End of training

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 6.0011893765477e-06,
-    "train_runtime": 939.4207,
+    "train_loss": 2.5437328076804986e-06,
+    "train_runtime": 1032.5939,
     "train_samples": 5,
-    "train_samples_per_second": 0.043,
-    "train_steps_per_second": 0.011
+    "train_samples_per_second": 0.039,
+    "train_steps_per_second": 0.01
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:181eba8ad9555bd9e4d1b07177e72a31c44e53743dfae606b9b7876752d6de01
+oid sha256:fe32a51b94ee072269400c356ea2fce2758558cc9a0664adc325f9b931289adc
 size 1976163472

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 6.0011893765477e-06,
-    "train_runtime": 939.4207,
+    "train_loss": 2.5437328076804986e-06,
+    "train_runtime": 1032.5939,
     "train_samples": 5,
-    "train_samples_per_second": 0.043,
-    "train_steps_per_second": 0.011
+    "train_samples_per_second": 0.039,
+    "train_steps_per_second": 0.01
 }

trainer_state.json

@@ -10,95 +10,95 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 83.0,
+      "completion_length": 51.625,
       "epoch": 0.8,
-      "grad_norm": 7.9807281494140625,
+      "grad_norm": 0.0,
       "kl": 0.0,
       "learning_rate": 5e-07,
-      "loss": -0.0,
-      "reward": 1.7760924957692623,
-      "reward_std": 0.022786720073781908,
+      "loss": -0.0001,
+      "reward": 1.7738320790231228,
+      "reward_std": 0.002120281533279922,
       "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 1.5,
-      "rewards/question_recreation_reward_func": 0.1510925330221653,
+      "rewards/question_recreation_reward_func": 0.14883213862776756,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
       "rewards/xmlcount_reward_func": 0.125,
       "step": 2
     },
     {
-      "completion_length": 56.833333333333336,
+      "completion_length": 53.166666666666664,
       "epoch": 1.4,
-      "grad_norm": 10.24837589263916,
-      "kl": 0.0015244192351625923,
+      "grad_norm": 6.040588855743408,
+      "kl": 0.006199638561687,
       "learning_rate": 4.415111107797445e-07,
       "loss": 0.0,
-      "reward": 2.550933281580607,
-      "reward_std": 0.07004699483513832,
+      "reward": 2.596501588821411,
+      "reward_std": 0.01238746444384257,
       "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 2.0,
-      "rewards/question_recreation_reward_func": 0.42593332131703693,
+      "rewards/question_recreation_reward_func": 0.47150157888730365,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
       "rewards/xmlcount_reward_func": 0.125,
       "step": 4
     },
     {
-      "completion_length": 77.66666666666667,
+      "completion_length": 66.5,
       "epoch": 2.0,
-      "grad_norm": 6.0542120933532715,
-      "kl": 0.0016350069102675964,
+      "grad_norm": 0.00015045171312522143,
+      "kl": 0.00036060751396386576,
       "learning_rate": 2.934120444167326e-07,
-      "loss": 0.0,
-      "reward": 1.5878510723511379,
-      "reward_std": 0.0029388506275912127,
+      "loss": -0.0,
+      "reward": 1.5683003862698872,
+      "reward_std": 0.010879872056345144,
       "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 1.3333333333333333,
-      "rewards/question_recreation_reward_func": 0.1295176992813746,
+      "rewards/question_recreation_reward_func": 0.10996711999177933,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
       "rewards/xmlcount_reward_func": 0.125,
       "step": 6
     },
     {
-      "completion_length": 72.25,
+      "completion_length": 71.75,
       "epoch": 2.8,
-      "grad_norm": 2.34584641456604,
-      "kl": 0.0029444374376907945,
+      "grad_norm": 6.939717769622803,
+      "kl": 0.01342857114536855,
       "learning_rate": 1.2500000000000005e-07,
-      "loss": -0.0,
-      "reward": 1.8496044538915157,
-      "reward_std": 0.0011594545212574303,
+      "loss": 0.0,
+      "reward": 1.8515289053320885,
+      "reward_std": 0.01689862343482673,
       "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 1.5,
-      "rewards/question_recreation_reward_func": 0.22460449486970901,
+      "rewards/question_recreation_reward_func": 0.2265289444476366,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
       "rewards/xmlcount_reward_func": 0.125,
       "step": 8
     },
     {
-      "completion_length": 75.83333333333333,
+      "completion_length": 53.166666666666664,
       "epoch": 3.4,
-      "grad_norm": 5.32132625579834,
-      "kl": 0.007413267778853576,
+      "grad_norm": 10.517523765563965,
+      "kl": 0.021501405475040276,
       "learning_rate": 1.507684480352292e-08,
-      "loss": 0.0,
-      "reward": 1.6325677633285522,
-      "reward_std": 0.042186157782756105,
+      "loss": 0.0001,
+      "reward": 1.6254764199256897,
+      "reward_std": 0.008517678147957971,
       "rewards/concensus_correctness_reward_func": 0.0,
       "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 1.3333333333333333,
-      "rewards/question_recreation_reward_func": 0.17423443992932638,
+      "rewards/question_recreation_reward_func": 0.16714302822947502,
       "rewards/soft_format_reward_func": 0.0,
       "rewards/strict_format_reward_func": 0.0,
       "rewards/xmlcount_reward_func": 0.125,
@@ -108,10 +108,10 @@
       "epoch": 3.4,
       "step": 10,
       "total_flos": 0.0,
-      "train_loss": 6.0011893765477e-06,
-      "train_runtime": 939.4207,
-      "train_samples_per_second": 0.043,
-      "train_steps_per_second": 0.011
+      "train_loss": 2.5437328076804986e-06,
+      "train_runtime": 1032.5939,
+      "train_samples_per_second": 0.039,
+      "train_steps_per_second": 0.01
     }
   ],
   "logging_steps": 2,