File size: 3,692 Bytes
581a708
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
{
  "best_metric": 0.47333332896232605,
  "best_model_checkpoint": "./zephyr/08-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.14-KTO_Hyperparameter search, altering lora params for KTO task.-2_max_steps-145_batch_16_2024-04-08_ppid_9/checkpoint-100",
  "epoch": 0.684931506849315,
  "eval_steps": 50,
  "global_step": 100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.14,
      "grad_norm": 0.0,
      "learning_rate": 0.0001785714285714286,
      "loss": 0.4159,
      "step": 20,
      "train/kl": 0.052512504160404205,
      "train/logps/chosen": -1660.366455078125,
      "train/logps/rejected": -1587.2052408854167,
      "train/rewards/chosen": -135.80235290527344,
      "train/rewards/margins": -6.56842549641928,
      "train/rewards/rejected": -129.23392740885416
    },
    {
      "epoch": 0.27,
      "grad_norm": 0.0,
      "learning_rate": 0.00015000000000000001,
      "loss": 0.45,
      "step": 40,
      "train/kl": 0.0,
      "train/logps/chosen": -4467.318576388889,
      "train/logps/rejected": -4453.0703125,
      "train/rewards/chosen": -420.4342447916667,
      "train/rewards/margins": -3.8474638967803116,
      "train/rewards/rejected": -416.5867808948864
    },
    {
      "epoch": 0.34,
      "eval/kl": 0.0,
      "eval/logps/chosen": -5048.772007042254,
      "eval/logps/rejected": -4424.3093354430375,
      "eval/rewards/chosen": -476.51050836267603,
      "eval/rewards/margins": -60.5168869228026,
      "eval/rewards/rejected": -415.99362143987344,
      "eval_loss": 0.47333332896232605,
      "eval_runtime": 139.068,
      "eval_samples_per_second": 2.157,
      "eval_steps_per_second": 0.539,
      "step": 50
    },
    {
      "epoch": 0.41,
      "grad_norm": 0.0,
      "learning_rate": 0.00012142857142857143,
      "loss": 0.4969,
      "step": 60,
      "train/kl": 0.0,
      "train/logps/chosen": -5034.042845911949,
      "train/logps/rejected": -5042.78998447205,
      "train/rewards/chosen": -474.4616745283019,
      "train/rewards/margins": -0.15693850345718374,
      "train/rewards/rejected": -474.3047360248447
    },
    {
      "epoch": 0.55,
      "grad_norm": 0.0,
      "learning_rate": 9.285714285714286e-05,
      "loss": 0.5281,
      "step": 80,
      "train/kl": 0.0,
      "train/logps/chosen": -5164.465606508876,
      "train/logps/rejected": -4831.312086092716,
      "train/rewards/chosen": -487.13577107988164,
      "train/rewards/margins": -31.98345319908691,
      "train/rewards/rejected": -455.1523178807947
    },
    {
      "epoch": 0.68,
      "grad_norm": 0.0,
      "learning_rate": 6.428571428571429e-05,
      "loss": 0.4813,
      "step": 100,
      "train/kl": 0.0,
      "train/logps/chosen": -5274.45413961039,
      "train/logps/rejected": -4860.5839608433735,
      "train/rewards/chosen": -499.15376420454544,
      "train/rewards/margins": -40.16430637322014,
      "train/rewards/rejected": -458.9894578313253
    },
    {
      "epoch": 0.68,
      "eval/kl": 0.0,
      "eval/logps/chosen": -5079.107394366197,
      "eval/logps/rejected": -4448.600079113924,
      "eval/rewards/chosen": -479.54401408450707,
      "eval/rewards/margins": -61.12139739146909,
      "eval/rewards/rejected": -418.422616693038,
      "eval_loss": 0.47333332896232605,
      "eval_runtime": 139.0173,
      "eval_samples_per_second": 2.158,
      "eval_steps_per_second": 0.54,
      "step": 100
    }
  ],
  "logging_steps": 20,
  "max_steps": 145,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 100,
  "total_flos": 0.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}