| {"step": 1, "epoch": 0.0008090614886731392, "time": "2026-06-26 20:13:51", "logs": {"loss": 0.748644232749939, "grad_norm": 4.001261234283447, "learning_rate": 0.0, "epoch": 0.0008090614886731392}} |
| {"step": 25, "epoch": 0.020226537216828478, "time": "2026-06-26 20:14:10", "logs": {"loss": 0.7263915538787842, "grad_norm": 2.0581886768341064, "learning_rate": 9.716599190283402e-07, "epoch": 0.020226537216828478}} |
| {"step": 50, "epoch": 0.040453074433656956, "time": "2026-06-26 20:14:31", "logs": {"loss": 0.7004936218261719, "grad_norm": 3.1348977088928223, "learning_rate": 1.9838056680161946e-06, "epoch": 0.040453074433656956}} |
| {"step": 75, "epoch": 0.06067961165048544, "time": "2026-06-26 20:14:51", "logs": {"loss": 0.6983621978759765, "grad_norm": 7.536289215087891, "learning_rate": 2.995951417004049e-06, "epoch": 0.06067961165048544}} |
| {"step": 100, "epoch": 0.08090614886731391, "time": "2026-06-26 20:15:13", "logs": {"loss": 0.6408506774902344, "grad_norm": 5.387710094451904, "learning_rate": 4.008097165991903e-06, "epoch": 0.08090614886731391}} |
| {"step": 125, "epoch": 0.1011326860841424, "time": "2026-06-26 20:15:35", "logs": {"loss": 0.5292000961303711, "grad_norm": 4.4866790771484375, "learning_rate": 5.020242914979757e-06, "epoch": 0.1011326860841424}} |
| {"step": 150, "epoch": 0.12135922330097088, "time": "2026-06-26 20:15:59", "logs": {"loss": 0.4315451049804688, "grad_norm": 5.9393534660339355, "learning_rate": 6.0323886639676124e-06, "epoch": 0.12135922330097088}} |
| {"step": 175, "epoch": 0.14158576051779936, "time": "2026-06-26 20:16:22", "logs": {"loss": 0.3674881362915039, "grad_norm": 4.814444065093994, "learning_rate": 7.044534412955466e-06, "epoch": 0.14158576051779936}} |
| {"step": 200, "epoch": 0.16181229773462782, "time": "2026-06-26 20:16:44", "logs": {"loss": 0.29549776077270506, "grad_norm": 2.463817834854126, "learning_rate": 8.056680161943322e-06, "epoch": 0.16181229773462782}} |
| {"step": 225, "epoch": 0.1820388349514563, "time": "2026-06-26 20:17:07", "logs": {"loss": 0.19513965606689454, "grad_norm": 1.6270427703857422, "learning_rate": 9.068825910931175e-06, "epoch": 0.1820388349514563}} |
| {"step": 250, "epoch": 0.2022653721682848, "time": "2026-06-26 20:17:30", "logs": {"loss": 0.258267822265625, "grad_norm": 15.206025123596191, "learning_rate": 1.008097165991903e-05, "epoch": 0.2022653721682848}} |
| {"step": 275, "epoch": 0.22249190938511326, "time": "2026-06-26 20:17:53", "logs": {"loss": 0.3252030563354492, "grad_norm": 5.6893134117126465, "learning_rate": 1.1093117408906884e-05, "epoch": 0.22249190938511326}} |
| {"step": 300, "epoch": 0.24271844660194175, "time": "2026-06-26 20:18:15", "logs": {"loss": 0.2558871841430664, "grad_norm": 15.90676498413086, "learning_rate": 1.2105263157894737e-05, "epoch": 0.24271844660194175}} |
| {"step": 325, "epoch": 0.26294498381877024, "time": "2026-06-26 20:18:38", "logs": {"loss": 0.20428178787231446, "grad_norm": 1.5257281064987183, "learning_rate": 1.3117408906882592e-05, "epoch": 0.26294498381877024}} |
| {"step": 350, "epoch": 0.28317152103559873, "time": "2026-06-26 20:19:01", "logs": {"loss": 0.2579817008972168, "grad_norm": 20.172571182250977, "learning_rate": 1.4129554655870446e-05, "epoch": 0.28317152103559873}} |
| {"step": 375, "epoch": 0.30339805825242716, "time": "2026-06-26 20:19:24", "logs": {"loss": 0.3013817977905273, "grad_norm": 0.42674902081489563, "learning_rate": 1.5141700404858302e-05, "epoch": 0.30339805825242716}} |
| {"step": 400, "epoch": 0.32362459546925565, "time": "2026-06-26 20:19:46", "logs": {"loss": 0.22500619888305665, "grad_norm": 0.6443495750427246, "learning_rate": 1.6153846153846154e-05, "epoch": 0.32362459546925565}} |
| {"step": 400, "epoch": 0.32362459546925565, "time": "2026-06-26 20:20:21", "logs": {"eval_loss": 0.16199301183223724, "eval_accuracy": 0.9560384032339565, "eval_precision": 0.9539936102236422, "eval_recall": 0.9900530503978779, "eval_f1": 0.9716889033517735, "eval_runtime": 34.745, "eval_samples_per_second": 56.958, "eval_steps_per_second": 3.569, "epoch": 0.32362459546925565}} |
| {"step": 425, "epoch": 0.34385113268608414, "time": "2026-06-26 20:21:03", "logs": {"loss": 0.23191883087158202, "grad_norm": 12.952281951904297, "learning_rate": 1.716599190283401e-05, "epoch": 0.34385113268608414}} |
| {"step": 450, "epoch": 0.3640776699029126, "time": "2026-06-26 20:21:27", "logs": {"loss": 0.26200836181640624, "grad_norm": 19.496694564819336, "learning_rate": 1.8178137651821864e-05, "epoch": 0.3640776699029126}} |
| {"step": 475, "epoch": 0.3843042071197411, "time": "2026-06-26 20:21:49", "logs": {"loss": 0.164515438079834, "grad_norm": 2.9963057041168213, "learning_rate": 1.9190283400809718e-05, "epoch": 0.3843042071197411}} |
| {"step": 500, "epoch": 0.4045307443365696, "time": "2026-06-26 20:22:11", "logs": {"loss": 0.2513391304016113, "grad_norm": 0.2180398553609848, "learning_rate": 1.999993769982229e-05, "epoch": 0.4045307443365696}} |
| {"step": 525, "epoch": 0.42475728155339804, "time": "2026-06-26 20:22:34", "logs": {"loss": 0.30474105834960935, "grad_norm": 15.512799263000488, "learning_rate": 1.9997757275108847e-05, "epoch": 0.42475728155339804}} |
| {"step": 550, "epoch": 0.4449838187702265, "time": "2026-06-26 20:22:57", "logs": {"loss": 0.214493408203125, "grad_norm": 23.762832641601562, "learning_rate": 1.9992462617728763e-05, "epoch": 0.4449838187702265}} |
| {"step": 575, "epoch": 0.465210355987055, "time": "2026-06-26 20:23:20", "logs": {"loss": 0.23740652084350586, "grad_norm": 0.5767886638641357, "learning_rate": 1.9984055376931414e-05, "epoch": 0.465210355987055}} |
| {"step": 600, "epoch": 0.4854368932038835, "time": "2026-06-26 20:23:43", "logs": {"loss": 0.2517130470275879, "grad_norm": 3.6927132606506348, "learning_rate": 1.9972538171514525e-05, "epoch": 0.4854368932038835}} |
| {"step": 625, "epoch": 0.5056634304207119, "time": "2026-06-26 20:24:05", "logs": {"loss": 0.21869991302490235, "grad_norm": 2.146721363067627, "learning_rate": 1.9957914589008405e-05, "epoch": 0.5056634304207119}} |
| {"step": 650, "epoch": 0.5258899676375405, "time": "2026-06-26 20:24:28", "logs": {"loss": 0.2582887077331543, "grad_norm": 9.424388885498047, "learning_rate": 1.9940189184558487e-05, "epoch": 0.5258899676375405}} |
| {"step": 675, "epoch": 0.5461165048543689, "time": "2026-06-26 20:24:51", "logs": {"loss": 0.2324064254760742, "grad_norm": 4.679764747619629, "learning_rate": 1.9919367479506413e-05, "epoch": 0.5461165048543689}} |
| {"step": 700, "epoch": 0.5663430420711975, "time": "2026-06-26 20:25:14", "logs": {"loss": 0.23698102951049804, "grad_norm": 3.0119147300720215, "learning_rate": 1.989545595967018e-05, "epoch": 0.5663430420711975}} |
| {"step": 725, "epoch": 0.5865695792880259, "time": "2026-06-26 20:25:37", "logs": {"loss": 0.23535886764526368, "grad_norm": 10.184627532958984, "learning_rate": 1.986846207332384e-05, "epoch": 0.5865695792880259}} |
| {"step": 750, "epoch": 0.6067961165048543, "time": "2026-06-26 20:26:00", "logs": {"loss": 0.13176690101623534, "grad_norm": 51.39314651489258, "learning_rate": 1.9838394228877443e-05, "epoch": 0.6067961165048543}} |
| {"step": 775, "epoch": 0.6270226537216829, "time": "2026-06-26 20:26:22", "logs": {"loss": 0.2888069725036621, "grad_norm": 18.441652297973633, "learning_rate": 1.980526179225785e-05, "epoch": 0.6270226537216829}} |
| {"step": 800, "epoch": 0.6472491909385113, "time": "2026-06-26 20:26:45", "logs": {"loss": 0.29139081954956053, "grad_norm": 0.9691627621650696, "learning_rate": 1.97690750839913e-05, "epoch": 0.6472491909385113}} |
| {"step": 800, "epoch": 0.6472491909385113, "time": "2026-06-26 20:27:20", "logs": {"eval_loss": 0.14249353110790253, "eval_accuracy": 0.965133906013138, "eval_precision": 0.967511371020143, "eval_recall": 0.9874005305039788, "eval_f1": 0.9773547751887102, "eval_runtime": 34.6417, "eval_samples_per_second": 57.128, "eval_steps_per_second": 3.58, "epoch": 0.6472491909385113}} |
| {"step": 825, "epoch": 0.6674757281553398, "time": "2026-06-26 20:28:03", "logs": {"loss": 0.1711592674255371, "grad_norm": 7.6904096603393555, "learning_rate": 1.9729845375988694e-05, "epoch": 0.6674757281553398}} |
| {"step": 850, "epoch": 0.6877022653721683, "time": "2026-06-26 20:28:26", "logs": {"loss": 0.2283180809020996, "grad_norm": 6.999588489532471, "learning_rate": 1.968758488803439e-05, "epoch": 0.6877022653721683}} |
| {"step": 875, "epoch": 0.7079288025889967, "time": "2026-06-26 20:28:49", "logs": {"loss": 0.26335405349731444, "grad_norm": 6.149271488189697, "learning_rate": 1.9642306783979902e-05, "epoch": 0.7079288025889967}} |
| {"step": 900, "epoch": 0.7281553398058253, "time": "2026-06-26 20:29:11", "logs": {"loss": 0.1903523635864258, "grad_norm": 9.907973289489746, "learning_rate": 1.959402516764339e-05, "epoch": 0.7281553398058253}} |
| {"step": 925, "epoch": 0.7483818770226537, "time": "2026-06-26 20:29:34", "logs": {"loss": 0.1740720748901367, "grad_norm": 12.104832649230957, "learning_rate": 1.954275507841646e-05, "epoch": 0.7483818770226537}} |
| {"step": 950, "epoch": 0.7686084142394822, "time": "2026-06-26 20:29:57", "logs": {"loss": 0.3097032928466797, "grad_norm": 2.8893074989318848, "learning_rate": 1.948851248657947e-05, "epoch": 0.7686084142394822}} |
| {"step": 975, "epoch": 0.7888349514563107, "time": "2026-06-26 20:30:20", "logs": {"loss": 0.14465970993041993, "grad_norm": 0.07049937546253204, "learning_rate": 1.9431314288326925e-05, "epoch": 0.7888349514563107}} |
| {"step": 1000, "epoch": 0.8090614886731392, "time": "2026-06-26 20:30:42", "logs": {"loss": 0.22320323944091797, "grad_norm": 18.34374237060547, "learning_rate": 1.9371178300504394e-05, "epoch": 0.8090614886731392}} |
| {"step": 1025, "epoch": 0.8292880258899676, "time": "2026-06-26 20:31:05", "logs": {"loss": 0.09155179023742675, "grad_norm": 5.898121356964111, "learning_rate": 1.930812325505871e-05, "epoch": 0.8292880258899676}} |
| {"step": 1050, "epoch": 0.8495145631067961, "time": "2026-06-26 20:31:28", "logs": {"loss": 0.28490724563598635, "grad_norm": 9.933682441711426, "learning_rate": 1.9242168793203088e-05, "epoch": 0.8495145631067961}} |
| {"step": 1075, "epoch": 0.8697411003236246, "time": "2026-06-26 20:31:51", "logs": {"loss": 0.14135926246643066, "grad_norm": 0.8073019981384277, "learning_rate": 1.9173335459299025e-05, "epoch": 0.8697411003236246}} |
| {"step": 1100, "epoch": 0.889967637540453, "time": "2026-06-26 20:32:14", "logs": {"loss": 0.2752204895019531, "grad_norm": 23.247236251831055, "learning_rate": 1.9101644694456873e-05, "epoch": 0.889967637540453}} |
| {"step": 1125, "epoch": 0.9101941747572816, "time": "2026-06-26 20:32:36", "logs": {"loss": 0.14879122734069825, "grad_norm": 10.088881492614746, "learning_rate": 1.902711882985708e-05, "epoch": 0.9101941747572816}} |
| {"step": 1150, "epoch": 0.93042071197411, "time": "2026-06-26 20:32:59", "logs": {"loss": 0.23470134735107423, "grad_norm": 0.5620437860488892, "learning_rate": 1.894978107979417e-05, "epoch": 0.93042071197411}} |
| {"step": 1175, "epoch": 0.9506472491909385, "time": "2026-06-26 20:33:22", "logs": {"loss": 0.19982004165649414, "grad_norm": 3.2695255279541016, "learning_rate": 1.886965553444568e-05, "epoch": 0.9506472491909385}} |
| {"step": 1200, "epoch": 0.970873786407767, "time": "2026-06-26 20:33:45", "logs": {"loss": 0.22285139083862304, "grad_norm": 19.842267990112305, "learning_rate": 1.8786767152368197e-05, "epoch": 0.970873786407767}} |
| {"step": 1200, "epoch": 0.970873786407767, "time": "2026-06-26 20:34:20", "logs": {"eval_loss": 0.14106644690036774, "eval_accuracy": 0.9641232945932289, "eval_precision": 0.9650485436893204, "eval_recall": 0.9887267904509284, "eval_f1": 0.9767441860465116, "eval_runtime": 34.6101, "eval_samples_per_second": 57.18, "eval_steps_per_second": 3.583, "epoch": 0.970873786407767}} |
| {"step": 1225, "epoch": 0.9911003236245954, "time": "2026-06-26 20:35:02", "logs": {"loss": 0.13252583503723145, "grad_norm": 17.05675506591797, "learning_rate": 1.8701141752722966e-05, "epoch": 0.9911003236245954}} |
| {"step": 1250, "epoch": 1.0113268608414239, "time": "2026-06-26 20:35:25", "logs": {"loss": 0.13512847900390626, "grad_norm": 14.713388442993164, "learning_rate": 1.8612806007233382e-05, "epoch": 1.0113268608414239}} |
| {"step": 1275, "epoch": 1.0315533980582525, "time": "2026-06-26 20:35:48", "logs": {"loss": 0.15763619422912598, "grad_norm": 16.86681365966797, "learning_rate": 1.8521787431876954e-05, "epoch": 1.0315533980582525}} |
| {"step": 1300, "epoch": 1.051779935275081, "time": "2026-06-26 20:36:11", "logs": {"loss": 0.18742298126220702, "grad_norm": 0.6122269630432129, "learning_rate": 1.8428114378314237e-05, "epoch": 1.051779935275081}} |
| {"step": 1325, "epoch": 1.0720064724919094, "time": "2026-06-26 20:36:33", "logs": {"loss": 0.16315383911132814, "grad_norm": 6.803472995758057, "learning_rate": 1.8331816025057508e-05, "epoch": 1.0720064724919094}} |
| {"step": 1350, "epoch": 1.0922330097087378, "time": "2026-06-26 20:36:57", "logs": {"loss": 0.1462923812866211, "grad_norm": 3.42342209815979, "learning_rate": 1.823292236838186e-05, "epoch": 1.0922330097087378}} |
| {"step": 1375, "epoch": 1.1124595469255663, "time": "2026-06-26 20:37:20", "logs": {"loss": 0.2096657371520996, "grad_norm": 6.770485877990723, "learning_rate": 1.813146421298154e-05, "epoch": 1.1124595469255663}} |
| {"step": 1400, "epoch": 1.132686084142395, "time": "2026-06-26 20:37:42", "logs": {"loss": 0.1903732681274414, "grad_norm": 6.892830848693848, "learning_rate": 1.802747316237454e-05, "epoch": 1.132686084142395}} |
| {"step": 1425, "epoch": 1.1529126213592233, "time": "2026-06-26 20:38:05", "logs": {"loss": 0.22764249801635741, "grad_norm": 1.7113957405090332, "learning_rate": 1.792098160905829e-05, "epoch": 1.1529126213592233}} |
| {"step": 1450, "epoch": 1.1731391585760518, "time": "2026-06-26 20:38:28", "logs": {"loss": 0.08850884437561035, "grad_norm": 0.2954396903514862, "learning_rate": 1.7812022724419624e-05, "epoch": 1.1731391585760518}} |
| {"step": 1475, "epoch": 1.1933656957928802, "time": "2026-06-26 20:38:50", "logs": {"loss": 0.15286758422851562, "grad_norm": 1.2150897979736328, "learning_rate": 1.7700630448402125e-05, "epoch": 1.1933656957928802}} |
| {"step": 1500, "epoch": 1.2135922330097086, "time": "2026-06-26 20:39:13", "logs": {"loss": 0.14677520751953124, "grad_norm": 2.9238908290863037, "learning_rate": 1.758683947893406e-05, "epoch": 1.2135922330097086}} |
| {"step": 1525, "epoch": 1.2338187702265373, "time": "2026-06-26 20:39:36", "logs": {"loss": 0.13224196434020996, "grad_norm": 9.288032531738281, "learning_rate": 1.747068526112022e-05, "epoch": 1.2338187702265373}} |
| {"step": 1550, "epoch": 1.2540453074433657, "time": "2026-06-26 20:39:59", "logs": {"loss": 0.08127058982849121, "grad_norm": 21.540422439575195, "learning_rate": 1.735220397620101e-05, "epoch": 1.2540453074433657}} |
| {"step": 1575, "epoch": 1.2742718446601942, "time": "2026-06-26 20:40:22", "logs": {"loss": 0.10571711540222167, "grad_norm": 0.07893255352973938, "learning_rate": 1.7231432530282246e-05, "epoch": 1.2742718446601942}} |
| {"step": 1600, "epoch": 1.2944983818770226, "time": "2026-06-26 20:40:44", "logs": {"loss": 0.1872186279296875, "grad_norm": 0.5910708904266357, "learning_rate": 1.7108408542839146e-05, "epoch": 1.2944983818770226}} |
| {"step": 1600, "epoch": 1.2944983818770226, "time": "2026-06-26 20:41:19", "logs": {"eval_loss": 0.1455843597650528, "eval_accuracy": 0.9671551288529561, "eval_precision": 0.9737360472751149, "eval_recall": 0.98342175066313, "eval_f1": 0.9785549323655559, "eval_runtime": 34.8064, "eval_samples_per_second": 56.857, "eval_steps_per_second": 3.563, "epoch": 1.2944983818770226}} |
| {"step": 1625, "epoch": 1.3147249190938513, "time": "2026-06-26 20:42:01", "logs": {"loss": 0.1566315174102783, "grad_norm": 0.48737797141075134, "learning_rate": 1.698317033499813e-05, "epoch": 1.3147249190938513}} |
| {"step": 1650, "epoch": 1.3349514563106797, "time": "2026-06-26 20:42:25", "logs": {"loss": 0.15158571243286134, "grad_norm": 17.18292808532715, "learning_rate": 1.6855756917600057e-05, "epoch": 1.3349514563106797}} |
| {"step": 1675, "epoch": 1.3551779935275081, "time": "2026-06-26 20:42:48", "logs": {"loss": 0.16150398254394532, "grad_norm": 10.572613716125488, "learning_rate": 1.6726207979048604e-05, "epoch": 1.3551779935275081}} |
| {"step": 1700, "epoch": 1.3754045307443366, "time": "2026-06-26 20:43:10", "logs": {"loss": 0.09849663734436036, "grad_norm": 0.1977110058069229, "learning_rate": 1.6594563872947613e-05, "epoch": 1.3754045307443366}} |
| {"step": 1725, "epoch": 1.395631067961165, "time": "2026-06-26 20:43:33", "logs": {"loss": 0.17041929244995116, "grad_norm": 2.6522152423858643, "learning_rate": 1.6460865605531214e-05, "epoch": 1.395631067961165}} |
| {"step": 1750, "epoch": 1.4158576051779934, "time": "2026-06-26 20:43:56", "logs": {"loss": 0.1783625602722168, "grad_norm": 0.18808063864707947, "learning_rate": 1.6325154822890663e-05, "epoch": 1.4158576051779934}} |
| {"step": 1775, "epoch": 1.436084142394822, "time": "2026-06-26 20:44:19", "logs": {"loss": 0.15096524238586426, "grad_norm": 33.89837646484375, "learning_rate": 1.618747379800188e-05, "epoch": 1.436084142394822}} |
| {"step": 1800, "epoch": 1.4563106796116505, "time": "2026-06-26 20:44:41", "logs": {"loss": 0.2288828659057617, "grad_norm": 10.100881576538086, "learning_rate": 1.6047865417557695e-05, "epoch": 1.4563106796116505}} |
| {"step": 1825, "epoch": 1.476537216828479, "time": "2026-06-26 20:45:04", "logs": {"loss": 0.22033729553222656, "grad_norm": 7.704436302185059, "learning_rate": 1.5906373168608952e-05, "epoch": 1.476537216828479}} |
| {"step": 1850, "epoch": 1.4967637540453074, "time": "2026-06-26 20:45:27", "logs": {"loss": 0.15806819915771483, "grad_norm": 9.382397651672363, "learning_rate": 1.576304112501858e-05, "epoch": 1.4967637540453074}} |
| {"step": 1875, "epoch": 1.516990291262136, "time": "2026-06-26 20:45:50", "logs": {"loss": 0.16990087509155274, "grad_norm": 6.415167808532715, "learning_rate": 1.5617913933732892e-05, "epoch": 1.516990291262136}} |
| {"step": 1900, "epoch": 1.5372168284789645, "time": "2026-06-26 20:46:13", "logs": {"loss": 0.1730282974243164, "grad_norm": 8.820036888122559, "learning_rate": 1.5471036800874375e-05, "epoch": 1.5372168284789645}} |
| {"step": 1925, "epoch": 1.557443365695793, "time": "2026-06-26 20:46:36", "logs": {"loss": 0.15791168212890624, "grad_norm": 28.52211570739746, "learning_rate": 1.5322455477660274e-05, "epoch": 1.557443365695793}} |
| {"step": 1950, "epoch": 1.5776699029126213, "time": "2026-06-26 20:46:58", "logs": {"loss": 0.24158744812011718, "grad_norm": 15.133056640625, "learning_rate": 1.5172216246151426e-05, "epoch": 1.5776699029126213}} |
| {"step": 1975, "epoch": 1.5978964401294498, "time": "2026-06-26 20:47:21", "logs": {"loss": 0.07609220504760743, "grad_norm": 13.767041206359863, "learning_rate": 1.50203659048357e-05, "epoch": 1.5978964401294498}} |
| {"step": 2000, "epoch": 1.6181229773462782, "time": "2026-06-26 20:47:44", "logs": {"loss": 0.2856024742126465, "grad_norm": 2.377875328063965, "learning_rate": 1.4866951754050605e-05, "epoch": 1.6181229773462782}} |
| {"step": 2000, "epoch": 1.6181229773462782, "time": "2026-06-26 20:48:19", "logs": {"eval_loss": 0.12881171703338623, "eval_accuracy": 0.9671551288529561, "eval_precision": 0.9819639278557114, "eval_recall": 0.9748010610079576, "eval_f1": 0.978369384359401, "eval_runtime": 34.7861, "eval_samples_per_second": 56.891, "eval_steps_per_second": 3.565, "epoch": 1.6181229773462782}} |
| {"step": 2025, "epoch": 1.6383495145631068, "time": "2026-06-26 20:49:02", "logs": {"loss": 0.18048593521118164, "grad_norm": 1.8338792324066162, "learning_rate": 1.4712021581249534e-05, "epoch": 1.6383495145631068}} |
| {"step": 2050, "epoch": 1.6585760517799353, "time": "2026-06-26 20:49:26", "logs": {"loss": 0.1664619827270508, "grad_norm": 0.2766644060611725, "learning_rate": 1.4555623646116325e-05, "epoch": 1.6585760517799353}} |
| {"step": 2075, "epoch": 1.6788025889967637, "time": "2026-06-26 20:49:48", "logs": {"loss": 0.16300722122192382, "grad_norm": 2.5218565464019775, "learning_rate": 1.4397806665532693e-05, "epoch": 1.6788025889967637}} |
| {"step": 2100, "epoch": 1.6990291262135924, "time": "2026-06-26 20:50:11", "logs": {"loss": 0.14837112426757812, "grad_norm": 0.9285559058189392, "learning_rate": 1.4238619798403266e-05, "epoch": 1.6990291262135924}} |
| {"step": 2125, "epoch": 1.7192556634304208, "time": "2026-06-26 20:50:33", "logs": {"loss": 0.15740616798400878, "grad_norm": 10.178064346313477, "learning_rate": 1.4078112630342891e-05, "epoch": 1.7192556634304208}} |
| {"step": 2150, "epoch": 1.7394822006472492, "time": "2026-06-26 20:50:56", "logs": {"loss": 0.22019144058227538, "grad_norm": 19.79058074951172, "learning_rate": 1.3916335158231101e-05, "epoch": 1.7394822006472492}} |
| {"step": 2175, "epoch": 1.7597087378640777, "time": "2026-06-26 20:51:19", "logs": {"loss": 0.10831342697143555, "grad_norm": 8.479653358459473, "learning_rate": 1.3753337774638397e-05, "epoch": 1.7597087378640777}} |
| {"step": 2200, "epoch": 1.779935275080906, "time": "2026-06-26 20:51:42", "logs": {"loss": 0.14887014389038086, "grad_norm": 0.27186402678489685, "learning_rate": 1.3589171252129329e-05, "epoch": 1.779935275080906}} |
| {"step": 2225, "epoch": 1.8001618122977345, "time": "2026-06-26 20:52:05", "logs": {"loss": 0.19739471435546874, "grad_norm": 5.499264240264893, "learning_rate": 1.3423886727447176e-05, "epoch": 1.8001618122977345}} |
| {"step": 2250, "epoch": 1.820388349514563, "time": "2026-06-26 20:52:27", "logs": {"loss": 0.19788772583007813, "grad_norm": 2.787201166152954, "learning_rate": 1.325753568558521e-05, "epoch": 1.820388349514563}} |
| {"step": 2275, "epoch": 1.8406148867313916, "time": "2026-06-26 20:52:50", "logs": {"loss": 0.11929478645324706, "grad_norm": 3.942160129547119, "learning_rate": 1.3090169943749475e-05, "epoch": 1.8406148867313916}} |
| {"step": 2300, "epoch": 1.86084142394822, "time": "2026-06-26 20:53:13", "logs": {"loss": 0.09196714401245117, "grad_norm": 0.19959710538387299, "learning_rate": 1.2921841635218082e-05, "epoch": 1.86084142394822}} |
| {"step": 2325, "epoch": 1.8810679611650487, "time": "2026-06-26 20:53:36", "logs": {"loss": 0.2154999542236328, "grad_norm": 0.05811980366706848, "learning_rate": 1.275260319310205e-05, "epoch": 1.8810679611650487}} |
| {"step": 2350, "epoch": 1.9012944983818771, "time": "2026-06-26 20:53:59", "logs": {"loss": 0.1920686912536621, "grad_norm": 1.1418527364730835, "learning_rate": 1.2582507334012755e-05, "epoch": 1.9012944983818771}} |
| {"step": 2375, "epoch": 1.9215210355987056, "time": "2026-06-26 20:54:21", "logs": {"loss": 0.15039829254150391, "grad_norm": 3.8587708473205566, "learning_rate": 1.2411607041641062e-05, "epoch": 1.9215210355987056}} |
| {"step": 2400, "epoch": 1.941747572815534, "time": "2026-06-26 20:54:44", "logs": {"loss": 0.1844302558898926, "grad_norm": 0.10201486200094223, "learning_rate": 1.2239955550253277e-05, "epoch": 1.941747572815534}} |
| {"step": 2400, "epoch": 1.941747572815534, "time": "2026-06-26 20:55:19", "logs": {"eval_loss": 0.14186660945415497, "eval_accuracy": 0.9631126831733199, "eval_precision": 0.9644012944983819, "eval_recall": 0.9880636604774535, "eval_f1": 0.9760890926957091, "eval_runtime": 34.6039, "eval_samples_per_second": 57.19, "eval_steps_per_second": 3.583, "epoch": 1.941747572815534}} |
| {"step": 2400, "epoch": 1.941747572815534, "time": "2026-06-26 20:55:37", "logs": {"train_runtime": 2509.4609, "train_samples_per_second": 31.503, "train_steps_per_second": 1.97, "total_flos": 1.010030719315968e+16, "train_loss": 0.2228707834581534, "epoch": 1.941747572815534}} |
| {"step": 2400, "epoch": 1.941747572815534, "time": "2026-06-26 20:56:13", "logs": {"eval_loss": 0.1455843597650528, "eval_accuracy": 0.9671551288529561, "eval_precision": 0.9737360472751149, "eval_recall": 0.98342175066313, "eval_f1": 0.9785549323655559, "eval_runtime": 35.6446, "eval_samples_per_second": 55.52, "eval_steps_per_second": 3.479, "epoch": 1.941747572815534}} |
|
|
