{ "pooled": { "epoch1_n_calls": 8, "epoch1_n_loss_tokens": 84496, "epoch1_n_nonzero_adv_tokens": 74749, "epoch1_dppo_masked_token_fraction": 0.0, "epoch1_dppo_masked_nonzero_adv_fraction": 0.0, "epoch1_ppo_clip_fraction": 0.001195322855519788, "epoch1_ppo_clip_nonzero_adv_fraction": 0.0013511886446641427, "epoch1_policy_drift_abs_p_q90": 0.012510707524517606, "epoch1_policy_drift_abs_p_q99": 0.03694757685554565, "epoch1_policy_drift_abs_p_max": 0.13695566826186, "epoch1_signed_drift_min": -0.13695566826186, "epoch1_signed_drift_max": 0.1303161921375165, "epoch1_pos_adv_drift_max": 0.1303161921375165, "epoch1_neg_adv_drift_min": -0.11192234652459221, "epoch1_abs_drift_gt_delta_tokens": 0, "epoch1_removed_nonzero_grad_pos": 0, "epoch1_removed_nonzero_grad_neg": 0, "epoch1_ratio_cap_hits": 0, "epoch1_ratio_cap_hits_nonzero_adv": 0, "epoch1_n_pos_adv": 36907, "epoch1_n_neg_adv": 37842, "epoch2_n_calls": 8, "epoch2_n_loss_tokens": 84496, "epoch2_n_nonzero_adv_tokens": 74749, "epoch2_dppo_masked_token_fraction": 0.0011361484567316796, "epoch2_dppo_masked_nonzero_adv_fraction": 0.0012842981177005712, "epoch2_ppo_clip_fraction": 0.013692955879568263, "epoch2_ppo_clip_nonzero_adv_fraction": 0.015478467939370426, "epoch2_policy_drift_abs_p_q90": 0.023858560402555296, "epoch2_policy_drift_abs_p_q99": 0.08628262678867459, "epoch2_policy_drift_abs_p_max": 0.41873542048368445, "epoch2_signed_drift_min": -0.41873542048368445, "epoch2_signed_drift_max": 0.3925847664298823, "epoch2_pos_adv_drift_max": 0.3925847664298823, "epoch2_neg_adv_drift_min": -0.41687695132880825, "epoch2_abs_drift_gt_delta_tokens": 224, "epoch2_removed_nonzero_grad_pos": 50, "epoch2_removed_nonzero_grad_neg": 46, "epoch2_ratio_cap_hits": 0, "epoch2_ratio_cap_hits_nonzero_adv": 0, "epoch2_n_pos_adv": 36907, "epoch2_n_neg_adv": 37842 }, "revised_checks": { "dppo_masked_in_range": true, "ppo_clip_ge_0.0005": true, "drift_q90_ge_0.005": true, "removed_grad_pos_present": true, "removed_grad_neg_present": true, "all_8_second_epoch_calls": true }, "sentinels": [ { "stage": "sentinel", "tag": "init", "n": 1976, "sha256": "10e156f726b8feacca2c615c4257f50c7cfa06f63a56b676bd9dfbb60806c1da", "vs_sampler_absdp_q90": 0.007642295211553574, "vs_sampler_absdp_max": 0.0621759295463562, "vs_sampler_mean_abs_dlogp": 0.006447661202400923, "t": 333.44 }, { "stage": "sentinel", "tag": "after_u1", "n": 1976, "sha256": "19f5893d062469b6459b6139ef88258ca479352966a813c076c0756d83e48154", "vs_sampler_absdp_q90": 0.006376773118972778, "vs_sampler_absdp_max": 0.06217598915100098, "vs_sampler_mean_abs_dlogp": 0.005680710077285767, "vs_init_absdp_q90": 0.005858182907104492, "vs_init_absdp_max": 0.06052061915397644, "vs_init_mean_abs_dlogp": 0.005724234972149134, "t": 382.85 }, { "stage": "sentinel", "tag": "after_u2", "n": 1976, "sha256": "99a3fb5d831822fab10f3c4e117d980b5eca7b60a7bf3a2d6eb9cbb78f0cc78a", "vs_sampler_absdp_q90": 0.006704002618789673, "vs_sampler_absdp_max": 0.062101125717163086, "vs_sampler_mean_abs_dlogp": 0.006335264537483454, "vs_init_absdp_q90": 0.007420867681503296, "vs_init_absdp_max": 0.061342209577560425, "vs_init_mean_abs_dlogp": 0.006251731887459755, "t": 397.41 }, { "stage": "sentinel", "tag": "after_u4", "n": 1976, "sha256": "ed2b5094849eaac7235c07597e38ab011235f4bac9afaf07bdeccab7218582a2", "vs_sampler_absdp_q90": 0.010810106992721558, "vs_sampler_absdp_max": 0.09195864200592041, "vs_sampler_mean_abs_dlogp": 0.008140268735587597, "vs_init_absdp_q90": 0.009358584880828857, "vs_init_absdp_max": 0.09145200252532959, "vs_init_mean_abs_dlogp": 0.007964221760630608, "t": 408.85 }, { "stage": "sentinel", "tag": "after_u8", "n": 1976, "sha256": "c8b000e2a3dd276266660ef42b819530c8220615c96375ce804d37b14305f658", "vs_sampler_absdp_q90": 0.015132233500480652, "vs_sampler_absdp_max": 0.15422680974006653, "vs_sampler_mean_abs_dlogp": 0.01190256793051958, "vs_init_absdp_q90": 0.015591561794281006, "vs_init_absdp_max": 0.1537201702594757, "vs_init_mean_abs_dlogp": 0.012130395509302616, "t": 430.08 }, { "stage": "sentinel", "tag": "after_u16", "n": 1976, "sha256": "53ddab1bf5d7be91b7476fd643edb0687a0c3a763ba36b168d5acddd75fae561", "vs_sampler_absdp_q90": 0.03070691227912903, "vs_sampler_absdp_max": 0.3288673162460327, "vs_sampler_mean_abs_dlogp": 0.027858657762408257, "vs_init_absdp_q90": 0.03011302649974823, "vs_init_absdp_max": 0.3283606767654419, "vs_init_mean_abs_dlogp": 0.027917437255382538, "t": 465.91 } ], "updates_executed": 16 }