{ "best_global_step": 20, "best_metric": 2.5787403924124583, "best_model_checkpoint": "/home3/s20245636/lab/OCR/VEHM/instr_tuning/output/fermat/red-600/v1-20250611-015104/checkpoint-20", "epoch": 0.4118404118404118, "eval_steps": 20, "global_step": 20, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 421.0, "completions/mean_length": 213.61459350585938, "completions/min_length": 159.0, "epoch": 0.02059202059202059, "grad_norm": 0.10479891461391005, "kl": 0.0, "learning_rate": 1.6666666666666667e-05, "loss": 2.023763954639435e-06, "memory(GiB)": 69.74, "reward": 0.9594758749008179, "reward_std": 0.4217739403247833, "rewards/CorrectnessMatchReward/mean": 0.375, "rewards/CorrectnessMatchReward/std": 0.4853886067867279, "rewards/CosineReward/mean": 0.06821195781230927, "rewards/CosineReward/std": 0.7002815008163452, "rewards/LocalizationReward/mean": 0.3333333432674408, "rewards/LocalizationReward/std": 0.47263696789741516, "rewards/RepetitionPenalty/mean": -0.0006631638389080763, "rewards/RepetitionPenalty/std": 0.0024723168462514877, "rewards/ThoughtLengthReward/mean": 0.18359375, "rewards/ThoughtLengthReward/std": 0.11070504039525986, "step": 1, "train_speed(iter/s)": 0.002402 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 399.0, "completions/mean_length": 221.2916717529297, "completions/min_length": 159.0, "epoch": 0.04118404118404118, "grad_norm": 0.10217217623805434, "kl": 0.0, "learning_rate": 3.3333333333333335e-05, "loss": 4.35501942774863e-06, "memory(GiB)": 71.58, "reward": 1.3502254486083984, "reward_std": 0.4391779899597168, "rewards/CorrectnessMatchReward/mean": 0.484375, "rewards/CorrectnessMatchReward/std": 0.5010623931884766, "rewards/CosineReward/mean": 0.2260473668575287, "rewards/CosineReward/std": 0.7226006388664246, "rewards/LocalizationReward/mean": 0.453125, "rewards/LocalizationReward/std": 0.49909934401512146, "rewards/RepetitionPenalty/mean": -0.0008220267482101917, "rewards/RepetitionPenalty/std": 0.003143024630844593, "rewards/ThoughtLengthReward/mean": 0.1875, "rewards/ThoughtLengthReward/std": 0.10853619128465652, "step": 2, "train_speed(iter/s)": 0.002552 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 389.0, "completions/mean_length": 245.65625, "completions/min_length": 155.0, "epoch": 0.06177606177606178, "grad_norm": 0.08675648900978587, "kl": 0.0011963844299316406, "learning_rate": 5e-05, "loss": 8.833594620227814e-06, "memory(GiB)": 73.82, "reward": 1.359039306640625, "reward_std": 0.5673330426216125, "rewards/CorrectnessMatchReward/mean": 0.4895833432674408, "rewards/CorrectnessMatchReward/std": 0.501198410987854, "rewards/CosineReward/mean": 0.23226088285446167, "rewards/CosineReward/std": 0.7217369675636292, "rewards/LocalizationReward/mean": 0.4375, "rewards/LocalizationReward/std": 0.4973752796649933, "rewards/RepetitionPenalty/mean": -0.0008257072186097503, "rewards/RepetitionPenalty/std": 0.00291589112021029, "rewards/ThoughtLengthReward/mean": 0.2005208432674408, "rewards/ThoughtLengthReward/std": 0.09986766427755356, "step": 3, "train_speed(iter/s)": 0.002411 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 401.0, "completions/mean_length": 228.4166717529297, "completions/min_length": 155.0, "epoch": 0.08236808236808237, "grad_norm": 0.1108443574397371, "kl": 0.0014638900756835938, "learning_rate": 4.993910125649561e-05, "loss": 2.1827097498317016e-06, "memory(GiB)": 74.29, "reward": 1.5590394735336304, "reward_std": 0.42595261335372925, "rewards/CorrectnessMatchReward/mean": 0.546875, "rewards/CorrectnessMatchReward/std": 0.49909934401512146, "rewards/CosineReward/mean": 0.31609195470809937, "rewards/CosineReward/std": 0.7199137210845947, "rewards/LocalizationReward/mean": 0.5052083730697632, "rewards/LocalizationReward/std": 0.5012800097465515, "rewards/RepetitionPenalty/mean": -0.0005419938242994249, "rewards/RepetitionPenalty/std": 0.002184933749958873, "rewards/ThoughtLengthReward/mean": 0.19140625, "rewards/ThoughtLengthReward/std": 0.10617875307798386, "step": 4, "train_speed(iter/s)": 0.002428 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 411.0, "completions/mean_length": 237.0625, "completions/min_length": 166.0, "epoch": 0.10296010296010295, "grad_norm": 0.11543720244360418, "kl": 0.005893707275390625, "learning_rate": 4.975670171853926e-05, "loss": 1.5038375750009436e-05, "memory(GiB)": 74.29, "reward": 1.4182860851287842, "reward_std": 0.2951998710632324, "rewards/CorrectnessMatchReward/mean": 0.5, "rewards/CorrectnessMatchReward/std": 0.5013071894645691, "rewards/CosineReward/mean": 0.24784688651561737, "rewards/CosineReward/std": 0.7203618288040161, "rewards/LocalizationReward/mean": 0.4427083432674408, "rewards/LocalizationReward/std": 0.4980054199695587, "rewards/RepetitionPenalty/mean": -0.001435890095308423, "rewards/RepetitionPenalty/std": 0.005375504959374666, "rewards/ThoughtLengthReward/mean": 0.2291666716337204, "rewards/ThoughtLengthReward/std": 0.06927699595689774, "step": 5, "train_speed(iter/s)": 0.002347 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 415.0, "completions/mean_length": 245.59375, "completions/min_length": 167.0, "epoch": 0.12355212355212356, "grad_norm": 0.1181069257034111, "kl": 0.0206451416015625, "learning_rate": 4.9453690018345144e-05, "loss": 4.468625411391258e-05, "memory(GiB)": 75.33, "reward": 1.221825122833252, "reward_std": 0.4320642650127411, "rewards/CorrectnessMatchReward/mean": 0.4479166865348816, "rewards/CorrectnessMatchReward/std": 0.4985799789428711, "rewards/CosineReward/mean": 0.17357566952705383, "rewards/CosineReward/std": 0.7135447263717651, "rewards/LocalizationReward/mean": 0.3541666865348816, "rewards/LocalizationReward/std": 0.4795103669166565, "rewards/RepetitionPenalty/mean": -0.0012296573258936405, "rewards/RepetitionPenalty/std": 0.003888001898303628, "rewards/ThoughtLengthReward/mean": 0.2473958432674408, "rewards/ThoughtLengthReward/std": 0.025448637083172798, "step": 6, "train_speed(iter/s)": 0.002324 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 458.0, "completions/mean_length": 261.07293701171875, "completions/min_length": 182.0, "epoch": 0.14414414414414414, "grad_norm": 0.11187858854446106, "kl": 0.0838165283203125, "learning_rate": 4.9031542398457974e-05, "loss": 0.00017542685964144766, "memory(GiB)": 75.33, "reward": 1.0862802267074585, "reward_std": 0.3004652261734009, "rewards/CorrectnessMatchReward/mean": 0.40625, "rewards/CorrectnessMatchReward/std": 0.49241629242897034, "rewards/CosineReward/mean": 0.11361651867628098, "rewards/CosineReward/std": 0.6996390223503113, "rewards/LocalizationReward/mean": 0.3177083432674408, "rewards/LocalizationReward/std": 0.46680259704589844, "rewards/RepetitionPenalty/mean": -0.0012946093920618296, "rewards/RepetitionPenalty/std": 0.0035169320181012154, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 7, "train_speed(iter/s)": 0.002296 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 498.0, "completions/mean_length": 315.33856201171875, "completions/min_length": 189.0, "epoch": 0.16473616473616473, "grad_norm": 0.6467505853332378, "kl": 0.88934326171875, "learning_rate": 4.849231551964771e-05, "loss": 0.0022380182053893805, "memory(GiB)": 75.33, "reward": 2.119154453277588, "reward_std": 0.23006069660186768, "rewards/CorrectnessMatchReward/mean": 0.7135416865348816, "rewards/CorrectnessMatchReward/std": 0.45328813791275024, "rewards/CosineReward/mean": 0.5472347140312195, "rewards/CosineReward/std": 0.6383499503135681, "rewards/LocalizationReward/mean": 0.609375, "rewards/LocalizationReward/std": 0.4891660213470459, "rewards/RepetitionPenalty/mean": -0.00099703180603683, "rewards/RepetitionPenalty/std": 0.003000867087393999, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 8, "train_speed(iter/s)": 0.002229 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 491.0, "completions/mean_length": 280.76043701171875, "completions/min_length": 182.0, "epoch": 0.18532818532818532, "grad_norm": 0.11481952665958169, "kl": 0.2429046630859375, "learning_rate": 4.783863644106502e-05, "loss": 0.000612734176684171, "memory(GiB)": 75.79, "reward": 1.5433845520019531, "reward_std": 0.21831923723220825, "rewards/CorrectnessMatchReward/mean": 0.5260416865348816, "rewards/CorrectnessMatchReward/std": 0.500626802444458, "rewards/CosineReward/mean": 0.28393489122390747, "rewards/CosineReward/std": 0.7069699168205261, "rewards/LocalizationReward/mean": 0.484375, "rewards/LocalizationReward/std": 0.5010623931884766, "rewards/RepetitionPenalty/mean": -0.0009670276194810867, "rewards/RepetitionPenalty/std": 0.0027709794230759144, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 9, "train_speed(iter/s)": 0.002219 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 499.0, "completions/mean_length": 269.4895935058594, "completions/min_length": 188.0, "epoch": 0.2059202059202059, "grad_norm": 0.09307591047318431, "kl": 0.1132965087890625, "learning_rate": 4.707368982147318e-05, "loss": 0.000245713337790221, "memory(GiB)": 76.07, "reward": 1.4684938192367554, "reward_std": 0.09292810410261154, "rewards/CorrectnessMatchReward/mean": 0.5052083730697632, "rewards/CorrectnessMatchReward/std": 0.5012800097465515, "rewards/CosineReward/mean": 0.2559816241264343, "rewards/CosineReward/std": 0.7106394171714783, "rewards/LocalizationReward/mean": 0.4583333432674408, "rewards/LocalizationReward/std": 0.4995635151863098, "rewards/RepetitionPenalty/mean": -0.0010295565007254481, "rewards/RepetitionPenalty/std": 0.0030646712984889746, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 10, "train_speed(iter/s)": 0.002168 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 457.0, "completions/mean_length": 297.13543701171875, "completions/min_length": 183.0, "epoch": 0.22651222651222652, "grad_norm": 0.09493513091521967, "kl": 0.08095550537109375, "learning_rate": 4.620120240391065e-05, "loss": 0.00018348329467698932, "memory(GiB)": 77.43, "reward": 1.725380301475525, "reward_std": 0.1333923041820526, "rewards/CorrectnessMatchReward/mean": 0.5885416865348816, "rewards/CorrectnessMatchReward/std": 0.493384450674057, "rewards/CosineReward/mean": 0.3725980222225189, "rewards/CosineReward/std": 0.6997078061103821, "rewards/LocalizationReward/mean": 0.515625, "rewards/LocalizationReward/std": 0.5010623335838318, "rewards/RepetitionPenalty/mean": -0.001384397386573255, "rewards/RepetitionPenalty/std": 0.003916753921657801, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 11, "train_speed(iter/s)": 0.00205 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 443.0, "completions/mean_length": 293.6302185058594, "completions/min_length": 181.0, "epoch": 0.2471042471042471, "grad_norm": 0.10466242902627128, "kl": 0.066253662109375, "learning_rate": 4.522542485937369e-05, "loss": 0.00015122009790502489, "memory(GiB)": 77.43, "reward": 1.68711256980896, "reward_std": 0.43765807151794434, "rewards/CorrectnessMatchReward/mean": 0.5833333730697632, "rewards/CorrectnessMatchReward/std": 0.49429556727409363, "rewards/CosineReward/mean": 0.3657306432723999, "rewards/CosineReward/std": 0.7009669542312622, "rewards/LocalizationReward/mean": 0.4895833432674408, "rewards/LocalizationReward/std": 0.501198410987854, "rewards/RepetitionPenalty/mean": -0.0015348497545346618, "rewards/RepetitionPenalty/std": 0.004266205709427595, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 12, "train_speed(iter/s)": 0.001976 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 467.0, "completions/mean_length": 285.25, "completions/min_length": 183.0, "epoch": 0.2676962676962677, "grad_norm": 0.09882070085915172, "kl": 0.0605621337890625, "learning_rate": 4.415111107797445e-05, "loss": 0.00013017794117331505, "memory(GiB)": 77.43, "reward": 1.4799035787582397, "reward_std": 0.22159945964813232, "rewards/CorrectnessMatchReward/mean": 0.5, "rewards/CorrectnessMatchReward/std": 0.5013071894645691, "rewards/CosineReward/mean": 0.24739837646484375, "rewards/CosineReward/std": 0.7114919424057007, "rewards/LocalizationReward/mean": 0.484375, "rewards/LocalizationReward/std": 0.5010623335838318, "rewards/RepetitionPenalty/mean": -0.0018697160994634032, "rewards/RepetitionPenalty/std": 0.006876891944557428, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 13, "train_speed(iter/s)": 0.0019 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 439.0, "completions/mean_length": 295.7552185058594, "completions/min_length": 181.0, "epoch": 0.2882882882882883, "grad_norm": 0.08776304365128487, "kl": 0.0581512451171875, "learning_rate": 4.2983495008466276e-05, "loss": 0.00012497154239099473, "memory(GiB)": 77.43, "reward": 1.776354193687439, "reward_std": 0.3223668038845062, "rewards/CorrectnessMatchReward/mean": 0.625, "rewards/CorrectnessMatchReward/std": 0.4853886067867279, "rewards/CosineReward/mean": 0.42455118894577026, "rewards/CosineReward/std": 0.6895366907119751, "rewards/LocalizationReward/mean": 0.4791666865348816, "rewards/LocalizationReward/std": 0.5008718967437744, "rewards/RepetitionPenalty/mean": -0.00106146652251482, "rewards/RepetitionPenalty/std": 0.003482910105958581, "rewards/ThoughtLengthReward/mean": 0.2486979216337204, "rewards/ThoughtLengthReward/std": 0.018042193725705147, "step": 14, "train_speed(iter/s)": 0.001821 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 461.0, "completions/mean_length": 299.93231201171875, "completions/min_length": 184.0, "epoch": 0.3088803088803089, "grad_norm": 0.08944460012633934, "kl": 0.05854034423828125, "learning_rate": 4.172826515897146e-05, "loss": 0.00013586971908807755, "memory(GiB)": 77.43, "reward": 1.760624647140503, "reward_std": 0.29807019233703613, "rewards/CorrectnessMatchReward/mean": 0.609375, "rewards/CorrectnessMatchReward/std": 0.4891660213470459, "rewards/CosineReward/mean": 0.4025270342826843, "rewards/CosineReward/std": 0.693856954574585, "rewards/LocalizationReward/mean": 0.5, "rewards/LocalizationReward/std": 0.5013071894645691, "rewards/RepetitionPenalty/mean": -0.001277304720133543, "rewards/RepetitionPenalty/std": 0.004088031593710184, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 15, "train_speed(iter/s)": 0.001785 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 476.0, "completions/mean_length": 301.24481201171875, "completions/min_length": 188.0, "epoch": 0.32947232947232946, "grad_norm": 0.09247588272626009, "kl": 0.0556640625, "learning_rate": 4.039153688314145e-05, "loss": 0.0001178596357931383, "memory(GiB)": 77.43, "reward": 1.8202136754989624, "reward_std": 0.35252445936203003, "rewards/CorrectnessMatchReward/mean": 0.625, "rewards/CorrectnessMatchReward/std": 0.4853886067867279, "rewards/CosineReward/mean": 0.4250945448875427, "rewards/CosineReward/std": 0.6889393329620361, "rewards/LocalizationReward/mean": 0.5208333730697632, "rewards/LocalizationReward/std": 0.5008718371391296, "rewards/RepetitionPenalty/mean": -0.0007142613176256418, "rewards/RepetitionPenalty/std": 0.0023925723508000374, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 16, "train_speed(iter/s)": 0.001738 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 468.0, "completions/mean_length": 308.96875, "completions/min_length": 187.0, "epoch": 0.35006435006435005, "grad_norm": 0.093285638447722, "kl": 0.05938720703125, "learning_rate": 3.897982258676867e-05, "loss": 0.00012818226241506636, "memory(GiB)": 77.43, "reward": 2.039015054702759, "reward_std": 0.36231595277786255, "rewards/CorrectnessMatchReward/mean": 0.6979166865348816, "rewards/CorrectnessMatchReward/std": 0.4603615999221802, "rewards/CosineReward/mean": 0.5298215746879578, "rewards/CosineReward/std": 0.6541029214859009, "rewards/LocalizationReward/mean": 0.5625, "rewards/LocalizationReward/std": 0.4973752796649933, "rewards/RepetitionPenalty/mean": -0.0012233429588377476, "rewards/RepetitionPenalty/std": 0.0036449579056352377, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 17, "train_speed(iter/s)": 0.001681 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 469.0, "completions/mean_length": 313.15106201171875, "completions/min_length": 185.0, "epoch": 0.37065637065637064, "grad_norm": 0.08803423910461482, "kl": 0.06447601318359375, "learning_rate": 3.7500000000000003e-05, "loss": 0.00014094838115852326, "memory(GiB)": 77.43, "reward": 1.7767027616500854, "reward_std": 0.3571460545063019, "rewards/CorrectnessMatchReward/mean": 0.6302083730697632, "rewards/CorrectnessMatchReward/std": 0.484010249376297, "rewards/CosineReward/mean": 0.43386849761009216, "rewards/CosineReward/std": 0.6876592040061951, "rewards/LocalizationReward/mean": 0.4635416865348816, "rewards/LocalizationReward/std": 0.49997273087501526, "rewards/RepetitionPenalty/mean": -0.0009157162858173251, "rewards/RepetitionPenalty/std": 0.0027353554032742977, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 18, "train_speed(iter/s)": 0.001639 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 452.0, "completions/mean_length": 339.6770935058594, "completions/min_length": 173.0, "epoch": 0.3912483912483912, "grad_norm": 0.10478084177122564, "kl": 0.0612640380859375, "learning_rate": 3.5959278669726935e-05, "loss": 0.0001375687716063112, "memory(GiB)": 77.88, "reward": 2.019371747970581, "reward_std": 0.3344690203666687, "rewards/CorrectnessMatchReward/mean": 0.703125, "rewards/CorrectnessMatchReward/std": 0.4580754339694977, "rewards/CosineReward/mean": 0.5363699197769165, "rewards/CosineReward/std": 0.6495335698127747, "rewards/LocalizationReward/mean": 0.53125, "rewards/LocalizationReward/std": 0.5003271102905273, "rewards/RepetitionPenalty/mean": -0.0013731191866099834, "rewards/RepetitionPenalty/std": 0.003995626699179411, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 19, "train_speed(iter/s)": 0.001608 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 461.0, "completions/mean_length": 351.08856201171875, "completions/min_length": 191.0, "epoch": 0.4118404118404118, "grad_norm": 0.10812396849475318, "kl": 0.0747222900390625, "learning_rate": 3.436516483539781e-05, "loss": 0.00018165742221754044, "memory(GiB)": 77.88, "reward": 2.520355701446533, "reward_std": 0.32913583517074585, "rewards/CorrectnessMatchReward/mean": 0.8541666865348816, "rewards/CorrectnessMatchReward/std": 0.3538617789745331, "rewards/CosineReward/mean": 0.7510079145431519, "rewards/CosineReward/std": 0.502041757106781, "rewards/LocalizationReward/mean": 0.6666666865348816, "rewards/LocalizationReward/std": 0.4726369380950928, "rewards/RepetitionPenalty/mean": -0.001485575339756906, "rewards/RepetitionPenalty/std": 0.004180458839982748, "rewards/ThoughtLengthReward/mean": 0.25, "rewards/ThoughtLengthReward/std": 0.0, "step": 20, "train_speed(iter/s)": 0.001539 }, { "epoch": 0.4118404118404118, "eval_clip_ratio/high_max": 0.0, "eval_clip_ratio/high_mean": 0.0, "eval_clip_ratio/low_mean": 0.0, "eval_clip_ratio/low_min": 0.0, "eval_clip_ratio/region_mean": 0.0, "eval_completions/clipped_ratio": 0.0, "eval_completions/max_length": 455.85714285714283, "eval_completions/mean_length": 373.452392578125, "eval_completions/min_length": 284.7142857142857, "eval_kl": 0.07278878348214286, "eval_loss": 0.00017117673996835947, "eval_reward": 2.5787403924124583, "eval_reward_std": 0.4637991892439978, "eval_rewards/CorrectnessMatchReward/mean": 0.9523809552192688, "eval_rewards/CorrectnessMatchReward/std": 0.10584887862205505, "eval_rewards/CosineReward/mean": 0.8896037680762154, "eval_rewards/CosineReward/std": 0.15317954731706, "eval_rewards/LocalizationReward/mean": 0.4880952558347157, "eval_rewards/LocalizationReward/std": 0.47584791268621174, "eval_rewards/RepetitionPenalty/mean": -0.001339557098357805, "eval_rewards/RepetitionPenalty/std": 0.0026112154259213378, "eval_rewards/ThoughtLengthReward/mean": 0.25, "eval_rewards/ThoughtLengthReward/std": 0.0, "eval_runtime": 1161.7833, "eval_samples_per_second": 0.013, "eval_steps_per_second": 0.002, "step": 20 } ], "logging_steps": 1, "max_steps": 48, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 20, "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": 12, "trial_name": null, "trial_params": null }