Copy to checkpoint/trainer_state.json
Browse files- checkpoint/trainer_state.json +664 -0
checkpoint/trainer_state.json
ADDED
|
@@ -0,0 +1,664 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 0.1568627450980392,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 30,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"accuracy_delta": -0.03125,
|
| 14 |
+
"baseline_accuracy": 0.5625,
|
| 15 |
+
"completion_length": 1660.1953125,
|
| 16 |
+
"degradation_rate": 0.1875,
|
| 17 |
+
"epoch": 0.00522875816993464,
|
| 18 |
+
"grad_norm": 1.8285036167620026,
|
| 19 |
+
"improvement_rate": 0.15625,
|
| 20 |
+
"kl": 0.0,
|
| 21 |
+
"learning_rate": 2e-06,
|
| 22 |
+
"loss": 0.0,
|
| 23 |
+
"reward": 0.24004681408405304,
|
| 24 |
+
"reward_std": 0.25635848194360733,
|
| 25 |
+
"rewards/AdaptiveTeachingReward": 0.24004681408405304,
|
| 26 |
+
"step": 1,
|
| 27 |
+
"student_accuracy": 0.53125,
|
| 28 |
+
"student_approach_length": 500.0,
|
| 29 |
+
"teaching_length_mean": 1349.5,
|
| 30 |
+
"teaching_length_std": 1533.2601122433693,
|
| 31 |
+
"token_efficiency": 0.01778511088825826
|
| 32 |
+
},
|
| 33 |
+
{
|
| 34 |
+
"accuracy_delta": -0.21875,
|
| 35 |
+
"baseline_accuracy": 1.0,
|
| 36 |
+
"completion_length": 2174.171875,
|
| 37 |
+
"degradation_rate": 0.21875,
|
| 38 |
+
"epoch": 0.01045751633986928,
|
| 39 |
+
"grad_norm": 1.2225533588422857,
|
| 40 |
+
"improvement_rate": 0.0,
|
| 41 |
+
"kl": 0.002572178840637207,
|
| 42 |
+
"learning_rate": 2e-06,
|
| 43 |
+
"loss": 0.0001,
|
| 44 |
+
"reward": 0.2656950503587723,
|
| 45 |
+
"reward_std": 0.17467603832483292,
|
| 46 |
+
"rewards/AdaptiveTeachingReward": 0.2656950503587723,
|
| 47 |
+
"step": 2,
|
| 48 |
+
"student_accuracy": 0.78125,
|
| 49 |
+
"student_approach_length": 489.875,
|
| 50 |
+
"teaching_length_mean": 1645.90625,
|
| 51 |
+
"teaching_length_std": 1777.2254866782437,
|
| 52 |
+
"token_efficiency": 0.016061369340544678
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"accuracy_delta": -0.0625,
|
| 56 |
+
"baseline_accuracy": 0.0625,
|
| 57 |
+
"completion_length": 2889.65625,
|
| 58 |
+
"degradation_rate": 0.0625,
|
| 59 |
+
"epoch": 0.01568627450980392,
|
| 60 |
+
"grad_norm": 0.4912196165940296,
|
| 61 |
+
"improvement_rate": 0.0,
|
| 62 |
+
"kl": 0.0021309852600097656,
|
| 63 |
+
"learning_rate": 2e-06,
|
| 64 |
+
"loss": 0.0001,
|
| 65 |
+
"reward": 0.1566198617219925,
|
| 66 |
+
"reward_std": 0.17452973127365112,
|
| 67 |
+
"rewards/AdaptiveTeachingReward": 0.1566198617219925,
|
| 68 |
+
"step": 3,
|
| 69 |
+
"student_accuracy": 0.0,
|
| 70 |
+
"student_approach_length": 500.0,
|
| 71 |
+
"teaching_length_mean": 3435.71875,
|
| 72 |
+
"teaching_length_std": 874.6175996443687,
|
| 73 |
+
"token_efficiency": 0.005029539554335019
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"accuracy_delta": 0.0,
|
| 77 |
+
"baseline_accuracy": 0.0,
|
| 78 |
+
"completion_length": 1765.4765625,
|
| 79 |
+
"degradation_rate": 0.0,
|
| 80 |
+
"epoch": 0.02091503267973856,
|
| 81 |
+
"grad_norm": 0.014919439029800962,
|
| 82 |
+
"improvement_rate": 0.0,
|
| 83 |
+
"kl": 0.0024797916412353516,
|
| 84 |
+
"learning_rate": 2e-06,
|
| 85 |
+
"loss": 0.0001,
|
| 86 |
+
"reward": 0.0,
|
| 87 |
+
"reward_std": 0.0,
|
| 88 |
+
"rewards/AdaptiveTeachingReward": 0.0,
|
| 89 |
+
"step": 4,
|
| 90 |
+
"student_accuracy": 0.0,
|
| 91 |
+
"student_approach_length": 500.0,
|
| 92 |
+
"teaching_length_mean": 1644.53125,
|
| 93 |
+
"teaching_length_std": 1544.9233390329066,
|
| 94 |
+
"token_efficiency": 0.0
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"accuracy_delta": 0.1875,
|
| 98 |
+
"baseline_accuracy": 0.375,
|
| 99 |
+
"completion_length": 2330.296875,
|
| 100 |
+
"degradation_rate": 0.09375,
|
| 101 |
+
"epoch": 0.026143790849673203,
|
| 102 |
+
"grad_norm": 1.6480657403271404,
|
| 103 |
+
"improvement_rate": 0.28125,
|
| 104 |
+
"kl": 0.002542257308959961,
|
| 105 |
+
"learning_rate": 2e-06,
|
| 106 |
+
"loss": 0.0001,
|
| 107 |
+
"reward": 0.293629452586174,
|
| 108 |
+
"reward_std": 0.31043318659067154,
|
| 109 |
+
"rewards/AdaptiveTeachingReward": 0.293629452586174,
|
| 110 |
+
"step": 5,
|
| 111 |
+
"student_accuracy": 0.5625,
|
| 112 |
+
"student_approach_length": 500.0,
|
| 113 |
+
"teaching_length_mean": 2543.6875,
|
| 114 |
+
"teaching_length_std": 1157.76463240177,
|
| 115 |
+
"token_efficiency": 0.015775285003662067
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"accuracy_delta": -0.125,
|
| 119 |
+
"baseline_accuracy": 0.375,
|
| 120 |
+
"completion_length": 2799.2265625,
|
| 121 |
+
"degradation_rate": 0.21875,
|
| 122 |
+
"epoch": 0.03137254901960784,
|
| 123 |
+
"grad_norm": 1.3626772466174568,
|
| 124 |
+
"improvement_rate": 0.09375,
|
| 125 |
+
"kl": 0.0024237632751464844,
|
| 126 |
+
"learning_rate": 2e-06,
|
| 127 |
+
"loss": 0.0001,
|
| 128 |
+
"reward": 0.297846183180809,
|
| 129 |
+
"reward_std": 0.1792445182800293,
|
| 130 |
+
"rewards/AdaptiveTeachingReward": 0.297846183180809,
|
| 131 |
+
"step": 6,
|
| 132 |
+
"student_accuracy": 0.25,
|
| 133 |
+
"student_approach_length": 500.0,
|
| 134 |
+
"teaching_length_mean": 3300.46875,
|
| 135 |
+
"teaching_length_std": 1322.463354762313,
|
| 136 |
+
"token_efficiency": 0.008984860526379333
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"accuracy_delta": 0.15625,
|
| 140 |
+
"baseline_accuracy": 0.09375,
|
| 141 |
+
"completion_length": 2839.59375,
|
| 142 |
+
"degradation_rate": 0.0,
|
| 143 |
+
"epoch": 0.036601307189542485,
|
| 144 |
+
"grad_norm": 1.0148015671601693,
|
| 145 |
+
"improvement_rate": 0.15625,
|
| 146 |
+
"kl": 0.0022563934326171875,
|
| 147 |
+
"learning_rate": 2e-06,
|
| 148 |
+
"loss": 0.0001,
|
| 149 |
+
"reward": 0.22244123369455338,
|
| 150 |
+
"reward_std": 0.32708095014095306,
|
| 151 |
+
"rewards/AdaptiveTeachingReward": 0.22244123369455338,
|
| 152 |
+
"step": 7,
|
| 153 |
+
"student_accuracy": 0.25,
|
| 154 |
+
"student_approach_length": 500.0,
|
| 155 |
+
"teaching_length_mean": 2663.28125,
|
| 156 |
+
"teaching_length_std": 1263.471554191286,
|
| 157 |
+
"token_efficiency": 0.009322577760442718
|
| 158 |
+
},
|
| 159 |
+
{
|
| 160 |
+
"accuracy_delta": 0.03125,
|
| 161 |
+
"baseline_accuracy": 0.34375,
|
| 162 |
+
"completion_length": 2997.0234375,
|
| 163 |
+
"degradation_rate": 0.0625,
|
| 164 |
+
"epoch": 0.04183006535947712,
|
| 165 |
+
"grad_norm": 1.0933058629769363,
|
| 166 |
+
"improvement_rate": 0.09375,
|
| 167 |
+
"kl": 0.00222015380859375,
|
| 168 |
+
"learning_rate": 2e-06,
|
| 169 |
+
"loss": 0.0001,
|
| 170 |
+
"reward": 0.4145798534154892,
|
| 171 |
+
"reward_std": 0.348370686173439,
|
| 172 |
+
"rewards/AdaptiveTeachingReward": 0.4145798534154892,
|
| 173 |
+
"step": 8,
|
| 174 |
+
"student_accuracy": 0.375,
|
| 175 |
+
"student_approach_length": 500.0,
|
| 176 |
+
"teaching_length_mean": 2859.9375,
|
| 177 |
+
"teaching_length_std": 1548.3925018635746,
|
| 178 |
+
"token_efficiency": 0.014895284304358753
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"accuracy_delta": 0.0,
|
| 182 |
+
"baseline_accuracy": 0.0,
|
| 183 |
+
"completion_length": 2069.3515625,
|
| 184 |
+
"degradation_rate": 0.0,
|
| 185 |
+
"epoch": 0.047058823529411764,
|
| 186 |
+
"grad_norm": 0.6907373861546955,
|
| 187 |
+
"improvement_rate": 0.0,
|
| 188 |
+
"kl": 0.0022356510162353516,
|
| 189 |
+
"learning_rate": 2e-06,
|
| 190 |
+
"loss": 0.0001,
|
| 191 |
+
"reward": 0.13787749409675598,
|
| 192 |
+
"reward_std": 0.08163860440254211,
|
| 193 |
+
"rewards/AdaptiveTeachingReward": 0.13787749409675598,
|
| 194 |
+
"step": 9,
|
| 195 |
+
"student_accuracy": 0.0,
|
| 196 |
+
"student_approach_length": 500.0,
|
| 197 |
+
"teaching_length_mean": 1586.78125,
|
| 198 |
+
"teaching_length_std": 1393.5538468081056,
|
| 199 |
+
"token_efficiency": 0.009196431155361413
|
| 200 |
+
},
|
| 201 |
+
{
|
| 202 |
+
"accuracy_delta": 0.0,
|
| 203 |
+
"baseline_accuracy": 0.0,
|
| 204 |
+
"completion_length": 2533.3203125,
|
| 205 |
+
"degradation_rate": 0.0,
|
| 206 |
+
"epoch": 0.05228758169934641,
|
| 207 |
+
"grad_norm": 0.9569337200656036,
|
| 208 |
+
"improvement_rate": 0.0,
|
| 209 |
+
"kl": 0.002455472946166992,
|
| 210 |
+
"learning_rate": 2e-06,
|
| 211 |
+
"loss": 0.0001,
|
| 212 |
+
"reward": 0.2650887817144394,
|
| 213 |
+
"reward_std": 0.1834174394607544,
|
| 214 |
+
"rewards/AdaptiveTeachingReward": 0.2650887817144394,
|
| 215 |
+
"step": 10,
|
| 216 |
+
"student_accuracy": 0.0,
|
| 217 |
+
"student_approach_length": 500.0,
|
| 218 |
+
"teaching_length_mean": 2718.53125,
|
| 219 |
+
"teaching_length_std": 1545.6357044636015,
|
| 220 |
+
"token_efficiency": 0.00983657689669804
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"accuracy_delta": 0.0,
|
| 224 |
+
"baseline_accuracy": 0.0,
|
| 225 |
+
"completion_length": 2366.4609375,
|
| 226 |
+
"degradation_rate": 0.0,
|
| 227 |
+
"epoch": 0.05751633986928104,
|
| 228 |
+
"grad_norm": 0.1992844149700769,
|
| 229 |
+
"improvement_rate": 0.0,
|
| 230 |
+
"kl": 0.0023772716522216797,
|
| 231 |
+
"learning_rate": 2e-06,
|
| 232 |
+
"loss": 0.0001,
|
| 233 |
+
"reward": 0.00799931213259697,
|
| 234 |
+
"reward_std": 0.045250944793224335,
|
| 235 |
+
"rewards/AdaptiveTeachingReward": 0.00799931213259697,
|
| 236 |
+
"step": 11,
|
| 237 |
+
"student_accuracy": 0.0,
|
| 238 |
+
"student_approach_length": 500.0,
|
| 239 |
+
"teaching_length_mean": 2560.28125,
|
| 240 |
+
"teaching_length_std": 1264.0513369895032,
|
| 241 |
+
"token_efficiency": 0.000244351732796639
|
| 242 |
+
},
|
| 243 |
+
{
|
| 244 |
+
"accuracy_delta": 0.0,
|
| 245 |
+
"baseline_accuracy": 0.0,
|
| 246 |
+
"completion_length": 2429.25,
|
| 247 |
+
"degradation_rate": 0.0,
|
| 248 |
+
"epoch": 0.06274509803921569,
|
| 249 |
+
"grad_norm": 0.02438642631727396,
|
| 250 |
+
"improvement_rate": 0.0,
|
| 251 |
+
"kl": 0.0028073787689208984,
|
| 252 |
+
"learning_rate": 2e-06,
|
| 253 |
+
"loss": 0.0001,
|
| 254 |
+
"reward": 0.0,
|
| 255 |
+
"reward_std": 0.0,
|
| 256 |
+
"rewards/AdaptiveTeachingReward": 0.0,
|
| 257 |
+
"step": 12,
|
| 258 |
+
"student_accuracy": 0.0,
|
| 259 |
+
"student_approach_length": 500.0,
|
| 260 |
+
"teaching_length_mean": 2684.0,
|
| 261 |
+
"teaching_length_std": 1537.9383137400418,
|
| 262 |
+
"token_efficiency": 0.0
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"accuracy_delta": 0.03125,
|
| 266 |
+
"baseline_accuracy": 0.03125,
|
| 267 |
+
"completion_length": 2985.0234375,
|
| 268 |
+
"degradation_rate": 0.03125,
|
| 269 |
+
"epoch": 0.06797385620915032,
|
| 270 |
+
"grad_norm": 1.9377263264415499,
|
| 271 |
+
"improvement_rate": 0.0625,
|
| 272 |
+
"kl": 0.002295255661010742,
|
| 273 |
+
"learning_rate": 2e-06,
|
| 274 |
+
"loss": 0.0001,
|
| 275 |
+
"reward": 0.3229658156633377,
|
| 276 |
+
"reward_std": 0.2836003005504608,
|
| 277 |
+
"rewards/AdaptiveTeachingReward": 0.3229658156633377,
|
| 278 |
+
"step": 13,
|
| 279 |
+
"student_accuracy": 0.0625,
|
| 280 |
+
"student_approach_length": 500.0,
|
| 281 |
+
"teaching_length_mean": 2764.78125,
|
| 282 |
+
"teaching_length_std": 1359.6938149421062,
|
| 283 |
+
"token_efficiency": 0.011864912871618307
|
| 284 |
+
},
|
| 285 |
+
{
|
| 286 |
+
"accuracy_delta": 0.0,
|
| 287 |
+
"baseline_accuracy": 0.40625,
|
| 288 |
+
"completion_length": 2860.7890625,
|
| 289 |
+
"degradation_rate": 0.03125,
|
| 290 |
+
"epoch": 0.07320261437908497,
|
| 291 |
+
"grad_norm": 0.7397348398480618,
|
| 292 |
+
"improvement_rate": 0.03125,
|
| 293 |
+
"kl": 0.0022339820861816406,
|
| 294 |
+
"learning_rate": 2e-06,
|
| 295 |
+
"loss": 0.0001,
|
| 296 |
+
"reward": 0.3183840811252594,
|
| 297 |
+
"reward_std": 0.2730633243918419,
|
| 298 |
+
"rewards/AdaptiveTeachingReward": 0.3183840811252594,
|
| 299 |
+
"step": 14,
|
| 300 |
+
"student_accuracy": 0.40625,
|
| 301 |
+
"student_approach_length": 500.0,
|
| 302 |
+
"teaching_length_mean": 3300.75,
|
| 303 |
+
"teaching_length_std": 1258.5522551939885,
|
| 304 |
+
"token_efficiency": 0.009677591351820222
|
| 305 |
+
},
|
| 306 |
+
{
|
| 307 |
+
"accuracy_delta": -0.25,
|
| 308 |
+
"baseline_accuracy": 0.59375,
|
| 309 |
+
"completion_length": 2667.375,
|
| 310 |
+
"degradation_rate": 0.375,
|
| 311 |
+
"epoch": 0.0784313725490196,
|
| 312 |
+
"grad_norm": 0.591091766464445,
|
| 313 |
+
"improvement_rate": 0.125,
|
| 314 |
+
"kl": 0.0020873546600341797,
|
| 315 |
+
"learning_rate": 2e-06,
|
| 316 |
+
"loss": 0.0001,
|
| 317 |
+
"reward": 0.14508739858865738,
|
| 318 |
+
"reward_std": 0.1915995106101036,
|
| 319 |
+
"rewards/AdaptiveTeachingReward": 0.14508739858865738,
|
| 320 |
+
"step": 15,
|
| 321 |
+
"student_accuracy": 0.34375,
|
| 322 |
+
"student_approach_length": 500.0,
|
| 323 |
+
"teaching_length_mean": 2580.1875,
|
| 324 |
+
"teaching_length_std": 1411.9177183341415,
|
| 325 |
+
"token_efficiency": 0.006345218750950054
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"accuracy_delta": -0.125,
|
| 329 |
+
"baseline_accuracy": 0.78125,
|
| 330 |
+
"completion_length": 2397.8046875,
|
| 331 |
+
"degradation_rate": 0.28125,
|
| 332 |
+
"epoch": 0.08366013071895424,
|
| 333 |
+
"grad_norm": 1.0467841728187868,
|
| 334 |
+
"improvement_rate": 0.15625,
|
| 335 |
+
"kl": 0.0020101070404052734,
|
| 336 |
+
"learning_rate": 2e-06,
|
| 337 |
+
"loss": 0.0001,
|
| 338 |
+
"reward": 0.29722827672958374,
|
| 339 |
+
"reward_std": 0.29581306129693985,
|
| 340 |
+
"rewards/AdaptiveTeachingReward": 0.29722827672958374,
|
| 341 |
+
"step": 16,
|
| 342 |
+
"student_accuracy": 0.65625,
|
| 343 |
+
"student_approach_length": 500.0,
|
| 344 |
+
"teaching_length_mean": 1867.65625,
|
| 345 |
+
"teaching_length_std": 1645.2406135166107,
|
| 346 |
+
"token_efficiency": 0.016315762169946447
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"accuracy_delta": 0.0,
|
| 350 |
+
"baseline_accuracy": 0.4375,
|
| 351 |
+
"completion_length": 2681.4296875,
|
| 352 |
+
"degradation_rate": 0.0625,
|
| 353 |
+
"epoch": 0.08888888888888889,
|
| 354 |
+
"grad_norm": 1.0178368263353954,
|
| 355 |
+
"improvement_rate": 0.0625,
|
| 356 |
+
"kl": 0.0022170543670654297,
|
| 357 |
+
"learning_rate": 2e-06,
|
| 358 |
+
"loss": 0.0001,
|
| 359 |
+
"reward": 0.27442795038223267,
|
| 360 |
+
"reward_std": 0.24267160892486572,
|
| 361 |
+
"rewards/AdaptiveTeachingReward": 0.27442795038223267,
|
| 362 |
+
"step": 17,
|
| 363 |
+
"student_accuracy": 0.4375,
|
| 364 |
+
"student_approach_length": 500.0,
|
| 365 |
+
"teaching_length_mean": 1946.4375,
|
| 366 |
+
"teaching_length_std": 1595.4173133278073,
|
| 367 |
+
"token_efficiency": 0.01725690617086827
|
| 368 |
+
},
|
| 369 |
+
{
|
| 370 |
+
"accuracy_delta": 0.0,
|
| 371 |
+
"baseline_accuracy": 0.0,
|
| 372 |
+
"completion_length": 2260.1484375,
|
| 373 |
+
"degradation_rate": 0.0,
|
| 374 |
+
"epoch": 0.09411764705882353,
|
| 375 |
+
"grad_norm": 0.39094525161531923,
|
| 376 |
+
"improvement_rate": 0.0,
|
| 377 |
+
"kl": 0.0026335716247558594,
|
| 378 |
+
"learning_rate": 2e-06,
|
| 379 |
+
"loss": 0.0001,
|
| 380 |
+
"reward": 0.15987491607666016,
|
| 381 |
+
"reward_std": 0.12795361876487732,
|
| 382 |
+
"rewards/AdaptiveTeachingReward": 0.15987491607666016,
|
| 383 |
+
"step": 18,
|
| 384 |
+
"student_accuracy": 0.0,
|
| 385 |
+
"student_approach_length": 500.0,
|
| 386 |
+
"teaching_length_mean": 2452.8125,
|
| 387 |
+
"teaching_length_std": 1687.6677585883976,
|
| 388 |
+
"token_efficiency": 0.00733118954839666
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"accuracy_delta": 0.09375,
|
| 392 |
+
"baseline_accuracy": 0.34375,
|
| 393 |
+
"completion_length": 2509.359375,
|
| 394 |
+
"degradation_rate": 0.0,
|
| 395 |
+
"epoch": 0.09934640522875816,
|
| 396 |
+
"grad_norm": 1.112027848483532,
|
| 397 |
+
"improvement_rate": 0.09375,
|
| 398 |
+
"kl": 0.0021845102310180664,
|
| 399 |
+
"learning_rate": 2e-06,
|
| 400 |
+
"loss": 0.0001,
|
| 401 |
+
"reward": 0.32069824635982513,
|
| 402 |
+
"reward_std": 0.1779022440314293,
|
| 403 |
+
"rewards/AdaptiveTeachingReward": 0.32069824635982513,
|
| 404 |
+
"step": 19,
|
| 405 |
+
"student_accuracy": 0.4375,
|
| 406 |
+
"student_approach_length": 500.0,
|
| 407 |
+
"teaching_length_mean": 1688.25,
|
| 408 |
+
"teaching_length_std": 1591.1127810270257,
|
| 409 |
+
"token_efficiency": 0.02000900419695598
|
| 410 |
+
},
|
| 411 |
+
{
|
| 412 |
+
"accuracy_delta": 0.15625,
|
| 413 |
+
"baseline_accuracy": 0.25,
|
| 414 |
+
"completion_length": 2566.109375,
|
| 415 |
+
"degradation_rate": 0.125,
|
| 416 |
+
"epoch": 0.10457516339869281,
|
| 417 |
+
"grad_norm": 0.635668741879516,
|
| 418 |
+
"improvement_rate": 0.28125,
|
| 419 |
+
"kl": 0.0020592212677001953,
|
| 420 |
+
"learning_rate": 2e-06,
|
| 421 |
+
"loss": 0.0001,
|
| 422 |
+
"reward": 0.3237999305129051,
|
| 423 |
+
"reward_std": 0.29250405728816986,
|
| 424 |
+
"rewards/AdaptiveTeachingReward": 0.3237999305129051,
|
| 425 |
+
"step": 20,
|
| 426 |
+
"student_accuracy": 0.40625,
|
| 427 |
+
"student_approach_length": 500.0,
|
| 428 |
+
"teaching_length_mean": 3000.8125,
|
| 429 |
+
"teaching_length_std": 1146.9664495361749,
|
| 430 |
+
"token_efficiency": 0.009248973352977438
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"accuracy_delta": -0.15625,
|
| 434 |
+
"baseline_accuracy": 0.5,
|
| 435 |
+
"completion_length": 2821.34375,
|
| 436 |
+
"degradation_rate": 0.15625,
|
| 437 |
+
"epoch": 0.10980392156862745,
|
| 438 |
+
"grad_norm": 1.038632983891241,
|
| 439 |
+
"improvement_rate": 0.0,
|
| 440 |
+
"kl": 0.0022677183151245117,
|
| 441 |
+
"learning_rate": 2e-06,
|
| 442 |
+
"loss": 0.0001,
|
| 443 |
+
"reward": 0.26167380064725876,
|
| 444 |
+
"reward_std": 0.20054005086421967,
|
| 445 |
+
"rewards/AdaptiveTeachingReward": 0.26167380064725876,
|
| 446 |
+
"step": 21,
|
| 447 |
+
"student_accuracy": 0.34375,
|
| 448 |
+
"student_approach_length": 500.0,
|
| 449 |
+
"teaching_length_mean": 2324.625,
|
| 450 |
+
"teaching_length_std": 1783.1465396131703,
|
| 451 |
+
"token_efficiency": 0.011391752427342916
|
| 452 |
+
},
|
| 453 |
+
{
|
| 454 |
+
"accuracy_delta": -0.0625,
|
| 455 |
+
"baseline_accuracy": 0.6875,
|
| 456 |
+
"completion_length": 1981.4453125,
|
| 457 |
+
"degradation_rate": 0.21875,
|
| 458 |
+
"epoch": 0.11503267973856209,
|
| 459 |
+
"grad_norm": 1.0591629807323937,
|
| 460 |
+
"improvement_rate": 0.15625,
|
| 461 |
+
"kl": 0.0025501251220703125,
|
| 462 |
+
"learning_rate": 2e-06,
|
| 463 |
+
"loss": 0.0001,
|
| 464 |
+
"reward": 0.3476671576499939,
|
| 465 |
+
"reward_std": 0.32001765072345734,
|
| 466 |
+
"rewards/AdaptiveTeachingReward": 0.3476671576499939,
|
| 467 |
+
"step": 22,
|
| 468 |
+
"student_accuracy": 0.625,
|
| 469 |
+
"student_approach_length": 500.0,
|
| 470 |
+
"teaching_length_mean": 1747.6875,
|
| 471 |
+
"teaching_length_std": 1544.1982674617702,
|
| 472 |
+
"token_efficiency": 0.019930595836054183
|
| 473 |
+
},
|
| 474 |
+
{
|
| 475 |
+
"accuracy_delta": 0.0,
|
| 476 |
+
"baseline_accuracy": 0.6875,
|
| 477 |
+
"completion_length": 2491.2109375,
|
| 478 |
+
"degradation_rate": 0.15625,
|
| 479 |
+
"epoch": 0.12026143790849673,
|
| 480 |
+
"grad_norm": 0.8736384977021919,
|
| 481 |
+
"improvement_rate": 0.15625,
|
| 482 |
+
"kl": 0.002077817916870117,
|
| 483 |
+
"learning_rate": 2e-06,
|
| 484 |
+
"loss": 0.0001,
|
| 485 |
+
"reward": 0.3966591954231262,
|
| 486 |
+
"reward_std": 0.35394637286663055,
|
| 487 |
+
"rewards/AdaptiveTeachingReward": 0.3966591954231262,
|
| 488 |
+
"step": 23,
|
| 489 |
+
"student_accuracy": 0.6875,
|
| 490 |
+
"student_approach_length": 500.0,
|
| 491 |
+
"teaching_length_mean": 2244.40625,
|
| 492 |
+
"teaching_length_std": 1464.1512714006012,
|
| 493 |
+
"token_efficiency": 0.020170202633726
|
| 494 |
+
},
|
| 495 |
+
{
|
| 496 |
+
"accuracy_delta": -0.0625,
|
| 497 |
+
"baseline_accuracy": 0.40625,
|
| 498 |
+
"completion_length": 3183.6796875,
|
| 499 |
+
"degradation_rate": 0.09375,
|
| 500 |
+
"epoch": 0.12549019607843137,
|
| 501 |
+
"grad_norm": 0.731937384244321,
|
| 502 |
+
"improvement_rate": 0.03125,
|
| 503 |
+
"kl": 0.002083301544189453,
|
| 504 |
+
"learning_rate": 2e-06,
|
| 505 |
+
"loss": 0.0001,
|
| 506 |
+
"reward": 0.24638524651527405,
|
| 507 |
+
"reward_std": 0.25337880849838257,
|
| 508 |
+
"rewards/AdaptiveTeachingReward": 0.24638524651527405,
|
| 509 |
+
"step": 24,
|
| 510 |
+
"student_accuracy": 0.34375,
|
| 511 |
+
"student_approach_length": 500.0,
|
| 512 |
+
"teaching_length_mean": 2833.75,
|
| 513 |
+
"teaching_length_std": 883.8529811162587,
|
| 514 |
+
"token_efficiency": 0.0154971457828618
|
| 515 |
+
},
|
| 516 |
+
{
|
| 517 |
+
"accuracy_delta": 0.09375,
|
| 518 |
+
"baseline_accuracy": 0.09375,
|
| 519 |
+
"completion_length": 2996.8203125,
|
| 520 |
+
"degradation_rate": 0.0625,
|
| 521 |
+
"epoch": 0.13071895424836602,
|
| 522 |
+
"grad_norm": 1.0366727211421645,
|
| 523 |
+
"improvement_rate": 0.15625,
|
| 524 |
+
"kl": 0.002071857452392578,
|
| 525 |
+
"learning_rate": 2e-06,
|
| 526 |
+
"loss": 0.0001,
|
| 527 |
+
"reward": 0.10002126544713974,
|
| 528 |
+
"reward_std": 0.12362907081842422,
|
| 529 |
+
"rewards/AdaptiveTeachingReward": 0.10002126544713974,
|
| 530 |
+
"step": 25,
|
| 531 |
+
"student_accuracy": 0.1875,
|
| 532 |
+
"student_approach_length": 500.0,
|
| 533 |
+
"teaching_length_mean": 2241.40625,
|
| 534 |
+
"teaching_length_std": 1680.2583046873353,
|
| 535 |
+
"token_efficiency": 0.005314979233325259
|
| 536 |
+
},
|
| 537 |
+
{
|
| 538 |
+
"accuracy_delta": 0.03125,
|
| 539 |
+
"baseline_accuracy": 0.125,
|
| 540 |
+
"completion_length": 2699.5078125,
|
| 541 |
+
"degradation_rate": 0.125,
|
| 542 |
+
"epoch": 0.13594771241830064,
|
| 543 |
+
"grad_norm": 0.8849958564728577,
|
| 544 |
+
"improvement_rate": 0.15625,
|
| 545 |
+
"kl": 0.00244140625,
|
| 546 |
+
"learning_rate": 2e-06,
|
| 547 |
+
"loss": 0.0001,
|
| 548 |
+
"reward": 0.21003766357898712,
|
| 549 |
+
"reward_std": 0.27024491131305695,
|
| 550 |
+
"rewards/AdaptiveTeachingReward": 0.21003766357898712,
|
| 551 |
+
"step": 26,
|
| 552 |
+
"student_accuracy": 0.15625,
|
| 553 |
+
"student_approach_length": 500.0,
|
| 554 |
+
"teaching_length_mean": 2865.75,
|
| 555 |
+
"teaching_length_std": 1575.6950824020187,
|
| 556 |
+
"token_efficiency": 0.00747702067329278
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"accuracy_delta": 0.09375,
|
| 560 |
+
"baseline_accuracy": 0.0,
|
| 561 |
+
"completion_length": 2216.78125,
|
| 562 |
+
"degradation_rate": 0.0,
|
| 563 |
+
"epoch": 0.1411764705882353,
|
| 564 |
+
"grad_norm": 0.6451950468247998,
|
| 565 |
+
"improvement_rate": 0.09375,
|
| 566 |
+
"kl": 0.0021767616271972656,
|
| 567 |
+
"learning_rate": 2e-06,
|
| 568 |
+
"loss": 0.0001,
|
| 569 |
+
"reward": 0.2089657336473465,
|
| 570 |
+
"reward_std": 0.21514078974723816,
|
| 571 |
+
"rewards/AdaptiveTeachingReward": 0.2089657336473465,
|
| 572 |
+
"step": 27,
|
| 573 |
+
"student_accuracy": 0.09375,
|
| 574 |
+
"student_approach_length": 499.96875,
|
| 575 |
+
"teaching_length_mean": 2619.40625,
|
| 576 |
+
"teaching_length_std": 1265.4369118665845,
|
| 577 |
+
"token_efficiency": 0.0060904088353781515
|
| 578 |
+
},
|
| 579 |
+
{
|
| 580 |
+
"accuracy_delta": 0.0,
|
| 581 |
+
"baseline_accuracy": 0.0,
|
| 582 |
+
"completion_length": 2668.4140625,
|
| 583 |
+
"degradation_rate": 0.0,
|
| 584 |
+
"epoch": 0.14640522875816994,
|
| 585 |
+
"grad_norm": 0.7896767066978999,
|
| 586 |
+
"improvement_rate": 0.0,
|
| 587 |
+
"kl": 0.002083301544189453,
|
| 588 |
+
"learning_rate": 2e-06,
|
| 589 |
+
"loss": 0.0001,
|
| 590 |
+
"reward": 0.14528799057006836,
|
| 591 |
+
"reward_std": 0.09566954523324966,
|
| 592 |
+
"rewards/AdaptiveTeachingReward": 0.14528799057006836,
|
| 593 |
+
"step": 28,
|
| 594 |
+
"student_accuracy": 0.0,
|
| 595 |
+
"student_approach_length": 500.0,
|
| 596 |
+
"teaching_length_mean": 3113.90625,
|
| 597 |
+
"teaching_length_std": 884.9974614855895,
|
| 598 |
+
"token_efficiency": 0.0035470700822770596
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"accuracy_delta": 0.03125,
|
| 602 |
+
"baseline_accuracy": 0.0,
|
| 603 |
+
"completion_length": 2467.4140625,
|
| 604 |
+
"degradation_rate": 0.0,
|
| 605 |
+
"epoch": 0.15163398692810456,
|
| 606 |
+
"grad_norm": 0.22872066233966626,
|
| 607 |
+
"improvement_rate": 0.03125,
|
| 608 |
+
"kl": 0.0025146007537841797,
|
| 609 |
+
"learning_rate": 2e-06,
|
| 610 |
+
"loss": 0.0001,
|
| 611 |
+
"reward": 0.007998689077794552,
|
| 612 |
+
"reward_std": 0.04524742066860199,
|
| 613 |
+
"rewards/AdaptiveTeachingReward": 0.007998689077794552,
|
| 614 |
+
"step": 29,
|
| 615 |
+
"student_accuracy": 0.03125,
|
| 616 |
+
"student_approach_length": 500.0,
|
| 617 |
+
"teaching_length_mean": 2764.9375,
|
| 618 |
+
"teaching_length_std": 1560.6896673437209,
|
| 619 |
+
"token_efficiency": 0.0002591402932910396
|
| 620 |
+
},
|
| 621 |
+
{
|
| 622 |
+
"accuracy_delta": 0.0,
|
| 623 |
+
"baseline_accuracy": 0.0,
|
| 624 |
+
"completion_length": 2854.59375,
|
| 625 |
+
"degradation_rate": 0.0,
|
| 626 |
+
"epoch": 0.1568627450980392,
|
| 627 |
+
"grad_norm": 0.5744778046050318,
|
| 628 |
+
"improvement_rate": 0.0,
|
| 629 |
+
"kl": 0.0022513866424560547,
|
| 630 |
+
"learning_rate": 2e-06,
|
| 631 |
+
"loss": 0.0001,
|
| 632 |
+
"reward": 0.16005077958106995,
|
| 633 |
+
"reward_std": 0.10021104663610458,
|
| 634 |
+
"rewards/AdaptiveTeachingReward": 0.16005077958106995,
|
| 635 |
+
"step": 30,
|
| 636 |
+
"student_accuracy": 0.0,
|
| 637 |
+
"student_approach_length": 500.0,
|
| 638 |
+
"teaching_length_mean": 2139.71875,
|
| 639 |
+
"teaching_length_std": 1711.9645016733543,
|
| 640 |
+
"token_efficiency": 0.007949871083127776
|
| 641 |
+
}
|
| 642 |
+
],
|
| 643 |
+
"logging_steps": 1,
|
| 644 |
+
"max_steps": 250,
|
| 645 |
+
"num_input_tokens_seen": 0,
|
| 646 |
+
"num_train_epochs": 2,
|
| 647 |
+
"save_steps": 10,
|
| 648 |
+
"stateful_callbacks": {
|
| 649 |
+
"TrainerControl": {
|
| 650 |
+
"args": {
|
| 651 |
+
"should_epoch_stop": false,
|
| 652 |
+
"should_evaluate": false,
|
| 653 |
+
"should_log": false,
|
| 654 |
+
"should_save": true,
|
| 655 |
+
"should_training_stop": false
|
| 656 |
+
},
|
| 657 |
+
"attributes": {}
|
| 658 |
+
}
|
| 659 |
+
},
|
| 660 |
+
"total_flos": 0.0,
|
| 661 |
+
"train_batch_size": 16,
|
| 662 |
+
"trial_name": null,
|
| 663 |
+
"trial_params": null
|
| 664 |
+
}
|