End of training

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 1.2860417473348208e-05,
-    "train_runtime": 555.7683,
-    "train_samples": 189,
-    "train_samples_per_second": 0.576,
-    "train_steps_per_second": 0.036
+    "train_loss": 3.397261410782448e-05,
+    "train_runtime": 331.6611,
+    "train_samples": 160,
+    "train_samples_per_second": 0.965,
+    "train_steps_per_second": 0.06
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:67cf7e6e18ab068c77c91c259ada264fa2c47f05b5c3ec45bfb9c9012c4a45d7
+oid sha256:3a9b7aeccb5585dec47b287ded621768218ecb4111e020a6f03cd1a09ebf3d74
 size 1976163472

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 1.2860417473348208e-05,
-    "train_runtime": 555.7683,
-    "train_samples": 189,
-    "train_samples_per_second": 0.576,
-    "train_steps_per_second": 0.036
+    "train_loss": 3.397261410782448e-05,
+    "train_runtime": 331.6611,
+    "train_samples": 160,
+    "train_samples_per_second": 0.965,
+    "train_steps_per_second": 0.06
 }

trainer_state.json

@@ -2,7 +2,7 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.8465608465608465,
+  "epoch": 1.0,
   "eval_steps": 500,
   "global_step": 20,
   "is_hyper_param_search": false,
@@ -10,203 +10,203 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 452.03125,
-      "epoch": 0.08465608465608465,
-      "grad_norm": 29.402589797973633,
+      "completion_length": 257.84375,
+      "epoch": 0.1,
+      "grad_norm": 71.31194305419922,
       "kl": 0.0,
       "learning_rate": 4.965903258506806e-07,
       "loss": -0.0,
-      "reward": 0.9871471562073566,
-      "reward_std": 1.2973561512553715,
-      "rewards/concensus_correctness_reward_func": 0.625,
+      "reward": 0.6542720094294054,
+      "reward_std": 0.671761574325501,
+      "rewards/concensus_correctness_reward_func": 0.109375,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0625,
-      "rewards/question_recreation_reward_func": 0.1523971924907528,
-      "rewards/soft_format_reward_func": 0.015625,
+      "rewards/final_correctness_reward_func": 0.375,
+      "rewards/question_recreation_reward_func": 0.11333451081009116,
+      "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.1316249966621399,
+      "rewards/xmlcount_reward_func": 0.05656249821186066,
       "step": 2
     },
     {
-      "completion_length": 320.875,
-      "epoch": 0.1693121693121693,
-      "grad_norm": 12.040271759033203,
-      "kl": 0.0018746699679468293,
+      "completion_length": 388.75,
+      "epoch": 0.2,
+      "grad_norm": 48.57069778442383,
+      "kl": 0.0015563661654596217,
       "learning_rate": 4.698684378016222e-07,
       "loss": 0.0,
-      "reward": 0.6975098436232656,
-      "reward_std": 0.697691164445132,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0625,
+      "reward": 1.34490574139636,
+      "reward_std": 1.4263234985410236,
+      "rewards/concensus_correctness_reward_func": 0.625,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.375,
-      "rewards/question_recreation_reward_func": 0.20275984914042056,
+      "rewards/final_correctness_reward_func": 0.5,
+      "rewards/question_recreation_reward_func": 0.13059324386995286,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.015625,
-      "rewards/xmlcount_reward_func": 0.041624996811151505,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.08931249612942338,
       "step": 4
     },
     {
-      "completion_length": 319.65625,
-      "epoch": 0.25396825396825395,
-      "grad_norm": 78.67212677001953,
-      "kl": 0.006561886559211416,
+      "completion_length": 355.125,
+      "epoch": 0.3,
+      "grad_norm": 11.513766288757324,
+      "kl": 0.0018346092207366382,
       "learning_rate": 4.193203929064353e-07,
       "loss": 0.0,
-      "reward": 0.4108218799519818,
-      "reward_std": 0.35047392526757903,
+      "reward": 0.3823581865581218,
+      "reward_std": 0.31888675036452696,
       "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.125,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.0625,
-      "rewards/question_recreation_reward_func": 0.14541563153034076,
+      "rewards/question_recreation_reward_func": 0.20129567771800794,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.07790625107008964,
+      "rewards/xmlcount_reward_func": 0.11856249894481152,
       "step": 6
     },
     {
-      "completion_length": 277.9375,
-      "epoch": 0.3386243386243386,
-      "grad_norm": 124.32317352294922,
-      "kl": 0.02648140014935052,
+      "completion_length": 332.78125,
+      "epoch": 0.4,
+      "grad_norm": 5.0440673828125,
+      "kl": 0.0017456002333346987,
       "learning_rate": 3.5042385616324236e-07,
       "loss": 0.0,
-      "reward": 1.175961585715413,
-      "reward_std": 1.4880700244102627,
-      "rewards/concensus_correctness_reward_func": 0.625,
-      "rewards/consensus_reward_func": 0.0625,
+      "reward": 0.7418190783355385,
+      "reward_std": 0.6216794796055183,
+      "rewards/concensus_correctness_reward_func": 0.0,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.1875,
-      "rewards/question_recreation_reward_func": 0.1822116176772397,
+      "rewards/final_correctness_reward_func": 0.5,
+      "rewards/question_recreation_reward_func": 0.16697532951366156,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.015625,
-      "rewards/xmlcount_reward_func": 0.10312500060535967,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.07484374335035682,
       "step": 8
     },
     {
-      "completion_length": 333.96875,
-      "epoch": 0.42328042328042326,
-      "grad_norm": 56.98233413696289,
-      "kl": 0.0017482519033364952,
+      "completion_length": 367.375,
+      "epoch": 0.5,
+      "grad_norm": 15.639695167541504,
+      "kl": 0.0024030036183830816,
       "learning_rate": 2.706448363680831e-07,
       "loss": 0.0,
-      "reward": 1.259338534262497,
-      "reward_std": 1.4753339138696902,
-      "rewards/concensus_correctness_reward_func": 0.625,
+      "reward": 0.8275067815557122,
+      "reward_std": 0.9870248432271183,
+      "rewards/concensus_correctness_reward_func": 0.0599374994635582,
       "rewards/consensus_reward_func": 0.0625,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.375,
-      "rewards/question_recreation_reward_func": 0.18893231771653518,
+      "rewards/final_correctness_reward_func": 0.5625,
+      "rewards/question_recreation_reward_func": 0.09656929643824697,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.007906252052634954,
+      "rewards/xmlcount_reward_func": 0.04600000171922147,
       "step": 10
     },
     {
-      "completion_length": 283.4375,
-      "epoch": 0.5079365079365079,
-      "grad_norm": 184.6020965576172,
-      "kl": 0.00985822520306101,
+      "completion_length": 377.71875,
+      "epoch": 0.6,
+      "grad_norm": 8.872749328613281,
+      "kl": 0.003477493522950681,
       "learning_rate": 1.886286282148002e-07,
       "loss": 0.0,
-      "reward": 1.576558955362998,
-      "reward_std": 2.3915363581618294,
-      "rewards/concensus_correctness_reward_func": 1.25,
-      "rewards/consensus_reward_func": 0.1875,
+      "reward": 0.5521137183532119,
+      "reward_std": 0.9777508705155924,
+      "rewards/concensus_correctness_reward_func": 0.05999999865889549,
+      "rewards/consensus_reward_func": 0.0625,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0625,
-      "rewards/question_recreation_reward_func": 0.0650276712840423,
+      "rewards/final_correctness_reward_func": 0.5,
+      "rewards/question_recreation_reward_func": 0.20373869099421427,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.011531248688697815,
+      "rewards/xmlcount_reward_func": -0.27412500232458115,
       "step": 12
     },
     {
-      "completion_length": 288.4375,
-      "epoch": 0.5925925925925926,
-      "grad_norm": 125.96855163574219,
-      "kl": 0.06800110284348193,
+      "completion_length": 343.59375,
+      "epoch": 0.7,
+      "grad_norm": 5.5229268074035645,
+      "kl": 0.002653340348842903,
       "learning_rate": 1.1326296046939333e-07,
-      "loss": 0.0001,
-      "reward": 1.2055153002147563,
-      "reward_std": 1.216392180234834,
-      "rewards/concensus_correctness_reward_func": 0.6851250007748604,
+      "loss": 0.0,
+      "reward": 1.1060769520699978,
+      "reward_std": 0.8273453344590962,
+      "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.1875,
-      "rewards/question_recreation_reward_func": 0.18382781816762872,
+      "rewards/final_correctness_reward_func": 0.875,
+      "rewards/question_recreation_reward_func": 0.18898318015271798,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.14906249940395355,
+      "rewards/xmlcount_reward_func": 0.04209375102072954,
       "step": 14
     },
     {
-      "completion_length": 241.90625,
-      "epoch": 0.6772486772486772,
-      "grad_norm": 70.70581817626953,
-      "kl": 0.010532293115829816,
+      "completion_length": 424.90625,
+      "epoch": 0.8,
+      "grad_norm": 19.185876846313477,
+      "kl": 0.27459875560816727,
       "learning_rate": 5.271487265090163e-08,
-      "loss": 0.0,
-      "reward": 2.7857588599435985,
-      "reward_std": 3.4491555616259575,
-      "rewards/concensus_correctness_reward_func": 1.875,
-      "rewards/consensus_reward_func": 0.125,
+      "loss": 0.0003,
+      "reward": 1.4826910288538784,
+      "reward_std": 0.36641132296063006,
+      "rewards/concensus_correctness_reward_func": 1.25,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.5,
-      "rewards/question_recreation_reward_func": 0.1783213593007531,
+      "rewards/final_correctness_reward_func": 0.125,
+      "rewards/question_recreation_reward_func": 0.14225358597468585,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.10743749956600368,
+      "rewards/xmlcount_reward_func": -0.03456250127055682,
       "step": 16
     },
     {
-      "completion_length": 440.34375,
-      "epoch": 0.7619047619047619,
-      "grad_norm": 25.547298431396484,
-      "kl": 0.0022962762595852837,
+      "completion_length": 284.46875,
+      "epoch": 0.9,
+      "grad_norm": 40.3567008972168,
+      "kl": 0.026134049770917045,
       "learning_rate": 1.3545689574841341e-08,
       "loss": 0.0,
-      "reward": 0.8677146534901112,
-      "reward_std": 0.7431555229704827,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.125,
+      "reward": 2.7290455140173435,
+      "reward_std": 1.5447391173802316,
+      "rewards/concensus_correctness_reward_func": 1.9351250007748604,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.1875,
-      "rewards/question_recreation_reward_func": 0.27558966260403395,
+      "rewards/final_correctness_reward_func": 0.5625,
+      "rewards/question_recreation_reward_func": 0.16785799397621304,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.27962499926798046,
+      "rewards/strict_format_reward_func": 0.015625,
+      "rewards/xmlcount_reward_func": 0.04793750052340329,
       "step": 18
     },
     {
-      "completion_length": 301.59375,
-      "epoch": 0.8465608465608465,
-      "grad_norm": 13.039204597473145,
-      "kl": 0.0017621701590542216,
+      "completion_length": 350.09375,
+      "epoch": 1.0,
+      "grad_norm": 52.060001373291016,
+      "kl": 0.025269348172514583,
       "learning_rate": 0.0,
       "loss": 0.0,
-      "reward": 0.7904947228962556,
-      "reward_std": 0.6705761743942276,
-      "rewards/concensus_correctness_reward_func": 0.0,
+      "reward": 1.9592706635594368,
+      "reward_std": 2.636638791358564,
+      "rewards/concensus_correctness_reward_func": 1.25,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.4375,
-      "rewards/question_recreation_reward_func": 0.19968222291208804,
+      "rewards/question_recreation_reward_func": 0.19583316356875002,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.15331249870359898,
+      "rewards/xmlcount_reward_func": 0.07593749847728759,
       "step": 20
     },
     {
-      "epoch": 0.8465608465608465,
+      "epoch": 1.0,
       "step": 20,
       "total_flos": 0.0,
-      "train_loss": 1.2860417473348208e-05,
-      "train_runtime": 555.7683,
-      "train_samples_per_second": 0.576,
-      "train_steps_per_second": 0.036
+      "train_loss": 3.397261410782448e-05,
+      "train_runtime": 331.6611,
+      "train_samples_per_second": 0.965,
+      "train_steps_per_second": 0.06
     }
   ],
   "logging_steps": 2,