134.23001098632812, | |
| "train/logps/rejected": -367.7132263183594, | |
| "train/rewards/chosen": 5.875685691833496, | |
| "train/rewards/margins": 24.18467617034912, | |
| "train/rewards/rejected": -18.308990478515625 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "grad_norm": 0.5995637774467468, | |
| "learning_rate": 0.00012834645669291338, | |
| "loss": 0.0257, | |
| "step": 100, | |
| "train/kl": 0.0, | |
| "train/logps/chosen": -145.93228149414062, | |
| "train/logps/rejected": -409.9689025878906, | |
| "train/rewards/chosen": 6.230437755584717, | |
| "train/rewards/margins": 27.460736751556396, | |
| "train/rewards/rejected": -21.23029899597168 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "eval/kl": 0.2827310264110565, | |
| "eval/logps/chosen": -131.14064025878906, | |
| "eval/logps/rejected": -441.502197265625, | |
| "eval/rewards/chosen": 6.376513957977295, | |
| "eval/rewards/margins": 30.501620769500732, | |
| "eval/rewards/rejected": -24.125106811523438, | |
| "eval_loss": 0.023628102615475655, | |
| "eval_runtime": 1026.1735, | |
| "eval_samples_per_second": 0.476, | |
| "eval_steps_per_second": 0.119, | |
| "step": 100 | |
| } | |
| ], | |
| "logging_steps": 20, | |
| "max_steps": 262, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 2, | |
| "save_steps": 100, | |
| "total_flos": 0.0, | |
| "train_batch_size": 4, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |