End of training

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 1.0292045772075653e-05,
-    "train_runtime": 191.3955,
+    "train_loss": 7.946134473968414e-06,
+    "train_runtime": 198.3078,
     "train_samples": 28,
-    "train_samples_per_second": 1.672,
-    "train_steps_per_second": 0.104
+    "train_samples_per_second": 1.614,
+    "train_steps_per_second": 0.101
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:269372e9aee00a429b27f6e5070e762525f1dddcc19afd826be0946b27d409c6
+oid sha256:e5a76669e4ad6e457be6aeb3907735d7b1cc74cb52cde7808113d27030de567d
 size 1976163472

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 1.0292045772075653e-05,
-    "train_runtime": 191.3955,
+    "train_loss": 7.946134473968414e-06,
+    "train_runtime": 198.3078,
     "train_samples": 28,
-    "train_samples_per_second": 1.672,
-    "train_steps_per_second": 0.104
+    "train_samples_per_second": 1.614,
+    "train_steps_per_second": 0.101
 }

trainer_state.json

@@ -10,203 +10,203 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 248.875,
+      "completion_length": 245.96875,
       "epoch": 0.5714285714285714,
-      "grad_norm": 7.0544843673706055,
+      "grad_norm": 7.138059616088867,
       "kl": 0.0,
       "learning_rate": 5e-07,
       "loss": 0.0,
-      "reward": 1.9025306217372417,
-      "reward_std": 1.6354162034549518,
-      "rewards/concensus_correctness_reward_func": 0.6328125,
-      "rewards/consensus_reward_func": 0.0625,
+      "reward": 1.978086668998003,
+      "reward_std": 1.549998540780507,
+      "rewards/concensus_correctness_reward_func": 0.7274375036358833,
+      "rewards/consensus_reward_func": 0.125,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.1875,
-      "rewards/question_recreation_reward_func": 0.38656184601131827,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.40799288742709905,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.633156256750226,
+      "rewards/xmlcount_reward_func": 0.7176562547683716,
       "step": 2
     },
     {
-      "completion_length": 158.125,
+      "completion_length": 346.8333333333333,
       "epoch": 1.0,
-      "grad_norm": 5.029088973999023,
-      "kl": 0.002313212321799559,
+      "grad_norm": 5.335450649261475,
+      "kl": 0.0022581211815122515,
       "learning_rate": 4.864543104251586e-07,
       "loss": 0.0,
-      "reward": 2.258645882209142,
-      "reward_std": 0.7633647699840367,
-      "rewards/concensus_correctness_reward_func": 0.23550000290075937,
-      "rewards/consensus_reward_func": 0.3333333333333333,
+      "reward": 1.0637057324250538,
+      "reward_std": 0.8543582057269911,
+      "rewards/concensus_correctness_reward_func": 0.0,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.16666666666666666,
-      "rewards/question_recreation_reward_func": 0.5354375538105766,
+      "rewards/question_recreation_reward_func": 0.462789058064421,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.9877083525061607,
+      "rewards/xmlcount_reward_func": 0.43425000955661136,
       "step": 4
     },
     {
-      "completion_length": 211.78125,
+      "completion_length": 237.0625,
       "epoch": 1.5714285714285714,
-      "grad_norm": 6.020397663116455,
-      "kl": 0.0041308358777314425,
+      "grad_norm": 5.941693305969238,
+      "kl": 0.0033424893372284714,
       "learning_rate": 4.472851273490984e-07,
       "loss": 0.0,
-      "reward": 1.4029428791254759,
-      "reward_std": 0.38589153878274374,
-      "rewards/concensus_correctness_reward_func": 0.06012500077486038,
-      "rewards/consensus_reward_func": 0.0,
+      "reward": 1.9579456727951765,
+      "reward_std": 1.5606304440298118,
+      "rewards/concensus_correctness_reward_func": 0.666812501847744,
+      "rewards/consensus_reward_func": 0.0625,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.125,
-      "rewards/question_recreation_reward_func": 0.46969289844855666,
+      "rewards/question_recreation_reward_func": 0.36688314378261566,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.748124998062849,
+      "rewards/xmlcount_reward_func": 0.7367500066757202,
       "step": 6
     },
     {
-      "completion_length": 228.66666666666666,
+      "completion_length": 219.70833333333334,
       "epoch": 2.0,
-      "grad_norm": 4.44202184677124,
-      "kl": 0.009342409209542287,
+      "grad_norm": 3.439047336578369,
+      "kl": 0.005905011979242166,
       "learning_rate": 3.867370395306068e-07,
       "loss": 0.0,
-      "reward": 1.8798749695221584,
-      "reward_std": 0.9903369223854194,
-      "rewards/concensus_correctness_reward_func": 0.172583336631457,
+      "reward": 1.7574667781591415,
+      "reward_std": 1.1081556143860023,
+      "rewards/concensus_correctness_reward_func": 0.17208333810170492,
       "rewards/consensus_reward_func": 0.25,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.16666666666666666,
-      "rewards/question_recreation_reward_func": 0.5592499550742408,
+      "rewards/question_recreation_reward_func": 0.354966773185879,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.7313749889532725,
+      "rewards/xmlcount_reward_func": 0.8137500037749609,
       "step": 8
     },
     {
-      "completion_length": 175.875,
+      "completion_length": 174.46875,
       "epoch": 2.571428571428571,
-      "grad_norm": 6.885485649108887,
-      "kl": 0.01059356433688663,
+      "grad_norm": 6.413847923278809,
+      "kl": 0.005604281301202718,
       "learning_rate": 3.1137137178519977e-07,
       "loss": 0.0,
-      "reward": 1.3127059005200863,
-      "reward_std": 0.5222314277198166,
+      "reward": 1.3979325704276562,
+      "reward_std": 0.33654866518918425,
       "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0625,
-      "rewards/question_recreation_reward_func": 0.38879964826628566,
+      "rewards/final_correctness_reward_func": 0.125,
+      "rewards/question_recreation_reward_func": 0.34633880454930477,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.8614062555134296,
+      "rewards/xmlcount_reward_func": 0.9265937581658363,
       "step": 10
     },
     {
-      "completion_length": 156.5,
+      "completion_length": 172.125,
       "epoch": 3.0,
-      "grad_norm": 4.863602638244629,
-      "kl": 0.012806190915095309,
+      "grad_norm": 4.729302406311035,
+      "kl": 0.012371947659024348,
       "learning_rate": 2.2935516363191693e-07,
       "loss": 0.0,
-      "reward": 1.9954389035701752,
-      "reward_std": 0.8163471951459845,
-      "rewards/concensus_correctness_reward_func": 0.1652500033378601,
-      "rewards/consensus_reward_func": 0.16666666666666666,
+      "reward": 2.351278011997541,
+      "reward_std": 1.1179534547651808,
+      "rewards/concensus_correctness_reward_func": 0.3329166720310847,
+      "rewards/consensus_reward_func": 0.4166666666666667,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.08333333333333333,
-      "rewards/question_recreation_reward_func": 0.5653555756434798,
+      "rewards/question_recreation_reward_func": 0.5985279902815819,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 1.0148333460092545,
+      "rewards/xmlcount_reward_func": 0.9198333447178205,
       "step": 12
     },
     {
-      "completion_length": 252.9375,
+      "completion_length": 154.5,
       "epoch": 3.571428571428571,
-      "grad_norm": 8.07120132446289,
-      "kl": 0.022088178637204692,
+      "grad_norm": 6.839085578918457,
+      "kl": 0.02112107082211878,
       "learning_rate": 1.4957614383675767e-07,
       "loss": 0.0,
-      "reward": 2.21107729524374,
-      "reward_std": 1.3267412132117897,
-      "rewards/concensus_correctness_reward_func": 0.35843750834465027,
-      "rewards/consensus_reward_func": 0.25,
+      "reward": 1.7222793698310852,
+      "reward_std": 0.7603381713852286,
+      "rewards/concensus_correctness_reward_func": 0.06843750178813934,
+      "rewards/consensus_reward_func": 0.125,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.1875,
-      "rewards/question_recreation_reward_func": 0.5949522890150547,
+      "rewards/final_correctness_reward_func": 0.0625,
+      "rewards/question_recreation_reward_func": 0.4770293580368161,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.8201874941587448,
+      "rewards/xmlcount_reward_func": 0.9893125109374523,
       "step": 14
     },
     {
-      "completion_length": 239.16666666666666,
+      "completion_length": 201.79166666666666,
       "epoch": 4.0,
-      "grad_norm": 4.181697845458984,
-      "kl": 0.020798012391120817,
+      "grad_norm": 7.2178497314453125,
+      "kl": 0.01169851414548854,
       "learning_rate": 8.067960709356478e-08,
       "loss": 0.0,
-      "reward": 1.5264141807953517,
-      "reward_std": 0.6949908261497816,
-      "rewards/concensus_correctness_reward_func": 0.16033333539962769,
-      "rewards/consensus_reward_func": 0.08333333333333333,
+      "reward": 1.5704542646805446,
+      "reward_std": 0.9336798458049694,
+      "rewards/concensus_correctness_reward_func": 0.16166667143503824,
+      "rewards/consensus_reward_func": 0.16666666666666666,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.08333333333333333,
-      "rewards/question_recreation_reward_func": 0.4137058320532863,
+      "rewards/final_correctness_reward_func": 0.16666666666666666,
+      "rewards/question_recreation_reward_func": 0.418662591992567,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.785708320637544,
+      "rewards/xmlcount_reward_func": 0.6567916870117188,
       "step": 16
     },
     {
-      "completion_length": 248.21875,
+      "completion_length": 246.5625,
       "epoch": 4.571428571428571,
-      "grad_norm": 7.192660331726074,
-      "kl": 0.02242492102959659,
+      "grad_norm": 5.255609512329102,
+      "kl": 0.012498048308771104,
       "learning_rate": 3.013156219837776e-08,
       "loss": 0.0,
-      "reward": 1.6895426772534847,
-      "reward_std": 0.7501816740259528,
-      "rewards/concensus_correctness_reward_func": 0.1593750026077032,
-      "rewards/consensus_reward_func": 0.25,
+      "reward": 1.578137394040823,
+      "reward_std": 0.5315014050574973,
+      "rewards/concensus_correctness_reward_func": 0.0078125,
+      "rewards/consensus_reward_func": 0.0625,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.4894801494665444,
+      "rewards/final_correctness_reward_func": 0.125,
+      "rewards/question_recreation_reward_func": 0.49882487556897104,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.7906875088810921,
+      "rewards/xmlcount_reward_func": 0.884000001475215,
       "step": 18
     },
     {
-      "completion_length": 165.29166666666666,
+      "completion_length": 170.33333333333334,
       "epoch": 5.0,
-      "grad_norm": 6.6531901359558105,
-      "kl": 0.010204441185730198,
+      "grad_norm": 7.401615619659424,
+      "kl": 0.01765511992077033,
       "learning_rate": 3.4096741493194193e-09,
       "loss": 0.0,
-      "reward": 1.9160203635692596,
-      "reward_std": 0.8500873198111852,
-      "rewards/concensus_correctness_reward_func": 0.12016666928927104,
+      "reward": 1.9397808909416199,
+      "reward_std": 1.1341329428056877,
+      "rewards/concensus_correctness_reward_func": 0.24175000687440237,
       "rewards/consensus_reward_func": 0.16666666666666666,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.25,
-      "rewards/question_recreation_reward_func": 0.4647703676794966,
+      "rewards/final_correctness_reward_func": 0.16666666666666666,
+      "rewards/question_recreation_reward_func": 0.48361418799807626,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.9144166658322016,
+      "rewards/xmlcount_reward_func": 0.8810833394527435,
       "step": 20
     },
     {
       "epoch": 5.0,
       "step": 20,
       "total_flos": 0.0,
-      "train_loss": 1.0292045772075653e-05,
-      "train_runtime": 191.3955,
-      "train_samples_per_second": 1.672,
-      "train_steps_per_second": 0.104
+      "train_loss": 7.946134473968414e-06,
+      "train_runtime": 198.3078,
+      "train_samples_per_second": 1.614,
+      "train_steps_per_second": 0.101
     }
   ],
   "logging_steps": 2,