{
  "best_metric": 0.6185767650604248,
  "best_model_checkpoint": "/storage/trained_grpo_distill_14b/v42-20250317-212613/checkpoint-2",
  "epoch": 1.784313725490196,
  "eval_steps": 2,
  "global_step": 22,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "clip_ratio": 0.0,
      "completion_length": 6417.984375,
      "epoch": 0.0784313725490196,
      "grad_norm": 0.002763764699921012,
      "kl": 0.0,
      "learning_rate": 1.6666666666666667e-05,
      "loss": 0.0836174339056015,
      "memory(GiB)": 184.97,
      "response_clip_ratio": 0.12890625,
      "reward": 0.4296856001019478,
      "reward_std": 0.14499560371041298,
      "rewards/CosineReward": 0.4296856001019478,
      "step": 1,
      "train_speed(iter/s)": 0.000129
    },
    {
      "epoch": 0.1568627450980392,
      "grad_norm": 0.0027679470367729664,
      "learning_rate": 3.3333333333333335e-05,
      "loss": 0.0836174339056015,
      "memory(GiB)": 184.97,
      "step": 2,
      "train_speed(iter/s)": 0.000249
    },
    {
      "epoch": 0.1568627450980392,
      "eval_clip_ratio": 0.00024764647241681814,
      "eval_completion_length": 5048.4375,
      "eval_kl": 0.0001354217529296875,
      "eval_loss": 0.0001584067940711975,
      "eval_response_clip_ratio": 0.0625,
      "eval_reward": 0.6185767650604248,
      "eval_reward_std": 0.12649827264249325,
      "eval_rewards/CosineReward": 0.6185767650604248,
      "eval_runtime": 1074.328,
      "eval_samples_per_second": 0.007,
      "eval_steps_per_second": 0.001,
      "step": 2
    },
    {
      "clip_ratio": 0.0001064514563040575,
      "completion_length": 6496.7421875,
      "epoch": 0.23529411764705882,
      "grad_norm": 0.0026931765023618937,
      "kl": 6.115436553955078e-05,
      "learning_rate": 5e-05,
      "loss": 0.08872900903224945,
      "memory(GiB)": 184.97,
      "response_clip_ratio": 0.15625,
      "reward": 0.4729522615671158,
      "reward_std": 0.11709235422313213,
      "rewards/CosineReward": 0.4729522615671158,
      "step": 3,
      "train_speed(iter/s)": 0.000176
    },
    {
      "epoch": 0.3137254901960784,
      "grad_norm": 0.002692870795726776,
      "learning_rate": 6.666666666666667e-05,
      "loss": 0.08870330452919006,
      "memory(GiB)": 184.97,
      "step": 4,
      "train_speed(iter/s)": 0.00023
    },
    {
      "epoch": 0.3137254901960784,
      "eval_clip_ratio": 0.0002070416376227513,
      "eval_completion_length": 5490.0,
      "eval_kl": 0.00012302398681640625,
      "eval_loss": 1.0617077350616455e-05,
      "eval_response_clip_ratio": 0.0,
      "eval_reward": 0.5434095412492752,
      "eval_reward_std": 0.1754817459732294,
      "eval_rewards/CosineReward": 0.5434095412492752,
      "eval_runtime": 1342.619,
      "eval_samples_per_second": 0.006,
      "eval_steps_per_second": 0.001,
      "step": 4
    },
    {
      "clip_ratio": 0.00019944452469644602,
      "completion_length": 6302.51953125,
      "epoch": 0.39215686274509803,
      "grad_norm": 0.003180823987349868,
      "kl": 0.00012123584747314453,
      "learning_rate": 8.333333333333334e-05,
      "loss": 0.08817961066961288,
      "memory(GiB)": 184.97,
      "response_clip_ratio": 0.1171875,
      "reward": 0.480031818151474,
      "reward_std": 0.12067002058029175,
      "rewards/CosineReward": 0.480031818151474,
      "step": 5,
      "train_speed(iter/s)": 0.000188
    },
    {
      "epoch": 0.47058823529411764,
      "grad_norm": 0.003291113767772913,
      "learning_rate": 0.0001,
      "loss": 0.08811932057142258,
      "memory(GiB)": 184.97,
      "step": 6,
      "train_speed(iter/s)": 0.000224
    },
    {
      "epoch": 0.47058823529411764,
      "eval_clip_ratio": 0.00020231583039276302,
      "eval_completion_length": 5615.6875,
      "eval_kl": 0.00018548965454101562,
      "eval_loss": 0.00017052143812179565,
      "eval_response_clip_ratio": 0.125,
      "eval_reward": 0.4852888882160187,
      "eval_reward_std": 0.14715369045734406,
      "eval_rewards/CosineReward": 0.4852888882160187,
      "eval_runtime": 1364.1653,
      "eval_samples_per_second": 0.006,
      "eval_steps_per_second": 0.001,
      "step": 6
    },
    {
      "clip_ratio": 0.0002164878969779238,
      "completion_length": 6660.578125,
      "epoch": 0.5490196078431373,
      "grad_norm": 0.003197300247848034,
      "kl": 0.00015372037887573242,
      "learning_rate": 9.991540791356342e-05,
      "loss": 0.086369089782238,
      "memory(GiB)": 184.97,
      "response_clip_ratio": 0.13671875,
      "reward": 0.43960579484701157,
      "reward_std": 0.12519060634076595,
      "rewards/CosineReward": 0.43960579484701157,
      "step": 7,
      "train_speed(iter/s)": 0.000194
    },
    {
      "epoch": 0.6274509803921569,
      "grad_norm": 0.0034806649200618267,
      "learning_rate": 9.966191788709716e-05,
      "loss": 0.08621430397033691,
      "memory(GiB)": 184.97,
      "step": 8,
      "train_speed(iter/s)": 0.00022
    },
    {
      "epoch": 0.6274509803921569,
      "eval_clip_ratio": 7.152849320846144e-05,
      "eval_completion_length": 5739.0625,
      "eval_kl": 0.001102447509765625,
      "eval_loss": 0.00010670721530914307,
      "eval_response_clip_ratio": 0.125,
      "eval_reward": 0.5124959498643875,
      "eval_reward_std": 0.21372128650546074,
      "eval_rewards/CosineReward": 0.5124959498643875,
      "eval_runtime": 1470.3009,
      "eval_samples_per_second": 0.005,
      "eval_steps_per_second": 0.001,
      "step": 8
    },
    {
      "clip_ratio": 0.0003287304334662622,
      "completion_length": 6587.18359375,
      "epoch": 0.7058823529411765,
      "grad_norm": 0.004174312110990286,
      "kl": 0.0007574558258056641,
      "learning_rate": 9.924038765061042e-05,
      "loss": 0.0881085991859436,
      "memory(GiB)": 184.97,
      "response_clip_ratio": 0.1953125,
      "reward": 0.4415409117937088,
      "reward_std": 0.10440743528306484,
      "rewards/CosineReward": 0.4415409117937088,
      "step": 9,
      "train_speed(iter/s)": 0.000196
    },
    {
      "epoch": 0.7843137254901961,
      "grad_norm": 0.003945177886635065,
      "learning_rate": 9.865224352899119e-05,
      "loss": 0.08792611956596375,
      "memory(GiB)": 184.97,
      "step": 10,
      "train_speed(iter/s)": 0.000217
    },
    {
      "epoch": 0.7843137254901961,
      "eval_clip_ratio": 0.00014116203965386376,
      "eval_completion_length": 6354.875,
      "eval_kl": 0.003173828125,
      "eval_loss": 3.191456198692322e-05,
      "eval_response_clip_ratio": 0.125,
      "eval_reward": 0.5644881427288055,
      "eval_reward_std": 0.11511430516839027,
      "eval_rewards/CosineReward": 0.5644881427288055,
      "eval_runtime": 1510.6041,
      "eval_samples_per_second": 0.005,
      "eval_steps_per_second": 0.001,
      "step": 10
    },
    {
      "clip_ratio": 0.0004681620275732712,
      "completion_length": 6903.640625,
      "epoch": 0.8627450980392157,
      "grad_norm": 0.00393508980050683,
      "kl": 0.002521514892578125,
      "learning_rate": 9.789947561577445e-05,
      "loss": 0.0897117406129837,
      "memory(GiB)": 184.97,
      "response_clip_ratio": 0.171875,
      "reward": 0.4390856474637985,
      "reward_std": 0.11577454954385757,
      "rewards/CosineReward": 0.4390856474637985,
      "step": 11,
      "train_speed(iter/s)": 0.000198
    },
    {
      "epoch": 0.9411764705882353,
      "grad_norm": 0.0034022172912955284,
      "learning_rate": 9.698463103929542e-05,
      "loss": 0.08947472274303436,
      "memory(GiB)": 184.97,
      "step": 12,
      "train_speed(iter/s)": 0.000215
    },
    {
      "epoch": 0.9411764705882353,
      "eval_clip_ratio": 7.602112054883037e-05,
      "eval_completion_length": 6181.5,
      "eval_kl": 0.00714111328125,
      "eval_loss": 0.00035206228494644165,
      "eval_response_clip_ratio": 0.1875,
      "eval_reward": 0.45696309208869934,
      "eval_reward_std": 0.04341535549610853,
      "eval_rewards/CosineReward": 0.45696309208869934,
      "eval_runtime": 1437.7604,
      "eval_samples_per_second": 0.006,
      "eval_steps_per_second": 0.001,
      "step": 12
    },
    {
      "clip_ratio": 0.0003813115013144852,
      "completion_length": 6900.296875,
      "epoch": 1.0784313725490196,
      "grad_norm": 0.013968409970402718,
      "kl": 0.0058441162109375,
      "learning_rate": 9.591080534401371e-05,
      "loss": 0.08859321475028992,
      "memory(GiB)": 184.97,
      "response_clip_ratio": 0.203125,
      "reward": 0.4131528064608574,
      "reward_std": 0.1398314479738474,
      "rewards/CosineReward": 0.4131528064608574,
      "step": 13,
      "train_speed(iter/s)": 0.000199
    },
    {
      "epoch": 1.156862745098039,
      "grad_norm": 0.0037056964356452227,
      "learning_rate": 9.468163201617062e-05,
      "loss": 0.08834587037563324,
      "memory(GiB)": 184.97,
      "step": 14,
      "train_speed(iter/s)": 0.000214
    },
    {
      "epoch": 1.156862745098039,
      "eval_clip_ratio": 0.00016801093079266138,
      "eval_completion_length": 6428.5,
      "eval_kl": 0.012603759765625,
      "eval_loss": 0.0005806386470794678,
      "eval_response_clip_ratio": 0.25,
      "eval_reward": 0.4538377672433853,
      "eval_reward_std": 0.1443052440881729,
      "eval_rewards/CosineReward": 0.4538377672433853,
      "eval_runtime": 1590.5873,
      "eval_samples_per_second": 0.005,
      "eval_steps_per_second": 0.001,
      "step": 14
    },
    {
      "clip_ratio": 0.00040143599017028464,
      "completion_length": 6398.2890625,
      "epoch": 1.2352941176470589,
      "grad_norm": 0.0041822106577456,
      "kl": 0.0112152099609375,
      "learning_rate": 9.330127018922194e-05,
      "loss": 0.09059401601552963,
      "memory(GiB)": 184.97,
      "response_clip_ratio": 0.1484375,
      "reward": 0.46379417926073074,
      "reward_std": 0.11776200495660305,
      "rewards/CosineReward": 0.46379417926073074,
      "step": 15,
      "train_speed(iter/s)": 0.0002
    },
    {
      "epoch": 1.3137254901960784,
      "grad_norm": 0.0046047731302678585,
      "learning_rate": 9.177439057064683e-05,
      "loss": 0.09031489491462708,
      "memory(GiB)": 184.97,
      "step": 16,
      "train_speed(iter/s)": 0.000212
    },
    {
      "epoch": 1.3137254901960784,
      "eval_clip_ratio": 5.4339221605914645e-05,
      "eval_completion_length": 5552.0625,
      "eval_kl": 0.01727294921875,
      "eval_loss": 0.0005940012633800507,
      "eval_response_clip_ratio": 0.0625,
      "eval_reward": 0.506618320941925,
      "eval_reward_std": 0.21169210970401764,
      "eval_rewards/CosineReward": 0.506618320941925,
      "eval_runtime": 1518.0802,
      "eval_samples_per_second": 0.005,
      "eval_steps_per_second": 0.001,
      "step": 16
    },
    {
      "clip_ratio": 6.063861019356409e-05,
      "completion_length": 6686.59765625,
      "epoch": 1.392156862745098,
      "grad_norm": 0.004301061388105154,
      "kl": 0.0179443359375,
      "learning_rate": 9.01061596377522e-05,
      "loss": 0.07838998734951019,
      "memory(GiB)": 184.28,
      "response_clip_ratio": 0.18359375,
      "reward": 0.4415987730026245,
      "reward_std": 0.09190415777266026,
      "rewards/CosineReward": 0.4415987730026245,
      "step": 17,
      "train_speed(iter/s)": 0.002176
    },
    {
      "epoch": 1.4705882352941178,
      "grad_norm": 0.004479076247662306,
      "learning_rate": 8.83022221559489e-05,
      "loss": 0.07815618067979813,
      "memory(GiB)": 184.28,
      "step": 18,
      "train_speed(iter/s)": 0.002225
    },
    {
      "epoch": 1.4705882352941178,
      "eval_clip_ratio": 9.80815566435922e-05,
      "eval_completion_length": 6907.125,
      "eval_kl": 0.02203369140625,
      "eval_loss": 0.0009158775210380554,
      "eval_response_clip_ratio": 0.1875,
      "eval_reward": 0.4723748713731766,
      "eval_reward_std": 0.3155394047498703,
      "eval_rewards/CosineReward": 0.4723748713731766,
      "eval_runtime": 1457.6003,
      "eval_samples_per_second": 0.005,
      "eval_steps_per_second": 0.001,
      "step": 18
    },
    {
      "clip_ratio": 0.00023432340185536304,
      "completion_length": 7038.578125,
      "epoch": 1.5490196078431373,
      "grad_norm": 0.005921878386288881,
      "kl": 0.0222320556640625,
      "learning_rate": 8.636868207865244e-05,
      "loss": 0.0942603051662445,
      "memory(GiB)": 184.28,
      "response_clip_ratio": 0.19921875,
      "reward": 0.42305945605039597,
      "reward_std": 0.17355207353830338,
      "rewards/CosineReward": 0.42305945605039597,
      "step": 19,
      "train_speed(iter/s)": 0.001092
    },
    {
      "epoch": 1.6274509803921569,
      "grad_norm": 0.006151494570076466,
      "learning_rate": 8.43120818934367e-05,
      "loss": 0.09388023614883423,
      "memory(GiB)": 184.28,
      "step": 20,
      "train_speed(iter/s)": 0.001131
    },
    {
      "epoch": 1.6274509803921569,
      "eval_clip_ratio": 6.59385500512144e-05,
      "eval_completion_length": 7039.0625,
      "eval_kl": 0.02911376953125,
      "eval_loss": 0.0012894980609416962,
      "eval_response_clip_ratio": 0.1875,
      "eval_reward": 0.4684200882911682,
      "eval_reward_std": 0.3149525970220566,
      "eval_rewards/CosineReward": 0.4684200882911682,
      "eval_runtime": 1460.114,
      "eval_samples_per_second": 0.005,
      "eval_steps_per_second": 0.001,
      "step": 20
    },
    {
      "clip_ratio": 0.00028659221607085783,
      "completion_length": 7469.6796875,
      "epoch": 1.7058823529411766,
      "grad_norm": 0.006666958797723055,
      "kl": 0.0279693603515625,
      "learning_rate": 8.213938048432697e-05,
      "loss": 0.075472891330719,
      "memory(GiB)": 184.28,
      "response_clip_ratio": 0.29296875,
      "reward": 0.3378320038318634,
      "reward_std": 0.1059294268488884,
      "rewards/CosineReward": 0.3378320038318634,
      "step": 21,
      "train_speed(iter/s)": 0.000775
    },
    {
      "epoch": 1.784313725490196,
      "grad_norm": 0.0069817290641367435,
      "learning_rate": 7.985792958513931e-05,
      "loss": 0.07505190372467041,
      "memory(GiB)": 184.28,
      "step": 22,
      "train_speed(iter/s)": 0.000803
    },
    {
      "epoch": 1.784313725490196,
      "eval_clip_ratio": 3.5660988942254335e-05,
      "eval_completion_length": 7378.4375,
      "eval_kl": 0.0372314453125,
      "eval_loss": 0.0015567503869533539,
      "eval_response_clip_ratio": 0.1875,
      "eval_reward": 0.4472413510084152,
      "eval_reward_std": 0.311752051115036,
      "eval_rewards/CosineReward": 0.4472413510084152,
      "eval_runtime": 1469.5869,
      "eval_samples_per_second": 0.005,
      "eval_steps_per_second": 0.001,
      "step": 22
    }
  ],
  "logging_steps": 1,
  "max_steps": 60,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 2,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}