End of training

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.00039611602551303804,
-    "train_runtime": 242.2421,
+    "train_loss": 8.675511101926531,
+    "train_runtime": 192.5653,
     "train_samples": 28,
-    "train_samples_per_second": 1.321,
-    "train_steps_per_second": 0.083
+    "train_samples_per_second": 1.662,
+    "train_steps_per_second": 0.104
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8e72770482bb761fd6372292fc12a68c23a73d93e4f464073bf0ec3bad48edb7
+oid sha256:6fb35785199e6d78014b99fb5d1cacb5f3b4c9b77b521bf7559aeb5b9de82413
 size 1976163472

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.00039611602551303804,
-    "train_runtime": 242.2421,
+    "train_loss": 8.675511101926531,
+    "train_runtime": 192.5653,
     "train_samples": 28,
-    "train_samples_per_second": 1.321,
-    "train_steps_per_second": 0.083
+    "train_samples_per_second": 1.662,
+    "train_steps_per_second": 0.104
 }

trainer_state.json

@@ -10,203 +10,203 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 320.90625,
+      "completion_length": 237.09375,
       "epoch": 0.5714285714285714,
-      "grad_norm": 14.3243408203125,
+      "grad_norm": 23.86282730102539,
       "kl": 0.0,
       "learning_rate": 5e-07,
       "loss": 0.0,
-      "reward": 1.8238091692328453,
-      "reward_std": 1.0591860833810642,
-      "rewards/concensus_correctness_reward_func": 0.25181249529123306,
-      "rewards/consensus_reward_func": 0.375,
+      "reward": 2.320491649210453,
+      "reward_std": 1.035187211818993,
+      "rewards/concensus_correctness_reward_func": 0.31562499701976776,
+      "rewards/consensus_reward_func": 0.4375,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0625,
-      "rewards/question_recreation_reward_func": 0.5330279082991183,
+      "rewards/final_correctness_reward_func": 0.125,
+      "rewards/question_recreation_reward_func": 0.6428666170686483,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.6014687430579215,
+      "rewards/xmlcount_reward_func": 0.7994999922811985,
       "step": 2
     },
     {
-      "completion_length": 394.8333333333333,
+      "completion_length": 202.75,
       "epoch": 1.0,
-      "grad_norm": 6.504026889801025,
-      "kl": 0.00567529887969916,
+      "grad_norm": 20.716642379760742,
+      "kl": 0.3580615110695362,
       "learning_rate": 4.864543104251586e-07,
-      "loss": 0.0,
-      "reward": 2.645399361848831,
-      "reward_std": 1.4961872142351542,
-      "rewards/concensus_correctness_reward_func": 0.5172500063975652,
-      "rewards/consensus_reward_func": 0.5,
+      "loss": 0.0003,
+      "reward": 2.4445405304431915,
+      "reward_std": 0.7943247662236294,
+      "rewards/concensus_correctness_reward_func": 0.31799999872843426,
+      "rewards/consensus_reward_func": 0.25,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.08333333333333333,
-      "rewards/question_recreation_reward_func": 0.740566020210584,
+      "rewards/question_recreation_reward_func": 0.7158738623062769,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.8042500068744024,
+      "rewards/xmlcount_reward_func": 1.0773333460092545,
       "step": 4
     },
     {
-      "completion_length": 264.84375,
+      "completion_length": 236.0,
       "epoch": 1.5714285714285714,
-      "grad_norm": 13.624690055847168,
-      "kl": 0.09184227080550045,
+      "grad_norm": 17.435176849365234,
+      "kl": 0.07964501332025975,
       "learning_rate": 4.472851273490984e-07,
       "loss": 0.0001,
-      "reward": 2.3658637814223766,
-      "reward_std": 0.9425615618820302,
-      "rewards/concensus_correctness_reward_func": 0.4235000014305115,
-      "rewards/consensus_reward_func": 0.375,
+      "reward": 3.0734478905797005,
+      "reward_std": 1.2019765962613747,
+      "rewards/concensus_correctness_reward_func": 0.818124994635582,
+      "rewards/consensus_reward_func": 0.625,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0625,
-      "rewards/question_recreation_reward_func": 0.6558637763373554,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.6496978905051947,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.8489999696612358,
+      "rewards/xmlcount_reward_func": 0.9806249961256981,
       "step": 6
     },
     {
-      "completion_length": 363.25,
+      "completion_length": 218.04166666666666,
       "epoch": 2.0,
-      "grad_norm": 11.709728240966797,
-      "kl": 0.03270413619854177,
+      "grad_norm": 10.406091690063477,
+      "kl": 81480.78907666805,
       "learning_rate": 3.867370395306068e-07,
-      "loss": 0.0,
-      "reward": 3.828521261612574,
-      "reward_std": 2.0679361606016755,
-      "rewards/concensus_correctness_reward_func": 1.4121666252613068,
-      "rewards/consensus_reward_func": 0.75,
+      "loss": 61.1106,
+      "reward": 3.46412992477417,
+      "reward_std": 2.2507151973744235,
+      "rewards/concensus_correctness_reward_func": 1.1319166670242946,
+      "rewards/consensus_reward_func": 0.4166666666666667,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.25,
-      "rewards/question_recreation_reward_func": 0.6656045454243819,
+      "rewards/final_correctness_reward_func": 0.08333333333333333,
+      "rewards/question_recreation_reward_func": 0.746754934390386,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.020833333333333332,
-      "rewards/xmlcount_reward_func": 0.7299166768789291,
+      "rewards/strict_format_reward_func": 0.041666666666666664,
+      "rewards/xmlcount_reward_func": 1.0437916765610378,
       "step": 8
     },
     {
-      "completion_length": 303.125,
+      "completion_length": 222.4375,
       "epoch": 2.571428571428571,
-      "grad_norm": 12.579998970031738,
-      "kl": 0.07898051122901961,
+      "grad_norm": 14.132076263427734,
+      "kl": 0.5107612958818208,
       "learning_rate": 3.1137137178519977e-07,
-      "loss": 0.0001,
-      "reward": 1.8908442594110966,
-      "reward_std": 1.1686942288652062,
-      "rewards/concensus_correctness_reward_func": 0.28337499499320984,
-      "rewards/consensus_reward_func": 0.3125,
+      "loss": 0.0005,
+      "reward": 3.7268150821328163,
+      "reward_std": 1.5598002884944435,
+      "rewards/concensus_correctness_reward_func": 0.9949999861419201,
+      "rewards/consensus_reward_func": 0.8125,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.527281790971756,
+      "rewards/final_correctness_reward_func": 0.1875,
+      "rewards/question_recreation_reward_func": 0.6768150591524318,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.7676874920725822,
+      "rewards/strict_format_reward_func": 0.015625,
+      "rewards/xmlcount_reward_func": 1.0393749959766865,
       "step": 10
     },
     {
-      "completion_length": 304.625,
+      "completion_length": 205.75,
       "epoch": 3.0,
-      "grad_norm": 6.953789234161377,
-      "kl": 0.07293941577275594,
+      "grad_norm": 8.287981033325195,
+      "kl": 0.11747636886623998,
       "learning_rate": 2.2935516363191693e-07,
       "loss": 0.0001,
-      "reward": 3.5938471953074136,
-      "reward_std": 2.2706022734443345,
-      "rewards/concensus_correctness_reward_func": 0.7098333388566971,
-      "rewards/consensus_reward_func": 0.75,
+      "reward": 4.006574511528015,
+      "reward_std": 1.4391742528726656,
+      "rewards/concensus_correctness_reward_func": 0.9271666631102562,
+      "rewards/consensus_reward_func": 1.0833333333333333,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.4166666666666667,
-      "rewards/question_recreation_reward_func": 0.8268054773410162,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.8568661361932755,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.890541652838389,
+      "rewards/xmlcount_reward_func": 1.1392083217700322,
       "step": 12
     },
     {
-      "completion_length": 256.96875,
+      "completion_length": 273.4375,
       "epoch": 3.571428571428571,
-      "grad_norm": 16.279869079589844,
-      "kl": 0.7254527114564553,
+      "grad_norm": 12.828157424926758,
+      "kl": 0.5077391383820213,
       "learning_rate": 1.4957614383675767e-07,
-      "loss": 0.0007,
-      "reward": 3.9172775223851204,
-      "reward_std": 1.1482481649145484,
-      "rewards/concensus_correctness_reward_func": 0.9585625063627958,
-      "rewards/consensus_reward_func": 0.8125,
+      "loss": 0.0005,
+      "reward": 3.0370506569743156,
+      "reward_std": 1.9715431984513998,
+      "rewards/concensus_correctness_reward_func": 0.613062497228384,
+      "rewards/consensus_reward_func": 0.6875,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.375,
-      "rewards/question_recreation_reward_func": 0.7574650766327977,
+      "rewards/final_correctness_reward_func": 0.1875,
+      "rewards/question_recreation_reward_func": 0.6411756332963705,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 1.0137499924749136,
+      "rewards/xmlcount_reward_func": 0.9078125022351742,
       "step": 14
     },
     {
-      "completion_length": 308.25,
+      "completion_length": 249.08333333333334,
       "epoch": 4.0,
-      "grad_norm": 11.026092529296875,
-      "kl": 0.3210177328437567,
+      "grad_norm": 12.552739143371582,
+      "kl": 34149.09689274756,
       "learning_rate": 8.067960709356478e-08,
-      "loss": 0.0002,
-      "reward": 2.663438856601715,
-      "reward_std": 1.4866043164705236,
-      "rewards/concensus_correctness_reward_func": 0.5567499945561091,
-      "rewards/consensus_reward_func": 0.5833333333333334,
+      "loss": 25.6118,
+      "reward": 3.7040178924798965,
+      "reward_std": 1.3390946853905916,
+      "rewards/concensus_correctness_reward_func": 0.9702499856551489,
+      "rewards/consensus_reward_func": 0.8333333333333334,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.7065221890807152,
+      "rewards/final_correctness_reward_func": 0.25,
+      "rewards/question_recreation_reward_func": 0.6871429421007633,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.8168333073457082,
+      "rewards/xmlcount_reward_func": 0.9632916674017906,
       "step": 16
     },
     {
-      "completion_length": 268.71875,
+      "completion_length": 212.78125,
       "epoch": 4.571428571428571,
-      "grad_norm": 18.54529571533203,
-      "kl": 2.3130286334780976,
+      "grad_norm": 16.37578773498535,
+      "kl": 19.238060696865432,
       "learning_rate": 3.013156219837776e-08,
-      "loss": 0.0023,
-      "reward": 3.231190398335457,
-      "reward_std": 0.8707997768069617,
-      "rewards/concensus_correctness_reward_func": 0.48849999718368053,
-      "rewards/consensus_reward_func": 0.625,
+      "loss": 0.0192,
+      "reward": 3.540576569736004,
+      "reward_std": 1.7284664756152779,
+      "rewards/concensus_correctness_reward_func": 0.9062499925494194,
+      "rewards/consensus_reward_func": 0.8125,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.3125,
-      "rewards/question_recreation_reward_func": 0.7732217563316226,
+      "rewards/final_correctness_reward_func": 0.1875,
+      "rewards/question_recreation_reward_func": 0.6683264966122806,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 1.031968742609024,
+      "rewards/xmlcount_reward_func": 0.96599998511374,
       "step": 18
     },
     {
-      "completion_length": 267.5,
+      "completion_length": 242.83333333333334,
       "epoch": 5.0,
-      "grad_norm": 11.74589729309082,
-      "kl": 0.5716485911980271,
+      "grad_norm": 244.17030334472656,
+      "kl": 15.989224990286553,
       "learning_rate": 3.4096741493194193e-09,
-      "loss": 0.0004,
-      "reward": 2.583486075202624,
-      "reward_std": 0.9447366874665022,
-      "rewards/concensus_correctness_reward_func": 0.415583332379659,
+      "loss": 0.012,
+      "reward": 3.2668827176094055,
+      "reward_std": 1.63679713383317,
+      "rewards/concensus_correctness_reward_func": 0.7084999978542328,
       "rewards/consensus_reward_func": 0.4166666666666667,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.08333333333333333,
-      "rewards/question_recreation_reward_func": 0.654319416731596,
+      "rewards/final_correctness_reward_func": 0.25,
+      "rewards/question_recreation_reward_func": 0.8270076736807823,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 1.013583317399025,
+      "rewards/xmlcount_reward_func": 1.0647083222866058,
       "step": 20
     },
     {
       "epoch": 5.0,
       "step": 20,
       "total_flos": 0.0,
-      "train_loss": 0.00039611602551303804,
-      "train_runtime": 242.2421,
-      "train_samples_per_second": 1.321,
-      "train_steps_per_second": 0.083
+      "train_loss": 8.675511101926531,
+      "train_runtime": 192.5653,
+      "train_samples_per_second": 1.662,
+      "train_steps_per_second": 0.104
     }
   ],
   "logging_steps": 2,