End of training

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 3.0612738328272826e-05,
-    "train_runtime": 915.7967,
+    "train_loss": 0.00011620071672950872,
+    "train_runtime": 657.8689,
     "train_samples": 160,
-    "train_samples_per_second": 0.349,
-    "train_steps_per_second": 0.022
+    "train_samples_per_second": 0.486,
+    "train_steps_per_second": 0.03
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3ae2fadc842ea4e318a3f0eee5416d82c0634d1218d1b4dc876a8027575ba18c
+oid sha256:e5a228fb920cc8ae208a5f1952a082f0345eb795d272b028e97c7a6eb310128a
 size 1976163472

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 3.0612738328272826e-05,
-    "train_runtime": 915.7967,
+    "train_loss": 0.00011620071672950872,
+    "train_runtime": 657.8689,
     "train_samples": 160,
-    "train_samples_per_second": 0.349,
-    "train_steps_per_second": 0.022
+    "train_samples_per_second": 0.486,
+    "train_steps_per_second": 0.03
 }

trainer_state.json

@@ -10,203 +10,203 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 237.1875,
+      "completion_length": 188.875,
       "epoch": 0.1,
-      "grad_norm": 8.398701667785645,
-      "kl": 0.0,
+      "grad_norm": 20.606088638305664,
+      "kl": -1.8626451769865326e-10,
       "learning_rate": 4.965903258506806e-07,
-      "loss": 0.0,
-      "reward": 0.09728920509223826,
-      "reward_std": 0.05032656194816809,
+      "loss": -0.0,
+      "reward": -0.08459861180745065,
+      "reward_std": 0.4505786441732198,
       "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.09338295520865358,
+      "rewards/final_correctness_reward_func": 0.0625,
+      "rewards/question_recreation_reward_func": 0.08155764313414693,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.00390625,
+      "rewards/xmlcount_reward_func": -0.22865625121630728,
       "step": 2
     },
     {
-      "completion_length": 254.8125,
+      "completion_length": 159.0,
       "epoch": 0.2,
-      "grad_norm": 4.527547359466553,
-      "kl": 0.0005934742662248027,
+      "grad_norm": 17.4925537109375,
+      "kl": 0.012501606208388694,
       "learning_rate": 4.698684378016222e-07,
       "loss": 0.0,
-      "reward": 0.10370710026472807,
-      "reward_std": 0.056229994814202655,
+      "reward": 0.15525958803482354,
+      "reward_std": 0.13993048649354023,
       "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.10933210048824549,
+      "rewards/final_correctness_reward_func": 0.0625,
+      "rewards/question_recreation_reward_func": 0.10338459047488868,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": -0.005625000223517418,
+      "rewards/xmlcount_reward_func": -0.010625000111758709,
       "step": 4
     },
     {
-      "completion_length": 198.375,
+      "completion_length": 103.90625,
       "epoch": 0.3,
-      "grad_norm": 16.620403289794922,
-      "kl": 0.011071403857840778,
+      "grad_norm": 44.36183547973633,
+      "kl": 0.116791727632517,
       "learning_rate": 4.193203929064353e-07,
-      "loss": 0.0,
-      "reward": 0.17157541977940127,
-      "reward_std": 0.12756577530672075,
+      "loss": 0.0001,
+      "reward": 0.09123604014166631,
+      "reward_std": 0.024207589820434805,
       "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": 0.1181379213812761,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.09123604014166631,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": -0.009062499972060323,
+      "rewards/xmlcount_reward_func": 0.0,
       "step": 6
     },
     {
-      "completion_length": 151.34375,
+      "completion_length": 65.5625,
       "epoch": 0.4,
-      "grad_norm": 5.440056324005127,
-      "kl": 0.0035489053407218307,
+      "grad_norm": 23.13948631286621,
+      "kl": 0.14583962078904733,
       "learning_rate": 3.5042385616324236e-07,
-      "loss": 0.0,
-      "reward": 0.12494469387456775,
-      "reward_std": 0.02955692503019236,
+      "loss": 0.0001,
+      "reward": 0.10612659784965217,
+      "reward_std": 0.07028272053867113,
       "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.12494469387456775,
+      "rewards/question_recreation_reward_func": 0.0766265980200842,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.02949999994598329,
       "step": 8
     },
     {
-      "completion_length": 159.4375,
+      "completion_length": 125.875,
       "epoch": 0.5,
-      "grad_norm": 17.10235595703125,
-      "kl": 0.019251713070843834,
+      "grad_norm": 18.663105010986328,
+      "kl": 0.2906235277914675,
       "learning_rate": 2.706448363680831e-07,
-      "loss": 0.0,
-      "reward": 0.2264014035463333,
-      "reward_std": 0.1991720616351813,
-      "rewards/concensus_correctness_reward_func": 0.020625000819563866,
-      "rewards/consensus_reward_func": 0.0625,
+      "loss": 0.0003,
+      "reward": 0.05466344498563558,
+      "reward_std": 0.022103407893155236,
+      "rewards/concensus_correctness_reward_func": 0.0,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.12343264813534915,
+      "rewards/question_recreation_reward_func": 0.05138219508808106,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.019843749701976776,
+      "rewards/xmlcount_reward_func": 0.003281249897554517,
       "step": 10
     },
     {
-      "completion_length": 161.6875,
+      "completion_length": 81.15625,
       "epoch": 0.6,
-      "grad_norm": 13.440507888793945,
-      "kl": 0.01697970710665686,
+      "grad_norm": 17.701265335083008,
+      "kl": 0.08585882172337733,
       "learning_rate": 1.886286282148002e-07,
-      "loss": 0.0,
-      "reward": 0.30120708944741637,
-      "reward_std": 0.30583049908454996,
-      "rewards/concensus_correctness_reward_func": 0.0364375002682209,
-      "rewards/consensus_reward_func": 0.125,
+      "loss": 0.0001,
+      "reward": 0.06056131387595087,
+      "reward_std": 0.0290565535397036,
+      "rewards/concensus_correctness_reward_func": 0.0,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.14223835291340947,
+      "rewards/question_recreation_reward_func": 0.05399881454650313,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": -0.0024687498807907104,
+      "rewards/xmlcount_reward_func": 0.006562499795109034,
       "step": 12
     },
     {
-      "completion_length": 142.25,
+      "completion_length": 103.40625,
       "epoch": 0.7,
-      "grad_norm": 24.540307998657227,
-      "kl": 0.0875492471823236,
+      "grad_norm": 33.02743148803711,
+      "kl": 0.103502563375514,
       "learning_rate": 1.1326296046939333e-07,
       "loss": 0.0001,
-      "reward": 0.5317345891380683,
-      "reward_std": 0.23181627465237398,
-      "rewards/concensus_correctness_reward_func": 0.04262499883770943,
-      "rewards/consensus_reward_func": 0.1875,
+      "reward": 0.12133114249445498,
+      "reward_std": 0.1054717963852454,
+      "rewards/concensus_correctness_reward_func": 0.0,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.1875,
-      "rewards/question_recreation_reward_func": 0.0995783495309297,
+      "rewards/final_correctness_reward_func": 0.0625,
+      "rewards/question_recreation_reward_func": 0.05883114255266264,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.014531250111758709,
+      "rewards/xmlcount_reward_func": 0.0,
       "step": 14
     },
     {
-      "completion_length": 130.9375,
+      "completion_length": 44.375,
       "epoch": 0.8,
-      "grad_norm": 24.440187454223633,
-      "kl": 0.062173418860766105,
+      "grad_norm": 30.598201751708984,
+      "kl": 0.15886804103502072,
       "learning_rate": 5.271487265090163e-08,
-      "loss": 0.0001,
-      "reward": 0.5183445231523365,
-      "reward_std": 0.6032340360834496,
-      "rewards/concensus_correctness_reward_func": 0.13956249877810478,
-      "rewards/consensus_reward_func": 0.1875,
+      "loss": 0.0002,
+      "reward": 0.3126172423362732,
+      "reward_std": 0.285806781725114,
+      "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": 0.14553201623493806,
+      "rewards/final_correctness_reward_func": 0.1875,
+      "rewards/question_recreation_reward_func": 0.11936723813414574,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": -0.016750000417232513,
+      "rewards/xmlcount_reward_func": 0.005750000011175871,
       "step": 16
     },
     {
-      "completion_length": 118.03125,
+      "completion_length": 77.59375,
       "epoch": 0.9,
-      "grad_norm": 27.92249870300293,
-      "kl": 0.03141187228175113,
+      "grad_norm": 8.833693504333496,
+      "kl": 0.06381005875300616,
       "learning_rate": 1.3545689574841341e-08,
-      "loss": 0.0,
-      "reward": 0.3571384238312021,
-      "reward_std": 0.3553317471523769,
-      "rewards/concensus_correctness_reward_func": 0.019187500700354576,
-      "rewards/consensus_reward_func": 0.1875,
+      "loss": 0.0001,
+      "reward": 0.23175003798678517,
+      "reward_std": 0.031403945409692824,
+      "rewards/concensus_correctness_reward_func": 0.0,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.15045091963838786,
+      "rewards/final_correctness_reward_func": 0.125,
+      "rewards/question_recreation_reward_func": 0.10675003798678517,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
       "rewards/xmlcount_reward_func": 0.0,
       "step": 18
     },
     {
-      "completion_length": 104.75,
+      "completion_length": 97.0625,
       "epoch": 1.0,
-      "grad_norm": 5.20966100692749,
-      "kl": 0.07331209685071371,
+      "grad_norm": 23.614116668701172,
+      "kl": 0.18440437169920187,
       "learning_rate": 0.0,
-      "loss": 0.0001,
-      "reward": 0.12898236550972797,
-      "reward_std": 0.06363634395529516,
+      "loss": 0.0002,
+      "reward": 0.15951462506200187,
+      "reward_std": 0.12634751566764635,
       "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.0,
-      "rewards/question_recreation_reward_func": 0.12898236550972797,
+      "rewards/final_correctness_reward_func": 0.0625,
+      "rewards/question_recreation_reward_func": 0.10354587234905921,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
-      "rewards/xmlcount_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": -0.006531250197440386,
       "step": 20
     },
     {
       "epoch": 1.0,
       "step": 20,
       "total_flos": 0.0,
-      "train_loss": 3.0612738328272826e-05,
-      "train_runtime": 915.7967,
-      "train_samples_per_second": 0.349,
-      "train_steps_per_second": 0.022
+      "train_loss": 0.00011620071672950872,
+      "train_runtime": 657.8689,
+      "train_samples_per_second": 0.486,
+      "train_steps_per_second": 0.03
     }
   ],
   "logging_steps": 2,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4465d29ac5e03c96dac45e1932a3e3e27d76a6974141bfedf5adbacbbbdf2ff8
+oid sha256:8d6dca5f2da01dab09e6e5302424155a551834b766d7a2acd3354b7cb16eb9db
 size 6008