End of training

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.00011620071672950872,
-    "train_runtime": 657.8689,
+    "train_loss": 1.3666093832398473e-05,
+    "train_runtime": 806.7775,
     "train_samples": 160,
-    "train_samples_per_second": 0.486,
-    "train_steps_per_second": 0.03
+    "train_samples_per_second": 0.397,
+    "train_steps_per_second": 0.025
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e5a228fb920cc8ae208a5f1952a082f0345eb795d272b028e97c7a6eb310128a
+oid sha256:5cd9f6190971fc6940d7b7203e2332c5ef183ce3b68dff60758644ae811d2c18
 size 1976163472

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.00011620071672950872,
-    "train_runtime": 657.8689,
+    "train_loss": 1.3666093832398473e-05,
+    "train_runtime": 806.7775,
     "train_samples": 160,
-    "train_samples_per_second": 0.486,
-    "train_steps_per_second": 0.03
+    "train_samples_per_second": 0.397,
+    "train_steps_per_second": 0.025
 }

trainer_state.json

@@ -10,203 +10,203 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 188.875,
+      "completion_length": 158.84375,
       "epoch": 0.1,
-      "grad_norm": 20.606088638305664,
-      "kl": -1.8626451769865326e-10,
+      "grad_norm": 92.82727813720703,
+      "kl": 0.0,
       "learning_rate": 4.965903258506806e-07,
-      "loss": -0.0,
-      "reward": -0.08459861180745065,
-      "reward_std": 0.4505786441732198,
+      "loss": 0.0,
+      "reward": 0.1468239967362024,
+      "reward_std": 0.17140711040701717,
       "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": 0.08155764313414693,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.1628240012214519,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": -0.22865625121630728,
+      "rewards/xmlcount_reward_func": -0.016000002156943083,
       "step": 2
     },
     {
-      "completion_length": 159.0,
+      "completion_length": 88.96875,
       "epoch": 0.2,
-      "grad_norm": 17.4925537109375,
-      "kl": 0.012501606208388694,
+      "grad_norm": 17.767974853515625,
+      "kl": 0.0021755212871994445,
       "learning_rate": 4.698684378016222e-07,
       "loss": 0.0,
-      "reward": 0.15525958803482354,
-      "reward_std": 0.13993048649354023,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0,
+      "reward": 0.5362368610221893,
+      "reward_std": 0.18039140166092693,
+      "rewards/concensus_correctness_reward_func": 0.1185000017285347,
+      "rewards/consensus_reward_func": 0.125,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0625,
-      "rewards/question_recreation_reward_func": 0.10338459047488868,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.17770562344230711,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": -0.010625000111758709,
+      "rewards/xmlcount_reward_func": 0.1150312521494925,
       "step": 4
     },
     {
-      "completion_length": 103.90625,
+      "completion_length": 161.3125,
       "epoch": 0.3,
-      "grad_norm": 44.36183547973633,
-      "kl": 0.116791727632517,
+      "grad_norm": 7.2867231369018555,
+      "kl": 0.004310879157856107,
       "learning_rate": 4.193203929064353e-07,
-      "loss": 0.0001,
-      "reward": 0.09123604014166631,
-      "reward_std": 0.024207589820434805,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0,
+      "loss": 0.0,
+      "reward": 0.4580355920479633,
+      "reward_std": 0.26154627031223754,
+      "rewards/concensus_correctness_reward_func": 0.0599374994635582,
+      "rewards/consensus_reward_func": 0.0625,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.09123604014166631,
-      "rewards/soft_format_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.179504333762452,
+      "rewards/soft_format_reward_func": 0.015625,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.14046875084750354,
       "step": 6
     },
     {
-      "completion_length": 65.5625,
+      "completion_length": 138.25,
       "epoch": 0.4,
-      "grad_norm": 23.13948631286621,
-      "kl": 0.14583962078904733,
+      "grad_norm": 7.924788951873779,
+      "kl": 0.011087903680163436,
       "learning_rate": 3.5042385616324236e-07,
-      "loss": 0.0001,
-      "reward": 0.10612659784965217,
-      "reward_std": 0.07028272053867113,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0,
+      "loss": 0.0,
+      "reward": 0.5225309516536072,
+      "reward_std": 0.49480345542542636,
+      "rewards/concensus_correctness_reward_func": 0.04700000025331974,
+      "rewards/consensus_reward_func": 0.125,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.0766265980200842,
+      "rewards/final_correctness_reward_func": 0.0625,
+      "rewards/question_recreation_reward_func": 0.12593720143195242,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.02949999994598329,
+      "rewards/xmlcount_reward_func": 0.16209374903701246,
       "step": 8
     },
     {
-      "completion_length": 125.875,
+      "completion_length": 166.46875,
       "epoch": 0.5,
-      "grad_norm": 18.663105010986328,
-      "kl": 0.2906235277914675,
+      "grad_norm": 10.63830852508545,
+      "kl": 0.013151258695870638,
       "learning_rate": 2.706448363680831e-07,
-      "loss": 0.0003,
-      "reward": 0.05466344498563558,
-      "reward_std": 0.022103407893155236,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0,
+      "loss": 0.0,
+      "reward": 0.4802669673226774,
+      "reward_std": 0.3484726174501702,
+      "rewards/concensus_correctness_reward_func": 0.02800000086426735,
+      "rewards/consensus_reward_func": 0.0625,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.05138219508808106,
+      "rewards/question_recreation_reward_func": 0.16017321776598692,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.003281249897554517,
+      "rewards/xmlcount_reward_func": 0.22959374566562474,
       "step": 10
     },
     {
-      "completion_length": 81.15625,
+      "completion_length": 173.4375,
       "epoch": 0.6,
-      "grad_norm": 17.701265335083008,
-      "kl": 0.08585882172337733,
+      "grad_norm": 9.655745506286621,
+      "kl": 0.016364283743314445,
       "learning_rate": 1.886286282148002e-07,
-      "loss": 0.0001,
-      "reward": 0.06056131387595087,
-      "reward_std": 0.0290565535397036,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0,
+      "loss": 0.0,
+      "reward": 0.8550984656903893,
+      "reward_std": 0.33143275073962286,
+      "rewards/concensus_correctness_reward_func": 0.2682499997317791,
+      "rewards/consensus_reward_func": 0.3125,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.05399881454650313,
+      "rewards/question_recreation_reward_func": 0.14825469313655049,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.006562499795109034,
+      "rewards/xmlcount_reward_func": 0.12609374476596713,
       "step": 12
     },
     {
-      "completion_length": 103.40625,
+      "completion_length": 109.84375,
       "epoch": 0.7,
-      "grad_norm": 33.02743148803711,
-      "kl": 0.103502563375514,
+      "grad_norm": 7.929982662200928,
+      "kl": 0.021056591620435938,
       "learning_rate": 1.1326296046939333e-07,
-      "loss": 0.0001,
-      "reward": 0.12133114249445498,
-      "reward_std": 0.1054717963852454,
+      "loss": 0.0,
+      "reward": 0.378742050845176,
+      "reward_std": 0.1508836179878017,
       "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": 0.05883114255266264,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.14089830702869222,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.23784374445676804,
       "step": 14
     },
     {
-      "completion_length": 44.375,
+      "completion_length": 146.1875,
       "epoch": 0.8,
-      "grad_norm": 30.598201751708984,
-      "kl": 0.15886804103502072,
+      "grad_norm": 6.844194412231445,
+      "kl": 0.016575219749938697,
       "learning_rate": 5.271487265090163e-08,
-      "loss": 0.0002,
-      "reward": 0.3126172423362732,
-      "reward_std": 0.285806781725114,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0,
+      "loss": 0.0,
+      "reward": 1.0233479777816683,
+      "reward_std": 0.34056491823866963,
+      "rewards/concensus_correctness_reward_func": 0.2397499978542328,
+      "rewards/consensus_reward_func": 0.25,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.1875,
-      "rewards/question_recreation_reward_func": 0.11936723813414574,
+      "rewards/final_correctness_reward_func": 0.125,
+      "rewards/question_recreation_reward_func": 0.24947296862956136,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.005750000011175871,
+      "rewards/xmlcount_reward_func": 0.15912500116974115,
       "step": 16
     },
     {
-      "completion_length": 77.59375,
+      "completion_length": 128.4375,
       "epoch": 0.9,
-      "grad_norm": 8.833693504333496,
-      "kl": 0.06381005875300616,
+      "grad_norm": 25.0986270904541,
+      "kl": 0.022347735619405285,
       "learning_rate": 1.3545689574841341e-08,
-      "loss": 0.0001,
-      "reward": 0.23175003798678517,
-      "reward_std": 0.031403945409692824,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0,
+      "loss": 0.0,
+      "reward": 0.8327681426890194,
+      "reward_std": 0.3936522727308329,
+      "rewards/concensus_correctness_reward_func": 0.15112500078976154,
+      "rewards/consensus_reward_func": 0.125,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.125,
-      "rewards/question_recreation_reward_func": 0.10675003798678517,
+      "rewards/final_correctness_reward_func": 0.1875,
+      "rewards/question_recreation_reward_func": 0.2066118831280619,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.16253125155344605,
       "step": 18
     },
     {
-      "completion_length": 97.0625,
+      "completion_length": 138.625,
       "epoch": 1.0,
-      "grad_norm": 23.614116668701172,
-      "kl": 0.18440437169920187,
+      "grad_norm": 20.88378143310547,
+      "kl": 0.02928804296971066,
       "learning_rate": 0.0,
-      "loss": 0.0002,
-      "reward": 0.15951462506200187,
-      "reward_std": 0.12634751566764635,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0,
+      "loss": 0.0,
+      "reward": 0.8610214402433485,
+      "reward_std": 0.6351559593313141,
+      "rewards/concensus_correctness_reward_func": 0.14743749983608723,
+      "rewards/consensus_reward_func": 0.1875,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0625,
-      "rewards/question_recreation_reward_func": 0.10354587234905921,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.19989642850123346,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": -0.006531250197440386,
+      "rewards/xmlcount_reward_func": 0.3261875007301569,
       "step": 20
     },
     {
       "epoch": 1.0,
       "step": 20,
       "total_flos": 0.0,
-      "train_loss": 0.00011620071672950872,
-      "train_runtime": 657.8689,
-      "train_samples_per_second": 0.486,
-      "train_steps_per_second": 0.03
+      "train_loss": 1.3666093832398473e-05,
+      "train_runtime": 806.7775,
+      "train_samples_per_second": 0.397,
+      "train_steps_per_second": 0.025
     }
   ],
   "logging_steps": 2,