10511 lines
402 KiB
JSON
10511 lines
402 KiB
JSON
[
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 0.0,
|
|
"num_tokens": 8999.0,
|
|
"completions/mean_length": 438.625,
|
|
"completions/min_length": 221.0,
|
|
"completions/max_length": 912.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 373.5714416503906,
|
|
"completions/min_terminated_length": 221.0,
|
|
"completions/max_terminated_length": 912.0,
|
|
"tools/call_frequency": 11.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005120682530105114,
|
|
"sampling/sampling_logp_difference/max": 0.744844913482666,
|
|
"sampling/importance_sampling_ratio/min": 0.5189680457115173,
|
|
"sampling/importance_sampling_ratio/mean": 0.8280686140060425,
|
|
"sampling/importance_sampling_ratio/max": 1.5097784996032715,
|
|
"kl": 0.0,
|
|
"entropy": 0.06919643934816122,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.612434606999159,
|
|
"epoch": 2.604166666666667e-05,
|
|
"step": 1
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 5e-08,
|
|
"num_tokens": 18088.0,
|
|
"completions/mean_length": 449.875,
|
|
"completions/min_length": 255.0,
|
|
"completions/max_length": 898.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 304.0,
|
|
"completions/min_terminated_length": 255.0,
|
|
"completions/max_terminated_length": 339.0,
|
|
"tools/call_frequency": 11.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004156216513365507,
|
|
"sampling/sampling_logp_difference/max": 0.5351243019104004,
|
|
"sampling/importance_sampling_ratio/min": 0.5063021183013916,
|
|
"sampling/importance_sampling_ratio/mean": 1.0779616832733154,
|
|
"sampling/importance_sampling_ratio/max": 1.8130271434783936,
|
|
"kl": 0.0,
|
|
"entropy": 0.05932584893889725,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.131927305832505,
|
|
"epoch": 5.208333333333334e-05,
|
|
"step": 2
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 1e-07,
|
|
"num_tokens": 28597.0,
|
|
"completions/mean_length": 627.875,
|
|
"completions/min_length": 282.0,
|
|
"completions/max_length": 1015.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 519.1666870117188,
|
|
"completions/min_terminated_length": 282.0,
|
|
"completions/max_terminated_length": 959.0,
|
|
"tools/call_frequency": 15.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003144721034914255,
|
|
"sampling/sampling_logp_difference/max": 0.5001346468925476,
|
|
"sampling/importance_sampling_ratio/min": 0.6824742555618286,
|
|
"sampling/importance_sampling_ratio/mean": 1.1974340677261353,
|
|
"sampling/importance_sampling_ratio/max": 1.4131324291229248,
|
|
"kl": 0.0,
|
|
"entropy": 0.0534731907537207,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.68013490922749,
|
|
"epoch": 7.8125e-05,
|
|
"step": 3
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 1.5e-07,
|
|
"num_tokens": 38545.0,
|
|
"completions/mean_length": 557.0,
|
|
"completions/min_length": 268.0,
|
|
"completions/max_length": 1010.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 433.5,
|
|
"completions/min_terminated_length": 268.0,
|
|
"completions/max_terminated_length": 1010.0,
|
|
"tools/call_frequency": 13.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003950539976358414,
|
|
"sampling/sampling_logp_difference/max": 0.49446868896484375,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.8259817361831665,
|
|
"sampling/importance_sampling_ratio/max": 1.6527460813522339,
|
|
"kl": 0.0,
|
|
"entropy": 0.0693533371668309,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.165325535461307,
|
|
"epoch": 0.00010416666666666667,
|
|
"step": 4
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 2e-07,
|
|
"num_tokens": 47761.0,
|
|
"completions/mean_length": 466.5,
|
|
"completions/min_length": 267.0,
|
|
"completions/max_length": 993.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 305.16668701171875,
|
|
"completions/min_terminated_length": 267.0,
|
|
"completions/max_terminated_length": 370.0,
|
|
"tools/call_frequency": 11.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004810507874935865,
|
|
"sampling/sampling_logp_difference/max": 0.9897897243499756,
|
|
"sampling/importance_sampling_ratio/min": 0.2608538568019867,
|
|
"sampling/importance_sampling_ratio/mean": 1.051840901374817,
|
|
"sampling/importance_sampling_ratio/max": 2.2131173610687256,
|
|
"kl": 0.0,
|
|
"entropy": 0.061791570857167244,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.275234065949917,
|
|
"epoch": 0.00013020833333333333,
|
|
"step": 5
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 2.5e-07,
|
|
"num_tokens": 56411.0,
|
|
"completions/mean_length": 395.0,
|
|
"completions/min_length": 285.0,
|
|
"completions/max_length": 913.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 395.0,
|
|
"completions/min_terminated_length": 285.0,
|
|
"completions/max_terminated_length": 913.0,
|
|
"tools/call_frequency": 9.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006951844785362482,
|
|
"sampling/sampling_logp_difference/max": 0.6398494243621826,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.49105578660964966,
|
|
"sampling/importance_sampling_ratio/max": 1.095594048500061,
|
|
"kl": 0.0,
|
|
"entropy": 0.08770200004801154,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.828022077679634,
|
|
"epoch": 0.00015625,
|
|
"step": 6
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 3e-07,
|
|
"num_tokens": 65712.0,
|
|
"completions/mean_length": 476.5,
|
|
"completions/min_length": 262.0,
|
|
"completions/max_length": 940.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 415.5714416503906,
|
|
"completions/min_terminated_length": 262.0,
|
|
"completions/max_terminated_length": 940.0,
|
|
"tools/call_frequency": 12.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004118240438401699,
|
|
"sampling/sampling_logp_difference/max": 0.7463555335998535,
|
|
"sampling/importance_sampling_ratio/min": 0.34403476119041443,
|
|
"sampling/importance_sampling_ratio/mean": 0.7522589564323425,
|
|
"sampling/importance_sampling_ratio/max": 2.080158233642578,
|
|
"kl": 0.0,
|
|
"entropy": 0.0710453613428399,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.89737706631422,
|
|
"epoch": 0.00018229166666666667,
|
|
"step": 7
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 3.5e-07,
|
|
"num_tokens": 74145.0,
|
|
"completions/mean_length": 369.0,
|
|
"completions/min_length": 260.0,
|
|
"completions/max_length": 745.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 315.2857360839844,
|
|
"completions/min_terminated_length": 260.0,
|
|
"completions/max_terminated_length": 353.0,
|
|
"tools/call_frequency": 10.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005332476459443569,
|
|
"sampling/sampling_logp_difference/max": 1.5807342529296875,
|
|
"sampling/importance_sampling_ratio/min": 0.35255640745162964,
|
|
"sampling/importance_sampling_ratio/mean": 0.9108841419219971,
|
|
"sampling/importance_sampling_ratio/max": 2.1552317142486572,
|
|
"kl": 0.0,
|
|
"entropy": 0.06978825887199491,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.949710752815008,
|
|
"epoch": 0.00020833333333333335,
|
|
"step": 8
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 4e-07,
|
|
"num_tokens": 82834.0,
|
|
"completions/mean_length": 400.625,
|
|
"completions/min_length": 288.0,
|
|
"completions/max_length": 931.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 400.625,
|
|
"completions/min_terminated_length": 288.0,
|
|
"completions/max_terminated_length": 931.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00760924955829978,
|
|
"sampling/sampling_logp_difference/max": 0.4636850357055664,
|
|
"sampling/importance_sampling_ratio/min": 0.4945988953113556,
|
|
"sampling/importance_sampling_ratio/mean": 0.9668939113616943,
|
|
"sampling/importance_sampling_ratio/max": 1.6802269220352173,
|
|
"kl": 0.0,
|
|
"entropy": 0.09935027034953237,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.877145735546947,
|
|
"epoch": 0.000234375,
|
|
"step": 9
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 4.5e-07,
|
|
"num_tokens": 91968.0,
|
|
"completions/mean_length": 455.25,
|
|
"completions/min_length": 247.0,
|
|
"completions/max_length": 912.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 303.8333435058594,
|
|
"completions/min_terminated_length": 247.0,
|
|
"completions/max_terminated_length": 331.0,
|
|
"tools/call_frequency": 12.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004774360917508602,
|
|
"sampling/sampling_logp_difference/max": 0.98126220703125,
|
|
"sampling/importance_sampling_ratio/min": 0.24541591107845306,
|
|
"sampling/importance_sampling_ratio/mean": 0.7404389381408691,
|
|
"sampling/importance_sampling_ratio/max": 1.0929973125457764,
|
|
"kl": 0.0,
|
|
"entropy": 0.07833283627405763,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.142524428665638,
|
|
"epoch": 0.00026041666666666666,
|
|
"step": 10
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 5e-07,
|
|
"num_tokens": 101783.0,
|
|
"completions/mean_length": 540.875,
|
|
"completions/min_length": 293.0,
|
|
"completions/max_length": 908.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 419.3333435058594,
|
|
"completions/min_terminated_length": 293.0,
|
|
"completions/max_terminated_length": 885.0,
|
|
"tools/call_frequency": 13.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00364117044955492,
|
|
"sampling/sampling_logp_difference/max": 0.4771270751953125,
|
|
"sampling/importance_sampling_ratio/min": 0.353768914937973,
|
|
"sampling/importance_sampling_ratio/mean": 0.8930596709251404,
|
|
"sampling/importance_sampling_ratio/max": 2.3585257530212402,
|
|
"kl": 0.0,
|
|
"entropy": 0.06662136141676456,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.38863966241479,
|
|
"epoch": 0.00028645833333333333,
|
|
"step": 11
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 4.982758620689655e-07,
|
|
"num_tokens": 112206.0,
|
|
"completions/mean_length": 616.625,
|
|
"completions/min_length": 265.0,
|
|
"completions/max_length": 993.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 562.857177734375,
|
|
"completions/min_terminated_length": 265.0,
|
|
"completions/max_terminated_length": 957.0,
|
|
"tools/call_frequency": 14.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0033395111095160246,
|
|
"sampling/sampling_logp_difference/max": 0.4717535972595215,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.9615724086761475,
|
|
"sampling/importance_sampling_ratio/max": 2.478001832962036,
|
|
"kl": 0.0,
|
|
"entropy": 0.06196625670418143,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.46426684781909,
|
|
"epoch": 0.0003125,
|
|
"step": 12
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 4.96551724137931e-07,
|
|
"num_tokens": 121414.0,
|
|
"completions/mean_length": 464.625,
|
|
"completions/min_length": 249.0,
|
|
"completions/max_length": 910.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 320.0,
|
|
"completions/min_terminated_length": 249.0,
|
|
"completions/max_terminated_length": 377.0,
|
|
"tools/call_frequency": 11.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004406385589390993,
|
|
"sampling/sampling_logp_difference/max": 0.46643543243408203,
|
|
"sampling/importance_sampling_ratio/min": 0.31518030166625977,
|
|
"sampling/importance_sampling_ratio/mean": 0.9380097389221191,
|
|
"sampling/importance_sampling_ratio/max": 1.669421672821045,
|
|
"kl": 0.0,
|
|
"entropy": 0.07358739827759564,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.102533174678683,
|
|
"epoch": 0.0003385416666666667,
|
|
"step": 13
|
|
},
|
|
{
|
|
"loss": 0.0,
|
|
"grad_norm": 0.0,
|
|
"learning_rate": 4.948275862068965e-07,
|
|
"num_tokens": 130065.0,
|
|
"completions/mean_length": 394.75,
|
|
"completions/min_length": 260.0,
|
|
"completions/max_length": 947.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 315.8571472167969,
|
|
"completions/min_terminated_length": 260.0,
|
|
"completions/max_terminated_length": 337.0,
|
|
"tools/call_frequency": 10.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006064712535589933,
|
|
"sampling/sampling_logp_difference/max": 0.6249533891677856,
|
|
"sampling/importance_sampling_ratio/min": 0.2008495181798935,
|
|
"sampling/importance_sampling_ratio/mean": 0.9442447423934937,
|
|
"sampling/importance_sampling_ratio/max": 1.682063102722168,
|
|
"kl": 0.0,
|
|
"entropy": 0.0793033498339355,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.007314339280128,
|
|
"epoch": 0.00036458333333333335,
|
|
"step": 14
|
|
},
|
|
{
|
|
"loss": 0.1824590265750885,
|
|
"grad_norm": 2.445009231567383,
|
|
"learning_rate": 4.93103448275862e-07,
|
|
"num_tokens": 139588.0,
|
|
"completions/mean_length": 504.25,
|
|
"completions/min_length": 227.0,
|
|
"completions/max_length": 966.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 396.5,
|
|
"completions/min_terminated_length": 227.0,
|
|
"completions/max_terminated_length": 966.0,
|
|
"tools/call_frequency": 14.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003295793430879712,
|
|
"sampling/sampling_logp_difference/max": 0.6283330917358398,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.9160506725311279,
|
|
"sampling/importance_sampling_ratio/max": 1.474215030670166,
|
|
"kl": 0.0,
|
|
"entropy": 0.05436322238529101,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.251120718196034,
|
|
"epoch": 0.000390625,
|
|
"step": 15
|
|
},
|
|
{
|
|
"loss": 3.104244024143554e-05,
|
|
"grad_norm": 0.08518494665622711,
|
|
"learning_rate": 4.913793103448275e-07,
|
|
"num_tokens": 149532.0,
|
|
"completions/mean_length": 557.125,
|
|
"completions/min_length": 286.0,
|
|
"completions/max_length": 1000.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 327.8000183105469,
|
|
"completions/min_terminated_length": 286.0,
|
|
"completions/max_terminated_length": 406.0,
|
|
"tools/call_frequency": 13.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0031099196057766676,
|
|
"sampling/sampling_logp_difference/max": 0.5189504623413086,
|
|
"sampling/importance_sampling_ratio/min": 0.4643911123275757,
|
|
"sampling/importance_sampling_ratio/mean": 0.9884326457977295,
|
|
"sampling/importance_sampling_ratio/max": 1.4223459959030151,
|
|
"kl": 0.00019533169177066156,
|
|
"entropy": 0.05777230981038883,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.190284309908748,
|
|
"epoch": 0.0004166666666666667,
|
|
"step": 16
|
|
},
|
|
{
|
|
"loss": 4.9894926632987335e-05,
|
|
"grad_norm": 0.13512587547302246,
|
|
"learning_rate": 4.89655172413793e-07,
|
|
"num_tokens": 158535.0,
|
|
"completions/mean_length": 439.375,
|
|
"completions/min_length": 256.0,
|
|
"completions/max_length": 902.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 290.8333435058594,
|
|
"completions/min_terminated_length": 256.0,
|
|
"completions/max_terminated_length": 342.0,
|
|
"tools/call_frequency": 11.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004964124411344528,
|
|
"sampling/sampling_logp_difference/max": 0.9403212070465088,
|
|
"sampling/importance_sampling_ratio/min": 0.25480666756629944,
|
|
"sampling/importance_sampling_ratio/mean": 0.745871365070343,
|
|
"sampling/importance_sampling_ratio/max": 1.3032686710357666,
|
|
"kl": 0.0003626156139091563,
|
|
"entropy": 0.06438471336150542,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.618618844076991,
|
|
"epoch": 0.0004427083333333333,
|
|
"step": 17
|
|
},
|
|
{
|
|
"loss": 4.5905577280791476e-05,
|
|
"grad_norm": 0.2815433442592621,
|
|
"learning_rate": 4.879310344827585e-07,
|
|
"num_tokens": 168364.0,
|
|
"completions/mean_length": 543.125,
|
|
"completions/min_length": 290.0,
|
|
"completions/max_length": 973.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 497.14288330078125,
|
|
"completions/min_terminated_length": 290.0,
|
|
"completions/max_terminated_length": 973.0,
|
|
"tools/call_frequency": 13.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0031487673986703157,
|
|
"sampling/sampling_logp_difference/max": 0.6201504468917847,
|
|
"sampling/importance_sampling_ratio/min": 0.22788047790527344,
|
|
"sampling/importance_sampling_ratio/mean": 0.7212487459182739,
|
|
"sampling/importance_sampling_ratio/max": 1.1352124214172363,
|
|
"kl": 0.0003103034136984206,
|
|
"entropy": 0.04502043721731752,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.573951479047537,
|
|
"epoch": 0.00046875,
|
|
"step": 18
|
|
},
|
|
{
|
|
"loss": 3.7292425986379385e-05,
|
|
"grad_norm": 0.08619094640016556,
|
|
"learning_rate": 4.86206896551724e-07,
|
|
"num_tokens": 177508.0,
|
|
"completions/mean_length": 457.375,
|
|
"completions/min_length": 257.0,
|
|
"completions/max_length": 893.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 317.3333435058594,
|
|
"completions/min_terminated_length": 257.0,
|
|
"completions/max_terminated_length": 358.0,
|
|
"tools/call_frequency": 11.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004081712570041418,
|
|
"sampling/sampling_logp_difference/max": 0.4247640371322632,
|
|
"sampling/importance_sampling_ratio/min": 0.5013089179992676,
|
|
"sampling/importance_sampling_ratio/mean": 0.9189894199371338,
|
|
"sampling/importance_sampling_ratio/max": 1.2129801511764526,
|
|
"kl": 0.0002287696368057368,
|
|
"entropy": 0.07021064823493361,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.501019734889269,
|
|
"epoch": 0.0004947916666666667,
|
|
"step": 19
|
|
},
|
|
{
|
|
"loss": 5.1855990022886544e-05,
|
|
"grad_norm": 0.12558245658874512,
|
|
"learning_rate": 4.844827586206897e-07,
|
|
"num_tokens": 186615.0,
|
|
"completions/mean_length": 452.5,
|
|
"completions/min_length": 237.0,
|
|
"completions/max_length": 899.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 389.5714416503906,
|
|
"completions/min_terminated_length": 237.0,
|
|
"completions/max_terminated_length": 899.0,
|
|
"tools/call_frequency": 11.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004194268956780434,
|
|
"sampling/sampling_logp_difference/max": 0.40123844146728516,
|
|
"sampling/importance_sampling_ratio/min": 0.37160754203796387,
|
|
"sampling/importance_sampling_ratio/mean": 0.9739571213722229,
|
|
"sampling/importance_sampling_ratio/max": 1.4154884815216064,
|
|
"kl": 0.0003337603102409048,
|
|
"entropy": 0.06498852674849331,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.881056571379304,
|
|
"epoch": 0.0005208333333333333,
|
|
"step": 20
|
|
},
|
|
{
|
|
"loss": 6.783201388316229e-05,
|
|
"grad_norm": 0.20701435208320618,
|
|
"learning_rate": 4.827586206896552e-07,
|
|
"num_tokens": 195820.0,
|
|
"completions/mean_length": 464.625,
|
|
"completions/min_length": 255.0,
|
|
"completions/max_length": 988.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 310.66668701171875,
|
|
"completions/min_terminated_length": 255.0,
|
|
"completions/max_terminated_length": 394.0,
|
|
"tools/call_frequency": 11.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004447505343705416,
|
|
"sampling/sampling_logp_difference/max": 0.6968374252319336,
|
|
"sampling/importance_sampling_ratio/min": 0.3282564580440521,
|
|
"sampling/importance_sampling_ratio/mean": 0.7744771242141724,
|
|
"sampling/importance_sampling_ratio/max": 1.401808738708496,
|
|
"kl": 0.0004668139523005266,
|
|
"entropy": 0.0781644499511458,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.718991735950112,
|
|
"epoch": 0.000546875,
|
|
"step": 21
|
|
},
|
|
{
|
|
"loss": 8.095511293504387e-05,
|
|
"grad_norm": 0.302961528301239,
|
|
"learning_rate": 4.810344827586207e-07,
|
|
"num_tokens": 204335.0,
|
|
"completions/mean_length": 378.25,
|
|
"completions/min_length": 269.0,
|
|
"completions/max_length": 889.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 305.2857360839844,
|
|
"completions/min_terminated_length": 269.0,
|
|
"completions/max_terminated_length": 326.0,
|
|
"tools/call_frequency": 10.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005118285771459341,
|
|
"sampling/sampling_logp_difference/max": 0.46361303329467773,
|
|
"sampling/importance_sampling_ratio/min": 0.33733904361724854,
|
|
"sampling/importance_sampling_ratio/mean": 0.7433972358703613,
|
|
"sampling/importance_sampling_ratio/max": 1.6031376123428345,
|
|
"kl": 0.0004206575677017099,
|
|
"entropy": 0.07559330854564905,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.768400074914098,
|
|
"epoch": 0.0005729166666666667,
|
|
"step": 22
|
|
},
|
|
{
|
|
"loss": 9.874640090856701e-05,
|
|
"grad_norm": 0.24912399053573608,
|
|
"learning_rate": 4.793103448275862e-07,
|
|
"num_tokens": 212844.0,
|
|
"completions/mean_length": 377.875,
|
|
"completions/min_length": 263.0,
|
|
"completions/max_length": 934.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 377.875,
|
|
"completions/min_terminated_length": 263.0,
|
|
"completions/max_terminated_length": 934.0,
|
|
"tools/call_frequency": 9.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006939517334103584,
|
|
"sampling/sampling_logp_difference/max": 0.7168383598327637,
|
|
"sampling/importance_sampling_ratio/min": 0.506744384765625,
|
|
"sampling/importance_sampling_ratio/mean": 1.0606098175048828,
|
|
"sampling/importance_sampling_ratio/max": 2.2798423767089844,
|
|
"kl": 0.0005632244392472785,
|
|
"entropy": 0.08980596379842609,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.027928613126278,
|
|
"epoch": 0.0005989583333333333,
|
|
"step": 23
|
|
},
|
|
{
|
|
"loss": 5.0750531954690814e-05,
|
|
"grad_norm": 0.11612534523010254,
|
|
"learning_rate": 4.775862068965517e-07,
|
|
"num_tokens": 221203.0,
|
|
"completions/mean_length": 358.375,
|
|
"completions/min_length": 224.0,
|
|
"completions/max_length": 935.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 276.0,
|
|
"completions/min_terminated_length": 224.0,
|
|
"completions/max_terminated_length": 336.0,
|
|
"tools/call_frequency": 9.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0046598236076533794,
|
|
"sampling/sampling_logp_difference/max": 0.4002344608306885,
|
|
"sampling/importance_sampling_ratio/min": 0.4000450372695923,
|
|
"sampling/importance_sampling_ratio/mean": 1.2596435546875,
|
|
"sampling/importance_sampling_ratio/max": 2.5778093338012695,
|
|
"kl": 0.0003016099472006317,
|
|
"entropy": 0.07953018648549914,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.955980291590095,
|
|
"epoch": 0.000625,
|
|
"step": 24
|
|
},
|
|
{
|
|
"loss": 4.622945925802924e-05,
|
|
"grad_norm": 0.28392016887664795,
|
|
"learning_rate": 4.7586206896551725e-07,
|
|
"num_tokens": 232154.0,
|
|
"completions/mean_length": 682.75,
|
|
"completions/min_length": 291.0,
|
|
"completions/max_length": 956.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 461.25,
|
|
"completions/min_terminated_length": 291.0,
|
|
"completions/max_terminated_length": 883.0,
|
|
"tools/call_frequency": 17.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.002912044757977128,
|
|
"sampling/sampling_logp_difference/max": 0.461759090423584,
|
|
"sampling/importance_sampling_ratio/min": 0.3053264617919922,
|
|
"sampling/importance_sampling_ratio/mean": 0.8728585839271545,
|
|
"sampling/importance_sampling_ratio/max": 1.527904987335205,
|
|
"kl": 0.0003561442101371881,
|
|
"entropy": 0.06057301053078845,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.196169601753354,
|
|
"epoch": 0.0006510416666666666,
|
|
"step": 25
|
|
},
|
|
{
|
|
"loss": 7.385913340840489e-05,
|
|
"grad_norm": 0.3044932782649994,
|
|
"learning_rate": 4.7413793103448276e-07,
|
|
"num_tokens": 242048.0,
|
|
"completions/mean_length": 550.75,
|
|
"completions/min_length": 271.0,
|
|
"completions/max_length": 992.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 317.6000061035156,
|
|
"completions/min_terminated_length": 271.0,
|
|
"completions/max_terminated_length": 354.0,
|
|
"tools/call_frequency": 13.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0039165630005300045,
|
|
"sampling/sampling_logp_difference/max": 1.0759506225585938,
|
|
"sampling/importance_sampling_ratio/min": 0.17456994950771332,
|
|
"sampling/importance_sampling_ratio/mean": 0.7464568614959717,
|
|
"sampling/importance_sampling_ratio/max": 1.5292515754699707,
|
|
"kl": 0.0005547249934352294,
|
|
"entropy": 0.06094788189511746,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.334168083965778,
|
|
"epoch": 0.0006770833333333334,
|
|
"step": 26
|
|
},
|
|
{
|
|
"loss": 6.715160998282954e-05,
|
|
"grad_norm": 0.17405003309249878,
|
|
"learning_rate": 4.7241379310344827e-07,
|
|
"num_tokens": 250573.0,
|
|
"completions/mean_length": 380.0,
|
|
"completions/min_length": 261.0,
|
|
"completions/max_length": 873.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 309.5714416503906,
|
|
"completions/min_terminated_length": 261.0,
|
|
"completions/max_terminated_length": 375.0,
|
|
"tools/call_frequency": 10.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005185483489185572,
|
|
"sampling/sampling_logp_difference/max": 0.6727774739265442,
|
|
"sampling/importance_sampling_ratio/min": 0.3460983633995056,
|
|
"sampling/importance_sampling_ratio/mean": 0.9702853560447693,
|
|
"sampling/importance_sampling_ratio/max": 2.2396013736724854,
|
|
"kl": 0.00040503288073523436,
|
|
"entropy": 0.06895103328861296,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.755330117419362,
|
|
"epoch": 0.000703125,
|
|
"step": 27
|
|
},
|
|
{
|
|
"loss": 5.5044383771019056e-05,
|
|
"grad_norm": 0.1485505998134613,
|
|
"learning_rate": 4.706896551724138e-07,
|
|
"num_tokens": 259712.0,
|
|
"completions/mean_length": 456.875,
|
|
"completions/min_length": 273.0,
|
|
"completions/max_length": 920.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 409.2857360839844,
|
|
"completions/min_terminated_length": 273.0,
|
|
"completions/max_terminated_length": 920.0,
|
|
"tools/call_frequency": 12.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005413999781012535,
|
|
"sampling/sampling_logp_difference/max": 0.4519634246826172,
|
|
"sampling/importance_sampling_ratio/min": 0.4429890811443329,
|
|
"sampling/importance_sampling_ratio/mean": 1.3726776838302612,
|
|
"sampling/importance_sampling_ratio/max": 2.2206027507781982,
|
|
"kl": 0.0003276469606134924,
|
|
"entropy": 0.08287549880333245,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.363507905974984,
|
|
"epoch": 0.0007291666666666667,
|
|
"step": 28
|
|
},
|
|
{
|
|
"loss": 4.951518712914549e-05,
|
|
"grad_norm": 0.25877827405929565,
|
|
"learning_rate": 4.689655172413793e-07,
|
|
"num_tokens": 269967.0,
|
|
"completions/mean_length": 596.5,
|
|
"completions/min_length": 256.0,
|
|
"completions/max_length": 907.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 504.66668701171875,
|
|
"completions/min_terminated_length": 256.0,
|
|
"completions/max_terminated_length": 907.0,
|
|
"tools/call_frequency": 15.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.002795945620164275,
|
|
"sampling/sampling_logp_difference/max": 0.6488215923309326,
|
|
"sampling/importance_sampling_ratio/min": 0.4588918089866638,
|
|
"sampling/importance_sampling_ratio/mean": 0.9158952236175537,
|
|
"sampling/importance_sampling_ratio/max": 1.317452311515808,
|
|
"kl": 0.00031874833871370356,
|
|
"entropy": 0.049767605727538466,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.826723217964172,
|
|
"epoch": 0.0007552083333333333,
|
|
"step": 29
|
|
},
|
|
{
|
|
"loss": 4.719300341093913e-05,
|
|
"grad_norm": 0.1595955491065979,
|
|
"learning_rate": 4.672413793103448e-07,
|
|
"num_tokens": 279960.0,
|
|
"completions/mean_length": 563.375,
|
|
"completions/min_length": 274.0,
|
|
"completions/max_length": 1018.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 320.0,
|
|
"completions/min_terminated_length": 274.0,
|
|
"completions/max_terminated_length": 379.0,
|
|
"tools/call_frequency": 12.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0034552644938230515,
|
|
"sampling/sampling_logp_difference/max": 0.42853736877441406,
|
|
"sampling/importance_sampling_ratio/min": 0.30411919951438904,
|
|
"sampling/importance_sampling_ratio/mean": 1.2200520038604736,
|
|
"sampling/importance_sampling_ratio/max": 2.2717125415802,
|
|
"kl": 0.0003433458432482439,
|
|
"entropy": 0.06305162096396089,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.507702138274908,
|
|
"epoch": 0.00078125,
|
|
"step": 30
|
|
},
|
|
{
|
|
"loss": 0.23939669132232666,
|
|
"grad_norm": 2.087970733642578,
|
|
"learning_rate": 4.655172413793103e-07,
|
|
"num_tokens": 289053.0,
|
|
"completions/mean_length": 450.375,
|
|
"completions/min_length": 86.0,
|
|
"completions/max_length": 904.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 305.66668701171875,
|
|
"completions/min_terminated_length": 86.0,
|
|
"completions/max_terminated_length": 381.0,
|
|
"tools/call_frequency": 11.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00436559459194541,
|
|
"sampling/sampling_logp_difference/max": 0.4428131580352783,
|
|
"sampling/importance_sampling_ratio/min": 0.551540195941925,
|
|
"sampling/importance_sampling_ratio/mean": 0.9210087656974792,
|
|
"sampling/importance_sampling_ratio/max": 1.3286274671554565,
|
|
"kl": 0.00042575068539463246,
|
|
"entropy": 0.07597333489684388,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.663079224526882,
|
|
"epoch": 0.0008072916666666667,
|
|
"step": 31
|
|
},
|
|
{
|
|
"loss": 6.424207094823942e-05,
|
|
"grad_norm": 0.1744643747806549,
|
|
"learning_rate": 4.637931034482758e-07,
|
|
"num_tokens": 298815.0,
|
|
"completions/mean_length": 534.5,
|
|
"completions/min_length": 278.0,
|
|
"completions/max_length": 919.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 407.5,
|
|
"completions/min_terminated_length": 278.0,
|
|
"completions/max_terminated_length": 910.0,
|
|
"tools/call_frequency": 13.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0042410315945744514,
|
|
"sampling/sampling_logp_difference/max": 0.6758233308792114,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 1.031713604927063,
|
|
"sampling/importance_sampling_ratio/max": 2.438713788986206,
|
|
"kl": 0.00040169546446122695,
|
|
"entropy": 0.0736081664217636,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.375019334256649,
|
|
"epoch": 0.0008333333333333334,
|
|
"step": 32
|
|
},
|
|
{
|
|
"loss": 7.172206824179739e-05,
|
|
"grad_norm": 0.6806848645210266,
|
|
"learning_rate": 4.620689655172413e-07,
|
|
"num_tokens": 309171.0,
|
|
"completions/mean_length": 608.75,
|
|
"completions/min_length": 234.0,
|
|
"completions/max_length": 947.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 437.3999938964844,
|
|
"completions/min_terminated_length": 234.0,
|
|
"completions/max_terminated_length": 947.0,
|
|
"tools/call_frequency": 15.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0037174178287386894,
|
|
"sampling/sampling_logp_difference/max": 1.891221523284912,
|
|
"sampling/importance_sampling_ratio/min": 0.32070663571357727,
|
|
"sampling/importance_sampling_ratio/mean": 0.7887095212936401,
|
|
"sampling/importance_sampling_ratio/max": 1.2726608514785767,
|
|
"kl": 0.0003718546931850142,
|
|
"entropy": 0.05699150753207505,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.154099933803082,
|
|
"epoch": 0.000859375,
|
|
"step": 33
|
|
},
|
|
{
|
|
"loss": -0.0677376314997673,
|
|
"grad_norm": 6.155096530914307,
|
|
"learning_rate": 4.603448275862069e-07,
|
|
"num_tokens": 317891.0,
|
|
"completions/mean_length": 403.75,
|
|
"completions/min_length": 271.0,
|
|
"completions/max_length": 865.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 337.8571472167969,
|
|
"completions/min_terminated_length": 271.0,
|
|
"completions/max_terminated_length": 396.0,
|
|
"tools/call_frequency": 10.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005503959953784943,
|
|
"sampling/sampling_logp_difference/max": 0.6555323600769043,
|
|
"sampling/importance_sampling_ratio/min": 0.5667561292648315,
|
|
"sampling/importance_sampling_ratio/mean": 1.0191190242767334,
|
|
"sampling/importance_sampling_ratio/max": 1.4016114473342896,
|
|
"kl": 0.0008165286690200446,
|
|
"entropy": 0.08068722242023796,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.613661583513021,
|
|
"epoch": 0.0008854166666666666,
|
|
"step": 34
|
|
},
|
|
{
|
|
"loss": 5.972979488433339e-05,
|
|
"grad_norm": 0.22135384380817413,
|
|
"learning_rate": 4.586206896551724e-07,
|
|
"num_tokens": 327596.0,
|
|
"completions/mean_length": 527.125,
|
|
"completions/min_length": 272.0,
|
|
"completions/max_length": 918.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 303.20001220703125,
|
|
"completions/min_terminated_length": 272.0,
|
|
"completions/max_terminated_length": 333.0,
|
|
"tools/call_frequency": 13.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0036802568938583136,
|
|
"sampling/sampling_logp_difference/max": 0.453754186630249,
|
|
"sampling/importance_sampling_ratio/min": 0.6890103220939636,
|
|
"sampling/importance_sampling_ratio/mean": 1.2899754047393799,
|
|
"sampling/importance_sampling_ratio/max": 2.662895917892456,
|
|
"kl": 0.0004022571424684429,
|
|
"entropy": 0.0653612285386771,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.79808066226542,
|
|
"epoch": 0.0009114583333333333,
|
|
"step": 35
|
|
},
|
|
{
|
|
"loss": 4.123644612263888e-05,
|
|
"grad_norm": 0.18219245970249176,
|
|
"learning_rate": 4.5689655172413795e-07,
|
|
"num_tokens": 336688.0,
|
|
"completions/mean_length": 450.875,
|
|
"completions/min_length": 243.0,
|
|
"completions/max_length": 916.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 385.2857360839844,
|
|
"completions/min_terminated_length": 243.0,
|
|
"completions/max_terminated_length": 916.0,
|
|
"tools/call_frequency": 11.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0044593727216124535,
|
|
"sampling/sampling_logp_difference/max": 0.6627933979034424,
|
|
"sampling/importance_sampling_ratio/min": 0.3981926441192627,
|
|
"sampling/importance_sampling_ratio/mean": 0.7554673552513123,
|
|
"sampling/importance_sampling_ratio/max": 1.9235453605651855,
|
|
"kl": 0.00021305169502738863,
|
|
"entropy": 0.06368815246969461,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.647657308727503,
|
|
"epoch": 0.0009375,
|
|
"step": 36
|
|
},
|
|
{
|
|
"loss": 7.420648034894839e-05,
|
|
"grad_norm": 0.14686107635498047,
|
|
"learning_rate": 4.5517241379310346e-07,
|
|
"num_tokens": 346528.0,
|
|
"completions/mean_length": 544.25,
|
|
"completions/min_length": 281.0,
|
|
"completions/max_length": 921.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 422.16668701171875,
|
|
"completions/min_terminated_length": 281.0,
|
|
"completions/max_terminated_length": 898.0,
|
|
"tools/call_frequency": 14.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004388547036796808,
|
|
"sampling/sampling_logp_difference/max": 0.6111435890197754,
|
|
"sampling/importance_sampling_ratio/min": 0.36445656418800354,
|
|
"sampling/importance_sampling_ratio/mean": 1.1810851097106934,
|
|
"sampling/importance_sampling_ratio/max": 2.149569034576416,
|
|
"kl": 0.0005221854676165094,
|
|
"entropy": 0.06803273968398571,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.245145002380013,
|
|
"epoch": 0.0009635416666666667,
|
|
"step": 37
|
|
},
|
|
{
|
|
"loss": 5.548145418288186e-05,
|
|
"grad_norm": 0.15423864126205444,
|
|
"learning_rate": 4.5344827586206896e-07,
|
|
"num_tokens": 356279.0,
|
|
"completions/mean_length": 532.5,
|
|
"completions/min_length": 266.0,
|
|
"completions/max_length": 965.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 491.71429443359375,
|
|
"completions/min_terminated_length": 266.0,
|
|
"completions/max_terminated_length": 965.0,
|
|
"tools/call_frequency": 13.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0037059136666357517,
|
|
"sampling/sampling_logp_difference/max": 0.40836501121520996,
|
|
"sampling/importance_sampling_ratio/min": 0.39679059386253357,
|
|
"sampling/importance_sampling_ratio/mean": 1.1560297012329102,
|
|
"sampling/importance_sampling_ratio/max": 2.809492349624634,
|
|
"kl": 0.00035619683512777556,
|
|
"entropy": 0.0782635185169056,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.42924958653748,
|
|
"epoch": 0.0009895833333333334,
|
|
"step": 38
|
|
},
|
|
{
|
|
"loss": 0.201189786195755,
|
|
"grad_norm": 4.789113998413086,
|
|
"learning_rate": 4.5172413793103447e-07,
|
|
"num_tokens": 364270.0,
|
|
"completions/mean_length": 313.125,
|
|
"completions/min_length": 127.0,
|
|
"completions/max_length": 485.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 313.125,
|
|
"completions/min_terminated_length": 127.0,
|
|
"completions/max_terminated_length": 485.0,
|
|
"tools/call_frequency": 7.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00890029314905405,
|
|
"sampling/sampling_logp_difference/max": 0.7430671453475952,
|
|
"sampling/importance_sampling_ratio/min": 0.08373568207025528,
|
|
"sampling/importance_sampling_ratio/mean": 1.0101125240325928,
|
|
"sampling/importance_sampling_ratio/max": 2.4743096828460693,
|
|
"kl": 0.0008517875030520372,
|
|
"entropy": 0.10258017480373383,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 7.318014806136489,
|
|
"epoch": 0.001015625,
|
|
"step": 39
|
|
},
|
|
{
|
|
"loss": 3.533359995344654e-05,
|
|
"grad_norm": 0.08979543298482895,
|
|
"learning_rate": 4.5e-07,
|
|
"num_tokens": 373772.0,
|
|
"completions/mean_length": 501.5,
|
|
"completions/min_length": 14.0,
|
|
"completions/max_length": 989.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 369.5,
|
|
"completions/min_terminated_length": 14.0,
|
|
"completions/max_terminated_length": 989.0,
|
|
"tools/call_frequency": 12.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.002634549979120493,
|
|
"sampling/sampling_logp_difference/max": 0.3736910820007324,
|
|
"sampling/importance_sampling_ratio/min": 0.4946852922439575,
|
|
"sampling/importance_sampling_ratio/mean": 0.8066096305847168,
|
|
"sampling/importance_sampling_ratio/max": 1.863921046257019,
|
|
"kl": 0.00030057956246309914,
|
|
"entropy": 0.03789114218670875,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.816683830693364,
|
|
"epoch": 0.0010416666666666667,
|
|
"step": 40
|
|
},
|
|
{
|
|
"loss": 8.965920278569683e-05,
|
|
"grad_norm": 0.34930890798568726,
|
|
"learning_rate": 4.482758620689655e-07,
|
|
"num_tokens": 383052.0,
|
|
"completions/mean_length": 474.125,
|
|
"completions/min_length": 272.0,
|
|
"completions/max_length": 922.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 410.14288330078125,
|
|
"completions/min_terminated_length": 272.0,
|
|
"completions/max_terminated_length": 882.0,
|
|
"tools/call_frequency": 11.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00594820873811841,
|
|
"sampling/sampling_logp_difference/max": 0.7374162673950195,
|
|
"sampling/importance_sampling_ratio/min": 0.12759791314601898,
|
|
"sampling/importance_sampling_ratio/mean": 0.6370245218276978,
|
|
"sampling/importance_sampling_ratio/max": 1.253360390663147,
|
|
"kl": 0.0005718860593333375,
|
|
"entropy": 0.08354273182339966,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.237549209967256,
|
|
"epoch": 0.0010677083333333333,
|
|
"step": 41
|
|
},
|
|
{
|
|
"loss": 5.2547664381563663e-05,
|
|
"grad_norm": 0.17010188102722168,
|
|
"learning_rate": 4.46551724137931e-07,
|
|
"num_tokens": 393369.0,
|
|
"completions/mean_length": 603.875,
|
|
"completions/min_length": 286.0,
|
|
"completions/max_length": 906.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 431.8000183105469,
|
|
"completions/min_terminated_length": 286.0,
|
|
"completions/max_terminated_length": 895.0,
|
|
"tools/call_frequency": 15.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0038773745764046907,
|
|
"sampling/sampling_logp_difference/max": 0.515815019607544,
|
|
"sampling/importance_sampling_ratio/min": 0.199687197804451,
|
|
"sampling/importance_sampling_ratio/mean": 0.7096465826034546,
|
|
"sampling/importance_sampling_ratio/max": 1.3539584875106812,
|
|
"kl": 0.0003507028708327198,
|
|
"entropy": 0.05310575076146051,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.425766199827194,
|
|
"epoch": 0.00109375,
|
|
"step": 42
|
|
},
|
|
{
|
|
"loss": 0.09767334163188934,
|
|
"grad_norm": 2.976389169692993,
|
|
"learning_rate": 4.4482758620689656e-07,
|
|
"num_tokens": 401293.0,
|
|
"completions/mean_length": 304.125,
|
|
"completions/min_length": 207.0,
|
|
"completions/max_length": 451.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 304.125,
|
|
"completions/min_terminated_length": 207.0,
|
|
"completions/max_terminated_length": 451.0,
|
|
"tools/call_frequency": 8.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007010959088802338,
|
|
"sampling/sampling_logp_difference/max": 0.3849409818649292,
|
|
"sampling/importance_sampling_ratio/min": 0.31409361958503723,
|
|
"sampling/importance_sampling_ratio/mean": 0.6507951021194458,
|
|
"sampling/importance_sampling_ratio/max": 1.1157318353652954,
|
|
"kl": 0.0006615748116018949,
|
|
"entropy": 0.09388655051589012,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.759804220870137,
|
|
"epoch": 0.0011197916666666667,
|
|
"step": 43
|
|
},
|
|
{
|
|
"loss": 8.672561671119183e-05,
|
|
"grad_norm": 0.19920027256011963,
|
|
"learning_rate": 4.4310344827586207e-07,
|
|
"num_tokens": 411055.0,
|
|
"completions/mean_length": 534.0,
|
|
"completions/min_length": 270.0,
|
|
"completions/max_length": 910.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 320.0,
|
|
"completions/min_terminated_length": 270.0,
|
|
"completions/max_terminated_length": 397.0,
|
|
"tools/call_frequency": 13.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004351199138909578,
|
|
"sampling/sampling_logp_difference/max": 0.49504995346069336,
|
|
"sampling/importance_sampling_ratio/min": 0.2550686001777649,
|
|
"sampling/importance_sampling_ratio/mean": 0.8352887630462646,
|
|
"sampling/importance_sampling_ratio/max": 1.773252010345459,
|
|
"kl": 0.0006050881270311947,
|
|
"entropy": 0.07111403869930655,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.923332823440433,
|
|
"epoch": 0.0011458333333333333,
|
|
"step": 44
|
|
},
|
|
{
|
|
"loss": 5.482061897055246e-05,
|
|
"grad_norm": 0.12083607167005539,
|
|
"learning_rate": 4.413793103448276e-07,
|
|
"num_tokens": 421319.0,
|
|
"completions/mean_length": 597.5,
|
|
"completions/min_length": 262.0,
|
|
"completions/max_length": 912.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 415.3999938964844,
|
|
"completions/min_terminated_length": 262.0,
|
|
"completions/max_terminated_length": 899.0,
|
|
"tools/call_frequency": 15.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003673094790428877,
|
|
"sampling/sampling_logp_difference/max": 0.6855030059814453,
|
|
"sampling/importance_sampling_ratio/min": 0.16135461628437042,
|
|
"sampling/importance_sampling_ratio/mean": 1.018993854522705,
|
|
"sampling/importance_sampling_ratio/max": 1.932885766029358,
|
|
"kl": 0.0004124365600546298,
|
|
"entropy": 0.061806486337445676,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.396770410239697,
|
|
"epoch": 0.001171875,
|
|
"step": 45
|
|
},
|
|
{
|
|
"loss": 3.1239018426276743e-05,
|
|
"grad_norm": 0.07350409775972366,
|
|
"learning_rate": 4.396551724137931e-07,
|
|
"num_tokens": 432132.0,
|
|
"completions/mean_length": 665.0,
|
|
"completions/min_length": 277.0,
|
|
"completions/max_length": 911.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 536.6000366210938,
|
|
"completions/min_terminated_length": 277.0,
|
|
"completions/max_terminated_length": 911.0,
|
|
"tools/call_frequency": 17.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0024923798628151417,
|
|
"sampling/sampling_logp_difference/max": 0.6201342344284058,
|
|
"sampling/importance_sampling_ratio/min": 0.49367356300354004,
|
|
"sampling/importance_sampling_ratio/mean": 0.9417571425437927,
|
|
"sampling/importance_sampling_ratio/max": 2.393481731414795,
|
|
"kl": 0.00022237552889237122,
|
|
"entropy": 0.04745159257436171,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.772759208455682,
|
|
"epoch": 0.0011979166666666666,
|
|
"step": 46
|
|
},
|
|
{
|
|
"loss": 0.17699739336967468,
|
|
"grad_norm": 1.7483922243118286,
|
|
"learning_rate": 4.379310344827586e-07,
|
|
"num_tokens": 441202.0,
|
|
"completions/mean_length": 448.125,
|
|
"completions/min_length": 169.0,
|
|
"completions/max_length": 1008.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 368.14288330078125,
|
|
"completions/min_terminated_length": 169.0,
|
|
"completions/max_terminated_length": 985.0,
|
|
"tools/call_frequency": 10.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004451942630112171,
|
|
"sampling/sampling_logp_difference/max": 0.6487927436828613,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.5886639952659607,
|
|
"sampling/importance_sampling_ratio/max": 0.947281539440155,
|
|
"kl": 0.0007134959905670257,
|
|
"entropy": 0.06826360616832972,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.997154900804162,
|
|
"epoch": 0.0012239583333333334,
|
|
"step": 47
|
|
},
|
|
{
|
|
"loss": 0.1560358703136444,
|
|
"grad_norm": 2.6841835975646973,
|
|
"learning_rate": 4.362068965517241e-07,
|
|
"num_tokens": 450134.0,
|
|
"completions/mean_length": 431.0,
|
|
"completions/min_length": 230.0,
|
|
"completions/max_length": 913.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 290.8333435058594,
|
|
"completions/min_terminated_length": 230.0,
|
|
"completions/max_terminated_length": 382.0,
|
|
"tools/call_frequency": 11.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0055829803459346294,
|
|
"sampling/sampling_logp_difference/max": 0.45156240463256836,
|
|
"sampling/importance_sampling_ratio/min": 0.3921710252761841,
|
|
"sampling/importance_sampling_ratio/mean": 0.7257195711135864,
|
|
"sampling/importance_sampling_ratio/max": 1.4624004364013672,
|
|
"kl": 0.0010254892185912468,
|
|
"entropy": 0.08866729470901191,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.637660548090935,
|
|
"epoch": 0.00125,
|
|
"step": 48
|
|
},
|
|
{
|
|
"loss": 0.00025182493845932186,
|
|
"grad_norm": 2.370086908340454,
|
|
"learning_rate": 4.344827586206896e-07,
|
|
"num_tokens": 459192.0,
|
|
"completions/mean_length": 446.375,
|
|
"completions/min_length": 256.0,
|
|
"completions/max_length": 896.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 382.14288330078125,
|
|
"completions/min_terminated_length": 256.0,
|
|
"completions/max_terminated_length": 895.0,
|
|
"tools/call_frequency": 12.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0026751491241157055,
|
|
"sampling/sampling_logp_difference/max": 0.7721126079559326,
|
|
"sampling/importance_sampling_ratio/min": 0.6026241779327393,
|
|
"sampling/importance_sampling_ratio/mean": 1.2482657432556152,
|
|
"sampling/importance_sampling_ratio/max": 2.4711217880249023,
|
|
"kl": 0.0009107709756790427,
|
|
"entropy": 0.046249379171058536,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.714190512895584,
|
|
"epoch": 0.0012760416666666666,
|
|
"step": 49
|
|
},
|
|
{
|
|
"loss": 3.876509072142653e-05,
|
|
"grad_norm": 0.06937248259782791,
|
|
"learning_rate": 4.327586206896551e-07,
|
|
"num_tokens": 469728.0,
|
|
"completions/mean_length": 631.125,
|
|
"completions/min_length": 307.0,
|
|
"completions/max_length": 961.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 536.8333740234375,
|
|
"completions/min_terminated_length": 307.0,
|
|
"completions/max_terminated_length": 961.0,
|
|
"tools/call_frequency": 15.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0031879059970378876,
|
|
"sampling/sampling_logp_difference/max": 0.4823262691497803,
|
|
"sampling/importance_sampling_ratio/min": 0.5109031200408936,
|
|
"sampling/importance_sampling_ratio/mean": 0.9676103591918945,
|
|
"sampling/importance_sampling_ratio/max": 1.9039955139160156,
|
|
"kl": 0.00026533159871178214,
|
|
"entropy": 0.0526639383751899,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.122338488698006,
|
|
"epoch": 0.0013020833333333333,
|
|
"step": 50
|
|
},
|
|
{
|
|
"loss": 0.00012122109910706058,
|
|
"grad_norm": 0.17886631190776825,
|
|
"learning_rate": 4.310344827586206e-07,
|
|
"num_tokens": 478215.0,
|
|
"completions/mean_length": 375.25,
|
|
"completions/min_length": 256.0,
|
|
"completions/max_length": 916.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 298.0,
|
|
"completions/min_terminated_length": 256.0,
|
|
"completions/max_terminated_length": 340.0,
|
|
"tools/call_frequency": 9.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006158121861517429,
|
|
"sampling/sampling_logp_difference/max": 0.7373216152191162,
|
|
"sampling/importance_sampling_ratio/min": 0.28068816661834717,
|
|
"sampling/importance_sampling_ratio/mean": 0.7527609467506409,
|
|
"sampling/importance_sampling_ratio/max": 1.4969127178192139,
|
|
"kl": 0.0006216938672878314,
|
|
"entropy": 0.0811998926801607,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.063926000148058,
|
|
"epoch": 0.001328125,
|
|
"step": 51
|
|
},
|
|
{
|
|
"loss": 9.272761235479265e-05,
|
|
"grad_norm": 0.20103920996189117,
|
|
"learning_rate": 4.2931034482758624e-07,
|
|
"num_tokens": 487234.0,
|
|
"completions/mean_length": 441.25,
|
|
"completions/min_length": 236.0,
|
|
"completions/max_length": 898.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 291.16668701171875,
|
|
"completions/min_terminated_length": 236.0,
|
|
"completions/max_terminated_length": 324.0,
|
|
"tools/call_frequency": 11.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003737924387678504,
|
|
"sampling/sampling_logp_difference/max": 0.8087606430053711,
|
|
"sampling/importance_sampling_ratio/min": 0.30201274156570435,
|
|
"sampling/importance_sampling_ratio/mean": 0.9798649549484253,
|
|
"sampling/importance_sampling_ratio/max": 2.1515817642211914,
|
|
"kl": 0.0004854137396250735,
|
|
"entropy": 0.06504985934589058,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.36140395142138,
|
|
"epoch": 0.0013541666666666667,
|
|
"step": 52
|
|
},
|
|
{
|
|
"loss": 9.33837000047788e-05,
|
|
"grad_norm": 0.17825999855995178,
|
|
"learning_rate": 4.2758620689655174e-07,
|
|
"num_tokens": 495823.0,
|
|
"completions/mean_length": 388.25,
|
|
"completions/min_length": 287.0,
|
|
"completions/max_length": 900.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 315.14288330078125,
|
|
"completions/min_terminated_length": 287.0,
|
|
"completions/max_terminated_length": 337.0,
|
|
"tools/call_frequency": 10.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0055551426485180855,
|
|
"sampling/sampling_logp_difference/max": 0.4654419422149658,
|
|
"sampling/importance_sampling_ratio/min": 0.5057454705238342,
|
|
"sampling/importance_sampling_ratio/mean": 0.953542172908783,
|
|
"sampling/importance_sampling_ratio/max": 1.6615447998046875,
|
|
"kl": 0.000554399394786742,
|
|
"entropy": 0.07758200215175748,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.93938085436821,
|
|
"epoch": 0.0013802083333333333,
|
|
"step": 53
|
|
},
|
|
{
|
|
"loss": 0.00011007356806658208,
|
|
"grad_norm": 0.21997272968292236,
|
|
"learning_rate": 4.2586206896551725e-07,
|
|
"num_tokens": 504456.0,
|
|
"completions/mean_length": 393.125,
|
|
"completions/min_length": 262.0,
|
|
"completions/max_length": 970.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 310.71429443359375,
|
|
"completions/min_terminated_length": 262.0,
|
|
"completions/max_terminated_length": 374.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006153465248644352,
|
|
"sampling/sampling_logp_difference/max": 0.49439287185668945,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 1.0548570156097412,
|
|
"sampling/importance_sampling_ratio/max": 2.3968632221221924,
|
|
"kl": 0.0006114366224210244,
|
|
"entropy": 0.09402138902805746,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.476956544443965,
|
|
"epoch": 0.00140625,
|
|
"step": 54
|
|
},
|
|
{
|
|
"loss": 8.359746425412595e-05,
|
|
"grad_norm": 0.19514411687850952,
|
|
"learning_rate": 4.2413793103448276e-07,
|
|
"num_tokens": 514079.0,
|
|
"completions/mean_length": 517.625,
|
|
"completions/min_length": 254.0,
|
|
"completions/max_length": 913.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 285.6000061035156,
|
|
"completions/min_terminated_length": 254.0,
|
|
"completions/max_terminated_length": 335.0,
|
|
"tools/call_frequency": 13.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004124087747186422,
|
|
"sampling/sampling_logp_difference/max": 0.884941816329956,
|
|
"sampling/importance_sampling_ratio/min": 0.14841300249099731,
|
|
"sampling/importance_sampling_ratio/mean": 1.1035406589508057,
|
|
"sampling/importance_sampling_ratio/max": 2.3862318992614746,
|
|
"kl": 0.0006481431400970905,
|
|
"entropy": 0.06848104018718004,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.327478913590312,
|
|
"epoch": 0.0014322916666666666,
|
|
"step": 55
|
|
},
|
|
{
|
|
"loss": 6.543206836795434e-05,
|
|
"grad_norm": 0.14221961796283722,
|
|
"learning_rate": 4.2241379310344827e-07,
|
|
"num_tokens": 524775.0,
|
|
"completions/mean_length": 651.5,
|
|
"completions/min_length": 14.0,
|
|
"completions/max_length": 1019.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 554.1666870117188,
|
|
"completions/min_terminated_length": 14.0,
|
|
"completions/max_terminated_length": 912.0,
|
|
"tools/call_frequency": 15.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003034706227481365,
|
|
"sampling/sampling_logp_difference/max": 0.4254109859466553,
|
|
"sampling/importance_sampling_ratio/min": 0.44151103496551514,
|
|
"sampling/importance_sampling_ratio/mean": 0.7225172519683838,
|
|
"sampling/importance_sampling_ratio/max": 1.2321521043777466,
|
|
"kl": 0.0009398069050803315,
|
|
"entropy": 0.0447555755963549,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.231991661712527,
|
|
"epoch": 0.0014583333333333334,
|
|
"step": 56
|
|
},
|
|
{
|
|
"loss": 0.12315325438976288,
|
|
"grad_norm": 1.9316191673278809,
|
|
"learning_rate": 4.206896551724138e-07,
|
|
"num_tokens": 533641.0,
|
|
"completions/mean_length": 423.0,
|
|
"completions/min_length": 236.0,
|
|
"completions/max_length": 880.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 357.71429443359375,
|
|
"completions/min_terminated_length": 236.0,
|
|
"completions/max_terminated_length": 760.0,
|
|
"tools/call_frequency": 12.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0037256774958223104,
|
|
"sampling/sampling_logp_difference/max": 0.7714500427246094,
|
|
"sampling/importance_sampling_ratio/min": 0.12824466824531555,
|
|
"sampling/importance_sampling_ratio/mean": 0.6545704007148743,
|
|
"sampling/importance_sampling_ratio/max": 1.468153715133667,
|
|
"kl": 0.00048176046402659267,
|
|
"entropy": 0.06308237928897142,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.971560696139932,
|
|
"epoch": 0.001484375,
|
|
"step": 57
|
|
},
|
|
{
|
|
"loss": 0.19779449701309204,
|
|
"grad_norm": 3.6533758640289307,
|
|
"learning_rate": 4.189655172413793e-07,
|
|
"num_tokens": 542392.0,
|
|
"completions/mean_length": 407.875,
|
|
"completions/min_length": 272.0,
|
|
"completions/max_length": 926.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 333.8571472167969,
|
|
"completions/min_terminated_length": 272.0,
|
|
"completions/max_terminated_length": 383.0,
|
|
"tools/call_frequency": 10.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005720178596675396,
|
|
"sampling/sampling_logp_difference/max": 0.47936201095581055,
|
|
"sampling/importance_sampling_ratio/min": 0.1948479562997818,
|
|
"sampling/importance_sampling_ratio/mean": 1.271311640739441,
|
|
"sampling/importance_sampling_ratio/max": 2.252943277359009,
|
|
"kl": 0.0006307259282039013,
|
|
"entropy": 0.07814169942867011,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.410670073702931,
|
|
"epoch": 0.0015104166666666666,
|
|
"step": 58
|
|
},
|
|
{
|
|
"loss": 4.250071651767939e-05,
|
|
"grad_norm": 0.10488645732402802,
|
|
"learning_rate": 4.172413793103448e-07,
|
|
"num_tokens": 553272.0,
|
|
"completions/mean_length": 674.125,
|
|
"completions/min_length": 285.0,
|
|
"completions/max_length": 941.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 454.75,
|
|
"completions/min_terminated_length": 285.0,
|
|
"completions/max_terminated_length": 862.0,
|
|
"tools/call_frequency": 17.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0027292598970234394,
|
|
"sampling/sampling_logp_difference/max": 0.8170576095581055,
|
|
"sampling/importance_sampling_ratio/min": 0.31252321600914,
|
|
"sampling/importance_sampling_ratio/mean": 0.7700244188308716,
|
|
"sampling/importance_sampling_ratio/max": 1.065280795097351,
|
|
"kl": 0.0003094102976319846,
|
|
"entropy": 0.0414471928961575,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.859119538217783,
|
|
"epoch": 0.0015364583333333333,
|
|
"step": 59
|
|
},
|
|
{
|
|
"loss": 0.00012963441258762032,
|
|
"grad_norm": 0.639519453048706,
|
|
"learning_rate": 4.155172413793103e-07,
|
|
"num_tokens": 561895.0,
|
|
"completions/mean_length": 392.25,
|
|
"completions/min_length": 269.0,
|
|
"completions/max_length": 893.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 320.71429443359375,
|
|
"completions/min_terminated_length": 269.0,
|
|
"completions/max_terminated_length": 357.0,
|
|
"tools/call_frequency": 9.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005912547465413809,
|
|
"sampling/sampling_logp_difference/max": 0.6218948364257812,
|
|
"sampling/importance_sampling_ratio/min": 0.5177650451660156,
|
|
"sampling/importance_sampling_ratio/mean": 1.3758962154388428,
|
|
"sampling/importance_sampling_ratio/max": 2.0427422523498535,
|
|
"kl": 0.0007609040894749342,
|
|
"entropy": 0.08540999540127814,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.43052283115685,
|
|
"epoch": 0.0015625,
|
|
"step": 60
|
|
},
|
|
{
|
|
"loss": 0.1635953038930893,
|
|
"grad_norm": 2.5187792778015137,
|
|
"learning_rate": 4.1379310344827586e-07,
|
|
"num_tokens": 571057.0,
|
|
"completions/mean_length": 459.5,
|
|
"completions/min_length": 240.0,
|
|
"completions/max_length": 916.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 307.8333435058594,
|
|
"completions/min_terminated_length": 240.0,
|
|
"completions/max_terminated_length": 339.0,
|
|
"tools/call_frequency": 12.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004072536714375019,
|
|
"sampling/sampling_logp_difference/max": 0.4885883331298828,
|
|
"sampling/importance_sampling_ratio/min": 0.22263440489768982,
|
|
"sampling/importance_sampling_ratio/mean": 0.6854640245437622,
|
|
"sampling/importance_sampling_ratio/max": 1.466776967048645,
|
|
"kl": 0.0017422621294826968,
|
|
"entropy": 0.05582418688572943,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.144758637994528,
|
|
"epoch": 0.0015885416666666667,
|
|
"step": 61
|
|
},
|
|
{
|
|
"loss": 5.091783532407135e-05,
|
|
"grad_norm": 0.13202157616615295,
|
|
"learning_rate": 4.1206896551724137e-07,
|
|
"num_tokens": 582184.0,
|
|
"completions/mean_length": 705.25,
|
|
"completions/min_length": 274.0,
|
|
"completions/max_length": 985.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 632.1666870117188,
|
|
"completions/min_terminated_length": 274.0,
|
|
"completions/max_terminated_length": 985.0,
|
|
"tools/call_frequency": 16.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0028828911017626524,
|
|
"sampling/sampling_logp_difference/max": 0.954443097114563,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.6391115188598633,
|
|
"sampling/importance_sampling_ratio/max": 1.1153470277786255,
|
|
"kl": 0.0003236138568354363,
|
|
"entropy": 0.04818637215066701,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.475913885980844,
|
|
"epoch": 0.0016145833333333333,
|
|
"step": 62
|
|
},
|
|
{
|
|
"loss": 0.00014178639685269445,
|
|
"grad_norm": 0.4381712079048157,
|
|
"learning_rate": 4.103448275862069e-07,
|
|
"num_tokens": 590866.0,
|
|
"completions/mean_length": 398.375,
|
|
"completions/min_length": 255.0,
|
|
"completions/max_length": 933.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 398.375,
|
|
"completions/min_terminated_length": 255.0,
|
|
"completions/max_terminated_length": 933.0,
|
|
"tools/call_frequency": 10.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0047070784494280815,
|
|
"sampling/sampling_logp_difference/max": 0.4186941385269165,
|
|
"sampling/importance_sampling_ratio/min": 0.34947139024734497,
|
|
"sampling/importance_sampling_ratio/mean": 0.7317827939987183,
|
|
"sampling/importance_sampling_ratio/max": 1.1116923093795776,
|
|
"kl": 0.0008102815772872418,
|
|
"entropy": 0.06618655566126108,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.919740242883563,
|
|
"epoch": 0.001640625,
|
|
"step": 63
|
|
},
|
|
{
|
|
"loss": 0.281381219625473,
|
|
"grad_norm": 4.33756160736084,
|
|
"learning_rate": 4.086206896551724e-07,
|
|
"num_tokens": 599325.0,
|
|
"completions/mean_length": 371.75,
|
|
"completions/min_length": 223.0,
|
|
"completions/max_length": 916.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 294.0,
|
|
"completions/min_terminated_length": 223.0,
|
|
"completions/max_terminated_length": 368.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.011250000447034836,
|
|
"rewards/reward_func/std": 0.018077217042446136,
|
|
"reward": 0.011250000447034836,
|
|
"reward_std": 0.018077215179800987,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005168641451746225,
|
|
"sampling/sampling_logp_difference/max": 1.2271291017532349,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 1.0040652751922607,
|
|
"sampling/importance_sampling_ratio/max": 2.3365116119384766,
|
|
"kl": 0.0015624821248820808,
|
|
"entropy": 0.08087504806462675,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.74105865135789,
|
|
"epoch": 0.0016666666666666668,
|
|
"step": 64
|
|
},
|
|
{
|
|
"loss": 6.748238956788555e-05,
|
|
"grad_norm": 0.24954332411289215,
|
|
"learning_rate": 4.068965517241379e-07,
|
|
"num_tokens": 609007.0,
|
|
"completions/mean_length": 524.75,
|
|
"completions/min_length": 269.0,
|
|
"completions/max_length": 910.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 469.71429443359375,
|
|
"completions/min_terminated_length": 269.0,
|
|
"completions/max_terminated_length": 904.0,
|
|
"tools/call_frequency": 13.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003576460061594844,
|
|
"sampling/sampling_logp_difference/max": 0.8774768114089966,
|
|
"sampling/importance_sampling_ratio/min": 0.14383213222026825,
|
|
"sampling/importance_sampling_ratio/mean": 0.9540280103683472,
|
|
"sampling/importance_sampling_ratio/max": 1.4263296127319336,
|
|
"kl": 0.000518176234436396,
|
|
"entropy": 0.05891842325218022,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.425749273970723,
|
|
"epoch": 0.0016927083333333334,
|
|
"step": 65
|
|
},
|
|
{
|
|
"loss": 0.00013024230429437011,
|
|
"grad_norm": 0.319911926984787,
|
|
"learning_rate": 4.051724137931034e-07,
|
|
"num_tokens": 618133.0,
|
|
"completions/mean_length": 454.5,
|
|
"completions/min_length": 252.0,
|
|
"completions/max_length": 918.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 309.3333435058594,
|
|
"completions/min_terminated_length": 252.0,
|
|
"completions/max_terminated_length": 370.0,
|
|
"tools/call_frequency": 11.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0041692317463457584,
|
|
"sampling/sampling_logp_difference/max": 0.49662017822265625,
|
|
"sampling/importance_sampling_ratio/min": 0.48036420345306396,
|
|
"sampling/importance_sampling_ratio/mean": 0.9337974190711975,
|
|
"sampling/importance_sampling_ratio/max": 1.5279669761657715,
|
|
"kl": 0.0007439369364874437,
|
|
"entropy": 0.06188662478234619,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.517954733222723,
|
|
"epoch": 0.00171875,
|
|
"step": 66
|
|
},
|
|
{
|
|
"loss": 8.756962779443711e-05,
|
|
"grad_norm": 0.23356978595256805,
|
|
"learning_rate": 4.034482758620689e-07,
|
|
"num_tokens": 627371.0,
|
|
"completions/mean_length": 468.875,
|
|
"completions/min_length": 252.0,
|
|
"completions/max_length": 987.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 406.0000305175781,
|
|
"completions/min_terminated_length": 252.0,
|
|
"completions/max_terminated_length": 987.0,
|
|
"tools/call_frequency": 11.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004394870717078447,
|
|
"sampling/sampling_logp_difference/max": 0.498384952545166,
|
|
"sampling/importance_sampling_ratio/min": 0.21863745152950287,
|
|
"sampling/importance_sampling_ratio/mean": 0.7562764286994934,
|
|
"sampling/importance_sampling_ratio/max": 1.966374397277832,
|
|
"kl": 0.0005482062315422809,
|
|
"entropy": 0.07061635795980692,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.15664903447032,
|
|
"epoch": 0.0017447916666666666,
|
|
"step": 67
|
|
},
|
|
{
|
|
"loss": 0.047497715801000595,
|
|
"grad_norm": 4.288607597351074,
|
|
"learning_rate": 4.017241379310345e-07,
|
|
"num_tokens": 637063.0,
|
|
"completions/mean_length": 526.0,
|
|
"completions/min_length": 257.0,
|
|
"completions/max_length": 904.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 304.20001220703125,
|
|
"completions/min_terminated_length": 257.0,
|
|
"completions/max_terminated_length": 405.0,
|
|
"tools/call_frequency": 13.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0028123678639531136,
|
|
"sampling/sampling_logp_difference/max": 0.4388679265975952,
|
|
"sampling/importance_sampling_ratio/min": 0.42804259061813354,
|
|
"sampling/importance_sampling_ratio/mean": 1.0825767517089844,
|
|
"sampling/importance_sampling_ratio/max": 1.7897344827651978,
|
|
"kl": 0.0006937088937775115,
|
|
"entropy": 0.05416678136680275,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.682278875261545,
|
|
"epoch": 0.0017708333333333332,
|
|
"step": 68
|
|
},
|
|
{
|
|
"loss": 0.3702544569969177,
|
|
"grad_norm": 1.6178300380706787,
|
|
"learning_rate": 4e-07,
|
|
"num_tokens": 647179.0,
|
|
"completions/mean_length": 578.625,
|
|
"completions/min_length": 105.0,
|
|
"completions/max_length": 923.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 387.6000061035156,
|
|
"completions/min_terminated_length": 105.0,
|
|
"completions/max_terminated_length": 923.0,
|
|
"tools/call_frequency": 15.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0023824286181479692,
|
|
"sampling/sampling_logp_difference/max": 0.3518613576889038,
|
|
"sampling/importance_sampling_ratio/min": 0.4380924701690674,
|
|
"sampling/importance_sampling_ratio/mean": 1.1113414764404297,
|
|
"sampling/importance_sampling_ratio/max": 2.210801601409912,
|
|
"kl": 0.00048735409018263454,
|
|
"entropy": 0.04531048284843564,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.324252692982554,
|
|
"epoch": 0.001796875,
|
|
"step": 69
|
|
},
|
|
{
|
|
"loss": 0.18352027237415314,
|
|
"grad_norm": 1.570194125175476,
|
|
"learning_rate": 3.9827586206896554e-07,
|
|
"num_tokens": 657611.0,
|
|
"completions/mean_length": 617.625,
|
|
"completions/min_length": 278.0,
|
|
"completions/max_length": 917.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 329.75,
|
|
"completions/min_terminated_length": 278.0,
|
|
"completions/max_terminated_length": 369.0,
|
|
"tools/call_frequency": 15.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0031271539628505707,
|
|
"sampling/sampling_logp_difference/max": 0.9900598526000977,
|
|
"sampling/importance_sampling_ratio/min": 0.2617069482803345,
|
|
"sampling/importance_sampling_ratio/mean": 0.6101282835006714,
|
|
"sampling/importance_sampling_ratio/max": 1.068818211555481,
|
|
"kl": 0.001243382570464746,
|
|
"entropy": 0.047085250611416996,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.614635061472654,
|
|
"epoch": 0.0018229166666666667,
|
|
"step": 70
|
|
},
|
|
{
|
|
"loss": 0.00019148402498103678,
|
|
"grad_norm": 0.3739577531814575,
|
|
"learning_rate": 3.9655172413793105e-07,
|
|
"num_tokens": 665432.0,
|
|
"completions/mean_length": 292.125,
|
|
"completions/min_length": 262.0,
|
|
"completions/max_length": 331.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 292.125,
|
|
"completions/min_terminated_length": 262.0,
|
|
"completions/max_terminated_length": 331.0,
|
|
"tools/call_frequency": 8.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007045305799692869,
|
|
"sampling/sampling_logp_difference/max": 0.4629080295562744,
|
|
"sampling/importance_sampling_ratio/min": 0.5882341861724854,
|
|
"sampling/importance_sampling_ratio/mean": 1.0942822694778442,
|
|
"sampling/importance_sampling_ratio/max": 2.002429485321045,
|
|
"kl": 0.0009478337087784894,
|
|
"entropy": 0.07637359201908112,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 7.603707263246179,
|
|
"epoch": 0.0018489583333333333,
|
|
"step": 71
|
|
},
|
|
{
|
|
"loss": 0.2256423979997635,
|
|
"grad_norm": 1.755808711051941,
|
|
"learning_rate": 3.9482758620689656e-07,
|
|
"num_tokens": 675870.0,
|
|
"completions/mean_length": 618.375,
|
|
"completions/min_length": 272.0,
|
|
"completions/max_length": 947.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 520.1666870117188,
|
|
"completions/min_terminated_length": 272.0,
|
|
"completions/max_terminated_length": 947.0,
|
|
"tools/call_frequency": 15.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0035377012100070715,
|
|
"sampling/sampling_logp_difference/max": 0.8321799039840698,
|
|
"sampling/importance_sampling_ratio/min": 0.2189892828464508,
|
|
"sampling/importance_sampling_ratio/mean": 0.8904492855072021,
|
|
"sampling/importance_sampling_ratio/max": 2.221644401550293,
|
|
"kl": 0.001277914462662011,
|
|
"entropy": 0.0552240404067561,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.137599328532815,
|
|
"epoch": 0.001875,
|
|
"step": 72
|
|
},
|
|
{
|
|
"loss": 0.0002286221570102498,
|
|
"grad_norm": 0.8774559497833252,
|
|
"learning_rate": 3.9310344827586207e-07,
|
|
"num_tokens": 685166.0,
|
|
"completions/mean_length": 475.875,
|
|
"completions/min_length": 243.0,
|
|
"completions/max_length": 1001.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 300.8333435058594,
|
|
"completions/min_terminated_length": 243.0,
|
|
"completions/max_terminated_length": 317.0,
|
|
"tools/call_frequency": 10.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0048016938380897045,
|
|
"sampling/sampling_logp_difference/max": 1.3366353511810303,
|
|
"sampling/importance_sampling_ratio/min": 0.15849082171916962,
|
|
"sampling/importance_sampling_ratio/mean": 0.7896101474761963,
|
|
"sampling/importance_sampling_ratio/max": 2.249459981918335,
|
|
"kl": 0.0015927808453852776,
|
|
"entropy": 0.06800470710732043,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.87636312842369,
|
|
"epoch": 0.0019010416666666668,
|
|
"step": 73
|
|
},
|
|
{
|
|
"loss": 0.0001120783417718485,
|
|
"grad_norm": 0.25287508964538574,
|
|
"learning_rate": 3.913793103448276e-07,
|
|
"num_tokens": 694380.0,
|
|
"completions/mean_length": 466.375,
|
|
"completions/min_length": 259.0,
|
|
"completions/max_length": 868.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 332.5,
|
|
"completions/min_terminated_length": 259.0,
|
|
"completions/max_terminated_length": 447.0,
|
|
"tools/call_frequency": 12.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004081926308572292,
|
|
"sampling/sampling_logp_difference/max": 0.5778262615203857,
|
|
"sampling/importance_sampling_ratio/min": 0.15823151171207428,
|
|
"sampling/importance_sampling_ratio/mean": 1.0285694599151611,
|
|
"sampling/importance_sampling_ratio/max": 2.9633452892303467,
|
|
"kl": 0.0007673393743061752,
|
|
"entropy": 0.07295832026284188,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.762696349993348,
|
|
"epoch": 0.0019270833333333334,
|
|
"step": 74
|
|
},
|
|
{
|
|
"loss": 0.09944476187229156,
|
|
"grad_norm": 4.064919471740723,
|
|
"learning_rate": 3.896551724137931e-07,
|
|
"num_tokens": 702847.0,
|
|
"completions/mean_length": 371.75,
|
|
"completions/min_length": 255.0,
|
|
"completions/max_length": 896.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 371.75,
|
|
"completions/min_terminated_length": 255.0,
|
|
"completions/max_terminated_length": 896.0,
|
|
"tools/call_frequency": 9.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005073638632893562,
|
|
"sampling/sampling_logp_difference/max": 0.5905663967132568,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.9095017910003662,
|
|
"sampling/importance_sampling_ratio/max": 2.2085916996002197,
|
|
"kl": 0.0021330338895495515,
|
|
"entropy": 0.06320790946483612,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.958335928618908,
|
|
"epoch": 0.001953125,
|
|
"step": 75
|
|
},
|
|
{
|
|
"loss": 0.00032947928411886096,
|
|
"grad_norm": 1.5639448165893555,
|
|
"learning_rate": 3.879310344827586e-07,
|
|
"num_tokens": 712068.0,
|
|
"completions/mean_length": 467.375,
|
|
"completions/min_length": 271.0,
|
|
"completions/max_length": 914.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 403.5714416503906,
|
|
"completions/min_terminated_length": 271.0,
|
|
"completions/max_terminated_length": 910.0,
|
|
"tools/call_frequency": 11.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00445514265447855,
|
|
"sampling/sampling_logp_difference/max": 0.49289560317993164,
|
|
"sampling/importance_sampling_ratio/min": 0.42830783128738403,
|
|
"sampling/importance_sampling_ratio/mean": 1.024637222290039,
|
|
"sampling/importance_sampling_ratio/max": 2.5617618560791016,
|
|
"kl": 0.002158068635253585,
|
|
"entropy": 0.06738780299201608,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.103176252916455,
|
|
"epoch": 0.001979166666666667,
|
|
"step": 76
|
|
},
|
|
{
|
|
"loss": 5.172662713448517e-05,
|
|
"grad_norm": 0.10409528017044067,
|
|
"learning_rate": 3.862068965517241e-07,
|
|
"num_tokens": 722361.0,
|
|
"completions/mean_length": 601.125,
|
|
"completions/min_length": 263.0,
|
|
"completions/max_length": 884.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 509.5,
|
|
"completions/min_terminated_length": 263.0,
|
|
"completions/max_terminated_length": 877.0,
|
|
"tools/call_frequency": 16.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003133634803816676,
|
|
"sampling/sampling_logp_difference/max": 0.3853919506072998,
|
|
"sampling/importance_sampling_ratio/min": 0.4599654972553253,
|
|
"sampling/importance_sampling_ratio/mean": 1.134650468826294,
|
|
"sampling/importance_sampling_ratio/max": 2.0692203044891357,
|
|
"kl": 0.0003829032320936676,
|
|
"entropy": 0.06137791194487363,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.306931715458632,
|
|
"epoch": 0.0020052083333333332,
|
|
"step": 77
|
|
},
|
|
{
|
|
"loss": 0.2285258173942566,
|
|
"grad_norm": 3.369677782058716,
|
|
"learning_rate": 3.844827586206896e-07,
|
|
"num_tokens": 730786.0,
|
|
"completions/mean_length": 367.5,
|
|
"completions/min_length": 244.0,
|
|
"completions/max_length": 865.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 367.5,
|
|
"completions/min_terminated_length": 244.0,
|
|
"completions/max_terminated_length": 865.0,
|
|
"tools/call_frequency": 10.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005530106835067272,
|
|
"sampling/sampling_logp_difference/max": 0.5499169826507568,
|
|
"sampling/importance_sampling_ratio/min": 0.29444000124931335,
|
|
"sampling/importance_sampling_ratio/mean": 0.8568315505981445,
|
|
"sampling/importance_sampling_ratio/max": 1.8049551248550415,
|
|
"kl": 0.0013267821695990278,
|
|
"entropy": 0.08332669688388705,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.634299155324697,
|
|
"epoch": 0.00203125,
|
|
"step": 78
|
|
},
|
|
{
|
|
"loss": 5.426750794867985e-05,
|
|
"grad_norm": 0.09505382925271988,
|
|
"learning_rate": 3.827586206896551e-07,
|
|
"num_tokens": 740973.0,
|
|
"completions/mean_length": 587.875,
|
|
"completions/min_length": 271.0,
|
|
"completions/max_length": 918.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 495.5,
|
|
"completions/min_terminated_length": 271.0,
|
|
"completions/max_terminated_length": 918.0,
|
|
"tools/call_frequency": 15.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.002152449917048216,
|
|
"sampling/sampling_logp_difference/max": 0.6485425233840942,
|
|
"sampling/importance_sampling_ratio/min": 0.4880044162273407,
|
|
"sampling/importance_sampling_ratio/mean": 0.9390411972999573,
|
|
"sampling/importance_sampling_ratio/max": 1.7447205781936646,
|
|
"kl": 0.00036825446841248777,
|
|
"entropy": 0.04440738243283704,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.775712365284562,
|
|
"epoch": 0.0020572916666666665,
|
|
"step": 79
|
|
},
|
|
{
|
|
"loss": 0.10231752693653107,
|
|
"grad_norm": 2.1638376712799072,
|
|
"learning_rate": 3.810344827586207e-07,
|
|
"num_tokens": 750104.0,
|
|
"completions/mean_length": 456.375,
|
|
"completions/min_length": 213.0,
|
|
"completions/max_length": 1017.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 298.16668701171875,
|
|
"completions/min_terminated_length": 213.0,
|
|
"completions/max_terminated_length": 395.0,
|
|
"tools/call_frequency": 10.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004695703275501728,
|
|
"sampling/sampling_logp_difference/max": 1.3471412658691406,
|
|
"sampling/importance_sampling_ratio/min": 0.2171163707971573,
|
|
"sampling/importance_sampling_ratio/mean": 0.843420684337616,
|
|
"sampling/importance_sampling_ratio/max": 1.2368226051330566,
|
|
"kl": 0.0010103452696057502,
|
|
"entropy": 0.0694628341589123,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.783677212893963,
|
|
"epoch": 0.0020833333333333333,
|
|
"step": 80
|
|
},
|
|
{
|
|
"loss": 7.571020978502929e-05,
|
|
"grad_norm": 0.13844043016433716,
|
|
"learning_rate": 3.793103448275862e-07,
|
|
"num_tokens": 759373.0,
|
|
"completions/mean_length": 472.5,
|
|
"completions/min_length": 256.0,
|
|
"completions/max_length": 1018.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 399.0000305175781,
|
|
"completions/min_terminated_length": 256.0,
|
|
"completions/max_terminated_length": 1018.0,
|
|
"tools/call_frequency": 10.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00414977828040719,
|
|
"sampling/sampling_logp_difference/max": 0.560932993888855,
|
|
"sampling/importance_sampling_ratio/min": 0.3751889169216156,
|
|
"sampling/importance_sampling_ratio/mean": 1.0670043230056763,
|
|
"sampling/importance_sampling_ratio/max": 1.625799298286438,
|
|
"kl": 0.0005246041882855934,
|
|
"entropy": 0.07239746814593673,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.92165930941701,
|
|
"epoch": 0.002109375,
|
|
"step": 81
|
|
},
|
|
{
|
|
"loss": 0.11225898563861847,
|
|
"grad_norm": 2.709763765335083,
|
|
"learning_rate": 3.775862068965517e-07,
|
|
"num_tokens": 769240.0,
|
|
"completions/mean_length": 547.25,
|
|
"completions/min_length": 17.0,
|
|
"completions/max_length": 913.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 226.5,
|
|
"completions/min_terminated_length": 17.0,
|
|
"completions/max_terminated_length": 317.0,
|
|
"tools/call_frequency": 14.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0024680851493030787,
|
|
"sampling/sampling_logp_difference/max": 0.35571300983428955,
|
|
"sampling/importance_sampling_ratio/min": 0.5412890911102295,
|
|
"sampling/importance_sampling_ratio/mean": 1.0960311889648438,
|
|
"sampling/importance_sampling_ratio/max": 1.7128722667694092,
|
|
"kl": 0.0014146232635319222,
|
|
"entropy": 0.032854694058187306,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.654339293017983,
|
|
"epoch": 0.0021354166666666665,
|
|
"step": 82
|
|
},
|
|
{
|
|
"loss": 0.3167492151260376,
|
|
"grad_norm": 2.9918735027313232,
|
|
"learning_rate": 3.758620689655172e-07,
|
|
"num_tokens": 777841.0,
|
|
"completions/mean_length": 390.25,
|
|
"completions/min_length": 14.0,
|
|
"completions/max_length": 918.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 223.33334350585938,
|
|
"completions/min_terminated_length": 14.0,
|
|
"completions/max_terminated_length": 381.0,
|
|
"tools/call_frequency": 10.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0036312020383775234,
|
|
"sampling/sampling_logp_difference/max": 0.6204323768615723,
|
|
"sampling/importance_sampling_ratio/min": 0.3736453652381897,
|
|
"sampling/importance_sampling_ratio/mean": 0.7378817200660706,
|
|
"sampling/importance_sampling_ratio/max": 1.590867042541504,
|
|
"kl": 0.001401739693392301,
|
|
"entropy": 0.05958460038527846,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.606628136709332,
|
|
"epoch": 0.0021614583333333334,
|
|
"step": 83
|
|
},
|
|
{
|
|
"loss": 0.0001225034357048571,
|
|
"grad_norm": 0.2906716763973236,
|
|
"learning_rate": 3.7413793103448276e-07,
|
|
"num_tokens": 787532.0,
|
|
"completions/mean_length": 525.125,
|
|
"completions/min_length": 258.0,
|
|
"completions/max_length": 903.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 303.8000183105469,
|
|
"completions/min_terminated_length": 258.0,
|
|
"completions/max_terminated_length": 346.0,
|
|
"tools/call_frequency": 13.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.0,
|
|
"reward": 0.0,
|
|
"reward_std": 0.0,
|
|
"frac_reward_zero_std": 1.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003803050145506859,
|
|
"sampling/sampling_logp_difference/max": 0.4165186882019043,
|
|
"sampling/importance_sampling_ratio/min": 0.27582019567489624,
|
|
"sampling/importance_sampling_ratio/mean": 0.8301675915718079,
|
|
"sampling/importance_sampling_ratio/max": 2.0671372413635254,
|
|
"kl": 0.0009060970432983595,
|
|
"entropy": 0.06211770256049931,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.076002571731806,
|
|
"epoch": 0.0021875,
|
|
"step": 84
|
|
},
|
|
{
|
|
"loss": 0.0410054549574852,
|
|
"grad_norm": 4.415999412536621,
|
|
"learning_rate": 3.7241379310344827e-07,
|
|
"num_tokens": 796038.0,
|
|
"completions/mean_length": 377.0,
|
|
"completions/min_length": 246.0,
|
|
"completions/max_length": 995.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 377.0,
|
|
"completions/min_terminated_length": 246.0,
|
|
"completions/max_terminated_length": 995.0,
|
|
"tools/call_frequency": 9.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004189161583781242,
|
|
"sampling/sampling_logp_difference/max": 0.415729820728302,
|
|
"sampling/importance_sampling_ratio/min": 0.5751233100891113,
|
|
"sampling/importance_sampling_ratio/mean": 0.8906446695327759,
|
|
"sampling/importance_sampling_ratio/max": 1.3490046262741089,
|
|
"kl": 0.0013324315741556347,
|
|
"entropy": 0.06493875803425908,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.188016356900334,
|
|
"epoch": 0.0022135416666666666,
|
|
"step": 85
|
|
},
|
|
{
|
|
"loss": 0.18268167972564697,
|
|
"grad_norm": 3.0740888118743896,
|
|
"learning_rate": 3.706896551724138e-07,
|
|
"num_tokens": 804447.0,
|
|
"completions/mean_length": 364.5,
|
|
"completions/min_length": 230.0,
|
|
"completions/max_length": 901.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 287.8571472167969,
|
|
"completions/min_terminated_length": 230.0,
|
|
"completions/max_terminated_length": 382.0,
|
|
"tools/call_frequency": 9.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005412322003394365,
|
|
"sampling/sampling_logp_difference/max": 0.870719313621521,
|
|
"sampling/importance_sampling_ratio/min": 0.32040998339653015,
|
|
"sampling/importance_sampling_ratio/mean": 0.7531617879867554,
|
|
"sampling/importance_sampling_ratio/max": 1.2364206314086914,
|
|
"kl": 0.0026889686960203107,
|
|
"entropy": 0.07111081504262984,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.888756342232227,
|
|
"epoch": 0.0022395833333333334,
|
|
"step": 86
|
|
},
|
|
{
|
|
"loss": 0.23346655070781708,
|
|
"grad_norm": 1.97305166721344,
|
|
"learning_rate": 3.689655172413793e-07,
|
|
"num_tokens": 814313.0,
|
|
"completions/mean_length": 547.5,
|
|
"completions/min_length": 271.0,
|
|
"completions/max_length": 969.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 319.3999938964844,
|
|
"completions/min_terminated_length": 271.0,
|
|
"completions/max_terminated_length": 358.0,
|
|
"tools/call_frequency": 13.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.08062499761581421,
|
|
"rewards/reward_func/std": 0.22007203102111816,
|
|
"reward": 0.08062499761581421,
|
|
"reward_std": 0.22007201611995697,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003668141085654497,
|
|
"sampling/sampling_logp_difference/max": 0.4533498287200928,
|
|
"sampling/importance_sampling_ratio/min": 0.3676047921180725,
|
|
"sampling/importance_sampling_ratio/mean": 0.8992157578468323,
|
|
"sampling/importance_sampling_ratio/max": 1.903110384941101,
|
|
"kl": 0.0009356735870369448,
|
|
"entropy": 0.05884273396804929,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.534716825932264,
|
|
"epoch": 0.002265625,
|
|
"step": 87
|
|
},
|
|
{
|
|
"loss": 0.06900765001773834,
|
|
"grad_norm": 2.70992374420166,
|
|
"learning_rate": 3.6724137931034485e-07,
|
|
"num_tokens": 824049.0,
|
|
"completions/mean_length": 531.5,
|
|
"completions/min_length": 248.0,
|
|
"completions/max_length": 916.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 308.3999938964844,
|
|
"completions/min_terminated_length": 248.0,
|
|
"completions/max_terminated_length": 338.0,
|
|
"tools/call_frequency": 13.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003263751044869423,
|
|
"sampling/sampling_logp_difference/max": 1.243861198425293,
|
|
"sampling/importance_sampling_ratio/min": 0.15755848586559296,
|
|
"sampling/importance_sampling_ratio/mean": 0.7777807116508484,
|
|
"sampling/importance_sampling_ratio/max": 1.8547863960266113,
|
|
"kl": 0.0025930205556505825,
|
|
"entropy": 0.04691084288060665,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.109845971688628,
|
|
"epoch": 0.0022916666666666667,
|
|
"step": 88
|
|
},
|
|
{
|
|
"loss": -0.018843263387680054,
|
|
"grad_norm": 6.513974189758301,
|
|
"learning_rate": 3.6551724137931036e-07,
|
|
"num_tokens": 833705.0,
|
|
"completions/mean_length": 520.625,
|
|
"completions/min_length": 236.0,
|
|
"completions/max_length": 924.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 399.16668701171875,
|
|
"completions/min_terminated_length": 236.0,
|
|
"completions/max_terminated_length": 924.0,
|
|
"tools/call_frequency": 13.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004058406688272953,
|
|
"sampling/sampling_logp_difference/max": 0.6853499412536621,
|
|
"sampling/importance_sampling_ratio/min": 0.2979694902896881,
|
|
"sampling/importance_sampling_ratio/mean": 1.156081199645996,
|
|
"sampling/importance_sampling_ratio/max": 2.4297683238983154,
|
|
"kl": 0.0015220412642520387,
|
|
"entropy": 0.05867711687460542,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.647199613973498,
|
|
"epoch": 0.0023177083333333335,
|
|
"step": 89
|
|
},
|
|
{
|
|
"loss": -0.03724433109164238,
|
|
"grad_norm": 4.886791706085205,
|
|
"learning_rate": 3.6379310344827587e-07,
|
|
"num_tokens": 842741.0,
|
|
"completions/mean_length": 443.25,
|
|
"completions/min_length": 255.0,
|
|
"completions/max_length": 885.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 380.2857360839844,
|
|
"completions/min_terminated_length": 255.0,
|
|
"completions/max_terminated_length": 885.0,
|
|
"tools/call_frequency": 11.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004349705297499895,
|
|
"sampling/sampling_logp_difference/max": 0.48426711559295654,
|
|
"sampling/importance_sampling_ratio/min": 0.2715539336204529,
|
|
"sampling/importance_sampling_ratio/mean": 0.943439245223999,
|
|
"sampling/importance_sampling_ratio/max": 2.5567474365234375,
|
|
"kl": 0.002848195182195923,
|
|
"entropy": 0.07823656545951962,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.210258081555367,
|
|
"epoch": 0.00234375,
|
|
"step": 90
|
|
},
|
|
{
|
|
"loss": 0.032495807856321335,
|
|
"grad_norm": 2.5754916667938232,
|
|
"learning_rate": 3.620689655172414e-07,
|
|
"num_tokens": 851729.0,
|
|
"completions/mean_length": 437.75,
|
|
"completions/min_length": 235.0,
|
|
"completions/max_length": 941.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 437.75,
|
|
"completions/min_terminated_length": 235.0,
|
|
"completions/max_terminated_length": 941.0,
|
|
"tools/call_frequency": 11.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0042322175577282906,
|
|
"sampling/sampling_logp_difference/max": 0.8520015478134155,
|
|
"sampling/importance_sampling_ratio/min": 0.3146383762359619,
|
|
"sampling/importance_sampling_ratio/mean": 0.6506801843643188,
|
|
"sampling/importance_sampling_ratio/max": 0.9957988858222961,
|
|
"kl": 0.0018773903993860586,
|
|
"entropy": 0.053038322599604726,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.307958789169788,
|
|
"epoch": 0.0023697916666666667,
|
|
"step": 91
|
|
},
|
|
{
|
|
"loss": 0.22543008625507355,
|
|
"grad_norm": 4.455987930297852,
|
|
"learning_rate": 3.603448275862069e-07,
|
|
"num_tokens": 859847.0,
|
|
"completions/mean_length": 328.875,
|
|
"completions/min_length": 279.0,
|
|
"completions/max_length": 415.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 328.875,
|
|
"completions/min_terminated_length": 279.0,
|
|
"completions/max_terminated_length": 415.0,
|
|
"tools/call_frequency": 8.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007280445657670498,
|
|
"sampling/sampling_logp_difference/max": 0.4824948310852051,
|
|
"sampling/importance_sampling_ratio/min": 0.5559368133544922,
|
|
"sampling/importance_sampling_ratio/mean": 0.9710679054260254,
|
|
"sampling/importance_sampling_ratio/max": 2.022029161453247,
|
|
"kl": 0.003891883166943444,
|
|
"entropy": 0.09541127784177661,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.783526951447129,
|
|
"epoch": 0.002395833333333333,
|
|
"step": 92
|
|
},
|
|
{
|
|
"loss": 0.2872963547706604,
|
|
"grad_norm": 5.897304058074951,
|
|
"learning_rate": 3.586206896551724e-07,
|
|
"num_tokens": 868733.0,
|
|
"completions/mean_length": 424.375,
|
|
"completions/min_length": 254.0,
|
|
"completions/max_length": 895.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 272.5,
|
|
"completions/min_terminated_length": 254.0,
|
|
"completions/max_terminated_length": 306.0,
|
|
"tools/call_frequency": 11.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0035844917874783278,
|
|
"sampling/sampling_logp_difference/max": 0.748805046081543,
|
|
"sampling/importance_sampling_ratio/min": 0.2048145979642868,
|
|
"sampling/importance_sampling_ratio/mean": 1.0041184425354004,
|
|
"sampling/importance_sampling_ratio/max": 2.300896167755127,
|
|
"kl": 0.0044286388210821315,
|
|
"entropy": 0.05557826370932162,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.38397977128625,
|
|
"epoch": 0.002421875,
|
|
"step": 93
|
|
},
|
|
{
|
|
"loss": 0.0941847562789917,
|
|
"grad_norm": 5.983191967010498,
|
|
"learning_rate": 3.568965517241379e-07,
|
|
"num_tokens": 877380.0,
|
|
"completions/mean_length": 395.125,
|
|
"completions/min_length": 253.0,
|
|
"completions/max_length": 889.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 324.5714416503906,
|
|
"completions/min_terminated_length": 253.0,
|
|
"completions/max_terminated_length": 450.0,
|
|
"tools/call_frequency": 9.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005478454288095236,
|
|
"sampling/sampling_logp_difference/max": 0.9258518218994141,
|
|
"sampling/importance_sampling_ratio/min": 0.28980544209480286,
|
|
"sampling/importance_sampling_ratio/mean": 1.2061667442321777,
|
|
"sampling/importance_sampling_ratio/max": 2.0317187309265137,
|
|
"kl": 0.00429094387072837,
|
|
"entropy": 0.08049777243286371,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.554597036913037,
|
|
"epoch": 0.002447916666666667,
|
|
"step": 94
|
|
},
|
|
{
|
|
"loss": 0.11975863575935364,
|
|
"grad_norm": 3.674995183944702,
|
|
"learning_rate": 3.551724137931034e-07,
|
|
"num_tokens": 886414.0,
|
|
"completions/mean_length": 443.625,
|
|
"completions/min_length": 245.0,
|
|
"completions/max_length": 897.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 378.8571472167969,
|
|
"completions/min_terminated_length": 245.0,
|
|
"completions/max_terminated_length": 795.0,
|
|
"tools/call_frequency": 12.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004570898599922657,
|
|
"sampling/sampling_logp_difference/max": 0.8744406700134277,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.8802112936973572,
|
|
"sampling/importance_sampling_ratio/max": 1.7730200290679932,
|
|
"kl": 0.006708486099796573,
|
|
"entropy": 0.04953509464394301,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.829270949587226,
|
|
"epoch": 0.0024739583333333332,
|
|
"step": 95
|
|
},
|
|
{
|
|
"loss": 0.580945611000061,
|
|
"grad_norm": 3.1283204555511475,
|
|
"learning_rate": 3.534482758620689e-07,
|
|
"num_tokens": 895625.0,
|
|
"completions/mean_length": 465.625,
|
|
"completions/min_length": 230.0,
|
|
"completions/max_length": 1005.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 395.71429443359375,
|
|
"completions/min_terminated_length": 230.0,
|
|
"completions/max_terminated_length": 1005.0,
|
|
"tools/call_frequency": 10.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003726779483258724,
|
|
"sampling/sampling_logp_difference/max": 0.5279864072799683,
|
|
"sampling/importance_sampling_ratio/min": 0.6423169374465942,
|
|
"sampling/importance_sampling_ratio/mean": 1.2203757762908936,
|
|
"sampling/importance_sampling_ratio/max": 2.5857577323913574,
|
|
"kl": 0.004018513996925321,
|
|
"entropy": 0.0660056765191257,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.495224792510271,
|
|
"epoch": 0.0025,
|
|
"step": 96
|
|
},
|
|
{
|
|
"loss": -0.205459862947464,
|
|
"grad_norm": 7.980952739715576,
|
|
"learning_rate": 3.517241379310344e-07,
|
|
"num_tokens": 904290.0,
|
|
"completions/mean_length": 396.625,
|
|
"completions/min_length": 269.0,
|
|
"completions/max_length": 895.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 325.4285888671875,
|
|
"completions/min_terminated_length": 269.0,
|
|
"completions/max_terminated_length": 484.0,
|
|
"tools/call_frequency": 9.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0053188251331448555,
|
|
"sampling/sampling_logp_difference/max": 0.4320927858352661,
|
|
"sampling/importance_sampling_ratio/min": 0.4420548379421234,
|
|
"sampling/importance_sampling_ratio/mean": 1.0731589794158936,
|
|
"sampling/importance_sampling_ratio/max": 2.2819418907165527,
|
|
"kl": 0.0034033727715723217,
|
|
"entropy": 0.08061148156411946,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.271773666143417,
|
|
"epoch": 0.0025260416666666665,
|
|
"step": 97
|
|
},
|
|
{
|
|
"loss": 0.13967831432819366,
|
|
"grad_norm": 3.22408390045166,
|
|
"learning_rate": 3.5e-07,
|
|
"num_tokens": 913405.0,
|
|
"completions/mean_length": 453.5,
|
|
"completions/min_length": 147.0,
|
|
"completions/max_length": 971.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 394.2857360839844,
|
|
"completions/min_terminated_length": 147.0,
|
|
"completions/max_terminated_length": 971.0,
|
|
"tools/call_frequency": 11.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00460997736081481,
|
|
"sampling/sampling_logp_difference/max": 0.4426417350769043,
|
|
"sampling/importance_sampling_ratio/min": 0.29462188482284546,
|
|
"sampling/importance_sampling_ratio/mean": 0.9616898894309998,
|
|
"sampling/importance_sampling_ratio/max": 1.754823923110962,
|
|
"kl": 0.0016035606804507552,
|
|
"entropy": 0.07924353866837919,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.487173549830914,
|
|
"epoch": 0.0025520833333333333,
|
|
"step": 98
|
|
},
|
|
{
|
|
"loss": -0.1679742932319641,
|
|
"grad_norm": 6.5833611488342285,
|
|
"learning_rate": 3.482758620689655e-07,
|
|
"num_tokens": 922017.0,
|
|
"completions/mean_length": 391.0,
|
|
"completions/min_length": 284.0,
|
|
"completions/max_length": 964.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 309.14288330078125,
|
|
"completions/min_terminated_length": 284.0,
|
|
"completions/max_terminated_length": 368.0,
|
|
"tools/call_frequency": 9.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0054966164752841,
|
|
"sampling/sampling_logp_difference/max": 0.6456756591796875,
|
|
"sampling/importance_sampling_ratio/min": 0.1540312021970749,
|
|
"sampling/importance_sampling_ratio/mean": 0.7367682456970215,
|
|
"sampling/importance_sampling_ratio/max": 1.4382715225219727,
|
|
"kl": 0.003967820608522743,
|
|
"entropy": 0.07931078062392771,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.588951583951712,
|
|
"epoch": 0.002578125,
|
|
"step": 99
|
|
},
|
|
{
|
|
"loss": -0.06838738918304443,
|
|
"grad_norm": 6.627912998199463,
|
|
"learning_rate": 3.4655172413793105e-07,
|
|
"num_tokens": 930625.0,
|
|
"completions/mean_length": 390.375,
|
|
"completions/min_length": 248.0,
|
|
"completions/max_length": 917.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 315.14288330078125,
|
|
"completions/min_terminated_length": 248.0,
|
|
"completions/max_terminated_length": 381.0,
|
|
"tools/call_frequency": 9.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006150159519165754,
|
|
"sampling/sampling_logp_difference/max": 0.4266536235809326,
|
|
"sampling/importance_sampling_ratio/min": 0.5355611443519592,
|
|
"sampling/importance_sampling_ratio/mean": 1.066882610321045,
|
|
"sampling/importance_sampling_ratio/max": 1.594473123550415,
|
|
"kl": 0.007232159354316536,
|
|
"entropy": 0.08850837219506502,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.784605333581567,
|
|
"epoch": 0.0026041666666666665,
|
|
"step": 100
|
|
},
|
|
{
|
|
"loss": 0.08135078847408295,
|
|
"grad_norm": 5.944860935211182,
|
|
"learning_rate": 3.4482758620689656e-07,
|
|
"num_tokens": 939171.0,
|
|
"completions/mean_length": 382.5,
|
|
"completions/min_length": 265.0,
|
|
"completions/max_length": 927.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 304.71429443359375,
|
|
"completions/min_terminated_length": 265.0,
|
|
"completions/max_terminated_length": 394.0,
|
|
"tools/call_frequency": 9.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005792862735688686,
|
|
"sampling/sampling_logp_difference/max": 0.6686649322509766,
|
|
"sampling/importance_sampling_ratio/min": 0.4069797694683075,
|
|
"sampling/importance_sampling_ratio/mean": 1.145476222038269,
|
|
"sampling/importance_sampling_ratio/max": 2.9042978286743164,
|
|
"kl": 0.007126070864615031,
|
|
"entropy": 0.08841658686287701,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.44231602922082,
|
|
"epoch": 0.0026302083333333334,
|
|
"step": 101
|
|
},
|
|
{
|
|
"loss": -0.02495439723134041,
|
|
"grad_norm": 2.7817928791046143,
|
|
"learning_rate": 3.4310344827586207e-07,
|
|
"num_tokens": 948151.0,
|
|
"completions/mean_length": 436.125,
|
|
"completions/min_length": 218.0,
|
|
"completions/max_length": 945.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 271.0,
|
|
"completions/min_terminated_length": 218.0,
|
|
"completions/max_terminated_length": 350.0,
|
|
"tools/call_frequency": 10.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0047197588719427586,
|
|
"sampling/sampling_logp_difference/max": 0.6686645746231079,
|
|
"sampling/importance_sampling_ratio/min": 0.18549923598766327,
|
|
"sampling/importance_sampling_ratio/mean": 0.665483832359314,
|
|
"sampling/importance_sampling_ratio/max": 1.5299997329711914,
|
|
"kl": 0.008183397418179084,
|
|
"entropy": 0.0706523151602596,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.575687261298299,
|
|
"epoch": 0.00265625,
|
|
"step": 102
|
|
},
|
|
{
|
|
"loss": -0.14288264513015747,
|
|
"grad_norm": 6.452617168426514,
|
|
"learning_rate": 3.413793103448276e-07,
|
|
"num_tokens": 955987.0,
|
|
"completions/mean_length": 293.0,
|
|
"completions/min_length": 259.0,
|
|
"completions/max_length": 319.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 293.0,
|
|
"completions/min_terminated_length": 259.0,
|
|
"completions/max_terminated_length": 319.0,
|
|
"tools/call_frequency": 7.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.011483734473586082,
|
|
"sampling/sampling_logp_difference/max": 0.7245371341705322,
|
|
"sampling/importance_sampling_ratio/min": 0.3339221775531769,
|
|
"sampling/importance_sampling_ratio/mean": 0.8164174556732178,
|
|
"sampling/importance_sampling_ratio/max": 2.5535342693328857,
|
|
"kl": 0.010253581189317629,
|
|
"entropy": 0.12678683595731854,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.359590545296669,
|
|
"epoch": 0.0026822916666666666,
|
|
"step": 103
|
|
},
|
|
{
|
|
"loss": -0.059778064489364624,
|
|
"grad_norm": 4.4524054527282715,
|
|
"learning_rate": 3.396551724137931e-07,
|
|
"num_tokens": 964481.0,
|
|
"completions/mean_length": 377.0,
|
|
"completions/min_length": 268.0,
|
|
"completions/max_length": 968.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 292.5714416503906,
|
|
"completions/min_terminated_length": 268.0,
|
|
"completions/max_terminated_length": 347.0,
|
|
"tools/call_frequency": 8.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.014999999664723873,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.014999999664723873,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005991534795612097,
|
|
"sampling/sampling_logp_difference/max": 0.5136802196502686,
|
|
"sampling/importance_sampling_ratio/min": 0.3697912395000458,
|
|
"sampling/importance_sampling_ratio/mean": 1.1754696369171143,
|
|
"sampling/importance_sampling_ratio/max": 2.991692543029785,
|
|
"kl": 0.01747839637027937,
|
|
"entropy": 0.09049951832275838,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.70468470454216,
|
|
"epoch": 0.0027083333333333334,
|
|
"step": 104
|
|
},
|
|
{
|
|
"loss": -0.004548046737909317,
|
|
"grad_norm": 7.1236138343811035,
|
|
"learning_rate": 3.379310344827586e-07,
|
|
"num_tokens": 972259.0,
|
|
"completions/mean_length": 286.5,
|
|
"completions/min_length": 252.0,
|
|
"completions/max_length": 313.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 286.5,
|
|
"completions/min_terminated_length": 252.0,
|
|
"completions/max_terminated_length": 313.0,
|
|
"tools/call_frequency": 7.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.008687643334269524,
|
|
"sampling/sampling_logp_difference/max": 0.9612791538238525,
|
|
"sampling/importance_sampling_ratio/min": 0.14305539429187775,
|
|
"sampling/importance_sampling_ratio/mean": 0.8711907863616943,
|
|
"sampling/importance_sampling_ratio/max": 2.4403645992279053,
|
|
"kl": 0.008136546683090273,
|
|
"entropy": 0.10206135828047991,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.820745028555393,
|
|
"epoch": 0.002734375,
|
|
"step": 105
|
|
},
|
|
{
|
|
"loss": 0.07031859457492828,
|
|
"grad_norm": 4.698828220367432,
|
|
"learning_rate": 3.362068965517241e-07,
|
|
"num_tokens": 980091.0,
|
|
"completions/mean_length": 293.0,
|
|
"completions/min_length": 258.0,
|
|
"completions/max_length": 380.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 293.0,
|
|
"completions/min_terminated_length": 258.0,
|
|
"completions/max_terminated_length": 380.0,
|
|
"tools/call_frequency": 7.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006427158135920763,
|
|
"sampling/sampling_logp_difference/max": 0.48835062980651855,
|
|
"sampling/importance_sampling_ratio/min": 0.33754444122314453,
|
|
"sampling/importance_sampling_ratio/mean": 0.8972164392471313,
|
|
"sampling/importance_sampling_ratio/max": 1.4706202745437622,
|
|
"kl": 0.009990831662435085,
|
|
"entropy": 0.07861929479986429,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.412420952692628,
|
|
"epoch": 0.0027604166666666667,
|
|
"step": 106
|
|
},
|
|
{
|
|
"loss": 0.46435877680778503,
|
|
"grad_norm": 5.831277847290039,
|
|
"learning_rate": 3.3448275862068966e-07,
|
|
"num_tokens": 988499.0,
|
|
"completions/mean_length": 365.5,
|
|
"completions/min_length": 14.0,
|
|
"completions/max_length": 1043.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 268.71429443359375,
|
|
"completions/min_terminated_length": 14.0,
|
|
"completions/max_terminated_length": 362.0,
|
|
"tools/call_frequency": 7.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005006421357393265,
|
|
"sampling/sampling_logp_difference/max": 0.8692312240600586,
|
|
"sampling/importance_sampling_ratio/min": 0.41967687010765076,
|
|
"sampling/importance_sampling_ratio/mean": 1.1174479722976685,
|
|
"sampling/importance_sampling_ratio/max": 1.9494823217391968,
|
|
"kl": 0.009601287310943007,
|
|
"entropy": 0.06292498670518398,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.616144863888621,
|
|
"epoch": 0.0027864583333333335,
|
|
"step": 107
|
|
},
|
|
{
|
|
"loss": 0.8400824666023254,
|
|
"grad_norm": 6.475160121917725,
|
|
"learning_rate": 3.3275862068965517e-07,
|
|
"num_tokens": 997526.0,
|
|
"completions/mean_length": 441.875,
|
|
"completions/min_length": 228.0,
|
|
"completions/max_length": 1052.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 375.2857360839844,
|
|
"completions/min_terminated_length": 228.0,
|
|
"completions/max_terminated_length": 1052.0,
|
|
"tools/call_frequency": 10.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.00624999962747097,
|
|
"rewards/reward_func/std": 0.024458419531583786,
|
|
"reward": 0.00624999962747097,
|
|
"reward_std": 0.024458419531583786,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004743676632642746,
|
|
"sampling/sampling_logp_difference/max": 0.49057912826538086,
|
|
"sampling/importance_sampling_ratio/min": 0.3813149333000183,
|
|
"sampling/importance_sampling_ratio/mean": 1.0575041770935059,
|
|
"sampling/importance_sampling_ratio/max": 2.4018681049346924,
|
|
"kl": 0.015303216827305732,
|
|
"entropy": 0.07703881501220167,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.497826302424073,
|
|
"epoch": 0.0028125,
|
|
"step": 108
|
|
},
|
|
{
|
|
"loss": 0.11117170751094818,
|
|
"grad_norm": 2.9853343963623047,
|
|
"learning_rate": 3.310344827586207e-07,
|
|
"num_tokens": 1007230.0,
|
|
"completions/mean_length": 527.25,
|
|
"completions/min_length": 258.0,
|
|
"completions/max_length": 957.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 400.66668701171875,
|
|
"completions/min_terminated_length": 258.0,
|
|
"completions/max_terminated_length": 957.0,
|
|
"tools/call_frequency": 13.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0028894334100186825,
|
|
"sampling/sampling_logp_difference/max": 0.476010799407959,
|
|
"sampling/importance_sampling_ratio/min": 0.19368234276771545,
|
|
"sampling/importance_sampling_ratio/mean": 0.8871850967407227,
|
|
"sampling/importance_sampling_ratio/max": 1.9029121398925781,
|
|
"kl": 0.007728074411716079,
|
|
"entropy": 0.05447922530584037,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.023359915241599,
|
|
"epoch": 0.0028385416666666667,
|
|
"step": 109
|
|
},
|
|
{
|
|
"loss": 0.18494461476802826,
|
|
"grad_norm": 3.6524336338043213,
|
|
"learning_rate": 3.293103448275862e-07,
|
|
"num_tokens": 1016286.0,
|
|
"completions/mean_length": 446.875,
|
|
"completions/min_length": 230.0,
|
|
"completions/max_length": 912.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 380.4285888671875,
|
|
"completions/min_terminated_length": 230.0,
|
|
"completions/max_terminated_length": 767.0,
|
|
"tools/call_frequency": 11.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0050048623234033585,
|
|
"sampling/sampling_logp_difference/max": 0.6685194969177246,
|
|
"sampling/importance_sampling_ratio/min": 0.20930853486061096,
|
|
"sampling/importance_sampling_ratio/mean": 0.9360355138778687,
|
|
"sampling/importance_sampling_ratio/max": 1.97806978225708,
|
|
"kl": 0.009379368471854832,
|
|
"entropy": 0.07528758549597114,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.349480360746384,
|
|
"epoch": 0.002864583333333333,
|
|
"step": 110
|
|
},
|
|
{
|
|
"loss": 0.12492945790290833,
|
|
"grad_norm": 2.254457712173462,
|
|
"learning_rate": 3.275862068965517e-07,
|
|
"num_tokens": 1027448.0,
|
|
"completions/mean_length": 708.75,
|
|
"completions/min_length": 270.0,
|
|
"completions/max_length": 1096.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 442.75,
|
|
"completions/min_terminated_length": 270.0,
|
|
"completions/max_terminated_length": 910.0,
|
|
"tools/call_frequency": 15.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": -0.003750000149011612,
|
|
"rewards/reward_func/std": 0.019955307245254517,
|
|
"reward": -0.003750000149011612,
|
|
"reward_std": 0.019955307245254517,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0034278081730008125,
|
|
"sampling/sampling_logp_difference/max": 0.7272565364837646,
|
|
"sampling/importance_sampling_ratio/min": 0.20217320322990417,
|
|
"sampling/importance_sampling_ratio/mean": 0.7357429265975952,
|
|
"sampling/importance_sampling_ratio/max": 1.39297616481781,
|
|
"kl": 0.00535351757571334,
|
|
"entropy": 0.060748465009965,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.653194352984428,
|
|
"epoch": 0.002890625,
|
|
"step": 111
|
|
},
|
|
{
|
|
"loss": 0.7069124579429626,
|
|
"grad_norm": 10.07013988494873,
|
|
"learning_rate": 3.258620689655172e-07,
|
|
"num_tokens": 1035705.0,
|
|
"completions/mean_length": 346.375,
|
|
"completions/min_length": 221.0,
|
|
"completions/max_length": 895.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 268.0,
|
|
"completions/min_terminated_length": 221.0,
|
|
"completions/max_terminated_length": 310.0,
|
|
"tools/call_frequency": 9.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.014999999664723873,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.014999999664723873,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007136066444218159,
|
|
"sampling/sampling_logp_difference/max": 0.587206244468689,
|
|
"sampling/importance_sampling_ratio/min": 0.5164411664009094,
|
|
"sampling/importance_sampling_ratio/mean": 1.238433599472046,
|
|
"sampling/importance_sampling_ratio/max": 2.1792774200439453,
|
|
"kl": 0.02077096902212361,
|
|
"entropy": 0.08839601022191346,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.868256894871593,
|
|
"epoch": 0.002916666666666667,
|
|
"step": 112
|
|
},
|
|
{
|
|
"loss": -0.23588499426841736,
|
|
"grad_norm": 11.169977188110352,
|
|
"learning_rate": 3.241379310344827e-07,
|
|
"num_tokens": 1043278.0,
|
|
"completions/mean_length": 260.875,
|
|
"completions/min_length": 44.0,
|
|
"completions/max_length": 388.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 260.875,
|
|
"completions/min_terminated_length": 44.0,
|
|
"completions/max_terminated_length": 388.0,
|
|
"tools/call_frequency": 7.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00828575063496828,
|
|
"sampling/sampling_logp_difference/max": 1.301053524017334,
|
|
"sampling/importance_sampling_ratio/min": 0.4418397545814514,
|
|
"sampling/importance_sampling_ratio/mean": 1.290097951889038,
|
|
"sampling/importance_sampling_ratio/max": 2.696202516555786,
|
|
"kl": 0.008347340093678213,
|
|
"entropy": 0.07820363715291023,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.294246409088373,
|
|
"epoch": 0.002942708333333333,
|
|
"step": 113
|
|
},
|
|
{
|
|
"loss": -0.1647491157054901,
|
|
"grad_norm": 9.056323051452637,
|
|
"learning_rate": 3.224137931034482e-07,
|
|
"num_tokens": 1050682.0,
|
|
"completions/mean_length": 239.75,
|
|
"completions/min_length": 14.0,
|
|
"completions/max_length": 382.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 239.75,
|
|
"completions/min_terminated_length": 14.0,
|
|
"completions/max_terminated_length": 382.0,
|
|
"tools/call_frequency": 6.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007003731559962034,
|
|
"sampling/sampling_logp_difference/max": 0.6249985694885254,
|
|
"sampling/importance_sampling_ratio/min": 0.21560606360435486,
|
|
"sampling/importance_sampling_ratio/mean": 0.8319985866546631,
|
|
"sampling/importance_sampling_ratio/max": 1.2620893716812134,
|
|
"kl": 0.013763066599494778,
|
|
"entropy": 0.08929688856005669,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.658404251560569,
|
|
"epoch": 0.00296875,
|
|
"step": 114
|
|
},
|
|
{
|
|
"loss": -0.28623083233833313,
|
|
"grad_norm": 9.122482299804688,
|
|
"learning_rate": 3.2068965517241373e-07,
|
|
"num_tokens": 1058590.0,
|
|
"completions/mean_length": 302.875,
|
|
"completions/min_length": 262.0,
|
|
"completions/max_length": 357.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 302.875,
|
|
"completions/min_terminated_length": 262.0,
|
|
"completions/max_terminated_length": 357.0,
|
|
"tools/call_frequency": 7.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.016249999403953552,
|
|
"rewards/reward_func/std": 0.01685018092393875,
|
|
"reward": 0.016249999403953552,
|
|
"reward_std": 0.01685018092393875,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.008012155070900917,
|
|
"sampling/sampling_logp_difference/max": 0.48839759826660156,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.8209759593009949,
|
|
"sampling/importance_sampling_ratio/max": 1.81101655960083,
|
|
"kl": 0.011069039232097566,
|
|
"entropy": 0.09217237355187535,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 7.51235125400126,
|
|
"epoch": 0.002994791666666667,
|
|
"step": 115
|
|
},
|
|
{
|
|
"loss": 0.21585524082183838,
|
|
"grad_norm": 5.234922885894775,
|
|
"learning_rate": 3.1896551724137934e-07,
|
|
"num_tokens": 1066885.0,
|
|
"completions/mean_length": 350.875,
|
|
"completions/min_length": 148.0,
|
|
"completions/max_length": 915.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 270.2857360839844,
|
|
"completions/min_terminated_length": 148.0,
|
|
"completions/max_terminated_length": 309.0,
|
|
"tools/call_frequency": 9.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0053081507794559,
|
|
"sampling/sampling_logp_difference/max": 0.5288757681846619,
|
|
"sampling/importance_sampling_ratio/min": 0.29975199699401855,
|
|
"sampling/importance_sampling_ratio/mean": 0.6770144701004028,
|
|
"sampling/importance_sampling_ratio/max": 1.2171082496643066,
|
|
"kl": 0.0078124696665327065,
|
|
"entropy": 0.07289047061931342,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.961262000724673,
|
|
"epoch": 0.0030208333333333333,
|
|
"step": 116
|
|
},
|
|
{
|
|
"loss": 0.1083274632692337,
|
|
"grad_norm": 5.7102837562561035,
|
|
"learning_rate": 3.1724137931034485e-07,
|
|
"num_tokens": 1075400.0,
|
|
"completions/mean_length": 379.375,
|
|
"completions/min_length": 247.0,
|
|
"completions/max_length": 870.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 309.2857360839844,
|
|
"completions/min_terminated_length": 247.0,
|
|
"completions/max_terminated_length": 345.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.01414213515818119,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.01414213515818119,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006049262825399637,
|
|
"sampling/sampling_logp_difference/max": 1.1611151695251465,
|
|
"sampling/importance_sampling_ratio/min": 0.17917907238006592,
|
|
"sampling/importance_sampling_ratio/mean": 0.9689471125602722,
|
|
"sampling/importance_sampling_ratio/max": 1.922648310661316,
|
|
"kl": 0.010483801223017508,
|
|
"entropy": 0.0830291856545955,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.169814247637987,
|
|
"epoch": 0.003046875,
|
|
"step": 117
|
|
},
|
|
{
|
|
"loss": 0.10439752787351608,
|
|
"grad_norm": 5.405170917510986,
|
|
"learning_rate": 3.1551724137931036e-07,
|
|
"num_tokens": 1084633.0,
|
|
"completions/mean_length": 468.0,
|
|
"completions/min_length": 257.0,
|
|
"completions/max_length": 921.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 411.2857360839844,
|
|
"completions/min_terminated_length": 257.0,
|
|
"completions/max_terminated_length": 921.0,
|
|
"tools/call_frequency": 11.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005020327400416136,
|
|
"sampling/sampling_logp_difference/max": 0.43385791778564453,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 1.0760118961334229,
|
|
"sampling/importance_sampling_ratio/max": 1.932090163230896,
|
|
"kl": 0.010577438220025215,
|
|
"entropy": 0.09225372283253819,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.673520376905799,
|
|
"epoch": 0.0030729166666666665,
|
|
"step": 118
|
|
},
|
|
{
|
|
"loss": 0.41595613956451416,
|
|
"grad_norm": 4.16091775894165,
|
|
"learning_rate": 3.1379310344827587e-07,
|
|
"num_tokens": 1093559.0,
|
|
"completions/mean_length": 430.0,
|
|
"completions/min_length": 226.0,
|
|
"completions/max_length": 931.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 367.4285888671875,
|
|
"completions/min_terminated_length": 226.0,
|
|
"completions/max_terminated_length": 931.0,
|
|
"tools/call_frequency": 11.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004744705278426409,
|
|
"sampling/sampling_logp_difference/max": 0.4831089973449707,
|
|
"sampling/importance_sampling_ratio/min": 0.5277978181838989,
|
|
"sampling/importance_sampling_ratio/mean": 1.1084668636322021,
|
|
"sampling/importance_sampling_ratio/max": 1.8976047039031982,
|
|
"kl": 0.015597150890243938,
|
|
"entropy": 0.06393339071655646,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.799420868977904,
|
|
"epoch": 0.0030989583333333333,
|
|
"step": 119
|
|
},
|
|
{
|
|
"loss": 0.2962050139904022,
|
|
"grad_norm": 2.9780967235565186,
|
|
"learning_rate": 3.120689655172414e-07,
|
|
"num_tokens": 1101923.0,
|
|
"completions/mean_length": 359.0,
|
|
"completions/min_length": 239.0,
|
|
"completions/max_length": 879.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 359.0,
|
|
"completions/min_terminated_length": 239.0,
|
|
"completions/max_terminated_length": 879.0,
|
|
"tools/call_frequency": 9.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005257544107735157,
|
|
"sampling/sampling_logp_difference/max": 0.6191132664680481,
|
|
"sampling/importance_sampling_ratio/min": 0.16893233358860016,
|
|
"sampling/importance_sampling_ratio/mean": 0.6383512020111084,
|
|
"sampling/importance_sampling_ratio/max": 1.0441057682037354,
|
|
"kl": 0.01099010584584903,
|
|
"entropy": 0.07132729887962341,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.635305689647794,
|
|
"epoch": 0.003125,
|
|
"step": 120
|
|
},
|
|
{
|
|
"loss": 0.2147320657968521,
|
|
"grad_norm": 4.304876804351807,
|
|
"learning_rate": 3.103448275862069e-07,
|
|
"num_tokens": 1110577.0,
|
|
"completions/mean_length": 395.625,
|
|
"completions/min_length": 278.0,
|
|
"completions/max_length": 890.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 395.625,
|
|
"completions/min_terminated_length": 278.0,
|
|
"completions/max_terminated_length": 890.0,
|
|
"tools/call_frequency": 9.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006748862098902464,
|
|
"sampling/sampling_logp_difference/max": 0.6192135810852051,
|
|
"sampling/importance_sampling_ratio/min": 0.2824941873550415,
|
|
"sampling/importance_sampling_ratio/mean": 0.7997482419013977,
|
|
"sampling/importance_sampling_ratio/max": 1.5805388689041138,
|
|
"kl": 0.011249663144553779,
|
|
"entropy": 0.09272133861668408,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.36530939117074,
|
|
"epoch": 0.0031510416666666666,
|
|
"step": 121
|
|
},
|
|
{
|
|
"loss": -0.00215265154838562,
|
|
"grad_norm": 6.253425121307373,
|
|
"learning_rate": 3.086206896551724e-07,
|
|
"num_tokens": 1118553.0,
|
|
"completions/mean_length": 310.875,
|
|
"completions/min_length": 221.0,
|
|
"completions/max_length": 422.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 310.875,
|
|
"completions/min_terminated_length": 221.0,
|
|
"completions/max_terminated_length": 422.0,
|
|
"tools/call_frequency": 7.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.014880476519465446,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.014880475588142872,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007874981500208378,
|
|
"sampling/sampling_logp_difference/max": 0.771700382232666,
|
|
"sampling/importance_sampling_ratio/min": 0.30450940132141113,
|
|
"sampling/importance_sampling_ratio/mean": 0.8644342422485352,
|
|
"sampling/importance_sampling_ratio/max": 1.2243115901947021,
|
|
"kl": 0.018789416033541784,
|
|
"entropy": 0.09023788291960955,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 9.312917228788137,
|
|
"epoch": 0.0031770833333333334,
|
|
"step": 122
|
|
},
|
|
{
|
|
"loss": -0.22130867838859558,
|
|
"grad_norm": 6.4872026443481445,
|
|
"learning_rate": 3.068965517241379e-07,
|
|
"num_tokens": 1126313.0,
|
|
"completions/mean_length": 283.625,
|
|
"completions/min_length": 234.0,
|
|
"completions/max_length": 345.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 283.625,
|
|
"completions/min_terminated_length": 234.0,
|
|
"completions/max_terminated_length": 345.0,
|
|
"tools/call_frequency": 7.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.014880476519465446,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.014880475588142872,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007742065470665693,
|
|
"sampling/sampling_logp_difference/max": 0.6588668823242188,
|
|
"sampling/importance_sampling_ratio/min": 0.3957456946372986,
|
|
"sampling/importance_sampling_ratio/mean": 1.0762112140655518,
|
|
"sampling/importance_sampling_ratio/max": 2.2421915531158447,
|
|
"kl": 0.017989573098020628,
|
|
"entropy": 0.08870360907167196,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.313219198957086,
|
|
"epoch": 0.003203125,
|
|
"step": 123
|
|
},
|
|
{
|
|
"loss": 0.48810526728630066,
|
|
"grad_norm": 2.247687816619873,
|
|
"learning_rate": 3.051724137931034e-07,
|
|
"num_tokens": 1135370.0,
|
|
"completions/mean_length": 446.375,
|
|
"completions/min_length": 230.0,
|
|
"completions/max_length": 949.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 375.2857360839844,
|
|
"completions/min_terminated_length": 230.0,
|
|
"completions/max_terminated_length": 949.0,
|
|
"tools/call_frequency": 11.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003948982339352369,
|
|
"sampling/sampling_logp_difference/max": 0.6286571025848389,
|
|
"sampling/importance_sampling_ratio/min": 0.12987120449543,
|
|
"sampling/importance_sampling_ratio/mean": 0.7765293121337891,
|
|
"sampling/importance_sampling_ratio/max": 2.045142650604248,
|
|
"kl": 0.010262770706503943,
|
|
"entropy": 0.0691474040504545,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.343333104625344,
|
|
"epoch": 0.0032291666666666666,
|
|
"step": 124
|
|
},
|
|
{
|
|
"loss": 0.4576762616634369,
|
|
"grad_norm": 2.9994709491729736,
|
|
"learning_rate": 3.0344827586206897e-07,
|
|
"num_tokens": 1143914.0,
|
|
"completions/mean_length": 382.75,
|
|
"completions/min_length": 269.0,
|
|
"completions/max_length": 883.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 382.75,
|
|
"completions/min_terminated_length": 269.0,
|
|
"completions/max_terminated_length": 883.0,
|
|
"tools/call_frequency": 9.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.016249999403953552,
|
|
"rewards/reward_func/std": 0.01685018092393875,
|
|
"reward": 0.016249999403953552,
|
|
"reward_std": 0.01685018092393875,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006361139938235283,
|
|
"sampling/sampling_logp_difference/max": 0.44611382484436035,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.778921902179718,
|
|
"sampling/importance_sampling_ratio/max": 1.2394492626190186,
|
|
"kl": 0.01161612311443605,
|
|
"entropy": 0.09289700607769191,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.000936675816774,
|
|
"epoch": 0.0032552083333333335,
|
|
"step": 125
|
|
},
|
|
{
|
|
"loss": 0.8930668234825134,
|
|
"grad_norm": 25.013158798217773,
|
|
"learning_rate": 3.017241379310345e-07,
|
|
"num_tokens": 1152933.0,
|
|
"completions/mean_length": 441.5,
|
|
"completions/min_length": 235.0,
|
|
"completions/max_length": 879.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 296.0,
|
|
"completions/min_terminated_length": 235.0,
|
|
"completions/max_terminated_length": 388.0,
|
|
"tools/call_frequency": 11.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005295595619827509,
|
|
"sampling/sampling_logp_difference/max": 1.4521937370300293,
|
|
"sampling/importance_sampling_ratio/min": 0.3781765401363373,
|
|
"sampling/importance_sampling_ratio/mean": 1.0722368955612183,
|
|
"sampling/importance_sampling_ratio/max": 2.3311986923217773,
|
|
"kl": 0.012714361888356507,
|
|
"entropy": 0.08642113872338086,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.945961307734251,
|
|
"epoch": 0.00328125,
|
|
"step": 126
|
|
},
|
|
{
|
|
"loss": 0.1633976548910141,
|
|
"grad_norm": 3.2046291828155518,
|
|
"learning_rate": 3e-07,
|
|
"num_tokens": 1162026.0,
|
|
"completions/mean_length": 451.125,
|
|
"completions/min_length": 229.0,
|
|
"completions/max_length": 890.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 392.0000305175781,
|
|
"completions/min_terminated_length": 229.0,
|
|
"completions/max_terminated_length": 890.0,
|
|
"tools/call_frequency": 12.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.01414213515818119,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.01414213515818119,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0038553159683942795,
|
|
"sampling/sampling_logp_difference/max": 0.831392765045166,
|
|
"sampling/importance_sampling_ratio/min": 0.24201825261116028,
|
|
"sampling/importance_sampling_ratio/mean": 0.8508594036102295,
|
|
"sampling/importance_sampling_ratio/max": 1.4487221240997314,
|
|
"kl": 0.011144677049742313,
|
|
"entropy": 0.06686653336510062,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.619546502828598,
|
|
"epoch": 0.0033072916666666667,
|
|
"step": 127
|
|
},
|
|
{
|
|
"loss": 0.19157162308692932,
|
|
"grad_norm": 2.2887234687805176,
|
|
"learning_rate": 2.982758620689655e-07,
|
|
"num_tokens": 1171264.0,
|
|
"completions/mean_length": 468.875,
|
|
"completions/min_length": 224.0,
|
|
"completions/max_length": 1112.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 268.16668701171875,
|
|
"completions/min_terminated_length": 224.0,
|
|
"completions/max_terminated_length": 303.0,
|
|
"tools/call_frequency": 9.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0037499996833503246,
|
|
"rewards/reward_func/std": 0.02386719174683094,
|
|
"reward": 0.0037499996833503246,
|
|
"reward_std": 0.02386719174683094,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005119456443935633,
|
|
"sampling/sampling_logp_difference/max": 0.6634714007377625,
|
|
"sampling/importance_sampling_ratio/min": 0.3455171585083008,
|
|
"sampling/importance_sampling_ratio/mean": 0.6479032039642334,
|
|
"sampling/importance_sampling_ratio/max": 1.010085940361023,
|
|
"kl": 0.013847738460754044,
|
|
"entropy": 0.08377158595249057,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.547344736754894,
|
|
"epoch": 0.0033333333333333335,
|
|
"step": 128
|
|
},
|
|
{
|
|
"loss": 0.16205757856369019,
|
|
"grad_norm": 2.6710877418518066,
|
|
"learning_rate": 2.96551724137931e-07,
|
|
"num_tokens": 1180518.0,
|
|
"completions/mean_length": 471.5,
|
|
"completions/min_length": 276.0,
|
|
"completions/max_length": 921.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 407.2857360839844,
|
|
"completions/min_terminated_length": 276.0,
|
|
"completions/max_terminated_length": 896.0,
|
|
"tools/call_frequency": 11.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005171550437808037,
|
|
"sampling/sampling_logp_difference/max": 0.6201157569885254,
|
|
"sampling/importance_sampling_ratio/min": 0.17714768648147583,
|
|
"sampling/importance_sampling_ratio/mean": 0.7121509909629822,
|
|
"sampling/importance_sampling_ratio/max": 1.6364480257034302,
|
|
"kl": 0.005411522004578728,
|
|
"entropy": 0.0826479981187731,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.734479563310742,
|
|
"epoch": 0.003359375,
|
|
"step": 129
|
|
},
|
|
{
|
|
"loss": -0.021296776831150055,
|
|
"grad_norm": 5.79312801361084,
|
|
"learning_rate": 2.948275862068965e-07,
|
|
"num_tokens": 1188183.0,
|
|
"completions/mean_length": 272.0,
|
|
"completions/min_length": 239.0,
|
|
"completions/max_length": 320.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 272.0,
|
|
"completions/min_terminated_length": 239.0,
|
|
"completions/max_terminated_length": 320.0,
|
|
"tools/call_frequency": 7.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006426252890378237,
|
|
"sampling/sampling_logp_difference/max": 0.4816563129425049,
|
|
"sampling/importance_sampling_ratio/min": 0.4499512314796448,
|
|
"sampling/importance_sampling_ratio/mean": 0.9402496218681335,
|
|
"sampling/importance_sampling_ratio/max": 1.3210963010787964,
|
|
"kl": 0.013666733517311513,
|
|
"entropy": 0.07675816770642996,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 7.068907061591744,
|
|
"epoch": 0.0033854166666666668,
|
|
"step": 130
|
|
},
|
|
{
|
|
"loss": 0.1784256398677826,
|
|
"grad_norm": 1.5753052234649658,
|
|
"learning_rate": 2.93103448275862e-07,
|
|
"num_tokens": 1197993.0,
|
|
"completions/mean_length": 539.875,
|
|
"completions/min_length": 242.0,
|
|
"completions/max_length": 955.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 411.5,
|
|
"completions/min_terminated_length": 242.0,
|
|
"completions/max_terminated_length": 955.0,
|
|
"tools/call_frequency": 13.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004044946748763323,
|
|
"sampling/sampling_logp_difference/max": 1.188810110092163,
|
|
"sampling/importance_sampling_ratio/min": 0.08638277649879456,
|
|
"sampling/importance_sampling_ratio/mean": 0.6029446125030518,
|
|
"sampling/importance_sampling_ratio/max": 1.113569736480713,
|
|
"kl": 0.008008234966837335,
|
|
"entropy": 0.0644446995574981,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.300636563450098,
|
|
"epoch": 0.003411458333333333,
|
|
"step": 131
|
|
},
|
|
{
|
|
"loss": 0.120160773396492,
|
|
"grad_norm": 3.3114044666290283,
|
|
"learning_rate": 2.913793103448276e-07,
|
|
"num_tokens": 1207029.0,
|
|
"completions/mean_length": 443.625,
|
|
"completions/min_length": 225.0,
|
|
"completions/max_length": 920.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 383.0000305175781,
|
|
"completions/min_terminated_length": 225.0,
|
|
"completions/max_terminated_length": 920.0,
|
|
"tools/call_frequency": 11.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.0,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0034094233997166157,
|
|
"sampling/sampling_logp_difference/max": 0.38815319538116455,
|
|
"sampling/importance_sampling_ratio/min": 0.20002828538417816,
|
|
"sampling/importance_sampling_ratio/mean": 1.1682875156402588,
|
|
"sampling/importance_sampling_ratio/max": 1.9055856466293335,
|
|
"kl": 0.0032579428207100136,
|
|
"entropy": 0.052077871980145574,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.740591928362846,
|
|
"epoch": 0.0034375,
|
|
"step": 132
|
|
},
|
|
{
|
|
"loss": 0.1864592432975769,
|
|
"grad_norm": 13.067723274230957,
|
|
"learning_rate": 2.896551724137931e-07,
|
|
"num_tokens": 1215866.0,
|
|
"completions/mean_length": 419.875,
|
|
"completions/min_length": 247.0,
|
|
"completions/max_length": 868.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 355.8571472167969,
|
|
"completions/min_terminated_length": 247.0,
|
|
"completions/max_terminated_length": 803.0,
|
|
"tools/call_frequency": 10.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.01414213515818119,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.01414213515818119,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003914377186447382,
|
|
"sampling/sampling_logp_difference/max": 0.5453448295593262,
|
|
"sampling/importance_sampling_ratio/min": 0.4023051857948303,
|
|
"sampling/importance_sampling_ratio/mean": 0.8168491125106812,
|
|
"sampling/importance_sampling_ratio/max": 1.3329473733901978,
|
|
"kl": 0.021923872642219067,
|
|
"entropy": 0.0557902313885279,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.70612084493041,
|
|
"epoch": 0.003463541666666667,
|
|
"step": 133
|
|
},
|
|
{
|
|
"loss": 0.7108664512634277,
|
|
"grad_norm": 3.699817657470703,
|
|
"learning_rate": 2.8793103448275865e-07,
|
|
"num_tokens": 1225634.0,
|
|
"completions/mean_length": 535.875,
|
|
"completions/min_length": 260.0,
|
|
"completions/max_length": 989.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 301.6000061035156,
|
|
"completions/min_terminated_length": 260.0,
|
|
"completions/max_terminated_length": 378.0,
|
|
"tools/call_frequency": 12.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005286019295454025,
|
|
"sampling/sampling_logp_difference/max": 1.2464321851730347,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.7495740652084351,
|
|
"sampling/importance_sampling_ratio/max": 2.121126890182495,
|
|
"kl": 0.012022783527299907,
|
|
"entropy": 0.0955480255652219,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.73185969144106,
|
|
"epoch": 0.0034895833333333333,
|
|
"step": 134
|
|
},
|
|
{
|
|
"loss": -0.12277907878160477,
|
|
"grad_norm": 1.638332486152649,
|
|
"learning_rate": 2.8620689655172416e-07,
|
|
"num_tokens": 1234919.0,
|
|
"completions/mean_length": 474.625,
|
|
"completions/min_length": 248.0,
|
|
"completions/max_length": 957.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 328.66668701171875,
|
|
"completions/min_terminated_length": 248.0,
|
|
"completions/max_terminated_length": 398.0,
|
|
"tools/call_frequency": 11.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": -0.003750000149011612,
|
|
"rewards/reward_func/std": 0.022638464346528053,
|
|
"reward": -0.003750000149011612,
|
|
"reward_std": 0.022638462483882904,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005199665669351816,
|
|
"sampling/sampling_logp_difference/max": 0.6200997233390808,
|
|
"sampling/importance_sampling_ratio/min": 0.22304587066173553,
|
|
"sampling/importance_sampling_ratio/mean": 0.687171459197998,
|
|
"sampling/importance_sampling_ratio/max": 1.0358991622924805,
|
|
"kl": 0.0067730720189729254,
|
|
"entropy": 0.07829280040459707,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.016188656911254,
|
|
"epoch": 0.003515625,
|
|
"step": 135
|
|
},
|
|
{
|
|
"loss": 0.45652079582214355,
|
|
"grad_norm": 2.1276769638061523,
|
|
"learning_rate": 2.8448275862068967e-07,
|
|
"num_tokens": 1245211.0,
|
|
"completions/mean_length": 601.125,
|
|
"completions/min_length": 246.0,
|
|
"completions/max_length": 931.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 430.20001220703125,
|
|
"completions/min_terminated_length": 246.0,
|
|
"completions/max_terminated_length": 931.0,
|
|
"tools/call_frequency": 15.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0034740508999675512,
|
|
"sampling/sampling_logp_difference/max": 0.4961376190185547,
|
|
"sampling/importance_sampling_ratio/min": 0.39931854605674744,
|
|
"sampling/importance_sampling_ratio/mean": 1.0868868827819824,
|
|
"sampling/importance_sampling_ratio/max": 1.9362167119979858,
|
|
"kl": 0.005297566302033374,
|
|
"entropy": 0.0645309763494879,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.862418776378036,
|
|
"epoch": 0.0035416666666666665,
|
|
"step": 136
|
|
},
|
|
{
|
|
"loss": 0.06947595626115799,
|
|
"grad_norm": 3.2402184009552,
|
|
"learning_rate": 2.827586206896552e-07,
|
|
"num_tokens": 1254580.0,
|
|
"completions/mean_length": 485.5,
|
|
"completions/min_length": 282.0,
|
|
"completions/max_length": 910.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 344.8333435058594,
|
|
"completions/min_terminated_length": 282.0,
|
|
"completions/max_terminated_length": 520.0,
|
|
"tools/call_frequency": 11.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005657813046127558,
|
|
"sampling/sampling_logp_difference/max": 0.48464536666870117,
|
|
"sampling/importance_sampling_ratio/min": 0.35901355743408203,
|
|
"sampling/importance_sampling_ratio/mean": 0.8260239362716675,
|
|
"sampling/importance_sampling_ratio/max": 1.3235818147659302,
|
|
"kl": 0.006206878111697733,
|
|
"entropy": 0.08789910678751767,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.070960972458124,
|
|
"epoch": 0.0035677083333333333,
|
|
"step": 137
|
|
},
|
|
{
|
|
"loss": 1.1847684383392334,
|
|
"grad_norm": 4.442853927612305,
|
|
"learning_rate": 2.810344827586207e-07,
|
|
"num_tokens": 1263624.0,
|
|
"completions/mean_length": 444.5,
|
|
"completions/min_length": 246.0,
|
|
"completions/max_length": 943.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 279.3333435058594,
|
|
"completions/min_terminated_length": 246.0,
|
|
"completions/max_terminated_length": 307.0,
|
|
"tools/call_frequency": 10.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.014999999664723873,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.014999999664723873,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004444813821464777,
|
|
"sampling/sampling_logp_difference/max": 0.5423450469970703,
|
|
"sampling/importance_sampling_ratio/min": 0.3824039399623871,
|
|
"sampling/importance_sampling_ratio/mean": 0.9495387077331543,
|
|
"sampling/importance_sampling_ratio/max": 1.7594578266143799,
|
|
"kl": 0.011284161275398219,
|
|
"entropy": 0.08384457719512284,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.585272112861276,
|
|
"epoch": 0.00359375,
|
|
"step": 138
|
|
},
|
|
{
|
|
"loss": 0.41621702909469604,
|
|
"grad_norm": 5.882610321044922,
|
|
"learning_rate": 2.793103448275862e-07,
|
|
"num_tokens": 1271627.0,
|
|
"completions/mean_length": 315.25,
|
|
"completions/min_length": 56.0,
|
|
"completions/max_length": 906.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 230.85714721679688,
|
|
"completions/min_terminated_length": 56.0,
|
|
"completions/max_terminated_length": 340.0,
|
|
"tools/call_frequency": 8.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005876841023564339,
|
|
"sampling/sampling_logp_difference/max": 0.6249297857284546,
|
|
"sampling/importance_sampling_ratio/min": 0.32682836055755615,
|
|
"sampling/importance_sampling_ratio/mean": 1.0201352834701538,
|
|
"sampling/importance_sampling_ratio/max": 2.958195924758911,
|
|
"kl": 0.010713011175539577,
|
|
"entropy": 0.07777869910933077,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.91103764437139,
|
|
"epoch": 0.0036197916666666666,
|
|
"step": 139
|
|
},
|
|
{
|
|
"loss": 0.12631958723068237,
|
|
"grad_norm": 1.6993509531021118,
|
|
"learning_rate": 2.775862068965517e-07,
|
|
"num_tokens": 1281594.0,
|
|
"completions/mean_length": 560.375,
|
|
"completions/min_length": 255.0,
|
|
"completions/max_length": 1015.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 330.20001220703125,
|
|
"completions/min_terminated_length": 255.0,
|
|
"completions/max_terminated_length": 428.0,
|
|
"tools/call_frequency": 12.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0039015121292322874,
|
|
"sampling/sampling_logp_difference/max": 0.4738839864730835,
|
|
"sampling/importance_sampling_ratio/min": 0.2868458032608032,
|
|
"sampling/importance_sampling_ratio/mean": 0.6188288331031799,
|
|
"sampling/importance_sampling_ratio/max": 1.1442465782165527,
|
|
"kl": 0.004594533118506661,
|
|
"entropy": 0.07008718093857169,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.056007090955973,
|
|
"epoch": 0.0036458333333333334,
|
|
"step": 140
|
|
},
|
|
{
|
|
"loss": 0.14388377964496613,
|
|
"grad_norm": 2.702698230743408,
|
|
"learning_rate": 2.758620689655172e-07,
|
|
"num_tokens": 1293040.0,
|
|
"completions/mean_length": 745.375,
|
|
"completions/min_length": 289.0,
|
|
"completions/max_length": 1028.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 529.75,
|
|
"completions/min_terminated_length": 289.0,
|
|
"completions/max_terminated_length": 965.0,
|
|
"tools/call_frequency": 15.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0028421031311154366,
|
|
"sampling/sampling_logp_difference/max": 1.0000008344650269,
|
|
"sampling/importance_sampling_ratio/min": 0.4013004004955292,
|
|
"sampling/importance_sampling_ratio/mean": 1.0830907821655273,
|
|
"sampling/importance_sampling_ratio/max": 2.23561954498291,
|
|
"kl": 0.0022632270301983226,
|
|
"entropy": 0.05287150805816054,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 14.416002927348018,
|
|
"epoch": 0.003671875,
|
|
"step": 141
|
|
},
|
|
{
|
|
"loss": -0.009116150438785553,
|
|
"grad_norm": 6.55360221862793,
|
|
"learning_rate": 2.741379310344827e-07,
|
|
"num_tokens": 1302147.0,
|
|
"completions/mean_length": 452.625,
|
|
"completions/min_length": 239.0,
|
|
"completions/max_length": 898.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 314.5,
|
|
"completions/min_terminated_length": 239.0,
|
|
"completions/max_terminated_length": 359.0,
|
|
"tools/call_frequency": 11.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.012817398644983768,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.012817397713661194,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0055464571341872215,
|
|
"sampling/sampling_logp_difference/max": 0.4591031074523926,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 1.057511329650879,
|
|
"sampling/importance_sampling_ratio/max": 2.5048530101776123,
|
|
"kl": 0.008584533527027816,
|
|
"entropy": 0.08057547151111066,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.956653794273734,
|
|
"epoch": 0.0036979166666666666,
|
|
"step": 142
|
|
},
|
|
{
|
|
"loss": -0.21052180230617523,
|
|
"grad_norm": 50.517608642578125,
|
|
"learning_rate": 2.724137931034483e-07,
|
|
"num_tokens": 1310055.0,
|
|
"completions/mean_length": 302.375,
|
|
"completions/min_length": 241.0,
|
|
"completions/max_length": 505.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 302.375,
|
|
"completions/min_terminated_length": 241.0,
|
|
"completions/max_terminated_length": 505.0,
|
|
"tools/call_frequency": 7.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.008566089905798435,
|
|
"sampling/sampling_logp_difference/max": 1.0783277750015259,
|
|
"sampling/importance_sampling_ratio/min": 0.20345060527324677,
|
|
"sampling/importance_sampling_ratio/mean": 0.6607866287231445,
|
|
"sampling/importance_sampling_ratio/max": 1.3724199533462524,
|
|
"kl": 0.0617900958750397,
|
|
"entropy": 0.07935992162674665,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 7.198723400011659,
|
|
"epoch": 0.0037239583333333335,
|
|
"step": 143
|
|
},
|
|
{
|
|
"loss": 0.19541701674461365,
|
|
"grad_norm": 2.0452730655670166,
|
|
"learning_rate": 2.706896551724138e-07,
|
|
"num_tokens": 1319640.0,
|
|
"completions/mean_length": 511.625,
|
|
"completions/min_length": 251.0,
|
|
"completions/max_length": 903.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 400.0,
|
|
"completions/min_terminated_length": 251.0,
|
|
"completions/max_terminated_length": 882.0,
|
|
"tools/call_frequency": 13.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004346841014921665,
|
|
"sampling/sampling_logp_difference/max": 0.7460143566131592,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.5977460741996765,
|
|
"sampling/importance_sampling_ratio/max": 1.1298850774765015,
|
|
"kl": 0.010761995104985544,
|
|
"entropy": 0.06976676208432764,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.253700632601976,
|
|
"epoch": 0.00375,
|
|
"step": 144
|
|
},
|
|
{
|
|
"loss": 0.15736129879951477,
|
|
"grad_norm": 2.452097177505493,
|
|
"learning_rate": 2.689655172413793e-07,
|
|
"num_tokens": 1328173.0,
|
|
"completions/mean_length": 380.625,
|
|
"completions/min_length": 228.0,
|
|
"completions/max_length": 943.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 380.625,
|
|
"completions/min_terminated_length": 228.0,
|
|
"completions/max_terminated_length": 943.0,
|
|
"tools/call_frequency": 8.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.016249999403953552,
|
|
"rewards/reward_func/std": 0.016850179061293602,
|
|
"reward": 0.016249999403953552,
|
|
"reward_std": 0.01685018092393875,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005344541743397713,
|
|
"sampling/sampling_logp_difference/max": 0.4617586135864258,
|
|
"sampling/importance_sampling_ratio/min": 0.48709753155708313,
|
|
"sampling/importance_sampling_ratio/mean": 0.7376953363418579,
|
|
"sampling/importance_sampling_ratio/max": 1.1122395992279053,
|
|
"kl": 0.00878529746842105,
|
|
"entropy": 0.08644933323375881,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.84443442709744,
|
|
"epoch": 0.0037760416666666667,
|
|
"step": 145
|
|
},
|
|
{
|
|
"loss": 0.355109840631485,
|
|
"grad_norm": 3.348142147064209,
|
|
"learning_rate": 2.672413793103448e-07,
|
|
"num_tokens": 1337910.0,
|
|
"completions/mean_length": 530.25,
|
|
"completions/min_length": 252.0,
|
|
"completions/max_length": 905.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 407.16668701171875,
|
|
"completions/min_terminated_length": 252.0,
|
|
"completions/max_terminated_length": 895.0,
|
|
"tools/call_frequency": 13.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0040186611004173756,
|
|
"sampling/sampling_logp_difference/max": 0.4408893585205078,
|
|
"sampling/importance_sampling_ratio/min": 0.3892063498497009,
|
|
"sampling/importance_sampling_ratio/mean": 1.077467918395996,
|
|
"sampling/importance_sampling_ratio/max": 2.069901943206787,
|
|
"kl": 0.007951671152113704,
|
|
"entropy": 0.0674425505567342,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.709803758189082,
|
|
"epoch": 0.0038020833333333335,
|
|
"step": 146
|
|
},
|
|
{
|
|
"loss": 0.26510369777679443,
|
|
"grad_norm": 3.6637167930603027,
|
|
"learning_rate": 2.655172413793103e-07,
|
|
"num_tokens": 1346995.0,
|
|
"completions/mean_length": 449.625,
|
|
"completions/min_length": 233.0,
|
|
"completions/max_length": 884.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 387.5714416503906,
|
|
"completions/min_terminated_length": 233.0,
|
|
"completions/max_terminated_length": 744.0,
|
|
"tools/call_frequency": 12.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005694272927939892,
|
|
"sampling/sampling_logp_difference/max": 0.7496256828308105,
|
|
"sampling/importance_sampling_ratio/min": 0.29766976833343506,
|
|
"sampling/importance_sampling_ratio/mean": 0.9870160222053528,
|
|
"sampling/importance_sampling_ratio/max": 1.642647624015808,
|
|
"kl": 0.010808795315824682,
|
|
"entropy": 0.07984890695661306,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.246438594534993,
|
|
"epoch": 0.003828125,
|
|
"step": 147
|
|
},
|
|
{
|
|
"loss": 0.28156739473342896,
|
|
"grad_norm": 5.059882164001465,
|
|
"learning_rate": 2.6379310344827587e-07,
|
|
"num_tokens": 1356490.0,
|
|
"completions/mean_length": 500.875,
|
|
"completions/min_length": 202.0,
|
|
"completions/max_length": 909.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 270.3999938964844,
|
|
"completions/min_terminated_length": 202.0,
|
|
"completions/max_terminated_length": 372.0,
|
|
"tools/call_frequency": 13.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.014880476519465446,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.014880475588142872,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003188743256032467,
|
|
"sampling/sampling_logp_difference/max": 0.44309818744659424,
|
|
"sampling/importance_sampling_ratio/min": 0.3629075288772583,
|
|
"sampling/importance_sampling_ratio/mean": 1.0272725820541382,
|
|
"sampling/importance_sampling_ratio/max": 2.483339548110962,
|
|
"kl": 0.013519551188437617,
|
|
"entropy": 0.06758475344395265,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.678413605317473,
|
|
"epoch": 0.0038541666666666668,
|
|
"step": 148
|
|
},
|
|
{
|
|
"loss": 0.16013890504837036,
|
|
"grad_norm": 2.207519054412842,
|
|
"learning_rate": 2.620689655172414e-07,
|
|
"num_tokens": 1364980.0,
|
|
"completions/mean_length": 375.125,
|
|
"completions/min_length": 232.0,
|
|
"completions/max_length": 885.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 302.2857360839844,
|
|
"completions/min_terminated_length": 232.0,
|
|
"completions/max_terminated_length": 342.0,
|
|
"tools/call_frequency": 9.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.013749999925494194,
|
|
"rewards/reward_func/std": 0.01767767034471035,
|
|
"reward": 0.013749999925494194,
|
|
"reward_std": 0.01767767034471035,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005511585623025894,
|
|
"sampling/sampling_logp_difference/max": 0.8766627311706543,
|
|
"sampling/importance_sampling_ratio/min": 0.33169519901275635,
|
|
"sampling/importance_sampling_ratio/mean": 0.6266351938247681,
|
|
"sampling/importance_sampling_ratio/max": 1.2635807991027832,
|
|
"kl": 0.008238287657150067,
|
|
"entropy": 0.0773795711575076,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.68444712460041,
|
|
"epoch": 0.003880208333333333,
|
|
"step": 149
|
|
},
|
|
{
|
|
"loss": 0.2061554491519928,
|
|
"grad_norm": 3.9576876163482666,
|
|
"learning_rate": 2.603448275862069e-07,
|
|
"num_tokens": 1373674.0,
|
|
"completions/mean_length": 400.875,
|
|
"completions/min_length": 236.0,
|
|
"completions/max_length": 944.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 323.2857360839844,
|
|
"completions/min_terminated_length": 236.0,
|
|
"completions/max_terminated_length": 366.0,
|
|
"tools/call_frequency": 9.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004895641468465328,
|
|
"sampling/sampling_logp_difference/max": 0.8447420597076416,
|
|
"sampling/importance_sampling_ratio/min": 0.4069957733154297,
|
|
"sampling/importance_sampling_ratio/mean": 0.9111777544021606,
|
|
"sampling/importance_sampling_ratio/max": 1.4833433628082275,
|
|
"kl": 0.011995875654974952,
|
|
"entropy": 0.0711302588460967,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.686782391741872,
|
|
"epoch": 0.00390625,
|
|
"step": 150
|
|
},
|
|
{
|
|
"loss": 0.29711028933525085,
|
|
"grad_norm": 2.27693510055542,
|
|
"learning_rate": 2.586206896551724e-07,
|
|
"num_tokens": 1383356.0,
|
|
"completions/mean_length": 524.375,
|
|
"completions/min_length": 229.0,
|
|
"completions/max_length": 944.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 397.16668701171875,
|
|
"completions/min_terminated_length": 229.0,
|
|
"completions/max_terminated_length": 916.0,
|
|
"tools/call_frequency": 13.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00325409765355289,
|
|
"sampling/sampling_logp_difference/max": 1.2148401737213135,
|
|
"sampling/importance_sampling_ratio/min": 0.19665421545505524,
|
|
"sampling/importance_sampling_ratio/mean": 0.9977468252182007,
|
|
"sampling/importance_sampling_ratio/max": 2.113764524459839,
|
|
"kl": 0.005079533737443853,
|
|
"entropy": 0.058041157375555485,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.013143414631486,
|
|
"epoch": 0.003932291666666666,
|
|
"step": 151
|
|
},
|
|
{
|
|
"loss": 0.15864849090576172,
|
|
"grad_norm": 3.052394390106201,
|
|
"learning_rate": 2.5689655172413796e-07,
|
|
"num_tokens": 1392458.0,
|
|
"completions/mean_length": 450.625,
|
|
"completions/min_length": 234.0,
|
|
"completions/max_length": 972.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 450.625,
|
|
"completions/min_terminated_length": 234.0,
|
|
"completions/max_terminated_length": 972.0,
|
|
"tools/call_frequency": 11.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.01414213515818119,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.01414213515818119,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003607141086831689,
|
|
"sampling/sampling_logp_difference/max": 0.596961259841919,
|
|
"sampling/importance_sampling_ratio/min": 0.5029790997505188,
|
|
"sampling/importance_sampling_ratio/mean": 1.0688074827194214,
|
|
"sampling/importance_sampling_ratio/max": 1.752822995185852,
|
|
"kl": 0.007834671720047481,
|
|
"entropy": 0.060456611681729555,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.8379608951509,
|
|
"epoch": 0.003958333333333334,
|
|
"step": 152
|
|
},
|
|
{
|
|
"loss": 0.784598171710968,
|
|
"grad_norm": 2.6799111366271973,
|
|
"learning_rate": 2.5517241379310346e-07,
|
|
"num_tokens": 1402010.0,
|
|
"completions/mean_length": 507.875,
|
|
"completions/min_length": 197.0,
|
|
"completions/max_length": 899.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 456.857177734375,
|
|
"completions/min_terminated_length": 197.0,
|
|
"completions/max_terminated_length": 899.0,
|
|
"tools/call_frequency": 13.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003934091422706842,
|
|
"sampling/sampling_logp_difference/max": 1.036914348602295,
|
|
"sampling/importance_sampling_ratio/min": 0.27182358503341675,
|
|
"sampling/importance_sampling_ratio/mean": 0.7754835486412048,
|
|
"sampling/importance_sampling_ratio/max": 1.4751989841461182,
|
|
"kl": 0.01072395659139147,
|
|
"entropy": 0.07619885372696444,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.678724393248558,
|
|
"epoch": 0.003984375,
|
|
"step": 153
|
|
},
|
|
{
|
|
"loss": 0.23067213594913483,
|
|
"grad_norm": 4.334474563598633,
|
|
"learning_rate": 2.5344827586206897e-07,
|
|
"num_tokens": 1412343.0,
|
|
"completions/mean_length": 606.0,
|
|
"completions/min_length": 221.0,
|
|
"completions/max_length": 986.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 561.7142944335938,
|
|
"completions/min_terminated_length": 221.0,
|
|
"completions/max_terminated_length": 986.0,
|
|
"tools/call_frequency": 14.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0024640304036438465,
|
|
"sampling/sampling_logp_difference/max": 0.6983184814453125,
|
|
"sampling/importance_sampling_ratio/min": 0.3580022156238556,
|
|
"sampling/importance_sampling_ratio/mean": 0.9473062753677368,
|
|
"sampling/importance_sampling_ratio/max": 2.753143072128296,
|
|
"kl": 0.0053151231822994305,
|
|
"entropy": 0.05121419607894495,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 14.257784998044372,
|
|
"epoch": 0.0040104166666666665,
|
|
"step": 154
|
|
},
|
|
{
|
|
"loss": 0.38859823346138,
|
|
"grad_norm": 2.8113319873809814,
|
|
"learning_rate": 2.517241379310345e-07,
|
|
"num_tokens": 1422276.0,
|
|
"completions/mean_length": 555.625,
|
|
"completions/min_length": 14.0,
|
|
"completions/max_length": 900.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 354.0,
|
|
"completions/min_terminated_length": 14.0,
|
|
"completions/max_terminated_length": 888.0,
|
|
"tools/call_frequency": 14.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003091561608016491,
|
|
"sampling/sampling_logp_difference/max": 0.6240787506103516,
|
|
"sampling/importance_sampling_ratio/min": 0.5358552932739258,
|
|
"sampling/importance_sampling_ratio/mean": 0.9822759628295898,
|
|
"sampling/importance_sampling_ratio/max": 1.294351577758789,
|
|
"kl": 0.00937311277812114,
|
|
"entropy": 0.04510764358565211,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.778962310403585,
|
|
"epoch": 0.004036458333333334,
|
|
"step": 155
|
|
},
|
|
{
|
|
"loss": 0.07130838185548782,
|
|
"grad_norm": 3.3684043884277344,
|
|
"learning_rate": 2.5e-07,
|
|
"num_tokens": 1430738.0,
|
|
"completions/mean_length": 371.5,
|
|
"completions/min_length": 222.0,
|
|
"completions/max_length": 906.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 295.14288330078125,
|
|
"completions/min_terminated_length": 222.0,
|
|
"completions/max_terminated_length": 376.0,
|
|
"tools/call_frequency": 9.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0048439837992191315,
|
|
"sampling/sampling_logp_difference/max": 0.4381091594696045,
|
|
"sampling/importance_sampling_ratio/min": 0.2527463138103485,
|
|
"sampling/importance_sampling_ratio/mean": 0.8320872187614441,
|
|
"sampling/importance_sampling_ratio/max": 1.5195680856704712,
|
|
"kl": 0.007922279299236834,
|
|
"entropy": 0.07430013502016664,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.923164380714297,
|
|
"epoch": 0.0040625,
|
|
"step": 156
|
|
},
|
|
{
|
|
"loss": 0.3246328830718994,
|
|
"grad_norm": 2.4954326152801514,
|
|
"learning_rate": 2.482758620689655e-07,
|
|
"num_tokens": 1441812.0,
|
|
"completions/mean_length": 698.125,
|
|
"completions/min_length": 267.0,
|
|
"completions/max_length": 998.0,
|
|
"completions/clipped_ratio": 0.625,
|
|
"completions/mean_terminated_length": 301.3333435058594,
|
|
"completions/min_terminated_length": 267.0,
|
|
"completions/max_terminated_length": 327.0,
|
|
"tools/call_frequency": 16.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0026307774242013693,
|
|
"sampling/sampling_logp_difference/max": 0.581841230392456,
|
|
"sampling/importance_sampling_ratio/min": 0.317604124546051,
|
|
"sampling/importance_sampling_ratio/mean": 1.14352548122406,
|
|
"sampling/importance_sampling_ratio/max": 2.0271735191345215,
|
|
"kl": 0.0052480718040897045,
|
|
"entropy": 0.03790102247148752,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.063249157741666,
|
|
"epoch": 0.0040885416666666665,
|
|
"step": 157
|
|
},
|
|
{
|
|
"loss": -0.02520151436328888,
|
|
"grad_norm": 7.141921520233154,
|
|
"learning_rate": 2.46551724137931e-07,
|
|
"num_tokens": 1452129.0,
|
|
"completions/mean_length": 603.625,
|
|
"completions/min_length": 248.0,
|
|
"completions/max_length": 969.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 415.20001220703125,
|
|
"completions/min_terminated_length": 248.0,
|
|
"completions/max_terminated_length": 921.0,
|
|
"tools/call_frequency": 15.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003178365994244814,
|
|
"sampling/sampling_logp_difference/max": 0.8997042179107666,
|
|
"sampling/importance_sampling_ratio/min": 0.30386480689048767,
|
|
"sampling/importance_sampling_ratio/mean": 0.9523824453353882,
|
|
"sampling/importance_sampling_ratio/max": 2.900059938430786,
|
|
"kl": 0.008052273029534263,
|
|
"entropy": 0.05091056425590068,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.20818499661982,
|
|
"epoch": 0.004114583333333333,
|
|
"step": 158
|
|
},
|
|
{
|
|
"loss": 0.45495691895484924,
|
|
"grad_norm": 3.886829137802124,
|
|
"learning_rate": 2.448275862068965e-07,
|
|
"num_tokens": 1461129.0,
|
|
"completions/mean_length": 438.5,
|
|
"completions/min_length": 248.0,
|
|
"completions/max_length": 941.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 373.14288330078125,
|
|
"completions/min_terminated_length": 248.0,
|
|
"completions/max_terminated_length": 941.0,
|
|
"tools/call_frequency": 10.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.014999999664723873,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.014999999664723873,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004588077310472727,
|
|
"sampling/sampling_logp_difference/max": 0.4519333839416504,
|
|
"sampling/importance_sampling_ratio/min": 0.43872809410095215,
|
|
"sampling/importance_sampling_ratio/mean": 1.0474553108215332,
|
|
"sampling/importance_sampling_ratio/max": 1.782901644706726,
|
|
"kl": 0.015291483468899969,
|
|
"entropy": 0.07987244543619454,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.494998019188643,
|
|
"epoch": 0.004140625,
|
|
"step": 159
|
|
},
|
|
{
|
|
"loss": 0.19334207475185394,
|
|
"grad_norm": 2.772022247314453,
|
|
"learning_rate": 2.43103448275862e-07,
|
|
"num_tokens": 1469492.0,
|
|
"completions/mean_length": 360.0,
|
|
"completions/min_length": 124.0,
|
|
"completions/max_length": 899.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 283.0,
|
|
"completions/min_terminated_length": 124.0,
|
|
"completions/max_terminated_length": 396.0,
|
|
"tools/call_frequency": 8.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.015118579380214214,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.015118577517569065,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006366310175508261,
|
|
"sampling/sampling_logp_difference/max": 0.6010000705718994,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.5896856784820557,
|
|
"sampling/importance_sampling_ratio/max": 1.035461187362671,
|
|
"kl": 0.008948636830609757,
|
|
"entropy": 0.08990807412192225,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.304909136146307,
|
|
"epoch": 0.004166666666666667,
|
|
"step": 160
|
|
},
|
|
{
|
|
"loss": 0.5132325887680054,
|
|
"grad_norm": 2.1742565631866455,
|
|
"learning_rate": 2.413793103448276e-07,
|
|
"num_tokens": 1480170.0,
|
|
"completions/mean_length": 649.625,
|
|
"completions/min_length": 207.0,
|
|
"completions/max_length": 895.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 413.0,
|
|
"completions/min_terminated_length": 207.0,
|
|
"completions/max_terminated_length": 883.0,
|
|
"tools/call_frequency": 17.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.002703455276787281,
|
|
"sampling/sampling_logp_difference/max": 0.569021463394165,
|
|
"sampling/importance_sampling_ratio/min": 0.3237609565258026,
|
|
"sampling/importance_sampling_ratio/mean": 0.9330805540084839,
|
|
"sampling/importance_sampling_ratio/max": 1.4763319492340088,
|
|
"kl": 0.007108663057806552,
|
|
"entropy": 0.053262577566783875,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.20669529773295,
|
|
"epoch": 0.004192708333333333,
|
|
"step": 161
|
|
},
|
|
{
|
|
"loss": 0.532462477684021,
|
|
"grad_norm": 8.624494552612305,
|
|
"learning_rate": 2.396551724137931e-07,
|
|
"num_tokens": 1488836.0,
|
|
"completions/mean_length": 397.875,
|
|
"completions/min_length": 222.0,
|
|
"completions/max_length": 933.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 321.4285888671875,
|
|
"completions/min_terminated_length": 222.0,
|
|
"completions/max_terminated_length": 473.0,
|
|
"tools/call_frequency": 9.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.001249999739229679,
|
|
"rewards/reward_func/std": 0.022951815277338028,
|
|
"reward": 0.001249999739229679,
|
|
"reward_std": 0.02295181341469288,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005203216802328825,
|
|
"sampling/sampling_logp_difference/max": 0.39698219299316406,
|
|
"sampling/importance_sampling_ratio/min": 0.45961838960647583,
|
|
"sampling/importance_sampling_ratio/mean": 1.2090280055999756,
|
|
"sampling/importance_sampling_ratio/max": 2.1715245246887207,
|
|
"kl": 0.011136812645418104,
|
|
"entropy": 0.07759461458772421,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.812960557639599,
|
|
"epoch": 0.00421875,
|
|
"step": 162
|
|
},
|
|
{
|
|
"loss": 0.21938498318195343,
|
|
"grad_norm": 2.7539706230163574,
|
|
"learning_rate": 2.3793103448275863e-07,
|
|
"num_tokens": 1498374.0,
|
|
"completions/mean_length": 507.375,
|
|
"completions/min_length": 225.0,
|
|
"completions/max_length": 959.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 372.5,
|
|
"completions/min_terminated_length": 225.0,
|
|
"completions/max_terminated_length": 914.0,
|
|
"tools/call_frequency": 12.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.016249999403953552,
|
|
"rewards/reward_func/std": 0.01685018092393875,
|
|
"reward": 0.016249999403953552,
|
|
"reward_std": 0.01685018092393875,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0036989289801567793,
|
|
"sampling/sampling_logp_difference/max": 1.2334673404693604,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.80265212059021,
|
|
"sampling/importance_sampling_ratio/max": 1.7188775539398193,
|
|
"kl": 0.009603490238077939,
|
|
"entropy": 0.05996394716203213,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.873560272157192,
|
|
"epoch": 0.004244791666666667,
|
|
"step": 163
|
|
},
|
|
{
|
|
"loss": 0.2591663897037506,
|
|
"grad_norm": 2.1876959800720215,
|
|
"learning_rate": 2.3620689655172413e-07,
|
|
"num_tokens": 1509289.0,
|
|
"completions/mean_length": 678.125,
|
|
"completions/min_length": 237.0,
|
|
"completions/max_length": 957.0,
|
|
"completions/clipped_ratio": 0.625,
|
|
"completions/mean_terminated_length": 290.3333435058594,
|
|
"completions/min_terminated_length": 237.0,
|
|
"completions/max_terminated_length": 376.0,
|
|
"tools/call_frequency": 16.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.002327412599697709,
|
|
"sampling/sampling_logp_difference/max": 0.479816198348999,
|
|
"sampling/importance_sampling_ratio/min": 0.3351881206035614,
|
|
"sampling/importance_sampling_ratio/mean": 0.8674755096435547,
|
|
"sampling/importance_sampling_ratio/max": 1.4120796918869019,
|
|
"kl": 0.007493237928429153,
|
|
"entropy": 0.04365518200211227,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.376963997259736,
|
|
"epoch": 0.004270833333333333,
|
|
"step": 164
|
|
},
|
|
{
|
|
"loss": 0.2494782656431198,
|
|
"grad_norm": 2.9401369094848633,
|
|
"learning_rate": 2.3448275862068964e-07,
|
|
"num_tokens": 1518329.0,
|
|
"completions/mean_length": 444.25,
|
|
"completions/min_length": 242.0,
|
|
"completions/max_length": 980.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 281.16668701171875,
|
|
"completions/min_terminated_length": 242.0,
|
|
"completions/max_terminated_length": 317.0,
|
|
"tools/call_frequency": 10.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.016249999403953552,
|
|
"rewards/reward_func/std": 0.01685018092393875,
|
|
"reward": 0.016249999403953552,
|
|
"reward_std": 0.01685018092393875,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0047093648463487625,
|
|
"sampling/sampling_logp_difference/max": 0.4828266501426697,
|
|
"sampling/importance_sampling_ratio/min": 0.4142991900444031,
|
|
"sampling/importance_sampling_ratio/mean": 0.6767470240592957,
|
|
"sampling/importance_sampling_ratio/max": 1.0061410665512085,
|
|
"kl": 0.013051618880126625,
|
|
"entropy": 0.07723805052228272,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.33225380629301,
|
|
"epoch": 0.004296875,
|
|
"step": 165
|
|
},
|
|
{
|
|
"loss": 0.5450688004493713,
|
|
"grad_norm": 4.9860124588012695,
|
|
"learning_rate": 2.3275862068965515e-07,
|
|
"num_tokens": 1526848.0,
|
|
"completions/mean_length": 379.125,
|
|
"completions/min_length": 244.0,
|
|
"completions/max_length": 891.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 306.0,
|
|
"completions/min_terminated_length": 244.0,
|
|
"completions/max_terminated_length": 401.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005283196922391653,
|
|
"sampling/sampling_logp_difference/max": 0.4645264148712158,
|
|
"sampling/importance_sampling_ratio/min": 0.582148551940918,
|
|
"sampling/importance_sampling_ratio/mean": 0.9555788040161133,
|
|
"sampling/importance_sampling_ratio/max": 1.731604814529419,
|
|
"kl": 0.011443080889876,
|
|
"entropy": 0.07826740574091673,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.328985534608364,
|
|
"epoch": 0.004322916666666667,
|
|
"step": 166
|
|
},
|
|
{
|
|
"loss": 0.3184610903263092,
|
|
"grad_norm": 2.1450023651123047,
|
|
"learning_rate": 2.3103448275862066e-07,
|
|
"num_tokens": 1536070.0,
|
|
"completions/mean_length": 466.75,
|
|
"completions/min_length": 258.0,
|
|
"completions/max_length": 928.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 316.66668701171875,
|
|
"completions/min_terminated_length": 258.0,
|
|
"completions/max_terminated_length": 344.0,
|
|
"tools/call_frequency": 10.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004849065560847521,
|
|
"sampling/sampling_logp_difference/max": 0.6581072807312012,
|
|
"sampling/importance_sampling_ratio/min": 0.22943541407585144,
|
|
"sampling/importance_sampling_ratio/mean": 0.6122701168060303,
|
|
"sampling/importance_sampling_ratio/max": 1.2230219841003418,
|
|
"kl": 0.007384931959677488,
|
|
"entropy": 0.07425842387601733,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.92301688529551,
|
|
"epoch": 0.004348958333333333,
|
|
"step": 167
|
|
},
|
|
{
|
|
"loss": 0.6113068461418152,
|
|
"grad_norm": 6.885270595550537,
|
|
"learning_rate": 2.293103448275862e-07,
|
|
"num_tokens": 1545282.0,
|
|
"completions/mean_length": 465.375,
|
|
"completions/min_length": 229.0,
|
|
"completions/max_length": 1023.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 296.5,
|
|
"completions/min_terminated_length": 229.0,
|
|
"completions/max_terminated_length": 377.0,
|
|
"tools/call_frequency": 10.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005654132459312677,
|
|
"sampling/sampling_logp_difference/max": 0.9898862838745117,
|
|
"sampling/importance_sampling_ratio/min": 0.06247289851307869,
|
|
"sampling/importance_sampling_ratio/mean": 1.055111050605774,
|
|
"sampling/importance_sampling_ratio/max": 2.72688364982605,
|
|
"kl": 0.012130821662140079,
|
|
"entropy": 0.0809087383095175,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.694183932617307,
|
|
"epoch": 0.004375,
|
|
"step": 168
|
|
},
|
|
{
|
|
"loss": 0.1832573264837265,
|
|
"grad_norm": 3.73466157913208,
|
|
"learning_rate": 2.2758620689655173e-07,
|
|
"num_tokens": 1554344.0,
|
|
"completions/mean_length": 447.125,
|
|
"completions/min_length": 263.0,
|
|
"completions/max_length": 908.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 447.125,
|
|
"completions/min_terminated_length": 263.0,
|
|
"completions/max_terminated_length": 908.0,
|
|
"tools/call_frequency": 11.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005070838611572981,
|
|
"sampling/sampling_logp_difference/max": 0.5936617851257324,
|
|
"sampling/importance_sampling_ratio/min": 0.588007926940918,
|
|
"sampling/importance_sampling_ratio/mean": 0.9101407527923584,
|
|
"sampling/importance_sampling_ratio/max": 1.5110055208206177,
|
|
"kl": 0.011189091823325725,
|
|
"entropy": 0.0827675883192569,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.566717436537147,
|
|
"epoch": 0.004401041666666667,
|
|
"step": 169
|
|
},
|
|
{
|
|
"loss": 0.10789915919303894,
|
|
"grad_norm": 3.300847053527832,
|
|
"learning_rate": 2.2586206896551724e-07,
|
|
"num_tokens": 1563371.0,
|
|
"completions/mean_length": 441.5,
|
|
"completions/min_length": 218.0,
|
|
"completions/max_length": 964.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 281.5,
|
|
"completions/min_terminated_length": 218.0,
|
|
"completions/max_terminated_length": 405.0,
|
|
"tools/call_frequency": 11.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.01414213515818119,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.01414213515818119,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005248270463198423,
|
|
"sampling/sampling_logp_difference/max": 0.6628669500350952,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 1.0210857391357422,
|
|
"sampling/importance_sampling_ratio/max": 2.1608545780181885,
|
|
"kl": 0.011338829084706958,
|
|
"entropy": 0.06719678780063987,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.913546672090888,
|
|
"epoch": 0.004427083333333333,
|
|
"step": 170
|
|
},
|
|
{
|
|
"loss": -0.15308786928653717,
|
|
"grad_norm": 9.328978538513184,
|
|
"learning_rate": 2.2413793103448274e-07,
|
|
"num_tokens": 1574477.0,
|
|
"completions/mean_length": 701.125,
|
|
"completions/min_length": 266.0,
|
|
"completions/max_length": 1009.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 458.25,
|
|
"completions/min_terminated_length": 266.0,
|
|
"completions/max_terminated_length": 965.0,
|
|
"tools/call_frequency": 15.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.11437499523162842,
|
|
"rewards/reward_func/std": 0.30747169256210327,
|
|
"reward": 0.11437499523162842,
|
|
"reward_std": 0.30747169256210327,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003461232176050544,
|
|
"sampling/sampling_logp_difference/max": 1.6048133373260498,
|
|
"sampling/importance_sampling_ratio/min": 0.019697746261954308,
|
|
"sampling/importance_sampling_ratio/mean": 1.164924144744873,
|
|
"sampling/importance_sampling_ratio/max": 2.911372423171997,
|
|
"kl": 0.00678792212420376,
|
|
"entropy": 0.049078432901296765,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 14.236025992780924,
|
|
"epoch": 0.004453125,
|
|
"step": 171
|
|
},
|
|
{
|
|
"loss": 0.06699332594871521,
|
|
"grad_norm": 3.498368978500366,
|
|
"learning_rate": 2.2241379310344828e-07,
|
|
"num_tokens": 1583088.0,
|
|
"completions/mean_length": 391.0,
|
|
"completions/min_length": 237.0,
|
|
"completions/max_length": 883.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 320.71429443359375,
|
|
"completions/min_terminated_length": 237.0,
|
|
"completions/max_terminated_length": 429.0,
|
|
"tools/call_frequency": 9.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.014999999664723873,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.014999999664723873,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007653472945094109,
|
|
"sampling/sampling_logp_difference/max": 0.7320506572723389,
|
|
"sampling/importance_sampling_ratio/min": 0.5328601598739624,
|
|
"sampling/importance_sampling_ratio/mean": 0.8372435569763184,
|
|
"sampling/importance_sampling_ratio/max": 1.0698771476745605,
|
|
"kl": 0.013877270212105941,
|
|
"entropy": 0.10430800006724894,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.402664301916957,
|
|
"epoch": 0.004479166666666667,
|
|
"step": 172
|
|
},
|
|
{
|
|
"loss": 0.14728188514709473,
|
|
"grad_norm": 2.8850626945495605,
|
|
"learning_rate": 2.206896551724138e-07,
|
|
"num_tokens": 1593074.0,
|
|
"completions/mean_length": 562.25,
|
|
"completions/min_length": 280.0,
|
|
"completions/max_length": 982.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 333.20001220703125,
|
|
"completions/min_terminated_length": 280.0,
|
|
"completions/max_terminated_length": 415.0,
|
|
"tools/call_frequency": 12.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.012817398644983768,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.012817397713661194,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003247647313401103,
|
|
"sampling/sampling_logp_difference/max": 0.7332391738891602,
|
|
"sampling/importance_sampling_ratio/min": 0.4890999495983124,
|
|
"sampling/importance_sampling_ratio/mean": 1.0704684257507324,
|
|
"sampling/importance_sampling_ratio/max": 1.5261626243591309,
|
|
"kl": 0.00994222271401668,
|
|
"entropy": 0.0633869837038219,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.608450036495924,
|
|
"epoch": 0.004505208333333333,
|
|
"step": 173
|
|
},
|
|
{
|
|
"loss": -0.09494911134243011,
|
|
"grad_norm": 4.086657524108887,
|
|
"learning_rate": 2.189655172413793e-07,
|
|
"num_tokens": 1601517.0,
|
|
"completions/mean_length": 370.0,
|
|
"completions/min_length": 249.0,
|
|
"completions/max_length": 891.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 295.5714416503906,
|
|
"completions/min_terminated_length": 249.0,
|
|
"completions/max_terminated_length": 426.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005537630990147591,
|
|
"sampling/sampling_logp_difference/max": 0.8397269248962402,
|
|
"sampling/importance_sampling_ratio/min": 0.2925705909729004,
|
|
"sampling/importance_sampling_ratio/mean": 0.8812037706375122,
|
|
"sampling/importance_sampling_ratio/max": 2.358135938644409,
|
|
"kl": 0.017766535296686925,
|
|
"entropy": 0.08754522632807493,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.887404596433043,
|
|
"epoch": 0.00453125,
|
|
"step": 174
|
|
},
|
|
{
|
|
"loss": 0.16594122350215912,
|
|
"grad_norm": 3.63606333732605,
|
|
"learning_rate": 2.172413793103448e-07,
|
|
"num_tokens": 1609916.0,
|
|
"completions/mean_length": 364.125,
|
|
"completions/min_length": 240.0,
|
|
"completions/max_length": 884.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 364.125,
|
|
"completions/min_terminated_length": 240.0,
|
|
"completions/max_terminated_length": 884.0,
|
|
"tools/call_frequency": 9.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006493411026895046,
|
|
"sampling/sampling_logp_difference/max": 0.7087509632110596,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.7516402006149292,
|
|
"sampling/importance_sampling_ratio/max": 1.6000678539276123,
|
|
"kl": 0.01043550155009143,
|
|
"entropy": 0.08659344958141446,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.824491767212749,
|
|
"epoch": 0.004557291666666667,
|
|
"step": 175
|
|
},
|
|
{
|
|
"loss": 0.10819478332996368,
|
|
"grad_norm": 4.050840854644775,
|
|
"learning_rate": 2.155172413793103e-07,
|
|
"num_tokens": 1617990.0,
|
|
"completions/mean_length": 323.125,
|
|
"completions/min_length": 248.0,
|
|
"completions/max_length": 496.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 323.125,
|
|
"completions/min_terminated_length": 248.0,
|
|
"completions/max_terminated_length": 496.0,
|
|
"tools/call_frequency": 7.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0037499996833503246,
|
|
"rewards/reward_func/std": 0.02386719174683094,
|
|
"reward": 0.0037499996833503246,
|
|
"reward_std": 0.02386719174683094,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007743766065686941,
|
|
"sampling/sampling_logp_difference/max": 0.46173572540283203,
|
|
"sampling/importance_sampling_ratio/min": 0.5169284343719482,
|
|
"sampling/importance_sampling_ratio/mean": 0.6347374320030212,
|
|
"sampling/importance_sampling_ratio/max": 0.8881505131721497,
|
|
"kl": 0.016461904189782217,
|
|
"entropy": 0.0991242746822536,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 7.148869596421719,
|
|
"epoch": 0.004583333333333333,
|
|
"step": 176
|
|
},
|
|
{
|
|
"loss": 0.4952075481414795,
|
|
"grad_norm": 4.714170932769775,
|
|
"learning_rate": 2.1379310344827587e-07,
|
|
"num_tokens": 1627008.0,
|
|
"completions/mean_length": 441.75,
|
|
"completions/min_length": 222.0,
|
|
"completions/max_length": 904.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 287.8333435058594,
|
|
"completions/min_terminated_length": 222.0,
|
|
"completions/max_terminated_length": 403.0,
|
|
"tools/call_frequency": 11.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0036747658159583807,
|
|
"sampling/sampling_logp_difference/max": 0.406902551651001,
|
|
"sampling/importance_sampling_ratio/min": 0.6819117665290833,
|
|
"sampling/importance_sampling_ratio/mean": 1.2496057748794556,
|
|
"sampling/importance_sampling_ratio/max": 2.0169427394866943,
|
|
"kl": 0.010367191109253326,
|
|
"entropy": 0.06905269168782979,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.003465969115496,
|
|
"epoch": 0.004609375,
|
|
"step": 177
|
|
},
|
|
{
|
|
"loss": 0.18410532176494598,
|
|
"grad_norm": 4.070309638977051,
|
|
"learning_rate": 2.1206896551724138e-07,
|
|
"num_tokens": 1635903.0,
|
|
"completions/mean_length": 424.75,
|
|
"completions/min_length": 294.0,
|
|
"completions/max_length": 912.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 355.14288330078125,
|
|
"completions/min_terminated_length": 294.0,
|
|
"completions/max_terminated_length": 511.0,
|
|
"tools/call_frequency": 9.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006847224198281765,
|
|
"sampling/sampling_logp_difference/max": 0.8087801933288574,
|
|
"sampling/importance_sampling_ratio/min": 0.202076718211174,
|
|
"sampling/importance_sampling_ratio/mean": 0.7071809768676758,
|
|
"sampling/importance_sampling_ratio/max": 1.6358869075775146,
|
|
"kl": 0.015008093221695162,
|
|
"entropy": 0.09249526518397033,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.032261852174997,
|
|
"epoch": 0.004635416666666667,
|
|
"step": 178
|
|
},
|
|
{
|
|
"loss": 0.004809200763702393,
|
|
"grad_norm": 4.912271976470947,
|
|
"learning_rate": 2.103448275862069e-07,
|
|
"num_tokens": 1644272.0,
|
|
"completions/mean_length": 360.25,
|
|
"completions/min_length": 54.0,
|
|
"completions/max_length": 1053.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 360.25,
|
|
"completions/min_terminated_length": 54.0,
|
|
"completions/max_terminated_length": 1053.0,
|
|
"tools/call_frequency": 7.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006806019227951765,
|
|
"sampling/sampling_logp_difference/max": 0.6759872436523438,
|
|
"sampling/importance_sampling_ratio/min": 0.4946882426738739,
|
|
"sampling/importance_sampling_ratio/mean": 0.8796041011810303,
|
|
"sampling/importance_sampling_ratio/max": 1.6375784873962402,
|
|
"kl": 0.014993524004239589,
|
|
"entropy": 0.1118375463411212,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.609144873917103,
|
|
"epoch": 0.004661458333333333,
|
|
"step": 179
|
|
},
|
|
{
|
|
"loss": -0.16764119267463684,
|
|
"grad_norm": 2.9542858600616455,
|
|
"learning_rate": 2.086206896551724e-07,
|
|
"num_tokens": 1652682.0,
|
|
"completions/mean_length": 365.375,
|
|
"completions/min_length": 236.0,
|
|
"completions/max_length": 921.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 286.0,
|
|
"completions/min_terminated_length": 236.0,
|
|
"completions/max_terminated_length": 322.0,
|
|
"tools/call_frequency": 9.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.0,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005468287970870733,
|
|
"sampling/sampling_logp_difference/max": 0.8208304047584534,
|
|
"sampling/importance_sampling_ratio/min": 0.19086091220378876,
|
|
"sampling/importance_sampling_ratio/mean": 0.6315317153930664,
|
|
"sampling/importance_sampling_ratio/max": 1.0772355794906616,
|
|
"kl": 0.007307734194910154,
|
|
"entropy": 0.06448584096506238,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.761766800656915,
|
|
"epoch": 0.0046875,
|
|
"step": 180
|
|
},
|
|
{
|
|
"loss": 0.11693141609430313,
|
|
"grad_norm": 2.7521212100982666,
|
|
"learning_rate": 2.0689655172413793e-07,
|
|
"num_tokens": 1661932.0,
|
|
"completions/mean_length": 470.25,
|
|
"completions/min_length": 275.0,
|
|
"completions/max_length": 1002.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 409.5714416503906,
|
|
"completions/min_terminated_length": 275.0,
|
|
"completions/max_terminated_length": 1002.0,
|
|
"tools/call_frequency": 11.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004386642016470432,
|
|
"sampling/sampling_logp_difference/max": 0.664912223815918,
|
|
"sampling/importance_sampling_ratio/min": 0.31256696581840515,
|
|
"sampling/importance_sampling_ratio/mean": 0.7364121675491333,
|
|
"sampling/importance_sampling_ratio/max": 1.0970759391784668,
|
|
"kl": 0.009388577123900177,
|
|
"entropy": 0.07613171765115112,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.147337261587381,
|
|
"epoch": 0.004713541666666667,
|
|
"step": 181
|
|
},
|
|
{
|
|
"loss": 0.20711863040924072,
|
|
"grad_norm": 2.328280448913574,
|
|
"learning_rate": 2.0517241379310344e-07,
|
|
"num_tokens": 1671273.0,
|
|
"completions/mean_length": 482.75,
|
|
"completions/min_length": 267.0,
|
|
"completions/max_length": 975.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 412.4285888671875,
|
|
"completions/min_terminated_length": 267.0,
|
|
"completions/max_terminated_length": 927.0,
|
|
"tools/call_frequency": 10.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005277763120830059,
|
|
"sampling/sampling_logp_difference/max": 0.6020386815071106,
|
|
"sampling/importance_sampling_ratio/min": 0.2552863359451294,
|
|
"sampling/importance_sampling_ratio/mean": 0.5943953394889832,
|
|
"sampling/importance_sampling_ratio/max": 1.0889952182769775,
|
|
"kl": 0.012548870632599574,
|
|
"entropy": 0.07835292292293161,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.10030797868967,
|
|
"epoch": 0.0047395833333333335,
|
|
"step": 182
|
|
},
|
|
{
|
|
"loss": -0.008056513965129852,
|
|
"grad_norm": 3.597985029220581,
|
|
"learning_rate": 2.0344827586206895e-07,
|
|
"num_tokens": 1679010.0,
|
|
"completions/mean_length": 281.125,
|
|
"completions/min_length": 260.0,
|
|
"completions/max_length": 305.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 281.125,
|
|
"completions/min_terminated_length": 260.0,
|
|
"completions/max_terminated_length": 305.0,
|
|
"tools/call_frequency": 7.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007184979040175676,
|
|
"sampling/sampling_logp_difference/max": 0.7120084762573242,
|
|
"sampling/importance_sampling_ratio/min": 0.07046766579151154,
|
|
"sampling/importance_sampling_ratio/mean": 0.6944385766983032,
|
|
"sampling/importance_sampling_ratio/max": 1.289284586906433,
|
|
"kl": 0.011193251179065555,
|
|
"entropy": 0.08541464293375611,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.182709155604243,
|
|
"epoch": 0.004765625,
|
|
"step": 183
|
|
},
|
|
{
|
|
"loss": 0.1550823152065277,
|
|
"grad_norm": 6.878434181213379,
|
|
"learning_rate": 2.0172413793103446e-07,
|
|
"num_tokens": 1687508.0,
|
|
"completions/mean_length": 375.75,
|
|
"completions/min_length": 267.0,
|
|
"completions/max_length": 873.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 304.71429443359375,
|
|
"completions/min_terminated_length": 267.0,
|
|
"completions/max_terminated_length": 362.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.014880476519465446,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.014880475588142872,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004603398963809013,
|
|
"sampling/sampling_logp_difference/max": 0.6487945318222046,
|
|
"sampling/importance_sampling_ratio/min": 0.2580660879611969,
|
|
"sampling/importance_sampling_ratio/mean": 1.0643993616104126,
|
|
"sampling/importance_sampling_ratio/max": 2.123004198074341,
|
|
"kl": 0.013062086174613796,
|
|
"entropy": 0.059464852791279554,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.084156369790435,
|
|
"epoch": 0.004791666666666666,
|
|
"step": 184
|
|
},
|
|
{
|
|
"loss": -0.02415241301059723,
|
|
"grad_norm": 3.178339958190918,
|
|
"learning_rate": 2e-07,
|
|
"num_tokens": 1696154.0,
|
|
"completions/mean_length": 394.375,
|
|
"completions/min_length": 264.0,
|
|
"completions/max_length": 914.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 320.14288330078125,
|
|
"completions/min_terminated_length": 264.0,
|
|
"completions/max_terminated_length": 388.0,
|
|
"tools/call_frequency": 9.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.00624999962747097,
|
|
"rewards/reward_func/std": 0.024458419531583786,
|
|
"reward": 0.00624999962747097,
|
|
"reward_std": 0.024458419531583786,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006546440068632364,
|
|
"sampling/sampling_logp_difference/max": 0.4030141830444336,
|
|
"sampling/importance_sampling_ratio/min": 0.25485801696777344,
|
|
"sampling/importance_sampling_ratio/mean": 0.7779988050460815,
|
|
"sampling/importance_sampling_ratio/max": 1.346533179283142,
|
|
"kl": 0.013216295345955587,
|
|
"entropy": 0.09463539719581604,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.202319471165538,
|
|
"epoch": 0.0048177083333333336,
|
|
"step": 185
|
|
},
|
|
{
|
|
"loss": 0.6004295945167542,
|
|
"grad_norm": 4.816537857055664,
|
|
"learning_rate": 1.9827586206896553e-07,
|
|
"num_tokens": 1704668.0,
|
|
"completions/mean_length": 377.875,
|
|
"completions/min_length": 260.0,
|
|
"completions/max_length": 915.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 377.875,
|
|
"completions/min_terminated_length": 260.0,
|
|
"completions/max_terminated_length": 915.0,
|
|
"tools/call_frequency": 9.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.014999999664723873,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.014999999664723873,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006655080709606409,
|
|
"sampling/sampling_logp_difference/max": 0.44252705574035645,
|
|
"sampling/importance_sampling_ratio/min": 0.4718688428401947,
|
|
"sampling/importance_sampling_ratio/mean": 0.7980417609214783,
|
|
"sampling/importance_sampling_ratio/max": 1.3584773540496826,
|
|
"kl": 0.01720045687397942,
|
|
"entropy": 0.09600168094038963,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.794458897784352,
|
|
"epoch": 0.00484375,
|
|
"step": 186
|
|
},
|
|
{
|
|
"loss": -0.08035939186811447,
|
|
"grad_norm": 6.85352897644043,
|
|
"learning_rate": 1.9655172413793103e-07,
|
|
"num_tokens": 1712444.0,
|
|
"completions/mean_length": 285.625,
|
|
"completions/min_length": 229.0,
|
|
"completions/max_length": 381.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 285.625,
|
|
"completions/min_terminated_length": 229.0,
|
|
"completions/max_terminated_length": 381.0,
|
|
"tools/call_frequency": 7.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0065433732233941555,
|
|
"sampling/sampling_logp_difference/max": 0.6092119216918945,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.745704174041748,
|
|
"sampling/importance_sampling_ratio/max": 1.5213173627853394,
|
|
"kl": 0.011352399364113808,
|
|
"entropy": 0.09392454428598285,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.627636021003127,
|
|
"epoch": 0.004869791666666666,
|
|
"step": 187
|
|
},
|
|
{
|
|
"loss": 0.46371421217918396,
|
|
"grad_norm": 9.625018119812012,
|
|
"learning_rate": 1.9482758620689654e-07,
|
|
"num_tokens": 1720214.0,
|
|
"completions/mean_length": 285.625,
|
|
"completions/min_length": 239.0,
|
|
"completions/max_length": 325.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 285.625,
|
|
"completions/min_terminated_length": 239.0,
|
|
"completions/max_terminated_length": 325.0,
|
|
"tools/call_frequency": 7.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.01414213515818119,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.01414213515818119,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.008557692170143127,
|
|
"sampling/sampling_logp_difference/max": 1.2061786651611328,
|
|
"sampling/importance_sampling_ratio/min": 0.09275034070014954,
|
|
"sampling/importance_sampling_ratio/mean": 0.8954262733459473,
|
|
"sampling/importance_sampling_ratio/max": 1.9961811304092407,
|
|
"kl": 0.02454577063326724,
|
|
"entropy": 0.09649293217808008,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.599423883482814,
|
|
"epoch": 0.004895833333333334,
|
|
"step": 188
|
|
},
|
|
{
|
|
"loss": 0.023951441049575806,
|
|
"grad_norm": 3.927997350692749,
|
|
"learning_rate": 1.9310344827586205e-07,
|
|
"num_tokens": 1728383.0,
|
|
"completions/mean_length": 335.125,
|
|
"completions/min_length": 230.0,
|
|
"completions/max_length": 873.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 335.125,
|
|
"completions/min_terminated_length": 230.0,
|
|
"completions/max_terminated_length": 873.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005485013592988253,
|
|
"sampling/sampling_logp_difference/max": 0.4885861873626709,
|
|
"sampling/importance_sampling_ratio/min": 0.22928155958652496,
|
|
"sampling/importance_sampling_ratio/mean": 0.7439508438110352,
|
|
"sampling/importance_sampling_ratio/max": 1.510162353515625,
|
|
"kl": 0.0063320136978290975,
|
|
"entropy": 0.0742957228794694,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.099998611956835,
|
|
"epoch": 0.004921875,
|
|
"step": 189
|
|
},
|
|
{
|
|
"loss": 0.3788870573043823,
|
|
"grad_norm": 4.601541996002197,
|
|
"learning_rate": 1.9137931034482756e-07,
|
|
"num_tokens": 1736844.0,
|
|
"completions/mean_length": 372.125,
|
|
"completions/min_length": 16.0,
|
|
"completions/max_length": 900.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 312.4285888671875,
|
|
"completions/min_terminated_length": 16.0,
|
|
"completions/max_terminated_length": 900.0,
|
|
"tools/call_frequency": 10.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005046514328569174,
|
|
"sampling/sampling_logp_difference/max": 0.7030599117279053,
|
|
"sampling/importance_sampling_ratio/min": 0.2662879228591919,
|
|
"sampling/importance_sampling_ratio/mean": 0.8889700174331665,
|
|
"sampling/importance_sampling_ratio/max": 1.668057918548584,
|
|
"kl": 0.011069814747315831,
|
|
"entropy": 0.06667471880791709,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.216938806697726,
|
|
"epoch": 0.0049479166666666664,
|
|
"step": 190
|
|
},
|
|
{
|
|
"loss": 0.15704400837421417,
|
|
"grad_norm": 8.36581039428711,
|
|
"learning_rate": 1.896551724137931e-07,
|
|
"num_tokens": 1744726.0,
|
|
"completions/mean_length": 298.5,
|
|
"completions/min_length": 223.0,
|
|
"completions/max_length": 417.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 298.5,
|
|
"completions/min_terminated_length": 223.0,
|
|
"completions/max_terminated_length": 417.0,
|
|
"tools/call_frequency": 7.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007063709665089846,
|
|
"sampling/sampling_logp_difference/max": 0.5444717407226562,
|
|
"sampling/importance_sampling_ratio/min": 0.4139198958873749,
|
|
"sampling/importance_sampling_ratio/mean": 1.2709922790527344,
|
|
"sampling/importance_sampling_ratio/max": 2.4579975605010986,
|
|
"kl": 0.010441938284202479,
|
|
"entropy": 0.08181588677689433,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.833273472264409,
|
|
"epoch": 0.004973958333333334,
|
|
"step": 191
|
|
},
|
|
{
|
|
"loss": 0.5298736095428467,
|
|
"grad_norm": 4.761072158813477,
|
|
"learning_rate": 1.879310344827586e-07,
|
|
"num_tokens": 1754335.0,
|
|
"completions/mean_length": 515.0,
|
|
"completions/min_length": 222.0,
|
|
"completions/max_length": 956.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 277.3999938964844,
|
|
"completions/min_terminated_length": 222.0,
|
|
"completions/max_terminated_length": 328.0,
|
|
"tools/call_frequency": 12.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003510719398036599,
|
|
"sampling/sampling_logp_difference/max": 1.5242676734924316,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 1.3581147193908691,
|
|
"sampling/importance_sampling_ratio/max": 2.3688228130340576,
|
|
"kl": 0.010152340613785782,
|
|
"entropy": 0.050915788335260004,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.738830855116248,
|
|
"epoch": 0.005,
|
|
"step": 192
|
|
},
|
|
{
|
|
"loss": 0.3305976986885071,
|
|
"grad_norm": 5.408193111419678,
|
|
"learning_rate": 1.8620689655172414e-07,
|
|
"num_tokens": 1762662.0,
|
|
"completions/mean_length": 355.375,
|
|
"completions/min_length": 235.0,
|
|
"completions/max_length": 868.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 282.14288330078125,
|
|
"completions/min_terminated_length": 235.0,
|
|
"completions/max_terminated_length": 335.0,
|
|
"tools/call_frequency": 9.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0055471546947956085,
|
|
"sampling/sampling_logp_difference/max": 0.4533047676086426,
|
|
"sampling/importance_sampling_ratio/min": 0.3283662796020508,
|
|
"sampling/importance_sampling_ratio/mean": 1.1884596347808838,
|
|
"sampling/importance_sampling_ratio/max": 1.8613319396972656,
|
|
"kl": 0.009093987868254771,
|
|
"entropy": 0.08643521228805184,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.757152412086725,
|
|
"epoch": 0.0050260416666666665,
|
|
"step": 193
|
|
},
|
|
{
|
|
"loss": 0.690026044845581,
|
|
"grad_norm": 4.415740489959717,
|
|
"learning_rate": 1.8448275862068964e-07,
|
|
"num_tokens": 1771717.0,
|
|
"completions/mean_length": 446.0,
|
|
"completions/min_length": 199.0,
|
|
"completions/max_length": 952.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 283.66668701171875,
|
|
"completions/min_terminated_length": 199.0,
|
|
"completions/max_terminated_length": 406.0,
|
|
"tools/call_frequency": 11.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0047027007676661015,
|
|
"sampling/sampling_logp_difference/max": 0.4399834871292114,
|
|
"sampling/importance_sampling_ratio/min": 0.2970791757106781,
|
|
"sampling/importance_sampling_ratio/mean": 0.9641309976577759,
|
|
"sampling/importance_sampling_ratio/max": 2.6016318798065186,
|
|
"kl": 0.010459220327902585,
|
|
"entropy": 0.08191939466632903,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.49713927321136,
|
|
"epoch": 0.005052083333333333,
|
|
"step": 194
|
|
},
|
|
{
|
|
"loss": 0.46394893527030945,
|
|
"grad_norm": 2.905545473098755,
|
|
"learning_rate": 1.8275862068965518e-07,
|
|
"num_tokens": 1780134.0,
|
|
"completions/mean_length": 366.375,
|
|
"completions/min_length": 251.0,
|
|
"completions/max_length": 731.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 366.375,
|
|
"completions/min_terminated_length": 251.0,
|
|
"completions/max_terminated_length": 731.0,
|
|
"tools/call_frequency": 10.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.013749999925494194,
|
|
"rewards/reward_func/std": 0.01767767034471035,
|
|
"reward": 0.013749999925494194,
|
|
"reward_std": 0.0176776684820652,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006837677676230669,
|
|
"sampling/sampling_logp_difference/max": 0.9431126117706299,
|
|
"sampling/importance_sampling_ratio/min": 0.07040684670209885,
|
|
"sampling/importance_sampling_ratio/mean": 0.7594817876815796,
|
|
"sampling/importance_sampling_ratio/max": 1.6868196725845337,
|
|
"kl": 0.009364684810861945,
|
|
"entropy": 0.09019348560832441,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.291951604187489,
|
|
"epoch": 0.005078125,
|
|
"step": 195
|
|
},
|
|
{
|
|
"loss": 0.7447500824928284,
|
|
"grad_norm": 7.7760009765625,
|
|
"learning_rate": 1.810344827586207e-07,
|
|
"num_tokens": 1789178.0,
|
|
"completions/mean_length": 444.875,
|
|
"completions/min_length": 228.0,
|
|
"completions/max_length": 934.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 379.14288330078125,
|
|
"completions/min_terminated_length": 228.0,
|
|
"completions/max_terminated_length": 934.0,
|
|
"tools/call_frequency": 10.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004274965729564428,
|
|
"sampling/sampling_logp_difference/max": 0.45672106742858887,
|
|
"sampling/importance_sampling_ratio/min": 0.41907447576522827,
|
|
"sampling/importance_sampling_ratio/mean": 0.8791233897209167,
|
|
"sampling/importance_sampling_ratio/max": 1.6555150747299194,
|
|
"kl": 0.014116399755948805,
|
|
"entropy": 0.07077089021913707,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.031839961186051,
|
|
"epoch": 0.005104166666666667,
|
|
"step": 196
|
|
},
|
|
{
|
|
"loss": 0.5225419402122498,
|
|
"grad_norm": 4.841485977172852,
|
|
"learning_rate": 1.793103448275862e-07,
|
|
"num_tokens": 1798021.0,
|
|
"completions/mean_length": 419.625,
|
|
"completions/min_length": 226.0,
|
|
"completions/max_length": 917.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 366.4285888671875,
|
|
"completions/min_terminated_length": 226.0,
|
|
"completions/max_terminated_length": 917.0,
|
|
"tools/call_frequency": 11.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004878918174654245,
|
|
"sampling/sampling_logp_difference/max": 0.4819919466972351,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.7972027063369751,
|
|
"sampling/importance_sampling_ratio/max": 2.156886339187622,
|
|
"kl": 0.014213947288226336,
|
|
"entropy": 0.07599374675191939,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.44630222581327,
|
|
"epoch": 0.005130208333333333,
|
|
"step": 197
|
|
},
|
|
{
|
|
"loss": 0.7860352993011475,
|
|
"grad_norm": 2.9424397945404053,
|
|
"learning_rate": 1.775862068965517e-07,
|
|
"num_tokens": 1808262.0,
|
|
"completions/mean_length": 594.125,
|
|
"completions/min_length": 212.0,
|
|
"completions/max_length": 950.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 485.5,
|
|
"completions/min_terminated_length": 212.0,
|
|
"completions/max_terminated_length": 918.0,
|
|
"tools/call_frequency": 14.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0037405744660645723,
|
|
"sampling/sampling_logp_difference/max": 0.6202025413513184,
|
|
"sampling/importance_sampling_ratio/min": 0.4386453628540039,
|
|
"sampling/importance_sampling_ratio/mean": 1.2641130685806274,
|
|
"sampling/importance_sampling_ratio/max": 2.8301429748535156,
|
|
"kl": 0.010706182043577428,
|
|
"entropy": 0.05549907009117305,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.233762234449387,
|
|
"epoch": 0.00515625,
|
|
"step": 198
|
|
},
|
|
{
|
|
"loss": -0.06379703432321548,
|
|
"grad_norm": 5.798716068267822,
|
|
"learning_rate": 1.758620689655172e-07,
|
|
"num_tokens": 1815670.0,
|
|
"completions/mean_length": 240.75,
|
|
"completions/min_length": 14.0,
|
|
"completions/max_length": 352.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 240.75,
|
|
"completions/min_terminated_length": 14.0,
|
|
"completions/max_terminated_length": 352.0,
|
|
"tools/call_frequency": 6.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.008095236495137215,
|
|
"sampling/sampling_logp_difference/max": 0.6814284324645996,
|
|
"sampling/importance_sampling_ratio/min": 0.31352534890174866,
|
|
"sampling/importance_sampling_ratio/mean": 0.7094199657440186,
|
|
"sampling/importance_sampling_ratio/max": 1.0965360403060913,
|
|
"kl": 0.010527506747166626,
|
|
"entropy": 0.09722990076988935,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.208026530221105,
|
|
"epoch": 0.005182291666666667,
|
|
"step": 199
|
|
},
|
|
{
|
|
"loss": 0.00420079380273819,
|
|
"grad_norm": 4.5245137214660645,
|
|
"learning_rate": 1.7413793103448275e-07,
|
|
"num_tokens": 1823292.0,
|
|
"completions/mean_length": 266.875,
|
|
"completions/min_length": 101.0,
|
|
"completions/max_length": 350.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 266.875,
|
|
"completions/min_terminated_length": 101.0,
|
|
"completions/max_terminated_length": 350.0,
|
|
"tools/call_frequency": 6.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.012500000186264515,
|
|
"rewards/reward_func/std": 0.027645720168948174,
|
|
"reward": 0.012500000186264515,
|
|
"reward_std": 0.027645718306303024,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.009022197686135769,
|
|
"sampling/sampling_logp_difference/max": 0.44314849376678467,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.8128917217254639,
|
|
"sampling/importance_sampling_ratio/max": 1.3757731914520264,
|
|
"kl": 0.011725208925781772,
|
|
"entropy": 0.10286683309823275,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.614142658188939,
|
|
"epoch": 0.005208333333333333,
|
|
"step": 200
|
|
},
|
|
{
|
|
"loss": 0.3259810507297516,
|
|
"grad_norm": 4.861057758331299,
|
|
"learning_rate": 1.7241379310344828e-07,
|
|
"num_tokens": 1832229.0,
|
|
"completions/mean_length": 432.0,
|
|
"completions/min_length": 211.0,
|
|
"completions/max_length": 885.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 369.71429443359375,
|
|
"completions/min_terminated_length": 211.0,
|
|
"completions/max_terminated_length": 885.0,
|
|
"tools/call_frequency": 11.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0054626488126814365,
|
|
"sampling/sampling_logp_difference/max": 0.6299458742141724,
|
|
"sampling/importance_sampling_ratio/min": 0.37601593136787415,
|
|
"sampling/importance_sampling_ratio/mean": 1.0792756080627441,
|
|
"sampling/importance_sampling_ratio/max": 2.5133056640625,
|
|
"kl": 0.010409980575786904,
|
|
"entropy": 0.0824297599028796,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.48865781724453,
|
|
"epoch": 0.005234375,
|
|
"step": 201
|
|
},
|
|
{
|
|
"loss": 0.22302263975143433,
|
|
"grad_norm": 8.754408836364746,
|
|
"learning_rate": 1.706896551724138e-07,
|
|
"num_tokens": 1840743.0,
|
|
"completions/mean_length": 378.5,
|
|
"completions/min_length": 236.0,
|
|
"completions/max_length": 895.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 304.71429443359375,
|
|
"completions/min_terminated_length": 236.0,
|
|
"completions/max_terminated_length": 365.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006168486550450325,
|
|
"sampling/sampling_logp_difference/max": 0.5609146356582642,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 1.1544127464294434,
|
|
"sampling/importance_sampling_ratio/max": 2.3783020973205566,
|
|
"kl": 0.014555628236848861,
|
|
"entropy": 0.09192336280830204,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.128158314153552,
|
|
"epoch": 0.005260416666666667,
|
|
"step": 202
|
|
},
|
|
{
|
|
"loss": 0.03685027360916138,
|
|
"grad_norm": 4.780720233917236,
|
|
"learning_rate": 1.689655172413793e-07,
|
|
"num_tokens": 1850545.0,
|
|
"completions/mean_length": 539.375,
|
|
"completions/min_length": 261.0,
|
|
"completions/max_length": 930.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 539.375,
|
|
"completions/min_terminated_length": 261.0,
|
|
"completions/max_terminated_length": 930.0,
|
|
"tools/call_frequency": 13.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.0,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003720005275681615,
|
|
"sampling/sampling_logp_difference/max": 1.2656283378601074,
|
|
"sampling/importance_sampling_ratio/min": 0.2228778898715973,
|
|
"sampling/importance_sampling_ratio/mean": 0.7731937170028687,
|
|
"sampling/importance_sampling_ratio/max": 1.7620240449905396,
|
|
"kl": 0.009604091101209633,
|
|
"entropy": 0.06727365998085588,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.728407178074121,
|
|
"epoch": 0.005286458333333333,
|
|
"step": 203
|
|
},
|
|
{
|
|
"loss": 0.34542468190193176,
|
|
"grad_norm": 3.677611827850342,
|
|
"learning_rate": 1.6724137931034483e-07,
|
|
"num_tokens": 1859499.0,
|
|
"completions/mean_length": 433.625,
|
|
"completions/min_length": 232.0,
|
|
"completions/max_length": 891.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 433.625,
|
|
"completions/min_terminated_length": 232.0,
|
|
"completions/max_terminated_length": 891.0,
|
|
"tools/call_frequency": 11.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004896019119769335,
|
|
"sampling/sampling_logp_difference/max": 0.8252008557319641,
|
|
"sampling/importance_sampling_ratio/min": 0.37006932497024536,
|
|
"sampling/importance_sampling_ratio/mean": 0.8097536563873291,
|
|
"sampling/importance_sampling_ratio/max": 1.3510611057281494,
|
|
"kl": 0.013208601259975694,
|
|
"entropy": 0.07727039302699268,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.187844906002283,
|
|
"epoch": 0.0053125,
|
|
"step": 204
|
|
},
|
|
{
|
|
"loss": 0.27115777134895325,
|
|
"grad_norm": 3.295074701309204,
|
|
"learning_rate": 1.6551724137931034e-07,
|
|
"num_tokens": 1869661.0,
|
|
"completions/mean_length": 584.25,
|
|
"completions/min_length": 248.0,
|
|
"completions/max_length": 950.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 268.5,
|
|
"completions/min_terminated_length": 248.0,
|
|
"completions/max_terminated_length": 296.0,
|
|
"tools/call_frequency": 14.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003074244363233447,
|
|
"sampling/sampling_logp_difference/max": 0.48839902877807617,
|
|
"sampling/importance_sampling_ratio/min": 0.5709440112113953,
|
|
"sampling/importance_sampling_ratio/mean": 1.0870229005813599,
|
|
"sampling/importance_sampling_ratio/max": 1.5152596235275269,
|
|
"kl": 0.008057826040385407,
|
|
"entropy": 0.05944025958888233,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.64357578381896,
|
|
"epoch": 0.005338541666666667,
|
|
"step": 205
|
|
},
|
|
{
|
|
"loss": 0.4306704103946686,
|
|
"grad_norm": 2.1126129627227783,
|
|
"learning_rate": 1.6379310344827585e-07,
|
|
"num_tokens": 1879239.0,
|
|
"completions/mean_length": 511.0,
|
|
"completions/min_length": 223.0,
|
|
"completions/max_length": 985.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 267.3999938964844,
|
|
"completions/min_terminated_length": 223.0,
|
|
"completions/max_terminated_length": 324.0,
|
|
"tools/call_frequency": 12.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003096927423030138,
|
|
"sampling/sampling_logp_difference/max": 0.4760732650756836,
|
|
"sampling/importance_sampling_ratio/min": 0.4842031002044678,
|
|
"sampling/importance_sampling_ratio/mean": 0.9973984360694885,
|
|
"sampling/importance_sampling_ratio/max": 1.7781884670257568,
|
|
"kl": 0.01244279180536978,
|
|
"entropy": 0.05290089244954288,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.096541641280055,
|
|
"epoch": 0.005364583333333333,
|
|
"step": 206
|
|
},
|
|
{
|
|
"loss": 0.25513696670532227,
|
|
"grad_norm": 2.6380980014801025,
|
|
"learning_rate": 1.6206896551724136e-07,
|
|
"num_tokens": 1888923.0,
|
|
"completions/mean_length": 525.625,
|
|
"completions/min_length": 239.0,
|
|
"completions/max_length": 955.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 390.0,
|
|
"completions/min_terminated_length": 239.0,
|
|
"completions/max_terminated_length": 856.0,
|
|
"tools/call_frequency": 13.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004550670739263296,
|
|
"sampling/sampling_logp_difference/max": 0.7925977110862732,
|
|
"sampling/importance_sampling_ratio/min": 0.20352283120155334,
|
|
"sampling/importance_sampling_ratio/mean": 0.7297013401985168,
|
|
"sampling/importance_sampling_ratio/max": 2.0595176219940186,
|
|
"kl": 0.009407802270288812,
|
|
"entropy": 0.07345355022698641,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.497568165883422,
|
|
"epoch": 0.005390625,
|
|
"step": 207
|
|
},
|
|
{
|
|
"loss": 0.05194547027349472,
|
|
"grad_norm": 2.0321733951568604,
|
|
"learning_rate": 1.6034482758620686e-07,
|
|
"num_tokens": 1897912.0,
|
|
"completions/mean_length": 437.75,
|
|
"completions/min_length": 224.0,
|
|
"completions/max_length": 920.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 284.0,
|
|
"completions/min_terminated_length": 224.0,
|
|
"completions/max_terminated_length": 342.0,
|
|
"tools/call_frequency": 10.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.01414213515818119,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.01414213515818119,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005642731674015522,
|
|
"sampling/sampling_logp_difference/max": 1.057381272315979,
|
|
"sampling/importance_sampling_ratio/min": 0.32317325472831726,
|
|
"sampling/importance_sampling_ratio/mean": 0.7260663509368896,
|
|
"sampling/importance_sampling_ratio/max": 1.594076156616211,
|
|
"kl": 0.013691804590052925,
|
|
"entropy": 0.08363189385272563,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.344809915870428,
|
|
"epoch": 0.005416666666666667,
|
|
"step": 208
|
|
},
|
|
{
|
|
"loss": 0.5507755279541016,
|
|
"grad_norm": 8.91094970703125,
|
|
"learning_rate": 1.5862068965517243e-07,
|
|
"num_tokens": 1906861.0,
|
|
"completions/mean_length": 432.625,
|
|
"completions/min_length": 232.0,
|
|
"completions/max_length": 919.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 363.14288330078125,
|
|
"completions/min_terminated_length": 232.0,
|
|
"completions/max_terminated_length": 885.0,
|
|
"tools/call_frequency": 10.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.014880476519465446,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.014880475588142872,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005022066179662943,
|
|
"sampling/sampling_logp_difference/max": 0.7026243209838867,
|
|
"sampling/importance_sampling_ratio/min": 0.7715235948562622,
|
|
"sampling/importance_sampling_ratio/mean": 1.3284590244293213,
|
|
"sampling/importance_sampling_ratio/max": 2.0784852504730225,
|
|
"kl": 0.010684128446882823,
|
|
"entropy": 0.07906584884040058,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.111095648258924,
|
|
"epoch": 0.005442708333333333,
|
|
"step": 209
|
|
},
|
|
{
|
|
"loss": 0.4049566686153412,
|
|
"grad_norm": 1.878878116607666,
|
|
"learning_rate": 1.5689655172413793e-07,
|
|
"num_tokens": 1917812.0,
|
|
"completions/mean_length": 683.5,
|
|
"completions/min_length": 232.0,
|
|
"completions/max_length": 990.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 542.0,
|
|
"completions/min_terminated_length": 232.0,
|
|
"completions/max_terminated_length": 899.0,
|
|
"tools/call_frequency": 16.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0021994865965098143,
|
|
"sampling/sampling_logp_difference/max": 0.4519634246826172,
|
|
"sampling/importance_sampling_ratio/min": 0.3885037302970886,
|
|
"sampling/importance_sampling_ratio/mean": 1.079639196395874,
|
|
"sampling/importance_sampling_ratio/max": 1.9972246885299683,
|
|
"kl": 0.004377339570055483,
|
|
"entropy": 0.04309751570690423,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.118454433977604,
|
|
"epoch": 0.00546875,
|
|
"step": 210
|
|
},
|
|
{
|
|
"loss": 0.30709511041641235,
|
|
"grad_norm": 1.7129255533218384,
|
|
"learning_rate": 1.5517241379310344e-07,
|
|
"num_tokens": 1928300.0,
|
|
"completions/mean_length": 625.625,
|
|
"completions/min_length": 306.0,
|
|
"completions/max_length": 930.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 454.6000061035156,
|
|
"completions/min_terminated_length": 306.0,
|
|
"completions/max_terminated_length": 908.0,
|
|
"tools/call_frequency": 15.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0036662286147475243,
|
|
"sampling/sampling_logp_difference/max": 0.9846904277801514,
|
|
"sampling/importance_sampling_ratio/min": 0.35380998253822327,
|
|
"sampling/importance_sampling_ratio/mean": 0.9664549827575684,
|
|
"sampling/importance_sampling_ratio/max": 2.491652727127075,
|
|
"kl": 0.0038907830930838827,
|
|
"entropy": 0.05476259533315897,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.252161404117942,
|
|
"epoch": 0.005494791666666667,
|
|
"step": 211
|
|
},
|
|
{
|
|
"loss": 0.19826537370681763,
|
|
"grad_norm": 5.184826850891113,
|
|
"learning_rate": 1.5344827586206895e-07,
|
|
"num_tokens": 1936775.0,
|
|
"completions/mean_length": 373.625,
|
|
"completions/min_length": 221.0,
|
|
"completions/max_length": 920.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 373.625,
|
|
"completions/min_terminated_length": 221.0,
|
|
"completions/max_terminated_length": 920.0,
|
|
"tools/call_frequency": 9.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0067168776877224445,
|
|
"sampling/sampling_logp_difference/max": 0.648794412612915,
|
|
"sampling/importance_sampling_ratio/min": 0.266431599855423,
|
|
"sampling/importance_sampling_ratio/mean": 0.9259718060493469,
|
|
"sampling/importance_sampling_ratio/max": 1.6832388639450073,
|
|
"kl": 0.010728500106779393,
|
|
"entropy": 0.07544480776414275,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.35914597287774,
|
|
"epoch": 0.005520833333333333,
|
|
"step": 212
|
|
},
|
|
{
|
|
"loss": 0.13625037670135498,
|
|
"grad_norm": 4.069812297821045,
|
|
"learning_rate": 1.5172413793103449e-07,
|
|
"num_tokens": 1945925.0,
|
|
"completions/mean_length": 458.125,
|
|
"completions/min_length": 248.0,
|
|
"completions/max_length": 977.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 384.0000305175781,
|
|
"completions/min_terminated_length": 248.0,
|
|
"completions/max_terminated_length": 906.0,
|
|
"tools/call_frequency": 11.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005248204339295626,
|
|
"sampling/sampling_logp_difference/max": 0.40825724601745605,
|
|
"sampling/importance_sampling_ratio/min": 0.3797793388366699,
|
|
"sampling/importance_sampling_ratio/mean": 1.012399673461914,
|
|
"sampling/importance_sampling_ratio/max": 1.6649028062820435,
|
|
"kl": 0.004666758032726648,
|
|
"entropy": 0.09495618473738432,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.016700683161616,
|
|
"epoch": 0.005546875,
|
|
"step": 213
|
|
},
|
|
{
|
|
"loss": 0.2760922312736511,
|
|
"grad_norm": 7.027612686157227,
|
|
"learning_rate": 1.5e-07,
|
|
"num_tokens": 1953497.0,
|
|
"completions/mean_length": 260.125,
|
|
"completions/min_length": 122.0,
|
|
"completions/max_length": 327.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 260.125,
|
|
"completions/min_terminated_length": 122.0,
|
|
"completions/max_terminated_length": 327.0,
|
|
"tools/call_frequency": 7.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.011250000447034836,
|
|
"rewards/reward_func/std": 0.021001702174544334,
|
|
"reward": 0.011250000447034836,
|
|
"reward_std": 0.021001700311899185,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006467821542173624,
|
|
"sampling/sampling_logp_difference/max": 0.7257360816001892,
|
|
"sampling/importance_sampling_ratio/min": 0.668561577796936,
|
|
"sampling/importance_sampling_ratio/mean": 1.1528606414794922,
|
|
"sampling/importance_sampling_ratio/max": 1.6346029043197632,
|
|
"kl": 0.008714034214790445,
|
|
"entropy": 0.0790727618150413,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.081474890932441,
|
|
"epoch": 0.005572916666666667,
|
|
"step": 214
|
|
},
|
|
{
|
|
"loss": 0.14856287837028503,
|
|
"grad_norm": 4.055591583251953,
|
|
"learning_rate": 1.482758620689655e-07,
|
|
"num_tokens": 1961766.0,
|
|
"completions/mean_length": 346.875,
|
|
"completions/min_length": 213.0,
|
|
"completions/max_length": 903.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 267.4285888671875,
|
|
"completions/min_terminated_length": 213.0,
|
|
"completions/max_terminated_length": 323.0,
|
|
"tools/call_frequency": 9.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0058244201354682446,
|
|
"sampling/sampling_logp_difference/max": 0.5278515815734863,
|
|
"sampling/importance_sampling_ratio/min": 0.45679914951324463,
|
|
"sampling/importance_sampling_ratio/mean": 0.7222980260848999,
|
|
"sampling/importance_sampling_ratio/max": 1.7350542545318604,
|
|
"kl": 0.007983293871802744,
|
|
"entropy": 0.09123679832555354,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.957878882065415,
|
|
"epoch": 0.005598958333333333,
|
|
"step": 215
|
|
},
|
|
{
|
|
"loss": 0.21726903319358826,
|
|
"grad_norm": 4.068999767303467,
|
|
"learning_rate": 1.46551724137931e-07,
|
|
"num_tokens": 1970670.0,
|
|
"completions/mean_length": 426.875,
|
|
"completions/min_length": 240.0,
|
|
"completions/max_length": 882.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 278.0,
|
|
"completions/min_terminated_length": 240.0,
|
|
"completions/max_terminated_length": 374.0,
|
|
"tools/call_frequency": 11.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0034573969896882772,
|
|
"sampling/sampling_logp_difference/max": 0.4873371720314026,
|
|
"sampling/importance_sampling_ratio/min": 0.4590717852115631,
|
|
"sampling/importance_sampling_ratio/mean": 1.2062077522277832,
|
|
"sampling/importance_sampling_ratio/max": 1.9513335227966309,
|
|
"kl": 0.009382102622112143,
|
|
"entropy": 0.05402011063415557,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.251953925937414,
|
|
"epoch": 0.005625,
|
|
"step": 216
|
|
},
|
|
{
|
|
"loss": 0.29229333996772766,
|
|
"grad_norm": 1.4313315153121948,
|
|
"learning_rate": 1.4482758620689654e-07,
|
|
"num_tokens": 1981479.0,
|
|
"completions/mean_length": 666.0,
|
|
"completions/min_length": 226.0,
|
|
"completions/max_length": 1020.0,
|
|
"completions/clipped_ratio": 0.625,
|
|
"completions/mean_terminated_length": 265.3333435058594,
|
|
"completions/min_terminated_length": 226.0,
|
|
"completions/max_terminated_length": 294.0,
|
|
"tools/call_frequency": 16.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0019905187655240297,
|
|
"sampling/sampling_logp_difference/max": 0.48215246200561523,
|
|
"sampling/importance_sampling_ratio/min": 0.5785024166107178,
|
|
"sampling/importance_sampling_ratio/mean": 0.8174154162406921,
|
|
"sampling/importance_sampling_ratio/max": 1.2137908935546875,
|
|
"kl": 0.004790996799783898,
|
|
"entropy": 0.03503893502056599,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.526190172880888,
|
|
"epoch": 0.005651041666666667,
|
|
"step": 217
|
|
},
|
|
{
|
|
"loss": -0.13751596212387085,
|
|
"grad_norm": 4.4416961669921875,
|
|
"learning_rate": 1.4310344827586208e-07,
|
|
"num_tokens": 1989232.0,
|
|
"completions/mean_length": 283.0,
|
|
"completions/min_length": 245.0,
|
|
"completions/max_length": 326.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 283.0,
|
|
"completions/min_terminated_length": 245.0,
|
|
"completions/max_terminated_length": 326.0,
|
|
"tools/call_frequency": 7.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.015118579380214214,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.015118579380214214,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007030034437775612,
|
|
"sampling/sampling_logp_difference/max": 0.43896234035491943,
|
|
"sampling/importance_sampling_ratio/min": 0.2981618344783783,
|
|
"sampling/importance_sampling_ratio/mean": 0.978378415107727,
|
|
"sampling/importance_sampling_ratio/max": 1.506421446800232,
|
|
"kl": 0.0177495478419587,
|
|
"entropy": 0.08014145749621093,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.690877456218004,
|
|
"epoch": 0.0056770833333333335,
|
|
"step": 218
|
|
},
|
|
{
|
|
"loss": 0.48176294565200806,
|
|
"grad_norm": 4.9243316650390625,
|
|
"learning_rate": 1.413793103448276e-07,
|
|
"num_tokens": 1998184.0,
|
|
"completions/mean_length": 432.75,
|
|
"completions/min_length": 228.0,
|
|
"completions/max_length": 916.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 273.3333435058594,
|
|
"completions/min_terminated_length": 228.0,
|
|
"completions/max_terminated_length": 337.0,
|
|
"tools/call_frequency": 10.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004023532383143902,
|
|
"sampling/sampling_logp_difference/max": 0.5054259896278381,
|
|
"sampling/importance_sampling_ratio/min": 0.41768375039100647,
|
|
"sampling/importance_sampling_ratio/mean": 0.9501832723617554,
|
|
"sampling/importance_sampling_ratio/max": 1.9075477123260498,
|
|
"kl": 0.015464820477063768,
|
|
"entropy": 0.06690557627007365,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.182410040870309,
|
|
"epoch": 0.005703125,
|
|
"step": 219
|
|
},
|
|
{
|
|
"loss": 0.5063859224319458,
|
|
"grad_norm": 7.252478122711182,
|
|
"learning_rate": 1.396551724137931e-07,
|
|
"num_tokens": 2006538.0,
|
|
"completions/mean_length": 358.5,
|
|
"completions/min_length": 211.0,
|
|
"completions/max_length": 909.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 279.8571472167969,
|
|
"completions/min_terminated_length": 211.0,
|
|
"completions/max_terminated_length": 351.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006370892748236656,
|
|
"sampling/sampling_logp_difference/max": 0.9805175065994263,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.7893059253692627,
|
|
"sampling/importance_sampling_ratio/max": 1.8430289030075073,
|
|
"kl": 0.013167917204555124,
|
|
"entropy": 0.08507121866568923,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.664595870301127,
|
|
"epoch": 0.005729166666666666,
|
|
"step": 220
|
|
},
|
|
{
|
|
"loss": 0.522027850151062,
|
|
"grad_norm": 2.863616704940796,
|
|
"learning_rate": 1.379310344827586e-07,
|
|
"num_tokens": 2016124.0,
|
|
"completions/mean_length": 512.5,
|
|
"completions/min_length": 228.0,
|
|
"completions/max_length": 888.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 391.16668701171875,
|
|
"completions/min_terminated_length": 228.0,
|
|
"completions/max_terminated_length": 870.0,
|
|
"tools/call_frequency": 13.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0035555653739720583,
|
|
"sampling/sampling_logp_difference/max": 1.4420175552368164,
|
|
"sampling/importance_sampling_ratio/min": 0.2985347509384155,
|
|
"sampling/importance_sampling_ratio/mean": 0.9967462420463562,
|
|
"sampling/importance_sampling_ratio/max": 1.8762089014053345,
|
|
"kl": 0.008223454749895609,
|
|
"entropy": 0.05731790023855865,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.40049036219716,
|
|
"epoch": 0.0057552083333333335,
|
|
"step": 221
|
|
},
|
|
{
|
|
"loss": 0.2064030021429062,
|
|
"grad_norm": 4.2190985679626465,
|
|
"learning_rate": 1.3620689655172414e-07,
|
|
"num_tokens": 2025032.0,
|
|
"completions/mean_length": 428.25,
|
|
"completions/min_length": 246.0,
|
|
"completions/max_length": 865.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 282.8333435058594,
|
|
"completions/min_terminated_length": 246.0,
|
|
"completions/max_terminated_length": 324.0,
|
|
"tools/call_frequency": 11.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004900211468338966,
|
|
"sampling/sampling_logp_difference/max": 0.5969611406326294,
|
|
"sampling/importance_sampling_ratio/min": 0.22761015594005585,
|
|
"sampling/importance_sampling_ratio/mean": 0.9077639579772949,
|
|
"sampling/importance_sampling_ratio/max": 2.0088999271392822,
|
|
"kl": 0.011889342466020025,
|
|
"entropy": 0.06723553303163499,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.138604583218694,
|
|
"epoch": 0.00578125,
|
|
"step": 222
|
|
},
|
|
{
|
|
"loss": 0.48906293511390686,
|
|
"grad_norm": 3.660278797149658,
|
|
"learning_rate": 1.3448275862068965e-07,
|
|
"num_tokens": 2033843.0,
|
|
"completions/mean_length": 415.375,
|
|
"completions/min_length": 220.0,
|
|
"completions/max_length": 905.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 258.3333435058594,
|
|
"completions/min_terminated_length": 220.0,
|
|
"completions/max_terminated_length": 294.0,
|
|
"tools/call_frequency": 11.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004095318261533976,
|
|
"sampling/sampling_logp_difference/max": 0.5458716154098511,
|
|
"sampling/importance_sampling_ratio/min": 0.44007760286331177,
|
|
"sampling/importance_sampling_ratio/mean": 1.0135478973388672,
|
|
"sampling/importance_sampling_ratio/max": 1.7006627321243286,
|
|
"kl": 0.009895527967273665,
|
|
"entropy": 0.06212259654421359,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.07671763561666,
|
|
"epoch": 0.005807291666666666,
|
|
"step": 223
|
|
},
|
|
{
|
|
"loss": 0.13212215900421143,
|
|
"grad_norm": 2.812335252761841,
|
|
"learning_rate": 1.3275862068965515e-07,
|
|
"num_tokens": 2043343.0,
|
|
"completions/mean_length": 501.875,
|
|
"completions/min_length": 225.0,
|
|
"completions/max_length": 901.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 450.0000305175781,
|
|
"completions/min_terminated_length": 225.0,
|
|
"completions/max_terminated_length": 901.0,
|
|
"tools/call_frequency": 13.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0031335896346718073,
|
|
"sampling/sampling_logp_difference/max": 0.4039543867111206,
|
|
"sampling/importance_sampling_ratio/min": 0.4327608644962311,
|
|
"sampling/importance_sampling_ratio/mean": 0.9170093536376953,
|
|
"sampling/importance_sampling_ratio/max": 1.7664108276367188,
|
|
"kl": 0.005495975594840274,
|
|
"entropy": 0.05750249361153692,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.750075383111835,
|
|
"epoch": 0.005833333333333334,
|
|
"step": 224
|
|
},
|
|
{
|
|
"loss": 0.30575570464134216,
|
|
"grad_norm": 2.3510024547576904,
|
|
"learning_rate": 1.310344827586207e-07,
|
|
"num_tokens": 2053124.0,
|
|
"completions/mean_length": 537.0,
|
|
"completions/min_length": 228.0,
|
|
"completions/max_length": 907.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 324.0,
|
|
"completions/min_terminated_length": 228.0,
|
|
"completions/max_terminated_length": 467.0,
|
|
"tools/call_frequency": 13.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003807642962783575,
|
|
"sampling/sampling_logp_difference/max": 0.6556974649429321,
|
|
"sampling/importance_sampling_ratio/min": 0.1374845653772354,
|
|
"sampling/importance_sampling_ratio/mean": 0.7388471364974976,
|
|
"sampling/importance_sampling_ratio/max": 1.6192972660064697,
|
|
"kl": 0.007617256833327701,
|
|
"entropy": 0.06444896757602692,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.134143091738224,
|
|
"epoch": 0.005859375,
|
|
"step": 225
|
|
},
|
|
{
|
|
"loss": -0.040912434458732605,
|
|
"grad_norm": 3.1499762535095215,
|
|
"learning_rate": 1.293103448275862e-07,
|
|
"num_tokens": 2063879.0,
|
|
"completions/mean_length": 658.5,
|
|
"completions/min_length": 248.0,
|
|
"completions/max_length": 963.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 444.0,
|
|
"completions/min_terminated_length": 248.0,
|
|
"completions/max_terminated_length": 963.0,
|
|
"tools/call_frequency": 17.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.012817398644983768,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.012817397713661194,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0025558623019605875,
|
|
"sampling/sampling_logp_difference/max": 0.9761226177215576,
|
|
"sampling/importance_sampling_ratio/min": 0.1592540591955185,
|
|
"sampling/importance_sampling_ratio/mean": 1.0013396739959717,
|
|
"sampling/importance_sampling_ratio/max": 2.1301157474517822,
|
|
"kl": 0.0060506322110995825,
|
|
"entropy": 0.046147919551003724,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.241912120953202,
|
|
"epoch": 0.005885416666666666,
|
|
"step": 226
|
|
},
|
|
{
|
|
"loss": 0.2730395495891571,
|
|
"grad_norm": 2.8770666122436523,
|
|
"learning_rate": 1.2758620689655173e-07,
|
|
"num_tokens": 2072295.0,
|
|
"completions/mean_length": 366.0,
|
|
"completions/min_length": 256.0,
|
|
"completions/max_length": 910.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 288.2857360839844,
|
|
"completions/min_terminated_length": 256.0,
|
|
"completions/max_terminated_length": 350.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006704138591885567,
|
|
"sampling/sampling_logp_difference/max": 0.7818012237548828,
|
|
"sampling/importance_sampling_ratio/min": 0.1917857974767685,
|
|
"sampling/importance_sampling_ratio/mean": 0.6862891912460327,
|
|
"sampling/importance_sampling_ratio/max": 1.4552291631698608,
|
|
"kl": 0.007856475676817354,
|
|
"entropy": 0.08110286644659936,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.04762289300561,
|
|
"epoch": 0.005911458333333334,
|
|
"step": 227
|
|
},
|
|
{
|
|
"loss": 0.4151316285133362,
|
|
"grad_norm": 1.7982980012893677,
|
|
"learning_rate": 1.2586206896551724e-07,
|
|
"num_tokens": 2082680.0,
|
|
"completions/mean_length": 611.875,
|
|
"completions/min_length": 251.0,
|
|
"completions/max_length": 1006.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 566.0,
|
|
"completions/min_terminated_length": 251.0,
|
|
"completions/max_terminated_length": 1006.0,
|
|
"tools/call_frequency": 14.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003361513838171959,
|
|
"sampling/sampling_logp_difference/max": 0.7221490144729614,
|
|
"sampling/importance_sampling_ratio/min": 0.2211952656507492,
|
|
"sampling/importance_sampling_ratio/mean": 0.8970264196395874,
|
|
"sampling/importance_sampling_ratio/max": 1.4118191003799438,
|
|
"kl": 0.006087072626542067,
|
|
"entropy": 0.062064238358289,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.644388776272535,
|
|
"epoch": 0.0059375,
|
|
"step": 228
|
|
},
|
|
{
|
|
"loss": 0.4544917047023773,
|
|
"grad_norm": 3.5075223445892334,
|
|
"learning_rate": 1.2413793103448275e-07,
|
|
"num_tokens": 2092511.0,
|
|
"completions/mean_length": 542.75,
|
|
"completions/min_length": 252.0,
|
|
"completions/max_length": 985.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 398.3333435058594,
|
|
"completions/min_terminated_length": 252.0,
|
|
"completions/max_terminated_length": 764.0,
|
|
"tools/call_frequency": 13.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004384452477097511,
|
|
"sampling/sampling_logp_difference/max": 0.6165733337402344,
|
|
"sampling/importance_sampling_ratio/min": 0.3656148612499237,
|
|
"sampling/importance_sampling_ratio/mean": 1.0641149282455444,
|
|
"sampling/importance_sampling_ratio/max": 2.1429593563079834,
|
|
"kl": 0.004549082615994848,
|
|
"entropy": 0.06837549060583115,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.185580670833588,
|
|
"epoch": 0.0059635416666666665,
|
|
"step": 229
|
|
},
|
|
{
|
|
"loss": 0.5729131698608398,
|
|
"grad_norm": 3.7511894702911377,
|
|
"learning_rate": 1.2241379310344826e-07,
|
|
"num_tokens": 2102682.0,
|
|
"completions/mean_length": 584.75,
|
|
"completions/min_length": 231.0,
|
|
"completions/max_length": 921.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 279.75,
|
|
"completions/min_terminated_length": 231.0,
|
|
"completions/max_terminated_length": 336.0,
|
|
"tools/call_frequency": 15.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0025345429312437773,
|
|
"sampling/sampling_logp_difference/max": 0.5872063040733337,
|
|
"sampling/importance_sampling_ratio/min": 0.4276439845561981,
|
|
"sampling/importance_sampling_ratio/mean": 0.9848134517669678,
|
|
"sampling/importance_sampling_ratio/max": 2.8137013912200928,
|
|
"kl": 0.006965349082747707,
|
|
"entropy": 0.04455703974235803,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.643966412171721,
|
|
"epoch": 0.005989583333333334,
|
|
"step": 230
|
|
},
|
|
{
|
|
"loss": 0.7138606309890747,
|
|
"grad_norm": 4.3516154289245605,
|
|
"learning_rate": 1.206896551724138e-07,
|
|
"num_tokens": 2111225.0,
|
|
"completions/mean_length": 383.125,
|
|
"completions/min_length": 234.0,
|
|
"completions/max_length": 883.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 311.71429443359375,
|
|
"completions/min_terminated_length": 234.0,
|
|
"completions/max_terminated_length": 385.0,
|
|
"tools/call_frequency": 9.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.014999999664723873,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.014999999664723873,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0064496430568397045,
|
|
"sampling/sampling_logp_difference/max": 0.8236088156700134,
|
|
"sampling/importance_sampling_ratio/min": 0.23686307668685913,
|
|
"sampling/importance_sampling_ratio/mean": 0.9143799543380737,
|
|
"sampling/importance_sampling_ratio/max": 1.8767057657241821,
|
|
"kl": 0.015334301198890898,
|
|
"entropy": 0.07671961770392954,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.216418443247676,
|
|
"epoch": 0.006015625,
|
|
"step": 231
|
|
},
|
|
{
|
|
"loss": 0.2530985176563263,
|
|
"grad_norm": 2.527024745941162,
|
|
"learning_rate": 1.1896551724137931e-07,
|
|
"num_tokens": 2119839.0,
|
|
"completions/mean_length": 391.25,
|
|
"completions/min_length": 14.0,
|
|
"completions/max_length": 888.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 324.4285888671875,
|
|
"completions/min_terminated_length": 14.0,
|
|
"completions/max_terminated_length": 888.0,
|
|
"tools/call_frequency": 10.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003528482746332884,
|
|
"sampling/sampling_logp_difference/max": 0.748659610748291,
|
|
"sampling/importance_sampling_ratio/min": 0.4568804204463959,
|
|
"sampling/importance_sampling_ratio/mean": 0.789121150970459,
|
|
"sampling/importance_sampling_ratio/max": 1.0676907300949097,
|
|
"kl": 0.014444019525399199,
|
|
"entropy": 0.061192034278064966,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.373705960810184,
|
|
"epoch": 0.0060416666666666665,
|
|
"step": 232
|
|
},
|
|
{
|
|
"loss": 0.08381801098585129,
|
|
"grad_norm": 5.8883185386657715,
|
|
"learning_rate": 1.1724137931034482e-07,
|
|
"num_tokens": 2127791.0,
|
|
"completions/mean_length": 308.5,
|
|
"completions/min_length": 232.0,
|
|
"completions/max_length": 391.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 308.5,
|
|
"completions/min_terminated_length": 232.0,
|
|
"completions/max_terminated_length": 391.0,
|
|
"tools/call_frequency": 7.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.007460989989340305,
|
|
"sampling/sampling_logp_difference/max": 0.901742696762085,
|
|
"sampling/importance_sampling_ratio/min": 0.5176633596420288,
|
|
"sampling/importance_sampling_ratio/mean": 1.071728229522705,
|
|
"sampling/importance_sampling_ratio/max": 2.1707139015197754,
|
|
"kl": 0.011356201925082132,
|
|
"entropy": 0.09658351447433233,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.816479025408626,
|
|
"epoch": 0.006067708333333333,
|
|
"step": 233
|
|
},
|
|
{
|
|
"loss": 0.4069221615791321,
|
|
"grad_norm": 7.694828510284424,
|
|
"learning_rate": 1.1551724137931033e-07,
|
|
"num_tokens": 2135377.0,
|
|
"completions/mean_length": 262.75,
|
|
"completions/min_length": 136.0,
|
|
"completions/max_length": 354.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 262.75,
|
|
"completions/min_terminated_length": 136.0,
|
|
"completions/max_terminated_length": 354.0,
|
|
"tools/call_frequency": 7.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.016249999403953552,
|
|
"rewards/reward_func/std": 0.01685018092393875,
|
|
"reward": 0.016249999403953552,
|
|
"reward_std": 0.01685018092393875,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0074232821352779865,
|
|
"sampling/sampling_logp_difference/max": 0.4771212339401245,
|
|
"sampling/importance_sampling_ratio/min": 0.5575864315032959,
|
|
"sampling/importance_sampling_ratio/mean": 1.113311767578125,
|
|
"sampling/importance_sampling_ratio/max": 1.8703396320343018,
|
|
"kl": 0.012179076002212241,
|
|
"entropy": 0.08267823932692409,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 7.0029825158417225,
|
|
"epoch": 0.00609375,
|
|
"step": 234
|
|
},
|
|
{
|
|
"loss": 0.12486986815929413,
|
|
"grad_norm": 1.6375352144241333,
|
|
"learning_rate": 1.1379310344827586e-07,
|
|
"num_tokens": 2145762.0,
|
|
"completions/mean_length": 612.625,
|
|
"completions/min_length": 285.0,
|
|
"completions/max_length": 912.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 449.0,
|
|
"completions/min_terminated_length": 285.0,
|
|
"completions/max_terminated_length": 911.0,
|
|
"tools/call_frequency": 15.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0038559294771403074,
|
|
"sampling/sampling_logp_difference/max": 0.7733533382415771,
|
|
"sampling/importance_sampling_ratio/min": 0.2591477036476135,
|
|
"sampling/importance_sampling_ratio/mean": 0.5573127865791321,
|
|
"sampling/importance_sampling_ratio/max": 0.9797956943511963,
|
|
"kl": 0.003839784400042845,
|
|
"entropy": 0.07239836419466883,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.27153348736465,
|
|
"epoch": 0.006119791666666667,
|
|
"step": 235
|
|
},
|
|
{
|
|
"loss": 0.6225030422210693,
|
|
"grad_norm": 4.492090225219727,
|
|
"learning_rate": 1.1206896551724137e-07,
|
|
"num_tokens": 2154709.0,
|
|
"completions/mean_length": 432.75,
|
|
"completions/min_length": 241.0,
|
|
"completions/max_length": 898.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 281.0,
|
|
"completions/min_terminated_length": 241.0,
|
|
"completions/max_terminated_length": 317.0,
|
|
"tools/call_frequency": 11.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00492401747033,
|
|
"sampling/sampling_logp_difference/max": 0.4237682819366455,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 1.0076963901519775,
|
|
"sampling/importance_sampling_ratio/max": 2.0787644386291504,
|
|
"kl": 0.010645602436852641,
|
|
"entropy": 0.08664517686702311,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.41959990747273,
|
|
"epoch": 0.006145833333333333,
|
|
"step": 236
|
|
},
|
|
{
|
|
"loss": 0.2281426191329956,
|
|
"grad_norm": 31.22895622253418,
|
|
"learning_rate": 1.103448275862069e-07,
|
|
"num_tokens": 2163695.0,
|
|
"completions/mean_length": 437.5,
|
|
"completions/min_length": 223.0,
|
|
"completions/max_length": 900.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 289.16668701171875,
|
|
"completions/min_terminated_length": 223.0,
|
|
"completions/max_terminated_length": 360.0,
|
|
"tools/call_frequency": 11.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004786212462931871,
|
|
"sampling/sampling_logp_difference/max": 1.5236012935638428,
|
|
"sampling/importance_sampling_ratio/min": 0.2738311290740967,
|
|
"sampling/importance_sampling_ratio/mean": 0.9567095041275024,
|
|
"sampling/importance_sampling_ratio/max": 1.8892822265625,
|
|
"kl": 0.044213560431671795,
|
|
"entropy": 0.06694216525647789,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.619100231677294,
|
|
"epoch": 0.006171875,
|
|
"step": 237
|
|
},
|
|
{
|
|
"loss": 0.19045661389827728,
|
|
"grad_norm": 5.344757556915283,
|
|
"learning_rate": 1.086206896551724e-07,
|
|
"num_tokens": 2171923.0,
|
|
"completions/mean_length": 343.5,
|
|
"completions/min_length": 228.0,
|
|
"completions/max_length": 881.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 343.5,
|
|
"completions/min_terminated_length": 228.0,
|
|
"completions/max_terminated_length": 881.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005176930222660303,
|
|
"sampling/sampling_logp_difference/max": 0.33557987213134766,
|
|
"sampling/importance_sampling_ratio/min": 0.5333213210105896,
|
|
"sampling/importance_sampling_ratio/mean": 0.9724828004837036,
|
|
"sampling/importance_sampling_ratio/max": 2.0568439960479736,
|
|
"kl": 0.008732907132070977,
|
|
"entropy": 0.07178128557279706,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.572901349514723,
|
|
"epoch": 0.006197916666666667,
|
|
"step": 238
|
|
},
|
|
{
|
|
"loss": 0.2778877019882202,
|
|
"grad_norm": 6.737323760986328,
|
|
"learning_rate": 1.0689655172413794e-07,
|
|
"num_tokens": 2179780.0,
|
|
"completions/mean_length": 296.125,
|
|
"completions/min_length": 236.0,
|
|
"completions/max_length": 459.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 296.125,
|
|
"completions/min_terminated_length": 236.0,
|
|
"completions/max_terminated_length": 459.0,
|
|
"tools/call_frequency": 7.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006709440145641565,
|
|
"sampling/sampling_logp_difference/max": 0.40797853469848633,
|
|
"sampling/importance_sampling_ratio/min": 0.4145645201206207,
|
|
"sampling/importance_sampling_ratio/mean": 0.9115852117538452,
|
|
"sampling/importance_sampling_ratio/max": 1.8955971002578735,
|
|
"kl": 0.020956200918590184,
|
|
"entropy": 0.07279529003426433,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 7.016493774950504,
|
|
"epoch": 0.006223958333333333,
|
|
"step": 239
|
|
},
|
|
{
|
|
"loss": 0.46434324979782104,
|
|
"grad_norm": 3.690675735473633,
|
|
"learning_rate": 1.0517241379310344e-07,
|
|
"num_tokens": 2188158.0,
|
|
"completions/mean_length": 361.0,
|
|
"completions/min_length": 226.0,
|
|
"completions/max_length": 917.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 361.0,
|
|
"completions/min_terminated_length": 226.0,
|
|
"completions/max_terminated_length": 917.0,
|
|
"tools/call_frequency": 9.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.017500000074505806,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.017500000074505806,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005031631328165531,
|
|
"sampling/sampling_logp_difference/max": 0.5336201786994934,
|
|
"sampling/importance_sampling_ratio/min": 0.4367794096469879,
|
|
"sampling/importance_sampling_ratio/mean": 0.8346672058105469,
|
|
"sampling/importance_sampling_ratio/max": 2.0089845657348633,
|
|
"kl": 0.01607281963515561,
|
|
"entropy": 0.070617281133309,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.543106522411108,
|
|
"epoch": 0.00625,
|
|
"step": 240
|
|
},
|
|
{
|
|
"loss": 0.021518878638744354,
|
|
"grad_norm": 6.071545124053955,
|
|
"learning_rate": 1.0344827586206897e-07,
|
|
"num_tokens": 2196638.0,
|
|
"completions/mean_length": 374.125,
|
|
"completions/min_length": 230.0,
|
|
"completions/max_length": 1012.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 283.0,
|
|
"completions/min_terminated_length": 230.0,
|
|
"completions/max_terminated_length": 308.0,
|
|
"tools/call_frequency": 8.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005964357405900955,
|
|
"sampling/sampling_logp_difference/max": 0.6650679111480713,
|
|
"sampling/importance_sampling_ratio/min": 0.2610292136669159,
|
|
"sampling/importance_sampling_ratio/mean": 1.1714744567871094,
|
|
"sampling/importance_sampling_ratio/max": 2.400663137435913,
|
|
"kl": 0.011474874594568973,
|
|
"entropy": 0.08578880922868848,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.897009413689375,
|
|
"epoch": 0.006276041666666667,
|
|
"step": 241
|
|
},
|
|
{
|
|
"loss": 0.35851266980171204,
|
|
"grad_norm": 2.9004931449890137,
|
|
"learning_rate": 1.0172413793103447e-07,
|
|
"num_tokens": 2206207.0,
|
|
"completions/mean_length": 509.75,
|
|
"completions/min_length": 251.0,
|
|
"completions/max_length": 932.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 373.8333435058594,
|
|
"completions/min_terminated_length": 251.0,
|
|
"completions/max_terminated_length": 900.0,
|
|
"tools/call_frequency": 12.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003657017834484577,
|
|
"sampling/sampling_logp_difference/max": 0.5031510591506958,
|
|
"sampling/importance_sampling_ratio/min": 0.7240630388259888,
|
|
"sampling/importance_sampling_ratio/mean": 1.02143394947052,
|
|
"sampling/importance_sampling_ratio/max": 1.3293250799179077,
|
|
"kl": 0.006794752807763871,
|
|
"entropy": 0.06060198391787708,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.663152404129505,
|
|
"epoch": 0.006302083333333333,
|
|
"step": 242
|
|
},
|
|
{
|
|
"loss": 0.22505345940589905,
|
|
"grad_norm": 2.789543628692627,
|
|
"learning_rate": 1e-07,
|
|
"num_tokens": 2215243.0,
|
|
"completions/mean_length": 443.75,
|
|
"completions/min_length": 222.0,
|
|
"completions/max_length": 967.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 279.8333435058594,
|
|
"completions/min_terminated_length": 222.0,
|
|
"completions/max_terminated_length": 335.0,
|
|
"tools/call_frequency": 11.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003608885919675231,
|
|
"sampling/sampling_logp_difference/max": 0.4483809471130371,
|
|
"sampling/importance_sampling_ratio/min": 0.3409290909767151,
|
|
"sampling/importance_sampling_ratio/mean": 0.8782253861427307,
|
|
"sampling/importance_sampling_ratio/max": 1.6534655094146729,
|
|
"kl": 0.0067834420333383605,
|
|
"entropy": 0.0630563497543335,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.94569875486195,
|
|
"epoch": 0.006328125,
|
|
"step": 243
|
|
},
|
|
{
|
|
"loss": 0.06894535571336746,
|
|
"grad_norm": 3.274223566055298,
|
|
"learning_rate": 9.827586206896552e-08,
|
|
"num_tokens": 2224724.0,
|
|
"completions/mean_length": 499.0,
|
|
"completions/min_length": 230.0,
|
|
"completions/max_length": 905.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 258.6000061035156,
|
|
"completions/min_terminated_length": 230.0,
|
|
"completions/max_terminated_length": 342.0,
|
|
"tools/call_frequency": 13.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.01414213515818119,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.01414213515818119,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0035293279215693474,
|
|
"sampling/sampling_logp_difference/max": 0.5808429718017578,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.8453118205070496,
|
|
"sampling/importance_sampling_ratio/max": 1.678948998451233,
|
|
"kl": 0.009609646389435511,
|
|
"entropy": 0.05383499653544277,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.233815779909492,
|
|
"epoch": 0.006354166666666667,
|
|
"step": 244
|
|
},
|
|
{
|
|
"loss": 0.23755919933319092,
|
|
"grad_norm": 3.3479530811309814,
|
|
"learning_rate": 9.655172413793103e-08,
|
|
"num_tokens": 2234785.0,
|
|
"completions/mean_length": 572.375,
|
|
"completions/min_length": 239.0,
|
|
"completions/max_length": 908.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 391.6000061035156,
|
|
"completions/min_terminated_length": 239.0,
|
|
"completions/max_terminated_length": 887.0,
|
|
"tools/call_frequency": 15.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003022867953404784,
|
|
"sampling/sampling_logp_difference/max": 0.4263339042663574,
|
|
"sampling/importance_sampling_ratio/min": 0.3625563979148865,
|
|
"sampling/importance_sampling_ratio/mean": 1.0036848783493042,
|
|
"sampling/importance_sampling_ratio/max": 1.3762973546981812,
|
|
"kl": 0.00560847119322716,
|
|
"entropy": 0.05764547025319189,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.687129236757755,
|
|
"epoch": 0.006380208333333333,
|
|
"step": 245
|
|
},
|
|
{
|
|
"loss": 0.1123998835682869,
|
|
"grad_norm": 2.6210505962371826,
|
|
"learning_rate": 9.482758620689655e-08,
|
|
"num_tokens": 2244309.0,
|
|
"completions/mean_length": 504.75,
|
|
"completions/min_length": 219.0,
|
|
"completions/max_length": 967.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 265.8000183105469,
|
|
"completions/min_terminated_length": 219.0,
|
|
"completions/max_terminated_length": 332.0,
|
|
"tools/call_frequency": 13.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0025708149187266827,
|
|
"sampling/sampling_logp_difference/max": 0.8236080408096313,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.8182451128959656,
|
|
"sampling/importance_sampling_ratio/max": 1.6612991094589233,
|
|
"kl": 0.009890891281429504,
|
|
"entropy": 0.04406958329491317,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.179574264213443,
|
|
"epoch": 0.00640625,
|
|
"step": 246
|
|
},
|
|
{
|
|
"loss": 0.3937268555164337,
|
|
"grad_norm": 2.7107176780700684,
|
|
"learning_rate": 9.310344827586207e-08,
|
|
"num_tokens": 2252498.0,
|
|
"completions/mean_length": 337.5,
|
|
"completions/min_length": 223.0,
|
|
"completions/max_length": 868.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 261.71429443359375,
|
|
"completions/min_terminated_length": 223.0,
|
|
"completions/max_terminated_length": 349.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.014999999664723873,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.014999999664723873,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004549999721348286,
|
|
"sampling/sampling_logp_difference/max": 0.7026820182800293,
|
|
"sampling/importance_sampling_ratio/min": 0.3919149935245514,
|
|
"sampling/importance_sampling_ratio/mean": 0.8184147477149963,
|
|
"sampling/importance_sampling_ratio/max": 1.300963282585144,
|
|
"kl": 0.0122263476655462,
|
|
"entropy": 0.06679779017576948,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.024746986106038,
|
|
"epoch": 0.006432291666666667,
|
|
"step": 247
|
|
},
|
|
{
|
|
"loss": -0.0668078064918518,
|
|
"grad_norm": 3.360973834991455,
|
|
"learning_rate": 9.137931034482759e-08,
|
|
"num_tokens": 2261740.0,
|
|
"completions/mean_length": 469.5,
|
|
"completions/min_length": 233.0,
|
|
"completions/max_length": 1058.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 298.0,
|
|
"completions/min_terminated_length": 233.0,
|
|
"completions/max_terminated_length": 352.0,
|
|
"tools/call_frequency": 10.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": -0.0062500000931322575,
|
|
"rewards/reward_func/std": 0.020658792927861214,
|
|
"reward": -0.0062500000931322575,
|
|
"reward_std": 0.020658792927861214,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005611687898635864,
|
|
"sampling/sampling_logp_difference/max": 0.5652132034301758,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.569216251373291,
|
|
"sampling/importance_sampling_ratio/max": 1.4573065042495728,
|
|
"kl": 0.009170733541395748,
|
|
"entropy": 0.07505051954649389,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 15.318448189646006,
|
|
"epoch": 0.006458333333333333,
|
|
"step": 248
|
|
},
|
|
{
|
|
"loss": 0.25679558515548706,
|
|
"grad_norm": 3.0161712169647217,
|
|
"learning_rate": 8.96551724137931e-08,
|
|
"num_tokens": 2271434.0,
|
|
"completions/mean_length": 526.25,
|
|
"completions/min_length": 226.0,
|
|
"completions/max_length": 1005.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 465.4285888671875,
|
|
"completions/min_terminated_length": 226.0,
|
|
"completions/max_terminated_length": 1005.0,
|
|
"tools/call_frequency": 12.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.01414213515818119,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.01414213515818119,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0031429186929017305,
|
|
"sampling/sampling_logp_difference/max": 1.1103920936584473,
|
|
"sampling/importance_sampling_ratio/min": 0.11957414448261261,
|
|
"sampling/importance_sampling_ratio/mean": 0.8082802891731262,
|
|
"sampling/importance_sampling_ratio/max": 1.6786141395568848,
|
|
"kl": 0.012878221445134841,
|
|
"entropy": 0.05151165463030338,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.512800469994545,
|
|
"epoch": 0.006484375,
|
|
"step": 249
|
|
},
|
|
{
|
|
"loss": 0.024655239656567574,
|
|
"grad_norm": 1.2700872421264648,
|
|
"learning_rate": 8.79310344827586e-08,
|
|
"num_tokens": 2281054.0,
|
|
"completions/mean_length": 516.5,
|
|
"completions/min_length": 237.0,
|
|
"completions/max_length": 947.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 464.2857360839844,
|
|
"completions/min_terminated_length": 237.0,
|
|
"completions/max_terminated_length": 947.0,
|
|
"tools/call_frequency": 13.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.0,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0035700220614671707,
|
|
"sampling/sampling_logp_difference/max": 0.7709542512893677,
|
|
"sampling/importance_sampling_ratio/min": 0.2625862658023834,
|
|
"sampling/importance_sampling_ratio/mean": 0.6785949468612671,
|
|
"sampling/importance_sampling_ratio/max": 1.040136694908142,
|
|
"kl": 0.005439797446342709,
|
|
"entropy": 0.048666520160622895,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.586364854127169,
|
|
"epoch": 0.006510416666666667,
|
|
"step": 250
|
|
},
|
|
{
|
|
"loss": 0.39852964878082275,
|
|
"grad_norm": 2.411869764328003,
|
|
"learning_rate": 8.620689655172414e-08,
|
|
"num_tokens": 2290618.0,
|
|
"completions/mean_length": 509.375,
|
|
"completions/min_length": 227.0,
|
|
"completions/max_length": 958.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 365.8333435058594,
|
|
"completions/min_terminated_length": 227.0,
|
|
"completions/max_terminated_length": 945.0,
|
|
"tools/call_frequency": 12.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0024800822138786316,
|
|
"sampling/sampling_logp_difference/max": 0.4458940029144287,
|
|
"sampling/importance_sampling_ratio/min": 0.4365721642971039,
|
|
"sampling/importance_sampling_ratio/mean": 0.9041198492050171,
|
|
"sampling/importance_sampling_ratio/max": 1.1539802551269531,
|
|
"kl": 0.00905564020285965,
|
|
"entropy": 0.05024857854004949,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.516810789704323,
|
|
"epoch": 0.006536458333333333,
|
|
"step": 251
|
|
},
|
|
{
|
|
"loss": 0.48702356219291687,
|
|
"grad_norm": 2.947753667831421,
|
|
"learning_rate": 8.448275862068965e-08,
|
|
"num_tokens": 2300156.0,
|
|
"completions/mean_length": 506.375,
|
|
"completions/min_length": 223.0,
|
|
"completions/max_length": 904.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 268.3999938964844,
|
|
"completions/min_terminated_length": 223.0,
|
|
"completions/max_terminated_length": 320.0,
|
|
"tools/call_frequency": 13.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0035441291984170675,
|
|
"sampling/sampling_logp_difference/max": 0.9285266399383545,
|
|
"sampling/importance_sampling_ratio/min": 0.21628588438034058,
|
|
"sampling/importance_sampling_ratio/mean": 0.7983152270317078,
|
|
"sampling/importance_sampling_ratio/max": 1.5297478437423706,
|
|
"kl": 0.01673757631215267,
|
|
"entropy": 0.058467789785936475,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.156265430152416,
|
|
"epoch": 0.0065625,
|
|
"step": 252
|
|
},
|
|
{
|
|
"loss": 0.3011172413825989,
|
|
"grad_norm": 2.1026012897491455,
|
|
"learning_rate": 8.275862068965517e-08,
|
|
"num_tokens": 2309919.0,
|
|
"completions/mean_length": 534.625,
|
|
"completions/min_length": 229.0,
|
|
"completions/max_length": 975.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 306.0,
|
|
"completions/min_terminated_length": 229.0,
|
|
"completions/max_terminated_length": 384.0,
|
|
"tools/call_frequency": 13.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0028093173168599606,
|
|
"sampling/sampling_logp_difference/max": 0.37752997875213623,
|
|
"sampling/importance_sampling_ratio/min": 0.6290064454078674,
|
|
"sampling/importance_sampling_ratio/mean": 1.0789461135864258,
|
|
"sampling/importance_sampling_ratio/max": 1.5364922285079956,
|
|
"kl": 0.012943116746100713,
|
|
"entropy": 0.04965827940031886,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.358246324583888,
|
|
"epoch": 0.006588541666666667,
|
|
"step": 253
|
|
},
|
|
{
|
|
"loss": 0.20058459043502808,
|
|
"grad_norm": 2.1985929012298584,
|
|
"learning_rate": 8.103448275862068e-08,
|
|
"num_tokens": 2319563.0,
|
|
"completions/mean_length": 519.125,
|
|
"completions/min_length": 245.0,
|
|
"completions/max_length": 919.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 391.5,
|
|
"completions/min_terminated_length": 245.0,
|
|
"completions/max_terminated_length": 882.0,
|
|
"tools/call_frequency": 13.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.011250000447034836,
|
|
"rewards/reward_func/std": 0.018077217042446136,
|
|
"reward": 0.011250000447034836,
|
|
"reward_std": 0.018077215179800987,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0033701013308018446,
|
|
"sampling/sampling_logp_difference/max": 0.42209482192993164,
|
|
"sampling/importance_sampling_ratio/min": 0.19347310066223145,
|
|
"sampling/importance_sampling_ratio/mean": 0.7575610876083374,
|
|
"sampling/importance_sampling_ratio/max": 1.4560166597366333,
|
|
"kl": 0.006214593318873085,
|
|
"entropy": 0.060175721533596516,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.539172381162643,
|
|
"epoch": 0.006614583333333333,
|
|
"step": 254
|
|
},
|
|
{
|
|
"loss": -0.16774767637252808,
|
|
"grad_norm": 2.9720096588134766,
|
|
"learning_rate": 7.931034482758621e-08,
|
|
"num_tokens": 2327247.0,
|
|
"completions/mean_length": 274.5,
|
|
"completions/min_length": 246.0,
|
|
"completions/max_length": 315.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 274.5,
|
|
"completions/min_terminated_length": 246.0,
|
|
"completions/max_terminated_length": 315.0,
|
|
"tools/call_frequency": 7.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.008589494973421097,
|
|
"sampling/sampling_logp_difference/max": 1.2420251369476318,
|
|
"sampling/importance_sampling_ratio/min": 0.15825656056404114,
|
|
"sampling/importance_sampling_ratio/mean": 0.6277849674224854,
|
|
"sampling/importance_sampling_ratio/max": 1.0720003843307495,
|
|
"kl": 0.011047484556911513,
|
|
"entropy": 0.08509592898190022,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.190464340150356,
|
|
"epoch": 0.006640625,
|
|
"step": 255
|
|
},
|
|
{
|
|
"loss": 0.678689181804657,
|
|
"grad_norm": 2.162055015563965,
|
|
"learning_rate": 7.758620689655172e-08,
|
|
"num_tokens": 2338212.0,
|
|
"completions/mean_length": 684.75,
|
|
"completions/min_length": 219.0,
|
|
"completions/max_length": 999.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 426.0,
|
|
"completions/min_terminated_length": 219.0,
|
|
"completions/max_terminated_length": 904.0,
|
|
"tools/call_frequency": 16.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0024466628674417734,
|
|
"sampling/sampling_logp_difference/max": 0.603243350982666,
|
|
"sampling/importance_sampling_ratio/min": 0.45637086033821106,
|
|
"sampling/importance_sampling_ratio/mean": 1.036811113357544,
|
|
"sampling/importance_sampling_ratio/max": 1.8374240398406982,
|
|
"kl": 0.006152577276225202,
|
|
"entropy": 0.03899537120014429,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.21735617145896,
|
|
"epoch": 0.006666666666666667,
|
|
"step": 256
|
|
},
|
|
{
|
|
"loss": 0.02569650672376156,
|
|
"grad_norm": 1.383256435394287,
|
|
"learning_rate": 7.586206896551724e-08,
|
|
"num_tokens": 2348714.0,
|
|
"completions/mean_length": 627.125,
|
|
"completions/min_length": 263.0,
|
|
"completions/max_length": 987.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 454.20001220703125,
|
|
"completions/min_terminated_length": 263.0,
|
|
"completions/max_terminated_length": 935.0,
|
|
"tools/call_frequency": 14.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.0,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003685368224978447,
|
|
"sampling/sampling_logp_difference/max": 0.6686651706695557,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.7703503370285034,
|
|
"sampling/importance_sampling_ratio/max": 2.2623162269592285,
|
|
"kl": 0.005646704201353714,
|
|
"entropy": 0.059528822312131524,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.834392301738262,
|
|
"epoch": 0.0066927083333333335,
|
|
"step": 257
|
|
},
|
|
{
|
|
"loss": 0.13307367265224457,
|
|
"grad_norm": 2.7584047317504883,
|
|
"learning_rate": 7.413793103448275e-08,
|
|
"num_tokens": 2357131.0,
|
|
"completions/mean_length": 366.75,
|
|
"completions/min_length": 234.0,
|
|
"completions/max_length": 915.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 288.4285888671875,
|
|
"completions/min_terminated_length": 234.0,
|
|
"completions/max_terminated_length": 390.0,
|
|
"tools/call_frequency": 9.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.014880476519465446,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.014880475588142872,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0042374273762106895,
|
|
"sampling/sampling_logp_difference/max": 0.43998217582702637,
|
|
"sampling/importance_sampling_ratio/min": 0.36689502000808716,
|
|
"sampling/importance_sampling_ratio/mean": 0.9189438819885254,
|
|
"sampling/importance_sampling_ratio/max": 1.5682750940322876,
|
|
"kl": 0.00996059033968777,
|
|
"entropy": 0.05985496856737882,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.818406570702791,
|
|
"epoch": 0.00671875,
|
|
"step": 258
|
|
},
|
|
{
|
|
"loss": 0.32452595233917236,
|
|
"grad_norm": 2.1824963092803955,
|
|
"learning_rate": 7.241379310344827e-08,
|
|
"num_tokens": 2366125.0,
|
|
"completions/mean_length": 438.25,
|
|
"completions/min_length": 227.0,
|
|
"completions/max_length": 992.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 366.8571472167969,
|
|
"completions/min_terminated_length": 227.0,
|
|
"completions/max_terminated_length": 992.0,
|
|
"tools/call_frequency": 10.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0035509171430021524,
|
|
"sampling/sampling_logp_difference/max": 0.4454996585845947,
|
|
"sampling/importance_sampling_ratio/min": 0.31019100546836853,
|
|
"sampling/importance_sampling_ratio/mean": 0.8306182622909546,
|
|
"sampling/importance_sampling_ratio/max": 1.7142812013626099,
|
|
"kl": 0.005876144152352936,
|
|
"entropy": 0.05782180209644139,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.770527290180326,
|
|
"epoch": 0.006744791666666666,
|
|
"step": 259
|
|
},
|
|
{
|
|
"loss": -0.0690397173166275,
|
|
"grad_norm": 5.6424031257629395,
|
|
"learning_rate": 7.06896551724138e-08,
|
|
"num_tokens": 2375184.0,
|
|
"completions/mean_length": 446.75,
|
|
"completions/min_length": 250.0,
|
|
"completions/max_length": 1036.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 362.5714416503906,
|
|
"completions/min_terminated_length": 250.0,
|
|
"completions/max_terminated_length": 904.0,
|
|
"tools/call_frequency": 10.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.09385416656732559,
|
|
"rewards/reward_func/std": 0.23334479331970215,
|
|
"reward": 0.09385416656732559,
|
|
"reward_std": 0.23334479331970215,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0032503274269402027,
|
|
"sampling/sampling_logp_difference/max": 0.4519524574279785,
|
|
"sampling/importance_sampling_ratio/min": 0.4732486605644226,
|
|
"sampling/importance_sampling_ratio/mean": 0.9928018450737,
|
|
"sampling/importance_sampling_ratio/max": 1.6393862962722778,
|
|
"kl": 0.00979274905694183,
|
|
"entropy": 0.05829371698200703,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.69159053079784,
|
|
"epoch": 0.0067708333333333336,
|
|
"step": 260
|
|
},
|
|
{
|
|
"loss": 0.4674473702907562,
|
|
"grad_norm": 3.0789122581481934,
|
|
"learning_rate": 6.89655172413793e-08,
|
|
"num_tokens": 2385294.0,
|
|
"completions/mean_length": 578.375,
|
|
"completions/min_length": 210.0,
|
|
"completions/max_length": 973.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 250.5,
|
|
"completions/min_terminated_length": 210.0,
|
|
"completions/max_terminated_length": 304.0,
|
|
"tools/call_frequency": 14.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.013749999925494194,
|
|
"rewards/reward_func/std": 0.01767767034471035,
|
|
"reward": 0.013749999925494194,
|
|
"reward_std": 0.0176776684820652,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0032551144249737263,
|
|
"sampling/sampling_logp_difference/max": 0.48508358001708984,
|
|
"sampling/importance_sampling_ratio/min": 0.2822646200656891,
|
|
"sampling/importance_sampling_ratio/mean": 1.0616158246994019,
|
|
"sampling/importance_sampling_ratio/max": 2.962817907333374,
|
|
"kl": 0.009575902957294602,
|
|
"entropy": 0.060567187960259616,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.325932217761874,
|
|
"epoch": 0.006796875,
|
|
"step": 261
|
|
},
|
|
{
|
|
"loss": 0.3925217390060425,
|
|
"grad_norm": 2.921971321105957,
|
|
"learning_rate": 6.724137931034482e-08,
|
|
"num_tokens": 2394880.0,
|
|
"completions/mean_length": 511.875,
|
|
"completions/min_length": 224.0,
|
|
"completions/max_length": 988.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 356.66668701171875,
|
|
"completions/min_terminated_length": 224.0,
|
|
"completions/max_terminated_length": 806.0,
|
|
"tools/call_frequency": 13.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003077454399317503,
|
|
"sampling/sampling_logp_difference/max": 0.8122477531433105,
|
|
"sampling/importance_sampling_ratio/min": 0.32636696100234985,
|
|
"sampling/importance_sampling_ratio/mean": 0.9069831371307373,
|
|
"sampling/importance_sampling_ratio/max": 2.403791904449463,
|
|
"kl": 0.007388941776298452,
|
|
"entropy": 0.04864441673271358,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.269655622541904,
|
|
"epoch": 0.006822916666666666,
|
|
"step": 262
|
|
},
|
|
{
|
|
"loss": 0.3128494322299957,
|
|
"grad_norm": 3.577718496322632,
|
|
"learning_rate": 6.551724137931034e-08,
|
|
"num_tokens": 2403912.0,
|
|
"completions/mean_length": 443.25,
|
|
"completions/min_length": 250.0,
|
|
"completions/max_length": 947.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 371.2857360839844,
|
|
"completions/min_terminated_length": 250.0,
|
|
"completions/max_terminated_length": 909.0,
|
|
"tools/call_frequency": 10.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.014880476519465446,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.014880475588142872,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00485264603048563,
|
|
"sampling/sampling_logp_difference/max": 0.7732473611831665,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.9245357513427734,
|
|
"sampling/importance_sampling_ratio/max": 2.9667270183563232,
|
|
"kl": 0.009359787838548073,
|
|
"entropy": 0.07874505454674363,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.381516905501485,
|
|
"epoch": 0.006848958333333334,
|
|
"step": 263
|
|
},
|
|
{
|
|
"loss": 0.4138055741786957,
|
|
"grad_norm": 3.5194525718688965,
|
|
"learning_rate": 6.379310344827587e-08,
|
|
"num_tokens": 2412924.0,
|
|
"completions/mean_length": 441.0,
|
|
"completions/min_length": 239.0,
|
|
"completions/max_length": 1009.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 273.0,
|
|
"completions/min_terminated_length": 239.0,
|
|
"completions/max_terminated_length": 300.0,
|
|
"tools/call_frequency": 10.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.014880476519465446,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.014880475588142872,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004156154580414295,
|
|
"sampling/sampling_logp_difference/max": 0.6608169674873352,
|
|
"sampling/importance_sampling_ratio/min": 0.31790050864219666,
|
|
"sampling/importance_sampling_ratio/mean": 0.9935593605041504,
|
|
"sampling/importance_sampling_ratio/max": 1.525246500968933,
|
|
"kl": 0.012059394881362095,
|
|
"entropy": 0.07312226248905063,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.104876002296805,
|
|
"epoch": 0.006875,
|
|
"step": 264
|
|
},
|
|
{
|
|
"loss": 0.1565098762512207,
|
|
"grad_norm": 3.6302406787872314,
|
|
"learning_rate": 6.206896551724137e-08,
|
|
"num_tokens": 2421900.0,
|
|
"completions/mean_length": 435.375,
|
|
"completions/min_length": 240.0,
|
|
"completions/max_length": 911.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 435.375,
|
|
"completions/min_terminated_length": 240.0,
|
|
"completions/max_terminated_length": 911.0,
|
|
"tools/call_frequency": 11.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004992013797163963,
|
|
"sampling/sampling_logp_difference/max": 0.46503114700317383,
|
|
"sampling/importance_sampling_ratio/min": 0.33158570528030396,
|
|
"sampling/importance_sampling_ratio/mean": 0.7800635695457458,
|
|
"sampling/importance_sampling_ratio/max": 1.6718090772628784,
|
|
"kl": 0.006607427443668712,
|
|
"entropy": 0.08226339193060994,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.492700284346938,
|
|
"epoch": 0.0069010416666666664,
|
|
"step": 265
|
|
},
|
|
{
|
|
"loss": 0.589852511882782,
|
|
"grad_norm": 2.495140314102173,
|
|
"learning_rate": 6.03448275862069e-08,
|
|
"num_tokens": 2431262.0,
|
|
"completions/mean_length": 485.375,
|
|
"completions/min_length": 229.0,
|
|
"completions/max_length": 910.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 350.8333435058594,
|
|
"completions/min_terminated_length": 229.0,
|
|
"completions/max_terminated_length": 883.0,
|
|
"tools/call_frequency": 13.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0026053327601403,
|
|
"sampling/sampling_logp_difference/max": 0.6101561784744263,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.9313764572143555,
|
|
"sampling/importance_sampling_ratio/max": 1.7153178453445435,
|
|
"kl": 0.009163391629726902,
|
|
"entropy": 0.055689023807644844,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.45149003341794,
|
|
"epoch": 0.006927083333333334,
|
|
"step": 266
|
|
},
|
|
{
|
|
"loss": 0.8367747664451599,
|
|
"grad_norm": 5.528039932250977,
|
|
"learning_rate": 5.862068965517241e-08,
|
|
"num_tokens": 2441103.0,
|
|
"completions/mean_length": 545.25,
|
|
"completions/min_length": 240.0,
|
|
"completions/max_length": 1062.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 299.20001220703125,
|
|
"completions/min_terminated_length": 240.0,
|
|
"completions/max_terminated_length": 392.0,
|
|
"tools/call_frequency": 12.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": -0.0012500002048909664,
|
|
"rewards/reward_func/std": 0.021671246737241745,
|
|
"reward": -0.0012500002048909664,
|
|
"reward_std": 0.021671246737241745,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004504072014242411,
|
|
"sampling/sampling_logp_difference/max": 0.43561410903930664,
|
|
"sampling/importance_sampling_ratio/min": 0.3998229503631592,
|
|
"sampling/importance_sampling_ratio/mean": 0.8630289435386658,
|
|
"sampling/importance_sampling_ratio/max": 1.7298011779785156,
|
|
"kl": 0.005085467862954829,
|
|
"entropy": 0.0693967689294368,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.293195264413953,
|
|
"epoch": 0.006953125,
|
|
"step": 267
|
|
},
|
|
{
|
|
"loss": 0.23700451850891113,
|
|
"grad_norm": 4.011788845062256,
|
|
"learning_rate": 5.689655172413793e-08,
|
|
"num_tokens": 2449592.0,
|
|
"completions/mean_length": 375.5,
|
|
"completions/min_length": 220.0,
|
|
"completions/max_length": 897.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 301.0,
|
|
"completions/min_terminated_length": 220.0,
|
|
"completions/max_terminated_length": 404.0,
|
|
"tools/call_frequency": 9.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006352824158966541,
|
|
"sampling/sampling_logp_difference/max": 1.225382924079895,
|
|
"sampling/importance_sampling_ratio/min": 0.3933752477169037,
|
|
"sampling/importance_sampling_ratio/mean": 0.7365532517433167,
|
|
"sampling/importance_sampling_ratio/max": 1.1142364740371704,
|
|
"kl": 0.011371326181688346,
|
|
"entropy": 0.08945641107857227,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.197782006114721,
|
|
"epoch": 0.0069791666666666665,
|
|
"step": 268
|
|
},
|
|
{
|
|
"loss": 0.07808616757392883,
|
|
"grad_norm": 3.0376551151275635,
|
|
"learning_rate": 5.517241379310345e-08,
|
|
"num_tokens": 2458719.0,
|
|
"completions/mean_length": 454.25,
|
|
"completions/min_length": 277.0,
|
|
"completions/max_length": 924.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 387.14288330078125,
|
|
"completions/min_terminated_length": 277.0,
|
|
"completions/max_terminated_length": 907.0,
|
|
"tools/call_frequency": 11.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005793869029730558,
|
|
"sampling/sampling_logp_difference/max": 1.2325255870819092,
|
|
"sampling/importance_sampling_ratio/min": 0.30067309737205505,
|
|
"sampling/importance_sampling_ratio/mean": 0.719024658203125,
|
|
"sampling/importance_sampling_ratio/max": 1.1819229125976562,
|
|
"kl": 0.0033787663414841518,
|
|
"entropy": 0.08568726177327335,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.479113213717937,
|
|
"epoch": 0.007005208333333333,
|
|
"step": 269
|
|
},
|
|
{
|
|
"loss": 0.0703253448009491,
|
|
"grad_norm": 4.206221580505371,
|
|
"learning_rate": 5.344827586206897e-08,
|
|
"num_tokens": 2467388.0,
|
|
"completions/mean_length": 398.125,
|
|
"completions/min_length": 224.0,
|
|
"completions/max_length": 968.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 316.71429443359375,
|
|
"completions/min_terminated_length": 224.0,
|
|
"completions/max_terminated_length": 367.0,
|
|
"tools/call_frequency": 9.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005903733894228935,
|
|
"sampling/sampling_logp_difference/max": 0.7299997806549072,
|
|
"sampling/importance_sampling_ratio/min": 0.2959112823009491,
|
|
"sampling/importance_sampling_ratio/mean": 0.6095614433288574,
|
|
"sampling/importance_sampling_ratio/max": 1.4996168613433838,
|
|
"kl": 0.0105119954332622,
|
|
"entropy": 0.08386149373836815,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.709685321897268,
|
|
"epoch": 0.00703125,
|
|
"step": 270
|
|
},
|
|
{
|
|
"loss": 0.2908468246459961,
|
|
"grad_norm": 1.7733089923858643,
|
|
"learning_rate": 5.172413793103448e-08,
|
|
"num_tokens": 2477017.0,
|
|
"completions/mean_length": 517.375,
|
|
"completions/min_length": 245.0,
|
|
"completions/max_length": 910.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 291.0,
|
|
"completions/min_terminated_length": 245.0,
|
|
"completions/max_terminated_length": 349.0,
|
|
"tools/call_frequency": 13.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0033942032605409622,
|
|
"sampling/sampling_logp_difference/max": 0.4444870948791504,
|
|
"sampling/importance_sampling_ratio/min": 0.6484279632568359,
|
|
"sampling/importance_sampling_ratio/mean": 1.2220649719238281,
|
|
"sampling/importance_sampling_ratio/max": 1.7466833591461182,
|
|
"kl": 0.0031487909436691552,
|
|
"entropy": 0.0599333334248513,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.078023090958595,
|
|
"epoch": 0.007057291666666667,
|
|
"step": 271
|
|
},
|
|
{
|
|
"loss": 0.19954326748847961,
|
|
"grad_norm": 4.745079040527344,
|
|
"learning_rate": 5e-08,
|
|
"num_tokens": 2485379.0,
|
|
"completions/mean_length": 359.875,
|
|
"completions/min_length": 223.0,
|
|
"completions/max_length": 865.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 287.71429443359375,
|
|
"completions/min_terminated_length": 223.0,
|
|
"completions/max_terminated_length": 354.0,
|
|
"tools/call_frequency": 9.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005116648506373167,
|
|
"sampling/sampling_logp_difference/max": 0.57706618309021,
|
|
"sampling/importance_sampling_ratio/min": 0.3768359422683716,
|
|
"sampling/importance_sampling_ratio/mean": 1.1508021354675293,
|
|
"sampling/importance_sampling_ratio/max": 2.214113235473633,
|
|
"kl": 0.011548115697223693,
|
|
"entropy": 0.0793663039803505,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.815610196441412,
|
|
"epoch": 0.007083333333333333,
|
|
"step": 272
|
|
},
|
|
{
|
|
"loss": 0.21407172083854675,
|
|
"grad_norm": 4.624972343444824,
|
|
"learning_rate": 4.827586206896551e-08,
|
|
"num_tokens": 2493931.0,
|
|
"completions/mean_length": 383.5,
|
|
"completions/min_length": 211.0,
|
|
"completions/max_length": 889.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 311.2857360839844,
|
|
"completions/min_terminated_length": 211.0,
|
|
"completions/max_terminated_length": 425.0,
|
|
"tools/call_frequency": 9.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0024999999441206455,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.0024999999441206455,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.008358258754014969,
|
|
"sampling/sampling_logp_difference/max": 0.5642658472061157,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.92418372631073,
|
|
"sampling/importance_sampling_ratio/max": 2.679272413253784,
|
|
"kl": 0.007251122107845731,
|
|
"entropy": 0.09258958837017417,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.082607511430979,
|
|
"epoch": 0.007109375,
|
|
"step": 273
|
|
},
|
|
{
|
|
"loss": 0.2847435474395752,
|
|
"grad_norm": 1.9598088264465332,
|
|
"learning_rate": 4.6551724137931034e-08,
|
|
"num_tokens": 2504689.0,
|
|
"completions/mean_length": 659.25,
|
|
"completions/min_length": 221.0,
|
|
"completions/max_length": 1035.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 560.3333740234375,
|
|
"completions/min_terminated_length": 221.0,
|
|
"completions/max_terminated_length": 931.0,
|
|
"tools/call_frequency": 16.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.002258891239762306,
|
|
"sampling/sampling_logp_difference/max": 0.808781623840332,
|
|
"sampling/importance_sampling_ratio/min": 0.2601649761199951,
|
|
"sampling/importance_sampling_ratio/mean": 0.7750483155250549,
|
|
"sampling/importance_sampling_ratio/max": 2.2202298641204834,
|
|
"kl": 0.0075916502755717374,
|
|
"entropy": 0.038621517829596996,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.542291646823287,
|
|
"epoch": 0.007135416666666667,
|
|
"step": 274
|
|
},
|
|
{
|
|
"loss": 0.26881924271583557,
|
|
"grad_norm": 1.6947553157806396,
|
|
"learning_rate": 4.482758620689655e-08,
|
|
"num_tokens": 2515095.0,
|
|
"completions/mean_length": 615.0,
|
|
"completions/min_length": 287.0,
|
|
"completions/max_length": 967.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 434.20001220703125,
|
|
"completions/min_terminated_length": 287.0,
|
|
"completions/max_terminated_length": 967.0,
|
|
"tools/call_frequency": 14.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.002989558968693018,
|
|
"sampling/sampling_logp_difference/max": 0.4621241092681885,
|
|
"sampling/importance_sampling_ratio/min": 0.25885090231895447,
|
|
"sampling/importance_sampling_ratio/mean": 0.7193624973297119,
|
|
"sampling/importance_sampling_ratio/max": 1.0958611965179443,
|
|
"kl": 0.00491372260148637,
|
|
"entropy": 0.051202194998040795,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.535184862092137,
|
|
"epoch": 0.007161458333333333,
|
|
"step": 275
|
|
},
|
|
{
|
|
"loss": 0.532142162322998,
|
|
"grad_norm": 5.226210117340088,
|
|
"learning_rate": 4.310344827586207e-08,
|
|
"num_tokens": 2523325.0,
|
|
"completions/mean_length": 342.5,
|
|
"completions/min_length": 233.0,
|
|
"completions/max_length": 879.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 342.5,
|
|
"completions/min_terminated_length": 233.0,
|
|
"completions/max_terminated_length": 879.0,
|
|
"tools/call_frequency": 9.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005580945871770382,
|
|
"sampling/sampling_logp_difference/max": 1.083496332168579,
|
|
"sampling/importance_sampling_ratio/min": 0.18701519072055817,
|
|
"sampling/importance_sampling_ratio/mean": 1.123509407043457,
|
|
"sampling/importance_sampling_ratio/max": 2.6058671474456787,
|
|
"kl": 0.007965543587488355,
|
|
"entropy": 0.07284831930883229,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.433826694265008,
|
|
"epoch": 0.0071875,
|
|
"step": 276
|
|
},
|
|
{
|
|
"loss": 0.23374122381210327,
|
|
"grad_norm": 1.8336155414581299,
|
|
"learning_rate": 4.1379310344827585e-08,
|
|
"num_tokens": 2532942.0,
|
|
"completions/mean_length": 515.25,
|
|
"completions/min_length": 228.0,
|
|
"completions/max_length": 974.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 393.66668701171875,
|
|
"completions/min_terminated_length": 228.0,
|
|
"completions/max_terminated_length": 974.0,
|
|
"tools/call_frequency": 12.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0032031123992055655,
|
|
"sampling/sampling_logp_difference/max": 0.4661426544189453,
|
|
"sampling/importance_sampling_ratio/min": 0.3372173607349396,
|
|
"sampling/importance_sampling_ratio/mean": 0.8848479390144348,
|
|
"sampling/importance_sampling_ratio/max": 1.833616852760315,
|
|
"kl": 0.008634808058559429,
|
|
"entropy": 0.0563446240266785,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.387338645756245,
|
|
"epoch": 0.007213541666666667,
|
|
"step": 277
|
|
},
|
|
{
|
|
"loss": 0.33201882243156433,
|
|
"grad_norm": 3.0642924308776855,
|
|
"learning_rate": 3.9655172413793106e-08,
|
|
"num_tokens": 2541372.0,
|
|
"completions/mean_length": 367.875,
|
|
"completions/min_length": 236.0,
|
|
"completions/max_length": 898.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 292.14288330078125,
|
|
"completions/min_terminated_length": 236.0,
|
|
"completions/max_terminated_length": 358.0,
|
|
"tools/call_frequency": 9.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350982658565044,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006460727192461491,
|
|
"sampling/sampling_logp_difference/max": 0.7220900058746338,
|
|
"sampling/importance_sampling_ratio/min": 0.3216889202594757,
|
|
"sampling/importance_sampling_ratio/mean": 0.8248634338378906,
|
|
"sampling/importance_sampling_ratio/max": 1.8941302299499512,
|
|
"kl": 0.0075324779318179935,
|
|
"entropy": 0.08785595803055912,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.174775030463934,
|
|
"epoch": 0.007239583333333333,
|
|
"step": 278
|
|
},
|
|
{
|
|
"loss": -0.11112722009420395,
|
|
"grad_norm": 3.914093255996704,
|
|
"learning_rate": 3.793103448275862e-08,
|
|
"num_tokens": 2549163.0,
|
|
"completions/mean_length": 287.75,
|
|
"completions/min_length": 240.0,
|
|
"completions/max_length": 367.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 287.75,
|
|
"completions/min_terminated_length": 240.0,
|
|
"completions/max_terminated_length": 367.0,
|
|
"tools/call_frequency": 7.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.015118579380214214,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.015118579380214214,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006510537583380938,
|
|
"sampling/sampling_logp_difference/max": 0.6504517793655396,
|
|
"sampling/importance_sampling_ratio/min": 0.19207413494586945,
|
|
"sampling/importance_sampling_ratio/mean": 0.9166412353515625,
|
|
"sampling/importance_sampling_ratio/max": 2.0869061946868896,
|
|
"kl": 0.009531883595627733,
|
|
"entropy": 0.07367783645167947,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 6.6603082697838545,
|
|
"epoch": 0.007265625,
|
|
"step": 279
|
|
},
|
|
{
|
|
"loss": 0.32302504777908325,
|
|
"grad_norm": 1.7986541986465454,
|
|
"learning_rate": 3.6206896551724136e-08,
|
|
"num_tokens": 2559050.0,
|
|
"completions/mean_length": 549.5,
|
|
"completions/min_length": 240.0,
|
|
"completions/max_length": 1017.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 315.6000061035156,
|
|
"completions/min_terminated_length": 240.0,
|
|
"completions/max_terminated_length": 386.0,
|
|
"tools/call_frequency": 12.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.008750000037252903,
|
|
"rewards/reward_func/std": 0.018077217042446136,
|
|
"reward": 0.008750000037252903,
|
|
"reward_std": 0.018077215179800987,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003989457618445158,
|
|
"sampling/sampling_logp_difference/max": 0.7379839420318604,
|
|
"sampling/importance_sampling_ratio/min": 0.36552587151527405,
|
|
"sampling/importance_sampling_ratio/mean": 0.9021133780479431,
|
|
"sampling/importance_sampling_ratio/max": 1.1511744260787964,
|
|
"kl": 0.0051152469459339045,
|
|
"entropy": 0.0793979907175526,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 13.537986971437931,
|
|
"epoch": 0.007291666666666667,
|
|
"step": 280
|
|
},
|
|
{
|
|
"loss": 0.3495161235332489,
|
|
"grad_norm": 6.53402042388916,
|
|
"learning_rate": 3.448275862068965e-08,
|
|
"num_tokens": 2567839.0,
|
|
"completions/mean_length": 413.25,
|
|
"completions/min_length": 73.0,
|
|
"completions/max_length": 906.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 251.83334350585938,
|
|
"completions/min_terminated_length": 73.0,
|
|
"completions/max_terminated_length": 360.0,
|
|
"tools/call_frequency": 10.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003817326854914427,
|
|
"sampling/sampling_logp_difference/max": 0.9216570854187012,
|
|
"sampling/importance_sampling_ratio/min": 0.3402063250541687,
|
|
"sampling/importance_sampling_ratio/mean": 0.9173711538314819,
|
|
"sampling/importance_sampling_ratio/max": 1.6381093263626099,
|
|
"kl": 0.008653688688355032,
|
|
"entropy": 0.06760681321611628,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.91777971573174,
|
|
"epoch": 0.007317708333333333,
|
|
"step": 281
|
|
},
|
|
{
|
|
"loss": -0.005241177976131439,
|
|
"grad_norm": 5.505411148071289,
|
|
"learning_rate": 3.275862068965517e-08,
|
|
"num_tokens": 2577588.0,
|
|
"completions/mean_length": 533.375,
|
|
"completions/min_length": 247.0,
|
|
"completions/max_length": 898.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 326.20001220703125,
|
|
"completions/min_terminated_length": 247.0,
|
|
"completions/max_terminated_length": 388.0,
|
|
"tools/call_frequency": 13.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0043061100877821445,
|
|
"sampling/sampling_logp_difference/max": 0.6723388433456421,
|
|
"sampling/importance_sampling_ratio/min": 0.3206723928451538,
|
|
"sampling/importance_sampling_ratio/mean": 1.013468861579895,
|
|
"sampling/importance_sampling_ratio/max": 2.7026264667510986,
|
|
"kl": 0.006114764335507061,
|
|
"entropy": 0.0807863858062774,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.989927288144827,
|
|
"epoch": 0.00734375,
|
|
"step": 282
|
|
},
|
|
{
|
|
"loss": 0.12377066910266876,
|
|
"grad_norm": 2.994549036026001,
|
|
"learning_rate": 3.103448275862069e-08,
|
|
"num_tokens": 2586009.0,
|
|
"completions/mean_length": 366.0,
|
|
"completions/min_length": 220.0,
|
|
"completions/max_length": 918.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 287.14288330078125,
|
|
"completions/min_terminated_length": 220.0,
|
|
"completions/max_terminated_length": 387.0,
|
|
"tools/call_frequency": 9.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004887273069471121,
|
|
"sampling/sampling_logp_difference/max": 0.3900187015533447,
|
|
"sampling/importance_sampling_ratio/min": 0.15042641758918762,
|
|
"sampling/importance_sampling_ratio/mean": 0.70041424036026,
|
|
"sampling/importance_sampling_ratio/max": 1.1667711734771729,
|
|
"kl": 0.007626379599969368,
|
|
"entropy": 0.06976995407603681,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.127276694402099,
|
|
"epoch": 0.007369791666666667,
|
|
"step": 283
|
|
},
|
|
{
|
|
"loss": 0.18977496027946472,
|
|
"grad_norm": 2.4539172649383545,
|
|
"learning_rate": 2.9310344827586205e-08,
|
|
"num_tokens": 2596194.0,
|
|
"completions/mean_length": 587.375,
|
|
"completions/min_length": 216.0,
|
|
"completions/max_length": 918.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 393.20001220703125,
|
|
"completions/min_terminated_length": 216.0,
|
|
"completions/max_terminated_length": 889.0,
|
|
"tools/call_frequency": 14.875,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00307518825866282,
|
|
"sampling/sampling_logp_difference/max": 1.657059907913208,
|
|
"sampling/importance_sampling_ratio/min": 0.15387243032455444,
|
|
"sampling/importance_sampling_ratio/mean": 0.9094022512435913,
|
|
"sampling/importance_sampling_ratio/max": 1.4928914308547974,
|
|
"kl": 0.004419457352923928,
|
|
"entropy": 0.0577328548533842,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.484683072194457,
|
|
"epoch": 0.007395833333333333,
|
|
"step": 284
|
|
},
|
|
{
|
|
"loss": -0.12873654067516327,
|
|
"grad_norm": 2.665879011154175,
|
|
"learning_rate": 2.7586206896551723e-08,
|
|
"num_tokens": 2605182.0,
|
|
"completions/mean_length": 437.875,
|
|
"completions/min_length": 228.0,
|
|
"completions/max_length": 912.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 437.875,
|
|
"completions/min_terminated_length": 228.0,
|
|
"completions/max_terminated_length": 912.0,
|
|
"tools/call_frequency": 11.375,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": -0.008750000037252903,
|
|
"rewards/reward_func/std": 0.021001702174544334,
|
|
"reward": -0.008750000037252903,
|
|
"reward_std": 0.021001700311899185,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0038780251052230597,
|
|
"sampling/sampling_logp_difference/max": 0.5645434856414795,
|
|
"sampling/importance_sampling_ratio/min": 0.5552787184715271,
|
|
"sampling/importance_sampling_ratio/mean": 0.8206003904342651,
|
|
"sampling/importance_sampling_ratio/max": 1.4064229726791382,
|
|
"kl": 0.009380330113344826,
|
|
"entropy": 0.05325771262869239,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.997083142399788,
|
|
"epoch": 0.007421875,
|
|
"step": 285
|
|
},
|
|
{
|
|
"loss": 0.23218244314193726,
|
|
"grad_norm": 1.84993577003479,
|
|
"learning_rate": 2.586206896551724e-08,
|
|
"num_tokens": 2614930.0,
|
|
"completions/mean_length": 533.25,
|
|
"completions/min_length": 249.0,
|
|
"completions/max_length": 974.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 409.8333435058594,
|
|
"completions/min_terminated_length": 249.0,
|
|
"completions/max_terminated_length": 974.0,
|
|
"tools/call_frequency": 12.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004786623176187277,
|
|
"sampling/sampling_logp_difference/max": 0.7971168160438538,
|
|
"sampling/importance_sampling_ratio/min": 0.18262535333633423,
|
|
"sampling/importance_sampling_ratio/mean": 0.4871298670768738,
|
|
"sampling/importance_sampling_ratio/max": 0.8146491646766663,
|
|
"kl": 0.008224239099945407,
|
|
"entropy": 0.07259130221791565,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.807900799438357,
|
|
"epoch": 0.007447916666666667,
|
|
"step": 286
|
|
},
|
|
{
|
|
"loss": 0.3082104027271271,
|
|
"grad_norm": 2.382488250732422,
|
|
"learning_rate": 2.4137931034482756e-08,
|
|
"num_tokens": 2625302.0,
|
|
"completions/mean_length": 610.875,
|
|
"completions/min_length": 223.0,
|
|
"completions/max_length": 976.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 401.3999938964844,
|
|
"completions/min_terminated_length": 223.0,
|
|
"completions/max_terminated_length": 924.0,
|
|
"tools/call_frequency": 14.125,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0026229447685182095,
|
|
"sampling/sampling_logp_difference/max": 0.45190882682800293,
|
|
"sampling/importance_sampling_ratio/min": 0.3924666941165924,
|
|
"sampling/importance_sampling_ratio/mean": 1.0087003707885742,
|
|
"sampling/importance_sampling_ratio/max": 1.9725937843322754,
|
|
"kl": 0.007161801992879191,
|
|
"entropy": 0.049780790228396654,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.230722200125456,
|
|
"epoch": 0.007473958333333333,
|
|
"step": 287
|
|
},
|
|
{
|
|
"loss": 0.16229654848575592,
|
|
"grad_norm": 3.0092933177948,
|
|
"learning_rate": 2.2413793103448274e-08,
|
|
"num_tokens": 2633754.0,
|
|
"completions/mean_length": 371.25,
|
|
"completions/min_length": 218.0,
|
|
"completions/max_length": 893.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 296.71429443359375,
|
|
"completions/min_terminated_length": 218.0,
|
|
"completions/max_terminated_length": 345.0,
|
|
"tools/call_frequency": 9.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.014999999664723873,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.014999999664723873,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.006459589581936598,
|
|
"sampling/sampling_logp_difference/max": 0.6311815977096558,
|
|
"sampling/importance_sampling_ratio/min": 0.2634403109550476,
|
|
"sampling/importance_sampling_ratio/mean": 0.5094240307807922,
|
|
"sampling/importance_sampling_ratio/max": 0.9145494103431702,
|
|
"kl": 0.011056321847718209,
|
|
"entropy": 0.07610198156908154,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.565595062449574,
|
|
"epoch": 0.0075,
|
|
"step": 288
|
|
},
|
|
{
|
|
"loss": 0.0985320508480072,
|
|
"grad_norm": 4.670811176300049,
|
|
"learning_rate": 2.0689655172413793e-08,
|
|
"num_tokens": 2641418.0,
|
|
"completions/mean_length": 271.75,
|
|
"completions/min_length": 229.0,
|
|
"completions/max_length": 355.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/mean_terminated_length": 271.75,
|
|
"completions/min_terminated_length": 229.0,
|
|
"completions/max_terminated_length": 355.0,
|
|
"tools/call_frequency": 7.75,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.004985018167644739,
|
|
"sampling/sampling_logp_difference/max": 0.7013615369796753,
|
|
"sampling/importance_sampling_ratio/min": 0.29009488224983215,
|
|
"sampling/importance_sampling_ratio/mean": 0.9498580694198608,
|
|
"sampling/importance_sampling_ratio/max": 1.3713898658752441,
|
|
"kl": 0.00903856082004495,
|
|
"entropy": 0.0653785951435566,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 5.979483006522059,
|
|
"epoch": 0.007526041666666667,
|
|
"step": 289
|
|
},
|
|
{
|
|
"loss": 0.4474698305130005,
|
|
"grad_norm": 3.3852388858795166,
|
|
"learning_rate": 1.896551724137931e-08,
|
|
"num_tokens": 2650475.0,
|
|
"completions/mean_length": 445.75,
|
|
"completions/min_length": 240.0,
|
|
"completions/max_length": 985.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 368.71429443359375,
|
|
"completions/min_terminated_length": 240.0,
|
|
"completions/max_terminated_length": 893.0,
|
|
"tools/call_frequency": 11.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0029322272166609764,
|
|
"sampling/sampling_logp_difference/max": 0.480452299118042,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 1.096142053604126,
|
|
"sampling/importance_sampling_ratio/max": 1.8924423456192017,
|
|
"kl": 0.008347948458322207,
|
|
"entropy": 0.04790501273237169,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.025753172114491,
|
|
"epoch": 0.007552083333333333,
|
|
"step": 290
|
|
},
|
|
{
|
|
"loss": 0.3999462127685547,
|
|
"grad_norm": 2.090080499649048,
|
|
"learning_rate": 1.7241379310344825e-08,
|
|
"num_tokens": 2660792.0,
|
|
"completions/mean_length": 604.0,
|
|
"completions/min_length": 223.0,
|
|
"completions/max_length": 984.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 275.0,
|
|
"completions/min_terminated_length": 223.0,
|
|
"completions/max_terminated_length": 310.0,
|
|
"tools/call_frequency": 14.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0024004140868782997,
|
|
"sampling/sampling_logp_difference/max": 0.532813549041748,
|
|
"sampling/importance_sampling_ratio/min": 0.7601610422134399,
|
|
"sampling/importance_sampling_ratio/mean": 1.1334989070892334,
|
|
"sampling/importance_sampling_ratio/max": 1.704061508178711,
|
|
"kl": 0.004228754736686824,
|
|
"entropy": 0.03497787821106613,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.161181304603815,
|
|
"epoch": 0.007578125,
|
|
"step": 291
|
|
},
|
|
{
|
|
"loss": -0.023790448904037476,
|
|
"grad_norm": 4.209896087646484,
|
|
"learning_rate": 1.5517241379310344e-08,
|
|
"num_tokens": 2670009.0,
|
|
"completions/mean_length": 466.25,
|
|
"completions/min_length": 241.0,
|
|
"completions/max_length": 957.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 310.8333435058594,
|
|
"completions/min_terminated_length": 241.0,
|
|
"completions/max_terminated_length": 405.0,
|
|
"tools/call_frequency": 11.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.005274425260722637,
|
|
"sampling/sampling_logp_difference/max": 0.6928887367248535,
|
|
"sampling/importance_sampling_ratio/min": 0.0,
|
|
"sampling/importance_sampling_ratio/mean": 0.6994878053665161,
|
|
"sampling/importance_sampling_ratio/max": 2.022887945175171,
|
|
"kl": 0.0069219160577631555,
|
|
"entropy": 0.08423585467971861,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.449313148856163,
|
|
"epoch": 0.007604166666666667,
|
|
"step": 292
|
|
},
|
|
{
|
|
"loss": 0.14826519787311554,
|
|
"grad_norm": 3.7165799140930176,
|
|
"learning_rate": 1.3793103448275862e-08,
|
|
"num_tokens": 2679032.0,
|
|
"completions/mean_length": 442.125,
|
|
"completions/min_length": 227.0,
|
|
"completions/max_length": 949.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 377.71429443359375,
|
|
"completions/min_terminated_length": 227.0,
|
|
"completions/max_terminated_length": 949.0,
|
|
"tools/call_frequency": 11.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.007499999832361937,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.007499999832361937,
|
|
"reward_std": 0.010350982658565044,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0030684417579323053,
|
|
"sampling/sampling_logp_difference/max": 0.5281970500946045,
|
|
"sampling/importance_sampling_ratio/min": 0.4313776195049286,
|
|
"sampling/importance_sampling_ratio/mean": 0.8659086227416992,
|
|
"sampling/importance_sampling_ratio/max": 1.475406527519226,
|
|
"kl": 0.008013775317522231,
|
|
"entropy": 0.051352149224840105,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.108693651854992,
|
|
"epoch": 0.0076302083333333335,
|
|
"step": 293
|
|
},
|
|
{
|
|
"loss": 0.37415972352027893,
|
|
"grad_norm": 2.002143621444702,
|
|
"learning_rate": 1.2068965517241378e-08,
|
|
"num_tokens": 2690042.0,
|
|
"completions/mean_length": 690.875,
|
|
"completions/min_length": 242.0,
|
|
"completions/max_length": 1030.0,
|
|
"completions/clipped_ratio": 0.625,
|
|
"completions/mean_terminated_length": 301.3333435058594,
|
|
"completions/min_terminated_length": 242.0,
|
|
"completions/max_terminated_length": 403.0,
|
|
"tools/call_frequency": 16.0,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0031540789641439915,
|
|
"sampling/sampling_logp_difference/max": 0.6719317436218262,
|
|
"sampling/importance_sampling_ratio/min": 0.526262640953064,
|
|
"sampling/importance_sampling_ratio/mean": 0.9227762222290039,
|
|
"sampling/importance_sampling_ratio/max": 1.7364429235458374,
|
|
"kl": 0.004388742865558015,
|
|
"entropy": 0.060072507709264755,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 14.383178193122149,
|
|
"epoch": 0.00765625,
|
|
"step": 294
|
|
},
|
|
{
|
|
"loss": 0.38500067591667175,
|
|
"grad_norm": 2.328819513320923,
|
|
"learning_rate": 1.0344827586206896e-08,
|
|
"num_tokens": 2698271.0,
|
|
"completions/mean_length": 342.875,
|
|
"completions/min_length": 224.0,
|
|
"completions/max_length": 865.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 268.2857360839844,
|
|
"completions/min_terminated_length": 224.0,
|
|
"completions/max_terminated_length": 329.0,
|
|
"tools/call_frequency": 9.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.017500000074505806,
|
|
"rewards/reward_func/std": 0.007071067579090595,
|
|
"reward": 0.017500000074505806,
|
|
"reward_std": 0.007071067579090595,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0034988841507583857,
|
|
"sampling/sampling_logp_difference/max": 0.48293352127075195,
|
|
"sampling/importance_sampling_ratio/min": 0.5720033645629883,
|
|
"sampling/importance_sampling_ratio/mean": 1.0393388271331787,
|
|
"sampling/importance_sampling_ratio/max": 2.0356533527374268,
|
|
"kl": 0.009147551409114385,
|
|
"entropy": 0.06304878962691873,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.597867541015148,
|
|
"epoch": 0.007682291666666666,
|
|
"step": 295
|
|
},
|
|
{
|
|
"loss": -0.05369969829916954,
|
|
"grad_norm": 3.707434892654419,
|
|
"learning_rate": 8.620689655172413e-09,
|
|
"num_tokens": 2706729.0,
|
|
"completions/mean_length": 371.0,
|
|
"completions/min_length": 226.0,
|
|
"completions/max_length": 949.0,
|
|
"completions/clipped_ratio": 0.125,
|
|
"completions/mean_terminated_length": 288.4285888671875,
|
|
"completions/min_terminated_length": 226.0,
|
|
"completions/max_terminated_length": 351.0,
|
|
"tools/call_frequency": 9.25,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": -0.0012500002048909664,
|
|
"rewards/reward_func/std": 0.021671246737241745,
|
|
"reward": -0.0012500002048909664,
|
|
"reward_std": 0.021671246737241745,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.00490680243819952,
|
|
"sampling/sampling_logp_difference/max": 0.6842975616455078,
|
|
"sampling/importance_sampling_ratio/min": 0.5149754881858826,
|
|
"sampling/importance_sampling_ratio/mean": 1.3425382375717163,
|
|
"sampling/importance_sampling_ratio/max": 2.8035032749176025,
|
|
"kl": 0.0062346317354240455,
|
|
"entropy": 0.06913655600510538,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 10.846539419144392,
|
|
"epoch": 0.0077083333333333335,
|
|
"step": 296
|
|
},
|
|
{
|
|
"loss": 0.1919277161359787,
|
|
"grad_norm": 3.5629501342773438,
|
|
"learning_rate": 6.896551724137931e-09,
|
|
"num_tokens": 2715601.0,
|
|
"completions/mean_length": 422.75,
|
|
"completions/min_length": 218.0,
|
|
"completions/max_length": 957.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 252.0,
|
|
"completions/min_terminated_length": 218.0,
|
|
"completions/max_terminated_length": 318.0,
|
|
"tools/call_frequency": 10.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.01249999925494194,
|
|
"rewards/reward_func/std": 0.010350983589887619,
|
|
"reward": 0.01249999925494194,
|
|
"reward_std": 0.010350983589887619,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.0038121493998914957,
|
|
"sampling/sampling_logp_difference/max": 0.45551300048828125,
|
|
"sampling/importance_sampling_ratio/min": 0.5735144019126892,
|
|
"sampling/importance_sampling_ratio/mean": 1.1412287950515747,
|
|
"sampling/importance_sampling_ratio/max": 2.110752820968628,
|
|
"kl": 0.01284753848449327,
|
|
"entropy": 0.060013891430571675,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.54495875351131,
|
|
"epoch": 0.007734375,
|
|
"step": 297
|
|
},
|
|
{
|
|
"loss": 0.0071893236599862576,
|
|
"grad_norm": 1.5247243642807007,
|
|
"learning_rate": 5.172413793103448e-09,
|
|
"num_tokens": 2726009.0,
|
|
"completions/mean_length": 615.125,
|
|
"completions/min_length": 234.0,
|
|
"completions/max_length": 1036.0,
|
|
"completions/clipped_ratio": 0.5,
|
|
"completions/mean_terminated_length": 297.75,
|
|
"completions/min_terminated_length": 234.0,
|
|
"completions/max_terminated_length": 402.0,
|
|
"tools/call_frequency": 14.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.0,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.0,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.002946035470813513,
|
|
"sampling/sampling_logp_difference/max": 0.6161698698997498,
|
|
"sampling/importance_sampling_ratio/min": 0.306558758020401,
|
|
"sampling/importance_sampling_ratio/mean": 0.8154085874557495,
|
|
"sampling/importance_sampling_ratio/max": 1.4422669410705566,
|
|
"kl": 0.006372960778207926,
|
|
"entropy": 0.0492012407630682,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 12.67117932997644,
|
|
"epoch": 0.007760416666666666,
|
|
"step": 298
|
|
},
|
|
{
|
|
"loss": 0.32350555062294006,
|
|
"grad_norm": 2.8527565002441406,
|
|
"learning_rate": 3.4482758620689654e-09,
|
|
"num_tokens": 2734900.0,
|
|
"completions/mean_length": 425.5,
|
|
"completions/min_length": 233.0,
|
|
"completions/max_length": 880.0,
|
|
"completions/clipped_ratio": 0.25,
|
|
"completions/mean_terminated_length": 276.0,
|
|
"completions/min_terminated_length": 233.0,
|
|
"completions/max_terminated_length": 385.0,
|
|
"tools/call_frequency": 11.625,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.009999999776482582,
|
|
"rewards/reward_func/std": 0.010690449737012386,
|
|
"reward": 0.009999999776482582,
|
|
"reward_std": 0.010690449737012386,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.003388202516362071,
|
|
"sampling/sampling_logp_difference/max": 0.40558838844299316,
|
|
"sampling/importance_sampling_ratio/min": 0.6585291028022766,
|
|
"sampling/importance_sampling_ratio/mean": 0.8299328088760376,
|
|
"sampling/importance_sampling_ratio/max": 1.0050992965698242,
|
|
"kl": 0.009260975187316944,
|
|
"entropy": 0.05885127425426617,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.24569141305983,
|
|
"epoch": 0.007786458333333334,
|
|
"step": 299
|
|
},
|
|
{
|
|
"loss": 0.3415127694606781,
|
|
"grad_norm": 2.9234366416931152,
|
|
"learning_rate": 1.7241379310344827e-09,
|
|
"num_tokens": 2744353.0,
|
|
"completions/mean_length": 496.0,
|
|
"completions/min_length": 207.0,
|
|
"completions/max_length": 887.0,
|
|
"completions/clipped_ratio": 0.375,
|
|
"completions/mean_terminated_length": 265.3999938964844,
|
|
"completions/min_terminated_length": 207.0,
|
|
"completions/max_terminated_length": 323.0,
|
|
"tools/call_frequency": 13.5,
|
|
"tools/failure_frequency": 0.0,
|
|
"rewards/reward_func/mean": 0.004999999888241291,
|
|
"rewards/reward_func/std": 0.009258201345801353,
|
|
"reward": 0.004999999888241291,
|
|
"reward_std": 0.009258200414478779,
|
|
"frac_reward_zero_std": 0.0,
|
|
"sampling/sampling_logp_difference/mean": 0.002770001767203212,
|
|
"sampling/sampling_logp_difference/max": 0.32262474298477173,
|
|
"sampling/importance_sampling_ratio/min": 0.6522640585899353,
|
|
"sampling/importance_sampling_ratio/mean": 1.1230385303497314,
|
|
"sampling/importance_sampling_ratio/max": 1.9639067649841309,
|
|
"kl": 0.0044916425831615925,
|
|
"entropy": 0.05473532422911376,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"step_time": 11.373993566259742,
|
|
"epoch": 0.0078125,
|
|
"step": 300
|
|
},
|
|
{
|
|
"train_runtime": 3864.7546,
|
|
"train_samples_per_second": 0.621,
|
|
"train_steps_per_second": 0.078,
|
|
"total_flos": 0.0,
|
|
"train_loss": 0.18028511427661822,
|
|
"epoch": 0.0078125,
|
|
"step": 300
|
|
}
|
|
] |