End of training

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 5.803685635328293e-05,
-    "train_runtime": 278.3631,
+    "train_loss": 4.274839545814757e-06,
+    "train_runtime": 281.202,
     "train_samples": 28,
-    "train_samples_per_second": 1.15,
-    "train_steps_per_second": 0.072
+    "train_samples_per_second": 1.138,
+    "train_steps_per_second": 0.071
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ca5c79c10ea549f21c7db2bf3224c741e3b7da826f35e0683051254345d2fd00
+oid sha256:e9b4de44b22d04b34afb0bf84ca06c2c505ce2b683b7ba897774227d4fb84be5
 size 1976163472

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 5.803685635328293e-05,
-    "train_runtime": 278.3631,
+    "train_loss": 4.274839545814757e-06,
+    "train_runtime": 281.202,
     "train_samples": 28,
-    "train_samples_per_second": 1.15,
-    "train_steps_per_second": 0.072
+    "train_samples_per_second": 1.138,
+    "train_steps_per_second": 0.071
 }

trainer_state.json

@@ -10,203 +10,203 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 280.1875,
+      "completion_length": 350.25,
       "epoch": 0.5714285714285714,
-      "grad_norm": 76.21530151367188,
+      "grad_norm": 6.531829833984375,
       "kl": 0.0,
       "learning_rate": 5e-07,
-      "loss": 0.0,
-      "reward": 1.1465849268715829,
-      "reward_std": 1.4973960118368268,
-      "rewards/concensus_correctness_reward_func": 0.625,
+      "loss": -0.0,
+      "reward": 0.4527950035408139,
+      "reward_std": 0.29492819524602965,
+      "rewards/concensus_correctness_reward_func": 0.0078125,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.125,
-      "rewards/question_recreation_reward_func": 0.32024113717488945,
+      "rewards/final_correctness_reward_func": 0.0625,
+      "rewards/question_recreation_reward_func": 0.23382625449448824,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.07634375477209687,
+      "rewards/xmlcount_reward_func": 0.1486562481150031,
       "step": 2
     },
     {
-      "completion_length": 374.375,
+      "completion_length": 381.625,
       "epoch": 1.0,
-      "grad_norm": 2.7701170444488525,
-      "kl": 0.0013190652583337699,
+      "grad_norm": 4.339130878448486,
+      "kl": 0.0009718153305584565,
       "learning_rate": 4.864543104251586e-07,
       "loss": 0.0,
-      "reward": 0.5202060750064751,
-      "reward_std": 0.5441535848658532,
-      "rewards/concensus_correctness_reward_func": 0.018833333005507786,
+      "reward": 0.008616745471954346,
+      "reward_std": 0.3103340204882746,
+      "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.08333333333333333,
-      "rewards/question_recreation_reward_func": 0.21845607459545135,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.16345008462667465,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.1995833346930643,
+      "rewards/xmlcount_reward_func": -0.1548333317041397,
       "step": 4
     },
     {
-      "completion_length": 322.15625,
+      "completion_length": 353.1875,
       "epoch": 1.5714285714285714,
-      "grad_norm": 64.42672729492188,
-      "kl": 0.07032692650682293,
+      "grad_norm": 86.73339080810547,
+      "kl": 0.0029522667755372822,
       "learning_rate": 4.472851273490984e-07,
-      "loss": 0.0001,
-      "reward": 0.3091004164889455,
-      "reward_std": 0.5367216910090065,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0,
+      "loss": 0.0,
+      "reward": 1.8719191025011241,
+      "reward_std": 0.7418497779872268,
+      "rewards/concensus_correctness_reward_func": 1.2578125,
+      "rewards/consensus_reward_func": 0.0625,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.25581916980445385,
+      "rewards/final_correctness_reward_func": 0.25,
+      "rewards/question_recreation_reward_func": 0.30138792924117297,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.05328125832602382,
+      "rewards/xmlcount_reward_func": 0.00021875090897083282,
       "step": 6
     },
     {
-      "completion_length": 374.7916666666667,
+      "completion_length": 328.125,
       "epoch": 2.0,
-      "grad_norm": 4.210530757904053,
-      "kl": 0.0011358094052411616,
+      "grad_norm": 8.744457244873047,
+      "kl": 0.0016929735332572211,
       "learning_rate": 3.867370395306068e-07,
       "loss": 0.0,
-      "reward": 0.28024414647370577,
-      "reward_std": 0.35849861746343475,
-      "rewards/concensus_correctness_reward_func": 0.03608333319425583,
-      "rewards/consensus_reward_func": 0.08333333333333333,
+      "reward": 0.20979913541426262,
+      "reward_std": 0.4416549120408793,
+      "rewards/concensus_correctness_reward_func": 0.0,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.08333333333333333,
-      "rewards/question_recreation_reward_func": 0.18024415026108423,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.26946580099562806,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": -0.10275000085433324,
+      "rewards/xmlcount_reward_func": -0.059666667288790144,
       "step": 8
     },
     {
-      "completion_length": 376.40625,
+      "completion_length": 315.6875,
       "epoch": 2.571428571428571,
-      "grad_norm": 7.121423721313477,
-      "kl": 0.0020210541697451845,
+      "grad_norm": 14.717121124267578,
+      "kl": 0.011745726533263223,
       "learning_rate": 3.1137137178519977e-07,
       "loss": 0.0,
-      "reward": 0.1576997386291623,
-      "reward_std": 0.5325253210030496,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0,
+      "reward": 0.3240256152348593,
+      "reward_std": 0.5225911034503952,
+      "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": 0.21501224458916113,
+      "rewards/final_correctness_reward_func": 0.125,
+      "rewards/question_recreation_reward_func": 0.23383812047541142,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": -0.057312499498948455,
+      "rewards/xmlcount_reward_func": -0.10512499790638685,
       "step": 10
     },
     {
-      "completion_length": 381.8333333333333,
+      "completion_length": 382.0416666666667,
       "epoch": 3.0,
-      "grad_norm": 2.8112943172454834,
-      "kl": 0.5838237675246395,
+      "grad_norm": 4.500463485717773,
+      "kl": 0.0023677585510692247,
       "learning_rate": 2.2935516363191693e-07,
-      "loss": 0.0004,
-      "reward": 0.5091939164946476,
-      "reward_std": 0.3211197915176551,
-      "rewards/concensus_correctness_reward_func": 0.0,
+      "loss": 0.0,
+      "reward": 0.27927784745891887,
+      "reward_std": 0.48502769274637103,
+      "rewards/concensus_correctness_reward_func": 0.022999999423821766,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.25,
-      "rewards/question_recreation_reward_func": 0.20627725295101604,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.24306950687120357,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.05291666463017464,
+      "rewards/xmlcount_reward_func": 0.013208337128162384,
       "step": 12
     },
     {
-      "completion_length": 373.9375,
+      "completion_length": 322.6875,
       "epoch": 3.571428571428571,
-      "grad_norm": 5.300664901733398,
-      "kl": 0.06384690720915387,
+      "grad_norm": 48.863433837890625,
+      "kl": 0.0057819957146421075,
       "learning_rate": 1.4957614383675767e-07,
-      "loss": 0.0001,
-      "reward": 0.4644634025171399,
-      "reward_std": 0.6556758279912174,
-      "rewards/concensus_correctness_reward_func": 0.04118749964982271,
-      "rewards/consensus_reward_func": 0.0625,
+      "loss": 0.0,
+      "reward": 0.24080068030161783,
+      "reward_std": 0.41197579625077196,
+      "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": 0.27574466867372394,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.2828319200198166,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.022531244670972228,
+      "rewards/xmlcount_reward_func": -0.04203125135973096,
       "step": 14
     },
     {
-      "completion_length": 364.1666666666667,
+      "completion_length": 397.3333333333333,
       "epoch": 4.0,
-      "grad_norm": 4.4072771072387695,
-      "kl": 0.001848986047358873,
+      "grad_norm": 32.489540100097656,
+      "kl": 0.0022319424897432327,
       "learning_rate": 8.067960709356478e-08,
       "loss": 0.0,
-      "reward": 0.4843742731027305,
-      "reward_std": 0.31728264258708805,
-      "rewards/concensus_correctness_reward_func": 0.010416666666666666,
+      "reward": 1.1202503452077508,
+      "reward_std": 1.5144341605870675,
+      "rewards/concensus_correctness_reward_func": 0.8333333333333334,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.08333333333333333,
-      "rewards/question_recreation_reward_func": 0.2724159355275333,
+      "rewards/final_correctness_reward_func": 0.16666666666666666,
+      "rewards/question_recreation_reward_func": 0.2893753442913294,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.1182083406408007,
+      "rewards/xmlcount_reward_func": -0.16912500405063233,
       "step": 16
     },
     {
-      "completion_length": 409.5625,
+      "completion_length": 442.15625,
       "epoch": 4.571428571428571,
-      "grad_norm": 5.115142822265625,
-      "kl": 0.0018609182243380928,
+      "grad_norm": 43.90248107910156,
+      "kl": 0.015148180060350569,
       "learning_rate": 3.013156219837776e-08,
       "loss": 0.0,
-      "reward": 0.18973279604688287,
-      "reward_std": 0.560065383091569,
-      "rewards/concensus_correctness_reward_func": 0.0,
-      "rewards/consensus_reward_func": 0.0,
+      "reward": 0.34479564492357895,
+      "reward_std": 0.4152768463245593,
+      "rewards/concensus_correctness_reward_func": 0.0078125,
+      "rewards/consensus_reward_func": 0.0625,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.125,
-      "rewards/question_recreation_reward_func": 0.2575453120516613,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.2232956478255801,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": -0.1928125019185245,
+      "rewards/xmlcount_reward_func": 0.05118750053225085,
       "step": 18
     },
     {
-      "completion_length": 374.9166666666667,
+      "completion_length": 396.7083333333333,
       "epoch": 5.0,
-      "grad_norm": 5.7314982414245605,
-      "kl": 0.0016712552460376173,
+      "grad_norm": 2.7734463214874268,
+      "kl": 0.002418578983148715,
       "learning_rate": 3.4096741493194193e-09,
       "loss": 0.0,
-      "reward": 0.31323465270300704,
-      "reward_std": 0.331939674836273,
-      "rewards/concensus_correctness_reward_func": 0.0,
+      "reward": 0.3893080285439889,
+      "reward_std": 1.1155760533486803,
+      "rewards/concensus_correctness_reward_func": 0.0503333310286204,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.2745679809401433,
+      "rewards/final_correctness_reward_func": 0.25,
+      "rewards/question_recreation_reward_func": 0.266849700982372,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.038666668658455215,
+      "rewards/xmlcount_reward_func": -0.17787500098347664,
       "step": 20
     },
     {
       "epoch": 5.0,
       "step": 20,
       "total_flos": 0.0,
-      "train_loss": 5.803685635328293e-05,
-      "train_runtime": 278.3631,
-      "train_samples_per_second": 1.15,
-      "train_steps_per_second": 0.072
+      "train_loss": 4.274839545814757e-06,
+      "train_runtime": 281.202,
+      "train_samples_per_second": 1.138,
+      "train_steps_per_second": 0.071
     }
   ],
   "logging_steps": 2,