e845312141
Model: cjiao/goldengoose-p3_goose_lowdiv_n128_indoc_tau0.10-25grp Source: Original Platform
2735 lines
98 KiB
JSON
2735 lines
98 KiB
JSON
{
|
|
"best_global_step": null,
|
|
"best_metric": null,
|
|
"best_model_checkpoint": null,
|
|
"epoch": 0.25,
|
|
"eval_steps": 500,
|
|
"global_step": 100,
|
|
"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": 685.0,
|
|
"completions/max_terminated_length": 685.0,
|
|
"completions/mean_length": 222.625,
|
|
"completions/mean_terminated_length": 222.625,
|
|
"completions/min_length": 2.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0025,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 59.533512115478516,
|
|
"learning_rate": 0.0,
|
|
"loss": 0.0,
|
|
"num_tokens": 104344.0,
|
|
"reward": 0.5445312261581421,
|
|
"reward_std": 0.47544848918914795,
|
|
"rewards/format_reward/mean": 0.2890625,
|
|
"rewards/format_reward/std": 0.2789533734321594,
|
|
"rewards/mcq_exact_match_reward/mean": 0.515625,
|
|
"rewards/mcq_exact_match_reward/std": 0.5037065148353577,
|
|
"step": 1
|
|
},
|
|
{
|
|
"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": 1001.0,
|
|
"completions/max_terminated_length": 1001.0,
|
|
"completions/mean_length": 286.765625,
|
|
"completions/mean_terminated_length": 286.765625,
|
|
"completions/min_length": 3.0,
|
|
"completions/min_terminated_length": 3.0,
|
|
"epoch": 0.005,
|
|
"frac_reward_zero_std": 0.25,
|
|
"grad_norm": 14.809494972229004,
|
|
"learning_rate": 2e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 208097.0,
|
|
"reward": 0.4679687023162842,
|
|
"reward_std": 0.25451505184173584,
|
|
"rewards/format_reward/mean": 0.3046875,
|
|
"rewards/format_reward/std": 0.2615155577659607,
|
|
"rewards/mcq_exact_match_reward/mean": 0.4375,
|
|
"rewards/mcq_exact_match_reward/std": 0.5,
|
|
"step": 2
|
|
},
|
|
{
|
|
"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": 1127.0,
|
|
"completions/max_terminated_length": 1127.0,
|
|
"completions/mean_length": 319.671875,
|
|
"completions/mean_terminated_length": 319.671875,
|
|
"completions/min_length": 3.0,
|
|
"completions/min_terminated_length": 3.0,
|
|
"epoch": 0.0075,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 8.160704612731934,
|
|
"learning_rate": 4e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 305484.0,
|
|
"reward": 0.5460937023162842,
|
|
"reward_std": 0.2842584550380707,
|
|
"rewards/format_reward/mean": 0.3046875,
|
|
"rewards/format_reward/std": 0.32874447107315063,
|
|
"rewards/mcq_exact_match_reward/mean": 0.515625,
|
|
"rewards/mcq_exact_match_reward/std": 0.5037065148353577,
|
|
"step": 3
|
|
},
|
|
{
|
|
"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": 1389.0,
|
|
"completions/max_terminated_length": 1389.0,
|
|
"completions/mean_length": 244.53125,
|
|
"completions/mean_terminated_length": 244.53125,
|
|
"completions/min_length": 3.0,
|
|
"completions/min_terminated_length": 3.0,
|
|
"epoch": 0.01,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 8.07497787475586,
|
|
"learning_rate": 6e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 420206.0,
|
|
"reward": 0.3843749761581421,
|
|
"reward_std": 0.36569860577583313,
|
|
"rewards/format_reward/mean": 0.25,
|
|
"rewards/format_reward/std": 0.28171807527542114,
|
|
"rewards/mcq_exact_match_reward/mean": 0.359375,
|
|
"rewards/mcq_exact_match_reward/std": 0.4836103618144989,
|
|
"step": 4
|
|
},
|
|
{
|
|
"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": 1638.0,
|
|
"completions/max_terminated_length": 1638.0,
|
|
"completions/mean_length": 349.03125,
|
|
"completions/mean_terminated_length": 349.03125,
|
|
"completions/min_length": 2.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0125,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 25.123981475830078,
|
|
"learning_rate": 8e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 551312.0,
|
|
"reward": 0.2671874761581421,
|
|
"reward_std": 0.34218358993530273,
|
|
"rewards/format_reward/mean": 0.328125,
|
|
"rewards/format_reward/std": 0.25539806485176086,
|
|
"rewards/mcq_exact_match_reward/mean": 0.234375,
|
|
"rewards/mcq_exact_match_reward/std": 0.42695629596710205,
|
|
"step": 5
|
|
},
|
|
{
|
|
"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": 1693.0,
|
|
"completions/max_terminated_length": 1693.0,
|
|
"completions/mean_length": 357.921875,
|
|
"completions/mean_terminated_length": 357.921875,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.015,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 12.899702072143555,
|
|
"learning_rate": 1e-06,
|
|
"loss": -0.0,
|
|
"num_tokens": 670235.0,
|
|
"reward": 0.45624998211860657,
|
|
"reward_std": 0.40883612632751465,
|
|
"rewards/format_reward/mean": 0.34375,
|
|
"rewards/format_reward/std": 0.233588308095932,
|
|
"rewards/mcq_exact_match_reward/mean": 0.421875,
|
|
"rewards/mcq_exact_match_reward/std": 0.49776285886764526,
|
|
"step": 6
|
|
},
|
|
{
|
|
"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": 1128.0,
|
|
"completions/max_terminated_length": 1128.0,
|
|
"completions/mean_length": 248.90625,
|
|
"completions/mean_terminated_length": 248.90625,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0175,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 9.06348705291748,
|
|
"learning_rate": 9.99726628670463e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 786077.0,
|
|
"reward": 0.23671872913837433,
|
|
"reward_std": 0.3678078055381775,
|
|
"rewards/format_reward/mean": 0.3359375,
|
|
"rewards/format_reward/std": 0.25283610820770264,
|
|
"rewards/mcq_exact_match_reward/mean": 0.203125,
|
|
"rewards/mcq_exact_match_reward/std": 0.40550529956817627,
|
|
"step": 7
|
|
},
|
|
{
|
|
"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": 1530.0,
|
|
"completions/max_terminated_length": 1530.0,
|
|
"completions/mean_length": 493.984375,
|
|
"completions/mean_terminated_length": 493.984375,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.02,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 3.943413734436035,
|
|
"learning_rate": 9.989068136093872e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 897284.0,
|
|
"reward": 0.296875,
|
|
"reward_std": 0.45596975088119507,
|
|
"rewards/format_reward/mean": 0.3125,
|
|
"rewards/format_reward/std": 0.2745848298072815,
|
|
"rewards/mcq_exact_match_reward/mean": 0.265625,
|
|
"rewards/mcq_exact_match_reward/std": 0.44515693187713623,
|
|
"step": 8
|
|
},
|
|
{
|
|
"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": 1452.0,
|
|
"completions/max_terminated_length": 1452.0,
|
|
"completions/mean_length": 219.5625,
|
|
"completions/mean_terminated_length": 219.5625,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0225,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 15.49644947052002,
|
|
"learning_rate": 9.975414512725056e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 994448.0,
|
|
"reward": 0.38515621423721313,
|
|
"reward_std": 0.35887449979782104,
|
|
"rewards/format_reward/mean": 0.4140625,
|
|
"rewards/format_reward/std": 0.209963858127594,
|
|
"rewards/mcq_exact_match_reward/mean": 0.34375,
|
|
"rewards/mcq_exact_match_reward/std": 0.4787135720252991,
|
|
"step": 9
|
|
},
|
|
{
|
|
"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": 1492.0,
|
|
"completions/max_terminated_length": 1492.0,
|
|
"completions/mean_length": 414.5625,
|
|
"completions/mean_terminated_length": 414.5625,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.025,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 17.415740966796875,
|
|
"learning_rate": 9.956320346634875e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 1119476.0,
|
|
"reward": 0.44140625,
|
|
"reward_std": 0.4423086643218994,
|
|
"rewards/format_reward/mean": 0.3515625,
|
|
"rewards/format_reward/std": 0.3044600784778595,
|
|
"rewards/mcq_exact_match_reward/mean": 0.40625,
|
|
"rewards/mcq_exact_match_reward/std": 0.49501484632492065,
|
|
"step": 10
|
|
},
|
|
{
|
|
"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": 939.0,
|
|
"completions/max_terminated_length": 939.0,
|
|
"completions/mean_length": 261.59375,
|
|
"completions/mean_terminated_length": 261.59375,
|
|
"completions/min_length": 3.0,
|
|
"completions/min_terminated_length": 3.0,
|
|
"epoch": 0.0275,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 21.028289794921875,
|
|
"learning_rate": 9.931806517013612e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 1218250.0,
|
|
"reward": 0.3359375,
|
|
"reward_std": 0.40134403109550476,
|
|
"rewards/format_reward/mean": 0.390625,
|
|
"rewards/format_reward/std": 0.3145764470100403,
|
|
"rewards/mcq_exact_match_reward/mean": 0.296875,
|
|
"rewards/mcq_exact_match_reward/std": 0.4604927599430084,
|
|
"step": 11
|
|
},
|
|
{
|
|
"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": 1675.0,
|
|
"completions/max_terminated_length": 1675.0,
|
|
"completions/mean_length": 503.96875,
|
|
"completions/mean_terminated_length": 503.96875,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.03,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 5.3941802978515625,
|
|
"learning_rate": 9.901899829374047e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 1353544.0,
|
|
"reward": 0.32343748211860657,
|
|
"reward_std": 0.39192718267440796,
|
|
"rewards/format_reward/mean": 0.421875,
|
|
"rewards/format_reward/std": 0.3238992393016815,
|
|
"rewards/mcq_exact_match_reward/mean": 0.28125,
|
|
"rewards/mcq_exact_match_reward/std": 0.4531635046005249,
|
|
"step": 12
|
|
},
|
|
{
|
|
"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": 0.03125,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 2036.0,
|
|
"completions/mean_length": 615.359375,
|
|
"completions/mean_terminated_length": 569.1451416015625,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0325,
|
|
"frac_reward_zero_std": 0.25,
|
|
"grad_norm": 3.4431583881378174,
|
|
"learning_rate": 9.866632986240029e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 1481711.0,
|
|
"reward": 0.29140621423721313,
|
|
"reward_std": 0.3188745379447937,
|
|
"rewards/format_reward/mean": 0.4140625,
|
|
"rewards/format_reward/std": 0.209963858127594,
|
|
"rewards/mcq_exact_match_reward/mean": 0.25,
|
|
"rewards/mcq_exact_match_reward/std": 0.4364357888698578,
|
|
"step": 13
|
|
},
|
|
{
|
|
"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": 0.03125,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1428.0,
|
|
"completions/mean_length": 489.390625,
|
|
"completions/mean_terminated_length": 439.1128845214844,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.035,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 4.944615840911865,
|
|
"learning_rate": 9.826044551386742e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 1601000.0,
|
|
"reward": 0.3140624761581421,
|
|
"reward_std": 0.3419710397720337,
|
|
"rewards/format_reward/mean": 0.328125,
|
|
"rewards/format_reward/std": 0.3238992393016815,
|
|
"rewards/mcq_exact_match_reward/mean": 0.28125,
|
|
"rewards/mcq_exact_match_reward/std": 0.4531635046005249,
|
|
"step": 14
|
|
},
|
|
{
|
|
"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": 0.046875,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1766.0,
|
|
"completions/mean_length": 571.1875,
|
|
"completions/mean_terminated_length": 498.5573425292969,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0375,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 3.2745373249053955,
|
|
"learning_rate": 9.780178907671788e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 1731604.0,
|
|
"reward": 0.2867187261581421,
|
|
"reward_std": 0.40185993909835815,
|
|
"rewards/format_reward/mean": 0.3671875,
|
|
"rewards/format_reward/std": 0.23974503576755524,
|
|
"rewards/mcq_exact_match_reward/mean": 0.25,
|
|
"rewards/mcq_exact_match_reward/std": 0.4364357888698578,
|
|
"step": 15
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1636.0,
|
|
"completions/mean_length": 601.015625,
|
|
"completions/mean_terminated_length": 578.0476684570312,
|
|
"completions/min_length": 17.0,
|
|
"completions/min_terminated_length": 17.0,
|
|
"epoch": 0.04,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 2.9144675731658936,
|
|
"learning_rate": 9.729086208503173e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 1859949.0,
|
|
"reward": 0.38749998807907104,
|
|
"reward_std": 0.43986088037490845,
|
|
"rewards/format_reward/mean": 0.4375,
|
|
"rewards/format_reward/std": 0.2597312331199646,
|
|
"rewards/mcq_exact_match_reward/mean": 0.34375,
|
|
"rewards/mcq_exact_match_reward/std": 0.4787135720252991,
|
|
"step": 16
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1056.0,
|
|
"completions/mean_length": 380.1875,
|
|
"completions/mean_terminated_length": 353.71429443359375,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0425,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 1.8824231624603271,
|
|
"learning_rate": 9.672822322997304e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 1970809.0,
|
|
"reward": 0.567187488079071,
|
|
"reward_std": 0.4544362723827362,
|
|
"rewards/format_reward/mean": 0.359375,
|
|
"rewards/format_reward/std": 0.28824523091316223,
|
|
"rewards/mcq_exact_match_reward/mean": 0.53125,
|
|
"rewards/mcq_exact_match_reward/std": 0.5029674172401428,
|
|
"step": 17
|
|
},
|
|
{
|
|
"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": 1673.0,
|
|
"completions/max_terminated_length": 1673.0,
|
|
"completions/mean_length": 516.09375,
|
|
"completions/mean_terminated_length": 516.09375,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.045,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 2.5847556591033936,
|
|
"learning_rate": 9.611448774886923e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 2115567.0,
|
|
"reward": 0.4624999761581421,
|
|
"reward_std": 0.33887431025505066,
|
|
"rewards/format_reward/mean": 0.40625,
|
|
"rewards/format_reward/std": 0.233588308095932,
|
|
"rewards/mcq_exact_match_reward/mean": 0.421875,
|
|
"rewards/mcq_exact_match_reward/std": 0.49776285886764526,
|
|
"step": 18
|
|
},
|
|
{
|
|
"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": 1511.0,
|
|
"completions/max_terminated_length": 1511.0,
|
|
"completions/mean_length": 406.84375,
|
|
"completions/mean_terminated_length": 406.84375,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0475,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 1.9844298362731934,
|
|
"learning_rate": 9.545032675245813e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 2224709.0,
|
|
"reward": 0.4781249761581421,
|
|
"reward_std": 0.3849598169326782,
|
|
"rewards/format_reward/mean": 0.40625,
|
|
"rewards/format_reward/std": 0.265398770570755,
|
|
"rewards/mcq_exact_match_reward/mean": 0.4375,
|
|
"rewards/mcq_exact_match_reward/std": 0.5,
|
|
"step": 19
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1541.0,
|
|
"completions/mean_length": 444.796875,
|
|
"completions/mean_terminated_length": 419.3492431640625,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.05,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 6.835641384124756,
|
|
"learning_rate": 9.473646649103817e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 2350360.0,
|
|
"reward": 0.46406248211860657,
|
|
"reward_std": 0.3951150178909302,
|
|
"rewards/format_reward/mean": 0.421875,
|
|
"rewards/format_reward/std": 0.25539806485176086,
|
|
"rewards/mcq_exact_match_reward/mean": 0.421875,
|
|
"rewards/mcq_exact_match_reward/std": 0.49776285886764526,
|
|
"step": 20
|
|
},
|
|
{
|
|
"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": 1713.0,
|
|
"completions/max_terminated_length": 1713.0,
|
|
"completions/mean_length": 531.78125,
|
|
"completions/mean_terminated_length": 531.78125,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0525,
|
|
"frac_reward_zero_std": 0.25,
|
|
"grad_norm": 2.273691177368164,
|
|
"learning_rate": 9.397368756032444e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 2476410.0,
|
|
"reward": 0.4195312261581421,
|
|
"reward_std": 0.19411513209342957,
|
|
"rewards/format_reward/mean": 0.4453125,
|
|
"rewards/format_reward/std": 0.2376670390367508,
|
|
"rewards/mcq_exact_match_reward/mean": 0.375,
|
|
"rewards/mcq_exact_match_reward/std": 0.48795005679130554,
|
|
"step": 21
|
|
},
|
|
{
|
|
"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": 1147.0,
|
|
"completions/max_terminated_length": 1147.0,
|
|
"completions/mean_length": 415.796875,
|
|
"completions/mean_terminated_length": 415.796875,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.055,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 5.067054748535156,
|
|
"learning_rate": 9.316282404787869e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 2610701.0,
|
|
"reward": 0.37109375,
|
|
"reward_std": 0.4409857988357544,
|
|
"rewards/format_reward/mean": 0.4296875,
|
|
"rewards/format_reward/std": 0.26528194546699524,
|
|
"rewards/mcq_exact_match_reward/mean": 0.328125,
|
|
"rewards/mcq_exact_match_reward/std": 0.4732423722743988,
|
|
"step": 22
|
|
},
|
|
{
|
|
"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": 1403.0,
|
|
"completions/max_terminated_length": 1403.0,
|
|
"completions/mean_length": 364.1875,
|
|
"completions/mean_terminated_length": 364.1875,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0575,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 13.165079116821289,
|
|
"learning_rate": 9.230476262104676e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 2738369.0,
|
|
"reward": 0.6960936784744263,
|
|
"reward_std": 0.44959956407546997,
|
|
"rewards/format_reward/mean": 0.3984375,
|
|
"rewards/format_reward/std": 0.22146137058734894,
|
|
"rewards/mcq_exact_match_reward/mean": 0.65625,
|
|
"rewards/mcq_exact_match_reward/std": 0.4787135720252991,
|
|
"step": 23
|
|
},
|
|
{
|
|
"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": 1887.0,
|
|
"completions/max_terminated_length": 1887.0,
|
|
"completions/mean_length": 396.875,
|
|
"completions/mean_terminated_length": 396.875,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.06,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 5.967652797698975,
|
|
"learning_rate": 9.1400441557401e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 2853985.0,
|
|
"reward": 0.49296873807907104,
|
|
"reward_std": 0.4339084029197693,
|
|
"rewards/format_reward/mean": 0.3984375,
|
|
"rewards/format_reward/std": 0.20275264978408813,
|
|
"rewards/mcq_exact_match_reward/mean": 0.453125,
|
|
"rewards/mcq_exact_match_reward/std": 0.501733124256134,
|
|
"step": 24
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1321.0,
|
|
"completions/mean_length": 527.609375,
|
|
"completions/mean_terminated_length": 503.4762268066406,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0625,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 5.454898357391357,
|
|
"learning_rate": 9.045084971874737e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 2987496.0,
|
|
"reward": 0.3882812261581421,
|
|
"reward_std": 0.4757910370826721,
|
|
"rewards/format_reward/mean": 0.4453125,
|
|
"rewards/format_reward/std": 0.26899561285972595,
|
|
"rewards/mcq_exact_match_reward/mean": 0.34375,
|
|
"rewards/mcq_exact_match_reward/std": 0.4787135720252991,
|
|
"step": 25
|
|
},
|
|
{
|
|
"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": 1604.0,
|
|
"completions/max_terminated_length": 1604.0,
|
|
"completions/mean_length": 433.25,
|
|
"completions/mean_terminated_length": 433.25,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.065,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 7.106506824493408,
|
|
"learning_rate": 8.945702546981968e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 3120200.0,
|
|
"reward": 0.47968748211860657,
|
|
"reward_std": 0.47255939245224,
|
|
"rewards/format_reward/mean": 0.421875,
|
|
"rewards/format_reward/std": 0.23935678601264954,
|
|
"rewards/mcq_exact_match_reward/mean": 0.4375,
|
|
"rewards/mcq_exact_match_reward/std": 0.5,
|
|
"step": 26
|
|
},
|
|
{
|
|
"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": 1906.0,
|
|
"completions/max_terminated_length": 1906.0,
|
|
"completions/mean_length": 574.765625,
|
|
"completions/mean_terminated_length": 574.765625,
|
|
"completions/min_length": 90.0,
|
|
"completions/min_terminated_length": 90.0,
|
|
"epoch": 0.0675,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 1.6191740036010742,
|
|
"learning_rate": 8.842005554284295e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 3249513.0,
|
|
"reward": 0.4820312261581421,
|
|
"reward_std": 0.4589237570762634,
|
|
"rewards/format_reward/mean": 0.4453125,
|
|
"rewards/format_reward/std": 0.2203386276960373,
|
|
"rewards/mcq_exact_match_reward/mean": 0.4375,
|
|
"rewards/mcq_exact_match_reward/std": 0.5,
|
|
"step": 27
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1529.0,
|
|
"completions/mean_length": 472.84375,
|
|
"completions/mean_terminated_length": 447.84130859375,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.07,
|
|
"frac_reward_zero_std": 0.25,
|
|
"grad_norm": 3.762904167175293,
|
|
"learning_rate": 8.734107384920769e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 3379007.0,
|
|
"reward": 0.3570312261581421,
|
|
"reward_std": 0.2997973561286926,
|
|
"rewards/format_reward/mean": 0.4453125,
|
|
"rewards/format_reward/std": 0.20152568817138672,
|
|
"rewards/mcq_exact_match_reward/mean": 0.3125,
|
|
"rewards/mcq_exact_match_reward/std": 0.467176616191864,
|
|
"step": 28
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1093.0,
|
|
"completions/mean_length": 472.203125,
|
|
"completions/mean_terminated_length": 447.19049072265625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.0725,
|
|
"frac_reward_zero_std": 0.0,
|
|
"grad_norm": 10.816900253295898,
|
|
"learning_rate": 8.622126023955445e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 3493564.0,
|
|
"reward": 0.5625,
|
|
"reward_std": 0.4124056100845337,
|
|
"rewards/format_reward/mean": 0.46875,
|
|
"rewards/format_reward/std": 0.265398770570755,
|
|
"rewards/mcq_exact_match_reward/mean": 0.515625,
|
|
"rewards/mcq_exact_match_reward/std": 0.5037065148353577,
|
|
"step": 29
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1644.0,
|
|
"completions/mean_length": 378.25,
|
|
"completions/mean_terminated_length": 351.7460632324219,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.075,
|
|
"frac_reward_zero_std": 0.25,
|
|
"grad_norm": 6.790505886077881,
|
|
"learning_rate": 8.506183921362442e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 3593036.0,
|
|
"reward": 0.703125,
|
|
"reward_std": 0.25315535068511963,
|
|
"rewards/format_reward/mean": 0.46875,
|
|
"rewards/format_reward/std": 0.233588308095932,
|
|
"rewards/mcq_exact_match_reward/mean": 0.65625,
|
|
"rewards/mcq_exact_match_reward/std": 0.4787135720252991,
|
|
"step": 30
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1901.0,
|
|
"completions/mean_length": 449.421875,
|
|
"completions/mean_terminated_length": 424.0476379394531,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0775,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 6.935174465179443,
|
|
"learning_rate": 8.386407858128706e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 3710671.0,
|
|
"reward": 0.40625,
|
|
"reward_std": 0.29695966839790344,
|
|
"rewards/format_reward/mean": 0.46875,
|
|
"rewards/format_reward/std": 0.19669894874095917,
|
|
"rewards/mcq_exact_match_reward/mean": 0.359375,
|
|
"rewards/mcq_exact_match_reward/std": 0.4836103618144989,
|
|
"step": 31
|
|
},
|
|
{
|
|
"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": 1866.0,
|
|
"completions/max_terminated_length": 1866.0,
|
|
"completions/mean_length": 242.046875,
|
|
"completions/mean_terminated_length": 242.046875,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.08,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 11.75405502319336,
|
|
"learning_rate": 8.262928807620843e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 3812746.0,
|
|
"reward": 0.6046874523162842,
|
|
"reward_std": 0.3665615916252136,
|
|
"rewards/format_reward/mean": 0.578125,
|
|
"rewards/format_reward/std": 0.2847827076911926,
|
|
"rewards/mcq_exact_match_reward/mean": 0.546875,
|
|
"rewards/mcq_exact_match_reward/std": 0.501733124256134,
|
|
"step": 32
|
|
},
|
|
{
|
|
"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": 0.03125,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1017.0,
|
|
"completions/mean_length": 385.109375,
|
|
"completions/mean_terminated_length": 331.4677429199219,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0825,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 9.297252655029297,
|
|
"learning_rate": 8.135881792367685e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 3935601.0,
|
|
"reward": 0.36015623807907104,
|
|
"reward_std": 0.3222334682941437,
|
|
"rewards/format_reward/mean": 0.4765625,
|
|
"rewards/format_reward/std": 0.25865477323532104,
|
|
"rewards/mcq_exact_match_reward/mean": 0.3125,
|
|
"rewards/mcq_exact_match_reward/std": 0.467176616191864,
|
|
"step": 33
|
|
},
|
|
{
|
|
"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": 0.03125,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1522.0,
|
|
"completions/mean_length": 492.140625,
|
|
"completions/mean_terminated_length": 441.95159912109375,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.085,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 9.278347969055176,
|
|
"learning_rate": 8.005405736415125e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 4055274.0,
|
|
"reward": 0.4937499761581421,
|
|
"reward_std": 0.3744146227836609,
|
|
"rewards/format_reward/mean": 0.5625,
|
|
"rewards/format_reward/std": 0.22712838649749756,
|
|
"rewards/mcq_exact_match_reward/mean": 0.4375,
|
|
"rewards/mcq_exact_match_reward/std": 0.5,
|
|
"step": 34
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1146.0,
|
|
"completions/mean_length": 302.875,
|
|
"completions/mean_terminated_length": 275.17462158203125,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0875,
|
|
"frac_reward_zero_std": 0.25,
|
|
"grad_norm": 5.606284141540527,
|
|
"learning_rate": 7.871643313414718e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 4176362.0,
|
|
"reward": 0.47187501192092896,
|
|
"reward_std": 0.24273529648780823,
|
|
"rewards/format_reward/mean": 0.5,
|
|
"rewards/format_reward/std": 0.34503278136253357,
|
|
"rewards/mcq_exact_match_reward/mean": 0.421875,
|
|
"rewards/mcq_exact_match_reward/std": 0.49776285886764526,
|
|
"step": 35
|
|
},
|
|
{
|
|
"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": 1482.0,
|
|
"completions/max_terminated_length": 1482.0,
|
|
"completions/mean_length": 410.40625,
|
|
"completions/mean_terminated_length": 410.40625,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.09,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 9.850747108459473,
|
|
"learning_rate": 7.734740790612136e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 4292476.0,
|
|
"reward": 0.6898437738418579,
|
|
"reward_std": 0.22759628295898438,
|
|
"rewards/format_reward/mean": 0.6484375,
|
|
"rewards/format_reward/std": 0.29113471508026123,
|
|
"rewards/mcq_exact_match_reward/mean": 0.625,
|
|
"rewards/mcq_exact_match_reward/std": 0.48795005679130554,
|
|
"step": 36
|
|
},
|
|
{
|
|
"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": 1006.0,
|
|
"completions/max_terminated_length": 1006.0,
|
|
"completions/mean_length": 126.859375,
|
|
"completions/mean_terminated_length": 126.859375,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0925,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 20.594633102416992,
|
|
"learning_rate": 7.594847868906076e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 4413923.0,
|
|
"reward": 0.5546875,
|
|
"reward_std": 0.30506113171577454,
|
|
"rewards/format_reward/mean": 0.703125,
|
|
"rewards/format_reward/std": 0.2777281701564789,
|
|
"rewards/mcq_exact_match_reward/mean": 0.484375,
|
|
"rewards/mcq_exact_match_reward/std": 0.5037065148353577,
|
|
"step": 37
|
|
},
|
|
{
|
|
"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": 1447.0,
|
|
"completions/max_terminated_length": 1447.0,
|
|
"completions/mean_length": 381.84375,
|
|
"completions/mean_terminated_length": 381.84375,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.095,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 4.137030124664307,
|
|
"learning_rate": 7.452117519152541e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 4531905.0,
|
|
"reward": 0.44062501192092896,
|
|
"reward_std": 0.25260162353515625,
|
|
"rewards/format_reward/mean": 0.65625,
|
|
"rewards/format_reward/std": 0.265398770570755,
|
|
"rewards/mcq_exact_match_reward/mean": 0.375,
|
|
"rewards/mcq_exact_match_reward/std": 0.48795005679130554,
|
|
"step": 38
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1119.0,
|
|
"completions/mean_length": 270.359375,
|
|
"completions/mean_terminated_length": 242.1428680419922,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.0975,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 20.815427780151367,
|
|
"learning_rate": 7.306705814893439e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 4643792.0,
|
|
"reward": 0.41874998807907104,
|
|
"reward_std": 0.4047183394432068,
|
|
"rewards/format_reward/mean": 0.75,
|
|
"rewards/format_reward/std": 0.29546841979026794,
|
|
"rewards/mcq_exact_match_reward/mean": 0.34375,
|
|
"rewards/mcq_exact_match_reward/std": 0.4787135720252991,
|
|
"step": 39
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 1158.0,
|
|
"completions/mean_length": 153.5625,
|
|
"completions/mean_terminated_length": 123.49207305908203,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.1,
|
|
"frac_reward_zero_std": 0.125,
|
|
"grad_norm": 16.95915412902832,
|
|
"learning_rate": 7.158771761692464e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 4758172.0,
|
|
"reward": 0.5648437738418579,
|
|
"reward_std": 0.20101891458034515,
|
|
"rewards/format_reward/mean": 0.8046875,
|
|
"rewards/format_reward/std": 0.2615155577659607,
|
|
"rewards/mcq_exact_match_reward/mean": 0.484375,
|
|
"rewards/mcq_exact_match_reward/std": 0.5037065148353577,
|
|
"step": 40
|
|
},
|
|
{
|
|
"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": 604.0,
|
|
"completions/max_terminated_length": 604.0,
|
|
"completions/mean_length": 59.65625,
|
|
"completions/mean_terminated_length": 59.65625,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.1025,
|
|
"frac_reward_zero_std": 0.25,
|
|
"grad_norm": 19.127452850341797,
|
|
"learning_rate": 7.008477123264847e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 4861310.0,
|
|
"reward": 0.48750001192092896,
|
|
"reward_std": 0.32531416416168213,
|
|
"rewards/format_reward/mean": 0.8125,
|
|
"rewards/format_reward/std": 0.24397502839565277,
|
|
"rewards/mcq_exact_match_reward/mean": 0.40625,
|
|
"rewards/mcq_exact_match_reward/std": 0.49501484632492065,
|
|
"step": 41
|
|
},
|
|
{
|
|
"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": 126.0,
|
|
"completions/max_terminated_length": 126.0,
|
|
"completions/mean_length": 18.4375,
|
|
"completions/mean_terminated_length": 18.4375,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.105,
|
|
"frac_reward_zero_std": 0.375,
|
|
"grad_norm": 19.860843658447266,
|
|
"learning_rate": 6.855986244591103e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 4955818.0,
|
|
"reward": 0.6742187738418579,
|
|
"reward_std": 0.18233226239681244,
|
|
"rewards/format_reward/mean": 0.8046875,
|
|
"rewards/format_reward/std": 0.2615155577659607,
|
|
"rewards/mcq_exact_match_reward/mean": 0.59375,
|
|
"rewards/mcq_exact_match_reward/std": 0.49501484632492065,
|
|
"step": 42
|
|
},
|
|
{
|
|
"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": 522.0,
|
|
"completions/max_terminated_length": 522.0,
|
|
"completions/mean_length": 38.796875,
|
|
"completions/mean_terminated_length": 38.796875,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.1075,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 11.83443832397461,
|
|
"learning_rate": 6.701465872208216e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 5052005.0,
|
|
"reward": 0.5335937738418579,
|
|
"reward_std": 0.2508957087993622,
|
|
"rewards/format_reward/mean": 0.8046875,
|
|
"rewards/format_reward/std": 0.24587368965148926,
|
|
"rewards/mcq_exact_match_reward/mean": 0.453125,
|
|
"rewards/mcq_exact_match_reward/std": 0.501733124256134,
|
|
"step": 43
|
|
},
|
|
{
|
|
"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": 865.0,
|
|
"completions/max_terminated_length": 865.0,
|
|
"completions/mean_length": 102.421875,
|
|
"completions/mean_terminated_length": 102.421875,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.11,
|
|
"frac_reward_zero_std": 0.375,
|
|
"grad_norm": 7.313492774963379,
|
|
"learning_rate": 6.545084971874736e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 5165008.0,
|
|
"reward": 0.6687500476837158,
|
|
"reward_std": 0.2736644446849823,
|
|
"rewards/format_reward/mean": 0.90625,
|
|
"rewards/format_reward/std": 0.233588308095932,
|
|
"rewards/mcq_exact_match_reward/mean": 0.578125,
|
|
"rewards/mcq_exact_match_reward/std": 0.49776285886764526,
|
|
"step": 44
|
|
},
|
|
{
|
|
"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": 1833.0,
|
|
"completions/max_terminated_length": 1833.0,
|
|
"completions/mean_length": 95.1875,
|
|
"completions/mean_terminated_length": 95.1875,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.1125,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 20.023258209228516,
|
|
"learning_rate": 6.387014543809223e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 5266132.0,
|
|
"reward": 0.47265625,
|
|
"reward_std": 0.1391630470752716,
|
|
"rewards/format_reward/mean": 0.8203125,
|
|
"rewards/format_reward/std": 0.32571274042129517,
|
|
"rewards/mcq_exact_match_reward/mean": 0.390625,
|
|
"rewards/mcq_exact_match_reward/std": 0.4917473793029785,
|
|
"step": 45
|
|
},
|
|
{
|
|
"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": 1079.0,
|
|
"completions/max_terminated_length": 1079.0,
|
|
"completions/mean_length": 71.578125,
|
|
"completions/mean_terminated_length": 71.578125,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.115,
|
|
"frac_reward_zero_std": 0.25,
|
|
"grad_norm": 19.259967803955078,
|
|
"learning_rate": 6.227427435703995e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 5373785.0,
|
|
"reward": 0.5859375,
|
|
"reward_std": 0.34790486097335815,
|
|
"rewards/format_reward/mean": 0.859375,
|
|
"rewards/format_reward/std": 0.2741328477859497,
|
|
"rewards/mcq_exact_match_reward/mean": 0.5,
|
|
"rewards/mcq_exact_match_reward/std": 0.5039526224136353,
|
|
"step": 46
|
|
},
|
|
{
|
|
"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": 947.0,
|
|
"completions/max_terminated_length": 947.0,
|
|
"completions/mean_length": 128.34375,
|
|
"completions/mean_terminated_length": 128.34375,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.1175,
|
|
"frac_reward_zero_std": 0.375,
|
|
"grad_norm": 28.760831832885742,
|
|
"learning_rate": 6.066498153718734e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 5500999.0,
|
|
"reward": 0.557812511920929,
|
|
"reward_std": 0.24416987597942352,
|
|
"rewards/format_reward/mean": 0.890625,
|
|
"rewards/format_reward/std": 0.2083333432674408,
|
|
"rewards/mcq_exact_match_reward/mean": 0.46875,
|
|
"rewards/mcq_exact_match_reward/std": 0.5029674172401428,
|
|
"step": 47
|
|
},
|
|
{
|
|
"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": 638.0,
|
|
"completions/max_terminated_length": 638.0,
|
|
"completions/mean_length": 38.078125,
|
|
"completions/mean_terminated_length": 38.078125,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.12,
|
|
"frac_reward_zero_std": 0.375,
|
|
"grad_norm": 8.37263298034668,
|
|
"learning_rate": 5.90440267166055e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 5585012.0,
|
|
"reward": 0.47421878576278687,
|
|
"reward_std": 0.20632557570934296,
|
|
"rewards/format_reward/mean": 0.9921875,
|
|
"rewards/format_reward/std": 0.0625,
|
|
"rewards/mcq_exact_match_reward/mean": 0.375,
|
|
"rewards/mcq_exact_match_reward/std": 0.48795005679130554,
|
|
"step": 48
|
|
},
|
|
{
|
|
"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": 533.0,
|
|
"completions/max_terminated_length": 533.0,
|
|
"completions/mean_length": 21.296875,
|
|
"completions/mean_terminated_length": 21.296875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1225,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 15.718689918518066,
|
|
"learning_rate": 5.741318238559209e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 5652815.0,
|
|
"reward": 0.8023437857627869,
|
|
"reward_std": 0.16939528286457062,
|
|
"rewards/format_reward/mean": 0.9921875,
|
|
"rewards/format_reward/std": 0.0625,
|
|
"rewards/mcq_exact_match_reward/mean": 0.703125,
|
|
"rewards/mcq_exact_match_reward/std": 0.4604927599430084,
|
|
"step": 49
|
|
},
|
|
{
|
|
"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": 396.0,
|
|
"completions/max_terminated_length": 396.0,
|
|
"completions/mean_length": 34.1875,
|
|
"completions/mean_terminated_length": 34.1875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.125,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 7.805715084075928,
|
|
"learning_rate": 5.577423184847931e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 5754403.0,
|
|
"reward": 0.69140625,
|
|
"reward_std": 0.18080364167690277,
|
|
"rewards/format_reward/mean": 0.9765625,
|
|
"rewards/format_reward/std": 0.13886408507823944,
|
|
"rewards/mcq_exact_match_reward/mean": 0.59375,
|
|
"rewards/mcq_exact_match_reward/std": 0.49501484632492065,
|
|
"step": 50
|
|
},
|
|
{
|
|
"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": 438.0,
|
|
"completions/max_terminated_length": 438.0,
|
|
"completions/mean_length": 19.296875,
|
|
"completions/mean_terminated_length": 19.296875,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.1275,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 19.45261001586914,
|
|
"learning_rate": 5.412896727361662e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 5845782.0,
|
|
"reward": 0.7835937738418579,
|
|
"reward_std": 0.24337685108184814,
|
|
"rewards/format_reward/mean": 0.9609375,
|
|
"rewards/format_reward/std": 0.18483558297157288,
|
|
"rewards/mcq_exact_match_reward/mean": 0.6875,
|
|
"rewards/mcq_exact_match_reward/std": 0.467176616191864,
|
|
"step": 51
|
|
},
|
|
{
|
|
"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": 13.0,
|
|
"completions/max_terminated_length": 13.0,
|
|
"completions/mean_length": 11.359375,
|
|
"completions/mean_terminated_length": 11.359375,
|
|
"completions/min_length": 6.0,
|
|
"completions/min_terminated_length": 6.0,
|
|
"epoch": 0.13,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 10.836705207824707,
|
|
"learning_rate": 5.247918773366111e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 5928429.0,
|
|
"reward": 0.671875,
|
|
"reward_std": 0.10205793380737305,
|
|
"rewards/format_reward/mean": 0.9375,
|
|
"rewards/format_reward/std": 0.1666666716337204,
|
|
"rewards/mcq_exact_match_reward/mean": 0.578125,
|
|
"rewards/mcq_exact_match_reward/std": 0.49776285886764526,
|
|
"step": 52
|
|
},
|
|
{
|
|
"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": 308.0,
|
|
"completions/max_terminated_length": 308.0,
|
|
"completions/mean_length": 20.9375,
|
|
"completions/mean_terminated_length": 20.9375,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1325,
|
|
"frac_reward_zero_std": 0.375,
|
|
"grad_norm": 17.49638557434082,
|
|
"learning_rate": 5.082669723831793e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 6053041.0,
|
|
"reward": 0.38046878576278687,
|
|
"reward_std": 0.2242286652326584,
|
|
"rewards/format_reward/mean": 0.9921875,
|
|
"rewards/format_reward/std": 0.0625,
|
|
"rewards/mcq_exact_match_reward/mean": 0.28125,
|
|
"rewards/mcq_exact_match_reward/std": 0.4531635046005249,
|
|
"step": 53
|
|
},
|
|
{
|
|
"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": 177.0,
|
|
"completions/max_terminated_length": 177.0,
|
|
"completions/mean_length": 18.125,
|
|
"completions/mean_terminated_length": 18.125,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.135,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 5.867814064025879,
|
|
"learning_rate": 4.917330276168208e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 6143761.0,
|
|
"reward": 0.7867187857627869,
|
|
"reward_std": 0.10862661898136139,
|
|
"rewards/format_reward/mean": 0.9921875,
|
|
"rewards/format_reward/std": 0.0625,
|
|
"rewards/mcq_exact_match_reward/mean": 0.6875,
|
|
"rewards/mcq_exact_match_reward/std": 0.467176616191864,
|
|
"step": 54
|
|
},
|
|
{
|
|
"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": 276.0,
|
|
"completions/max_terminated_length": 276.0,
|
|
"completions/mean_length": 26.875,
|
|
"completions/mean_terminated_length": 26.875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1375,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 6.905524253845215,
|
|
"learning_rate": 4.752081226633888e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 6237969.0,
|
|
"reward": 0.6468750238418579,
|
|
"reward_std": 0.1530819982290268,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.546875,
|
|
"rewards/mcq_exact_match_reward/std": 0.501733124256134,
|
|
"step": 55
|
|
},
|
|
{
|
|
"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": 429.0,
|
|
"completions/max_terminated_length": 429.0,
|
|
"completions/mean_length": 32.140625,
|
|
"completions/mean_terminated_length": 32.140625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.14,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 14.239409446716309,
|
|
"learning_rate": 4.5871032726383385e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 6324170.0,
|
|
"reward": 0.551562488079071,
|
|
"reward_std": 0.15467961132526398,
|
|
"rewards/format_reward/mean": 0.984375,
|
|
"rewards/format_reward/std": 0.08768405020236969,
|
|
"rewards/mcq_exact_match_reward/mean": 0.453125,
|
|
"rewards/mcq_exact_match_reward/std": 0.501733124256134,
|
|
"step": 56
|
|
},
|
|
{
|
|
"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": 382.0,
|
|
"completions/max_terminated_length": 382.0,
|
|
"completions/mean_length": 17.90625,
|
|
"completions/mean_terminated_length": 17.90625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1425,
|
|
"frac_reward_zero_std": 0.25,
|
|
"grad_norm": 14.875107765197754,
|
|
"learning_rate": 4.4225768151520694e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 6433140.0,
|
|
"reward": 0.6781250238418579,
|
|
"reward_std": 0.36507582664489746,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.578125,
|
|
"rewards/mcq_exact_match_reward/std": 0.49776285886764526,
|
|
"step": 57
|
|
},
|
|
{
|
|
"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": 147.0,
|
|
"completions/max_terminated_length": 147.0,
|
|
"completions/mean_length": 16.390625,
|
|
"completions/mean_terminated_length": 16.390625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.145,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 8.728854179382324,
|
|
"learning_rate": 4.258681761440789e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 6518901.0,
|
|
"reward": 0.7718750238418579,
|
|
"reward_std": 0.1530819982290268,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.671875,
|
|
"rewards/mcq_exact_match_reward/std": 0.4732423722743988,
|
|
"step": 58
|
|
},
|
|
{
|
|
"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": 1584.0,
|
|
"completions/max_terminated_length": 1584.0,
|
|
"completions/mean_length": 55.421875,
|
|
"completions/mean_terminated_length": 55.421875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1475,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 12.296089172363281,
|
|
"learning_rate": 4.095597328339452e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 6611552.0,
|
|
"reward": 0.753125011920929,
|
|
"reward_std": 0.24601811170578003,
|
|
"rewards/format_reward/mean": 0.96875,
|
|
"rewards/format_reward/std": 0.17536810040473938,
|
|
"rewards/mcq_exact_match_reward/mean": 0.65625,
|
|
"rewards/mcq_exact_match_reward/std": 0.4787135720252991,
|
|
"step": 59
|
|
},
|
|
{
|
|
"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": 12.0,
|
|
"completions/max_terminated_length": 12.0,
|
|
"completions/mean_length": 12.0,
|
|
"completions/mean_terminated_length": 12.0,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.15,
|
|
"frac_reward_zero_std": 1.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 3.9335018462812664e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 6700512.0,
|
|
"reward": 0.7250000238418579,
|
|
"reward_std": 0.0,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.625,
|
|
"rewards/mcq_exact_match_reward/std": 0.48795005679130554,
|
|
"step": 60
|
|
},
|
|
{
|
|
"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": 404.0,
|
|
"completions/max_terminated_length": 404.0,
|
|
"completions/mean_length": 21.3125,
|
|
"completions/mean_terminated_length": 21.3125,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1525,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 9.406328201293945,
|
|
"learning_rate": 3.772572564296004e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 6785340.0,
|
|
"reward": 0.4906250238418579,
|
|
"reward_std": 0.16887323558330536,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.390625,
|
|
"rewards/mcq_exact_match_reward/std": 0.4917473793029785,
|
|
"step": 61
|
|
},
|
|
{
|
|
"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": 13.0,
|
|
"completions/max_terminated_length": 13.0,
|
|
"completions/mean_length": 12.046875,
|
|
"completions/mean_terminated_length": 12.046875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.155,
|
|
"frac_reward_zero_std": 0.875,
|
|
"grad_norm": 3.72755765914917,
|
|
"learning_rate": 3.612985456190778e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 6893447.0,
|
|
"reward": 0.8343750238418579,
|
|
"reward_std": 0.04419417306780815,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.734375,
|
|
"rewards/mcq_exact_match_reward/std": 0.44515693187713623,
|
|
"step": 62
|
|
},
|
|
{
|
|
"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": 600.0,
|
|
"completions/max_terminated_length": 600.0,
|
|
"completions/mean_length": 28.1875,
|
|
"completions/mean_terminated_length": 28.1875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1575,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 10.368932723999023,
|
|
"learning_rate": 3.454915028125263e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 6985387.0,
|
|
"reward": 0.8812500238418579,
|
|
"reward_std": 0.2177756428718567,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.78125,
|
|
"rewards/mcq_exact_match_reward/std": 0.4166666865348816,
|
|
"step": 63
|
|
},
|
|
{
|
|
"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": 13.0,
|
|
"completions/max_terminated_length": 13.0,
|
|
"completions/mean_length": 12.15625,
|
|
"completions/mean_terminated_length": 12.15625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.16,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 12.464853286743164,
|
|
"learning_rate": 3.2985341277917846e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 7078181.0,
|
|
"reward": 0.6296875476837158,
|
|
"reward_std": 0.1776757538318634,
|
|
"rewards/format_reward/mean": 0.984375,
|
|
"rewards/format_reward/std": 0.08768405020236969,
|
|
"rewards/mcq_exact_match_reward/mean": 0.53125,
|
|
"rewards/mcq_exact_match_reward/std": 0.5029674172401428,
|
|
"step": 64
|
|
},
|
|
{
|
|
"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": 114.0,
|
|
"completions/max_terminated_length": 114.0,
|
|
"completions/mean_length": 13.8125,
|
|
"completions/mean_terminated_length": 13.8125,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1625,
|
|
"frac_reward_zero_std": 0.875,
|
|
"grad_norm": 6.173940181732178,
|
|
"learning_rate": 3.1440137554088953e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 7178929.0,
|
|
"reward": 0.7242187261581421,
|
|
"reward_std": 0.00220970856025815,
|
|
"rewards/format_reward/mean": 0.9921875,
|
|
"rewards/format_reward/std": 0.0625,
|
|
"rewards/mcq_exact_match_reward/mean": 0.625,
|
|
"rewards/mcq_exact_match_reward/std": 0.48795005679130554,
|
|
"step": 65
|
|
},
|
|
{
|
|
"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": 380.0,
|
|
"completions/max_terminated_length": 380.0,
|
|
"completions/mean_length": 22.984375,
|
|
"completions/mean_terminated_length": 22.984375,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.165,
|
|
"frac_reward_zero_std": 0.875,
|
|
"grad_norm": 5.9279961585998535,
|
|
"learning_rate": 2.9915228767351535e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 7263416.0,
|
|
"reward": 0.9125000238418579,
|
|
"reward_std": 0.06681530922651291,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.8125,
|
|
"rewards/mcq_exact_match_reward/std": 0.39339789748191833,
|
|
"step": 66
|
|
},
|
|
{
|
|
"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": 179.0,
|
|
"completions/max_terminated_length": 179.0,
|
|
"completions/mean_length": 24.0,
|
|
"completions/mean_terminated_length": 24.0,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1675,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 10.434481620788574,
|
|
"learning_rate": 2.841228238307536e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 7358456.0,
|
|
"reward": 0.7718750238418579,
|
|
"reward_std": 0.11100947856903076,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.671875,
|
|
"rewards/mcq_exact_match_reward/std": 0.4732423722743988,
|
|
"step": 67
|
|
},
|
|
{
|
|
"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": 20.0,
|
|
"completions/max_terminated_length": 20.0,
|
|
"completions/mean_length": 12.296875,
|
|
"completions/mean_terminated_length": 12.296875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.17,
|
|
"frac_reward_zero_std": 1.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 2.6932941851065615e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 7466395.0,
|
|
"reward": 0.9750000238418579,
|
|
"reward_std": 0.0,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.875,
|
|
"rewards/mcq_exact_match_reward/std": 0.3333333432674408,
|
|
"step": 68
|
|
},
|
|
{
|
|
"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": 539.0,
|
|
"completions/max_terminated_length": 539.0,
|
|
"completions/mean_length": 54.09375,
|
|
"completions/mean_terminated_length": 54.09375,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1725,
|
|
"frac_reward_zero_std": 0.375,
|
|
"grad_norm": 10.782182693481445,
|
|
"learning_rate": 2.547882480847461e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 7536617.0,
|
|
"reward": 0.7875000238418579,
|
|
"reward_std": 0.2619796395301819,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.6875,
|
|
"rewards/mcq_exact_match_reward/std": 0.467176616191864,
|
|
"step": 69
|
|
},
|
|
{
|
|
"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": 13.0,
|
|
"completions/max_terminated_length": 13.0,
|
|
"completions/mean_length": 12.015625,
|
|
"completions/mean_terminated_length": 12.015625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.175,
|
|
"frac_reward_zero_std": 0.875,
|
|
"grad_norm": 5.800319194793701,
|
|
"learning_rate": 2.4051521310939254e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 7625498.0,
|
|
"reward": 0.7406250238418579,
|
|
"reward_std": 0.04419417306780815,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.640625,
|
|
"rewards/mcq_exact_match_reward/std": 0.4836103618144989,
|
|
"step": 70
|
|
},
|
|
{
|
|
"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": 13.0,
|
|
"completions/max_terminated_length": 13.0,
|
|
"completions/mean_length": 12.03125,
|
|
"completions/mean_terminated_length": 12.03125,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1775,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 17.38172721862793,
|
|
"learning_rate": 2.2652592093878665e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 7713524.0,
|
|
"reward": 0.7093750238418579,
|
|
"reward_std": 0.1804211586713791,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.609375,
|
|
"rewards/mcq_exact_match_reward/std": 0.4917473793029785,
|
|
"step": 71
|
|
},
|
|
{
|
|
"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": 239.0,
|
|
"completions/max_terminated_length": 239.0,
|
|
"completions/mean_length": 15.609375,
|
|
"completions/mean_terminated_length": 15.609375,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.18,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 9.46229362487793,
|
|
"learning_rate": 2.128356686585282e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 7791747.0,
|
|
"reward": 0.7867187857627869,
|
|
"reward_std": 0.11109752953052521,
|
|
"rewards/format_reward/mean": 0.9921875,
|
|
"rewards/format_reward/std": 0.0625,
|
|
"rewards/mcq_exact_match_reward/mean": 0.6875,
|
|
"rewards/mcq_exact_match_reward/std": 0.467176616191864,
|
|
"step": 72
|
|
},
|
|
{
|
|
"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": 113.0,
|
|
"completions/max_terminated_length": 113.0,
|
|
"completions/mean_length": 13.578125,
|
|
"completions/mean_terminated_length": 13.578125,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1825,
|
|
"frac_reward_zero_std": 0.875,
|
|
"grad_norm": 9.975543975830078,
|
|
"learning_rate": 1.9945942635848745e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 7868464.0,
|
|
"reward": 0.9437500238418579,
|
|
"reward_std": 0.0578637570142746,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.84375,
|
|
"rewards/mcq_exact_match_reward/std": 0.36596253514289856,
|
|
"step": 73
|
|
},
|
|
{
|
|
"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": 891.0,
|
|
"completions/max_terminated_length": 891.0,
|
|
"completions/mean_length": 25.8125,
|
|
"completions/mean_terminated_length": 25.8125,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.185,
|
|
"frac_reward_zero_std": 0.875,
|
|
"grad_norm": 3.036553382873535,
|
|
"learning_rate": 1.8641182076323148e-07,
|
|
"loss": -0.0,
|
|
"num_tokens": 7952236.0,
|
|
"reward": 0.7406250238418579,
|
|
"reward_std": 0.04419417306780815,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.640625,
|
|
"rewards/mcq_exact_match_reward/std": 0.4836103618144989,
|
|
"step": 74
|
|
},
|
|
{
|
|
"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": 12.0,
|
|
"completions/max_terminated_length": 12.0,
|
|
"completions/mean_length": 12.0,
|
|
"completions/mean_terminated_length": 12.0,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1875,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 14.528362274169922,
|
|
"learning_rate": 1.7370711923791564e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 8041828.0,
|
|
"reward": 0.8656250238418579,
|
|
"reward_std": 0.10205793380737305,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.765625,
|
|
"rewards/mcq_exact_match_reward/std": 0.42695629596710205,
|
|
"step": 75
|
|
},
|
|
{
|
|
"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": 159.0,
|
|
"completions/max_terminated_length": 159.0,
|
|
"completions/mean_length": 14.375,
|
|
"completions/mean_terminated_length": 14.375,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.19,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 15.905014991760254,
|
|
"learning_rate": 1.6135921418712955e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 8134972.0,
|
|
"reward": 0.7953125238418579,
|
|
"reward_std": 0.1753426045179367,
|
|
"rewards/format_reward/mean": 0.921875,
|
|
"rewards/format_reward/std": 0.27048972249031067,
|
|
"rewards/mcq_exact_match_reward/mean": 0.703125,
|
|
"rewards/mcq_exact_match_reward/std": 0.4604927599430084,
|
|
"step": 76
|
|
},
|
|
{
|
|
"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": 317.0,
|
|
"completions/max_terminated_length": 317.0,
|
|
"completions/mean_length": 16.78125,
|
|
"completions/mean_terminated_length": 16.78125,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1925,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 11.540793418884277,
|
|
"learning_rate": 1.493816078637557e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 8244030.0,
|
|
"reward": 0.7406250238418579,
|
|
"reward_std": 0.2109457552433014,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.640625,
|
|
"rewards/mcq_exact_match_reward/std": 0.4836103618144989,
|
|
"step": 77
|
|
},
|
|
{
|
|
"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/max_terminated_length": 498.0,
|
|
"completions/mean_length": 22.5,
|
|
"completions/mean_terminated_length": 22.5,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.195,
|
|
"frac_reward_zero_std": 0.875,
|
|
"grad_norm": 5.036442756652832,
|
|
"learning_rate": 1.3778739760445552e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 8332454.0,
|
|
"reward": 0.9437500238418579,
|
|
"reward_std": 0.0578637570142746,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.84375,
|
|
"rewards/mcq_exact_match_reward/std": 0.36596253514289856,
|
|
"step": 78
|
|
},
|
|
{
|
|
"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": 135.0,
|
|
"completions/max_terminated_length": 135.0,
|
|
"completions/mean_length": 14.5,
|
|
"completions/mean_terminated_length": 14.5,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.1975,
|
|
"frac_reward_zero_std": 0.375,
|
|
"grad_norm": 24.74799919128418,
|
|
"learning_rate": 1.2658926150792322e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 8426206.0,
|
|
"reward": 0.6031249761581421,
|
|
"reward_std": 0.2323707640171051,
|
|
"rewards/format_reward/mean": 0.875,
|
|
"rewards/format_reward/std": 0.3333333432674408,
|
|
"rewards/mcq_exact_match_reward/mean": 0.515625,
|
|
"rewards/mcq_exact_match_reward/std": 0.5037065148353577,
|
|
"step": 79
|
|
},
|
|
{
|
|
"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": 1192.0,
|
|
"completions/max_terminated_length": 1192.0,
|
|
"completions/mean_length": 54.09375,
|
|
"completions/mean_terminated_length": 54.09375,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.2,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 9.330410957336426,
|
|
"learning_rate": 1.1579944457157059e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 8530628.0,
|
|
"reward": 0.5687500238418579,
|
|
"reward_std": 0.10888782143592834,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.46875,
|
|
"rewards/mcq_exact_match_reward/std": 0.5029674172401428,
|
|
"step": 80
|
|
},
|
|
{
|
|
"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": 284.0,
|
|
"completions/max_terminated_length": 284.0,
|
|
"completions/mean_length": 22.078125,
|
|
"completions/mean_terminated_length": 22.078125,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.2025,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 4.5020928382873535,
|
|
"learning_rate": 1.0542974530180327e-07,
|
|
"loss": 0.0,
|
|
"num_tokens": 8629129.0,
|
|
"reward": 0.7796875238418579,
|
|
"reward_std": 0.11535719037055969,
|
|
"rewards/format_reward/mean": 0.921875,
|
|
"rewards/format_reward/std": 0.27048972249031067,
|
|
"rewards/mcq_exact_match_reward/mean": 0.6875,
|
|
"rewards/mcq_exact_match_reward/std": 0.467176616191864,
|
|
"step": 81
|
|
},
|
|
{
|
|
"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": 106.0,
|
|
"completions/max_terminated_length": 106.0,
|
|
"completions/mean_length": 13.46875,
|
|
"completions/mean_terminated_length": 13.46875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.205,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 8.712493896484375,
|
|
"learning_rate": 9.549150281252632e-08,
|
|
"loss": 0.0,
|
|
"num_tokens": 8714991.0,
|
|
"reward": 0.7718750238418579,
|
|
"reward_std": 0.15992169082164764,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.671875,
|
|
"rewards/mcq_exact_match_reward/std": 0.4732423722743988,
|
|
"step": 82
|
|
},
|
|
{
|
|
"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": 201.0,
|
|
"completions/max_terminated_length": 201.0,
|
|
"completions/mean_length": 14.96875,
|
|
"completions/mean_terminated_length": 14.96875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.2075,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 11.34872817993164,
|
|
"learning_rate": 8.599558442598998e-08,
|
|
"loss": 0.0,
|
|
"num_tokens": 8816477.0,
|
|
"reward": 0.6625000238418579,
|
|
"reward_std": 0.24511480331420898,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.5625,
|
|
"rewards/mcq_exact_match_reward/std": 0.5,
|
|
"step": 83
|
|
},
|
|
{
|
|
"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": 13.0,
|
|
"completions/max_terminated_length": 13.0,
|
|
"completions/mean_length": 12.015625,
|
|
"completions/mean_terminated_length": 12.015625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.21,
|
|
"frac_reward_zero_std": 0.875,
|
|
"grad_norm": 9.087535858154297,
|
|
"learning_rate": 7.695237378953224e-08,
|
|
"loss": 0.0,
|
|
"num_tokens": 8898886.0,
|
|
"reward": 1.021875023841858,
|
|
"reward_std": 0.0646936446428299,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.921875,
|
|
"rewards/mcq_exact_match_reward/std": 0.27048972249031067,
|
|
"step": 84
|
|
},
|
|
{
|
|
"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": 146.0,
|
|
"completions/max_terminated_length": 146.0,
|
|
"completions/mean_length": 14.09375,
|
|
"completions/mean_terminated_length": 14.09375,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.2125,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 7.845055103302002,
|
|
"learning_rate": 6.837175952121304e-08,
|
|
"loss": -0.0,
|
|
"num_tokens": 8994764.0,
|
|
"reward": 0.9125000238418579,
|
|
"reward_std": 0.10888782143592834,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.8125,
|
|
"rewards/mcq_exact_match_reward/std": 0.39339789748191833,
|
|
"step": 85
|
|
},
|
|
{
|
|
"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": 13.0,
|
|
"completions/max_terminated_length": 13.0,
|
|
"completions/mean_length": 12.015625,
|
|
"completions/mean_terminated_length": 12.015625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.215,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 7.6561279296875,
|
|
"learning_rate": 6.026312439675551e-08,
|
|
"loss": -0.0,
|
|
"num_tokens": 9079917.0,
|
|
"reward": 0.2875000238418579,
|
|
"reward_std": 0.10888782143592834,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.1875,
|
|
"rewards/mcq_exact_match_reward/std": 0.39339789748191833,
|
|
"step": 86
|
|
},
|
|
{
|
|
"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": 0.015625,
|
|
"completions/max_length": 2048.0,
|
|
"completions/max_terminated_length": 539.0,
|
|
"completions/mean_length": 54.09375,
|
|
"completions/mean_terminated_length": 22.444446563720703,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.2175,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 9.46330451965332,
|
|
"learning_rate": 5.263533508961826e-08,
|
|
"loss": 0.0,
|
|
"num_tokens": 9162947.0,
|
|
"reward": 0.989062488079071,
|
|
"reward_std": 0.13377662003040314,
|
|
"rewards/format_reward/mean": 0.984375,
|
|
"rewards/format_reward/std": 0.125,
|
|
"rewards/mcq_exact_match_reward/mean": 0.890625,
|
|
"rewards/mcq_exact_match_reward/std": 0.3145764470100403,
|
|
"step": 87
|
|
},
|
|
{
|
|
"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": 480.0,
|
|
"completions/max_terminated_length": 480.0,
|
|
"completions/mean_length": 37.21875,
|
|
"completions/mean_terminated_length": 37.21875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.22,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 14.467491149902344,
|
|
"learning_rate": 4.549673247541874e-08,
|
|
"loss": 0.0,
|
|
"num_tokens": 9246449.0,
|
|
"reward": 0.9437500238418579,
|
|
"reward_std": 0.1767766922712326,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.84375,
|
|
"rewards/mcq_exact_match_reward/std": 0.36596253514289856,
|
|
"step": 88
|
|
},
|
|
{
|
|
"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": 197.0,
|
|
"completions/max_terminated_length": 197.0,
|
|
"completions/mean_length": 21.90625,
|
|
"completions/mean_terminated_length": 21.90625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.2225,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 13.93521785736084,
|
|
"learning_rate": 3.8855122511307626e-08,
|
|
"loss": -0.0,
|
|
"num_tokens": 9332955.0,
|
|
"reward": 0.5218750238418579,
|
|
"reward_std": 0.13258251547813416,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.421875,
|
|
"rewards/mcq_exact_match_reward/std": 0.49776285886764526,
|
|
"step": 89
|
|
},
|
|
{
|
|
"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": 336.0,
|
|
"completions/max_terminated_length": 336.0,
|
|
"completions/mean_length": 30.0625,
|
|
"completions/mean_terminated_length": 30.0625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.225,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 13.635700225830078,
|
|
"learning_rate": 3.271776770026963e-08,
|
|
"loss": 0.0,
|
|
"num_tokens": 9428823.0,
|
|
"reward": 0.67578125,
|
|
"reward_std": 0.20194020867347717,
|
|
"rewards/format_reward/mean": 0.9765625,
|
|
"rewards/format_reward/std": 0.13886408507823944,
|
|
"rewards/mcq_exact_match_reward/mean": 0.578125,
|
|
"rewards/mcq_exact_match_reward/std": 0.49776285886764526,
|
|
"step": 90
|
|
},
|
|
{
|
|
"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": 456.0,
|
|
"completions/max_terminated_length": 456.0,
|
|
"completions/mean_length": 38.0,
|
|
"completions/mean_terminated_length": 38.0,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.2275,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 7.340695858001709,
|
|
"learning_rate": 2.7091379149682682e-08,
|
|
"loss": -0.0,
|
|
"num_tokens": 9520103.0,
|
|
"reward": 0.667187511920929,
|
|
"reward_std": 0.1133689433336258,
|
|
"rewards/format_reward/mean": 0.890625,
|
|
"rewards/format_reward/std": 0.301698237657547,
|
|
"rewards/mcq_exact_match_reward/mean": 0.578125,
|
|
"rewards/mcq_exact_match_reward/std": 0.49776285886764526,
|
|
"step": 91
|
|
},
|
|
{
|
|
"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": 135.0,
|
|
"completions/max_terminated_length": 135.0,
|
|
"completions/mean_length": 15.65625,
|
|
"completions/mean_terminated_length": 15.65625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.23,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 13.138660430908203,
|
|
"learning_rate": 2.1982109232821176e-08,
|
|
"loss": 0.0,
|
|
"num_tokens": 9616769.0,
|
|
"reward": 0.8968750238418579,
|
|
"reward_std": 0.2109457552433014,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.796875,
|
|
"rewards/mcq_exact_match_reward/std": 0.40550529956817627,
|
|
"step": 92
|
|
},
|
|
{
|
|
"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": 12.0,
|
|
"completions/max_terminated_length": 12.0,
|
|
"completions/mean_length": 12.0,
|
|
"completions/mean_terminated_length": 12.0,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.2325,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 9.40440845489502,
|
|
"learning_rate": 1.7395544861325718e-08,
|
|
"loss": 0.0,
|
|
"num_tokens": 9715273.0,
|
|
"reward": 0.7093750238418579,
|
|
"reward_std": 0.1315089464187622,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.609375,
|
|
"rewards/mcq_exact_match_reward/std": 0.4917473793029785,
|
|
"step": 93
|
|
},
|
|
{
|
|
"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": 319.0,
|
|
"completions/max_terminated_length": 319.0,
|
|
"completions/mean_length": 16.875,
|
|
"completions/mean_terminated_length": 16.875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.235,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 11.90187931060791,
|
|
"learning_rate": 1.3336701375997127e-08,
|
|
"loss": 0.0,
|
|
"num_tokens": 9823953.0,
|
|
"reward": 0.7093750238418579,
|
|
"reward_std": 0.23144522309303284,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.609375,
|
|
"rewards/mcq_exact_match_reward/std": 0.4917473793029785,
|
|
"step": 94
|
|
},
|
|
{
|
|
"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": 317.0,
|
|
"completions/max_terminated_length": 317.0,
|
|
"completions/mean_length": 20.828125,
|
|
"completions/mean_terminated_length": 20.828125,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.2375,
|
|
"frac_reward_zero_std": 0.5,
|
|
"grad_norm": 10.480510711669922,
|
|
"learning_rate": 9.810017062595321e-09,
|
|
"loss": 0.0,
|
|
"num_tokens": 9926510.0,
|
|
"reward": 0.6312500238418579,
|
|
"reward_std": 0.2130674123764038,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.53125,
|
|
"rewards/mcq_exact_match_reward/std": 0.5029674172401428,
|
|
"step": 95
|
|
},
|
|
{
|
|
"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": 935.0,
|
|
"completions/max_terminated_length": 935.0,
|
|
"completions/mean_length": 46.328125,
|
|
"completions/mean_terminated_length": 46.328125,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.24,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 9.879074096679688,
|
|
"learning_rate": 6.819348298638839e-09,
|
|
"loss": -0.0,
|
|
"num_tokens": 10022339.0,
|
|
"reward": 0.6468750238418579,
|
|
"reward_std": 0.1530819833278656,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.546875,
|
|
"rewards/mcq_exact_match_reward/std": 0.501733124256134,
|
|
"step": 96
|
|
},
|
|
{
|
|
"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": 120.0,
|
|
"completions/max_terminated_length": 120.0,
|
|
"completions/mean_length": 15.359375,
|
|
"completions/mean_terminated_length": 15.359375,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.2425,
|
|
"frac_reward_zero_std": 0.75,
|
|
"grad_norm": 12.736063957214355,
|
|
"learning_rate": 4.367965336512403e-09,
|
|
"loss": 0.0,
|
|
"num_tokens": 10134146.0,
|
|
"reward": 0.754687488079071,
|
|
"reward_std": 0.13165494799613953,
|
|
"rewards/format_reward/mean": 0.984375,
|
|
"rewards/format_reward/std": 0.125,
|
|
"rewards/mcq_exact_match_reward/mean": 0.65625,
|
|
"rewards/mcq_exact_match_reward/std": 0.4787135720252991,
|
|
"step": 97
|
|
},
|
|
{
|
|
"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": 630.0,
|
|
"completions/max_terminated_length": 630.0,
|
|
"completions/mean_length": 34.84375,
|
|
"completions/mean_terminated_length": 34.84375,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.245,
|
|
"frac_reward_zero_std": 1.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 2.458548727494292e-09,
|
|
"loss": 0.0,
|
|
"num_tokens": 10217352.0,
|
|
"reward": 0.4750000238418579,
|
|
"reward_std": 0.0,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.375,
|
|
"rewards/mcq_exact_match_reward/std": 0.48795005679130554,
|
|
"step": 98
|
|
},
|
|
{
|
|
"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": 13.0,
|
|
"completions/max_terminated_length": 13.0,
|
|
"completions/mean_length": 12.046875,
|
|
"completions/mean_terminated_length": 12.046875,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.2475,
|
|
"frac_reward_zero_std": 1.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 1.0931863906127325e-09,
|
|
"loss": 0.0,
|
|
"num_tokens": 10289723.0,
|
|
"reward": 0.9750000238418579,
|
|
"reward_std": 0.0,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.875,
|
|
"rewards/mcq_exact_match_reward/std": 0.3333333432674408,
|
|
"step": 99
|
|
},
|
|
{
|
|
"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": 13.0,
|
|
"completions/max_terminated_length": 13.0,
|
|
"completions/mean_length": 12.015625,
|
|
"completions/mean_terminated_length": 12.015625,
|
|
"completions/min_length": 12.0,
|
|
"completions/min_terminated_length": 12.0,
|
|
"epoch": 0.25,
|
|
"frac_reward_zero_std": 0.625,
|
|
"grad_norm": 11.199357986450195,
|
|
"learning_rate": 2.733713295369755e-10,
|
|
"loss": 0.0,
|
|
"num_tokens": 10406076.0,
|
|
"reward": 0.7875000238418579,
|
|
"reward_std": 0.16675157845020294,
|
|
"rewards/format_reward/mean": 1.0,
|
|
"rewards/format_reward/std": 0.0,
|
|
"rewards/mcq_exact_match_reward/mean": 0.6875,
|
|
"rewards/mcq_exact_match_reward/std": 0.467176616191864,
|
|
"step": 100
|
|
}
|
|
],
|
|
"logging_steps": 1,
|
|
"max_steps": 100,
|
|
"num_input_tokens_seen": 10406076,
|
|
"num_train_epochs": 1,
|
|
"save_steps": 50,
|
|
"stateful_callbacks": {
|
|
"TrainerControl": {
|
|
"args": {
|
|
"should_epoch_stop": false,
|
|
"should_evaluate": false,
|
|
"should_log": false,
|
|
"should_save": true,
|
|
"should_training_stop": true
|
|
},
|
|
"attributes": {}
|
|
}
|
|
},
|
|
"total_flos": 0.0,
|
|
"train_batch_size": 1,
|
|
"trial_name": null,
|
|
"trial_params": null
|
|
}
|