| {"lr": 9.723256244335045e-08, "step": 10, "loss": 2.6261255741119385, "prob_real_tokens": 0.5996527777777778, "percent_done": 0.3333333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12023.794532430335, "avg_wps": 7909.395521492682, "eta_in_seconds": 27218.262054920197, "at": "2026-05-01T18:03:52.523341"} |
| {"lr": 1.5600807173903563e-07, "step": 20, "loss": 2.5958285331726074, "prob_real_tokens": 0.5732361111111111, "percent_done": 0.6666666666666666, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12020.47019285301, "avg_wps": 9510.901331815166, "eta_in_seconds": 22559.376079559326, "at": "2026-05-01T18:04:52.902873"} |
| {"lr": 2.539366678223751e-07, "step": 30, "loss": 2.6436898708343506, "prob_real_tokens": 0.668125, "percent_done": 1.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11937.552492895436, "avg_wps": 10210.181079466158, "eta_in_seconds": 20943.800931215286, "at": "2026-05-01T18:05:53.055327"} |
| {"lr": 3.866809730557847e-07, "step": 40, "loss": 2.5177183151245117, "prob_real_tokens": 0.6114444444444445, "percent_done": 1.3333333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12131.654877409941, "avg_wps": 10619.82088532557, "eta_in_seconds": 20068.135075092316, "at": "2026-05-01T18:06:52.696614"} |
| {"lr": 5.483615793392435e-07, "step": 50, "loss": 2.579017162322998, "prob_real_tokens": 0.5165555555555555, "percent_done": 1.6666666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12054.742498485723, "avg_wps": 10871.80250922879, "eta_in_seconds": 19536.77872824669, "at": "2026-05-01T18:07:52.641166"} |
| {"lr": 7.318174543065956e-07, "step": 60, "loss": 2.653716802597046, "prob_real_tokens": 0.6576805555555556, "percent_done": 2.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11989.647267853004, "avg_wps": 11051.366473878414, "eta_in_seconds": 19154.19242501259, "at": "2026-05-01T18:08:52.414981"} |
| {"lr": 9.289231122326131e-07, "step": 70, "loss": 2.232825517654419, "prob_real_tokens": 0.44355555555555554, "percent_done": 2.3333333333333335, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12296.218493040637, "avg_wps": 11182.342055201194, "eta_in_seconds": 18865.457607950484, "at": "2026-05-01T18:09:52.227341"} |
| {"lr": 1.1309485017623056e-06, "step": 80, "loss": 2.33628511428833, "prob_real_tokens": 0.39131944444444444, "percent_done": 2.6666666666666665, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12276.50127023358, "avg_wps": 11300.829714118316, "eta_in_seconds": 18603.943720817566, "at": "2026-05-01T18:10:51.219713"} |
| {"lr": 1.3289456705935219e-06, "step": 90, "loss": 2.215456962585449, "prob_real_tokens": 0.4839027777777778, "percent_done": 3.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12180.682516981125, "avg_wps": 11391.654072309659, "eta_in_seconds": 18392.41243370374, "at": "2026-05-01T18:11:50.363694"} |
| {"lr": 1.514145081259773e-06, "step": 100, "loss": 2.3091251850128174, "prob_real_tokens": 0.5497777777777778, "percent_done": 3.3333333333333335, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11787.446487787347, "avg_wps": 11462.9773373768, "eta_in_seconds": 18215.16294193268, "at": "2026-05-01T18:12:49.638913"} |
| {"lr": 1.6783440247011435e-06, "step": 110, "loss": 2.352792739868164, "prob_real_tokens": 0.6556944444444445, "percent_done": 3.6666666666666665, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12160.250081097893, "avg_wps": 11526.122393133759, "eta_in_seconds": 18052.905643614857, "at": "2026-05-01T18:13:48.667796"} |
| {"lr": 1.8142699283098177e-06, "step": 120, "loss": 2.155761957168579, "prob_real_tokens": 0.6116666666666667, "percent_done": 4.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12137.773811924364, "avg_wps": 11579.523573614006, "eta_in_seconds": 17907.47250366211, "at": "2026-05-01T18:14:47.681395"} |
| {"lr": 1.915902467091256e-06, "step": 130, "loss": 2.151371717453003, "prob_real_tokens": 0.6074583333333333, "percent_done": 4.333333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12187.638939195973, "avg_wps": 11621.691860156478, "eta_in_seconds": 17780.543701080176, "at": "2026-05-01T18:15:46.930956"} |
| {"lr": 1.978740211242601e-06, "step": 140, "loss": 2.114699125289917, "prob_real_tokens": 0.5224444444444445, "percent_done": 4.666666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12201.047676826574, "avg_wps": 11658.838687931126, "eta_in_seconds": 17662.136471033096, "at": "2026-05-01T18:16:46.124575"} |
| {"lr": 2e-06, "step": 150, "loss": 2.066315174102783, "prob_real_tokens": 0.4477777777777778, "percent_done": 5.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11681.102194538717, "avg_wps": 11687.227544301317, "eta_in_seconds": 17557.628549814224, "at": "2026-05-01T18:17:45.637487"} |
| {"lr": 1.9999392461373595e-06, "step": 160, "loss": 1.8919814825057983, "prob_real_tokens": 0.526, "percent_done": 5.333333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12263.884918672898, "avg_wps": 11716.721161081405, "eta_in_seconds": 17451.981419444084, "at": "2026-05-01T18:18:44.770950"} |
| {"lr": 1.9997569919315307e-06, "step": 170, "loss": 2.259298801422119, "prob_real_tokens": 0.668125, "percent_done": 5.666666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12136.5513847244, "avg_wps": 11741.91996286355, "eta_in_seconds": 17353.209751423667, "at": "2026-05-01T18:19:43.983227"} |
| {"lr": 1.9994532595278964e-06, "step": 180, "loss": 1.9328594207763672, "prob_real_tokens": 0.47733333333333333, "percent_done": 6.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12174.105045323655, "avg_wps": 11763.32751973727, "eta_in_seconds": 17260.422245264053, "at": "2026-05-01T18:20:43.296658"} |
| {"lr": 1.999028085832438e-06, "step": 190, "loss": 1.8491747379302979, "prob_real_tokens": 0.4801111111111111, "percent_done": 6.333333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12167.416525605191, "avg_wps": 11779.874760135355, "eta_in_seconds": 17175.055263293416, "at": "2026-05-01T18:21:42.874139"} |
| {"lr": 1.9984815225072504e-06, "step": 200, "loss": 2.0131406784057617, "prob_real_tokens": 0.5551944444444444, "percent_done": 6.666666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12084.562275095064, "avg_wps": 11796.995768341552, "eta_in_seconds": 17089.096576690674, "at": "2026-05-01T18:22:42.225932"} |
| {"lr": 1.9978136359642655e-06, "step": 210, "loss": 1.8793667554855347, "prob_real_tokens": 0.5498888888888889, "percent_done": 7.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12048.226276138395, "avg_wps": 11810.588563092251, "eta_in_seconds": 17008.46650671959, "at": "2026-05-01T18:23:41.787371"} |
| {"lr": 1.997024507357183e-06, "step": 220, "loss": 2.1023807525634766, "prob_real_tokens": 0.7625555555555555, "percent_done": 7.333333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12109.06462403163, "avg_wps": 11823.82012564329, "eta_in_seconds": 16928.538989348846, "at": "2026-05-01T18:24:41.252274"} |
| {"lr": 1.9961142325716077e-06, "step": 230, "loss": 1.8086228370666504, "prob_real_tokens": 0.5661111111111111, "percent_done": 7.666666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12052.923364827506, "avg_wps": 11836.203337687475, "eta_in_seconds": 16849.997783070026, "at": "2026-05-01T18:25:40.684961"} |
| {"lr": 1.9950829222134013e-06, "step": 240, "loss": 1.7714561223983765, "prob_real_tokens": 0.5083333333333333, "percent_done": 8.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12096.96918714658, "avg_wps": 11848.220234767232, "eta_in_seconds": 16772.139280200005, "at": "2026-05-01T18:26:40.049230"} |
| {"lr": 1.99393070159524e-06, "step": 250, "loss": 2.028064250946045, "prob_real_tokens": 0.636875, "percent_done": 8.333333333333334, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12138.289490534109, "avg_wps": 11859.581536764066, "eta_in_seconds": 16695.361416101456, "at": "2026-05-01T18:27:39.367013"} |
| {"lr": 1.992657710721391e-06, "step": 260, "loss": 1.9634120464324951, "prob_real_tokens": 0.6586944444444445, "percent_done": 8.666666666666666, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12074.282835552356, "avg_wps": 11869.413706070389, "eta_in_seconds": 16620.871500932255, "at": "2026-05-01T18:28:38.773841"} |
| {"lr": 1.991264104270698e-06, "step": 270, "loss": 1.794603943824768, "prob_real_tokens": 0.5837777777777777, "percent_done": 9.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12080.532486099622, "avg_wps": 11878.562906153897, "eta_in_seconds": 16547.45624979337, "at": "2026-05-01T18:29:38.176365"} |
| {"lr": 1.9897500515777876e-06, "step": 280, "loss": 1.833031177520752, "prob_real_tokens": 0.5467777777777778, "percent_done": 9.333333333333334, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12146.602495599021, "avg_wps": 11886.906948164144, "eta_in_seconds": 16475.269879204887, "at": "2026-05-01T18:30:37.602198"} |
| {"lr": 1.988115736612495e-06, "step": 290, "loss": 1.9127681255340576, "prob_real_tokens": 0.5453472222222222, "percent_done": 9.666666666666666, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12193.984853110054, "avg_wps": 11895.361658272612, "eta_in_seconds": 16403.03217383089, "at": "2026-05-01T18:31:36.930982"} |
| {"lr": 1.9863613579575085e-06, "step": 300, "loss": 1.9145724773406982, "prob_real_tokens": 0.5354444444444444, "percent_done": 10.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12110.816235389535, "avg_wps": 11903.286442503293, "eta_in_seconds": 16331.62412238121, "at": "2026-05-01T18:32:36.253617"} |
| {"lr": 1.98448712878424e-06, "step": 310, "loss": 1.8228719234466553, "prob_real_tokens": 0.682, "percent_done": 10.333333333333334, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12135.50866261212, "avg_wps": 11910.874030339404, "eta_in_seconds": 16260.771418340744, "at": "2026-05-01T18:33:35.550523"} |
| {"lr": 1.982493276826924e-06, "step": 320, "loss": 1.7028560638427734, "prob_real_tokens": 0.5085555555555555, "percent_done": 10.666666666666666, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12183.927465602148, "avg_wps": 11918.456204348773, "eta_in_seconds": 16190.016281604767, "at": "2026-05-01T18:34:34.773001"} |
| {"lr": 1.980380044354946e-06, "step": 330, "loss": 1.786156177520752, "prob_real_tokens": 0.628125, "percent_done": 11.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12053.325538056317, "avg_wps": 11924.468948581158, "eta_in_seconds": 16121.47264829549, "at": "2026-05-01T18:35:34.181783"} |
| {"lr": 1.9781476881434026e-06, "step": 340, "loss": 1.772057294845581, "prob_real_tokens": 0.6168055555555556, "percent_done": 11.333333333333334, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12119.239005094345, "avg_wps": 11929.884758953602, "eta_in_seconds": 16053.801345923366, "at": "2026-05-01T18:36:33.633895"} |
| {"lr": 1.975796479441905e-06, "step": 350, "loss": 1.814256191253662, "prob_real_tokens": 0.4981111111111111, "percent_done": 11.666666666666666, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12143.677707261217, "avg_wps": 11935.398351286496, "eta_in_seconds": 15986.060488666808, "at": "2026-05-01T18:37:33.015117"} |
| {"lr": 1.9733267039416167e-06, "step": 360, "loss": 1.7071473598480225, "prob_real_tokens": 0.5338888888888889, "percent_done": 12.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12099.732863409916, "avg_wps": 11940.110727253668, "eta_in_seconds": 15919.4503587087, "at": "2026-05-01T18:38:32.486752"} |
| {"lr": 1.970738661740541e-06, "step": 370, "loss": 1.7102806568145752, "prob_real_tokens": 0.5811111111111111, "percent_done": 12.333333333333334, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12156.058587240776, "avg_wps": 11944.634577624729, "eta_in_seconds": 15853.142996498056, "at": "2026-05-01T18:39:31.946982"} |
| {"lr": 1.968032667307057e-06, "step": 380, "loss": 1.6926220655441284, "prob_real_tokens": 0.5667777777777778, "percent_done": 12.666666666666666, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12145.624478741207, "avg_wps": 11949.713741717826, "eta_in_seconds": 15786.152210612046, "at": "2026-05-01T18:40:31.254955"} |
| {"lr": 1.9652090494417073e-06, "step": 390, "loss": 1.8597700595855713, "prob_real_tokens": 0.6877777777777778, "percent_done": 13.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12036.263047227892, "avg_wps": 11954.659709583966, "eta_in_seconds": 15719.393488828953, "at": "2026-05-01T18:41:30.538478"} |
| {"lr": 1.962268151237249e-06, "step": 400, "loss": 1.5466258525848389, "prob_real_tokens": 0.44633333333333336, "percent_done": 13.333333333333334, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12204.898822446481, "avg_wps": 11959.64587615225, "eta_in_seconds": 15652.637372255325, "at": "2026-05-01T18:42:29.765692"} |
| {"lr": 1.959210330036962e-06, "step": 410, "loss": 1.7562555074691772, "prob_real_tokens": 0.6102222222222222, "percent_done": 13.666666666666666, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12137.909923403204, "avg_wps": 11964.212754908804, "eta_in_seconds": 15586.483107590093, "at": "2026-05-01T18:43:29.029874"} |
| {"lr": 1.9560359573912313e-06, "step": 420, "loss": 1.8284966945648193, "prob_real_tokens": 0.5383055555555556, "percent_done": 14.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12056.16604914673, "avg_wps": 11967.70953350638, "eta_in_seconds": 15521.76709168298, "at": "2026-05-01T18:44:28.474745"} |
| {"lr": 1.952745419012399e-06, "step": 430, "loss": 1.586488962173462, "prob_real_tokens": 0.5125694444444444, "percent_done": 14.333333333333334, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12158.374491681401, "avg_wps": 11971.157473790137, "eta_in_seconds": 15457.15194250262, "at": "2026-05-01T18:45:27.895432"} |
| {"lr": 1.9493391147278977e-06, "step": 440, "loss": 1.6054918766021729, "prob_real_tokens": 0.5045555555555555, "percent_done": 14.666666666666666, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12144.204143347104, "avg_wps": 11974.35580570633, "eta_in_seconds": 15392.894865556198, "at": "2026-05-01T18:46:27.336994"} |
| {"lr": 1.9458174584316676e-06, "step": 450, "loss": 1.6624916791915894, "prob_real_tokens": 0.5512361111111111, "percent_done": 15.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12087.361891762695, "avg_wps": 11977.503179578309, "eta_in_seconds": 15328.737320899963, "at": "2026-05-01T18:47:26.759052"} |
| {"lr": 1.9421808780338674e-06, "step": 460, "loss": 1.640034556388855, "prob_real_tokens": 0.5065555555555555, "percent_done": 15.333333333333334, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12178.773115288894, "avg_wps": 11980.635625998413, "eta_in_seconds": 15264.632504401, "at": "2026-05-01T18:48:26.153096"} |
| {"lr": 1.9384298154088753e-06, "step": 470, "loss": 1.693464994430542, "prob_real_tokens": 0.5613333333333334, "percent_done": 15.666666666666666, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12132.348914255008, "avg_wps": 11983.8425122307, "eta_in_seconds": 15200.466779673354, "at": "2026-05-01T18:49:25.498258"} |
| {"lr": 1.9345647263416023e-06, "step": 480, "loss": 1.6127969026565552, "prob_real_tokens": 0.6043611111111111, "percent_done": 16.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12085.84390129697, "avg_wps": 11986.971554490048, "eta_in_seconds": 15136.43368345499, "at": "2026-05-01T18:50:24.832185"} |
| {"lr": 1.930586080472107e-06, "step": 490, "loss": 1.6159451007843018, "prob_real_tokens": 0.46143055555555557, "percent_done": 16.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12204.575252414828, "avg_wps": 11990.24659455393, "eta_in_seconds": 15072.25048082702, "at": "2026-05-01T18:51:24.098668"} |
| {"lr": 1.926494361238533e-06, "step": 500, "loss": 1.635718584060669, "prob_real_tokens": 0.6003333333333334, "percent_done": 16.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12117.16308775074, "avg_wps": 11992.9644363781, "eta_in_seconds": 15008.799613714218, "at": "2026-05-01T18:52:23.470181"} |
| {"lr": 1.9222900658183648e-06, "step": 510, "loss": 1.450430989265442, "prob_real_tokens": 0.35254166666666664, "percent_done": 17.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12159.963638075313, "avg_wps": 11995.860961327091, "eta_in_seconds": 14945.154881168814, "at": "2026-05-01T18:53:24.297239"} |
| {"lr": 1.917973705068019e-06, "step": 520, "loss": 1.555511236190796, "prob_real_tokens": 0.49966666666666665, "percent_done": 17.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12180.981237514581, "avg_wps": 11998.21192447516, "eta_in_seconds": 14882.217544078827, "at": "2026-05-01T18:54:23.710473"} |
| {"lr": 1.91354580346077e-06, "step": 530, "loss": 1.578878402709961, "prob_real_tokens": 0.5530138888888889, "percent_done": 17.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12152.837267416284, "avg_wps": 12000.64223685579, "eta_in_seconds": 14819.206879930676, "at": "2026-05-01T18:55:23.078889"} |
| {"lr": 1.9090068990230223e-06, "step": 540, "loss": 1.6546475887298584, "prob_real_tokens": 0.7090138888888889, "percent_done": 18.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12019.18947752386, "avg_wps": 12001.965744943547, "eta_in_seconds": 14757.582529730267, "at": "2026-05-01T18:56:22.722841"} |
| {"lr": 1.9043575432689348e-06, "step": 550, "loss": 1.4221340417861938, "prob_real_tokens": 0.44966666666666666, "percent_done": 18.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12237.85844246397, "avg_wps": 12004.664537730045, "eta_in_seconds": 14694.288161537863, "at": "2026-05-01T18:57:21.975410"} |
| {"lr": 1.8995983011334088e-06, "step": 560, "loss": 1.6760753393173218, "prob_real_tokens": 0.5751111111111111, "percent_done": 18.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12154.204346464861, "avg_wps": 12007.096845067685, "eta_in_seconds": 14631.34696645396, "at": "2026-05-01T18:58:21.276069"} |
| {"lr": 1.8947297509034442e-06, "step": 570, "loss": 1.6924524307250977, "prob_real_tokens": 0.6142222222222222, "percent_done": 19.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12060.98638289011, "avg_wps": 12009.284523400669, "eta_in_seconds": 14568.728025310917, "at": "2026-05-01T18:59:20.624993"} |
| {"lr": 1.8897524841478707e-06, "step": 580, "loss": 1.5508928298950195, "prob_real_tokens": 0.5081111111111111, "percent_done": 19.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12185.191905646658, "avg_wps": 12011.540003995691, "eta_in_seconds": 14506.050010409848, "at": "2026-05-01T19:00:19.930261"} |
| {"lr": 1.8846671056454687e-06, "step": 590, "loss": 1.4657219648361206, "prob_real_tokens": 0.5774583333333333, "percent_done": 19.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12143.536583266998, "avg_wps": 12013.594627049577, "eta_in_seconds": 14443.63701179876, "at": "2026-05-01T19:01:19.271951"} |
| {"lr": 1.8794742333114843e-06, "step": 600, "loss": 1.5696967840194702, "prob_real_tokens": 0.5064444444444445, "percent_done": 20.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12087.478006157375, "avg_wps": 12015.648030507278, "eta_in_seconds": 14381.246817588806, "at": "2026-05-01T19:02:18.593748"} |
| {"lr": 1.8741744981225464e-06, "step": 610, "loss": 1.557704210281372, "prob_real_tokens": 0.503125, "percent_done": 20.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12223.546358157771, "avg_wps": 12017.686554291939, "eta_in_seconds": 14318.895672856785, "at": "2026-05-01T19:03:17.899607"} |
| {"lr": 1.868768544039998e-06, "step": 620, "loss": 1.5956988334655762, "prob_real_tokens": 0.6478888888888888, "percent_done": 20.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12095.897402847766, "avg_wps": 12019.191799901559, "eta_in_seconds": 14257.198225375145, "at": "2026-05-01T19:04:17.350353"} |
| {"lr": 1.8632570279316504e-06, "step": 630, "loss": 1.718435287475586, "prob_real_tokens": 0.5778888888888889, "percent_done": 21.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12098.464284898728, "avg_wps": 12020.709224475357, "eta_in_seconds": 14195.501847142265, "at": "2026-05-01T19:05:16.782226"} |
| {"lr": 1.857640619491967e-06, "step": 640, "loss": 1.5155446529388428, "prob_real_tokens": 0.4937916666666667, "percent_done": 21.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12189.966893443972, "avg_wps": 12022.769981840222, "eta_in_seconds": 14133.182307958603, "at": "2026-05-01T19:06:16.026294"} |
| {"lr": 1.851920001160692e-06, "step": 650, "loss": 1.5322139263153076, "prob_real_tokens": 0.5694444444444444, "percent_done": 21.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12134.04437464144, "avg_wps": 12024.592443511187, "eta_in_seconds": 14071.162976613412, "at": "2026-05-01T19:07:15.327194"} |
| {"lr": 1.8460958680399258e-06, "step": 660, "loss": 1.5891509056091309, "prob_real_tokens": 0.525875, "percent_done": 22.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12126.822216439703, "avg_wps": 12026.595112737865, "eta_in_seconds": 14008.952527349646, "at": "2026-05-01T19:08:14.550573"} |
| {"lr": 1.8401689278096672e-06, "step": 670, "loss": 1.4961044788360596, "prob_real_tokens": 0.5313333333333333, "percent_done": 22.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12186.522013994309, "avg_wps": 12028.529706266949, "eta_in_seconds": 13946.841725185737, "at": "2026-05-01T19:09:13.776042"} |
| {"lr": 1.8341399006418213e-06, "step": 680, "loss": 1.4696369171142578, "prob_real_tokens": 0.5406805555555556, "percent_done": 22.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12142.680143294858, "avg_wps": 12030.238879204366, "eta_in_seconds": 13885.0110689612, "at": "2026-05-01T19:10:13.059678"} |
| {"lr": 1.8280095191126956e-06, "step": 690, "loss": 1.5702919960021973, "prob_real_tokens": 0.632125, "percent_done": 23.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12054.181929322049, "avg_wps": 12031.464913323505, "eta_in_seconds": 13823.753067327581, "at": "2026-05-01T19:11:12.491769"} |
| {"lr": 1.8217785281139843e-06, "step": 700, "loss": 1.5026861429214478, "prob_real_tokens": 0.5573333333333333, "percent_done": 23.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12177.064642055699, "avg_wps": 12032.460950487754, "eta_in_seconds": 13762.770615373338, "at": "2026-05-01T19:12:11.991648"} |
| {"lr": 1.8154476847622587e-06, "step": 710, "loss": 1.415621280670166, "prob_real_tokens": 0.5672222222222222, "percent_done": 23.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12138.536856183306, "avg_wps": 12033.834971007276, "eta_in_seconds": 13701.367884572124, "at": "2026-05-01T19:13:11.348484"} |
| {"lr": 1.80901775830697e-06, "step": 720, "loss": 1.539616346359253, "prob_real_tokens": 0.5898888888888889, "percent_done": 24.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12078.401207688583, "avg_wps": 12035.093281357118, "eta_in_seconds": 13640.11031424999, "at": "2026-05-01T19:14:10.733664"} |
| {"lr": 1.802489530036981e-06, "step": 730, "loss": 1.4959543943405151, "prob_real_tokens": 0.5255555555555556, "percent_done": 24.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12200.349701202624, "avg_wps": 12036.324053901351, "eta_in_seconds": 13578.896618941058, "at": "2026-05-01T19:15:10.116627"} |
| {"lr": 1.795863793185632e-06, "step": 740, "loss": 1.4252872467041016, "prob_real_tokens": 0.5442222222222223, "percent_done": 24.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11364.752154868675, "avg_wps": 12036.310717664655, "eta_in_seconds": 13519.09267024736, "at": "2026-05-01T19:16:09.944203"} |
| {"lr": 1.7891413528343563e-06, "step": 750, "loss": 1.593334674835205, "prob_real_tokens": 0.7221527777777778, "percent_done": 25.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12036.964923548165, "avg_wps": 12037.534386573632, "eta_in_seconds": 13457.905481100082, "at": "2026-05-01T19:17:09.311786"} |
| {"lr": 1.7823230258148561e-06, "step": 760, "loss": 1.4781315326690674, "prob_real_tokens": 0.517, "percent_done": 25.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12188.816580057919, "avg_wps": 12038.893264413528, "eta_in_seconds": 13396.580271771081, "at": "2026-05-01T19:18:08.615517"} |
| {"lr": 1.7754096406098527e-06, "step": 770, "loss": 1.5694516897201538, "prob_real_tokens": 0.5757777777777778, "percent_done": 25.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12160.106613076652, "avg_wps": 12040.176280578109, "eta_in_seconds": 13335.352926601063, "at": "2026-05-01T19:19:07.935367"} |
| {"lr": 1.7684020372524158e-06, "step": 780, "loss": 1.560455083847046, "prob_real_tokens": 0.5617777777777778, "percent_done": 26.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12087.049361600311, "avg_wps": 12041.31678219029, "eta_in_seconds": 13274.295734533896, "at": "2026-05-01T19:20:07.297501"} |
| {"lr": 1.761301067223895e-06, "step": 790, "loss": 1.5703930854797363, "prob_real_tokens": 0.5153333333333333, "percent_done": 26.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12198.752938705273, "avg_wps": 12042.08148874878, "eta_in_seconds": 13213.662451019769, "at": "2026-05-01T19:21:06.795321"} |
| {"lr": 1.7541075933504573e-06, "step": 800, "loss": 1.5794358253479004, "prob_real_tokens": 0.546875, "percent_done": 26.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12123.913272576576, "avg_wps": 12043.335963149928, "eta_in_seconds": 13152.50197160244, "at": "2026-05-01T19:22:06.090925"} |
| {"lr": 1.746822489698246e-06, "step": 810, "loss": 1.5084044933319092, "prob_real_tokens": 0.575, "percent_done": 27.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12074.509253721963, "avg_wps": 12043.60923523175, "eta_in_seconds": 13092.420795149274, "at": "2026-05-01T19:23:05.768614"} |
| {"lr": 1.7394466414671752e-06, "step": 820, "loss": 1.534797191619873, "prob_real_tokens": 0.6903611111111111, "percent_done": 27.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12099.008136574972, "avg_wps": 12044.797920193023, "eta_in_seconds": 13031.35187821272, "at": "2026-05-01T19:24:05.071103"} |
| {"lr": 1.7319809448833707e-06, "step": 830, "loss": 1.588489294052124, "prob_real_tokens": 0.6412222222222222, "percent_done": 27.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12069.248794393543, "avg_wps": 12045.258356260347, "eta_in_seconds": 12971.079189745777, "at": "2026-05-01T19:25:04.662673"} |
| {"lr": 1.72442630709027e-06, "step": 840, "loss": 1.457984209060669, "prob_real_tokens": 0.539, "percent_done": 28.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12131.345413263902, "avg_wps": 12045.412152869767, "eta_in_seconds": 12911.139778886523, "at": "2026-05-01T19:26:04.378306"} |
| {"lr": 1.7167836460383987e-06, "step": 850, "loss": 1.3119593858718872, "prob_real_tokens": 0.39952777777777776, "percent_done": 28.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12221.663559278853, "avg_wps": 12046.697184454712, "eta_in_seconds": 12849.995117313723, "at": "2026-05-01T19:27:03.614259"} |
| {"lr": 1.709053890373831e-06, "step": 860, "loss": 1.5186548233032227, "prob_real_tokens": 0.6879166666666666, "percent_done": 28.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12115.620590889775, "avg_wps": 12047.810784418127, "eta_in_seconds": 12789.045475322146, "at": "2026-05-01T19:28:02.911029"} |
| {"lr": 1.7012379793253514e-06, "step": 870, "loss": 1.393080711364746, "prob_real_tokens": 0.5452361111111111, "percent_done": 29.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12106.271932328324, "avg_wps": 12048.822550774863, "eta_in_seconds": 12728.214674398816, "at": "2026-05-01T19:29:02.240254"} |
| {"lr": 1.6933368625903318e-06, "step": 880, "loss": 1.4822235107421875, "prob_real_tokens": 0.6226805555555556, "percent_done": 29.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12163.622309458633, "avg_wps": 12049.884134929278, "eta_in_seconds": 12667.341718045147, "at": "2026-05-01T19:30:01.539189"} |
| {"lr": 1.6853515002193335e-06, "step": 890, "loss": 1.340285301208496, "prob_real_tokens": 0.46044444444444443, "percent_done": 29.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12226.169193904509, "avg_wps": 12050.971225014435, "eta_in_seconds": 12606.452804787776, "at": "2026-05-01T19:31:00.814907"} |
| {"lr": 1.6772828624994548e-06, "step": 900, "loss": 1.4219485521316528, "prob_real_tokens": 0.650125, "percent_done": 30.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12075.400043009056, "avg_wps": 12052.045292758854, "eta_in_seconds": 12545.58843143781, "at": "2026-05-01T19:32:00.086132"} |
| {"lr": 1.6691319298364329e-06, "step": 910, "loss": 1.6348272562026978, "prob_real_tokens": 0.6165694444444444, "percent_done": 30.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12145.819385618144, "avg_wps": 12053.018643559919, "eta_in_seconds": 12484.839229913858, "at": "2026-05-01T19:32:59.394491"} |
| {"lr": 1.6608996926355163e-06, "step": 920, "loss": 1.4275784492492676, "prob_real_tokens": 0.5882222222222222, "percent_done": 30.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12143.057567459055, "avg_wps": 12053.593918593211, "eta_in_seconds": 12424.510151199673, "at": "2026-05-01T19:33:58.872586"} |
| {"lr": 1.6525871511811227e-06, "step": 930, "loss": 1.2660408020019531, "prob_real_tokens": 0.41444444444444445, "percent_done": 31.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11743.856512956474, "avg_wps": 12052.023360229272, "eta_in_seconds": 12366.388244136688, "at": "2026-05-01T19:34:59.333813"} |
| {"lr": 1.6441953155152952e-06, "step": 940, "loss": 1.3833093643188477, "prob_real_tokens": 0.4965277777777778, "percent_done": 31.333333333333332, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12215.514619447078, "avg_wps": 12052.770453002158, "eta_in_seconds": 12305.884408763115, "at": "2026-05-01T19:35:58.730446"} |
| {"lr": 1.6357252053149755e-06, "step": 950, "loss": 1.4041342735290527, "prob_real_tokens": 0.45418055555555553, "percent_done": 31.666666666666668, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12214.691967511391, "avg_wps": 12053.680908947792, "eta_in_seconds": 12245.222112228996, "at": "2026-05-01T19:36:58.043688"} |
| {"lr": 1.6271778497681033e-06, "step": 960, "loss": 1.5199459791183472, "prob_real_tokens": 0.6706805555555555, "percent_done": 32.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12059.413849734194, "avg_wps": 12054.916949184038, "eta_in_seconds": 12184.239893078804, "at": "2026-05-01T19:37:57.192641"} |
| {"lr": 1.6185542874485627e-06, "step": 970, "loss": 1.4819310903549194, "prob_real_tokens": 0.703125, "percent_done": 32.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12135.918804608942, "avg_wps": 12055.8194916654, "eta_in_seconds": 12123.605541792112, "at": "2026-05-01T19:38:56.489043"} |
| {"lr": 1.6098555661899867e-06, "step": 980, "loss": 1.5395146608352661, "prob_real_tokens": 0.5861111111111111, "percent_done": 32.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12117.555458932871, "avg_wps": 12056.613391537798, "eta_in_seconds": 12063.088968422948, "at": "2026-05-01T19:39:55.829980"} |
| {"lr": 1.601082742958436e-06, "step": 990, "loss": 1.5307248830795288, "prob_real_tokens": 0.7369305555555555, "percent_done": 33.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12034.881122676807, "avg_wps": 12057.277068520008, "eta_in_seconds": 12002.709996425745, "at": "2026-05-01T19:40:55.227042"} |
| {"lr": 1.5922368837239695e-06, "step": 1000, "loss": 1.4094603061676025, "prob_real_tokens": 0.5426805555555556, "percent_done": 33.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11965.61395322889, "avg_wps": 12058.141040238337, "eta_in_seconds": 11942.139299869537, "at": "2026-05-01T19:41:54.518149"} |
| {"lr": 1.5833190633311204e-06, "step": 1010, "loss": 1.512271523475647, "prob_real_tokens": 0.6009722222222222, "percent_done": 33.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12158.911993787313, "avg_wps": 12058.636051754404, "eta_in_seconds": 11881.940825235726, "at": "2026-05-01T19:42:55.629223"} |
| {"lr": 1.574330365368293e-06, "step": 1020, "loss": 1.5038570165634155, "prob_real_tokens": 0.658125, "percent_done": 34.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12083.969433576178, "avg_wps": 12059.694613367808, "eta_in_seconds": 11821.194861930959, "at": "2026-05-01T19:43:54.806909"} |
| {"lr": 1.5652718820360994e-06, "step": 1030, "loss": 1.450741171836853, "prob_real_tokens": 0.5843472222222222, "percent_done": 34.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12202.559738277847, "avg_wps": 12060.74782591069, "eta_in_seconds": 11760.464777754349, "at": "2026-05-01T19:44:53.976932"} |
| {"lr": 1.5561447140146463e-06, "step": 1040, "loss": 1.428680419921875, "prob_real_tokens": 0.6365694444444444, "percent_done": 34.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11697.81849684584, "avg_wps": 12061.212829300208, "eta_in_seconds": 11700.315880106047, "at": "2026-05-01T19:45:53.439400"} |
| {"lr": 1.5469499703297948e-06, "step": 1050, "loss": 1.4217331409454346, "prob_real_tokens": 0.5695555555555556, "percent_done": 35.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12133.119076167844, "avg_wps": 12062.048076123934, "eta_in_seconds": 11639.814326218197, "at": "2026-05-01T19:46:52.704230"} |
| {"lr": 1.5376887682184024e-06, "step": 1060, "loss": 1.5705152750015259, "prob_real_tokens": 0.6775555555555556, "percent_done": 35.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12161.435658032013, "avg_wps": 12062.838325719502, "eta_in_seconds": 11579.36434430896, "at": "2026-05-01T19:47:51.984884"} |
| {"lr": 1.5283622329925734e-06, "step": 1070, "loss": 1.3451398611068726, "prob_real_tokens": 0.5385555555555556, "percent_done": 35.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12150.410053859501, "avg_wps": 12063.401439441684, "eta_in_seconds": 11519.13916631057, "at": "2026-05-01T19:48:51.379957"} |
| {"lr": 1.5189714979029202e-06, "step": 1080, "loss": 1.3972558975219727, "prob_real_tokens": 0.49125, "percent_done": 36.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12122.855688065749, "avg_wps": 12064.024356420688, "eta_in_seconds": 11458.862806955973, "at": "2026-05-01T19:49:50.736204"} |
| {"lr": 1.5095177040008667e-06, "step": 1090, "loss": 1.3031773567199707, "prob_real_tokens": 0.43633333333333335, "percent_done": 36.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12088.320870003969, "avg_wps": 12064.661560049082, "eta_in_seconds": 11398.579174022063, "at": "2026-05-01T19:50:50.080006"} |
| {"lr": 1.500002e-06, "step": 1100, "loss": 1.3764522075653076, "prob_real_tokens": 0.502, "percent_done": 36.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12219.505930219591, "avg_wps": 12065.56432168858, "eta_in_seconds": 11338.052357326853, "at": "2026-05-01T19:51:49.271890"} |
| {"lr": 1.490425542136493e-06, "step": 1110, "loss": 1.4342670440673828, "prob_real_tokens": 0.5205555555555555, "percent_done": 37.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12132.589213775063, "avg_wps": 12066.51399736976, "eta_in_seconds": 11277.490750821862, "at": "2026-05-01T19:52:48.431307"} |
| {"lr": 1.4807894940286107e-06, "step": 1120, "loss": 1.4680265188217163, "prob_real_tokens": 0.5394305555555555, "percent_done": 37.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12229.899022198982, "avg_wps": 12067.582222933726, "eta_in_seconds": 11216.828483070647, "at": "2026-05-01T19:53:47.512462"} |
| {"lr": 1.4710950265353237e-06, "step": 1130, "loss": 1.3665461540222168, "prob_real_tokens": 0.5505694444444444, "percent_done": 37.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12184.857088628678, "avg_wps": 12068.534795863774, "eta_in_seconds": 11156.283863567673, "at": "2026-05-01T19:54:46.648354"} |
| {"lr": 1.461343317614036e-06, "step": 1140, "loss": 1.4020676612854004, "prob_real_tokens": 0.5149027777777778, "percent_done": 38.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11999.558465038195, "avg_wps": 12069.188717621446, "eta_in_seconds": 11096.023364393335, "at": "2026-05-01T19:55:45.944322"} |
| {"lr": 1.4515355521774562e-06, "step": 1150, "loss": 1.4182010889053345, "prob_real_tokens": 0.5739166666666666, "percent_done": 38.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12186.248101030445, "avg_wps": 12069.624224925452, "eta_in_seconds": 11035.969100423481, "at": "2026-05-01T19:56:45.358314"} |
| {"lr": 1.4416729219496197e-06, "step": 1160, "loss": 1.4415645599365234, "prob_real_tokens": 0.6287916666666666, "percent_done": 38.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12150.411520456446, "avg_wps": 12070.437951351718, "eta_in_seconds": 10975.575247057552, "at": "2026-05-01T19:57:44.549512"} |
| {"lr": 1.4317566253210836e-06, "step": 1170, "loss": 1.324491024017334, "prob_real_tokens": 0.4829027777777778, "percent_done": 39.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12159.944052720519, "avg_wps": 12071.432954464812, "eta_in_seconds": 10915.025622642957, "at": "2026-05-01T19:58:43.628357"} |
| {"lr": 1.421787867203315e-06, "step": 1180, "loss": 1.3003603219985962, "prob_real_tokens": 0.5095555555555555, "percent_done": 39.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12244.98063667192, "avg_wps": 12072.281025597138, "eta_in_seconds": 10854.61808933646, "at": "2026-05-01T19:59:42.782636"} |
| {"lr": 1.4117678588822814e-06, "step": 1190, "loss": 1.2646369934082031, "prob_real_tokens": 0.5386666666666666, "percent_done": 39.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12205.591398892488, "avg_wps": 12073.063759161363, "eta_in_seconds": 10794.277459282835, "at": "2026-05-01T20:00:41.966601"} |
| {"lr": 1.401697817871271e-06, "step": 1200, "loss": 1.2742161750793457, "prob_real_tokens": 0.5404583333333334, "percent_done": 40.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12091.949138918284, "avg_wps": 12073.587760281252, "eta_in_seconds": 10734.17467725277, "at": "2026-05-01T20:01:41.297310"} |
| {"lr": 1.3915789677629554e-06, "step": 1210, "loss": 1.3747882843017578, "prob_real_tokens": 0.6297916666666666, "percent_done": 40.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12193.888840098241, "avg_wps": 12074.471471552166, "eta_in_seconds": 10673.759120939192, "at": "2026-05-01T20:02:40.407736"} |
| {"lr": 1.3814125380807112e-06, "step": 1220, "loss": 1.2545289993286133, "prob_real_tokens": 0.4687777777777778, "percent_done": 40.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12231.035306754107, "avg_wps": 12075.298605757072, "eta_in_seconds": 10613.402134742893, "at": "2026-05-01T20:03:39.543525"} |
| {"lr": 1.3711997641292251e-06, "step": 1230, "loss": 1.490968108177185, "prob_real_tokens": 0.6463333333333333, "percent_done": 41.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12104.431392330784, "avg_wps": 12076.281000687894, "eta_in_seconds": 10552.91773955415, "at": "2026-05-01T20:04:38.576896"} |
| {"lr": 1.3609418868443924e-06, "step": 1240, "loss": 1.3621050119400024, "prob_real_tokens": 0.47556944444444443, "percent_done": 41.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12233.840267445565, "avg_wps": 12077.129462884866, "eta_in_seconds": 10492.559543178928, "at": "2026-05-01T20:05:37.682654"} |
| {"lr": 1.3506401526425338e-06, "step": 1250, "loss": 1.3833402395248413, "prob_real_tokens": 0.49977777777777777, "percent_done": 41.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12237.369477195587, "avg_wps": 12078.031972574756, "eta_in_seconds": 10432.163144302369, "at": "2026-05-01T20:06:36.746417"} |
| {"lr": 1.3402958132689453e-06, "step": 1260, "loss": 1.3808565139770508, "prob_real_tokens": 0.5717916666666667, "percent_done": 42.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12130.354746831019, "avg_wps": 12078.776759925602, "eta_in_seconds": 10371.911203429812, "at": "2026-05-01T20:07:35.901030"} |
| {"lr": 1.3299101256458018e-06, "step": 1270, "loss": 1.3875703811645508, "prob_real_tokens": 0.5767916666666667, "percent_done": 42.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12223.979790737078, "avg_wps": 12079.707982271546, "eta_in_seconds": 10311.507544951177, "at": "2026-05-01T20:08:34.930038"} |
| {"lr": 1.3194843517194297e-06, "step": 1280, "loss": 1.3721925020217896, "prob_real_tokens": 0.5182222222222223, "percent_done": 42.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12205.116847294972, "avg_wps": 12080.486350164068, "eta_in_seconds": 10251.242906153202, "at": "2026-05-01T20:09:34.045886"} |
| {"lr": 1.30901975830697e-06, "step": 1290, "loss": 1.3715840578079224, "prob_real_tokens": 0.41665277777777776, "percent_done": 43.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12204.404595863878, "avg_wps": 12081.275507333023, "eta_in_seconds": 10190.976931638497, "at": "2026-05-01T20:10:33.147553"} |
| {"lr": 1.2985176169424507e-06, "step": 1300, "loss": 1.3929275274276733, "prob_real_tokens": 0.6309027777777778, "percent_done": 43.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12182.998475904267, "avg_wps": 12082.101331830672, "eta_in_seconds": 10130.688084657375, "at": "2026-05-01T20:11:32.218788"} |
| {"lr": 1.2879792037222837e-06, "step": 1310, "loss": 1.4336661100387573, "prob_real_tokens": 0.6594444444444445, "percent_done": 43.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12158.187992493557, "avg_wps": 12082.84551361016, "eta_in_seconds": 10070.47552357921, "at": "2026-05-01T20:12:31.333963"} |
| {"lr": 1.2774057991502097e-06, "step": 1320, "loss": 1.5263334512710571, "prob_real_tokens": 0.797375, "percent_done": 44.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12032.244786447, "avg_wps": 12083.5104008774, "eta_in_seconds": 10010.336068500172, "at": "2026-05-01T20:13:30.492997"} |
| {"lr": 1.2667986879817056e-06, "step": 1330, "loss": 1.3969571590423584, "prob_real_tokens": 0.6356666666666667, "percent_done": 44.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12196.625557277232, "avg_wps": 12084.356451879914, "eta_in_seconds": 9950.054061943427, "at": "2026-05-01T20:14:29.526974"} |
| {"lr": 1.2561591590678767e-06, "step": 1340, "loss": 1.1773649454116821, "prob_real_tokens": 0.41252777777777777, "percent_done": 44.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12251.430158951807, "avg_wps": 12085.16131182132, "eta_in_seconds": 9889.81420405943, "at": "2026-05-01T20:15:28.582689"} |
| {"lr": 1.2454885051988508e-06, "step": 1350, "loss": 1.404482126235962, "prob_real_tokens": 0.5608888888888889, "percent_done": 45.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12141.054509209982, "avg_wps": 12085.814070260538, "eta_in_seconds": 9829.706076012719, "at": "2026-05-01T20:16:27.729182"} |
| {"lr": 1.2347880229466927e-06, "step": 1360, "loss": 1.3764160871505737, "prob_real_tokens": 0.6642222222222223, "percent_done": 45.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12206.111869527655, "avg_wps": 12086.584548767156, "eta_in_seconds": 9769.509287224097, "at": "2026-05-01T20:17:26.790179"} |
| {"lr": 1.224059012507861e-06, "step": 1370, "loss": 1.337385892868042, "prob_real_tokens": 0.6038888888888889, "percent_done": 45.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12205.808955152213, "avg_wps": 12087.290296413883, "eta_in_seconds": 9709.372168782855, "at": "2026-05-01T20:18:25.887684"} |
| {"lr": 1.213302777545224e-06, "step": 1380, "loss": 1.3679434061050415, "prob_real_tokens": 0.6549027777777777, "percent_done": 46.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12125.306957204102, "avg_wps": 12088.155375825196, "eta_in_seconds": 9649.114887558895, "at": "2026-05-01T20:19:24.870322"} |
| {"lr": 1.2025206250296518e-06, "step": 1390, "loss": 1.3085877895355225, "prob_real_tokens": 0.5956666666666667, "percent_done": 46.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12206.272213026792, "avg_wps": 12088.940157679855, "eta_in_seconds": 9588.929921731675, "at": "2026-05-01T20:20:23.898897"} |
| {"lr": 1.191713865081211e-06, "step": 1400, "loss": 1.2333073616027832, "prob_real_tokens": 0.5909166666666666, "percent_done": 46.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12141.866295483014, "avg_wps": 12089.651065166661, "eta_in_seconds": 9528.810995374406, "at": "2026-05-01T20:21:22.970512"} |
| {"lr": 1.1808838108099728e-06, "step": 1410, "loss": 1.3174691200256348, "prob_real_tokens": 0.5292222222222223, "percent_done": 47.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12184.600948824858, "avg_wps": 12090.523152626258, "eta_in_seconds": 9468.572910770457, "at": "2026-05-01T20:22:21.923274"} |
| {"lr": 1.1700317781564602e-06, "step": 1420, "loss": 1.3648560047149658, "prob_real_tokens": 0.517, "percent_done": 47.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12230.6088037979, "avg_wps": 12091.26643509017, "eta_in_seconds": 9408.443740008583, "at": "2026-05-01T20:23:20.958582"} |
| {"lr": 1.15915908573175e-06, "step": 1430, "loss": 1.350665807723999, "prob_real_tokens": 0.5833472222222222, "percent_done": 47.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12199.20925471081, "avg_wps": 12091.842374043312, "eta_in_seconds": 9348.451336303791, "at": "2026-05-01T20:24:20.103810"} |
| {"lr": 1.1482670546572512e-06, "step": 1440, "loss": 1.3492231369018555, "prob_real_tokens": 0.6576666666666666, "percent_done": 48.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12146.470586034944, "avg_wps": 12092.407764950283, "eta_in_seconds": 9288.472749451796, "at": "2026-05-01T20:25:19.252924"} |
| {"lr": 1.137357008404177e-06, "step": 1450, "loss": 1.3289828300476074, "prob_real_tokens": 0.6436944444444445, "percent_done": 48.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12191.540192695764, "avg_wps": 12093.206014034622, "eta_in_seconds": 9228.32207360761, "at": "2026-05-01T20:26:18.228293"} |
| {"lr": 1.126430272632733e-06, "step": 1460, "loss": 1.3213653564453125, "prob_real_tokens": 0.44519444444444445, "percent_done": 48.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12232.528053629847, "avg_wps": 12094.059276328513, "eta_in_seconds": 9168.137634071585, "at": "2026-05-01T20:27:17.157946"} |
| {"lr": 1.1154881750310386e-06, "step": 1470, "loss": 1.217204213142395, "prob_real_tokens": 0.4552222222222222, "percent_done": 49.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12184.284353277482, "avg_wps": 12094.663599875033, "eta_in_seconds": 9108.149151096539, "at": "2026-05-01T20:28:16.258061"} |
| {"lr": 1.1045320451538005e-06, "step": 1480, "loss": 1.4656403064727783, "prob_real_tokens": 0.6771805555555556, "percent_done": 49.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12179.662162769857, "avg_wps": 12095.255931727115, "eta_in_seconds": 9048.175633301606, "at": "2026-05-01T20:29:15.360999"} |
| {"lr": 1.093563214260764e-06, "step": 1490, "loss": 1.2531245946884155, "prob_real_tokens": 0.5117777777777778, "percent_done": 49.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12200.888456429673, "avg_wps": 12095.943514656437, "eta_in_seconds": 8988.13721048432, "at": "2026-05-01T20:30:14.387886"} |
| {"lr": 1.0825830151549506e-06, "step": 1500, "loss": 1.32451331615448, "prob_real_tokens": 0.4811111111111111, "percent_done": 50.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12133.74453907966, "avg_wps": 12096.535704726064, "eta_in_seconds": 8928.17601966858, "at": "2026-05-01T20:31:13.479495"} |
| {"lr": 1.0715927820207137e-06, "step": 1510, "loss": 1.2765754461288452, "prob_real_tokens": 0.5980138888888888, "percent_done": 50.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12198.997847047156, "avg_wps": 12096.672316043445, "eta_in_seconds": 8868.55468984787, "at": "2026-05-01T20:32:14.905416"} |
| {"lr": 1.0605938502616233e-06, "step": 1520, "loss": 1.2733150720596313, "prob_real_tokens": 0.4403888888888889, "percent_done": 50.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12219.116322761205, "avg_wps": 12097.21459510598, "eta_in_seconds": 8808.639308019689, "at": "2026-05-01T20:33:14.024942"} |
| {"lr": 1.0495875563382036e-06, "step": 1530, "loss": 1.2976866960525513, "prob_real_tokens": 0.6182222222222222, "percent_done": 51.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12095.279226667635, "avg_wps": 12097.615992466757, "eta_in_seconds": 8748.831180119048, "at": "2026-05-01T20:34:13.244599"} |
| {"lr": 1.038575237605543e-06, "step": 1540, "loss": 1.2257877588272095, "prob_real_tokens": 0.5233333333333333, "percent_done": 51.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12244.326774236004, "avg_wps": 12097.828694958303, "eta_in_seconds": 8689.162547309677, "at": "2026-05-01T20:35:12.602877"} |
| {"lr": 1.0275582321507926e-06, "step": 1550, "loss": 1.2989863157272339, "prob_real_tokens": 0.5987777777777777, "percent_done": 51.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12168.937424785288, "avg_wps": 12098.117096933776, "eta_in_seconds": 8629.442016763072, "at": "2026-05-01T20:36:11.905859"} |
| {"lr": 1.0165378786305797e-06, "step": 1560, "loss": 1.2371087074279785, "prob_real_tokens": 0.4182222222222222, "percent_done": 52.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12188.39842754709, "avg_wps": 12098.442971117001, "eta_in_seconds": 8569.697790659391, "at": "2026-05-01T20:37:11.172997"} |
| {"lr": 1.005515516108348e-06, "step": 1570, "loss": 1.3413125276565552, "prob_real_tokens": 0.6266666666666667, "percent_done": 52.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12190.473237906988, "avg_wps": 12098.959531126027, "eta_in_seconds": 8509.822661620037, "at": "2026-05-01T20:38:10.289516"} |
| {"lr": 9.944924838916522e-07, "step": 1580, "loss": 1.3670834302902222, "prob_real_tokens": 0.6400138888888889, "percent_done": 52.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12117.233586276348, "avg_wps": 12099.251120097855, "eta_in_seconds": 8450.109761766242, "at": "2026-05-01T20:39:09.575976"} |
| {"lr": 9.834701213694206e-07, "step": 1590, "loss": 1.3309540748596191, "prob_real_tokens": 0.5382361111111111, "percent_done": 53.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12138.389508961185, "avg_wps": 12099.339026616664, "eta_in_seconds": 8390.540985476297, "at": "2026-05-01T20:40:09.018991"} |
| {"lr": 9.724497678492077e-07, "step": 1600, "loss": 1.2217007875442505, "prob_real_tokens": 0.4652222222222222, "percent_done": 53.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12213.62540244131, "avg_wps": 12099.46976177591, "eta_in_seconds": 8330.943585515022, "at": "2026-05-01T20:41:08.429372"} |
| {"lr": 9.61432762394457e-07, "step": 1610, "loss": 1.292655110359192, "prob_real_tokens": 0.5385694444444444, "percent_done": 53.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12179.736829018719, "avg_wps": 12099.918685918321, "eta_in_seconds": 8271.129963581607, "at": "2026-05-01T20:42:07.583950"} |
| {"lr": 9.504204436617963e-07, "step": 1620, "loss": 1.1626288890838623, "prob_real_tokens": 0.4756666666666667, "percent_done": 54.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12162.151229574252, "avg_wps": 12100.065102736244, "eta_in_seconds": 8211.526066709448, "at": "2026-05-01T20:43:06.975204"} |
| {"lr": 9.394141497383771e-07, "step": 1630, "loss": 1.3697890043258667, "prob_real_tokens": 0.6221111111111111, "percent_done": 54.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12192.516757802725, "avg_wps": 12099.702075363612, "eta_in_seconds": 8152.266839763139, "at": "2026-05-01T20:44:06.773863"} |
| {"lr": 9.284152179792868e-07, "step": 1640, "loss": 1.252271056175232, "prob_real_tokens": 0.5298888888888889, "percent_done": 54.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12159.929363745825, "avg_wps": 12099.779613902312, "eta_in_seconds": 8092.709381871107, "at": "2026-05-01T20:45:06.220627"} |
| {"lr": 9.174249848450497e-07, "step": 1650, "loss": 1.4715330600738525, "prob_real_tokens": 0.6214583333333333, "percent_done": 55.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12116.684205686237, "avg_wps": 12099.893583989478, "eta_in_seconds": 8033.128500288183, "at": "2026-05-01T20:46:05.637220"} |
| {"lr": 9.064447857392363e-07, "step": 1660, "loss": 1.2795664072036743, "prob_real_tokens": 0.6009027777777778, "percent_done": 55.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12167.798430688248, "avg_wps": 12100.24073322627, "eta_in_seconds": 7973.395085857575, "at": "2026-05-01T20:47:04.863186"} |
| {"lr": 8.954759548461998e-07, "step": 1670, "loss": 1.2674802541732788, "prob_real_tokens": 0.4533055555555556, "percent_done": 55.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12183.68316198951, "avg_wps": 12100.332848136806, "eta_in_seconds": 7913.831892214849, "at": "2026-05-01T20:48:04.294909"} |
| {"lr": 8.845198249689618e-07, "step": 1680, "loss": 1.240209698677063, "prob_real_tokens": 0.4862222222222222, "percent_done": 56.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12080.636387631159, "avg_wps": 12100.357239485054, "eta_in_seconds": 7854.313564385686, "at": "2026-05-01T20:49:03.782480"} |
| {"lr": 8.735777273672671e-07, "step": 1690, "loss": 1.2173148393630981, "prob_real_tokens": 0.44666666666666666, "percent_done": 56.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12271.343608394824, "avg_wps": 12100.483391327827, "eta_in_seconds": 7794.729925219124, "at": "2026-05-01T20:50:03.186401"} |
| {"lr": 8.626509915958231e-07, "step": 1700, "loss": 1.3892433643341064, "prob_real_tokens": 0.6024583333333333, "percent_done": 56.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12131.83130340645, "avg_wps": 12100.927886444637, "eta_in_seconds": 7734.944037213045, "at": "2026-05-01T20:51:02.320096"} |
| {"lr": 8.517409453427488e-07, "step": 1710, "loss": 1.3169645071029663, "prob_real_tokens": 0.53, "percent_done": 57.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12153.197716007686, "avg_wps": 12101.334807650124, "eta_in_seconds": 7675.186372108627, "at": "2026-05-01T20:52:01.483950"} |
| {"lr": 8.4084891426825e-07, "step": 1720, "loss": 1.2941845655441284, "prob_real_tokens": 0.63325, "percent_done": 57.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12189.586055647647, "avg_wps": 12101.003823207559, "eta_in_seconds": 7615.897106259368, "at": "2026-05-01T20:53:01.265796"} |
| {"lr": 8.299762218435402e-07, "step": 1730, "loss": 1.309218406677246, "prob_real_tokens": 0.5442222222222223, "percent_done": 57.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12176.418503485798, "avg_wps": 12101.36470946411, "eta_in_seconds": 7556.172563619007, "at": "2026-05-01T20:54:00.461957"} |
| {"lr": 8.191241891900278e-07, "step": 1740, "loss": 1.2055740356445312, "prob_real_tokens": 0.4740138888888889, "percent_done": 58.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12156.98102585973, "avg_wps": 12101.793755155222, "eta_in_seconds": 7496.409361740639, "at": "2026-05-01T20:54:59.595866"} |
| {"lr": 8.082941349187896e-07, "step": 1750, "loss": 1.2302355766296387, "prob_real_tokens": 0.4792222222222222, "percent_done": 58.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12216.752017560995, "avg_wps": 12102.016079848023, "eta_in_seconds": 7436.777426685606, "at": "2026-05-01T20:55:58.903449"} |
| {"lr": 7.974873749703486e-07, "step": 1760, "loss": 1.4238805770874023, "prob_real_tokens": 0.6418055555555555, "percent_done": 58.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12139.924144641347, "avg_wps": 12102.387746328099, "eta_in_seconds": 7377.05664959279, "at": "2026-05-01T20:56:58.079857"} |
| {"lr": 7.867052224547763e-07, "step": 1770, "loss": 1.345263123512268, "prob_real_tokens": 0.6719027777777777, "percent_done": 59.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12103.349553834356, "avg_wps": 12102.457204701152, "eta_in_seconds": 7317.522260322409, "at": "2026-05-01T20:57:57.516038"} |
| {"lr": 7.759489874921392e-07, "step": 1780, "loss": 1.2554117441177368, "prob_real_tokens": 0.6669166666666667, "percent_done": 59.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12150.82169263073, "avg_wps": 12102.852079334321, "eta_in_seconds": 7257.793404745252, "at": "2026-05-01T20:58:56.666826"} |
| {"lr": 7.652199770533075e-07, "step": 1790, "loss": 1.1241222620010376, "prob_real_tokens": 0.4817777777777778, "percent_done": 59.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12231.865611526271, "avg_wps": 12103.35650655719, "eta_in_seconds": 7198.003293780642, "at": "2026-05-01T20:59:55.716719"} |
| {"lr": 7.545194948011495e-07, "step": 1800, "loss": 1.3100450038909912, "prob_real_tokens": 0.5496666666666666, "percent_done": 60.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12126.846077984415, "avg_wps": 12103.49767369635, "eta_in_seconds": 7138.432404359181, "at": "2026-05-01T21:00:55.082988"} |
| {"lr": 7.438488409321232e-07, "step": 1810, "loss": 1.2452832460403442, "prob_real_tokens": 0.6261388888888889, "percent_done": 60.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12186.190566119965, "avg_wps": 12103.858755570727, "eta_in_seconds": 7078.734288812342, "at": "2026-05-01T21:01:54.252967"} |
| {"lr": 7.332093120182945e-07, "step": 1820, "loss": 1.259029746055603, "prob_real_tokens": 0.583125, "percent_done": 60.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12135.581813136725, "avg_wps": 12104.007730603667, "eta_in_seconds": 7019.162734437775, "at": "2026-05-01T21:02:53.608827"} |
| {"lr": 7.226022008497907e-07, "step": 1830, "loss": 1.337355375289917, "prob_real_tokens": 0.557, "percent_done": 61.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12140.006132918383, "avg_wps": 12104.42117534399, "eta_in_seconds": 6959.4405861878, "at": "2026-05-01T21:03:52.725210"} |
| {"lr": 7.120287962777166e-07, "step": 1840, "loss": 1.3332364559173584, "prob_real_tokens": 0.6719305555555556, "percent_done": 61.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12127.10661320495, "avg_wps": 12104.3199868857, "eta_in_seconds": 6900.0158695811815, "at": "2026-05-01T21:04:52.302504"} |
| {"lr": 7.014903830575498e-07, "step": 1850, "loss": 1.254623532295227, "prob_real_tokens": 0.5415555555555556, "percent_done": 61.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12212.961548629495, "avg_wps": 12104.439611162463, "eta_in_seconds": 6840.465371370316, "at": "2026-05-01T21:05:51.680814"} |
| {"lr": 6.909882416930302e-07, "step": 1860, "loss": 1.3134076595306396, "prob_real_tokens": 0.6556805555555556, "percent_done": 62.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12099.707169268471, "avg_wps": 12104.823614764398, "eta_in_seconds": 6780.767949388874, "at": "2026-05-01T21:06:50.816353"} |
| {"lr": 6.805236482805706e-07, "step": 1870, "loss": 1.3353734016418457, "prob_real_tokens": 0.6570138888888889, "percent_done": 62.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12131.387324019195, "avg_wps": 12105.216828719309, "eta_in_seconds": 6721.069201088207, "at": "2026-05-01T21:07:49.939148"} |
| {"lr": 6.700978743541983e-07, "step": 1880, "loss": 1.2248189449310303, "prob_real_tokens": 0.476, "percent_done": 62.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12198.738155820265, "avg_wps": 12105.45614634203, "eta_in_seconds": 6661.459016921673, "at": "2026-05-01T21:08:49.201185"} |
| {"lr": 6.597121867310547e-07, "step": 1890, "loss": 1.2324867248535156, "prob_real_tokens": 0.5198888888888888, "percent_done": 63.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12095.392102085323, "avg_wps": 12105.826925901014, "eta_in_seconds": 6601.779497525049, "at": "2026-05-01T21:09:48.338190"} |
| {"lr": 6.493678473574666e-07, "step": 1900, "loss": 1.1928637027740479, "prob_real_tokens": 0.6257916666666666, "percent_done": 63.333333333333336, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12143.104930233516, "avg_wps": 12106.105251624978, "eta_in_seconds": 6542.153595547927, "at": "2026-05-01T21:10:47.558211"} |
| {"lr": 6.390661131556078e-07, "step": 1910, "loss": 1.212174654006958, "prob_real_tokens": 0.49733333333333335, "percent_done": 63.666666666666664, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12231.708558363465, "avg_wps": 12106.423621974998, "eta_in_seconds": 6482.5089927918, "at": "2026-05-01T21:11:46.737299"} |
| {"lr": 6.288082358707748e-07, "step": 1920, "loss": 1.324737548828125, "prob_real_tokens": 0.6077777777777778, "percent_done": 64.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12093.936025581694, "avg_wps": 12106.586256658515, "eta_in_seconds": 6422.950148910284, "at": "2026-05-01T21:12:46.060113"} |
| {"lr": 6.185954619192888e-07, "step": 1930, "loss": 1.188004493713379, "prob_real_tokens": 0.5614444444444444, "percent_done": 64.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12205.194785669199, "avg_wps": 12106.713343202779, "eta_in_seconds": 6363.411589591615, "at": "2026-05-01T21:13:45.415676"} |
| {"lr": 6.084290322370451e-07, "step": 1940, "loss": 1.3539929389953613, "prob_real_tokens": 0.6304444444444445, "percent_done": 64.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12132.008709306294, "avg_wps": 12106.96901723369, "eta_in_seconds": 6303.807327115659, "at": "2026-05-01T21:14:44.649876"} |
| {"lr": 5.983101821287294e-07, "step": 1950, "loss": 1.1857472658157349, "prob_real_tokens": 0.5253472222222222, "percent_done": 65.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12128.805479897479, "avg_wps": 12107.137502220527, "eta_in_seconds": 6244.250549408106, "at": "2026-05-01T21:15:43.961824"} |
| {"lr": 5.88240141117719e-07, "step": 1960, "loss": 1.2169647216796875, "prob_real_tokens": 0.5725555555555556, "percent_done": 65.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12198.786939476762, "avg_wps": 12107.464250923864, "eta_in_seconds": 6184.614585526135, "at": "2026-05-01T21:16:43.119919"} |
| {"lr": 5.782201327966854e-07, "step": 1970, "loss": 1.2282533645629883, "prob_real_tokens": 0.577, "percent_done": 65.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12198.21634293388, "avg_wps": 12107.722689829081, "eta_in_seconds": 6125.016396543096, "at": "2026-05-01T21:17:42.341080"} |
| {"lr": 5.682513746789166e-07, "step": 1980, "loss": 1.329833984375, "prob_real_tokens": 0.6243472222222223, "percent_done": 66.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12107.24216533971, "avg_wps": 12107.95246992237, "eta_in_seconds": 6065.435108242613, "at": "2026-05-01T21:18:41.587373"} |
| {"lr": 5.583350780503806e-07, "step": 1990, "loss": 1.2157623767852783, "prob_real_tokens": 0.49922222222222223, "percent_done": 66.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12246.22649497369, "avg_wps": 12108.252425486227, "eta_in_seconds": 6005.821273343648, "at": "2026-05-01T21:19:40.762989"} |
| {"lr": 5.484724478225442e-07, "step": 2000, "loss": 1.3252580165863037, "prob_real_tokens": 0.6575277777777778, "percent_done": 66.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12101.856643258141, "avg_wps": 12108.26922835689, "eta_in_seconds": 5946.349444508553, "at": "2026-05-01T21:20:40.214913"} |
| {"lr": 5.386646823859642e-07, "step": 2010, "loss": 1.2083008289337158, "prob_real_tokens": 0.5136944444444445, "percent_done": 67.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12127.812792989856, "avg_wps": 12108.518836774605, "eta_in_seconds": 5886.764596138428, "at": "2026-05-01T21:21:41.006538"} |
| {"lr": 5.289129734646764e-07, "step": 2020, "loss": 1.2622307538986206, "prob_real_tokens": 0.7085833333333333, "percent_done": 67.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12160.722619165885, "avg_wps": 12108.76595992527, "eta_in_seconds": 5827.183400316994, "at": "2026-05-01T21:22:40.229398"} |
| {"lr": 5.19218505971389e-07, "step": 2030, "loss": 1.24886155128479, "prob_real_tokens": 0.5814722222222222, "percent_done": 67.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12167.362109759082, "avg_wps": 12108.995109327132, "eta_in_seconds": 5767.613197415921, "at": "2026-05-01T21:23:39.467299"} |
| {"lr": 5.095824578635076e-07, "step": 2040, "loss": 1.181299090385437, "prob_real_tokens": 0.45655555555555555, "percent_done": 68.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12158.380855271693, "avg_wps": 12109.272633560604, "eta_in_seconds": 5708.022446239696, "at": "2026-05-01T21:24:38.654671"} |
| {"lr": 5.000060000000002e-07, "step": 2050, "loss": 1.2296390533447266, "prob_real_tokens": 0.5774583333333333, "percent_done": 68.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12215.161334920764, "avg_wps": 12109.399366999887, "eta_in_seconds": 5648.50476287051, "at": "2026-05-01T21:25:37.991002"} |
| {"lr": 4.904902959991335e-07, "step": 2060, "loss": 1.2735894918441772, "prob_real_tokens": 0.5033611111111111, "percent_done": 68.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12196.834912290713, "avg_wps": 12109.682993821394, "eta_in_seconds": 5588.91591419294, "at": "2026-05-01T21:26:37.167011"} |
| {"lr": 4.810365020970802e-07, "step": 2070, "loss": 1.2515358924865723, "prob_real_tokens": 0.5506805555555555, "percent_done": 69.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12140.968601921457, "avg_wps": 12109.935883940325, "eta_in_seconds": 5529.3438909779425, "at": "2026-05-01T21:27:36.371003"} |
| {"lr": 4.7164576700742696e-07, "step": 2080, "loss": 1.1986793279647827, "prob_real_tokens": 0.4826666666666667, "percent_done": 69.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12248.40946846734, "avg_wps": 12110.183700058056, "eta_in_seconds": 5469.7766475402395, "at": "2026-05-01T21:28:35.578771"} |
| {"lr": 4.623192317815978e-07, "step": 2090, "loss": 1.248482584953308, "prob_real_tokens": 0.6214444444444445, "percent_done": 69.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12131.859571032006, "avg_wps": 12110.422346052732, "eta_in_seconds": 5410.215938617168, "at": "2026-05-01T21:29:34.793049"} |
| {"lr": 4.530580296702058e-07, "step": 2100, "loss": 1.238773226737976, "prob_real_tokens": 0.5923333333333334, "percent_done": 70.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11860.114611031178, "avg_wps": 12110.520474734734, "eta_in_seconds": 5350.719660248075, "at": "2026-05-01T21:30:34.148801"} |
| {"lr": 4.4386328598535376e-07, "step": 2110, "loss": 1.1015173196792603, "prob_real_tokens": 0.4518888888888889, "percent_done": 70.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12231.035306754107, "avg_wps": 12110.782447250498, "eta_in_seconds": 5291.152762350878, "at": "2026-05-01T21:31:33.333664"} |
| {"lr": 4.347361179639006e-07, "step": 2120, "loss": 1.1892691850662231, "prob_real_tokens": 0.5505555555555556, "percent_done": 70.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12158.437148860528, "avg_wps": 12110.998393949127, "eta_in_seconds": 5231.608323196196, "at": "2026-05-01T21:32:32.564768"} |
| {"lr": 4.256776346317072e-07, "step": 2130, "loss": 1.258378028869629, "prob_real_tokens": 0.5248888888888888, "percent_done": 71.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12084.7271785491, "avg_wps": 12111.020781879715, "eta_in_seconds": 5172.148667577287, "at": "2026-05-01T21:33:31.996593"} |
| {"lr": 4.166889366688803e-07, "step": 2140, "loss": 1.2664952278137207, "prob_real_tokens": 0.5695694444444445, "percent_done": 71.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12180.912943205964, "avg_wps": 12111.17602769073, "eta_in_seconds": 5112.633146312749, "at": "2026-05-01T21:34:31.287800"} |
| {"lr": 4.077711162760308e-07, "step": 2150, "loss": 1.2321288585662842, "prob_real_tokens": 0.523, "percent_done": 71.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12195.53259452046, "avg_wps": 12111.501905818795, "eta_in_seconds": 5053.047960187113, "at": "2026-05-01T21:35:30.397203"} |
| {"lr": 3.9892525704156405e-07, "step": 2160, "loss": 1.2148308753967285, "prob_real_tokens": 0.43568055555555557, "percent_done": 72.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12196.355623067651, "avg_wps": 12111.529024591144, "eta_in_seconds": 4993.589156018363, "at": "2026-05-01T21:36:29.820217"} |
| {"lr": 3.9015243381001345e-07, "step": 2170, "loss": 1.191974401473999, "prob_real_tokens": 0.501, "percent_done": 72.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12226.667164926785, "avg_wps": 12111.789668076834, "eta_in_seconds": 4934.035484244747, "at": "2026-05-01T21:37:28.994097"} |
| {"lr": 3.814537125514374e-07, "step": 2180, "loss": 1.154778003692627, "prob_real_tokens": 0.5930138888888888, "percent_done": 72.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12177.125527948487, "avg_wps": 12111.846343740983, "eta_in_seconds": 4874.5664636432575, "at": "2026-05-01T21:38:28.383850"} |
| {"lr": 3.7283015023189687e-07, "step": 2190, "loss": 1.3154877424240112, "prob_real_tokens": 0.7037916666666667, "percent_done": 73.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12046.177496237038, "avg_wps": 12112.045834456703, "eta_in_seconds": 4815.04122400937, "at": "2026-05-01T21:39:27.620327"} |
| {"lr": 3.642827946850251e-07, "step": 2200, "loss": 1.2440673112869263, "prob_real_tokens": 0.5525555555555556, "percent_done": 73.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12226.049409711022, "avg_wps": 12112.354889032766, "eta_in_seconds": 4755.474928509106, "at": "2026-05-01T21:40:26.736502"} |
| {"lr": 3.558126844847051e-07, "step": 2210, "loss": 1.255953311920166, "prob_real_tokens": 0.6403472222222222, "percent_done": 73.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12170.04916703299, "avg_wps": 12112.472524970408, "eta_in_seconds": 4695.985884198236, "at": "2026-05-01T21:41:26.056437"} |
| {"lr": 3.4742084881887744e-07, "step": 2220, "loss": 1.3122426271438599, "prob_real_tokens": 0.6990138888888889, "percent_done": 74.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12054.427803454835, "avg_wps": 12112.735226558118, "eta_in_seconds": 4636.442467335108, "at": "2026-05-01T21:42:25.217243"} |
| {"lr": 3.3910830736448386e-07, "step": 2230, "loss": 1.2551305294036865, "prob_real_tokens": 0.535, "percent_done": 74.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11874.835358938748, "avg_wps": 12112.875630803914, "eta_in_seconds": 4576.947843748358, "at": "2026-05-01T21:43:24.509711"} |
| {"lr": 3.3087607016356756e-07, "step": 2240, "loss": 1.1537933349609375, "prob_real_tokens": 0.5194583333333334, "percent_done": 74.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12182.286346402729, "avg_wps": 12113.20588810078, "eta_in_seconds": 4517.383796287435, "at": "2026-05-01T21:44:23.591728"} |
| {"lr": 3.2272513750054567e-07, "step": 2250, "loss": 1.167872667312622, "prob_real_tokens": 0.48033333333333333, "percent_done": 75.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12170.673047859429, "avg_wps": 12113.4114747399, "eta_in_seconds": 4457.868876377741, "at": "2026-05-01T21:45:22.808449"} |
| {"lr": 3.1465649978066685e-07, "step": 2260, "loss": 1.2355475425720215, "prob_real_tokens": 0.5322361111111111, "percent_done": 75.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12220.894479645216, "avg_wps": 12113.649595484378, "eta_in_seconds": 4398.3441637490705, "at": "2026-05-01T21:46:21.987676"} |
| {"lr": 3.0667113740966857e-07, "step": 2270, "loss": 1.231582522392273, "prob_real_tokens": 0.5717916666666667, "percent_done": 75.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12181.434750803668, "avg_wps": 12113.834576056579, "eta_in_seconds": 4338.840824513708, "at": "2026-05-01T21:47:21.222942"} |
| {"lr": 2.9877002067464855e-07, "step": 2280, "loss": 1.1817396879196167, "prob_real_tokens": 0.5411111111111111, "percent_done": 76.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12136.450908843643, "avg_wps": 12114.143734513762, "eta_in_seconds": 4279.29543648268, "at": "2026-05-01T21:48:20.318873"} |
| {"lr": 2.909541096261691e-07, "step": 2290, "loss": 1.274882435798645, "prob_real_tokens": 0.5426666666666666, "percent_done": 76.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12233.116235509146, "avg_wps": 12114.287207237812, "eta_in_seconds": 4219.810800709579, "at": "2026-05-01T21:49:19.596448"} |
| {"lr": 2.832243539616016e-07, "step": 2300, "loss": 1.200040578842163, "prob_real_tokens": 0.5042083333333334, "percent_done": 76.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12198.021228902928, "avg_wps": 12114.537829728913, "eta_in_seconds": 4160.29077694727, "at": "2026-05-01T21:50:18.753518"} |
| {"lr": 2.7558169290973025e-07, "step": 2310, "loss": 1.2079274654388428, "prob_real_tokens": 0.5847777777777777, "percent_done": 77.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12138.912558997794, "avg_wps": 12114.645085669461, "eta_in_seconds": 4100.821744977654, "at": "2026-05-01T21:51:18.070090"} |
| {"lr": 2.6802705511662956e-07, "step": 2320, "loss": 1.2281467914581299, "prob_real_tokens": 0.66025, "percent_done": 77.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12135.064901665513, "avg_wps": 12114.910505428858, "eta_in_seconds": 4041.301004911291, "at": "2026-05-01T21:52:17.205221"} |
| {"lr": 2.605613585328249e-07, "step": 2330, "loss": 1.291995644569397, "prob_real_tokens": 0.5695555555555556, "percent_done": 77.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12152.879326801303, "avg_wps": 12115.119244906471, "eta_in_seconds": 3981.8015014818293, "at": "2026-05-01T21:53:16.404148"} |
| {"lr": 2.531855103017544e-07, "step": 2340, "loss": 1.2208168506622314, "prob_real_tokens": 0.7255, "percent_done": 78.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12064.287387020358, "avg_wps": 12115.182646737476, "eta_in_seconds": 3922.3511015574136, "at": "2026-05-01T21:54:15.765837"} |
| {"lr": 2.4590040664954297e-07, "step": 2350, "loss": 1.360719084739685, "prob_real_tokens": 0.582125, "percent_done": 78.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12203.434015843335, "avg_wps": 12115.395979153904, "eta_in_seconds": 3862.8535196476796, "at": "2026-05-01T21:55:14.953809"} |
| {"lr": 2.3870693277610515e-07, "step": 2360, "loss": 1.196913719177246, "prob_real_tokens": 0.6195833333333334, "percent_done": 78.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11857.936069684278, "avg_wps": 12115.52555030279, "eta_in_seconds": 3803.384327710685, "at": "2026-05-01T21:56:14.236881"} |
| {"lr": 2.3160596274758465e-07, "step": 2370, "loss": 1.162071943283081, "prob_real_tokens": 0.5417777777777778, "percent_done": 79.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12137.550381061508, "avg_wps": 12115.719780360863, "eta_in_seconds": 3743.896427311475, "at": "2026-05-01T21:57:13.444142"} |
| {"lr": 2.2459835939014752e-07, "step": 2380, "loss": 1.3375706672668457, "prob_real_tokens": 0.7125694444444445, "percent_done": 79.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12129.268756925612, "avg_wps": 12115.94056786591, "eta_in_seconds": 3684.402358195361, "at": "2026-05-01T21:58:12.618361"} |
| {"lr": 2.1768497418514384e-07, "step": 2390, "loss": 1.2047615051269531, "prob_real_tokens": 0.48344444444444445, "percent_done": 79.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12193.6909103529, "avg_wps": 12115.851895129827, "eta_in_seconds": 3625.0030439588318, "at": "2026-05-01T21:59:12.151819"} |
| {"lr": 2.1086664716564414e-07, "step": 2400, "loss": 1.2549245357513428, "prob_real_tokens": 0.5517777777777778, "percent_done": 80.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12130.112586879317, "avg_wps": 12116.015916316725, "eta_in_seconds": 3565.5284953713417, "at": "2026-05-01T22:00:11.388941"} |
| {"lr": 2.041442068143681e-07, "step": 2410, "loss": 1.1935161352157593, "prob_real_tokens": 0.47273611111111113, "percent_done": 80.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12209.131968357884, "avg_wps": 12115.881376956624, "eta_in_seconds": 3506.1419535514233, "at": "2026-05-01T22:01:10.977482"} |
| {"lr": 1.9751846996301896e-07, "step": 2420, "loss": 1.1582236289978027, "prob_real_tokens": 0.4697777777777778, "percent_done": 80.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12214.498796245105, "avg_wps": 12116.089095643587, "eta_in_seconds": 3446.6567281198895, "at": "2026-05-01T22:02:10.161100"} |
| {"lr": 1.9099024169303017e-07, "step": 2430, "loss": 1.1771974563598633, "prob_real_tokens": 0.48469444444444443, "percent_done": 81.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11636.231815370744, "avg_wps": 12116.03422208218, "eta_in_seconds": 3387.2469529016516, "at": "2026-05-01T22:03:09.657210"} |
| {"lr": 1.8456031523774143e-07, "step": 2440, "loss": 1.212212085723877, "prob_real_tokens": 0.5343333333333333, "percent_done": 81.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12225.208511556124, "avg_wps": 12116.286788417543, "eta_in_seconds": 3327.7521986805023, "at": "2026-05-01T22:04:08.784770"} |
| {"lr": 1.7822947188601585e-07, "step": 2450, "loss": 1.1940064430236816, "prob_real_tokens": 0.47641666666666665, "percent_done": 81.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12209.832923865706, "avg_wps": 12116.595656586916, "eta_in_seconds": 3268.2447382381984, "at": "2026-05-01T22:05:07.842180"} |
| {"lr": 1.7199848088730457e-07, "step": 2460, "loss": 1.3196537494659424, "prob_real_tokens": 0.5853472222222222, "percent_done": 82.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12051.64774887037, "avg_wps": 12116.52483416175, "eta_in_seconds": 3208.8408625532943, "at": "2026-05-01T22:06:07.354596"} |
| {"lr": 1.658680993581791e-07, "step": 2470, "loss": 1.1612741947174072, "prob_real_tokens": 0.5403472222222222, "percent_done": 82.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12221.955389745104, "avg_wps": 12116.746250857801, "eta_in_seconds": 3149.3603323828356, "at": "2026-05-01T22:07:06.513235"} |
| {"lr": 1.5983907219033318e-07, "step": 2480, "loss": 1.1877046823501587, "prob_real_tokens": 0.5704444444444444, "percent_done": 82.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12185.558701355245, "avg_wps": 12116.861240219221, "eta_in_seconds": 3089.909115714411, "at": "2026-05-01T22:08:05.799216"} |
| {"lr": 1.5391213196007416e-07, "step": 2490, "loss": 1.171887755393982, "prob_real_tokens": 0.4722777777777778, "percent_done": 83.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12154.816819368343, "avg_wps": 12117.137400951346, "eta_in_seconds": 3030.4187189556033, "at": "2026-05-01T22:09:04.888609"} |
| {"lr": 1.4808799883930842e-07, "step": 2500, "loss": 1.2361663579940796, "prob_real_tokens": 0.5069583333333333, "percent_done": 83.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12220.394011644565, "avg_wps": 12117.219877603626, "eta_in_seconds": 2970.978521776199, "at": "2026-05-01T22:10:04.212905"} |
| {"lr": 1.4236738050803318e-07, "step": 2510, "loss": 1.2186540365219116, "prob_real_tokens": 0.575, "percent_done": 83.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12165.3324032585, "avg_wps": 12117.455639426544, "eta_in_seconds": 2911.5023029429976, "at": "2026-05-01T22:11:05.243588"} |
| {"lr": 1.3675097206834996e-07, "step": 2520, "loss": 1.2401750087738037, "prob_real_tokens": 0.5756388888888889, "percent_done": 84.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12105.133963224895, "avg_wps": 12117.693825208247, "eta_in_seconds": 2852.0278279440745, "at": "2026-05-01T22:12:04.372512"} |
| {"lr": 1.3123945596000222e-07, "step": 2530, "loss": 1.2108324766159058, "prob_real_tokens": 0.6653472222222222, "percent_done": 84.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12182.080930451704, "avg_wps": 12117.663016786952, "eta_in_seconds": 2792.617681571146, "at": "2026-05-01T22:13:03.833430"} |
| {"lr": 1.2583350187745377e-07, "step": 2540, "loss": 1.1912890672683716, "prob_real_tokens": 0.5886666666666667, "percent_done": 84.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12162.589626083298, "avg_wps": 12117.882344253105, "eta_in_seconds": 2733.150814565148, "at": "2026-05-01T22:14:02.982225"} |
| {"lr": 1.2053376668851588e-07, "step": 2550, "loss": 1.263681411743164, "prob_real_tokens": 0.5970138888888888, "percent_done": 85.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12108.565505142364, "avg_wps": 12117.732476351968, "eta_in_seconds": 2673.7675603277544, "at": "2026-05-01T22:15:02.590238"} |
| {"lr": 1.1534089435453136e-07, "step": 2560, "loss": 1.097151756286621, "prob_real_tokens": 0.4538611111111111, "percent_done": 85.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12226.073168372714, "avg_wps": 12117.900752993908, "eta_in_seconds": 2614.314198948443, "at": "2026-05-01T22:16:01.800140"} |
| {"lr": 1.1025551585212956e-07, "step": 2570, "loss": 1.2054604291915894, "prob_real_tokens": 0.6107777777777778, "percent_done": 85.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12119.343573504895, "avg_wps": 12117.758073953191, "eta_in_seconds": 2554.9280494836516, "at": "2026-05-01T22:17:01.400934"} |
| {"lr": 1.0527824909655581e-07, "step": 2580, "loss": 1.1822400093078613, "prob_real_tokens": 0.47431944444444446, "percent_done": 86.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11943.35386063846, "avg_wps": 12117.644488020795, "eta_in_seconds": 2495.5345100192135, "at": "2026-05-01T22:18:00.965523"} |
| {"lr": 1.0040969886659098e-07, "step": 2590, "loss": 1.2395758628845215, "prob_real_tokens": 0.6175694444444444, "percent_done": 86.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11981.464450598132, "avg_wps": 12117.010903087306, "eta_in_seconds": 2436.2444035169237, "at": "2026-05-01T22:19:01.191830"} |
| {"lr": 9.565045673106536e-08, "step": 2600, "loss": 1.0670342445373535, "prob_real_tokens": 0.4386388888888889, "percent_done": 86.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 11868.84471714181, "avg_wps": 12116.449607304938, "eta_in_seconds": 2376.9339149181656, "at": "2026-05-01T22:20:01.332449"} |
| {"lr": 9.100110097697769e-08, "step": 2610, "loss": 1.287595510482788, "prob_real_tokens": 0.6701388888888888, "percent_done": 87.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12071.570822579983, "avg_wps": 12116.57887459298, "eta_in_seconds": 2317.485842384141, "at": "2026-05-01T22:21:00.595632"} |
| {"lr": 8.64621965392296e-08, "step": 2620, "loss": 1.1764854192733765, "prob_real_tokens": 0.5576666666666666, "percent_done": 87.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12198.081831835167, "avg_wps": 12116.555352291247, "eta_in_seconds": 2258.0675121354693, "at": "2026-05-01T22:22:00.052673"} |
| {"lr": 8.203429493198081e-08, "step": 2630, "loss": 1.0971014499664307, "prob_real_tokens": 0.5827916666666667, "percent_done": 87.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12153.104300548124, "avg_wps": 12116.603567078919, "eta_in_seconds": 2198.635933949467, "at": "2026-05-01T22:22:59.417578"} |
| {"lr": 7.771793418163516e-08, "step": 2640, "loss": 1.2456961870193481, "prob_real_tokens": 0.4767777777777778, "percent_done": 88.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12022.778753079845, "avg_wps": 12116.464268339847, "eta_in_seconds": 2139.2379349253397, "at": "2026-05-01T22:23:59.025495"} |
| {"lr": 7.35136387614671e-08, "step": 2650, "loss": 1.1770060062408447, "prob_real_tokens": 0.5333472222222222, "percent_done": 88.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12177.775179048525, "avg_wps": 12116.313384430066, "eta_in_seconds": 2079.8405587943093, "at": "2026-05-01T22:24:58.652962"} |
| {"lr": 6.942191952789292e-08, "step": 2660, "loss": 1.232515573501587, "prob_real_tokens": 0.6088888888888889, "percent_done": 88.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12108.314990158859, "avg_wps": 12116.425994279823, "eta_in_seconds": 2020.397765112999, "at": "2026-05-01T22:25:57.933523"} |
| {"lr": 6.544327365839766e-08, "step": 2670, "loss": 1.2437316179275513, "prob_real_tokens": 0.5826805555555555, "percent_done": 89.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12079.994159807273, "avg_wps": 12116.359018100455, "eta_in_seconds": 1960.985141204984, "at": "2026-05-01T22:26:57.450061"} |
| {"lr": 6.157818459112448e-08, "step": 2680, "loss": 1.199005126953125, "prob_real_tokens": 0.49545833333333333, "percent_done": 89.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12192.64917711195, "avg_wps": 12116.480613363708, "eta_in_seconds": 1901.5422658777948, "at": "2026-05-01T22:27:56.717660"} |
| {"lr": 5.7827121966132736e-08, "step": 2690, "loss": 1.2351887226104736, "prob_real_tokens": 0.573625, "percent_done": 89.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12159.593485541376, "avg_wps": 12116.393652348544, "eta_in_seconds": 1842.1322912097419, "at": "2026-05-01T22:28:56.259810"} |
| {"lr": 5.419054156833217e-08, "step": 2700, "loss": 1.1971439123153687, "prob_real_tokens": 0.641, "percent_done": 90.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12091.390913664098, "avg_wps": 12116.495886208384, "eta_in_seconds": 1782.6936271720463, "at": "2026-05-01T22:29:55.552442"} |
| {"lr": 5.0668885272102414e-08, "step": 2710, "loss": 1.229919195175171, "prob_real_tokens": 0.47733333333333333, "percent_done": 90.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12205.562293314413, "avg_wps": 12116.399925679232, "eta_in_seconds": 1723.2841543755408, "at": "2026-05-01T22:30:55.106755"} |
| {"lr": 4.726258098760108e-08, "step": 2720, "loss": 1.145553469657898, "prob_real_tokens": 0.4727777777777778, "percent_done": 90.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12213.534513429453, "avg_wps": 12116.582892984408, "eta_in_seconds": 1663.8354376028567, "at": "2026-05-01T22:31:54.291034"} |
| {"lr": 4.39720426087685e-08, "step": 2730, "loss": 1.2468335628509521, "prob_real_tokens": 0.48597222222222225, "percent_done": 91.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12128.465960742215, "avg_wps": 12116.771480125788, "eta_in_seconds": 1604.3877720963824, "at": "2026-05-01T22:32:53.464788"} |
| {"lr": 4.0797669963037686e-08, "step": 2740, "loss": 1.1568303108215332, "prob_real_tokens": 0.5950138888888888, "percent_done": 91.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12215.802697339552, "avg_wps": 12116.937380873242, "eta_in_seconds": 1544.9448496407836, "at": "2026-05-01T22:33:52.667927"} |
| {"lr": 3.7739848762750823e-08, "step": 2750, "loss": 1.1542396545410156, "prob_real_tokens": 0.5262222222222223, "percent_done": 91.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12166.526814069746, "avg_wps": 12117.087620319675, "eta_in_seconds": 1485.5054749142041, "at": "2026-05-01T22:34:51.889777"} |
| {"lr": 3.479895055829255e-08, "step": 2760, "loss": 1.3438825607299805, "prob_real_tokens": 0.5406805555555556, "percent_done": 92.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12113.65718582138, "avg_wps": 12117.191344490062, "eta_in_seconds": 1426.0730485086856, "at": "2026-05-01T22:35:51.174202"} |
| {"lr": 3.1975332692943076e-08, "step": 2770, "loss": 1.1631507873535156, "prob_real_tokens": 0.5592222222222222, "percent_done": 92.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12165.3324032585, "avg_wps": 12117.316904778612, "eta_in_seconds": 1366.6391768188685, "at": "2026-05-01T22:36:50.427899"} |
| {"lr": 2.9269338259458795e-08, "step": 2780, "loss": 1.314759612083435, "prob_real_tokens": 0.6061111111111112, "percent_done": 92.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12130.579374665369, "avg_wps": 12117.36266471425, "eta_in_seconds": 1307.2151455964972, "at": "2026-05-01T22:37:49.824901"} |
| {"lr": 2.668129605838325e-08, "step": 2790, "loss": 1.148567795753479, "prob_real_tokens": 0.4846666666666667, "percent_done": 93.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12154.26109056109, "avg_wps": 12117.502703010918, "eta_in_seconds": 1247.7818549397173, "at": "2026-05-01T22:38:49.055718"} |
| {"lr": 2.421152055809489e-08, "step": 2800, "loss": 1.24281907081604, "prob_real_tokens": 0.6686944444444445, "percent_done": 93.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12135.001022267428, "avg_wps": 12117.43958749514, "eta_in_seconds": 1188.3698611429759, "at": "2026-05-01T22:39:48.564168"} |
| {"lr": 2.1860311856597248e-08, "step": 2810, "loss": 1.1883509159088135, "prob_real_tokens": 0.6662361111111111, "percent_done": 93.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12123.308290235938, "avg_wps": 12117.567437219604, "eta_in_seconds": 1128.9394567742468, "at": "2026-05-01T22:40:47.810365"} |
| {"lr": 1.9627955645054144e-08, "step": 2820, "loss": 1.1803470849990845, "prob_real_tokens": 0.514, "percent_done": 94.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12151.911056621513, "avg_wps": 12117.658881162417, "eta_in_seconds": 1069.5135196574192, "at": "2026-05-01T22:41:47.106474"} |
| {"lr": 1.7514723173075704e-08, "step": 2830, "loss": 1.2907562255859375, "prob_real_tokens": 0.625125, "percent_done": 94.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12180.251166802267, "avg_wps": 12117.821267860758, "eta_in_seconds": 1010.082565952948, "at": "2026-05-01T22:42:46.302932"} |
| {"lr": 1.5520871215759742e-08, "step": 2840, "loss": 1.1639047861099243, "prob_real_tokens": 0.5227777777777778, "percent_done": 94.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12173.884691714266, "avg_wps": 12117.985356834974, "eta_in_seconds": 950.6530715109596, "at": "2026-05-01T22:43:45.496122"} |
| {"lr": 1.364664204249129e-08, "step": 2850, "loss": 1.2796859741210938, "prob_real_tokens": 0.6552222222222223, "percent_done": 95.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12074.93024994863, "avg_wps": 12118.05647690327, "eta_in_seconds": 891.2320239292948, "at": "2026-05-01T22:44:44.816637"} |
| {"lr": 1.18922633875047e-08, "step": 2860, "loss": 1.1713956594467163, "prob_real_tokens": 0.47055555555555556, "percent_done": 95.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12206.80014292944, "avg_wps": 12118.198314789477, "eta_in_seconds": 831.8068196406731, "at": "2026-05-01T22:45:44.037236"} |
| {"lr": 1.0257948422212238e-08, "step": 2870, "loss": 1.270655870437622, "prob_real_tokens": 0.6178055555555556, "percent_done": 95.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12151.987338975747, "avg_wps": 12118.297330799534, "eta_in_seconds": 772.3857357593364, "at": "2026-05-01T22:46:43.316908"} |
| {"lr": 8.743895729302114e-09, "step": 2880, "loss": 1.1884366273880005, "prob_real_tokens": 0.48633333333333334, "percent_done": 96.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12141.49236240384, "avg_wps": 12118.406199348305, "eta_in_seconds": 712.9650432467461, "at": "2026-05-01T22:47:42.581306"} |
| {"lr": 7.3502892786090035e-09, "step": 2890, "loss": 1.1798913478851318, "prob_real_tokens": 0.6227777777777778, "percent_done": 96.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12193.91985952134, "avg_wps": 12118.552045014632, "eta_in_seconds": 653.5434242128089, "at": "2026-05-01T22:48:41.792182"} |
| {"lr": 6.077298404759787e-09, "step": 2900, "loss": 1.1856943368911743, "prob_real_tokens": 0.5299027777777777, "percent_done": 96.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12183.22702372194, "avg_wps": 12118.780001641104, "eta_in_seconds": 594.119209939036, "at": "2026-05-01T22:49:40.884775"} |
| {"lr": 4.9250777865987644e-09, "step": 2910, "loss": 1.3068063259124756, "prob_real_tokens": 0.609, "percent_done": 97.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12161.737353493982, "avg_wps": 12118.972862270602, "eta_in_seconds": 534.6987796444254, "at": "2026-05-01T22:50:40.024998"} |
| {"lr": 3.893767428392201e-09, "step": 2920, "loss": 1.3094083070755005, "prob_real_tokens": 0.5355555555555556, "percent_done": 97.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12250.237903935204, "avg_wps": 12119.22520179922, "eta_in_seconds": 475.2779079594024, "at": "2026-05-01T22:51:39.078056"} |
| {"lr": 2.9834926428170155e-09, "step": 2930, "loss": 1.2298805713653564, "prob_real_tokens": 0.5639027777777778, "percent_done": 97.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12211.359014492538, "avg_wps": 12119.517325059387, "eta_in_seconds": 415.8581455697785, "at": "2026-05-01T22:52:38.072535"} |
| {"lr": 2.1943640357343405e-09, "step": 2940, "loss": 1.1970161199569702, "prob_real_tokens": 0.6095833333333334, "percent_done": 98.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12118.99047997344, "avg_wps": 12119.787614633144, "eta_in_seconds": 356.4418896898931, "at": "2026-05-01T22:53:37.095439"} |
| {"lr": 1.5264774927495592e-09, "step": 2950, "loss": 1.2413218021392822, "prob_real_tokens": 0.5642361111111112, "percent_done": 98.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12239.347894889337, "avg_wps": 12120.124420488932, "eta_in_seconds": 297.0266537787551, "at": "2026-05-01T22:54:36.020963"} |
| {"lr": 9.799141675618903e-10, "step": 2960, "loss": 1.1992884874343872, "prob_real_tokens": 0.6378055555555555, "percent_done": 98.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12152.959533877349, "avg_wps": 12120.300182006895, "eta_in_seconds": 237.61787717728998, "at": "2026-05-01T22:55:35.175044"} |
| {"lr": 5.547404721033148e-10, "step": 2970, "loss": 1.1481807231903076, "prob_real_tokens": 0.5414444444444444, "percent_done": 99.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12157.536513988258, "avg_wps": 12120.505870811512, "eta_in_seconds": 178.21038354526866, "at": "2026-05-01T22:56:34.283988"} |
| {"lr": 2.510080684692866e-10, "step": 2980, "loss": 1.3371645212173462, "prob_real_tokens": 0.6954583333333333, "percent_done": 99.33333333333333, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12163.708537634426, "avg_wps": 12120.775534858669, "eta_in_seconds": 118.80427913697774, "at": "2026-05-01T22:57:33.297485"} |
| {"lr": 6.875386264055674e-11, "step": 2990, "loss": 1.1283477544784546, "prob_real_tokens": 0.5195694444444444, "percent_done": 99.66666666666667, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12216.51232622643, "avg_wps": 12120.98125423993, "eta_in_seconds": 59.40113138514618, "at": "2026-05-01T22:58:32.404113"} |
| {"lr": 8e-12, "step": 3000, "loss": 1.1747231483459473, "prob_real_tokens": 0.6268055555555555, "percent_done": 100.0, "peak_allocated_mem": 24.519168853759766, "allocated_mem": 18.36881446838379, "wps": 12123.999912158279, "avg_wps": 12121.187691039946, "eta_in_seconds": 0.0, "at": "2026-05-01T22:59:31.507913"} |
|
|