End of training

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 4.6607106924057e-06,
-    "train_runtime": 528.4481,
-    "train_samples": 28,
-    "train_samples_per_second": 0.606,
-    "train_steps_per_second": 0.038
+    "train_loss": 7.716044910921482e-06,
+    "train_runtime": 943.9196,
+    "train_samples": 51,
+    "train_samples_per_second": 0.339,
+    "train_steps_per_second": 0.021
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:61ab4f847ebc3d81a28f6bed48fd929a834a56ceb7613054608622428531a563
+oid sha256:1f1a4572f5abe2eff668f5bbae2f855a40bdd09e1b4adf2747aa992dbec6945b
 size 1976163472

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 4.6607106924057e-06,
-    "train_runtime": 528.4481,
-    "train_samples": 28,
-    "train_samples_per_second": 0.606,
-    "train_steps_per_second": 0.038
+    "train_loss": 7.716044910921482e-06,
+    "train_runtime": 943.9196,
+    "train_samples": 51,
+    "train_samples_per_second": 0.339,
+    "train_steps_per_second": 0.021
 }

trainer_state.json

@@ -2,7 +2,7 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 5.0,
+  "epoch": 2.9411764705882355,
   "eval_steps": 500,
   "global_step": 20,
   "is_hyper_param_search": false,
@@ -10,209 +10,209 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 291.71875,
-      "epoch": 0.5714285714285714,
-      "grad_norm": 5.5191755294799805,
+      "completion_length": 320.21875,
+      "epoch": 0.3137254901960784,
+      "grad_norm": 18.987831115722656,
       "kl": 0.0,
       "learning_rate": 5e-07,
       "loss": 0.0,
-      "reward": 1.5525079816579819,
-      "reward_std": 1.210987733094953,
-      "rewards/concensus_correctness_reward_func": 0.3031249977648258,
+      "reward": 0.9052386409603059,
+      "reward_std": 1.0872065802104771,
+      "rewards/concensus_correctness_reward_func": 0.0078125,
       "rewards/consensus_reward_func": 0.25,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.125,
-      "rewards/question_recreation_reward_func": 0.4316017273813486,
+      "rewards/final_correctness_reward_func": 0.0625,
+      "rewards/question_recreation_reward_func": 0.32886362238787115,
       "rewards/soft_format_reward_func": 0.015625,
-      "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.42715625558048487,
+      "rewards/strict_format_reward_func": 0.015625,
+      "rewards/xmlcount_reward_func": 0.22481249924749136,
       "step": 2
     },
     {
-      "completion_length": 187.91666666666666,
-      "epoch": 1.0,
-      "grad_norm": 3.369981050491333,
-      "kl": 0.0018891085104163115,
+      "completion_length": 255.65625,
+      "epoch": 0.6274509803921569,
+      "grad_norm": 7.594056606292725,
+      "kl": 0.0013956625989521854,
       "learning_rate": 4.864543104251586e-07,
       "loss": 0.0,
-      "reward": 1.733353316783905,
-      "reward_std": 0.8906166382754842,
-      "rewards/concensus_correctness_reward_func": 0.1579166700442632,
-      "rewards/consensus_reward_func": 0.16666666666666666,
+      "reward": 1.7687878240831196,
+      "reward_std": 0.8872564652119763,
+      "rewards/concensus_correctness_reward_func": 0.19506250228732824,
+      "rewards/consensus_reward_func": 0.1875,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.16666666666666666,
-      "rewards/question_recreation_reward_func": 0.477644973124067,
+      "rewards/final_correctness_reward_func": 0.1875,
+      "rewards/question_recreation_reward_func": 0.5405378290452063,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.7644583284854889,
+      "rewards/xmlcount_reward_func": 0.6581875095143914,
       "step": 4
     },
     {
-      "completion_length": 226.65625,
-      "epoch": 1.5714285714285714,
-      "grad_norm": 5.8894524574279785,
-      "kl": 0.003135249629849568,
+      "completion_length": 241.71875,
+      "epoch": 0.9411764705882353,
+      "grad_norm": 5.940536022186279,
+      "kl": 0.0032457121487823315,
       "learning_rate": 4.472851273490984e-07,
       "loss": 0.0,
-      "reward": 1.976506132632494,
-      "reward_std": 1.3569977073930204,
-      "rewards/concensus_correctness_reward_func": 0.3553749993443489,
-      "rewards/consensus_reward_func": 0.3125,
+      "reward": 3.927860476076603,
+      "reward_std": 1.7695390642620623,
+      "rewards/concensus_correctness_reward_func": 2.1019999980926514,
+      "rewards/consensus_reward_func": 0.375,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.125,
-      "rewards/question_recreation_reward_func": 0.524193589692004,
+      "rewards/final_correctness_reward_func": 0.25,
+      "rewards/question_recreation_reward_func": 0.43264187331078574,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.6594374980777502,
+      "rewards/xmlcount_reward_func": 0.7682187557220459,
       "step": 6
     },
     {
-      "completion_length": 240.20833333333334,
-      "epoch": 2.0,
-      "grad_norm": 4.913305759429932,
-      "kl": 0.003807866412292545,
+      "completion_length": 256.0,
+      "epoch": 1.156862745098039,
+      "grad_norm": 9.757232666015625,
+      "kl": 0.00435605255717581,
       "learning_rate": 3.867370395306068e-07,
       "loss": 0.0,
-      "reward": 1.831045498450597,
-      "reward_std": 0.44153941185989726,
-      "rewards/concensus_correctness_reward_func": 0.010416666666666666,
-      "rewards/consensus_reward_func": 0.08333333333333333,
+      "reward": 1.873472045768391,
+      "reward_std": 0.48061720840632915,
+      "rewards/concensus_correctness_reward_func": 0.15699999982660467,
+      "rewards/consensus_reward_func": 0.09090909090909091,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.3333333333333333,
-      "rewards/question_recreation_reward_func": 0.5784205005814632,
+      "rewards/final_correctness_reward_func": 0.2727272727272727,
+      "rewards/question_recreation_reward_func": 0.7095629688013684,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.8255416651566824,
+      "rewards/xmlcount_reward_func": 0.643272725018588,
       "step": 8
     },
     {
-      "completion_length": 225.03125,
-      "epoch": 2.571428571428571,
-      "grad_norm": 8.09153938293457,
-      "kl": 0.007186611874203663,
+      "completion_length": 158.59375,
+      "epoch": 1.4705882352941178,
+      "grad_norm": 6.622710227966309,
+      "kl": 0.010859893664019182,
       "learning_rate": 3.1137137178519977e-07,
       "loss": 0.0,
-      "reward": 1.4572835750877857,
-      "reward_std": 1.1924903796752915,
-      "rewards/concensus_correctness_reward_func": 0.12212499976158142,
-      "rewards/consensus_reward_func": 0.1875,
+      "reward": 3.0381277231499553,
+      "reward_std": 0.5335058589698747,
+      "rewards/concensus_correctness_reward_func": 1.4662500023841858,
+      "rewards/consensus_reward_func": 0.4375,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0625,
-      "rewards/question_recreation_reward_func": 0.3975648246705532,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.36175276967696846,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.6875937469303608,
+      "rewards/xmlcount_reward_func": 0.77262500859797,
       "step": 10
     },
     {
-      "completion_length": 167.0,
-      "epoch": 3.0,
-      "grad_norm": 4.35850715637207,
-      "kl": 0.009038338272754723,
+      "completion_length": 283.375,
+      "epoch": 1.784313725490196,
+      "grad_norm": 5.638906002044678,
+      "kl": 0.007406034710584208,
       "learning_rate": 2.2935516363191693e-07,
       "loss": 0.0,
-      "reward": 2.172236522038778,
-      "reward_std": 1.3352671524820228,
-      "rewards/concensus_correctness_reward_func": 0.32199999690055847,
-      "rewards/consensus_reward_func": 0.25,
+      "reward": 1.6740143587812781,
+      "reward_std": 1.191217671148479,
+      "rewards/concensus_correctness_reward_func": 0.1471250019967556,
+      "rewards/consensus_reward_func": 0.1875,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.25,
-      "rewards/question_recreation_reward_func": 0.5439448431134224,
+      "rewards/question_recreation_reward_func": 0.48260811483487487,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.8062916770577431,
+      "rewards/xmlcount_reward_func": 0.6067812591791153,
       "step": 12
     },
     {
-      "completion_length": 176.78125,
-      "epoch": 3.571428571428571,
-      "grad_norm": 6.705428123474121,
-      "kl": 0.0063926830916898325,
+      "completion_length": 249.0,
+      "epoch": 2.0,
+      "grad_norm": 3.3914954662323,
+      "kl": 0.011619402540170333,
       "learning_rate": 1.4957614383675767e-07,
       "loss": 0.0,
-      "reward": 2.621943838894367,
-      "reward_std": 1.5366092347539961,
-      "rewards/concensus_correctness_reward_func": 0.43143749982118607,
-      "rewards/consensus_reward_func": 0.4375,
+      "reward": 4.049619214101271,
+      "reward_std": 2.009789251807061,
+      "rewards/concensus_correctness_reward_func": 2.756545454263687,
+      "rewards/consensus_reward_func": 0.09090909090909091,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.25,
-      "rewards/question_recreation_reward_func": 0.7003813628107309,
+      "rewards/final_correctness_reward_func": 0.09090909090909091,
+      "rewards/question_recreation_reward_func": 0.4135735107755119,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.8026250153779984,
+      "rewards/strict_format_reward_func": 0.022727272727272728,
+      "rewards/xmlcount_reward_func": 0.6749545498327776,
       "step": 14
     },
     {
-      "completion_length": 205.91666666666666,
-      "epoch": 4.0,
-      "grad_norm": 3.3670480251312256,
-      "kl": 0.007393967578536831,
+      "completion_length": 223.875,
+      "epoch": 2.313725490196078,
+      "grad_norm": 8.215188026428223,
+      "kl": 0.014186236599925905,
       "learning_rate": 8.067960709356478e-08,
       "loss": 0.0,
-      "reward": 1.4719673146804173,
-      "reward_std": 0.7381301919619242,
-      "rewards/concensus_correctness_reward_func": 0.0742499977350235,
-      "rewards/consensus_reward_func": 0.08333333333333333,
+      "reward": 2.642803938128054,
+      "reward_std": 1.4382841207552701,
+      "rewards/concensus_correctness_reward_func": 0.8331250017508864,
+      "rewards/consensus_reward_func": 0.25,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.571050651371479,
+      "rewards/final_correctness_reward_func": 0.3125,
+      "rewards/question_recreation_reward_func": 0.4023352946387604,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.7433333223064741,
+      "rewards/strict_format_reward_func": 0.015625,
+      "rewards/xmlcount_reward_func": 0.829218752682209,
       "step": 16
     },
     {
-      "completion_length": 174.0,
-      "epoch": 4.571428571428571,
-      "grad_norm": 8.952384948730469,
-      "kl": 0.0078432568552671,
+      "completion_length": 247.875,
+      "epoch": 2.627450980392157,
+      "grad_norm": 7.052781105041504,
+      "kl": 0.015200054476736113,
       "learning_rate": 3.013156219837776e-08,
       "loss": 0.0,
-      "reward": 2.2049832567572594,
-      "reward_std": 1.262722612824291,
-      "rewards/concensus_correctness_reward_func": 0.35725000500679016,
+      "reward": 1.8882133960723877,
+      "reward_std": 1.274847088381648,
+      "rewards/concensus_correctness_reward_func": 0.06793750077486038,
       "rewards/consensus_reward_func": 0.25,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.25,
-      "rewards/question_recreation_reward_func": 0.5689207338728011,
+      "rewards/final_correctness_reward_func": 0.3125,
+      "rewards/question_recreation_reward_func": 0.5125571433454752,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.7788124959915876,
+      "rewards/strict_format_reward_func": 0.046875,
+      "rewards/xmlcount_reward_func": 0.698343763127923,
       "step": 18
     },
     {
-      "completion_length": 207.25,
-      "epoch": 5.0,
-      "grad_norm": 4.44536828994751,
-      "kl": 0.00663318302637587,
+      "completion_length": 253.28125,
+      "epoch": 2.9411764705882355,
+      "grad_norm": 7.9368438720703125,
+      "kl": 0.013815053927828558,
       "learning_rate": 3.4096741493194193e-09,
       "loss": 0.0,
-      "reward": 1.7914074063301086,
-      "reward_std": 0.7179869742443165,
-      "rewards/concensus_correctness_reward_func": 0.08216666678587596,
-      "rewards/consensus_reward_func": 0.08333333333333333,
+      "reward": 3.0817041154950857,
+      "reward_std": 2.74389247270301,
+      "rewards/concensus_correctness_reward_func": 1.3802500003948808,
+      "rewards/consensus_reward_func": 0.25,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.16666666666666666,
-      "rewards/question_recreation_reward_func": 0.5659907013177872,
+      "rewards/final_correctness_reward_func": 0.1875,
+      "rewards/question_recreation_reward_func": 0.48770411510486156,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.8932499984900156,
+      "rewards/xmlcount_reward_func": 0.7762499954551458,
       "step": 20
     },
     {
-      "epoch": 5.0,
+      "epoch": 2.9411764705882355,
       "step": 20,
       "total_flos": 0.0,
-      "train_loss": 4.6607106924057e-06,
-      "train_runtime": 528.4481,
-      "train_samples_per_second": 0.606,
-      "train_steps_per_second": 0.038
+      "train_loss": 7.716044910921482e-06,
+      "train_runtime": 943.9196,
+      "train_samples_per_second": 0.339,
+      "train_steps_per_second": 0.021
     }
   ],
   "logging_steps": 2,
   "max_steps": 20,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 7,
+  "num_train_epochs": 4,
   "save_steps": 25,
   "stateful_callbacks": {
     "TrainerControl": {