File size: 5,780 Bytes
3d648bb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
{
  "best_metric": 0.09643030911684036,
  "best_model_checkpoint": "./StableLM-WI-DPO/checkpoint-45",
  "epoch": 0.2,
  "eval_steps": 45,
  "global_step": 45,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.022222222222222223,
      "grad_norm": 4.194765090942383,
      "learning_rate": 1.0869565217391305e-05,
      "logits/chosen": 1.6977390050888062,
      "logits/rejected": 1.6941993236541748,
      "logps/chosen": -87.9476318359375,
      "logps/rejected": -238.43380737304688,
      "loss": 0.6612,
      "rewards/accuracies": 0.4000000059604645,
      "rewards/chosen": 0.0017920683603733778,
      "rewards/margins": 0.06744769960641861,
      "rewards/rejected": -0.06565563380718231,
      "step": 5
    },
    {
      "epoch": 0.044444444444444446,
      "grad_norm": 1.8380119800567627,
      "learning_rate": 2.173913043478261e-05,
      "logits/chosen": 1.9384359121322632,
      "logits/rejected": 1.8992607593536377,
      "logps/chosen": -104.84549713134766,
      "logps/rejected": -238.51876831054688,
      "loss": 0.6276,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": -0.0692732185125351,
      "rewards/margins": 0.14073553681373596,
      "rewards/rejected": -0.21000878512859344,
      "step": 10
    },
    {
      "epoch": 0.06666666666666667,
      "grad_norm": 2.176408290863037,
      "learning_rate": 3.260869565217392e-05,
      "logits/chosen": 1.7943300008773804,
      "logits/rejected": 1.785248041152954,
      "logps/chosen": -135.22303771972656,
      "logps/rejected": -216.9629364013672,
      "loss": 0.5573,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": -0.08426995575428009,
      "rewards/margins": 0.3454127907752991,
      "rewards/rejected": -0.42968273162841797,
      "step": 15
    },
    {
      "epoch": 0.08888888888888889,
      "grad_norm": 1.9848376512527466,
      "learning_rate": 4.347826086956522e-05,
      "logits/chosen": 1.613822340965271,
      "logits/rejected": 1.584276795387268,
      "logps/chosen": -82.66014099121094,
      "logps/rejected": -186.2796173095703,
      "loss": 0.4525,
      "rewards/accuracies": 1.0,
      "rewards/chosen": -0.11625717580318451,
      "rewards/margins": 0.613094687461853,
      "rewards/rejected": -0.7293518781661987,
      "step": 20
    },
    {
      "epoch": 0.1111111111111111,
      "grad_norm": 1.4978930950164795,
      "learning_rate": 4.9997293511641216e-05,
      "logits/chosen": 1.6724803447723389,
      "logits/rejected": 1.6774917840957642,
      "logps/chosen": -121.98885345458984,
      "logps/rejected": -262.0728759765625,
      "loss": 0.217,
      "rewards/accuracies": 1.0,
      "rewards/chosen": -0.4427986741065979,
      "rewards/margins": 1.5746344327926636,
      "rewards/rejected": -2.0174331665039062,
      "step": 25
    },
    {
      "epoch": 0.13333333333333333,
      "grad_norm": 0.20637647807598114,
      "learning_rate": 4.9966852247120764e-05,
      "logits/chosen": 1.5546514987945557,
      "logits/rejected": 1.556323528289795,
      "logps/chosen": -102.07676696777344,
      "logps/rejected": -206.6514434814453,
      "loss": 0.2027,
      "rewards/accuracies": 1.0,
      "rewards/chosen": -0.9394552111625671,
      "rewards/margins": 2.5691986083984375,
      "rewards/rejected": -3.5086536407470703,
      "step": 30
    },
    {
      "epoch": 0.15555555555555556,
      "grad_norm": 7.478163719177246,
      "learning_rate": 4.9902627935540205e-05,
      "logits/chosen": 1.541826844215393,
      "logits/rejected": 1.5505328178405762,
      "logps/chosen": -151.88739013671875,
      "logps/rejected": -329.95953369140625,
      "loss": 0.0533,
      "rewards/accuracies": 1.0,
      "rewards/chosen": -3.0381178855895996,
      "rewards/margins": 4.932946681976318,
      "rewards/rejected": -7.971064567565918,
      "step": 35
    },
    {
      "epoch": 0.17777777777777778,
      "grad_norm": 7.0293235694407485e-06,
      "learning_rate": 4.980470747984265e-05,
      "logits/chosen": 1.455780029296875,
      "logits/rejected": 1.4680273532867432,
      "logps/chosen": -117.02046203613281,
      "logps/rejected": -331.1054992675781,
      "loss": 0.0518,
      "rewards/accuracies": 1.0,
      "rewards/chosen": -2.2030303478240967,
      "rewards/margins": 7.786482334136963,
      "rewards/rejected": -9.989511489868164,
      "step": 40
    },
    {
      "epoch": 0.2,
      "grad_norm": 0.7971689105033875,
      "learning_rate": 4.9673223377762715e-05,
      "logits/chosen": 1.3849093914031982,
      "logits/rejected": 1.408332109451294,
      "logps/chosen": -101.07164001464844,
      "logps/rejected": -294.77874755859375,
      "loss": 0.0546,
      "rewards/accuracies": 1.0,
      "rewards/chosen": -1.74613356590271,
      "rewards/margins": 8.281633377075195,
      "rewards/rejected": -10.027766227722168,
      "step": 45
    },
    {
      "epoch": 0.2,
      "eval_logits/chosen": 1.615716576576233,
      "eval_logits/rejected": 1.645801305770874,
      "eval_logps/chosen": -124.92742156982422,
      "eval_logps/rejected": -353.66717529296875,
      "eval_loss": 0.09643030911684036,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": -2.8166019916534424,
      "eval_rewards/margins": 10.488396644592285,
      "eval_rewards/rejected": -13.305000305175781,
      "eval_runtime": 26.191,
      "eval_samples_per_second": 1.909,
      "eval_steps_per_second": 0.267,
      "step": 45
    }
  ],
  "logging_steps": 5,
  "max_steps": 450,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 45,
  "total_flos": 0.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}