xiulinyang's picture
Adding model checkpoints and config files
5906368
{"num_parameters": 124439808, "trainable_parameters": 124439808, "step": 0}
{"train_info/time_between_train_steps": 3.5239946842193604, "step": 0}
{"info/global_step": 1, "train_info/time_within_train_step": 33.30619430541992, "step": 1}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 19761.71484375, "train_info/memory_reserved": 22624.0, "train_info/memory_max_reserved": 22624.0, "_timestamp": 1733727949, "_runtime": 39}, "step": 1}
{"logs": {"train/loss": 11.0188, "train/learning_rate": 4.9999999999999996e-06, "train/epoch": 0.0, "_timestamp": 1733727949, "_runtime": 39}, "step": 1}
{"train_info/time_between_train_steps": 0.12121915817260742, "step": 1}
{"info/global_step": 2, "train_info/time_within_train_step": 32.82634258270264, "step": 2}
{"train_info/time_between_train_steps": 0.006198883056640625, "step": 2}
{"info/global_step": 3, "train_info/time_within_train_step": 32.07644701004028, "step": 3}
{"train_info/time_between_train_steps": 0.0060307979583740234, "step": 3}
{"info/global_step": 4, "train_info/time_within_train_step": 32.430400371551514, "step": 4}
{"train_info/time_between_train_steps": 0.005764007568359375, "step": 4}
{"info/global_step": 5, "train_info/time_within_train_step": 32.437405586242676, "step": 5}
{"train_info/time_between_train_steps": 0.005811214447021484, "step": 5}
{"info/global_step": 6, "train_info/time_within_train_step": 32.627691984176636, "step": 6}
{"train_info/time_between_train_steps": 0.006096839904785156, "step": 6}
{"info/global_step": 7, "train_info/time_within_train_step": 32.49433374404907, "step": 7}
{"train_info/time_between_train_steps": 0.0057942867279052734, "step": 7}
{"info/global_step": 8, "train_info/time_within_train_step": 32.59205651283264, "step": 8}
{"train_info/time_between_train_steps": 0.005841493606567383, "step": 8}
{"info/global_step": 9, "train_info/time_within_train_step": 32.66184616088867, "step": 9}
{"train_info/time_between_train_steps": 0.00537872314453125, "step": 9}
{"info/global_step": 10, "train_info/time_within_train_step": 32.65320420265198, "step": 10}
{"train_info/time_between_train_steps": 0.008213520050048828, "step": 10}
{"info/global_step": 11, "train_info/time_within_train_step": 32.47786545753479, "step": 11}
{"train_info/time_between_train_steps": 0.0070493221282958984, "step": 11}
{"info/global_step": 12, "train_info/time_within_train_step": 32.47931623458862, "step": 12}
{"train_info/time_between_train_steps": 0.007341146469116211, "step": 12}
{"info/global_step": 13, "train_info/time_within_train_step": 31.828296899795532, "step": 13}
{"train_info/time_between_train_steps": 0.0064716339111328125, "step": 13}
{"info/global_step": 14, "train_info/time_within_train_step": 32.465394496917725, "step": 14}
{"train_info/time_between_train_steps": 0.006506443023681641, "step": 14}
{"info/global_step": 15, "train_info/time_within_train_step": 32.43285250663757, "step": 15}
{"train_info/time_between_train_steps": 0.0065310001373291016, "step": 15}
{"info/global_step": 16, "train_info/time_within_train_step": 32.523743867874146, "step": 16}
{"train_info/time_between_train_steps": 0.0063457489013671875, "step": 16}
{"info/global_step": 17, "train_info/time_within_train_step": 32.51143479347229, "step": 17}
{"train_info/time_between_train_steps": 0.006667375564575195, "step": 17}
{"info/global_step": 18, "train_info/time_within_train_step": 32.27540111541748, "step": 18}
{"train_info/time_between_train_steps": 0.0063588619232177734, "step": 18}
{"info/global_step": 19, "train_info/time_within_train_step": 32.51773524284363, "step": 19}
{"train_info/time_between_train_steps": 0.006559133529663086, "step": 19}
{"info/global_step": 20, "train_info/time_within_train_step": 32.49087309837341, "step": 20}
{"train_info/time_between_train_steps": 0.006365060806274414, "step": 20}
{"info/global_step": 21, "train_info/time_within_train_step": 32.4338755607605, "step": 21}
{"train_info/time_between_train_steps": 0.0062084197998046875, "step": 21}
{"info/global_step": 22, "train_info/time_within_train_step": 32.579148292541504, "step": 22}
{"train_info/time_between_train_steps": 0.009507179260253906, "step": 22}
{"info/global_step": 23, "train_info/time_within_train_step": 32.1200532913208, "step": 23}
{"train_info/time_between_train_steps": 0.0068285465240478516, "step": 23}
{"info/global_step": 24, "train_info/time_within_train_step": 32.45702576637268, "step": 24}
{"train_info/time_between_train_steps": 0.006601095199584961, "step": 24}
{"info/global_step": 25, "train_info/time_within_train_step": 32.46128463745117, "step": 25}
{"train_info/time_between_train_steps": 0.006630420684814453, "step": 25}
{"info/global_step": 26, "train_info/time_within_train_step": 32.39973592758179, "step": 26}
{"train_info/time_between_train_steps": 0.006768465042114258, "step": 26}
{"info/global_step": 27, "train_info/time_within_train_step": 32.35210990905762, "step": 27}
{"train_info/time_between_train_steps": 0.0067596435546875, "step": 27}
{"train_info/time_between_train_steps": 24.301878452301025, "step": 27}
{"info/global_step": 28, "train_info/time_within_train_step": 32.57335186004639, "step": 28}
{"train_info/time_between_train_steps": 0.008227348327636719, "step": 28}
{"info/global_step": 29, "train_info/time_within_train_step": 32.75170397758484, "step": 29}
{"train_info/time_between_train_steps": 0.007695913314819336, "step": 29}
{"info/global_step": 30, "train_info/time_within_train_step": 32.794291734695435, "step": 30}
{"train_info/time_between_train_steps": 0.006203651428222656, "step": 30}
{"info/global_step": 31, "train_info/time_within_train_step": 32.81736469268799, "step": 31}
{"train_info/time_between_train_steps": 0.0059735774993896484, "step": 31}
{"info/global_step": 32, "train_info/time_within_train_step": 32.635573625564575, "step": 32}
{"train_info/time_between_train_steps": 0.008797883987426758, "step": 32}
{"info/global_step": 33, "train_info/time_within_train_step": 32.026989221572876, "step": 33}
{"train_info/time_between_train_steps": 0.011559009552001953, "step": 33}
{"info/global_step": 34, "train_info/time_within_train_step": 32.53317165374756, "step": 34}
{"train_info/time_between_train_steps": 0.006182432174682617, "step": 34}
{"info/global_step": 35, "train_info/time_within_train_step": 32.571816205978394, "step": 35}
{"train_info/time_between_train_steps": 0.011111021041870117, "step": 35}
{"info/global_step": 36, "train_info/time_within_train_step": 32.48813843727112, "step": 36}
{"train_info/time_between_train_steps": 0.00572514533996582, "step": 36}
{"info/global_step": 37, "train_info/time_within_train_step": 32.6136531829834, "step": 37}
{"train_info/time_between_train_steps": 0.0054225921630859375, "step": 37}
{"info/global_step": 38, "train_info/time_within_train_step": 32.36243271827698, "step": 38}
{"train_info/time_between_train_steps": 0.007222652435302734, "step": 38}
{"info/global_step": 39, "train_info/time_within_train_step": 32.72102117538452, "step": 39}
{"train_info/time_between_train_steps": 0.005423784255981445, "step": 39}
{"info/global_step": 40, "train_info/time_within_train_step": 32.632346630096436, "step": 40}
{"train_info/time_between_train_steps": 0.005859851837158203, "step": 40}
{"info/global_step": 41, "train_info/time_within_train_step": 32.51607942581177, "step": 41}
{"train_info/time_between_train_steps": 0.009982109069824219, "step": 41}
{"info/global_step": 42, "train_info/time_within_train_step": 32.463184118270874, "step": 42}
{"train_info/time_between_train_steps": 0.007023334503173828, "step": 42}
{"info/global_step": 43, "train_info/time_within_train_step": 32.310779094696045, "step": 43}
{"train_info/time_between_train_steps": 0.011095523834228516, "step": 43}
{"info/global_step": 44, "train_info/time_within_train_step": 32.01535606384277, "step": 44}
{"train_info/time_between_train_steps": 0.005925893783569336, "step": 44}
{"info/global_step": 45, "train_info/time_within_train_step": 31.94600248336792, "step": 45}
{"train_info/time_between_train_steps": 0.00558781623840332, "step": 45}
{"info/global_step": 46, "train_info/time_within_train_step": 32.36760330200195, "step": 46}
{"train_info/time_between_train_steps": 0.010713338851928711, "step": 46}
{"info/global_step": 47, "train_info/time_within_train_step": 32.64608645439148, "step": 47}
{"train_info/time_between_train_steps": 0.005426168441772461, "step": 47}
{"info/global_step": 48, "train_info/time_within_train_step": 32.54395151138306, "step": 48}
{"train_info/time_between_train_steps": 0.006079435348510742, "step": 48}
{"info/global_step": 49, "train_info/time_within_train_step": 32.75273299217224, "step": 49}
{"train_info/time_between_train_steps": 0.0075681209564208984, "step": 49}
{"info/global_step": 50, "train_info/time_within_train_step": 32.38865685462952, "step": 50}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733729565, "_runtime": 1655}, "step": 50}
{"logs": {"train/loss": 8.1559, "train/learning_rate": 0.00025, "train/epoch": 1.02, "_timestamp": 1733729565, "_runtime": 1655}, "step": 50}
{"train_info/time_between_train_steps": 0.007920026779174805, "step": 50}
{"info/global_step": 51, "train_info/time_within_train_step": 32.54572081565857, "step": 51}
{"train_info/time_between_train_steps": 0.007287740707397461, "step": 51}
{"info/global_step": 52, "train_info/time_within_train_step": 32.8617730140686, "step": 52}
{"train_info/time_between_train_steps": 0.00644993782043457, "step": 52}
{"info/global_step": 53, "train_info/time_within_train_step": 32.58287835121155, "step": 53}
{"train_info/time_between_train_steps": 0.008871316909790039, "step": 53}
{"info/global_step": 54, "train_info/time_within_train_step": 31.799750804901123, "step": 54}
{"train_info/time_between_train_steps": 0.011792898178100586, "step": 54}
{"train_info/time_between_train_steps": 23.713664054870605, "step": 54}
{"info/global_step": 55, "train_info/time_within_train_step": 32.577099084854126, "step": 55}
{"train_info/time_between_train_steps": 0.006056785583496094, "step": 55}
{"info/global_step": 56, "train_info/time_within_train_step": 32.46104288101196, "step": 56}
{"train_info/time_between_train_steps": 0.00634455680847168, "step": 56}
{"info/global_step": 57, "train_info/time_within_train_step": 32.511727809906006, "step": 57}
{"train_info/time_between_train_steps": 0.0059435367584228516, "step": 57}
{"info/global_step": 58, "train_info/time_within_train_step": 32.62384510040283, "step": 58}
{"train_info/time_between_train_steps": 0.00543975830078125, "step": 58}
{"info/global_step": 59, "train_info/time_within_train_step": 32.556527853012085, "step": 59}
{"train_info/time_between_train_steps": 0.006226778030395508, "step": 59}
{"info/global_step": 60, "train_info/time_within_train_step": 32.79471302032471, "step": 60}
{"train_info/time_between_train_steps": 0.005613088607788086, "step": 60}
{"info/global_step": 61, "train_info/time_within_train_step": 32.66909432411194, "step": 61}
{"train_info/time_between_train_steps": 0.007854938507080078, "step": 61}
{"info/global_step": 62, "train_info/time_within_train_step": 32.683597564697266, "step": 62}
{"train_info/time_between_train_steps": 0.005547523498535156, "step": 62}
{"info/global_step": 63, "train_info/time_within_train_step": 32.381056785583496, "step": 63}
{"train_info/time_between_train_steps": 0.00529026985168457, "step": 63}
{"info/global_step": 64, "train_info/time_within_train_step": 31.830498456954956, "step": 64}
{"train_info/time_between_train_steps": 0.005256175994873047, "step": 64}
{"info/global_step": 65, "train_info/time_within_train_step": 32.5351083278656, "step": 65}
{"train_info/time_between_train_steps": 0.005312204360961914, "step": 65}
{"info/global_step": 66, "train_info/time_within_train_step": 32.37010860443115, "step": 66}
{"train_info/time_between_train_steps": 0.005232334136962891, "step": 66}
{"info/global_step": 67, "train_info/time_within_train_step": 32.368629455566406, "step": 67}
{"train_info/time_between_train_steps": 0.005313873291015625, "step": 67}
{"info/global_step": 68, "train_info/time_within_train_step": 32.44748783111572, "step": 68}
{"train_info/time_between_train_steps": 0.009017705917358398, "step": 68}
{"info/global_step": 69, "train_info/time_within_train_step": 32.630438566207886, "step": 69}
{"train_info/time_between_train_steps": 0.005511283874511719, "step": 69}
{"info/global_step": 70, "train_info/time_within_train_step": 32.32601737976074, "step": 70}
{"train_info/time_between_train_steps": 0.010171651840209961, "step": 70}
{"info/global_step": 71, "train_info/time_within_train_step": 32.7625629901886, "step": 71}
{"train_info/time_between_train_steps": 0.005700826644897461, "step": 71}
{"info/global_step": 72, "train_info/time_within_train_step": 32.367431640625, "step": 72}
{"train_info/time_between_train_steps": 0.00532078742980957, "step": 72}
{"info/global_step": 73, "train_info/time_within_train_step": 32.528573513031006, "step": 73}
{"train_info/time_between_train_steps": 0.005346775054931641, "step": 73}
{"info/global_step": 74, "train_info/time_within_train_step": 32.59959602355957, "step": 74}
{"train_info/time_between_train_steps": 0.005369901657104492, "step": 74}
{"info/global_step": 75, "train_info/time_within_train_step": 31.951497316360474, "step": 75}
{"train_info/time_between_train_steps": 0.005460262298583984, "step": 75}
{"info/global_step": 76, "train_info/time_within_train_step": 32.49175143241882, "step": 76}
{"train_info/time_between_train_steps": 0.00544428825378418, "step": 76}
{"info/global_step": 77, "train_info/time_within_train_step": 32.441633224487305, "step": 77}
{"train_info/time_between_train_steps": 0.005497455596923828, "step": 77}
{"info/global_step": 78, "train_info/time_within_train_step": 32.555604219436646, "step": 78}
{"train_info/time_between_train_steps": 0.005654573440551758, "step": 78}
{"info/global_step": 79, "train_info/time_within_train_step": 32.24544954299927, "step": 79}
{"train_info/time_between_train_steps": 0.010485410690307617, "step": 79}
{"info/global_step": 80, "train_info/time_within_train_step": 32.26199007034302, "step": 80}
{"train_info/time_between_train_steps": 0.005857706069946289, "step": 80}
{"info/global_step": 81, "train_info/time_within_train_step": 32.513444900512695, "step": 81}
{"train_info/time_between_train_steps": 0.0059604644775390625, "step": 81}
{"train_info/time_between_train_steps": 23.5996356010437, "step": 81}
{"info/global_step": 82, "train_info/time_within_train_step": 32.43805122375488, "step": 82}
{"train_info/time_between_train_steps": 0.00764155387878418, "step": 82}
{"info/global_step": 83, "train_info/time_within_train_step": 32.99715709686279, "step": 83}
{"train_info/time_between_train_steps": 0.005538463592529297, "step": 83}
{"info/global_step": 84, "train_info/time_within_train_step": 32.53729271888733, "step": 84}
{"train_info/time_between_train_steps": 0.011019706726074219, "step": 84}
{"info/global_step": 85, "train_info/time_within_train_step": 32.15237903594971, "step": 85}
{"train_info/time_between_train_steps": 0.005612850189208984, "step": 85}
{"info/global_step": 86, "train_info/time_within_train_step": 32.40459156036377, "step": 86}
{"train_info/time_between_train_steps": 0.005631923675537109, "step": 86}
{"info/global_step": 87, "train_info/time_within_train_step": 32.72761392593384, "step": 87}
{"train_info/time_between_train_steps": 0.005728006362915039, "step": 87}
{"info/global_step": 88, "train_info/time_within_train_step": 32.451345920562744, "step": 88}
{"train_info/time_between_train_steps": 0.005524873733520508, "step": 88}
{"info/global_step": 89, "train_info/time_within_train_step": 32.73307275772095, "step": 89}
{"train_info/time_between_train_steps": 0.0055828094482421875, "step": 89}
{"info/global_step": 90, "train_info/time_within_train_step": 32.50501465797424, "step": 90}
{"train_info/time_between_train_steps": 0.005242824554443359, "step": 90}
{"info/global_step": 91, "train_info/time_within_train_step": 32.67951679229736, "step": 91}
{"train_info/time_between_train_steps": 0.005095720291137695, "step": 91}
{"info/global_step": 92, "train_info/time_within_train_step": 32.52647089958191, "step": 92}
{"train_info/time_between_train_steps": 0.011179208755493164, "step": 92}
{"info/global_step": 93, "train_info/time_within_train_step": 32.63813638687134, "step": 93}
{"train_info/time_between_train_steps": 0.010184526443481445, "step": 93}
{"info/global_step": 94, "train_info/time_within_train_step": 32.47755694389343, "step": 94}
{"train_info/time_between_train_steps": 0.005214214324951172, "step": 94}
{"info/global_step": 95, "train_info/time_within_train_step": 32.209585189819336, "step": 95}
{"train_info/time_between_train_steps": 0.007891416549682617, "step": 95}
{"info/global_step": 96, "train_info/time_within_train_step": 31.819197416305542, "step": 96}
{"train_info/time_between_train_steps": 0.005233049392700195, "step": 96}
{"info/global_step": 97, "train_info/time_within_train_step": 32.28827404975891, "step": 97}
{"train_info/time_between_train_steps": 0.005190610885620117, "step": 97}
{"info/global_step": 98, "train_info/time_within_train_step": 32.579933166503906, "step": 98}
{"train_info/time_between_train_steps": 0.005327939987182617, "step": 98}
{"info/global_step": 99, "train_info/time_within_train_step": 32.31179189682007, "step": 99}
{"train_info/time_between_train_steps": 0.005142688751220703, "step": 99}
{"info/global_step": 100, "train_info/time_within_train_step": 32.38883376121521, "step": 100}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733731236, "_runtime": 3326}, "step": 100}
{"logs": {"train/loss": 6.8029, "train/learning_rate": 0.0005, "train/epoch": 3.02, "_timestamp": 1733731236, "_runtime": 3326}, "step": 100}
{"train_info/time_between_train_steps": 2.4117114543914795, "step": 100}
{"info/global_step": 101, "train_info/time_within_train_step": 32.610857248306274, "step": 101}
{"train_info/time_between_train_steps": 0.005159854888916016, "step": 101}
{"info/global_step": 102, "train_info/time_within_train_step": 32.398664236068726, "step": 102}
{"train_info/time_between_train_steps": 0.005468130111694336, "step": 102}
{"info/global_step": 103, "train_info/time_within_train_step": 32.531535387039185, "step": 103}
{"train_info/time_between_train_steps": 0.005291938781738281, "step": 103}
{"info/global_step": 104, "train_info/time_within_train_step": 32.50773310661316, "step": 104}
{"train_info/time_between_train_steps": 0.007621288299560547, "step": 104}
{"info/global_step": 105, "train_info/time_within_train_step": 32.501198530197144, "step": 105}
{"train_info/time_between_train_steps": 0.005516767501831055, "step": 105}
{"info/global_step": 106, "train_info/time_within_train_step": 32.503612756729126, "step": 106}
{"train_info/time_between_train_steps": 0.005551576614379883, "step": 106}
{"info/global_step": 107, "train_info/time_within_train_step": 31.9939444065094, "step": 107}
{"train_info/time_between_train_steps": 0.006025791168212891, "step": 107}
{"info/global_step": 108, "train_info/time_within_train_step": 32.67536425590515, "step": 108}
{"train_info/time_between_train_steps": 0.011773109436035156, "step": 108}
{"train_info/time_between_train_steps": 23.90402841567993, "step": 108}
{"info/global_step": 109, "train_info/time_within_train_step": 32.447810888290405, "step": 109}
{"train_info/time_between_train_steps": 0.0057830810546875, "step": 109}
{"info/global_step": 110, "train_info/time_within_train_step": 32.77566480636597, "step": 110}
{"train_info/time_between_train_steps": 0.005543708801269531, "step": 110}
{"info/global_step": 111, "train_info/time_within_train_step": 32.46916460990906, "step": 111}
{"train_info/time_between_train_steps": 0.005990505218505859, "step": 111}
{"info/global_step": 112, "train_info/time_within_train_step": 32.79366755485535, "step": 112}
{"train_info/time_between_train_steps": 0.0054931640625, "step": 112}
{"info/global_step": 113, "train_info/time_within_train_step": 32.70435667037964, "step": 113}
{"train_info/time_between_train_steps": 0.010869026184082031, "step": 113}
{"info/global_step": 114, "train_info/time_within_train_step": 32.74010729789734, "step": 114}
{"train_info/time_between_train_steps": 0.008101224899291992, "step": 114}
{"info/global_step": 115, "train_info/time_within_train_step": 32.58435606956482, "step": 115}
{"train_info/time_between_train_steps": 0.005648612976074219, "step": 115}
{"info/global_step": 116, "train_info/time_within_train_step": 32.583110332489014, "step": 116}
{"train_info/time_between_train_steps": 0.008628368377685547, "step": 116}
{"info/global_step": 117, "train_info/time_within_train_step": 31.99394154548645, "step": 117}
{"train_info/time_between_train_steps": 0.005329608917236328, "step": 117}
{"info/global_step": 118, "train_info/time_within_train_step": 32.25212860107422, "step": 118}
{"train_info/time_between_train_steps": 0.010208368301391602, "step": 118}
{"info/global_step": 119, "train_info/time_within_train_step": 32.298194885253906, "step": 119}
{"train_info/time_between_train_steps": 0.010580778121948242, "step": 119}
{"info/global_step": 120, "train_info/time_within_train_step": 32.52668595314026, "step": 120}
{"train_info/time_between_train_steps": 0.005211830139160156, "step": 120}
{"info/global_step": 121, "train_info/time_within_train_step": 32.607372522354126, "step": 121}
{"train_info/time_between_train_steps": 0.005261659622192383, "step": 121}
{"info/global_step": 122, "train_info/time_within_train_step": 32.66001892089844, "step": 122}
{"train_info/time_between_train_steps": 0.01020359992980957, "step": 122}
{"info/global_step": 123, "train_info/time_within_train_step": 32.54248309135437, "step": 123}
{"train_info/time_between_train_steps": 0.010389089584350586, "step": 123}
{"info/global_step": 124, "train_info/time_within_train_step": 32.63837122917175, "step": 124}
{"train_info/time_between_train_steps": 0.006478786468505859, "step": 124}
{"info/global_step": 125, "train_info/time_within_train_step": 32.637964963912964, "step": 125}
{"train_info/time_between_train_steps": 0.007994890213012695, "step": 125}
{"info/global_step": 126, "train_info/time_within_train_step": 32.648393869400024, "step": 126}
{"train_info/time_between_train_steps": 0.005290985107421875, "step": 126}
{"info/global_step": 127, "train_info/time_within_train_step": 32.568790674209595, "step": 127}
{"train_info/time_between_train_steps": 0.007803440093994141, "step": 127}
{"info/global_step": 128, "train_info/time_within_train_step": 32.35832738876343, "step": 128}
{"train_info/time_between_train_steps": 0.005267620086669922, "step": 128}
{"info/global_step": 129, "train_info/time_within_train_step": 32.122462034225464, "step": 129}
{"train_info/time_between_train_steps": 0.005375862121582031, "step": 129}
{"info/global_step": 130, "train_info/time_within_train_step": 32.14583444595337, "step": 130}
{"train_info/time_between_train_steps": 0.005419254302978516, "step": 130}
{"info/global_step": 131, "train_info/time_within_train_step": 32.50440216064453, "step": 131}
{"train_info/time_between_train_steps": 0.00587916374206543, "step": 131}
{"info/global_step": 132, "train_info/time_within_train_step": 32.26692748069763, "step": 132}
{"train_info/time_between_train_steps": 0.005728006362915039, "step": 132}
{"info/global_step": 133, "train_info/time_within_train_step": 32.70529556274414, "step": 133}
{"train_info/time_between_train_steps": 0.005585908889770508, "step": 133}
{"info/global_step": 134, "train_info/time_within_train_step": 32.60060000419617, "step": 134}
{"train_info/time_between_train_steps": 0.005721092224121094, "step": 134}
{"info/global_step": 135, "train_info/time_within_train_step": 32.567824363708496, "step": 135}
{"train_info/time_between_train_steps": 0.005921363830566406, "step": 135}
{"train_info/time_between_train_steps": 23.553526878356934, "step": 135}
{"info/global_step": 136, "train_info/time_within_train_step": 32.48542785644531, "step": 136}
{"train_info/time_between_train_steps": 0.0057413578033447266, "step": 136}
{"info/global_step": 137, "train_info/time_within_train_step": 32.84094500541687, "step": 137}
{"train_info/time_between_train_steps": 0.00548243522644043, "step": 137}
{"info/global_step": 138, "train_info/time_within_train_step": 32.80242848396301, "step": 138}
{"train_info/time_between_train_steps": 0.005598545074462891, "step": 138}
{"info/global_step": 139, "train_info/time_within_train_step": 32.0906035900116, "step": 139}
{"train_info/time_between_train_steps": 0.011701583862304688, "step": 139}
{"info/global_step": 140, "train_info/time_within_train_step": 32.467095375061035, "step": 140}
{"train_info/time_between_train_steps": 0.011652708053588867, "step": 140}
{"info/global_step": 141, "train_info/time_within_train_step": 32.713786602020264, "step": 141}
{"train_info/time_between_train_steps": 0.005730152130126953, "step": 141}
{"info/global_step": 142, "train_info/time_within_train_step": 32.53188753128052, "step": 142}
{"train_info/time_between_train_steps": 0.005874156951904297, "step": 142}
{"info/global_step": 143, "train_info/time_within_train_step": 32.51975703239441, "step": 143}
{"train_info/time_between_train_steps": 0.007168769836425781, "step": 143}
{"info/global_step": 144, "train_info/time_within_train_step": 32.66787314414978, "step": 144}
{"train_info/time_between_train_steps": 0.007295131683349609, "step": 144}
{"info/global_step": 145, "train_info/time_within_train_step": 32.36415910720825, "step": 145}
{"train_info/time_between_train_steps": 0.005171060562133789, "step": 145}
{"info/global_step": 146, "train_info/time_within_train_step": 32.26220726966858, "step": 146}
{"train_info/time_between_train_steps": 0.008688688278198242, "step": 146}
{"info/global_step": 147, "train_info/time_within_train_step": 32.70548629760742, "step": 147}
{"train_info/time_between_train_steps": 0.0052607059478759766, "step": 147}
{"info/global_step": 148, "train_info/time_within_train_step": 32.567047357559204, "step": 148}
{"train_info/time_between_train_steps": 0.007469654083251953, "step": 148}
{"info/global_step": 149, "train_info/time_within_train_step": 32.56263780593872, "step": 149}
{"train_info/time_between_train_steps": 0.005278110504150391, "step": 149}
{"info/global_step": 150, "train_info/time_within_train_step": 31.864696502685547, "step": 150}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733732911, "_runtime": 5001}, "step": 150}
{"logs": {"train/loss": 6.7121, "train/learning_rate": 0.0005833333333333333, "train/epoch": 5.01, "_timestamp": 1733732911, "_runtime": 5001}, "step": 150}
{"train_info/time_between_train_steps": 0.008316993713378906, "step": 150}
{"info/global_step": 151, "train_info/time_within_train_step": 32.42517447471619, "step": 151}
{"train_info/time_between_train_steps": 0.005280256271362305, "step": 151}
{"info/global_step": 152, "train_info/time_within_train_step": 32.679200410842896, "step": 152}
{"train_info/time_between_train_steps": 0.0053784847259521484, "step": 152}
{"info/global_step": 153, "train_info/time_within_train_step": 32.432223081588745, "step": 153}
{"train_info/time_between_train_steps": 0.010762691497802734, "step": 153}
{"info/global_step": 154, "train_info/time_within_train_step": 32.51068091392517, "step": 154}
{"train_info/time_between_train_steps": 0.005321502685546875, "step": 154}
{"info/global_step": 155, "train_info/time_within_train_step": 32.25856614112854, "step": 155}
{"train_info/time_between_train_steps": 0.005299568176269531, "step": 155}
{"info/global_step": 156, "train_info/time_within_train_step": 32.73437857627869, "step": 156}
{"train_info/time_between_train_steps": 0.005466938018798828, "step": 156}
{"info/global_step": 157, "train_info/time_within_train_step": 32.552634954452515, "step": 157}
{"train_info/time_between_train_steps": 0.005405902862548828, "step": 157}
{"info/global_step": 158, "train_info/time_within_train_step": 32.19252252578735, "step": 158}
{"train_info/time_between_train_steps": 0.0071604251861572266, "step": 158}
{"info/global_step": 159, "train_info/time_within_train_step": 32.48271560668945, "step": 159}
{"train_info/time_between_train_steps": 0.0056285858154296875, "step": 159}
{"info/global_step": 160, "train_info/time_within_train_step": 32.578574895858765, "step": 160}
{"train_info/time_between_train_steps": 0.005590200424194336, "step": 160}
{"info/global_step": 161, "train_info/time_within_train_step": 32.155471086502075, "step": 161}
{"train_info/time_between_train_steps": 0.006051540374755859, "step": 161}
{"info/global_step": 162, "train_info/time_within_train_step": 32.072571754455566, "step": 162}
{"train_info/time_between_train_steps": 0.010983943939208984, "step": 162}
{"train_info/time_between_train_steps": 23.380637884140015, "step": 162}
{"info/global_step": 163, "train_info/time_within_train_step": 32.33048152923584, "step": 163}
{"train_info/time_between_train_steps": 0.0056073665618896484, "step": 163}
{"info/global_step": 164, "train_info/time_within_train_step": 32.619622230529785, "step": 164}
{"train_info/time_between_train_steps": 0.005545139312744141, "step": 164}
{"info/global_step": 165, "train_info/time_within_train_step": 32.23409390449524, "step": 165}
{"train_info/time_between_train_steps": 0.008275747299194336, "step": 165}
{"info/global_step": 166, "train_info/time_within_train_step": 32.78898763656616, "step": 166}
{"train_info/time_between_train_steps": 0.006147623062133789, "step": 166}
{"info/global_step": 167, "train_info/time_within_train_step": 32.59888005256653, "step": 167}
{"train_info/time_between_train_steps": 0.006243228912353516, "step": 167}
{"info/global_step": 168, "train_info/time_within_train_step": 32.50809454917908, "step": 168}
{"train_info/time_between_train_steps": 0.008335590362548828, "step": 168}
{"info/global_step": 169, "train_info/time_within_train_step": 32.757012128829956, "step": 169}
{"train_info/time_between_train_steps": 0.005786895751953125, "step": 169}
{"info/global_step": 170, "train_info/time_within_train_step": 32.833399295806885, "step": 170}
{"train_info/time_between_train_steps": 0.005609989166259766, "step": 170}
{"info/global_step": 171, "train_info/time_within_train_step": 32.6693902015686, "step": 171}
{"train_info/time_between_train_steps": 0.0054225921630859375, "step": 171}
{"info/global_step": 172, "train_info/time_within_train_step": 31.96688461303711, "step": 172}
{"train_info/time_between_train_steps": 0.005198240280151367, "step": 172}
{"info/global_step": 173, "train_info/time_within_train_step": 32.425636768341064, "step": 173}
{"train_info/time_between_train_steps": 0.005220174789428711, "step": 173}
{"info/global_step": 174, "train_info/time_within_train_step": 32.53635501861572, "step": 174}
{"train_info/time_between_train_steps": 0.0052797794342041016, "step": 174}
{"info/global_step": 175, "train_info/time_within_train_step": 32.24324917793274, "step": 175}
{"train_info/time_between_train_steps": 0.0051958560943603516, "step": 175}
{"info/global_step": 176, "train_info/time_within_train_step": 32.4649703502655, "step": 176}
{"train_info/time_between_train_steps": 0.005271196365356445, "step": 176}
{"info/global_step": 177, "train_info/time_within_train_step": 32.54072165489197, "step": 177}
{"train_info/time_between_train_steps": 0.005265712738037109, "step": 177}
{"info/global_step": 178, "train_info/time_within_train_step": 32.47904944419861, "step": 178}
{"train_info/time_between_train_steps": 0.005516767501831055, "step": 178}
{"info/global_step": 179, "train_info/time_within_train_step": 32.5939302444458, "step": 179}
{"train_info/time_between_train_steps": 0.0053865909576416016, "step": 179}
{"info/global_step": 180, "train_info/time_within_train_step": 32.60504961013794, "step": 180}
{"train_info/time_between_train_steps": 0.009828567504882812, "step": 180}
{"info/global_step": 181, "train_info/time_within_train_step": 32.66117310523987, "step": 181}
{"train_info/time_between_train_steps": 0.0052721500396728516, "step": 181}
{"info/global_step": 182, "train_info/time_within_train_step": 32.428831577301025, "step": 182}
{"train_info/time_between_train_steps": 0.006880044937133789, "step": 182}
{"info/global_step": 183, "train_info/time_within_train_step": 32.068360328674316, "step": 183}
{"train_info/time_between_train_steps": 0.005422353744506836, "step": 183}
{"info/global_step": 184, "train_info/time_within_train_step": 32.429110050201416, "step": 184}
{"train_info/time_between_train_steps": 0.005430698394775391, "step": 184}
{"info/global_step": 185, "train_info/time_within_train_step": 32.207594871520996, "step": 185}
{"train_info/time_between_train_steps": 0.007444143295288086, "step": 185}
{"info/global_step": 186, "train_info/time_within_train_step": 32.69059205055237, "step": 186}
{"train_info/time_between_train_steps": 0.012813329696655273, "step": 186}
{"info/global_step": 187, "train_info/time_within_train_step": 32.29391813278198, "step": 187}
{"train_info/time_between_train_steps": 0.005809783935546875, "step": 187}
{"info/global_step": 188, "train_info/time_within_train_step": 32.441277742385864, "step": 188}
{"train_info/time_between_train_steps": 0.009009122848510742, "step": 188}
{"info/global_step": 189, "train_info/time_within_train_step": 32.47599148750305, "step": 189}
{"train_info/time_between_train_steps": 0.007350921630859375, "step": 189}
{"train_info/time_between_train_steps": 23.625925064086914, "step": 189}
{"info/global_step": 190, "train_info/time_within_train_step": 32.666550636291504, "step": 190}
{"train_info/time_between_train_steps": 0.010537862777709961, "step": 190}
{"info/global_step": 191, "train_info/time_within_train_step": 32.40535306930542, "step": 191}
{"train_info/time_between_train_steps": 0.005440473556518555, "step": 191}
{"info/global_step": 192, "train_info/time_within_train_step": 32.600043296813965, "step": 192}
{"train_info/time_between_train_steps": 0.005279541015625, "step": 192}
{"info/global_step": 193, "train_info/time_within_train_step": 32.59039044380188, "step": 193}
{"train_info/time_between_train_steps": 0.0065975189208984375, "step": 193}
{"info/global_step": 194, "train_info/time_within_train_step": 31.956440687179565, "step": 194}
{"train_info/time_between_train_steps": 0.005442380905151367, "step": 194}
{"info/global_step": 195, "train_info/time_within_train_step": 32.46985173225403, "step": 195}
{"train_info/time_between_train_steps": 0.013511419296264648, "step": 195}
{"info/global_step": 196, "train_info/time_within_train_step": 32.37282586097717, "step": 196}
{"train_info/time_between_train_steps": 0.0056285858154296875, "step": 196}
{"info/global_step": 197, "train_info/time_within_train_step": 32.42748737335205, "step": 197}
{"train_info/time_between_train_steps": 0.006499290466308594, "step": 197}
{"info/global_step": 198, "train_info/time_within_train_step": 32.314234018325806, "step": 198}
{"train_info/time_between_train_steps": 0.005264997482299805, "step": 198}
{"info/global_step": 199, "train_info/time_within_train_step": 32.78767991065979, "step": 199}
{"train_info/time_between_train_steps": 0.008960723876953125, "step": 199}
{"info/global_step": 200, "train_info/time_within_train_step": 32.405858516693115, "step": 200}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733734581, "_runtime": 6671}, "step": 200}
{"logs": {"train/loss": 6.6454, "train/learning_rate": 0.0005555555555555556, "train/epoch": 7.01, "_timestamp": 1733734581, "_runtime": 6671}, "step": 200}
{"train_info/time_between_train_steps": 2.5998942852020264, "step": 200}
{"info/global_step": 201, "train_info/time_within_train_step": 32.40394830703735, "step": 201}
{"train_info/time_between_train_steps": 0.01089787483215332, "step": 201}
{"info/global_step": 202, "train_info/time_within_train_step": 32.63244819641113, "step": 202}
{"train_info/time_between_train_steps": 0.005293846130371094, "step": 202}
{"info/global_step": 203, "train_info/time_within_train_step": 32.45321273803711, "step": 203}
{"train_info/time_between_train_steps": 0.005435943603515625, "step": 203}
{"info/global_step": 204, "train_info/time_within_train_step": 32.53374671936035, "step": 204}
{"train_info/time_between_train_steps": 0.005318880081176758, "step": 204}
{"info/global_step": 205, "train_info/time_within_train_step": 32.56669545173645, "step": 205}
{"train_info/time_between_train_steps": 0.005262613296508789, "step": 205}
{"info/global_step": 206, "train_info/time_within_train_step": 31.935059309005737, "step": 206}
{"train_info/time_between_train_steps": 0.005326986312866211, "step": 206}
{"info/global_step": 207, "train_info/time_within_train_step": 32.30827617645264, "step": 207}
{"train_info/time_between_train_steps": 0.01011967658996582, "step": 207}
{"info/global_step": 208, "train_info/time_within_train_step": 32.43762707710266, "step": 208}
{"train_info/time_between_train_steps": 0.00531315803527832, "step": 208}
{"info/global_step": 209, "train_info/time_within_train_step": 32.48708176612854, "step": 209}
{"train_info/time_between_train_steps": 0.00531458854675293, "step": 209}
{"info/global_step": 210, "train_info/time_within_train_step": 32.048933267593384, "step": 210}
{"train_info/time_between_train_steps": 0.005308389663696289, "step": 210}
{"info/global_step": 211, "train_info/time_within_train_step": 32.55812120437622, "step": 211}
{"train_info/time_between_train_steps": 0.005739450454711914, "step": 211}
{"info/global_step": 212, "train_info/time_within_train_step": 32.49545383453369, "step": 212}
{"train_info/time_between_train_steps": 0.009584903717041016, "step": 212}
{"info/global_step": 213, "train_info/time_within_train_step": 32.828173875808716, "step": 213}
{"train_info/time_between_train_steps": 0.005856037139892578, "step": 213}
{"info/global_step": 214, "train_info/time_within_train_step": 32.46050024032593, "step": 214}
{"train_info/time_between_train_steps": 0.010124444961547852, "step": 214}
{"info/global_step": 215, "train_info/time_within_train_step": 32.90741229057312, "step": 215}
{"train_info/time_between_train_steps": 0.006058692932128906, "step": 215}
{"info/global_step": 216, "train_info/time_within_train_step": 32.634214639663696, "step": 216}
{"train_info/time_between_train_steps": 0.005964040756225586, "step": 216}
{"train_info/time_between_train_steps": 23.442328453063965, "step": 216}
{"info/global_step": 217, "train_info/time_within_train_step": 31.778103351593018, "step": 217}
{"train_info/time_between_train_steps": 0.005990743637084961, "step": 217}
{"info/global_step": 218, "train_info/time_within_train_step": 32.667701959609985, "step": 218}
{"train_info/time_between_train_steps": 0.0056421756744384766, "step": 218}
{"info/global_step": 219, "train_info/time_within_train_step": 32.44559359550476, "step": 219}
{"train_info/time_between_train_steps": 0.005861520767211914, "step": 219}
{"info/global_step": 220, "train_info/time_within_train_step": 32.606083393096924, "step": 220}
{"train_info/time_between_train_steps": 0.005445957183837891, "step": 220}
{"info/global_step": 221, "train_info/time_within_train_step": 32.26065993309021, "step": 221}
{"train_info/time_between_train_steps": 0.005754232406616211, "step": 221}
{"info/global_step": 222, "train_info/time_within_train_step": 32.58308506011963, "step": 222}
{"train_info/time_between_train_steps": 0.012792587280273438, "step": 222}
{"info/global_step": 223, "train_info/time_within_train_step": 32.71584153175354, "step": 223}
{"train_info/time_between_train_steps": 0.005893230438232422, "step": 223}
{"info/global_step": 224, "train_info/time_within_train_step": 32.639975786209106, "step": 224}
{"train_info/time_between_train_steps": 0.005465030670166016, "step": 224}
{"info/global_step": 225, "train_info/time_within_train_step": 32.609822511672974, "step": 225}
{"train_info/time_between_train_steps": 0.005136013031005859, "step": 225}
{"info/global_step": 226, "train_info/time_within_train_step": 32.4238555431366, "step": 226}
{"train_info/time_between_train_steps": 0.0052318572998046875, "step": 226}
{"info/global_step": 227, "train_info/time_within_train_step": 32.622902393341064, "step": 227}
{"train_info/time_between_train_steps": 0.007096290588378906, "step": 227}
{"info/global_step": 228, "train_info/time_within_train_step": 32.67429733276367, "step": 228}
{"train_info/time_between_train_steps": 0.0051839351654052734, "step": 228}
{"info/global_step": 229, "train_info/time_within_train_step": 31.78616189956665, "step": 229}
{"train_info/time_between_train_steps": 0.0052568912506103516, "step": 229}
{"info/global_step": 230, "train_info/time_within_train_step": 32.57581281661987, "step": 230}
{"train_info/time_between_train_steps": 0.005162954330444336, "step": 230}
{"info/global_step": 231, "train_info/time_within_train_step": 32.222917795181274, "step": 231}
{"train_info/time_between_train_steps": 0.0053958892822265625, "step": 231}
{"info/global_step": 232, "train_info/time_within_train_step": 32.0310537815094, "step": 232}
{"train_info/time_between_train_steps": 0.005434274673461914, "step": 232}
{"info/global_step": 233, "train_info/time_within_train_step": 32.28780126571655, "step": 233}
{"train_info/time_between_train_steps": 0.007433652877807617, "step": 233}
{"info/global_step": 234, "train_info/time_within_train_step": 32.628666162490845, "step": 234}
{"train_info/time_between_train_steps": 0.0053272247314453125, "step": 234}
{"info/global_step": 235, "train_info/time_within_train_step": 32.52692794799805, "step": 235}
{"train_info/time_between_train_steps": 0.008682727813720703, "step": 235}
{"info/global_step": 236, "train_info/time_within_train_step": 32.0473473072052, "step": 236}
{"train_info/time_between_train_steps": 0.005234241485595703, "step": 236}
{"info/global_step": 237, "train_info/time_within_train_step": 27.725189924240112, "step": 237}
{"train_info/time_between_train_steps": 0.005344390869140625, "step": 237}
{"info/global_step": 238, "train_info/time_within_train_step": 27.742503881454468, "step": 238}
{"train_info/time_between_train_steps": 0.010245561599731445, "step": 238}
{"info/global_step": 239, "train_info/time_within_train_step": 27.716397285461426, "step": 239}
{"train_info/time_between_train_steps": 0.010390281677246094, "step": 239}
{"info/global_step": 240, "train_info/time_within_train_step": 27.723726272583008, "step": 240}
{"train_info/time_between_train_steps": 0.010617494583129883, "step": 240}
{"info/global_step": 241, "train_info/time_within_train_step": 28.08528447151184, "step": 241}
{"train_info/time_between_train_steps": 0.005606412887573242, "step": 241}
{"info/global_step": 242, "train_info/time_within_train_step": 32.017632246017456, "step": 242}
{"train_info/time_between_train_steps": 0.006241798400878906, "step": 242}
{"info/global_step": 243, "train_info/time_within_train_step": 33.11028790473938, "step": 243}
{"train_info/time_between_train_steps": 0.0060465335845947266, "step": 243}
{"train_info/time_between_train_steps": 23.988424062728882, "step": 243}
{"info/global_step": 244, "train_info/time_within_train_step": 33.28675556182861, "step": 244}
{"train_info/time_between_train_steps": 0.010713338851928711, "step": 244}
{"info/global_step": 245, "train_info/time_within_train_step": 33.713327407836914, "step": 245}
{"train_info/time_between_train_steps": 0.008015632629394531, "step": 245}
{"info/global_step": 246, "train_info/time_within_train_step": 33.40636730194092, "step": 246}
{"train_info/time_between_train_steps": 0.009162425994873047, "step": 246}
{"info/global_step": 247, "train_info/time_within_train_step": 33.54464244842529, "step": 247}
{"train_info/time_between_train_steps": 0.010618925094604492, "step": 247}
{"info/global_step": 248, "train_info/time_within_train_step": 33.29127216339111, "step": 248}
{"train_info/time_between_train_steps": 0.005620718002319336, "step": 248}
{"info/global_step": 249, "train_info/time_within_train_step": 33.27141785621643, "step": 249}
{"train_info/time_between_train_steps": 0.009466409683227539, "step": 249}
{"info/global_step": 250, "train_info/time_within_train_step": 31.888800382614136, "step": 250}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733736236, "_runtime": 8326}, "step": 250}
{"logs": {"train/loss": 6.5652, "train/learning_rate": 0.0005277777777777777, "train/epoch": 9.01, "_timestamp": 1733736236, "_runtime": 8326}, "step": 250}
{"train_info/time_between_train_steps": 0.007446765899658203, "step": 250}
{"info/global_step": 251, "train_info/time_within_train_step": 33.05218052864075, "step": 251}
{"train_info/time_between_train_steps": 0.005509376525878906, "step": 251}
{"info/global_step": 252, "train_info/time_within_train_step": 33.398144483566284, "step": 252}
{"train_info/time_between_train_steps": 0.008928060531616211, "step": 252}
{"info/global_step": 253, "train_info/time_within_train_step": 33.13544750213623, "step": 253}
{"train_info/time_between_train_steps": 0.005171537399291992, "step": 253}
{"info/global_step": 254, "train_info/time_within_train_step": 33.314164876937866, "step": 254}
{"train_info/time_between_train_steps": 0.005129814147949219, "step": 254}
{"info/global_step": 255, "train_info/time_within_train_step": 33.28737449645996, "step": 255}
{"train_info/time_between_train_steps": 0.005349397659301758, "step": 255}
{"info/global_step": 256, "train_info/time_within_train_step": 33.140156745910645, "step": 256}
{"train_info/time_between_train_steps": 0.007890462875366211, "step": 256}
{"info/global_step": 257, "train_info/time_within_train_step": 33.242947816848755, "step": 257}
{"train_info/time_between_train_steps": 0.005260467529296875, "step": 257}
{"info/global_step": 258, "train_info/time_within_train_step": 33.063705921173096, "step": 258}
{"train_info/time_between_train_steps": 0.008007287979125977, "step": 258}
{"info/global_step": 259, "train_info/time_within_train_step": 33.35023331642151, "step": 259}
{"train_info/time_between_train_steps": 0.005239248275756836, "step": 259}
{"info/global_step": 260, "train_info/time_within_train_step": 33.23584818840027, "step": 260}
{"train_info/time_between_train_steps": 0.005283355712890625, "step": 260}
{"info/global_step": 261, "train_info/time_within_train_step": 33.048805236816406, "step": 261}
{"train_info/time_between_train_steps": 0.005037784576416016, "step": 261}
{"info/global_step": 262, "train_info/time_within_train_step": 33.04968976974487, "step": 262}
{"train_info/time_between_train_steps": 0.008376121520996094, "step": 262}
{"info/global_step": 263, "train_info/time_within_train_step": 33.3747615814209, "step": 263}
{"train_info/time_between_train_steps": 0.005288362503051758, "step": 263}
{"info/global_step": 264, "train_info/time_within_train_step": 33.499961137771606, "step": 264}
{"train_info/time_between_train_steps": 0.005263328552246094, "step": 264}
{"info/global_step": 265, "train_info/time_within_train_step": 33.380064964294434, "step": 265}
{"train_info/time_between_train_steps": 0.005234718322753906, "step": 265}
{"info/global_step": 266, "train_info/time_within_train_step": 33.232165575027466, "step": 266}
{"train_info/time_between_train_steps": 0.005263328552246094, "step": 266}
{"info/global_step": 267, "train_info/time_within_train_step": 33.16843509674072, "step": 267}
{"train_info/time_between_train_steps": 0.005673885345458984, "step": 267}
{"info/global_step": 268, "train_info/time_within_train_step": 28.51536536216736, "step": 268}
{"train_info/time_between_train_steps": 0.01188516616821289, "step": 268}
{"info/global_step": 269, "train_info/time_within_train_step": 28.10352110862732, "step": 269}
{"train_info/time_between_train_steps": 0.011636018753051758, "step": 269}
{"info/global_step": 270, "train_info/time_within_train_step": 27.780179023742676, "step": 270}
{"train_info/time_between_train_steps": 0.0058362483978271484, "step": 270}
{"train_info/time_between_train_steps": 20.707744359970093, "step": 270}
{"info/global_step": 271, "train_info/time_within_train_step": 27.732093334197998, "step": 271}
{"train_info/time_between_train_steps": 0.0056607723236083984, "step": 271}
{"info/global_step": 272, "train_info/time_within_train_step": 27.910317182540894, "step": 272}
{"train_info/time_between_train_steps": 0.005409717559814453, "step": 272}
{"info/global_step": 273, "train_info/time_within_train_step": 27.744736671447754, "step": 273}
{"train_info/time_between_train_steps": 0.005286216735839844, "step": 273}
{"info/global_step": 274, "train_info/time_within_train_step": 27.917640447616577, "step": 274}
{"train_info/time_between_train_steps": 0.005404233932495117, "step": 274}
{"info/global_step": 275, "train_info/time_within_train_step": 27.82577872276306, "step": 275}
{"train_info/time_between_train_steps": 0.005451202392578125, "step": 275}
{"info/global_step": 276, "train_info/time_within_train_step": 27.891828298568726, "step": 276}
{"train_info/time_between_train_steps": 0.010924577713012695, "step": 276}
{"info/global_step": 277, "train_info/time_within_train_step": 27.717331409454346, "step": 277}
{"train_info/time_between_train_steps": 0.008262872695922852, "step": 277}
{"info/global_step": 278, "train_info/time_within_train_step": 27.779268741607666, "step": 278}
{"train_info/time_between_train_steps": 0.010529041290283203, "step": 278}
{"info/global_step": 279, "train_info/time_within_train_step": 27.70915961265564, "step": 279}
{"train_info/time_between_train_steps": 0.010143041610717773, "step": 279}
{"info/global_step": 280, "train_info/time_within_train_step": 27.702081203460693, "step": 280}
{"train_info/time_between_train_steps": 0.010216236114501953, "step": 280}
{"info/global_step": 281, "train_info/time_within_train_step": 27.70747685432434, "step": 281}
{"train_info/time_between_train_steps": 0.005129098892211914, "step": 281}
{"info/global_step": 282, "train_info/time_within_train_step": 27.696120023727417, "step": 282}
{"train_info/time_between_train_steps": 0.005152225494384766, "step": 282}
{"info/global_step": 283, "train_info/time_within_train_step": 27.69448685646057, "step": 283}
{"train_info/time_between_train_steps": 0.005094289779663086, "step": 283}
{"info/global_step": 284, "train_info/time_within_train_step": 27.695181131362915, "step": 284}
{"train_info/time_between_train_steps": 0.005211591720581055, "step": 284}
{"info/global_step": 285, "train_info/time_within_train_step": 27.694873571395874, "step": 285}
{"train_info/time_between_train_steps": 0.005132198333740234, "step": 285}
{"info/global_step": 286, "train_info/time_within_train_step": 27.69694757461548, "step": 286}
{"train_info/time_between_train_steps": 0.005182027816772461, "step": 286}
{"info/global_step": 287, "train_info/time_within_train_step": 27.69467568397522, "step": 287}
{"train_info/time_between_train_steps": 0.005160331726074219, "step": 287}
{"info/global_step": 288, "train_info/time_within_train_step": 27.69716501235962, "step": 288}
{"train_info/time_between_train_steps": 0.005303144454956055, "step": 288}
{"info/global_step": 289, "train_info/time_within_train_step": 27.69486951828003, "step": 289}
{"train_info/time_between_train_steps": 0.005105018615722656, "step": 289}
{"info/global_step": 290, "train_info/time_within_train_step": 27.689449548721313, "step": 290}
{"train_info/time_between_train_steps": 0.0050506591796875, "step": 290}
{"info/global_step": 291, "train_info/time_within_train_step": 27.794764757156372, "step": 291}
{"train_info/time_between_train_steps": 0.005352497100830078, "step": 291}
{"info/global_step": 292, "train_info/time_within_train_step": 27.699248790740967, "step": 292}
{"train_info/time_between_train_steps": 0.005152225494384766, "step": 292}
{"info/global_step": 293, "train_info/time_within_train_step": 27.699862480163574, "step": 293}
{"train_info/time_between_train_steps": 0.005132913589477539, "step": 293}
{"info/global_step": 294, "train_info/time_within_train_step": 27.700517177581787, "step": 294}
{"train_info/time_between_train_steps": 0.005300283432006836, "step": 294}
{"info/global_step": 295, "train_info/time_within_train_step": 27.702357292175293, "step": 295}
{"train_info/time_between_train_steps": 0.00556492805480957, "step": 295}
{"info/global_step": 296, "train_info/time_within_train_step": 27.711589097976685, "step": 296}
{"train_info/time_between_train_steps": 0.010672807693481445, "step": 296}
{"info/global_step": 297, "train_info/time_within_train_step": 27.731338500976562, "step": 297}
{"train_info/time_between_train_steps": 0.005391836166381836, "step": 297}
{"train_info/time_between_train_steps": 20.44498300552368, "step": 297}
{"info/global_step": 298, "train_info/time_within_train_step": 27.69527316093445, "step": 298}
{"train_info/time_between_train_steps": 0.004997730255126953, "step": 298}
{"info/global_step": 299, "train_info/time_within_train_step": 27.78650689125061, "step": 299}
{"train_info/time_between_train_steps": 0.004981517791748047, "step": 299}
{"info/global_step": 300, "train_info/time_within_train_step": 27.68814754486084, "step": 300}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733737759, "_runtime": 9849}, "step": 300}
{"logs": {"train/loss": 6.3824, "train/learning_rate": 0.0005, "train/epoch": 11.0, "_timestamp": 1733737759, "_runtime": 9849}, "step": 300}
{"train_info/time_between_train_steps": 2.4416441917419434, "step": 300}
{"info/global_step": 301, "train_info/time_within_train_step": 27.838789463043213, "step": 301}
{"train_info/time_between_train_steps": 0.005288362503051758, "step": 301}
{"info/global_step": 302, "train_info/time_within_train_step": 27.706849575042725, "step": 302}
{"train_info/time_between_train_steps": 0.005296468734741211, "step": 302}
{"info/global_step": 303, "train_info/time_within_train_step": 27.824729681015015, "step": 303}
{"train_info/time_between_train_steps": 0.0052928924560546875, "step": 303}
{"info/global_step": 304, "train_info/time_within_train_step": 27.70437264442444, "step": 304}
{"train_info/time_between_train_steps": 0.005217552185058594, "step": 304}
{"info/global_step": 305, "train_info/time_within_train_step": 27.735100984573364, "step": 305}
{"train_info/time_between_train_steps": 0.005272865295410156, "step": 305}
{"info/global_step": 306, "train_info/time_within_train_step": 27.81517243385315, "step": 306}
{"train_info/time_between_train_steps": 0.005239963531494141, "step": 306}
{"info/global_step": 307, "train_info/time_within_train_step": 27.69383430480957, "step": 307}
{"train_info/time_between_train_steps": 0.0050771236419677734, "step": 307}
{"info/global_step": 308, "train_info/time_within_train_step": 27.695647478103638, "step": 308}
{"train_info/time_between_train_steps": 0.005199909210205078, "step": 308}
{"info/global_step": 309, "train_info/time_within_train_step": 27.696410179138184, "step": 309}
{"train_info/time_between_train_steps": 0.005011320114135742, "step": 309}
{"info/global_step": 310, "train_info/time_within_train_step": 27.698092460632324, "step": 310}
{"train_info/time_between_train_steps": 0.010306358337402344, "step": 310}
{"info/global_step": 311, "train_info/time_within_train_step": 27.70191216468811, "step": 311}
{"train_info/time_between_train_steps": 0.0052716732025146484, "step": 311}
{"info/global_step": 312, "train_info/time_within_train_step": 27.69701361656189, "step": 312}
{"train_info/time_between_train_steps": 0.005107879638671875, "step": 312}
{"info/global_step": 313, "train_info/time_within_train_step": 27.69958209991455, "step": 313}
{"train_info/time_between_train_steps": 0.005086660385131836, "step": 313}
{"info/global_step": 314, "train_info/time_within_train_step": 27.702370405197144, "step": 314}
{"train_info/time_between_train_steps": 0.0052947998046875, "step": 314}
{"info/global_step": 315, "train_info/time_within_train_step": 27.700125455856323, "step": 315}
{"train_info/time_between_train_steps": 0.005105257034301758, "step": 315}
{"info/global_step": 316, "train_info/time_within_train_step": 27.69654083251953, "step": 316}
{"train_info/time_between_train_steps": 0.005054950714111328, "step": 316}
{"info/global_step": 317, "train_info/time_within_train_step": 27.697227001190186, "step": 317}
{"train_info/time_between_train_steps": 0.005094766616821289, "step": 317}
{"info/global_step": 318, "train_info/time_within_train_step": 27.702545881271362, "step": 318}
{"train_info/time_between_train_steps": 0.0051708221435546875, "step": 318}
{"info/global_step": 319, "train_info/time_within_train_step": 27.70536160469055, "step": 319}
{"train_info/time_between_train_steps": 0.005309343338012695, "step": 319}
{"info/global_step": 320, "train_info/time_within_train_step": 27.69805669784546, "step": 320}
{"train_info/time_between_train_steps": 0.005056619644165039, "step": 320}
{"info/global_step": 321, "train_info/time_within_train_step": 27.708109855651855, "step": 321}
{"train_info/time_between_train_steps": 0.005479335784912109, "step": 321}
{"info/global_step": 322, "train_info/time_within_train_step": 27.814167499542236, "step": 322}
{"train_info/time_between_train_steps": 0.0054361820220947266, "step": 322}
{"info/global_step": 323, "train_info/time_within_train_step": 27.714041233062744, "step": 323}
{"train_info/time_between_train_steps": 0.005610466003417969, "step": 323}
{"info/global_step": 324, "train_info/time_within_train_step": 27.729286432266235, "step": 324}
{"train_info/time_between_train_steps": 0.005563259124755859, "step": 324}
{"train_info/time_between_train_steps": 20.568724632263184, "step": 324}
{"info/global_step": 325, "train_info/time_within_train_step": 27.70559859275818, "step": 325}
{"train_info/time_between_train_steps": 0.005355119705200195, "step": 325}
{"info/global_step": 326, "train_info/time_within_train_step": 27.839161157608032, "step": 326}
{"train_info/time_between_train_steps": 0.0052013397216796875, "step": 326}
{"info/global_step": 327, "train_info/time_within_train_step": 27.704756498336792, "step": 327}
{"train_info/time_between_train_steps": 0.005225419998168945, "step": 327}
{"info/global_step": 328, "train_info/time_within_train_step": 27.81689953804016, "step": 328}
{"train_info/time_between_train_steps": 0.005124807357788086, "step": 328}
{"info/global_step": 329, "train_info/time_within_train_step": 27.702466011047363, "step": 329}
{"train_info/time_between_train_steps": 0.005099296569824219, "step": 329}
{"info/global_step": 330, "train_info/time_within_train_step": 27.82805299758911, "step": 330}
{"train_info/time_between_train_steps": 0.005224704742431641, "step": 330}
{"info/global_step": 331, "train_info/time_within_train_step": 27.7210636138916, "step": 331}
{"train_info/time_between_train_steps": 0.005377531051635742, "step": 331}
{"info/global_step": 332, "train_info/time_within_train_step": 27.763846397399902, "step": 332}
{"train_info/time_between_train_steps": 0.005309104919433594, "step": 332}
{"info/global_step": 333, "train_info/time_within_train_step": 27.70367193222046, "step": 333}
{"train_info/time_between_train_steps": 0.005035400390625, "step": 333}
{"info/global_step": 334, "train_info/time_within_train_step": 27.697436332702637, "step": 334}
{"train_info/time_between_train_steps": 0.005209207534790039, "step": 334}
{"info/global_step": 335, "train_info/time_within_train_step": 27.69590950012207, "step": 335}
{"train_info/time_between_train_steps": 0.005129098892211914, "step": 335}
{"info/global_step": 336, "train_info/time_within_train_step": 27.695242881774902, "step": 336}
{"train_info/time_between_train_steps": 0.005077362060546875, "step": 336}
{"info/global_step": 337, "train_info/time_within_train_step": 27.8159236907959, "step": 337}
{"train_info/time_between_train_steps": 0.00512385368347168, "step": 337}
{"info/global_step": 338, "train_info/time_within_train_step": 27.70241093635559, "step": 338}
{"train_info/time_between_train_steps": 0.005064249038696289, "step": 338}
{"info/global_step": 339, "train_info/time_within_train_step": 27.69467329978943, "step": 339}
{"train_info/time_between_train_steps": 0.005089759826660156, "step": 339}
{"info/global_step": 340, "train_info/time_within_train_step": 27.708123445510864, "step": 340}
{"train_info/time_between_train_steps": 0.0051958560943603516, "step": 340}
{"info/global_step": 341, "train_info/time_within_train_step": 27.704259157180786, "step": 341}
{"train_info/time_between_train_steps": 0.005144596099853516, "step": 341}
{"info/global_step": 342, "train_info/time_within_train_step": 27.695696592330933, "step": 342}
{"train_info/time_between_train_steps": 0.005051851272583008, "step": 342}
{"info/global_step": 343, "train_info/time_within_train_step": 27.707722663879395, "step": 343}
{"train_info/time_between_train_steps": 0.0050847530364990234, "step": 343}
{"info/global_step": 344, "train_info/time_within_train_step": 27.714982509613037, "step": 344}
{"train_info/time_between_train_steps": 0.005219936370849609, "step": 344}
{"info/global_step": 345, "train_info/time_within_train_step": 27.70806384086609, "step": 345}
{"train_info/time_between_train_steps": 0.009747028350830078, "step": 345}
{"info/global_step": 346, "train_info/time_within_train_step": 27.7022123336792, "step": 346}
{"train_info/time_between_train_steps": 0.0049762725830078125, "step": 346}
{"info/global_step": 347, "train_info/time_within_train_step": 27.702218055725098, "step": 347}
{"train_info/time_between_train_steps": 0.005182504653930664, "step": 347}
{"info/global_step": 348, "train_info/time_within_train_step": 27.722214698791504, "step": 348}
{"train_info/time_between_train_steps": 0.010913610458374023, "step": 348}
{"info/global_step": 349, "train_info/time_within_train_step": 27.734841346740723, "step": 349}
{"train_info/time_between_train_steps": 0.0052149295806884766, "step": 349}
{"info/global_step": 350, "train_info/time_within_train_step": 27.716193437576294, "step": 350}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733739169, "_runtime": 11259}, "step": 350}
{"logs": {"train/loss": 6.0627, "train/learning_rate": 0.00047222222222222224, "train/epoch": 12.02, "_timestamp": 1733739169, "_runtime": 11259}, "step": 350}
{"train_info/time_between_train_steps": 0.0070459842681884766, "step": 350}
{"info/global_step": 351, "train_info/time_within_train_step": 27.735996961593628, "step": 351}
{"train_info/time_between_train_steps": 0.0057621002197265625, "step": 351}
{"train_info/time_between_train_steps": 20.672895908355713, "step": 351}
{"info/global_step": 352, "train_info/time_within_train_step": 27.80048656463623, "step": 352}
{"train_info/time_between_train_steps": 0.004905223846435547, "step": 352}
{"info/global_step": 353, "train_info/time_within_train_step": 27.90076184272766, "step": 353}
{"train_info/time_between_train_steps": 0.005190372467041016, "step": 353}
{"info/global_step": 354, "train_info/time_within_train_step": 27.728652000427246, "step": 354}
{"train_info/time_between_train_steps": 0.005037069320678711, "step": 354}
{"info/global_step": 355, "train_info/time_within_train_step": 27.848462343215942, "step": 355}
{"train_info/time_between_train_steps": 0.005212306976318359, "step": 355}
{"info/global_step": 356, "train_info/time_within_train_step": 27.713457107543945, "step": 356}
{"train_info/time_between_train_steps": 0.005370378494262695, "step": 356}
{"info/global_step": 357, "train_info/time_within_train_step": 27.822561502456665, "step": 357}
{"train_info/time_between_train_steps": 0.005113363265991211, "step": 357}
{"info/global_step": 358, "train_info/time_within_train_step": 27.716047763824463, "step": 358}
{"train_info/time_between_train_steps": 0.005155801773071289, "step": 358}
{"info/global_step": 359, "train_info/time_within_train_step": 27.828569412231445, "step": 359}
{"train_info/time_between_train_steps": 0.005539655685424805, "step": 359}
{"info/global_step": 360, "train_info/time_within_train_step": 27.726001262664795, "step": 360}
{"train_info/time_between_train_steps": 0.00511932373046875, "step": 360}
{"info/global_step": 361, "train_info/time_within_train_step": 27.703295707702637, "step": 361}
{"train_info/time_between_train_steps": 0.005077838897705078, "step": 361}
{"info/global_step": 362, "train_info/time_within_train_step": 27.708306550979614, "step": 362}
{"train_info/time_between_train_steps": 0.005114555358886719, "step": 362}
{"info/global_step": 363, "train_info/time_within_train_step": 27.710381269454956, "step": 363}
{"train_info/time_between_train_steps": 0.0053861141204833984, "step": 363}
{"info/global_step": 364, "train_info/time_within_train_step": 27.713927507400513, "step": 364}
{"train_info/time_between_train_steps": 0.005120515823364258, "step": 364}
{"info/global_step": 365, "train_info/time_within_train_step": 27.731797218322754, "step": 365}
{"train_info/time_between_train_steps": 0.00531315803527832, "step": 365}
{"info/global_step": 366, "train_info/time_within_train_step": 27.731630325317383, "step": 366}
{"train_info/time_between_train_steps": 0.005357980728149414, "step": 366}
{"info/global_step": 367, "train_info/time_within_train_step": 27.819567918777466, "step": 367}
{"train_info/time_between_train_steps": 0.005049228668212891, "step": 367}
{"info/global_step": 368, "train_info/time_within_train_step": 27.73243498802185, "step": 368}
{"train_info/time_between_train_steps": 0.0052642822265625, "step": 368}
{"info/global_step": 369, "train_info/time_within_train_step": 27.731257438659668, "step": 369}
{"train_info/time_between_train_steps": 0.005258798599243164, "step": 369}
{"info/global_step": 370, "train_info/time_within_train_step": 27.726239919662476, "step": 370}
{"train_info/time_between_train_steps": 0.0053768157958984375, "step": 370}
{"info/global_step": 371, "train_info/time_within_train_step": 27.7173912525177, "step": 371}
{"train_info/time_between_train_steps": 0.005208492279052734, "step": 371}
{"info/global_step": 372, "train_info/time_within_train_step": 27.709816217422485, "step": 372}
{"train_info/time_between_train_steps": 0.005143404006958008, "step": 372}
{"info/global_step": 373, "train_info/time_within_train_step": 27.70932698249817, "step": 373}
{"train_info/time_between_train_steps": 0.005320072174072266, "step": 373}
{"info/global_step": 374, "train_info/time_within_train_step": 27.728046894073486, "step": 374}
{"train_info/time_between_train_steps": 0.004962444305419922, "step": 374}
{"info/global_step": 375, "train_info/time_within_train_step": 27.70909357070923, "step": 375}
{"train_info/time_between_train_steps": 0.005471706390380859, "step": 375}
{"info/global_step": 376, "train_info/time_within_train_step": 27.731749534606934, "step": 376}
{"train_info/time_between_train_steps": 0.005420684814453125, "step": 376}
{"info/global_step": 377, "train_info/time_within_train_step": 27.730865240097046, "step": 377}
{"train_info/time_between_train_steps": 0.005728483200073242, "step": 377}
{"info/global_step": 378, "train_info/time_within_train_step": 27.739713191986084, "step": 378}
{"train_info/time_between_train_steps": 0.005807638168334961, "step": 378}
{"train_info/time_between_train_steps": 20.89559006690979, "step": 378}
{"info/global_step": 379, "train_info/time_within_train_step": 27.702512979507446, "step": 379}
{"train_info/time_between_train_steps": 0.005133390426635742, "step": 379}
{"info/global_step": 380, "train_info/time_within_train_step": 27.853283882141113, "step": 380}
{"train_info/time_between_train_steps": 0.005710124969482422, "step": 380}
{"info/global_step": 381, "train_info/time_within_train_step": 27.72013831138611, "step": 381}
{"train_info/time_between_train_steps": 0.005322933197021484, "step": 381}
{"info/global_step": 382, "train_info/time_within_train_step": 27.93631625175476, "step": 382}
{"train_info/time_between_train_steps": 0.00554347038269043, "step": 382}
{"info/global_step": 383, "train_info/time_within_train_step": 27.717047929763794, "step": 383}
{"train_info/time_between_train_steps": 0.005241870880126953, "step": 383}
{"info/global_step": 384, "train_info/time_within_train_step": 27.84017539024353, "step": 384}
{"train_info/time_between_train_steps": 0.005233287811279297, "step": 384}
{"info/global_step": 385, "train_info/time_within_train_step": 27.720888137817383, "step": 385}
{"train_info/time_between_train_steps": 0.005511045455932617, "step": 385}
{"info/global_step": 386, "train_info/time_within_train_step": 27.795899629592896, "step": 386}
{"train_info/time_between_train_steps": 0.0053942203521728516, "step": 386}
{"info/global_step": 387, "train_info/time_within_train_step": 27.709355115890503, "step": 387}
{"train_info/time_between_train_steps": 0.004968166351318359, "step": 387}
{"info/global_step": 388, "train_info/time_within_train_step": 27.711721658706665, "step": 388}
{"train_info/time_between_train_steps": 0.0053102970123291016, "step": 388}
{"info/global_step": 389, "train_info/time_within_train_step": 27.746033906936646, "step": 389}
{"train_info/time_between_train_steps": 0.0052492618560791016, "step": 389}
{"info/global_step": 390, "train_info/time_within_train_step": 27.709453344345093, "step": 390}
{"train_info/time_between_train_steps": 0.005071401596069336, "step": 390}
{"info/global_step": 391, "train_info/time_within_train_step": 27.71274495124817, "step": 391}
{"train_info/time_between_train_steps": 0.005170106887817383, "step": 391}
{"info/global_step": 392, "train_info/time_within_train_step": 27.7165846824646, "step": 392}
{"train_info/time_between_train_steps": 0.005210399627685547, "step": 392}
{"info/global_step": 393, "train_info/time_within_train_step": 27.710926294326782, "step": 393}
{"train_info/time_between_train_steps": 0.005129575729370117, "step": 393}
{"info/global_step": 394, "train_info/time_within_train_step": 27.70778226852417, "step": 394}
{"train_info/time_between_train_steps": 0.00524449348449707, "step": 394}
{"info/global_step": 395, "train_info/time_within_train_step": 27.716696739196777, "step": 395}
{"train_info/time_between_train_steps": 0.0053043365478515625, "step": 395}
{"info/global_step": 396, "train_info/time_within_train_step": 27.70863127708435, "step": 396}
{"train_info/time_between_train_steps": 0.005094289779663086, "step": 396}
{"info/global_step": 397, "train_info/time_within_train_step": 27.709385871887207, "step": 397}
{"train_info/time_between_train_steps": 0.005167722702026367, "step": 397}
{"info/global_step": 398, "train_info/time_within_train_step": 27.819565057754517, "step": 398}
{"train_info/time_between_train_steps": 0.00518488883972168, "step": 398}
{"info/global_step": 399, "train_info/time_within_train_step": 27.71512746810913, "step": 399}
{"train_info/time_between_train_steps": 0.005328655242919922, "step": 399}
{"info/global_step": 400, "train_info/time_within_train_step": 27.716816663742065, "step": 400}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733740598, "_runtime": 12688}, "step": 400}
{"logs": {"train/loss": 5.9225, "train/learning_rate": 0.00044444444444444436, "train/epoch": 14.02, "_timestamp": 1733740598, "_runtime": 12688}, "step": 400}
{"train_info/time_between_train_steps": 2.3425838947296143, "step": 400}
{"info/global_step": 401, "train_info/time_within_train_step": 27.748517513275146, "step": 401}
{"train_info/time_between_train_steps": 0.005206584930419922, "step": 401}
{"info/global_step": 402, "train_info/time_within_train_step": 27.73351788520813, "step": 402}
{"train_info/time_between_train_steps": 0.005624532699584961, "step": 402}
{"info/global_step": 403, "train_info/time_within_train_step": 27.736296892166138, "step": 403}
{"train_info/time_between_train_steps": 0.005444526672363281, "step": 403}
{"info/global_step": 404, "train_info/time_within_train_step": 27.73905372619629, "step": 404}
{"train_info/time_between_train_steps": 0.005486726760864258, "step": 404}
{"info/global_step": 405, "train_info/time_within_train_step": 27.748311281204224, "step": 405}
{"train_info/time_between_train_steps": 0.005709171295166016, "step": 405}
{"train_info/time_between_train_steps": 20.54804563522339, "step": 405}
{"info/global_step": 406, "train_info/time_within_train_step": 27.707446098327637, "step": 406}
{"train_info/time_between_train_steps": 0.005021810531616211, "step": 406}
{"info/global_step": 407, "train_info/time_within_train_step": 27.83842658996582, "step": 407}
{"train_info/time_between_train_steps": 0.004973649978637695, "step": 407}
{"info/global_step": 408, "train_info/time_within_train_step": 27.710415840148926, "step": 408}
{"train_info/time_between_train_steps": 0.005414009094238281, "step": 408}
{"info/global_step": 409, "train_info/time_within_train_step": 27.851643562316895, "step": 409}
{"train_info/time_between_train_steps": 0.00533747673034668, "step": 409}
{"info/global_step": 410, "train_info/time_within_train_step": 27.755775690078735, "step": 410}
{"train_info/time_between_train_steps": 0.005555868148803711, "step": 410}
{"info/global_step": 411, "train_info/time_within_train_step": 27.98302388191223, "step": 411}
{"train_info/time_between_train_steps": 0.00580286979675293, "step": 411}
{"info/global_step": 412, "train_info/time_within_train_step": 27.80103302001953, "step": 412}
{"train_info/time_between_train_steps": 0.006266593933105469, "step": 412}
{"info/global_step": 413, "train_info/time_within_train_step": 27.94971466064453, "step": 413}
{"train_info/time_between_train_steps": 0.005638599395751953, "step": 413}
{"info/global_step": 414, "train_info/time_within_train_step": 27.784892320632935, "step": 414}
{"train_info/time_between_train_steps": 0.005320549011230469, "step": 414}
{"info/global_step": 415, "train_info/time_within_train_step": 27.70481824874878, "step": 415}
{"train_info/time_between_train_steps": 0.005147695541381836, "step": 415}
{"info/global_step": 416, "train_info/time_within_train_step": 27.706912755966187, "step": 416}
{"train_info/time_between_train_steps": 0.005299568176269531, "step": 416}
{"info/global_step": 417, "train_info/time_within_train_step": 27.720171689987183, "step": 417}
{"train_info/time_between_train_steps": 0.005167245864868164, "step": 417}
{"info/global_step": 418, "train_info/time_within_train_step": 27.713656187057495, "step": 418}
{"train_info/time_between_train_steps": 0.005190849304199219, "step": 418}
{"info/global_step": 419, "train_info/time_within_train_step": 27.718324661254883, "step": 419}
{"train_info/time_between_train_steps": 0.005313873291015625, "step": 419}
{"info/global_step": 420, "train_info/time_within_train_step": 27.73099184036255, "step": 420}
{"train_info/time_between_train_steps": 0.0053806304931640625, "step": 420}
{"info/global_step": 421, "train_info/time_within_train_step": 27.71865177154541, "step": 421}
{"train_info/time_between_train_steps": 0.005211353302001953, "step": 421}
{"info/global_step": 422, "train_info/time_within_train_step": 27.737178087234497, "step": 422}
{"train_info/time_between_train_steps": 0.005346536636352539, "step": 422}
{"info/global_step": 423, "train_info/time_within_train_step": 27.712944984436035, "step": 423}
{"train_info/time_between_train_steps": 0.0052988529205322266, "step": 423}
{"info/global_step": 424, "train_info/time_within_train_step": 27.728582859039307, "step": 424}
{"train_info/time_between_train_steps": 0.005321979522705078, "step": 424}
{"info/global_step": 425, "train_info/time_within_train_step": 27.792518138885498, "step": 425}
{"train_info/time_between_train_steps": 0.005298137664794922, "step": 425}
{"info/global_step": 426, "train_info/time_within_train_step": 27.71582317352295, "step": 426}
{"train_info/time_between_train_steps": 0.00552678108215332, "step": 426}
{"info/global_step": 427, "train_info/time_within_train_step": 27.72941827774048, "step": 427}
{"train_info/time_between_train_steps": 0.005261898040771484, "step": 427}
{"info/global_step": 428, "train_info/time_within_train_step": 27.71209144592285, "step": 428}
{"train_info/time_between_train_steps": 0.005296468734741211, "step": 428}
{"info/global_step": 429, "train_info/time_within_train_step": 27.807803869247437, "step": 429}
{"train_info/time_between_train_steps": 0.005538463592529297, "step": 429}
{"info/global_step": 430, "train_info/time_within_train_step": 27.71668815612793, "step": 430}
{"train_info/time_between_train_steps": 0.005404233932495117, "step": 430}
{"info/global_step": 431, "train_info/time_within_train_step": 27.730345249176025, "step": 431}
{"train_info/time_between_train_steps": 0.005890607833862305, "step": 431}
{"info/global_step": 432, "train_info/time_within_train_step": 27.736531734466553, "step": 432}
{"train_info/time_between_train_steps": 0.005769252777099609, "step": 432}
{"train_info/time_between_train_steps": 20.662100076675415, "step": 432}
{"info/global_step": 433, "train_info/time_within_train_step": 27.698712587356567, "step": 433}
{"train_info/time_between_train_steps": 0.005647182464599609, "step": 433}
{"info/global_step": 434, "train_info/time_within_train_step": 27.86001944541931, "step": 434}
{"train_info/time_between_train_steps": 0.005408525466918945, "step": 434}
{"info/global_step": 435, "train_info/time_within_train_step": 27.715007066726685, "step": 435}
{"train_info/time_between_train_steps": 0.005178689956665039, "step": 435}
{"info/global_step": 436, "train_info/time_within_train_step": 27.853134632110596, "step": 436}
{"train_info/time_between_train_steps": 0.005448341369628906, "step": 436}
{"info/global_step": 437, "train_info/time_within_train_step": 27.71509099006653, "step": 437}
{"train_info/time_between_train_steps": 0.010447025299072266, "step": 437}
{"info/global_step": 438, "train_info/time_within_train_step": 27.86115550994873, "step": 438}
{"train_info/time_between_train_steps": 0.005326747894287109, "step": 438}
{"info/global_step": 439, "train_info/time_within_train_step": 27.71311330795288, "step": 439}
{"train_info/time_between_train_steps": 0.005457162857055664, "step": 439}
{"info/global_step": 440, "train_info/time_within_train_step": 27.769365787506104, "step": 440}
{"train_info/time_between_train_steps": 0.00528717041015625, "step": 440}
{"info/global_step": 441, "train_info/time_within_train_step": 27.703694581985474, "step": 441}
{"train_info/time_between_train_steps": 0.005103588104248047, "step": 441}
{"info/global_step": 442, "train_info/time_within_train_step": 27.69909405708313, "step": 442}
{"train_info/time_between_train_steps": 0.0051686763763427734, "step": 442}
{"info/global_step": 443, "train_info/time_within_train_step": 27.69779086112976, "step": 443}
{"train_info/time_between_train_steps": 0.005045652389526367, "step": 443}
{"info/global_step": 444, "train_info/time_within_train_step": 27.800395727157593, "step": 444}
{"train_info/time_between_train_steps": 0.005181312561035156, "step": 444}
{"info/global_step": 445, "train_info/time_within_train_step": 27.69755506515503, "step": 445}
{"train_info/time_between_train_steps": 0.0050470829010009766, "step": 445}
{"info/global_step": 446, "train_info/time_within_train_step": 27.93005895614624, "step": 446}
{"train_info/time_between_train_steps": 0.009610414505004883, "step": 446}
{"info/global_step": 447, "train_info/time_within_train_step": 27.712183713912964, "step": 447}
{"train_info/time_between_train_steps": 0.005285978317260742, "step": 447}
{"info/global_step": 448, "train_info/time_within_train_step": 27.71355700492859, "step": 448}
{"train_info/time_between_train_steps": 0.005053997039794922, "step": 448}
{"info/global_step": 449, "train_info/time_within_train_step": 27.701140642166138, "step": 449}
{"train_info/time_between_train_steps": 0.005071878433227539, "step": 449}
{"info/global_step": 450, "train_info/time_within_train_step": 27.703421592712402, "step": 450}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733742030, "_runtime": 14120}, "step": 450}
{"logs": {"train/loss": 5.7523, "train/learning_rate": 0.00041666666666666664, "train/epoch": 16.02, "_timestamp": 1733742030, "_runtime": 14120}, "step": 450}
{"train_info/time_between_train_steps": 0.007102251052856445, "step": 450}
{"info/global_step": 451, "train_info/time_within_train_step": 27.700176000595093, "step": 451}
{"train_info/time_between_train_steps": 0.005112886428833008, "step": 451}
{"info/global_step": 452, "train_info/time_within_train_step": 27.696335315704346, "step": 452}
{"train_info/time_between_train_steps": 0.0050160884857177734, "step": 452}
{"info/global_step": 453, "train_info/time_within_train_step": 27.704971313476562, "step": 453}
{"train_info/time_between_train_steps": 0.005339384078979492, "step": 453}
{"info/global_step": 454, "train_info/time_within_train_step": 27.701698064804077, "step": 454}
{"train_info/time_between_train_steps": 0.005234956741333008, "step": 454}
{"info/global_step": 455, "train_info/time_within_train_step": 27.705267190933228, "step": 455}
{"train_info/time_between_train_steps": 0.005002498626708984, "step": 455}
{"info/global_step": 456, "train_info/time_within_train_step": 27.732173204421997, "step": 456}
{"train_info/time_between_train_steps": 0.005354642868041992, "step": 456}
{"info/global_step": 457, "train_info/time_within_train_step": 27.743441343307495, "step": 457}
{"train_info/time_between_train_steps": 0.010037660598754883, "step": 457}
{"info/global_step": 458, "train_info/time_within_train_step": 27.732728242874146, "step": 458}
{"train_info/time_between_train_steps": 0.011459112167358398, "step": 458}
{"info/global_step": 459, "train_info/time_within_train_step": 27.852314949035645, "step": 459}
{"train_info/time_between_train_steps": 0.005675792694091797, "step": 459}
{"train_info/time_between_train_steps": 20.79153800010681, "step": 459}
{"info/global_step": 460, "train_info/time_within_train_step": 27.70421814918518, "step": 460}
{"train_info/time_between_train_steps": 0.00490117073059082, "step": 460}
{"info/global_step": 461, "train_info/time_within_train_step": 27.888232946395874, "step": 461}
{"train_info/time_between_train_steps": 0.005513191223144531, "step": 461}
{"info/global_step": 462, "train_info/time_within_train_step": 27.755210161209106, "step": 462}
{"train_info/time_between_train_steps": 0.005387306213378906, "step": 462}
{"info/global_step": 463, "train_info/time_within_train_step": 27.916567087173462, "step": 463}
{"train_info/time_between_train_steps": 0.005487680435180664, "step": 463}
{"info/global_step": 464, "train_info/time_within_train_step": 27.728205919265747, "step": 464}
{"train_info/time_between_train_steps": 0.00541996955871582, "step": 464}
{"info/global_step": 465, "train_info/time_within_train_step": 27.932093381881714, "step": 465}
{"train_info/time_between_train_steps": 0.0054666996002197266, "step": 465}
{"info/global_step": 466, "train_info/time_within_train_step": 27.850825309753418, "step": 466}
{"train_info/time_between_train_steps": 0.006041288375854492, "step": 466}
{"info/global_step": 467, "train_info/time_within_train_step": 27.84208035469055, "step": 467}
{"train_info/time_between_train_steps": 0.005485057830810547, "step": 467}
{"info/global_step": 468, "train_info/time_within_train_step": 27.72421622276306, "step": 468}
{"train_info/time_between_train_steps": 0.005306243896484375, "step": 468}
{"info/global_step": 469, "train_info/time_within_train_step": 27.725179433822632, "step": 469}
{"train_info/time_between_train_steps": 0.00531458854675293, "step": 469}
{"info/global_step": 470, "train_info/time_within_train_step": 27.725547790527344, "step": 470}
{"train_info/time_between_train_steps": 0.005190849304199219, "step": 470}
{"info/global_step": 471, "train_info/time_within_train_step": 27.713530778884888, "step": 471}
{"train_info/time_between_train_steps": 0.005124330520629883, "step": 471}
{"info/global_step": 472, "train_info/time_within_train_step": 27.722431898117065, "step": 472}
{"train_info/time_between_train_steps": 0.011475086212158203, "step": 472}
{"info/global_step": 473, "train_info/time_within_train_step": 27.71898365020752, "step": 473}
{"train_info/time_between_train_steps": 0.00542140007019043, "step": 473}
{"info/global_step": 474, "train_info/time_within_train_step": 27.859305381774902, "step": 474}
{"train_info/time_between_train_steps": 0.005273103713989258, "step": 474}
{"info/global_step": 475, "train_info/time_within_train_step": 27.71573543548584, "step": 475}
{"train_info/time_between_train_steps": 0.00527191162109375, "step": 475}
{"info/global_step": 476, "train_info/time_within_train_step": 27.715715885162354, "step": 476}
{"train_info/time_between_train_steps": 0.005150794982910156, "step": 476}
{"info/global_step": 477, "train_info/time_within_train_step": 27.718459606170654, "step": 477}
{"train_info/time_between_train_steps": 0.005297660827636719, "step": 477}
{"info/global_step": 478, "train_info/time_within_train_step": 27.71787714958191, "step": 478}
{"train_info/time_between_train_steps": 0.0051000118255615234, "step": 478}
{"info/global_step": 479, "train_info/time_within_train_step": 27.722283124923706, "step": 479}
{"train_info/time_between_train_steps": 0.005255222320556641, "step": 479}
{"info/global_step": 480, "train_info/time_within_train_step": 27.719072580337524, "step": 480}
{"train_info/time_between_train_steps": 0.005140066146850586, "step": 480}
{"info/global_step": 481, "train_info/time_within_train_step": 27.712521076202393, "step": 481}
{"train_info/time_between_train_steps": 0.005403041839599609, "step": 481}
{"info/global_step": 482, "train_info/time_within_train_step": 27.728392124176025, "step": 482}
{"train_info/time_between_train_steps": 0.005295753479003906, "step": 482}
{"info/global_step": 483, "train_info/time_within_train_step": 27.7807457447052, "step": 483}
{"train_info/time_between_train_steps": 0.006061077117919922, "step": 483}
{"info/global_step": 484, "train_info/time_within_train_step": 27.794713497161865, "step": 484}
{"train_info/time_between_train_steps": 0.005728244781494141, "step": 484}
{"info/global_step": 485, "train_info/time_within_train_step": 27.747291326522827, "step": 485}
{"train_info/time_between_train_steps": 0.005705356597900391, "step": 485}
{"info/global_step": 486, "train_info/time_within_train_step": 27.75239372253418, "step": 486}
{"train_info/time_between_train_steps": 0.0058023929595947266, "step": 486}
{"train_info/time_between_train_steps": 20.4704008102417, "step": 486}
{"info/global_step": 487, "train_info/time_within_train_step": 27.71794557571411, "step": 487}
{"train_info/time_between_train_steps": 0.0056879520416259766, "step": 487}
{"info/global_step": 488, "train_info/time_within_train_step": 27.879119634628296, "step": 488}
{"train_info/time_between_train_steps": 0.0055217742919921875, "step": 488}
{"info/global_step": 489, "train_info/time_within_train_step": 27.838226795196533, "step": 489}
{"train_info/time_between_train_steps": 0.005625486373901367, "step": 489}
{"info/global_step": 490, "train_info/time_within_train_step": 27.885476112365723, "step": 490}
{"train_info/time_between_train_steps": 0.005483865737915039, "step": 490}
{"info/global_step": 491, "train_info/time_within_train_step": 27.73456883430481, "step": 491}
{"train_info/time_between_train_steps": 0.005675077438354492, "step": 491}
{"info/global_step": 492, "train_info/time_within_train_step": 27.9324631690979, "step": 492}
{"train_info/time_between_train_steps": 0.005396366119384766, "step": 492}
{"info/global_step": 493, "train_info/time_within_train_step": 27.721055269241333, "step": 493}
{"train_info/time_between_train_steps": 0.01053619384765625, "step": 493}
{"info/global_step": 494, "train_info/time_within_train_step": 27.78808832168579, "step": 494}
{"train_info/time_between_train_steps": 0.005347013473510742, "step": 494}
{"info/global_step": 495, "train_info/time_within_train_step": 27.708015203475952, "step": 495}
{"train_info/time_between_train_steps": 0.0050351619720458984, "step": 495}
{"info/global_step": 496, "train_info/time_within_train_step": 27.704185009002686, "step": 496}
{"train_info/time_between_train_steps": 0.005154609680175781, "step": 496}
{"info/global_step": 497, "train_info/time_within_train_step": 27.701451063156128, "step": 497}
{"train_info/time_between_train_steps": 0.005515098571777344, "step": 497}
{"info/global_step": 498, "train_info/time_within_train_step": 27.700594902038574, "step": 498}
{"train_info/time_between_train_steps": 0.005179405212402344, "step": 498}
{"info/global_step": 499, "train_info/time_within_train_step": 27.702489614486694, "step": 499}
{"train_info/time_between_train_steps": 0.0050563812255859375, "step": 499}
{"info/global_step": 500, "train_info/time_within_train_step": 27.7050998210907, "step": 500}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733743459, "_runtime": 15549}, "step": 500}
{"logs": {"train/loss": 5.6018, "train/learning_rate": 0.00038888888888888887, "train/epoch": 18.01, "_timestamp": 1733743459, "_runtime": 15549}, "step": 500}
{"train_info/time_between_train_steps": 2.4564011096954346, "step": 500}
{"info/global_step": 501, "train_info/time_within_train_step": 27.798463106155396, "step": 501}
{"train_info/time_between_train_steps": 0.00516200065612793, "step": 501}
{"info/global_step": 502, "train_info/time_within_train_step": 27.7069354057312, "step": 502}
{"train_info/time_between_train_steps": 0.005138874053955078, "step": 502}
{"info/global_step": 503, "train_info/time_within_train_step": 27.716140270233154, "step": 503}
{"train_info/time_between_train_steps": 0.005346775054931641, "step": 503}
{"info/global_step": 504, "train_info/time_within_train_step": 27.715003490447998, "step": 504}
{"train_info/time_between_train_steps": 0.00525975227355957, "step": 504}
{"info/global_step": 505, "train_info/time_within_train_step": 27.8082594871521, "step": 505}
{"train_info/time_between_train_steps": 0.005167245864868164, "step": 505}
{"info/global_step": 506, "train_info/time_within_train_step": 27.72254967689514, "step": 506}
{"train_info/time_between_train_steps": 0.0052373409271240234, "step": 506}
{"info/global_step": 507, "train_info/time_within_train_step": 27.755998134613037, "step": 507}
{"train_info/time_between_train_steps": 0.014172554016113281, "step": 507}
{"info/global_step": 508, "train_info/time_within_train_step": 27.80328607559204, "step": 508}
{"train_info/time_between_train_steps": 0.005537986755371094, "step": 508}
{"info/global_step": 509, "train_info/time_within_train_step": 27.73272967338562, "step": 509}
{"train_info/time_between_train_steps": 0.0054454803466796875, "step": 509}
{"info/global_step": 510, "train_info/time_within_train_step": 27.7487633228302, "step": 510}
{"train_info/time_between_train_steps": 0.005813121795654297, "step": 510}
{"info/global_step": 511, "train_info/time_within_train_step": 27.776302099227905, "step": 511}
{"train_info/time_between_train_steps": 0.005407810211181641, "step": 511}
{"info/global_step": 512, "train_info/time_within_train_step": 27.743057012557983, "step": 512}
{"train_info/time_between_train_steps": 0.005794525146484375, "step": 512}
{"info/global_step": 513, "train_info/time_within_train_step": 27.782750844955444, "step": 513}
{"train_info/time_between_train_steps": 0.005957603454589844, "step": 513}
{"train_info/time_between_train_steps": 20.633215188980103, "step": 513}
{"info/global_step": 514, "train_info/time_within_train_step": 27.72807478904724, "step": 514}
{"train_info/time_between_train_steps": 0.005647420883178711, "step": 514}
{"info/global_step": 515, "train_info/time_within_train_step": 27.920689344406128, "step": 515}
{"train_info/time_between_train_steps": 0.005562782287597656, "step": 515}
{"info/global_step": 516, "train_info/time_within_train_step": 27.767961978912354, "step": 516}
{"train_info/time_between_train_steps": 0.005240440368652344, "step": 516}
{"info/global_step": 517, "train_info/time_within_train_step": 27.921474933624268, "step": 517}
{"train_info/time_between_train_steps": 0.006395578384399414, "step": 517}
{"info/global_step": 518, "train_info/time_within_train_step": 27.740720987319946, "step": 518}
{"train_info/time_between_train_steps": 0.005425930023193359, "step": 518}
{"info/global_step": 519, "train_info/time_within_train_step": 27.900226354599, "step": 519}
{"train_info/time_between_train_steps": 0.005454063415527344, "step": 519}
{"info/global_step": 520, "train_info/time_within_train_step": 27.836931228637695, "step": 520}
{"train_info/time_between_train_steps": 0.005644559860229492, "step": 520}
{"info/global_step": 521, "train_info/time_within_train_step": 27.80302906036377, "step": 521}
{"train_info/time_between_train_steps": 0.0053386688232421875, "step": 521}
{"info/global_step": 522, "train_info/time_within_train_step": 27.7175931930542, "step": 522}
{"train_info/time_between_train_steps": 0.005173206329345703, "step": 522}
{"info/global_step": 523, "train_info/time_within_train_step": 27.759201526641846, "step": 523}
{"train_info/time_between_train_steps": 0.005149364471435547, "step": 523}
{"info/global_step": 524, "train_info/time_within_train_step": 27.728859186172485, "step": 524}
{"train_info/time_between_train_steps": 0.005212306976318359, "step": 524}
{"info/global_step": 525, "train_info/time_within_train_step": 27.705049514770508, "step": 525}
{"train_info/time_between_train_steps": 0.005047798156738281, "step": 525}
{"info/global_step": 526, "train_info/time_within_train_step": 27.72705602645874, "step": 526}
{"train_info/time_between_train_steps": 0.005125284194946289, "step": 526}
{"info/global_step": 527, "train_info/time_within_train_step": 27.70362401008606, "step": 527}
{"train_info/time_between_train_steps": 0.005137205123901367, "step": 527}
{"info/global_step": 528, "train_info/time_within_train_step": 27.71230125427246, "step": 528}
{"train_info/time_between_train_steps": 0.005290031433105469, "step": 528}
{"info/global_step": 529, "train_info/time_within_train_step": 27.71342372894287, "step": 529}
{"train_info/time_between_train_steps": 0.00507807731628418, "step": 529}
{"info/global_step": 530, "train_info/time_within_train_step": 27.709923028945923, "step": 530}
{"train_info/time_between_train_steps": 0.0052661895751953125, "step": 530}
{"info/global_step": 531, "train_info/time_within_train_step": 27.696926593780518, "step": 531}
{"train_info/time_between_train_steps": 0.005142688751220703, "step": 531}
{"info/global_step": 532, "train_info/time_within_train_step": 27.704379081726074, "step": 532}
{"train_info/time_between_train_steps": 0.005259990692138672, "step": 532}
{"info/global_step": 533, "train_info/time_within_train_step": 27.69907808303833, "step": 533}
{"train_info/time_between_train_steps": 0.004967212677001953, "step": 533}
{"info/global_step": 534, "train_info/time_within_train_step": 27.702842712402344, "step": 534}
{"train_info/time_between_train_steps": 0.005276203155517578, "step": 534}
{"info/global_step": 535, "train_info/time_within_train_step": 27.80285930633545, "step": 535}
{"train_info/time_between_train_steps": 0.005124568939208984, "step": 535}
{"info/global_step": 536, "train_info/time_within_train_step": 27.703996419906616, "step": 536}
{"train_info/time_between_train_steps": 0.005034208297729492, "step": 536}
{"info/global_step": 537, "train_info/time_within_train_step": 27.709589958190918, "step": 537}
{"train_info/time_between_train_steps": 0.005351543426513672, "step": 537}
{"info/global_step": 538, "train_info/time_within_train_step": 27.71022081375122, "step": 538}
{"train_info/time_between_train_steps": 0.005042076110839844, "step": 538}
{"info/global_step": 539, "train_info/time_within_train_step": 27.71142888069153, "step": 539}
{"train_info/time_between_train_steps": 0.005312919616699219, "step": 539}
{"info/global_step": 540, "train_info/time_within_train_step": 27.731322765350342, "step": 540}
{"train_info/time_between_train_steps": 0.005846500396728516, "step": 540}
{"train_info/time_between_train_steps": 20.675651788711548, "step": 540}
{"info/global_step": 541, "train_info/time_within_train_step": 27.69237971305847, "step": 541}
{"train_info/time_between_train_steps": 0.0048770904541015625, "step": 541}
{"info/global_step": 542, "train_info/time_within_train_step": 27.83979892730713, "step": 542}
{"train_info/time_between_train_steps": 0.005072116851806641, "step": 542}
{"info/global_step": 543, "train_info/time_within_train_step": 27.71096134185791, "step": 543}
{"train_info/time_between_train_steps": 0.0050656795501708984, "step": 543}
{"info/global_step": 544, "train_info/time_within_train_step": 27.821919441223145, "step": 544}
{"train_info/time_between_train_steps": 0.005350351333618164, "step": 544}
{"info/global_step": 545, "train_info/time_within_train_step": 27.70271396636963, "step": 545}
{"train_info/time_between_train_steps": 0.00509953498840332, "step": 545}
{"info/global_step": 546, "train_info/time_within_train_step": 27.82548975944519, "step": 546}
{"train_info/time_between_train_steps": 0.005022525787353516, "step": 546}
{"info/global_step": 547, "train_info/time_within_train_step": 27.70707058906555, "step": 547}
{"train_info/time_between_train_steps": 0.0054357051849365234, "step": 547}
{"info/global_step": 548, "train_info/time_within_train_step": 27.76925802230835, "step": 548}
{"train_info/time_between_train_steps": 0.005295276641845703, "step": 548}
{"info/global_step": 549, "train_info/time_within_train_step": 27.70400094985962, "step": 549}
{"train_info/time_between_train_steps": 0.004941463470458984, "step": 549}
{"info/global_step": 550, "train_info/time_within_train_step": 27.796264171600342, "step": 550}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733744891, "_runtime": 16981}, "step": 550}
{"logs": {"train/loss": 5.4548, "train/learning_rate": 0.0003611111111111111, "train/epoch": 20.01, "_timestamp": 1733744891, "_runtime": 16981}, "step": 550}
{"train_info/time_between_train_steps": 0.00683903694152832, "step": 550}
{"info/global_step": 551, "train_info/time_within_train_step": 27.69482445716858, "step": 551}
{"train_info/time_between_train_steps": 0.005122661590576172, "step": 551}
{"info/global_step": 552, "train_info/time_within_train_step": 27.69240140914917, "step": 552}
{"train_info/time_between_train_steps": 0.0050160884857177734, "step": 552}
{"info/global_step": 553, "train_info/time_within_train_step": 27.69811773300171, "step": 553}
{"train_info/time_between_train_steps": 0.005159616470336914, "step": 553}
{"info/global_step": 554, "train_info/time_within_train_step": 27.697506427764893, "step": 554}
{"train_info/time_between_train_steps": 0.005103349685668945, "step": 554}
{"info/global_step": 555, "train_info/time_within_train_step": 27.73779845237732, "step": 555}
{"train_info/time_between_train_steps": 0.005062580108642578, "step": 555}
{"info/global_step": 556, "train_info/time_within_train_step": 27.71693205833435, "step": 556}
{"train_info/time_between_train_steps": 0.005169391632080078, "step": 556}
{"info/global_step": 557, "train_info/time_within_train_step": 27.704394817352295, "step": 557}
{"train_info/time_between_train_steps": 0.0052585601806640625, "step": 557}
{"info/global_step": 558, "train_info/time_within_train_step": 27.701675415039062, "step": 558}
{"train_info/time_between_train_steps": 0.005133390426635742, "step": 558}
{"info/global_step": 559, "train_info/time_within_train_step": 27.704636812210083, "step": 559}
{"train_info/time_between_train_steps": 0.005179882049560547, "step": 559}
{"info/global_step": 560, "train_info/time_within_train_step": 27.703506469726562, "step": 560}
{"train_info/time_between_train_steps": 0.0051021575927734375, "step": 560}
{"info/global_step": 561, "train_info/time_within_train_step": 27.71056056022644, "step": 561}
{"train_info/time_between_train_steps": 0.005138397216796875, "step": 561}
{"info/global_step": 562, "train_info/time_within_train_step": 27.72364354133606, "step": 562}
{"train_info/time_between_train_steps": 0.005222320556640625, "step": 562}
{"info/global_step": 563, "train_info/time_within_train_step": 27.713943004608154, "step": 563}
{"train_info/time_between_train_steps": 0.005182504653930664, "step": 563}
{"info/global_step": 564, "train_info/time_within_train_step": 27.716552257537842, "step": 564}
{"train_info/time_between_train_steps": 0.005384683609008789, "step": 564}
{"info/global_step": 565, "train_info/time_within_train_step": 27.7224280834198, "step": 565}
{"train_info/time_between_train_steps": 0.0051898956298828125, "step": 565}
{"info/global_step": 566, "train_info/time_within_train_step": 27.82451629638672, "step": 566}
{"train_info/time_between_train_steps": 0.005671977996826172, "step": 566}
{"info/global_step": 567, "train_info/time_within_train_step": 27.750494718551636, "step": 567}
{"train_info/time_between_train_steps": 0.005863189697265625, "step": 567}
{"train_info/time_between_train_steps": 20.442688465118408, "step": 567}
{"info/global_step": 568, "train_info/time_within_train_step": 27.733638525009155, "step": 568}
{"train_info/time_between_train_steps": 0.005449056625366211, "step": 568}
{"info/global_step": 569, "train_info/time_within_train_step": 27.854480743408203, "step": 569}
{"train_info/time_between_train_steps": 0.00518345832824707, "step": 569}
{"info/global_step": 570, "train_info/time_within_train_step": 27.73526906967163, "step": 570}
{"train_info/time_between_train_steps": 0.005147695541381836, "step": 570}
{"info/global_step": 571, "train_info/time_within_train_step": 27.82264542579651, "step": 571}
{"train_info/time_between_train_steps": 0.005299091339111328, "step": 571}
{"info/global_step": 572, "train_info/time_within_train_step": 27.72020173072815, "step": 572}
{"train_info/time_between_train_steps": 0.005692720413208008, "step": 572}
{"info/global_step": 573, "train_info/time_within_train_step": 27.89142370223999, "step": 573}
{"train_info/time_between_train_steps": 0.005415678024291992, "step": 573}
{"info/global_step": 574, "train_info/time_within_train_step": 27.72462034225464, "step": 574}
{"train_info/time_between_train_steps": 0.005295276641845703, "step": 574}
{"info/global_step": 575, "train_info/time_within_train_step": 27.78157329559326, "step": 575}
{"train_info/time_between_train_steps": 0.005408525466918945, "step": 575}
{"info/global_step": 576, "train_info/time_within_train_step": 27.737837076187134, "step": 576}
{"train_info/time_between_train_steps": 0.006111621856689453, "step": 576}
{"info/global_step": 577, "train_info/time_within_train_step": 27.70204210281372, "step": 577}
{"train_info/time_between_train_steps": 0.005062103271484375, "step": 577}
{"info/global_step": 578, "train_info/time_within_train_step": 27.707814931869507, "step": 578}
{"train_info/time_between_train_steps": 0.005148887634277344, "step": 578}
{"info/global_step": 579, "train_info/time_within_train_step": 27.69997501373291, "step": 579}
{"train_info/time_between_train_steps": 0.00525665283203125, "step": 579}
{"info/global_step": 580, "train_info/time_within_train_step": 27.701221704483032, "step": 580}
{"train_info/time_between_train_steps": 0.005045413970947266, "step": 580}
{"info/global_step": 581, "train_info/time_within_train_step": 27.80289387702942, "step": 581}
{"train_info/time_between_train_steps": 0.005194664001464844, "step": 581}
{"info/global_step": 582, "train_info/time_within_train_step": 27.710023880004883, "step": 582}
{"train_info/time_between_train_steps": 0.005129337310791016, "step": 582}
{"info/global_step": 583, "train_info/time_within_train_step": 27.704015493392944, "step": 583}
{"train_info/time_between_train_steps": 0.00504612922668457, "step": 583}
{"info/global_step": 584, "train_info/time_within_train_step": 27.714833736419678, "step": 584}
{"train_info/time_between_train_steps": 0.0051496028900146484, "step": 584}
{"info/global_step": 585, "train_info/time_within_train_step": 27.71823263168335, "step": 585}
{"train_info/time_between_train_steps": 0.0051898956298828125, "step": 585}
{"info/global_step": 586, "train_info/time_within_train_step": 27.70519185066223, "step": 586}
{"train_info/time_between_train_steps": 0.00518798828125, "step": 586}
{"info/global_step": 587, "train_info/time_within_train_step": 27.7046320438385, "step": 587}
{"train_info/time_between_train_steps": 0.005168914794921875, "step": 587}
{"info/global_step": 588, "train_info/time_within_train_step": 27.71388864517212, "step": 588}
{"train_info/time_between_train_steps": 0.00534510612487793, "step": 588}
{"info/global_step": 589, "train_info/time_within_train_step": 27.72013568878174, "step": 589}
{"train_info/time_between_train_steps": 0.00510859489440918, "step": 589}
{"info/global_step": 590, "train_info/time_within_train_step": 27.709012985229492, "step": 590}
{"train_info/time_between_train_steps": 0.005206584930419922, "step": 590}
{"info/global_step": 591, "train_info/time_within_train_step": 27.70493245124817, "step": 591}
{"train_info/time_between_train_steps": 0.005637168884277344, "step": 591}
{"info/global_step": 592, "train_info/time_within_train_step": 27.734673261642456, "step": 592}
{"train_info/time_between_train_steps": 0.005349636077880859, "step": 592}
{"info/global_step": 593, "train_info/time_within_train_step": 27.726218700408936, "step": 593}
{"train_info/time_between_train_steps": 0.005453824996948242, "step": 593}
{"info/global_step": 594, "train_info/time_within_train_step": 27.726088523864746, "step": 594}
{"train_info/time_between_train_steps": 0.005694150924682617, "step": 594}
{"train_info/time_between_train_steps": 20.62260913848877, "step": 594}
{"info/global_step": 595, "train_info/time_within_train_step": 27.69162917137146, "step": 595}
{"train_info/time_between_train_steps": 0.004937648773193359, "step": 595}
{"info/global_step": 596, "train_info/time_within_train_step": 27.88989281654358, "step": 596}
{"train_info/time_between_train_steps": 0.004865884780883789, "step": 596}
{"info/global_step": 597, "train_info/time_within_train_step": 27.69988179206848, "step": 597}
{"train_info/time_between_train_steps": 0.00492095947265625, "step": 597}
{"info/global_step": 598, "train_info/time_within_train_step": 27.854976654052734, "step": 598}
{"train_info/time_between_train_steps": 0.00508427619934082, "step": 598}
{"info/global_step": 599, "train_info/time_within_train_step": 27.72650980949402, "step": 599}
{"train_info/time_between_train_steps": 0.005140781402587891, "step": 599}
{"info/global_step": 600, "train_info/time_within_train_step": 27.862191915512085, "step": 600}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733746319, "_runtime": 18409}, "step": 600}
{"logs": {"train/loss": 5.3217, "train/learning_rate": 0.0003333333333333333, "train/epoch": 22.0, "_timestamp": 1733746319, "_runtime": 18409}, "step": 600}
{"train_info/time_between_train_steps": 2.232936382293701, "step": 600}
{"info/global_step": 601, "train_info/time_within_train_step": 27.77298879623413, "step": 601}
{"train_info/time_between_train_steps": 0.0051746368408203125, "step": 601}
{"info/global_step": 602, "train_info/time_within_train_step": 27.755280017852783, "step": 602}
{"train_info/time_between_train_steps": 0.005356311798095703, "step": 602}
{"info/global_step": 603, "train_info/time_within_train_step": 27.747859954833984, "step": 603}
{"train_info/time_between_train_steps": 0.005194664001464844, "step": 603}
{"info/global_step": 604, "train_info/time_within_train_step": 27.694424390792847, "step": 604}
{"train_info/time_between_train_steps": 0.005030155181884766, "step": 604}
{"info/global_step": 605, "train_info/time_within_train_step": 27.712035417556763, "step": 605}
{"train_info/time_between_train_steps": 0.009885311126708984, "step": 605}
{"info/global_step": 606, "train_info/time_within_train_step": 27.71350121498108, "step": 606}
{"train_info/time_between_train_steps": 0.0051572322845458984, "step": 606}
{"info/global_step": 607, "train_info/time_within_train_step": 27.705052375793457, "step": 607}
{"train_info/time_between_train_steps": 0.005003690719604492, "step": 607}
{"info/global_step": 608, "train_info/time_within_train_step": 27.708004474639893, "step": 608}
{"train_info/time_between_train_steps": 0.0051577091217041016, "step": 608}
{"info/global_step": 609, "train_info/time_within_train_step": 27.716510772705078, "step": 609}
{"train_info/time_between_train_steps": 0.00520014762878418, "step": 609}
{"info/global_step": 610, "train_info/time_within_train_step": 27.70281744003296, "step": 610}
{"train_info/time_between_train_steps": 0.005151271820068359, "step": 610}
{"info/global_step": 611, "train_info/time_within_train_step": 27.7075457572937, "step": 611}
{"train_info/time_between_train_steps": 0.005189418792724609, "step": 611}
{"info/global_step": 612, "train_info/time_within_train_step": 27.8025643825531, "step": 612}
{"train_info/time_between_train_steps": 0.0051975250244140625, "step": 612}
{"info/global_step": 613, "train_info/time_within_train_step": 27.72065544128418, "step": 613}
{"train_info/time_between_train_steps": 0.0051152706146240234, "step": 613}
{"info/global_step": 614, "train_info/time_within_train_step": 27.70588493347168, "step": 614}
{"train_info/time_between_train_steps": 0.009940385818481445, "step": 614}
{"info/global_step": 615, "train_info/time_within_train_step": 27.71058988571167, "step": 615}
{"train_info/time_between_train_steps": 0.005430698394775391, "step": 615}
{"info/global_step": 616, "train_info/time_within_train_step": 27.708781003952026, "step": 616}
{"train_info/time_between_train_steps": 0.005288600921630859, "step": 616}
{"info/global_step": 617, "train_info/time_within_train_step": 27.705565690994263, "step": 617}
{"train_info/time_between_train_steps": 0.00511932373046875, "step": 617}
{"info/global_step": 618, "train_info/time_within_train_step": 27.712252855300903, "step": 618}
{"train_info/time_between_train_steps": 0.0055315494537353516, "step": 618}
{"info/global_step": 619, "train_info/time_within_train_step": 27.716534852981567, "step": 619}
{"train_info/time_between_train_steps": 0.00536036491394043, "step": 619}
{"info/global_step": 620, "train_info/time_within_train_step": 27.73241114616394, "step": 620}
{"train_info/time_between_train_steps": 0.0055980682373046875, "step": 620}
{"info/global_step": 621, "train_info/time_within_train_step": 27.74711537361145, "step": 621}
{"train_info/time_between_train_steps": 0.0055582523345947266, "step": 621}
{"train_info/time_between_train_steps": 20.70741105079651, "step": 621}
{"info/global_step": 622, "train_info/time_within_train_step": 27.698144674301147, "step": 622}
{"train_info/time_between_train_steps": 0.004864931106567383, "step": 622}
{"info/global_step": 623, "train_info/time_within_train_step": 27.822139263153076, "step": 623}
{"train_info/time_between_train_steps": 0.0050275325775146484, "step": 623}
{"info/global_step": 624, "train_info/time_within_train_step": 27.7480685710907, "step": 624}
{"train_info/time_between_train_steps": 0.005309343338012695, "step": 624}
{"info/global_step": 625, "train_info/time_within_train_step": 27.89151120185852, "step": 625}
{"train_info/time_between_train_steps": 0.005338907241821289, "step": 625}
{"info/global_step": 626, "train_info/time_within_train_step": 27.727776050567627, "step": 626}
{"train_info/time_between_train_steps": 0.00536036491394043, "step": 626}
{"info/global_step": 627, "train_info/time_within_train_step": 27.981335639953613, "step": 627}
{"train_info/time_between_train_steps": 0.005311489105224609, "step": 627}
{"info/global_step": 628, "train_info/time_within_train_step": 27.73680090904236, "step": 628}
{"train_info/time_between_train_steps": 0.005437612533569336, "step": 628}
{"info/global_step": 629, "train_info/time_within_train_step": 27.827748775482178, "step": 629}
{"train_info/time_between_train_steps": 0.0052967071533203125, "step": 629}
{"info/global_step": 630, "train_info/time_within_train_step": 27.74321222305298, "step": 630}
{"train_info/time_between_train_steps": 0.005252361297607422, "step": 630}
{"info/global_step": 631, "train_info/time_within_train_step": 27.752689599990845, "step": 631}
{"train_info/time_between_train_steps": 0.010286808013916016, "step": 631}
{"info/global_step": 632, "train_info/time_within_train_step": 27.733551263809204, "step": 632}
{"train_info/time_between_train_steps": 0.005273103713989258, "step": 632}
{"info/global_step": 633, "train_info/time_within_train_step": 27.738322973251343, "step": 633}
{"train_info/time_between_train_steps": 0.005201101303100586, "step": 633}
{"info/global_step": 634, "train_info/time_within_train_step": 27.73198699951172, "step": 634}
{"train_info/time_between_train_steps": 0.0051364898681640625, "step": 634}
{"info/global_step": 635, "train_info/time_within_train_step": 27.749095678329468, "step": 635}
{"train_info/time_between_train_steps": 0.005232572555541992, "step": 635}
{"info/global_step": 636, "train_info/time_within_train_step": 27.726736307144165, "step": 636}
{"train_info/time_between_train_steps": 0.005301237106323242, "step": 636}
{"info/global_step": 637, "train_info/time_within_train_step": 27.718153476715088, "step": 637}
{"train_info/time_between_train_steps": 0.005083560943603516, "step": 637}
{"info/global_step": 638, "train_info/time_within_train_step": 27.7117338180542, "step": 638}
{"train_info/time_between_train_steps": 0.005097389221191406, "step": 638}
{"info/global_step": 639, "train_info/time_within_train_step": 27.72206974029541, "step": 639}
{"train_info/time_between_train_steps": 0.009993553161621094, "step": 639}
{"info/global_step": 640, "train_info/time_within_train_step": 27.71079993247986, "step": 640}
{"train_info/time_between_train_steps": 0.01044607162475586, "step": 640}
{"info/global_step": 641, "train_info/time_within_train_step": 27.72025728225708, "step": 641}
{"train_info/time_between_train_steps": 0.005048513412475586, "step": 641}
{"info/global_step": 642, "train_info/time_within_train_step": 27.868391275405884, "step": 642}
{"train_info/time_between_train_steps": 0.005321502685546875, "step": 642}
{"info/global_step": 643, "train_info/time_within_train_step": 27.738252639770508, "step": 643}
{"train_info/time_between_train_steps": 0.011014461517333984, "step": 643}
{"info/global_step": 644, "train_info/time_within_train_step": 27.748359203338623, "step": 644}
{"train_info/time_between_train_steps": 0.01107335090637207, "step": 644}
{"info/global_step": 645, "train_info/time_within_train_step": 27.723795890808105, "step": 645}
{"train_info/time_between_train_steps": 0.010962724685668945, "step": 645}
{"info/global_step": 646, "train_info/time_within_train_step": 27.735137462615967, "step": 646}
{"train_info/time_between_train_steps": 0.005439281463623047, "step": 646}
{"info/global_step": 647, "train_info/time_within_train_step": 27.751171827316284, "step": 647}
{"train_info/time_between_train_steps": 0.005773782730102539, "step": 647}
{"info/global_step": 648, "train_info/time_within_train_step": 27.771984100341797, "step": 648}
{"train_info/time_between_train_steps": 0.005951404571533203, "step": 648}
{"train_info/time_between_train_steps": 20.52196979522705, "step": 648}
{"info/global_step": 649, "train_info/time_within_train_step": 27.736770153045654, "step": 649}
{"train_info/time_between_train_steps": 0.006063222885131836, "step": 649}
{"info/global_step": 650, "train_info/time_within_train_step": 27.938352584838867, "step": 650}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733747750, "_runtime": 19840}, "step": 650}
{"logs": {"train/loss": 5.2073, "train/learning_rate": 0.00030555555555555555, "train/epoch": 24.0, "_timestamp": 1733747750, "_runtime": 19840}, "step": 650}
{"train_info/time_between_train_steps": 0.007683753967285156, "step": 650}
{"info/global_step": 651, "train_info/time_within_train_step": 27.777586936950684, "step": 651}
{"train_info/time_between_train_steps": 0.005845546722412109, "step": 651}
{"info/global_step": 652, "train_info/time_within_train_step": 27.93612241744995, "step": 652}
{"train_info/time_between_train_steps": 0.005616903305053711, "step": 652}
{"info/global_step": 653, "train_info/time_within_train_step": 27.737373113632202, "step": 653}
{"train_info/time_between_train_steps": 0.005827188491821289, "step": 653}
{"info/global_step": 654, "train_info/time_within_train_step": 27.936732053756714, "step": 654}
{"train_info/time_between_train_steps": 0.005739688873291016, "step": 654}
{"info/global_step": 655, "train_info/time_within_train_step": 27.756995916366577, "step": 655}
{"train_info/time_between_train_steps": 0.005701780319213867, "step": 655}
{"info/global_step": 656, "train_info/time_within_train_step": 27.886625051498413, "step": 656}
{"train_info/time_between_train_steps": 0.005354404449462891, "step": 656}
{"info/global_step": 657, "train_info/time_within_train_step": 27.817912340164185, "step": 657}
{"train_info/time_between_train_steps": 0.005190134048461914, "step": 657}
{"info/global_step": 658, "train_info/time_within_train_step": 27.705742597579956, "step": 658}
{"train_info/time_between_train_steps": 0.005223274230957031, "step": 658}
{"info/global_step": 659, "train_info/time_within_train_step": 27.706010341644287, "step": 659}
{"train_info/time_between_train_steps": 0.005174160003662109, "step": 659}
{"info/global_step": 660, "train_info/time_within_train_step": 27.713493824005127, "step": 660}
{"train_info/time_between_train_steps": 0.0051517486572265625, "step": 660}
{"info/global_step": 661, "train_info/time_within_train_step": 27.707471132278442, "step": 661}
{"train_info/time_between_train_steps": 0.005204200744628906, "step": 661}
{"info/global_step": 662, "train_info/time_within_train_step": 27.704869508743286, "step": 662}
{"train_info/time_between_train_steps": 0.0052204132080078125, "step": 662}
{"info/global_step": 663, "train_info/time_within_train_step": 27.70941162109375, "step": 663}
{"train_info/time_between_train_steps": 0.005290031433105469, "step": 663}
{"info/global_step": 664, "train_info/time_within_train_step": 27.712816953659058, "step": 664}
{"train_info/time_between_train_steps": 0.0052337646484375, "step": 664}
{"info/global_step": 665, "train_info/time_within_train_step": 27.708250761032104, "step": 665}
{"train_info/time_between_train_steps": 0.0052945613861083984, "step": 665}
{"info/global_step": 666, "train_info/time_within_train_step": 27.708393812179565, "step": 666}
{"train_info/time_between_train_steps": 0.005190610885620117, "step": 666}
{"info/global_step": 667, "train_info/time_within_train_step": 27.72179889678955, "step": 667}
{"train_info/time_between_train_steps": 0.005450248718261719, "step": 667}
{"info/global_step": 668, "train_info/time_within_train_step": 27.72565007209778, "step": 668}
{"train_info/time_between_train_steps": 0.00513768196105957, "step": 668}
{"info/global_step": 669, "train_info/time_within_train_step": 27.733118534088135, "step": 669}
{"train_info/time_between_train_steps": 0.0052280426025390625, "step": 669}
{"info/global_step": 670, "train_info/time_within_train_step": 27.718170404434204, "step": 670}
{"train_info/time_between_train_steps": 0.005237579345703125, "step": 670}
{"info/global_step": 671, "train_info/time_within_train_step": 27.8020761013031, "step": 671}
{"train_info/time_between_train_steps": 0.010764837265014648, "step": 671}
{"info/global_step": 672, "train_info/time_within_train_step": 27.727130651474, "step": 672}
{"train_info/time_between_train_steps": 0.011396169662475586, "step": 672}
{"info/global_step": 673, "train_info/time_within_train_step": 27.820732831954956, "step": 673}
{"train_info/time_between_train_steps": 0.011188030242919922, "step": 673}
{"info/global_step": 674, "train_info/time_within_train_step": 27.734482049942017, "step": 674}
{"train_info/time_between_train_steps": 0.005802631378173828, "step": 674}
{"info/global_step": 675, "train_info/time_within_train_step": 27.74208426475525, "step": 675}
{"train_info/time_between_train_steps": 0.010628461837768555, "step": 675}
{"train_info/time_between_train_steps": 20.620266914367676, "step": 675}
{"info/global_step": 676, "train_info/time_within_train_step": 27.789583444595337, "step": 676}
{"train_info/time_between_train_steps": 0.0050716400146484375, "step": 676}
{"info/global_step": 677, "train_info/time_within_train_step": 27.9071626663208, "step": 677}
{"train_info/time_between_train_steps": 0.006676673889160156, "step": 677}
{"info/global_step": 678, "train_info/time_within_train_step": 27.73547124862671, "step": 678}
{"train_info/time_between_train_steps": 0.005365133285522461, "step": 678}
{"info/global_step": 679, "train_info/time_within_train_step": 27.868631601333618, "step": 679}
{"train_info/time_between_train_steps": 0.005340576171875, "step": 679}
{"info/global_step": 680, "train_info/time_within_train_step": 27.824342966079712, "step": 680}
{"train_info/time_between_train_steps": 0.005694389343261719, "step": 680}
{"info/global_step": 681, "train_info/time_within_train_step": 27.93387794494629, "step": 681}
{"train_info/time_between_train_steps": 0.00553131103515625, "step": 681}
{"info/global_step": 682, "train_info/time_within_train_step": 27.748826503753662, "step": 682}
{"train_info/time_between_train_steps": 0.010857820510864258, "step": 682}
{"info/global_step": 683, "train_info/time_within_train_step": 27.835777759552002, "step": 683}
{"train_info/time_between_train_steps": 0.009760618209838867, "step": 683}
{"info/global_step": 684, "train_info/time_within_train_step": 27.719922065734863, "step": 684}
{"train_info/time_between_train_steps": 0.0102691650390625, "step": 684}
{"info/global_step": 685, "train_info/time_within_train_step": 27.69807744026184, "step": 685}
{"train_info/time_between_train_steps": 0.009925603866577148, "step": 685}
{"info/global_step": 686, "train_info/time_within_train_step": 27.717633724212646, "step": 686}
{"train_info/time_between_train_steps": 0.009858846664428711, "step": 686}
{"info/global_step": 687, "train_info/time_within_train_step": 27.759849309921265, "step": 687}
{"train_info/time_between_train_steps": 0.005406856536865234, "step": 687}
{"info/global_step": 688, "train_info/time_within_train_step": 27.848172903060913, "step": 688}
{"train_info/time_between_train_steps": 0.005194902420043945, "step": 688}
{"info/global_step": 689, "train_info/time_within_train_step": 27.723135471343994, "step": 689}
{"train_info/time_between_train_steps": 0.005533695220947266, "step": 689}
{"info/global_step": 690, "train_info/time_within_train_step": 27.734174013137817, "step": 690}
{"train_info/time_between_train_steps": 0.011259078979492188, "step": 690}
{"info/global_step": 691, "train_info/time_within_train_step": 27.73047399520874, "step": 691}
{"train_info/time_between_train_steps": 0.00521087646484375, "step": 691}
{"info/global_step": 692, "train_info/time_within_train_step": 27.71221423149109, "step": 692}
{"train_info/time_between_train_steps": 0.005326032638549805, "step": 692}
{"info/global_step": 693, "train_info/time_within_train_step": 27.73199725151062, "step": 693}
{"train_info/time_between_train_steps": 0.005127429962158203, "step": 693}
{"info/global_step": 694, "train_info/time_within_train_step": 27.743197679519653, "step": 694}
{"train_info/time_between_train_steps": 0.005228281021118164, "step": 694}
{"info/global_step": 695, "train_info/time_within_train_step": 27.73246693611145, "step": 695}
{"train_info/time_between_train_steps": 0.005229949951171875, "step": 695}
{"info/global_step": 696, "train_info/time_within_train_step": 27.742820978164673, "step": 696}
{"train_info/time_between_train_steps": 0.005159616470336914, "step": 696}
{"info/global_step": 697, "train_info/time_within_train_step": 27.73163604736328, "step": 697}
{"train_info/time_between_train_steps": 0.005234956741333008, "step": 697}
{"info/global_step": 698, "train_info/time_within_train_step": 27.729117155075073, "step": 698}
{"train_info/time_between_train_steps": 0.0052378177642822266, "step": 698}
{"info/global_step": 699, "train_info/time_within_train_step": 27.733240604400635, "step": 699}
{"train_info/time_between_train_steps": 0.005555391311645508, "step": 699}
{"info/global_step": 700, "train_info/time_within_train_step": 27.751805067062378, "step": 700}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733749159, "_runtime": 21249}, "step": 700}
{"logs": {"train/loss": 5.0416, "train/learning_rate": 0.0002777777777777778, "train/epoch": 25.02, "_timestamp": 1733749159, "_runtime": 21249}, "step": 700}
{"train_info/time_between_train_steps": 2.3544883728027344, "step": 700}
{"info/global_step": 701, "train_info/time_within_train_step": 28.279643535614014, "step": 701}
{"train_info/time_between_train_steps": 0.012490987777709961, "step": 701}
{"info/global_step": 702, "train_info/time_within_train_step": 28.023571968078613, "step": 702}
{"train_info/time_between_train_steps": 0.00564122200012207, "step": 702}
{"train_info/time_between_train_steps": 20.72688627243042, "step": 702}
{"info/global_step": 703, "train_info/time_within_train_step": 27.812146186828613, "step": 703}
{"train_info/time_between_train_steps": 0.004929304122924805, "step": 703}
{"info/global_step": 704, "train_info/time_within_train_step": 27.909993410110474, "step": 704}
{"train_info/time_between_train_steps": 0.005687713623046875, "step": 704}
{"info/global_step": 705, "train_info/time_within_train_step": 27.756365060806274, "step": 705}
{"train_info/time_between_train_steps": 0.005329608917236328, "step": 705}
{"info/global_step": 706, "train_info/time_within_train_step": 27.913452863693237, "step": 706}
{"train_info/time_between_train_steps": 0.005452394485473633, "step": 706}
{"info/global_step": 707, "train_info/time_within_train_step": 27.734355449676514, "step": 707}
{"train_info/time_between_train_steps": 0.005632638931274414, "step": 707}
{"info/global_step": 708, "train_info/time_within_train_step": 27.927942752838135, "step": 708}
{"train_info/time_between_train_steps": 0.005441188812255859, "step": 708}
{"info/global_step": 709, "train_info/time_within_train_step": 27.735204696655273, "step": 709}
{"train_info/time_between_train_steps": 0.010732650756835938, "step": 709}
{"info/global_step": 710, "train_info/time_within_train_step": 27.82544994354248, "step": 710}
{"train_info/time_between_train_steps": 0.01052546501159668, "step": 710}
{"info/global_step": 711, "train_info/time_within_train_step": 27.728829860687256, "step": 711}
{"train_info/time_between_train_steps": 0.0052318572998046875, "step": 711}
{"info/global_step": 712, "train_info/time_within_train_step": 27.71099615097046, "step": 712}
{"train_info/time_between_train_steps": 0.008224725723266602, "step": 712}
{"info/global_step": 713, "train_info/time_within_train_step": 27.727489948272705, "step": 713}
{"train_info/time_between_train_steps": 0.005092144012451172, "step": 713}
{"info/global_step": 714, "train_info/time_within_train_step": 27.729954719543457, "step": 714}
{"train_info/time_between_train_steps": 0.005551338195800781, "step": 714}
{"info/global_step": 715, "train_info/time_within_train_step": 27.71587324142456, "step": 715}
{"train_info/time_between_train_steps": 0.0051174163818359375, "step": 715}
{"info/global_step": 716, "train_info/time_within_train_step": 27.71274209022522, "step": 716}
{"train_info/time_between_train_steps": 0.005277156829833984, "step": 716}
{"info/global_step": 717, "train_info/time_within_train_step": 27.70921516418457, "step": 717}
{"train_info/time_between_train_steps": 0.005329608917236328, "step": 717}
{"info/global_step": 718, "train_info/time_within_train_step": 27.704463481903076, "step": 718}
{"train_info/time_between_train_steps": 0.005251169204711914, "step": 718}
{"info/global_step": 719, "train_info/time_within_train_step": 27.810842275619507, "step": 719}
{"train_info/time_between_train_steps": 0.005279541015625, "step": 719}
{"info/global_step": 720, "train_info/time_within_train_step": 27.717429637908936, "step": 720}
{"train_info/time_between_train_steps": 0.0053310394287109375, "step": 720}
{"info/global_step": 721, "train_info/time_within_train_step": 27.710137844085693, "step": 721}
{"train_info/time_between_train_steps": 0.005146026611328125, "step": 721}
{"info/global_step": 722, "train_info/time_within_train_step": 27.735445737838745, "step": 722}
{"train_info/time_between_train_steps": 0.005383491516113281, "step": 722}
{"info/global_step": 723, "train_info/time_within_train_step": 27.714158296585083, "step": 723}
{"train_info/time_between_train_steps": 0.00561976432800293, "step": 723}
{"info/global_step": 724, "train_info/time_within_train_step": 27.729043006896973, "step": 724}
{"train_info/time_between_train_steps": 0.005274295806884766, "step": 724}
{"info/global_step": 725, "train_info/time_within_train_step": 27.719409465789795, "step": 725}
{"train_info/time_between_train_steps": 0.005254983901977539, "step": 725}
{"info/global_step": 726, "train_info/time_within_train_step": 27.741563081741333, "step": 726}
{"train_info/time_between_train_steps": 0.005807638168334961, "step": 726}
{"info/global_step": 727, "train_info/time_within_train_step": 27.74349594116211, "step": 727}
{"train_info/time_between_train_steps": 0.005467891693115234, "step": 727}
{"info/global_step": 728, "train_info/time_within_train_step": 27.74764633178711, "step": 728}
{"train_info/time_between_train_steps": 0.005576610565185547, "step": 728}
{"info/global_step": 729, "train_info/time_within_train_step": 27.793423891067505, "step": 729}
{"train_info/time_between_train_steps": 0.005595684051513672, "step": 729}
{"train_info/time_between_train_steps": 20.54286289215088, "step": 729}
{"info/global_step": 730, "train_info/time_within_train_step": 27.701030015945435, "step": 730}
{"train_info/time_between_train_steps": 0.005185604095458984, "step": 730}
{"info/global_step": 731, "train_info/time_within_train_step": 27.86791229248047, "step": 731}
{"train_info/time_between_train_steps": 0.005342960357666016, "step": 731}
{"info/global_step": 732, "train_info/time_within_train_step": 27.714997053146362, "step": 732}
{"train_info/time_between_train_steps": 0.0057220458984375, "step": 732}
{"info/global_step": 733, "train_info/time_within_train_step": 27.936159372329712, "step": 733}
{"train_info/time_between_train_steps": 0.005449533462524414, "step": 733}
{"info/global_step": 734, "train_info/time_within_train_step": 27.84874415397644, "step": 734}
{"train_info/time_between_train_steps": 0.005690097808837891, "step": 734}
{"info/global_step": 735, "train_info/time_within_train_step": 27.911348581314087, "step": 735}
{"train_info/time_between_train_steps": 0.010672807693481445, "step": 735}
{"info/global_step": 736, "train_info/time_within_train_step": 27.739991426467896, "step": 736}
{"train_info/time_between_train_steps": 0.005733489990234375, "step": 736}
{"info/global_step": 737, "train_info/time_within_train_step": 27.808342933654785, "step": 737}
{"train_info/time_between_train_steps": 0.010499954223632812, "step": 737}
{"info/global_step": 738, "train_info/time_within_train_step": 27.718589305877686, "step": 738}
{"train_info/time_between_train_steps": 0.0054187774658203125, "step": 738}
{"info/global_step": 739, "train_info/time_within_train_step": 27.70942497253418, "step": 739}
{"train_info/time_between_train_steps": 0.00508570671081543, "step": 739}
{"info/global_step": 740, "train_info/time_within_train_step": 27.715193271636963, "step": 740}
{"train_info/time_between_train_steps": 0.005404472351074219, "step": 740}
{"info/global_step": 741, "train_info/time_within_train_step": 27.705963850021362, "step": 741}
{"train_info/time_between_train_steps": 0.005165576934814453, "step": 741}
{"info/global_step": 742, "train_info/time_within_train_step": 27.70718216896057, "step": 742}
{"train_info/time_between_train_steps": 0.005079984664916992, "step": 742}
{"info/global_step": 743, "train_info/time_within_train_step": 27.705836296081543, "step": 743}
{"train_info/time_between_train_steps": 0.005255937576293945, "step": 743}
{"info/global_step": 744, "train_info/time_within_train_step": 27.708012104034424, "step": 744}
{"train_info/time_between_train_steps": 0.005357027053833008, "step": 744}
{"info/global_step": 745, "train_info/time_within_train_step": 27.704447031021118, "step": 745}
{"train_info/time_between_train_steps": 0.005138874053955078, "step": 745}
{"info/global_step": 746, "train_info/time_within_train_step": 27.704854011535645, "step": 746}
{"train_info/time_between_train_steps": 0.0052318572998046875, "step": 746}
{"info/global_step": 747, "train_info/time_within_train_step": 27.70173478126526, "step": 747}
{"train_info/time_between_train_steps": 0.005090951919555664, "step": 747}
{"info/global_step": 748, "train_info/time_within_train_step": 27.708436727523804, "step": 748}
{"train_info/time_between_train_steps": 0.005382061004638672, "step": 748}
{"info/global_step": 749, "train_info/time_within_train_step": 27.804311275482178, "step": 749}
{"train_info/time_between_train_steps": 0.005223751068115234, "step": 749}
{"info/global_step": 750, "train_info/time_within_train_step": 27.72121024131775, "step": 750}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733750592, "_runtime": 22682}, "step": 750}
{"logs": {"train/loss": 5.0168, "train/learning_rate": 0.00025, "train/epoch": 27.02, "_timestamp": 1733750592, "_runtime": 22682}, "step": 750}
{"train_info/time_between_train_steps": 0.007235527038574219, "step": 750}
{"info/global_step": 751, "train_info/time_within_train_step": 27.72299337387085, "step": 751}
{"train_info/time_between_train_steps": 0.005159854888916016, "step": 751}
{"info/global_step": 752, "train_info/time_within_train_step": 27.732471466064453, "step": 752}
{"train_info/time_between_train_steps": 0.0051190853118896484, "step": 752}
{"info/global_step": 753, "train_info/time_within_train_step": 27.728899478912354, "step": 753}
{"train_info/time_between_train_steps": 0.00549006462097168, "step": 753}
{"info/global_step": 754, "train_info/time_within_train_step": 27.730022430419922, "step": 754}
{"train_info/time_between_train_steps": 0.0051784515380859375, "step": 754}
{"info/global_step": 755, "train_info/time_within_train_step": 27.740049123764038, "step": 755}
{"train_info/time_between_train_steps": 0.0055577754974365234, "step": 755}
{"info/global_step": 756, "train_info/time_within_train_step": 27.756422996520996, "step": 756}
{"train_info/time_between_train_steps": 0.005881071090698242, "step": 756}
{"train_info/time_between_train_steps": 20.611700773239136, "step": 756}
{"info/global_step": 757, "train_info/time_within_train_step": 27.7342209815979, "step": 757}
{"train_info/time_between_train_steps": 0.005555868148803711, "step": 757}
{"info/global_step": 758, "train_info/time_within_train_step": 27.967235565185547, "step": 758}
{"train_info/time_between_train_steps": 0.010384559631347656, "step": 758}
{"info/global_step": 759, "train_info/time_within_train_step": 27.749695301055908, "step": 759}
{"train_info/time_between_train_steps": 0.005337238311767578, "step": 759}
{"info/global_step": 760, "train_info/time_within_train_step": 27.907946348190308, "step": 760}
{"train_info/time_between_train_steps": 0.005440235137939453, "step": 760}
{"info/global_step": 761, "train_info/time_within_train_step": 27.730759143829346, "step": 761}
{"train_info/time_between_train_steps": 0.0054779052734375, "step": 761}
{"info/global_step": 762, "train_info/time_within_train_step": 27.88257598876953, "step": 762}
{"train_info/time_between_train_steps": 0.010481119155883789, "step": 762}
{"info/global_step": 763, "train_info/time_within_train_step": 27.712940216064453, "step": 763}
{"train_info/time_between_train_steps": 0.005191326141357422, "step": 763}
{"info/global_step": 764, "train_info/time_within_train_step": 27.855641841888428, "step": 764}
{"train_info/time_between_train_steps": 0.0051996707916259766, "step": 764}
{"info/global_step": 765, "train_info/time_within_train_step": 27.715752124786377, "step": 765}
{"train_info/time_between_train_steps": 0.005199909210205078, "step": 765}
{"info/global_step": 766, "train_info/time_within_train_step": 27.790464878082275, "step": 766}
{"train_info/time_between_train_steps": 0.00499272346496582, "step": 766}
{"info/global_step": 767, "train_info/time_within_train_step": 27.78814196586609, "step": 767}
{"train_info/time_between_train_steps": 0.0050792694091796875, "step": 767}
{"info/global_step": 768, "train_info/time_within_train_step": 27.72385883331299, "step": 768}
{"train_info/time_between_train_steps": 0.0051326751708984375, "step": 768}
{"info/global_step": 769, "train_info/time_within_train_step": 27.724589586257935, "step": 769}
{"train_info/time_between_train_steps": 0.00519108772277832, "step": 769}
{"info/global_step": 770, "train_info/time_within_train_step": 27.712842226028442, "step": 770}
{"train_info/time_between_train_steps": 0.005265235900878906, "step": 770}
{"info/global_step": 771, "train_info/time_within_train_step": 27.725268602371216, "step": 771}
{"train_info/time_between_train_steps": 0.005365133285522461, "step": 771}
{"info/global_step": 772, "train_info/time_within_train_step": 27.704548120498657, "step": 772}
{"train_info/time_between_train_steps": 0.005095481872558594, "step": 772}
{"info/global_step": 773, "train_info/time_within_train_step": 27.70856261253357, "step": 773}
{"train_info/time_between_train_steps": 0.005137920379638672, "step": 773}
{"info/global_step": 774, "train_info/time_within_train_step": 27.718549966812134, "step": 774}
{"train_info/time_between_train_steps": 0.005258083343505859, "step": 774}
{"info/global_step": 775, "train_info/time_within_train_step": 27.714104413986206, "step": 775}
{"train_info/time_between_train_steps": 0.005175113677978516, "step": 775}
{"info/global_step": 776, "train_info/time_within_train_step": 27.701802968978882, "step": 776}
{"train_info/time_between_train_steps": 0.005059719085693359, "step": 776}
{"info/global_step": 777, "train_info/time_within_train_step": 27.70001530647278, "step": 777}
{"train_info/time_between_train_steps": 0.0050585269927978516, "step": 777}
{"info/global_step": 778, "train_info/time_within_train_step": 27.699514389038086, "step": 778}
{"train_info/time_between_train_steps": 0.005114316940307617, "step": 778}
{"info/global_step": 779, "train_info/time_within_train_step": 27.70265507698059, "step": 779}
{"train_info/time_between_train_steps": 0.010106086730957031, "step": 779}
{"info/global_step": 780, "train_info/time_within_train_step": 27.804004669189453, "step": 780}
{"train_info/time_between_train_steps": 0.005341291427612305, "step": 780}
{"info/global_step": 781, "train_info/time_within_train_step": 27.710665464401245, "step": 781}
{"train_info/time_between_train_steps": 0.005124330520629883, "step": 781}
{"info/global_step": 782, "train_info/time_within_train_step": 27.718562364578247, "step": 782}
{"train_info/time_between_train_steps": 0.010215520858764648, "step": 782}
{"info/global_step": 783, "train_info/time_within_train_step": 27.738779544830322, "step": 783}
{"train_info/time_between_train_steps": 0.010881423950195312, "step": 783}
{"train_info/time_between_train_steps": 20.687994718551636, "step": 783}
{"info/global_step": 784, "train_info/time_within_train_step": 27.691002368927002, "step": 784}
{"train_info/time_between_train_steps": 0.004877328872680664, "step": 784}
{"info/global_step": 785, "train_info/time_within_train_step": 27.838944673538208, "step": 785}
{"train_info/time_between_train_steps": 0.00529026985168457, "step": 785}
{"info/global_step": 786, "train_info/time_within_train_step": 27.724732160568237, "step": 786}
{"train_info/time_between_train_steps": 0.00510406494140625, "step": 786}
{"info/global_step": 787, "train_info/time_within_train_step": 27.827451944351196, "step": 787}
{"train_info/time_between_train_steps": 0.005472898483276367, "step": 787}
{"info/global_step": 788, "train_info/time_within_train_step": 27.705137252807617, "step": 788}
{"train_info/time_between_train_steps": 0.005099058151245117, "step": 788}
{"info/global_step": 789, "train_info/time_within_train_step": 27.822210788726807, "step": 789}
{"train_info/time_between_train_steps": 0.0050542354583740234, "step": 789}
{"info/global_step": 790, "train_info/time_within_train_step": 27.72640872001648, "step": 790}
{"train_info/time_between_train_steps": 0.005434989929199219, "step": 790}
{"info/global_step": 791, "train_info/time_within_train_step": 27.791432857513428, "step": 791}
{"train_info/time_between_train_steps": 0.0053060054779052734, "step": 791}
{"info/global_step": 792, "train_info/time_within_train_step": 27.705663919448853, "step": 792}
{"train_info/time_between_train_steps": 0.0049893856048583984, "step": 792}
{"info/global_step": 793, "train_info/time_within_train_step": 27.69251537322998, "step": 793}
{"train_info/time_between_train_steps": 0.005044698715209961, "step": 793}
{"info/global_step": 794, "train_info/time_within_train_step": 27.701319932937622, "step": 794}
{"train_info/time_between_train_steps": 0.00525355339050293, "step": 794}
{"info/global_step": 795, "train_info/time_within_train_step": 27.79950451850891, "step": 795}
{"train_info/time_between_train_steps": 0.0049893856048583984, "step": 795}
{"info/global_step": 796, "train_info/time_within_train_step": 27.700388193130493, "step": 796}
{"train_info/time_between_train_steps": 0.005009889602661133, "step": 796}
{"info/global_step": 797, "train_info/time_within_train_step": 27.702900409698486, "step": 797}
{"train_info/time_between_train_steps": 0.00545811653137207, "step": 797}
{"info/global_step": 798, "train_info/time_within_train_step": 27.704137086868286, "step": 798}
{"train_info/time_between_train_steps": 0.005080461502075195, "step": 798}
{"info/global_step": 799, "train_info/time_within_train_step": 27.731144905090332, "step": 799}
{"train_info/time_between_train_steps": 0.0049839019775390625, "step": 799}
{"info/global_step": 800, "train_info/time_within_train_step": 27.697519063949585, "step": 800}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733752021, "_runtime": 24111}, "step": 800}
{"logs": {"train/loss": 4.9351, "train/learning_rate": 0.00022222222222222218, "train/epoch": 29.01, "_timestamp": 1733752021, "_runtime": 24111}, "step": 800}
{"train_info/time_between_train_steps": 2.3555991649627686, "step": 800}
{"info/global_step": 801, "train_info/time_within_train_step": 27.71208620071411, "step": 801}
{"train_info/time_between_train_steps": 0.005163908004760742, "step": 801}
{"info/global_step": 802, "train_info/time_within_train_step": 27.703977823257446, "step": 802}
{"train_info/time_between_train_steps": 0.005080461502075195, "step": 802}
{"info/global_step": 803, "train_info/time_within_train_step": 27.71625852584839, "step": 803}
{"train_info/time_between_train_steps": 0.0051381587982177734, "step": 803}
{"info/global_step": 804, "train_info/time_within_train_step": 27.708868265151978, "step": 804}
{"train_info/time_between_train_steps": 0.010104894638061523, "step": 804}
{"info/global_step": 805, "train_info/time_within_train_step": 27.717331171035767, "step": 805}
{"train_info/time_between_train_steps": 0.0052623748779296875, "step": 805}
{"info/global_step": 806, "train_info/time_within_train_step": 27.712244510650635, "step": 806}
{"train_info/time_between_train_steps": 0.0050160884857177734, "step": 806}
{"info/global_step": 807, "train_info/time_within_train_step": 27.71908211708069, "step": 807}
{"train_info/time_between_train_steps": 0.005202531814575195, "step": 807}
{"info/global_step": 808, "train_info/time_within_train_step": 27.730326890945435, "step": 808}
{"train_info/time_between_train_steps": 0.005337953567504883, "step": 808}
{"info/global_step": 809, "train_info/time_within_train_step": 27.73056721687317, "step": 809}
{"train_info/time_between_train_steps": 0.01042485237121582, "step": 809}
{"info/global_step": 810, "train_info/time_within_train_step": 27.839595794677734, "step": 810}
{"train_info/time_between_train_steps": 0.010894536972045898, "step": 810}
{"train_info/time_between_train_steps": 20.597439527511597, "step": 810}
{"info/global_step": 811, "train_info/time_within_train_step": 27.711843729019165, "step": 811}
{"train_info/time_between_train_steps": 0.0050432682037353516, "step": 811}
{"info/global_step": 812, "train_info/time_within_train_step": 27.823791027069092, "step": 812}
{"train_info/time_between_train_steps": 0.009823799133300781, "step": 812}
{"info/global_step": 813, "train_info/time_within_train_step": 27.724287033081055, "step": 813}
{"train_info/time_between_train_steps": 0.005103349685668945, "step": 813}
{"info/global_step": 814, "train_info/time_within_train_step": 27.885627508163452, "step": 814}
{"train_info/time_between_train_steps": 0.0052073001861572266, "step": 814}
{"info/global_step": 815, "train_info/time_within_train_step": 27.72677493095398, "step": 815}
{"train_info/time_between_train_steps": 0.005335330963134766, "step": 815}
{"info/global_step": 816, "train_info/time_within_train_step": 27.843819856643677, "step": 816}
{"train_info/time_between_train_steps": 0.005197286605834961, "step": 816}
{"info/global_step": 817, "train_info/time_within_train_step": 27.72234296798706, "step": 817}
{"train_info/time_between_train_steps": 0.009819746017456055, "step": 817}
{"info/global_step": 818, "train_info/time_within_train_step": 27.76283621788025, "step": 818}
{"train_info/time_between_train_steps": 0.010358095169067383, "step": 818}
{"info/global_step": 819, "train_info/time_within_train_step": 27.724074840545654, "step": 819}
{"train_info/time_between_train_steps": 0.005196571350097656, "step": 819}
{"info/global_step": 820, "train_info/time_within_train_step": 27.71767020225525, "step": 820}
{"train_info/time_between_train_steps": 0.005205869674682617, "step": 820}
{"info/global_step": 821, "train_info/time_within_train_step": 27.714492321014404, "step": 821}
{"train_info/time_between_train_steps": 0.0051310062408447266, "step": 821}
{"info/global_step": 822, "train_info/time_within_train_step": 27.71778106689453, "step": 822}
{"train_info/time_between_train_steps": 0.005192279815673828, "step": 822}
{"info/global_step": 823, "train_info/time_within_train_step": 27.714547872543335, "step": 823}
{"train_info/time_between_train_steps": 0.005215167999267578, "step": 823}
{"info/global_step": 824, "train_info/time_within_train_step": 27.726882457733154, "step": 824}
{"train_info/time_between_train_steps": 0.005132913589477539, "step": 824}
{"info/global_step": 825, "train_info/time_within_train_step": 27.907797813415527, "step": 825}
{"train_info/time_between_train_steps": 0.009750127792358398, "step": 825}
{"info/global_step": 826, "train_info/time_within_train_step": 28.074145793914795, "step": 826}
{"train_info/time_between_train_steps": 0.009995698928833008, "step": 826}
{"info/global_step": 827, "train_info/time_within_train_step": 28.058331966400146, "step": 827}
{"train_info/time_between_train_steps": 0.005308628082275391, "step": 827}
{"info/global_step": 828, "train_info/time_within_train_step": 27.74130916595459, "step": 828}
{"train_info/time_between_train_steps": 0.005101680755615234, "step": 828}
{"info/global_step": 829, "train_info/time_within_train_step": 27.734254598617554, "step": 829}
{"train_info/time_between_train_steps": 0.0051877498626708984, "step": 829}
{"info/global_step": 830, "train_info/time_within_train_step": 27.720829010009766, "step": 830}
{"train_info/time_between_train_steps": 0.005107879638671875, "step": 830}
{"info/global_step": 831, "train_info/time_within_train_step": 27.720646858215332, "step": 831}
{"train_info/time_between_train_steps": 0.005264997482299805, "step": 831}
{"info/global_step": 832, "train_info/time_within_train_step": 27.718143463134766, "step": 832}
{"train_info/time_between_train_steps": 0.00506591796875, "step": 832}
{"info/global_step": 833, "train_info/time_within_train_step": 27.70400309562683, "step": 833}
{"train_info/time_between_train_steps": 0.005142688751220703, "step": 833}
{"info/global_step": 834, "train_info/time_within_train_step": 27.71899914741516, "step": 834}
{"train_info/time_between_train_steps": 0.0054094791412353516, "step": 834}
{"info/global_step": 835, "train_info/time_within_train_step": 27.732805490493774, "step": 835}
{"train_info/time_between_train_steps": 0.00531768798828125, "step": 835}
{"info/global_step": 836, "train_info/time_within_train_step": 27.780145406723022, "step": 836}
{"train_info/time_between_train_steps": 0.0052220821380615234, "step": 836}
{"info/global_step": 837, "train_info/time_within_train_step": 27.77595019340515, "step": 837}
{"train_info/time_between_train_steps": 0.005728483200073242, "step": 837}
{"train_info/time_between_train_steps": 20.523865222930908, "step": 837}
{"info/global_step": 838, "train_info/time_within_train_step": 27.720449209213257, "step": 838}
{"train_info/time_between_train_steps": 0.005013942718505859, "step": 838}
{"info/global_step": 839, "train_info/time_within_train_step": 27.811813831329346, "step": 839}
{"train_info/time_between_train_steps": 0.009808540344238281, "step": 839}
{"info/global_step": 840, "train_info/time_within_train_step": 27.818087577819824, "step": 840}
{"train_info/time_between_train_steps": 0.0051343441009521484, "step": 840}
{"info/global_step": 841, "train_info/time_within_train_step": 27.836276292800903, "step": 841}
{"train_info/time_between_train_steps": 0.005032539367675781, "step": 841}
{"info/global_step": 842, "train_info/time_within_train_step": 27.731199741363525, "step": 842}
{"train_info/time_between_train_steps": 0.010403871536254883, "step": 842}
{"info/global_step": 843, "train_info/time_within_train_step": 27.86820363998413, "step": 843}
{"train_info/time_between_train_steps": 0.010268688201904297, "step": 843}
{"info/global_step": 844, "train_info/time_within_train_step": 27.7256863117218, "step": 844}
{"train_info/time_between_train_steps": 0.010345697402954102, "step": 844}
{"info/global_step": 845, "train_info/time_within_train_step": 27.754212856292725, "step": 845}
{"train_info/time_between_train_steps": 0.010195255279541016, "step": 845}
{"info/global_step": 846, "train_info/time_within_train_step": 27.71847128868103, "step": 846}
{"train_info/time_between_train_steps": 0.009917020797729492, "step": 846}
{"info/global_step": 847, "train_info/time_within_train_step": 27.711312294006348, "step": 847}
{"train_info/time_between_train_steps": 0.00514674186706543, "step": 847}
{"info/global_step": 848, "train_info/time_within_train_step": 27.715456247329712, "step": 848}
{"train_info/time_between_train_steps": 0.009788751602172852, "step": 848}
{"info/global_step": 849, "train_info/time_within_train_step": 27.73181915283203, "step": 849}
{"train_info/time_between_train_steps": 0.005105733871459961, "step": 849}
{"info/global_step": 850, "train_info/time_within_train_step": 27.725518465042114, "step": 850}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733753453, "_runtime": 25543}, "step": 850}
{"logs": {"train/loss": 4.8656, "train/learning_rate": 0.00019444444444444443, "train/epoch": 31.01, "_timestamp": 1733753453, "_runtime": 25543}, "step": 850}
{"train_info/time_between_train_steps": 0.007051944732666016, "step": 850}
{"info/global_step": 851, "train_info/time_within_train_step": 27.740758180618286, "step": 851}
{"train_info/time_between_train_steps": 0.005272388458251953, "step": 851}
{"info/global_step": 852, "train_info/time_within_train_step": 27.73004460334778, "step": 852}
{"train_info/time_between_train_steps": 0.010231971740722656, "step": 852}
{"info/global_step": 853, "train_info/time_within_train_step": 27.719752550125122, "step": 853}
{"train_info/time_between_train_steps": 0.005142688751220703, "step": 853}
{"info/global_step": 854, "train_info/time_within_train_step": 27.714351892471313, "step": 854}
{"train_info/time_between_train_steps": 0.0051343441009521484, "step": 854}
{"info/global_step": 855, "train_info/time_within_train_step": 27.73592710494995, "step": 855}
{"train_info/time_between_train_steps": 0.005236148834228516, "step": 855}
{"info/global_step": 856, "train_info/time_within_train_step": 27.818369150161743, "step": 856}
{"train_info/time_between_train_steps": 0.005064964294433594, "step": 856}
{"info/global_step": 857, "train_info/time_within_train_step": 28.005573511123657, "step": 857}
{"train_info/time_between_train_steps": 0.005345344543457031, "step": 857}
{"info/global_step": 858, "train_info/time_within_train_step": 27.74093270301819, "step": 858}
{"train_info/time_between_train_steps": 0.00543522834777832, "step": 858}
{"info/global_step": 859, "train_info/time_within_train_step": 27.74711799621582, "step": 859}
{"train_info/time_between_train_steps": 0.005370140075683594, "step": 859}
{"info/global_step": 860, "train_info/time_within_train_step": 27.73588538169861, "step": 860}
{"train_info/time_between_train_steps": 0.005107879638671875, "step": 860}
{"info/global_step": 861, "train_info/time_within_train_step": 27.768044471740723, "step": 861}
{"train_info/time_between_train_steps": 0.005491733551025391, "step": 861}
{"info/global_step": 862, "train_info/time_within_train_step": 27.73136878013611, "step": 862}
{"train_info/time_between_train_steps": 0.005176067352294922, "step": 862}
{"info/global_step": 863, "train_info/time_within_train_step": 27.746469736099243, "step": 863}
{"train_info/time_between_train_steps": 0.005597829818725586, "step": 863}
{"info/global_step": 864, "train_info/time_within_train_step": 27.773672819137573, "step": 864}
{"train_info/time_between_train_steps": 0.00548553466796875, "step": 864}
{"train_info/time_between_train_steps": 20.797255754470825, "step": 864}
{"info/global_step": 865, "train_info/time_within_train_step": 27.752293348312378, "step": 865}
{"train_info/time_between_train_steps": 0.0051610469818115234, "step": 865}
{"info/global_step": 866, "train_info/time_within_train_step": 27.874290704727173, "step": 866}
{"train_info/time_between_train_steps": 0.00516963005065918, "step": 866}
{"info/global_step": 867, "train_info/time_within_train_step": 27.750015258789062, "step": 867}
{"train_info/time_between_train_steps": 0.005584239959716797, "step": 867}
{"info/global_step": 868, "train_info/time_within_train_step": 27.919322967529297, "step": 868}
{"train_info/time_between_train_steps": 0.0054247379302978516, "step": 868}
{"info/global_step": 869, "train_info/time_within_train_step": 27.747961282730103, "step": 869}
{"train_info/time_between_train_steps": 0.00540471076965332, "step": 869}
{"info/global_step": 870, "train_info/time_within_train_step": 27.954587936401367, "step": 870}
{"train_info/time_between_train_steps": 0.005473613739013672, "step": 870}
{"info/global_step": 871, "train_info/time_within_train_step": 27.82039761543274, "step": 871}
{"train_info/time_between_train_steps": 0.005512237548828125, "step": 871}
{"info/global_step": 872, "train_info/time_within_train_step": 27.906551122665405, "step": 872}
{"train_info/time_between_train_steps": 0.005404949188232422, "step": 872}
{"info/global_step": 873, "train_info/time_within_train_step": 27.74438452720642, "step": 873}
{"train_info/time_between_train_steps": 0.0052449703216552734, "step": 873}
{"info/global_step": 874, "train_info/time_within_train_step": 27.758524417877197, "step": 874}
{"train_info/time_between_train_steps": 0.006946086883544922, "step": 874}
{"info/global_step": 875, "train_info/time_within_train_step": 27.7564697265625, "step": 875}
{"train_info/time_between_train_steps": 0.005316019058227539, "step": 875}
{"info/global_step": 876, "train_info/time_within_train_step": 27.84838628768921, "step": 876}
{"train_info/time_between_train_steps": 0.005823612213134766, "step": 876}
{"info/global_step": 877, "train_info/time_within_train_step": 27.844353914260864, "step": 877}
{"train_info/time_between_train_steps": 0.0051116943359375, "step": 877}
{"info/global_step": 878, "train_info/time_within_train_step": 27.803064823150635, "step": 878}
{"train_info/time_between_train_steps": 0.010304927825927734, "step": 878}
{"info/global_step": 879, "train_info/time_within_train_step": 27.724650859832764, "step": 879}
{"train_info/time_between_train_steps": 0.005745410919189453, "step": 879}
{"info/global_step": 880, "train_info/time_within_train_step": 27.815582513809204, "step": 880}
{"train_info/time_between_train_steps": 0.009806394577026367, "step": 880}
{"info/global_step": 881, "train_info/time_within_train_step": 27.775553226470947, "step": 881}
{"train_info/time_between_train_steps": 0.005480289459228516, "step": 881}
{"info/global_step": 882, "train_info/time_within_train_step": 27.765264749526978, "step": 882}
{"train_info/time_between_train_steps": 0.01000523567199707, "step": 882}
{"info/global_step": 883, "train_info/time_within_train_step": 27.786093950271606, "step": 883}
{"train_info/time_between_train_steps": 0.005460977554321289, "step": 883}
{"info/global_step": 884, "train_info/time_within_train_step": 27.760204076766968, "step": 884}
{"train_info/time_between_train_steps": 0.00536656379699707, "step": 884}
{"info/global_step": 885, "train_info/time_within_train_step": 27.762107610702515, "step": 885}
{"train_info/time_between_train_steps": 0.005326986312866211, "step": 885}
{"info/global_step": 886, "train_info/time_within_train_step": 27.760233879089355, "step": 886}
{"train_info/time_between_train_steps": 0.01019430160522461, "step": 886}
{"info/global_step": 887, "train_info/time_within_train_step": 27.851397037506104, "step": 887}
{"train_info/time_between_train_steps": 0.005823612213134766, "step": 887}
{"info/global_step": 888, "train_info/time_within_train_step": 27.751134157180786, "step": 888}
{"train_info/time_between_train_steps": 0.005751609802246094, "step": 888}
{"info/global_step": 889, "train_info/time_within_train_step": 27.72193694114685, "step": 889}
{"train_info/time_between_train_steps": 0.005343198776245117, "step": 889}
{"info/global_step": 890, "train_info/time_within_train_step": 27.71769905090332, "step": 890}
{"train_info/time_between_train_steps": 0.0057048797607421875, "step": 890}
{"info/global_step": 891, "train_info/time_within_train_step": 27.737043380737305, "step": 891}
{"train_info/time_between_train_steps": 0.0059814453125, "step": 891}
{"train_info/time_between_train_steps": 20.608973503112793, "step": 891}
{"info/global_step": 892, "train_info/time_within_train_step": 28.01405930519104, "step": 892}
{"train_info/time_between_train_steps": 0.005148887634277344, "step": 892}
{"info/global_step": 893, "train_info/time_within_train_step": 27.89219880104065, "step": 893}
{"train_info/time_between_train_steps": 0.005072355270385742, "step": 893}
{"info/global_step": 894, "train_info/time_within_train_step": 27.760690212249756, "step": 894}
{"train_info/time_between_train_steps": 0.005579233169555664, "step": 894}
{"info/global_step": 895, "train_info/time_within_train_step": 27.955694675445557, "step": 895}
{"train_info/time_between_train_steps": 0.006352901458740234, "step": 895}
{"info/global_step": 896, "train_info/time_within_train_step": 27.747830629348755, "step": 896}
{"train_info/time_between_train_steps": 0.005683183670043945, "step": 896}
{"info/global_step": 897, "train_info/time_within_train_step": 27.92625880241394, "step": 897}
{"train_info/time_between_train_steps": 0.00555419921875, "step": 897}
{"info/global_step": 898, "train_info/time_within_train_step": 27.762584686279297, "step": 898}
{"train_info/time_between_train_steps": 0.0059680938720703125, "step": 898}
{"info/global_step": 899, "train_info/time_within_train_step": 27.804712057113647, "step": 899}
{"train_info/time_between_train_steps": 0.005370378494262695, "step": 899}
{"info/global_step": 900, "train_info/time_within_train_step": 27.753782510757446, "step": 900}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733754884, "_runtime": 26974}, "step": 900}
{"logs": {"train/loss": 4.8022, "train/learning_rate": 0.00016666666666666666, "train/epoch": 33.01, "_timestamp": 1733754884, "_runtime": 26974}, "step": 900}
{"train_info/time_between_train_steps": 2.9574942588806152, "step": 900}
{"info/global_step": 901, "train_info/time_within_train_step": 27.870670795440674, "step": 901}
{"train_info/time_between_train_steps": 0.007602214813232422, "step": 901}
{"info/global_step": 902, "train_info/time_within_train_step": 28.117194890975952, "step": 902}
{"train_info/time_between_train_steps": 0.010053634643554688, "step": 902}
{"info/global_step": 903, "train_info/time_within_train_step": 28.08033013343811, "step": 903}
{"train_info/time_between_train_steps": 0.009603500366210938, "step": 903}
{"info/global_step": 904, "train_info/time_within_train_step": 28.08544683456421, "step": 904}
{"train_info/time_between_train_steps": 0.012723207473754883, "step": 904}
{"info/global_step": 905, "train_info/time_within_train_step": 28.053199291229248, "step": 905}
{"train_info/time_between_train_steps": 0.013211727142333984, "step": 905}
{"info/global_step": 906, "train_info/time_within_train_step": 28.24870753288269, "step": 906}
{"train_info/time_between_train_steps": 0.009729862213134766, "step": 906}
{"info/global_step": 907, "train_info/time_within_train_step": 28.120388746261597, "step": 907}
{"train_info/time_between_train_steps": 0.005358695983886719, "step": 907}
{"info/global_step": 908, "train_info/time_within_train_step": 28.209376573562622, "step": 908}
{"train_info/time_between_train_steps": 0.005248069763183594, "step": 908}
{"info/global_step": 909, "train_info/time_within_train_step": 27.864581823349, "step": 909}
{"train_info/time_between_train_steps": 0.012825727462768555, "step": 909}
{"info/global_step": 910, "train_info/time_within_train_step": 28.163811445236206, "step": 910}
{"train_info/time_between_train_steps": 0.009636878967285156, "step": 910}
{"info/global_step": 911, "train_info/time_within_train_step": 28.022528409957886, "step": 911}
{"train_info/time_between_train_steps": 0.005316019058227539, "step": 911}
{"info/global_step": 912, "train_info/time_within_train_step": 28.00480818748474, "step": 912}
{"train_info/time_between_train_steps": 0.01289820671081543, "step": 912}
{"info/global_step": 913, "train_info/time_within_train_step": 28.00410485267639, "step": 913}
{"train_info/time_between_train_steps": 0.01044607162475586, "step": 913}
{"info/global_step": 914, "train_info/time_within_train_step": 28.194491147994995, "step": 914}
{"train_info/time_between_train_steps": 0.013077735900878906, "step": 914}
{"info/global_step": 915, "train_info/time_within_train_step": 28.171555042266846, "step": 915}
{"train_info/time_between_train_steps": 0.010096311569213867, "step": 915}
{"info/global_step": 916, "train_info/time_within_train_step": 28.159927129745483, "step": 916}
{"train_info/time_between_train_steps": 0.005877256393432617, "step": 916}
{"info/global_step": 917, "train_info/time_within_train_step": 28.14809799194336, "step": 917}
{"train_info/time_between_train_steps": 0.010483741760253906, "step": 917}
{"info/global_step": 918, "train_info/time_within_train_step": 28.099886417388916, "step": 918}
{"train_info/time_between_train_steps": 0.0165402889251709, "step": 918}
{"train_info/time_between_train_steps": 21.22126817703247, "step": 918}
{"info/global_step": 919, "train_info/time_within_train_step": 28.23785972595215, "step": 919}
{"train_info/time_between_train_steps": 0.007148265838623047, "step": 919}
{"info/global_step": 920, "train_info/time_within_train_step": 28.2588369846344, "step": 920}
{"train_info/time_between_train_steps": 0.0057451725006103516, "step": 920}
{"info/global_step": 921, "train_info/time_within_train_step": 28.031396627426147, "step": 921}
{"train_info/time_between_train_steps": 0.01442575454711914, "step": 921}
{"info/global_step": 922, "train_info/time_within_train_step": 28.400695323944092, "step": 922}
{"train_info/time_between_train_steps": 0.014317750930786133, "step": 922}
{"info/global_step": 923, "train_info/time_within_train_step": 28.29048442840576, "step": 923}
{"train_info/time_between_train_steps": 0.01000666618347168, "step": 923}
{"info/global_step": 924, "train_info/time_within_train_step": 28.538283824920654, "step": 924}
{"train_info/time_between_train_steps": 0.00986337661743164, "step": 924}
{"info/global_step": 925, "train_info/time_within_train_step": 28.168827772140503, "step": 925}
{"train_info/time_between_train_steps": 0.010832548141479492, "step": 925}
{"info/global_step": 926, "train_info/time_within_train_step": 27.86163568496704, "step": 926}
{"train_info/time_between_train_steps": 0.005661725997924805, "step": 926}
{"info/global_step": 927, "train_info/time_within_train_step": 27.764634609222412, "step": 927}
{"train_info/time_between_train_steps": 0.00516057014465332, "step": 927}
{"info/global_step": 928, "train_info/time_within_train_step": 28.075195789337158, "step": 928}
{"train_info/time_between_train_steps": 0.009793519973754883, "step": 928}
{"info/global_step": 929, "train_info/time_within_train_step": 28.205484867095947, "step": 929}
{"train_info/time_between_train_steps": 0.00951695442199707, "step": 929}
{"info/global_step": 930, "train_info/time_within_train_step": 27.841464042663574, "step": 930}
{"train_info/time_between_train_steps": 0.010114431381225586, "step": 930}
{"info/global_step": 931, "train_info/time_within_train_step": 28.274325132369995, "step": 931}
{"train_info/time_between_train_steps": 0.012579679489135742, "step": 931}
{"info/global_step": 932, "train_info/time_within_train_step": 28.385294914245605, "step": 932}
{"train_info/time_between_train_steps": 0.0055353641510009766, "step": 932}
{"info/global_step": 933, "train_info/time_within_train_step": 28.142786741256714, "step": 933}
{"train_info/time_between_train_steps": 0.012909173965454102, "step": 933}
{"info/global_step": 934, "train_info/time_within_train_step": 28.112680912017822, "step": 934}
{"train_info/time_between_train_steps": 0.011272192001342773, "step": 934}
{"info/global_step": 935, "train_info/time_within_train_step": 28.026901960372925, "step": 935}
{"train_info/time_between_train_steps": 0.005838871002197266, "step": 935}
{"info/global_step": 936, "train_info/time_within_train_step": 27.913586854934692, "step": 936}
{"train_info/time_between_train_steps": 0.009721040725708008, "step": 936}
{"info/global_step": 937, "train_info/time_within_train_step": 28.047208309173584, "step": 937}
{"train_info/time_between_train_steps": 0.0057528018951416016, "step": 937}
{"info/global_step": 938, "train_info/time_within_train_step": 27.966993808746338, "step": 938}
{"train_info/time_between_train_steps": 0.00983119010925293, "step": 938}
{"info/global_step": 939, "train_info/time_within_train_step": 28.19424080848694, "step": 939}
{"train_info/time_between_train_steps": 0.013045549392700195, "step": 939}
{"info/global_step": 940, "train_info/time_within_train_step": 28.260804176330566, "step": 940}
{"train_info/time_between_train_steps": 0.010981082916259766, "step": 940}
{"info/global_step": 941, "train_info/time_within_train_step": 28.20106339454651, "step": 941}
{"train_info/time_between_train_steps": 0.009920597076416016, "step": 941}
{"info/global_step": 942, "train_info/time_within_train_step": 28.049383401870728, "step": 942}
{"train_info/time_between_train_steps": 0.007566928863525391, "step": 942}
{"info/global_step": 943, "train_info/time_within_train_step": 28.182292938232422, "step": 943}
{"train_info/time_between_train_steps": 0.013134241104125977, "step": 943}
{"info/global_step": 944, "train_info/time_within_train_step": 28.22872042655945, "step": 944}
{"train_info/time_between_train_steps": 0.013620138168334961, "step": 944}
{"info/global_step": 945, "train_info/time_within_train_step": 28.277465343475342, "step": 945}
{"train_info/time_between_train_steps": 0.006234884262084961, "step": 945}
{"train_info/time_between_train_steps": 21.10333490371704, "step": 945}
{"info/global_step": 946, "train_info/time_within_train_step": 27.894716024398804, "step": 946}
{"train_info/time_between_train_steps": 0.005265474319458008, "step": 946}
{"info/global_step": 947, "train_info/time_within_train_step": 28.29163408279419, "step": 947}
{"train_info/time_between_train_steps": 0.011679649353027344, "step": 947}
{"info/global_step": 948, "train_info/time_within_train_step": 28.44573211669922, "step": 948}
{"train_info/time_between_train_steps": 0.01494288444519043, "step": 948}
{"info/global_step": 949, "train_info/time_within_train_step": 28.583067178726196, "step": 949}
{"train_info/time_between_train_steps": 0.010570287704467773, "step": 949}
{"info/global_step": 950, "train_info/time_within_train_step": 28.226224660873413, "step": 950}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733756337, "_runtime": 28427}, "step": 950}
{"logs": {"train/loss": 4.7449, "train/learning_rate": 0.0001388888888888889, "train/epoch": 35.0, "_timestamp": 1733756337, "_runtime": 28427}, "step": 950}
{"train_info/time_between_train_steps": 0.007596015930175781, "step": 950}
{"info/global_step": 951, "train_info/time_within_train_step": 28.367090702056885, "step": 951}
{"train_info/time_between_train_steps": 0.012933015823364258, "step": 951}
{"info/global_step": 952, "train_info/time_within_train_step": 28.194637298583984, "step": 952}
{"train_info/time_between_train_steps": 0.01331639289855957, "step": 952}
{"info/global_step": 953, "train_info/time_within_train_step": 28.198424577713013, "step": 953}
{"train_info/time_between_train_steps": 0.013239383697509766, "step": 953}
{"info/global_step": 954, "train_info/time_within_train_step": 27.910306215286255, "step": 954}
{"train_info/time_between_train_steps": 0.00532078742980957, "step": 954}
{"info/global_step": 955, "train_info/time_within_train_step": 28.071291208267212, "step": 955}
{"train_info/time_between_train_steps": 0.005220651626586914, "step": 955}
{"info/global_step": 956, "train_info/time_within_train_step": 28.148531198501587, "step": 956}
{"train_info/time_between_train_steps": 0.012646198272705078, "step": 956}
{"info/global_step": 957, "train_info/time_within_train_step": 28.19058084487915, "step": 957}
{"train_info/time_between_train_steps": 0.013000965118408203, "step": 957}
{"info/global_step": 958, "train_info/time_within_train_step": 28.306593418121338, "step": 958}
{"train_info/time_between_train_steps": 0.012982368469238281, "step": 958}
{"info/global_step": 959, "train_info/time_within_train_step": 28.147261381149292, "step": 959}
{"train_info/time_between_train_steps": 0.009898900985717773, "step": 959}
{"info/global_step": 960, "train_info/time_within_train_step": 28.089917421340942, "step": 960}
{"train_info/time_between_train_steps": 0.009948015213012695, "step": 960}
{"info/global_step": 961, "train_info/time_within_train_step": 28.130083322525024, "step": 961}
{"train_info/time_between_train_steps": 0.012966632843017578, "step": 961}
{"info/global_step": 962, "train_info/time_within_train_step": 28.310118198394775, "step": 962}
{"train_info/time_between_train_steps": 0.015857696533203125, "step": 962}
{"info/global_step": 963, "train_info/time_within_train_step": 28.401059865951538, "step": 963}
{"train_info/time_between_train_steps": 0.011188268661499023, "step": 963}
{"info/global_step": 964, "train_info/time_within_train_step": 28.157625675201416, "step": 964}
{"train_info/time_between_train_steps": 0.015714168548583984, "step": 964}
{"info/global_step": 965, "train_info/time_within_train_step": 28.134469747543335, "step": 965}
{"train_info/time_between_train_steps": 0.013096809387207031, "step": 965}
{"info/global_step": 966, "train_info/time_within_train_step": 28.28162717819214, "step": 966}
{"train_info/time_between_train_steps": 0.012695550918579102, "step": 966}
{"info/global_step": 967, "train_info/time_within_train_step": 28.26954483985901, "step": 967}
{"train_info/time_between_train_steps": 0.012966156005859375, "step": 967}
{"info/global_step": 968, "train_info/time_within_train_step": 28.15302801132202, "step": 968}
{"train_info/time_between_train_steps": 0.012058019638061523, "step": 968}
{"info/global_step": 969, "train_info/time_within_train_step": 28.1940016746521, "step": 969}
{"train_info/time_between_train_steps": 0.005910158157348633, "step": 969}
{"info/global_step": 970, "train_info/time_within_train_step": 28.142170906066895, "step": 970}
{"train_info/time_between_train_steps": 0.010223150253295898, "step": 970}
{"info/global_step": 971, "train_info/time_within_train_step": 28.180810689926147, "step": 971}
{"train_info/time_between_train_steps": 0.014508962631225586, "step": 971}
{"info/global_step": 972, "train_info/time_within_train_step": 28.287696599960327, "step": 972}
{"train_info/time_between_train_steps": 0.014188051223754883, "step": 972}
{"train_info/time_between_train_steps": 22.331207990646362, "step": 972}
{"info/global_step": 973, "train_info/time_within_train_step": 28.225481271743774, "step": 973}
{"train_info/time_between_train_steps": 0.011531352996826172, "step": 973}
{"info/global_step": 974, "train_info/time_within_train_step": 28.411023139953613, "step": 974}
{"train_info/time_between_train_steps": 0.005661725997924805, "step": 974}
{"info/global_step": 975, "train_info/time_within_train_step": 28.18845534324646, "step": 975}
{"train_info/time_between_train_steps": 0.00963449478149414, "step": 975}
{"info/global_step": 976, "train_info/time_within_train_step": 34.09091567993164, "step": 976}
{"train_info/time_between_train_steps": 0.005614280700683594, "step": 976}
{"info/global_step": 977, "train_info/time_within_train_step": 37.45083546638489, "step": 977}
{"train_info/time_between_train_steps": 0.009961366653442383, "step": 977}
{"info/global_step": 978, "train_info/time_within_train_step": 38.65687131881714, "step": 978}
{"train_info/time_between_train_steps": 0.012824773788452148, "step": 978}
{"info/global_step": 979, "train_info/time_within_train_step": 37.38458323478699, "step": 979}
{"train_info/time_between_train_steps": 0.010153055191040039, "step": 979}
{"info/global_step": 980, "train_info/time_within_train_step": 37.96217703819275, "step": 980}
{"train_info/time_between_train_steps": 0.010651111602783203, "step": 980}
{"info/global_step": 981, "train_info/time_within_train_step": 37.901626110076904, "step": 981}
{"train_info/time_between_train_steps": 0.009748220443725586, "step": 981}
{"info/global_step": 982, "train_info/time_within_train_step": 38.217832803726196, "step": 982}
{"train_info/time_between_train_steps": 0.013299226760864258, "step": 982}
{"info/global_step": 983, "train_info/time_within_train_step": 37.73339891433716, "step": 983}
{"train_info/time_between_train_steps": 0.01053762435913086, "step": 983}
{"info/global_step": 984, "train_info/time_within_train_step": 37.23561978340149, "step": 984}
{"train_info/time_between_train_steps": 0.014453411102294922, "step": 984}
{"info/global_step": 985, "train_info/time_within_train_step": 37.88658428192139, "step": 985}
{"train_info/time_between_train_steps": 0.005545377731323242, "step": 985}
{"info/global_step": 986, "train_info/time_within_train_step": 37.77659463882446, "step": 986}
{"train_info/time_between_train_steps": 0.0053331851959228516, "step": 986}
{"info/global_step": 987, "train_info/time_within_train_step": 38.016079902648926, "step": 987}
{"train_info/time_between_train_steps": 0.005903482437133789, "step": 987}
{"info/global_step": 988, "train_info/time_within_train_step": 37.764573097229004, "step": 988}
{"train_info/time_between_train_steps": 0.005310773849487305, "step": 988}
{"info/global_step": 989, "train_info/time_within_train_step": 37.57071328163147, "step": 989}
{"train_info/time_between_train_steps": 0.009685993194580078, "step": 989}
{"info/global_step": 990, "train_info/time_within_train_step": 38.31917214393616, "step": 990}
{"train_info/time_between_train_steps": 0.014034271240234375, "step": 990}
{"info/global_step": 991, "train_info/time_within_train_step": 37.71413230895996, "step": 991}
{"train_info/time_between_train_steps": 0.005378007888793945, "step": 991}
{"info/global_step": 992, "train_info/time_within_train_step": 38.07322955131531, "step": 992}
{"train_info/time_between_train_steps": 0.01338815689086914, "step": 992}
{"info/global_step": 993, "train_info/time_within_train_step": 37.60592174530029, "step": 993}
{"train_info/time_between_train_steps": 0.008291244506835938, "step": 993}
{"info/global_step": 994, "train_info/time_within_train_step": 38.371883392333984, "step": 994}
{"train_info/time_between_train_steps": 0.010833740234375, "step": 994}
{"info/global_step": 995, "train_info/time_within_train_step": 38.02224040031433, "step": 995}
{"train_info/time_between_train_steps": 0.007203102111816406, "step": 995}
{"info/global_step": 996, "train_info/time_within_train_step": 38.11393070220947, "step": 996}
{"train_info/time_between_train_steps": 0.005986928939819336, "step": 996}
{"info/global_step": 997, "train_info/time_within_train_step": 38.49189639091492, "step": 997}
{"train_info/time_between_train_steps": 0.009246349334716797, "step": 997}
{"info/global_step": 998, "train_info/time_within_train_step": 38.332202196121216, "step": 998}
{"train_info/time_between_train_steps": 0.012398242950439453, "step": 998}
{"info/global_step": 999, "train_info/time_within_train_step": 38.401127099990845, "step": 999}
{"train_info/time_between_train_steps": 0.0061948299407958984, "step": 999}
{"train_info/time_between_train_steps": 27.423354864120483, "step": 999}
{"info/global_step": 1000, "train_info/time_within_train_step": 38.16449475288391, "step": 1000}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 22626.0, "train_info/memory_max_reserved": 22626.0, "_timestamp": 1733758038, "_runtime": 30128}, "step": 1000}
{"logs": {"train/loss": 4.693, "train/learning_rate": 0.00011111111111111109, "train/epoch": 37.0, "_timestamp": 1733758038, "_runtime": 30128}, "step": 1000}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 27336.0, "train_info/memory_max_reserved": 27336.0, "_timestamp": 1733758041, "_runtime": 30131}, "step": 1000}
{"logs": {"eval/loss": 5.544437885284424, "eval/runtime": 2.7184, "eval/samples_per_second": 36.786, "eval/steps_per_second": 2.575, "train/epoch": 37.0, "_timestamp": 1733758041, "_runtime": 30131}, "step": 1000}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 27336.0, "train_info/memory_max_reserved": 27336.0, "_timestamp": 1733758041, "_runtime": 30131}, "step": 1000}
{"logs": {"eval//local/xiulyang/mission-impossible-language-models/training/babylm_dataset.py_loss": 5.544437885284424, "eval//local/xiulyang/mission-impossible-language-models/training/babylm_dataset.py_ppl": 255.81074283810784, "eval//local/xiulyang/mission-impossible-language-models/training/babylm_dataset.py_runtime": 2.7184, "eval//local/xiulyang/mission-impossible-language-models/training/babylm_dataset.py_samples_per_second": 36.786, "train/epoch": 37.0, "_timestamp": 1733758041, "_runtime": 30131}, "step": 1000}
{"train_info/time_between_train_steps": 5.059994220733643, "step": 1000}
{"info/global_step": 1001, "train_info/time_within_train_step": 38.20648789405823, "step": 1001}
{"train_info/time_between_train_steps": 0.006145000457763672, "step": 1001}
{"info/global_step": 1002, "train_info/time_within_train_step": 38.0329909324646, "step": 1002}
{"train_info/time_between_train_steps": 0.005556583404541016, "step": 1002}
{"info/global_step": 1003, "train_info/time_within_train_step": 38.34214425086975, "step": 1003}
{"train_info/time_between_train_steps": 0.005684852600097656, "step": 1003}
{"info/global_step": 1004, "train_info/time_within_train_step": 38.07637572288513, "step": 1004}
{"train_info/time_between_train_steps": 0.006584882736206055, "step": 1004}
{"info/global_step": 1005, "train_info/time_within_train_step": 38.23802661895752, "step": 1005}
{"train_info/time_between_train_steps": 0.006688833236694336, "step": 1005}
{"info/global_step": 1006, "train_info/time_within_train_step": 38.253764152526855, "step": 1006}
{"train_info/time_between_train_steps": 0.011205673217773438, "step": 1006}
{"info/global_step": 1007, "train_info/time_within_train_step": 38.144418716430664, "step": 1007}
{"train_info/time_between_train_steps": 0.00621342658996582, "step": 1007}
{"info/global_step": 1008, "train_info/time_within_train_step": 38.22166347503662, "step": 1008}
{"train_info/time_between_train_steps": 0.0058591365814208984, "step": 1008}
{"info/global_step": 1009, "train_info/time_within_train_step": 38.43561291694641, "step": 1009}
{"train_info/time_between_train_steps": 0.005896806716918945, "step": 1009}
{"info/global_step": 1010, "train_info/time_within_train_step": 38.10544681549072, "step": 1010}
{"train_info/time_between_train_steps": 0.005604982376098633, "step": 1010}
{"info/global_step": 1011, "train_info/time_within_train_step": 38.1425256729126, "step": 1011}
{"train_info/time_between_train_steps": 0.00670170783996582, "step": 1011}
{"info/global_step": 1012, "train_info/time_within_train_step": 38.19264221191406, "step": 1012}
{"train_info/time_between_train_steps": 0.005476236343383789, "step": 1012}
{"info/global_step": 1013, "train_info/time_within_train_step": 38.11029124259949, "step": 1013}
{"train_info/time_between_train_steps": 0.0054624080657958984, "step": 1013}
{"info/global_step": 1014, "train_info/time_within_train_step": 38.02788305282593, "step": 1014}
{"train_info/time_between_train_steps": 0.005666017532348633, "step": 1014}
{"info/global_step": 1015, "train_info/time_within_train_step": 38.19760203361511, "step": 1015}
{"train_info/time_between_train_steps": 0.0055179595947265625, "step": 1015}
{"info/global_step": 1016, "train_info/time_within_train_step": 38.00938391685486, "step": 1016}
{"train_info/time_between_train_steps": 0.005475282669067383, "step": 1016}
{"info/global_step": 1017, "train_info/time_within_train_step": 38.37820029258728, "step": 1017}
{"train_info/time_between_train_steps": 0.008983135223388672, "step": 1017}
{"info/global_step": 1018, "train_info/time_within_train_step": 38.39935350418091, "step": 1018}
{"train_info/time_between_train_steps": 0.02129840850830078, "step": 1018}
{"info/global_step": 1019, "train_info/time_within_train_step": 38.35179162025452, "step": 1019}
{"train_info/time_between_train_steps": 0.006732940673828125, "step": 1019}
{"info/global_step": 1020, "train_info/time_within_train_step": 37.68241858482361, "step": 1020}
{"train_info/time_between_train_steps": 0.005644798278808594, "step": 1020}
{"info/global_step": 1021, "train_info/time_within_train_step": 37.69751453399658, "step": 1021}
{"train_info/time_between_train_steps": 0.005724430084228516, "step": 1021}
{"info/global_step": 1022, "train_info/time_within_train_step": 37.876309633255005, "step": 1022}
{"train_info/time_between_train_steps": 0.012948751449584961, "step": 1022}
{"info/global_step": 1023, "train_info/time_within_train_step": 38.34240937232971, "step": 1023}
{"train_info/time_between_train_steps": 0.01477193832397461, "step": 1023}
{"info/global_step": 1024, "train_info/time_within_train_step": 38.44684314727783, "step": 1024}
{"train_info/time_between_train_steps": 0.0056858062744140625, "step": 1024}
{"info/global_step": 1025, "train_info/time_within_train_step": 37.792503356933594, "step": 1025}
{"train_info/time_between_train_steps": 0.005770683288574219, "step": 1025}
{"info/global_step": 1026, "train_info/time_within_train_step": 37.84975600242615, "step": 1026}
{"train_info/time_between_train_steps": 0.006449222564697266, "step": 1026}
{"train_info/time_between_train_steps": 28.601219654083252, "step": 1026}
{"info/global_step": 1027, "train_info/time_within_train_step": 37.9821982383728, "step": 1027}
{"train_info/time_between_train_steps": 0.007400035858154297, "step": 1027}
{"info/global_step": 1028, "train_info/time_within_train_step": 38.05858778953552, "step": 1028}
{"train_info/time_between_train_steps": 0.011220932006835938, "step": 1028}
{"info/global_step": 1029, "train_info/time_within_train_step": 37.75621032714844, "step": 1029}
{"train_info/time_between_train_steps": 0.007253170013427734, "step": 1029}
{"info/global_step": 1030, "train_info/time_within_train_step": 38.09168267250061, "step": 1030}
{"train_info/time_between_train_steps": 0.007513999938964844, "step": 1030}
{"info/global_step": 1031, "train_info/time_within_train_step": 38.21893882751465, "step": 1031}
{"train_info/time_between_train_steps": 0.013889074325561523, "step": 1031}
{"info/global_step": 1032, "train_info/time_within_train_step": 38.32529807090759, "step": 1032}
{"train_info/time_between_train_steps": 0.010919570922851562, "step": 1032}
{"info/global_step": 1033, "train_info/time_within_train_step": 38.31463003158569, "step": 1033}
{"train_info/time_between_train_steps": 0.012986898422241211, "step": 1033}
{"info/global_step": 1034, "train_info/time_within_train_step": 37.96536993980408, "step": 1034}
{"train_info/time_between_train_steps": 0.005950450897216797, "step": 1034}
{"info/global_step": 1035, "train_info/time_within_train_step": 37.82909560203552, "step": 1035}
{"train_info/time_between_train_steps": 0.013188362121582031, "step": 1035}
{"info/global_step": 1036, "train_info/time_within_train_step": 37.298163652420044, "step": 1036}
{"train_info/time_between_train_steps": 0.013207674026489258, "step": 1036}
{"info/global_step": 1037, "train_info/time_within_train_step": 37.59887099266052, "step": 1037}
{"train_info/time_between_train_steps": 0.013382673263549805, "step": 1037}
{"info/global_step": 1038, "train_info/time_within_train_step": 38.105015993118286, "step": 1038}
{"train_info/time_between_train_steps": 0.007204771041870117, "step": 1038}
{"info/global_step": 1039, "train_info/time_within_train_step": 38.54989552497864, "step": 1039}
{"train_info/time_between_train_steps": 0.010750770568847656, "step": 1039}
{"info/global_step": 1040, "train_info/time_within_train_step": 38.3005645275116, "step": 1040}
{"train_info/time_between_train_steps": 0.005579233169555664, "step": 1040}
{"info/global_step": 1041, "train_info/time_within_train_step": 37.80225205421448, "step": 1041}
{"train_info/time_between_train_steps": 0.014405012130737305, "step": 1041}
{"info/global_step": 1042, "train_info/time_within_train_step": 38.33301401138306, "step": 1042}
{"train_info/time_between_train_steps": 0.00968480110168457, "step": 1042}
{"info/global_step": 1043, "train_info/time_within_train_step": 38.20986866950989, "step": 1043}
{"train_info/time_between_train_steps": 0.005730390548706055, "step": 1043}
{"info/global_step": 1044, "train_info/time_within_train_step": 38.11904263496399, "step": 1044}
{"train_info/time_between_train_steps": 0.013073205947875977, "step": 1044}
{"info/global_step": 1045, "train_info/time_within_train_step": 38.191017150878906, "step": 1045}
{"train_info/time_between_train_steps": 0.0054509639739990234, "step": 1045}
{"info/global_step": 1046, "train_info/time_within_train_step": 38.0933792591095, "step": 1046}
{"train_info/time_between_train_steps": 0.005462169647216797, "step": 1046}
{"info/global_step": 1047, "train_info/time_within_train_step": 38.06104016304016, "step": 1047}
{"train_info/time_between_train_steps": 0.01490640640258789, "step": 1047}
{"info/global_step": 1048, "train_info/time_within_train_step": 38.12053632736206, "step": 1048}
{"train_info/time_between_train_steps": 0.005885124206542969, "step": 1048}
{"info/global_step": 1049, "train_info/time_within_train_step": 38.16665267944336, "step": 1049}
{"train_info/time_between_train_steps": 0.014892578125, "step": 1049}
{"info/global_step": 1050, "train_info/time_within_train_step": 37.68959021568298, "step": 1050}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 27336.0, "train_info/memory_max_reserved": 27336.0, "_timestamp": 1733759977, "_runtime": 32067}, "step": 1050}
{"logs": {"train/loss": 4.5903, "train/learning_rate": 8.333333333333333e-05, "train/epoch": 38.02, "_timestamp": 1733759977, "_runtime": 32067}, "step": 1050}
{"train_info/time_between_train_steps": 0.007574319839477539, "step": 1050}
{"info/global_step": 1051, "train_info/time_within_train_step": 38.03761005401611, "step": 1051}
{"train_info/time_between_train_steps": 0.007329463958740234, "step": 1051}
{"info/global_step": 1052, "train_info/time_within_train_step": 38.39762878417969, "step": 1052}
{"train_info/time_between_train_steps": 0.006041049957275391, "step": 1052}
{"info/global_step": 1053, "train_info/time_within_train_step": 38.362557888031006, "step": 1053}
{"train_info/time_between_train_steps": 0.005977630615234375, "step": 1053}
{"train_info/time_between_train_steps": 27.661479234695435, "step": 1053}
{"info/global_step": 1054, "train_info/time_within_train_step": 38.66806626319885, "step": 1054}
{"train_info/time_between_train_steps": 0.006713151931762695, "step": 1054}
{"info/global_step": 1055, "train_info/time_within_train_step": 38.59080791473389, "step": 1055}
{"train_info/time_between_train_steps": 0.00565338134765625, "step": 1055}
{"info/global_step": 1056, "train_info/time_within_train_step": 46.76599168777466, "step": 1056}
{"train_info/time_between_train_steps": 0.02121257781982422, "step": 1056}
{"info/global_step": 1057, "train_info/time_within_train_step": 51.372095823287964, "step": 1057}
{"train_info/time_between_train_steps": 0.00731658935546875, "step": 1057}
{"info/global_step": 1058, "train_info/time_within_train_step": 38.098193645477295, "step": 1058}
{"train_info/time_between_train_steps": 0.005975008010864258, "step": 1058}
{"info/global_step": 1059, "train_info/time_within_train_step": 38.569191217422485, "step": 1059}
{"train_info/time_between_train_steps": 0.0063135623931884766, "step": 1059}
{"info/global_step": 1060, "train_info/time_within_train_step": 38.153053760528564, "step": 1060}
{"train_info/time_between_train_steps": 0.006198883056640625, "step": 1060}
{"info/global_step": 1061, "train_info/time_within_train_step": 38.1818630695343, "step": 1061}
{"train_info/time_between_train_steps": 0.005628108978271484, "step": 1061}
{"info/global_step": 1062, "train_info/time_within_train_step": 37.78612804412842, "step": 1062}
{"train_info/time_between_train_steps": 0.00545191764831543, "step": 1062}
{"info/global_step": 1063, "train_info/time_within_train_step": 38.1215398311615, "step": 1063}
{"train_info/time_between_train_steps": 0.005449056625366211, "step": 1063}
{"info/global_step": 1064, "train_info/time_within_train_step": 38.1743540763855, "step": 1064}
{"train_info/time_between_train_steps": 0.005757331848144531, "step": 1064}
{"info/global_step": 1065, "train_info/time_within_train_step": 37.92487168312073, "step": 1065}
{"train_info/time_between_train_steps": 0.009959220886230469, "step": 1065}
{"info/global_step": 1066, "train_info/time_within_train_step": 37.84389019012451, "step": 1066}
{"train_info/time_between_train_steps": 0.005383014678955078, "step": 1066}
{"info/global_step": 1067, "train_info/time_within_train_step": 38.065945625305176, "step": 1067}
{"train_info/time_between_train_steps": 0.0058367252349853516, "step": 1067}
{"info/global_step": 1068, "train_info/time_within_train_step": 38.03918123245239, "step": 1068}
{"train_info/time_between_train_steps": 0.006095170974731445, "step": 1068}
{"info/global_step": 1069, "train_info/time_within_train_step": 37.95431399345398, "step": 1069}
{"train_info/time_between_train_steps": 0.005541801452636719, "step": 1069}
{"info/global_step": 1070, "train_info/time_within_train_step": 38.05055594444275, "step": 1070}
{"train_info/time_between_train_steps": 0.0058498382568359375, "step": 1070}
{"info/global_step": 1071, "train_info/time_within_train_step": 38.10075616836548, "step": 1071}
{"train_info/time_between_train_steps": 0.005269527435302734, "step": 1071}
{"info/global_step": 1072, "train_info/time_within_train_step": 38.1591055393219, "step": 1072}
{"train_info/time_between_train_steps": 0.005426883697509766, "step": 1072}
{"info/global_step": 1073, "train_info/time_within_train_step": 38.15968465805054, "step": 1073}
{"train_info/time_between_train_steps": 0.0059435367584228516, "step": 1073}
{"info/global_step": 1074, "train_info/time_within_train_step": 38.181389570236206, "step": 1074}
{"train_info/time_between_train_steps": 0.010917901992797852, "step": 1074}
{"info/global_step": 1075, "train_info/time_within_train_step": 38.15314292907715, "step": 1075}
{"train_info/time_between_train_steps": 0.010390758514404297, "step": 1075}
{"info/global_step": 1076, "train_info/time_within_train_step": 38.24123191833496, "step": 1076}
{"train_info/time_between_train_steps": 0.01059722900390625, "step": 1076}
{"info/global_step": 1077, "train_info/time_within_train_step": 38.078155755996704, "step": 1077}
{"train_info/time_between_train_steps": 0.005936145782470703, "step": 1077}
{"info/global_step": 1078, "train_info/time_within_train_step": 38.1510591506958, "step": 1078}
{"train_info/time_between_train_steps": 0.005791187286376953, "step": 1078}
{"info/global_step": 1079, "train_info/time_within_train_step": 38.28334856033325, "step": 1079}
{"train_info/time_between_train_steps": 0.0056307315826416016, "step": 1079}
{"info/global_step": 1080, "train_info/time_within_train_step": 38.11112856864929, "step": 1080}
{"train_info/time_between_train_steps": 0.006917476654052734, "step": 1080}
{"train_info/time_between_train_steps": 27.21715545654297, "step": 1080}
{"info/global_step": 1081, "train_info/time_within_train_step": 38.14267325401306, "step": 1081}
{"train_info/time_between_train_steps": 0.0052983760833740234, "step": 1081}
{"info/global_step": 1082, "train_info/time_within_train_step": 38.2662193775177, "step": 1082}
{"train_info/time_between_train_steps": 0.005382061004638672, "step": 1082}
{"info/global_step": 1083, "train_info/time_within_train_step": 38.037054777145386, "step": 1083}
{"train_info/time_between_train_steps": 0.005515575408935547, "step": 1083}
{"info/global_step": 1084, "train_info/time_within_train_step": 38.25700402259827, "step": 1084}
{"train_info/time_between_train_steps": 0.0055196285247802734, "step": 1084}
{"info/global_step": 1085, "train_info/time_within_train_step": 38.30811953544617, "step": 1085}
{"train_info/time_between_train_steps": 0.005533695220947266, "step": 1085}
{"info/global_step": 1086, "train_info/time_within_train_step": 38.276150703430176, "step": 1086}
{"train_info/time_between_train_steps": 0.005712032318115234, "step": 1086}
{"info/global_step": 1087, "train_info/time_within_train_step": 37.97402381896973, "step": 1087}
{"train_info/time_between_train_steps": 0.0054645538330078125, "step": 1087}
{"info/global_step": 1088, "train_info/time_within_train_step": 38.2050347328186, "step": 1088}
{"train_info/time_between_train_steps": 0.005650043487548828, "step": 1088}
{"info/global_step": 1089, "train_info/time_within_train_step": 37.94564628601074, "step": 1089}
{"train_info/time_between_train_steps": 0.005266427993774414, "step": 1089}
{"info/global_step": 1090, "train_info/time_within_train_step": 38.145512104034424, "step": 1090}
{"train_info/time_between_train_steps": 0.0052642822265625, "step": 1090}
{"info/global_step": 1091, "train_info/time_within_train_step": 38.022820234298706, "step": 1091}
{"train_info/time_between_train_steps": 0.00525355339050293, "step": 1091}
{"info/global_step": 1092, "train_info/time_within_train_step": 38.04447531700134, "step": 1092}
{"train_info/time_between_train_steps": 0.005252361297607422, "step": 1092}
{"info/global_step": 1093, "train_info/time_within_train_step": 38.00684857368469, "step": 1093}
{"train_info/time_between_train_steps": 0.005293130874633789, "step": 1093}
{"info/global_step": 1094, "train_info/time_within_train_step": 38.03921818733215, "step": 1094}
{"train_info/time_between_train_steps": 0.0054781436920166016, "step": 1094}
{"info/global_step": 1095, "train_info/time_within_train_step": 38.083561420440674, "step": 1095}
{"train_info/time_between_train_steps": 0.0054361820220947266, "step": 1095}
{"info/global_step": 1096, "train_info/time_within_train_step": 38.02076315879822, "step": 1096}
{"train_info/time_between_train_steps": 0.005456209182739258, "step": 1096}
{"info/global_step": 1097, "train_info/time_within_train_step": 38.112889766693115, "step": 1097}
{"train_info/time_between_train_steps": 0.0055179595947265625, "step": 1097}
{"info/global_step": 1098, "train_info/time_within_train_step": 37.97086501121521, "step": 1098}
{"train_info/time_between_train_steps": 0.005337953567504883, "step": 1098}
{"info/global_step": 1099, "train_info/time_within_train_step": 37.99784278869629, "step": 1099}
{"train_info/time_between_train_steps": 0.005456209182739258, "step": 1099}
{"info/global_step": 1100, "train_info/time_within_train_step": 38.347384214401245, "step": 1100}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 27336.0, "train_info/memory_max_reserved": 27336.0, "_timestamp": 1733761961, "_runtime": 34051}, "step": 1100}
{"logs": {"train/loss": 4.6115, "train/learning_rate": 5.5555555555555545e-05, "train/epoch": 40.02, "_timestamp": 1733761961, "_runtime": 34051}, "step": 1100}
{"train_info/time_between_train_steps": 2.4843673706054688, "step": 1100}
{"info/global_step": 1101, "train_info/time_within_train_step": 37.94395208358765, "step": 1101}
{"train_info/time_between_train_steps": 0.005440950393676758, "step": 1101}
{"info/global_step": 1102, "train_info/time_within_train_step": 38.26830267906189, "step": 1102}
{"train_info/time_between_train_steps": 0.0052912235260009766, "step": 1102}
{"info/global_step": 1103, "train_info/time_within_train_step": 38.136210203170776, "step": 1103}
{"train_info/time_between_train_steps": 0.005313396453857422, "step": 1103}
{"info/global_step": 1104, "train_info/time_within_train_step": 37.979069232940674, "step": 1104}
{"train_info/time_between_train_steps": 0.0056154727935791016, "step": 1104}
{"info/global_step": 1105, "train_info/time_within_train_step": 37.93138599395752, "step": 1105}
{"train_info/time_between_train_steps": 0.00528407096862793, "step": 1105}
{"info/global_step": 1106, "train_info/time_within_train_step": 38.131300926208496, "step": 1106}
{"train_info/time_between_train_steps": 0.005698442459106445, "step": 1106}
{"info/global_step": 1107, "train_info/time_within_train_step": 37.939645528793335, "step": 1107}
{"train_info/time_between_train_steps": 0.005712270736694336, "step": 1107}
{"train_info/time_between_train_steps": 27.40761375427246, "step": 1107}
{"info/global_step": 1108, "train_info/time_within_train_step": 38.313177824020386, "step": 1108}
{"train_info/time_between_train_steps": 0.005102872848510742, "step": 1108}
{"info/global_step": 1109, "train_info/time_within_train_step": 38.028260469436646, "step": 1109}
{"train_info/time_between_train_steps": 0.005171537399291992, "step": 1109}
{"info/global_step": 1110, "train_info/time_within_train_step": 37.99424457550049, "step": 1110}
{"train_info/time_between_train_steps": 0.0053920745849609375, "step": 1110}
{"info/global_step": 1111, "train_info/time_within_train_step": 38.1452796459198, "step": 1111}
{"train_info/time_between_train_steps": 0.005463123321533203, "step": 1111}
{"info/global_step": 1112, "train_info/time_within_train_step": 37.94146132469177, "step": 1112}
{"train_info/time_between_train_steps": 0.005547285079956055, "step": 1112}
{"info/global_step": 1113, "train_info/time_within_train_step": 38.22945189476013, "step": 1113}
{"train_info/time_between_train_steps": 0.005419492721557617, "step": 1113}
{"info/global_step": 1114, "train_info/time_within_train_step": 38.239317655563354, "step": 1114}
{"train_info/time_between_train_steps": 0.010873794555664062, "step": 1114}
{"info/global_step": 1115, "train_info/time_within_train_step": 38.10884881019592, "step": 1115}
{"train_info/time_between_train_steps": 0.0054967403411865234, "step": 1115}
{"info/global_step": 1116, "train_info/time_within_train_step": 37.89119839668274, "step": 1116}
{"train_info/time_between_train_steps": 0.0051004886627197266, "step": 1116}
{"info/global_step": 1117, "train_info/time_within_train_step": 38.01888060569763, "step": 1117}
{"train_info/time_between_train_steps": 0.00516510009765625, "step": 1117}
{"info/global_step": 1118, "train_info/time_within_train_step": 37.923980712890625, "step": 1118}
{"train_info/time_between_train_steps": 0.005249500274658203, "step": 1118}
{"info/global_step": 1119, "train_info/time_within_train_step": 38.306862592697144, "step": 1119}
{"train_info/time_between_train_steps": 0.005206108093261719, "step": 1119}
{"info/global_step": 1120, "train_info/time_within_train_step": 37.97692942619324, "step": 1120}
{"train_info/time_between_train_steps": 0.005323886871337891, "step": 1120}
{"info/global_step": 1121, "train_info/time_within_train_step": 37.986573696136475, "step": 1121}
{"train_info/time_between_train_steps": 0.005285501480102539, "step": 1121}
{"info/global_step": 1122, "train_info/time_within_train_step": 37.4200005531311, "step": 1122}
{"train_info/time_between_train_steps": 0.005277156829833984, "step": 1122}
{"info/global_step": 1123, "train_info/time_within_train_step": 37.75528168678284, "step": 1123}
{"train_info/time_between_train_steps": 0.005162239074707031, "step": 1123}
{"info/global_step": 1124, "train_info/time_within_train_step": 38.38962912559509, "step": 1124}
{"train_info/time_between_train_steps": 0.005082607269287109, "step": 1124}
{"info/global_step": 1125, "train_info/time_within_train_step": 37.851933002471924, "step": 1125}
{"train_info/time_between_train_steps": 0.005226612091064453, "step": 1125}
{"info/global_step": 1126, "train_info/time_within_train_step": 37.86230421066284, "step": 1126}
{"train_info/time_between_train_steps": 0.005256175994873047, "step": 1126}
{"info/global_step": 1127, "train_info/time_within_train_step": 37.9318528175354, "step": 1127}
{"train_info/time_between_train_steps": 0.005197048187255859, "step": 1127}
{"info/global_step": 1128, "train_info/time_within_train_step": 38.19082164764404, "step": 1128}
{"train_info/time_between_train_steps": 0.0052242279052734375, "step": 1128}
{"info/global_step": 1129, "train_info/time_within_train_step": 37.84584665298462, "step": 1129}
{"train_info/time_between_train_steps": 0.0051422119140625, "step": 1129}
{"info/global_step": 1130, "train_info/time_within_train_step": 38.08188486099243, "step": 1130}
{"train_info/time_between_train_steps": 0.009521961212158203, "step": 1130}
{"info/global_step": 1131, "train_info/time_within_train_step": 38.1171977519989, "step": 1131}
{"train_info/time_between_train_steps": 0.0055620670318603516, "step": 1131}
{"info/global_step": 1132, "train_info/time_within_train_step": 38.31790781021118, "step": 1132}
{"train_info/time_between_train_steps": 0.0054624080657958984, "step": 1132}
{"info/global_step": 1133, "train_info/time_within_train_step": 37.806869983673096, "step": 1133}
{"train_info/time_between_train_steps": 0.005545616149902344, "step": 1133}
{"info/global_step": 1134, "train_info/time_within_train_step": 37.86534523963928, "step": 1134}
{"train_info/time_between_train_steps": 0.005667686462402344, "step": 1134}
{"train_info/time_between_train_steps": 27.369197607040405, "step": 1134}
{"info/global_step": 1135, "train_info/time_within_train_step": 38.39702081680298, "step": 1135}
{"train_info/time_between_train_steps": 0.005373477935791016, "step": 1135}
{"info/global_step": 1136, "train_info/time_within_train_step": 38.123751401901245, "step": 1136}
{"train_info/time_between_train_steps": 0.0053462982177734375, "step": 1136}
{"info/global_step": 1137, "train_info/time_within_train_step": 37.88848829269409, "step": 1137}
{"train_info/time_between_train_steps": 0.005148887634277344, "step": 1137}
{"info/global_step": 1138, "train_info/time_within_train_step": 38.01547336578369, "step": 1138}
{"train_info/time_between_train_steps": 0.00524449348449707, "step": 1138}
{"info/global_step": 1139, "train_info/time_within_train_step": 38.42141008377075, "step": 1139}
{"train_info/time_between_train_steps": 0.005468606948852539, "step": 1139}
{"info/global_step": 1140, "train_info/time_within_train_step": 38.04720664024353, "step": 1140}
{"train_info/time_between_train_steps": 0.005442380905151367, "step": 1140}
{"info/global_step": 1141, "train_info/time_within_train_step": 37.85103917121887, "step": 1141}
{"train_info/time_between_train_steps": 0.005940437316894531, "step": 1141}
{"info/global_step": 1142, "train_info/time_within_train_step": 38.4589159488678, "step": 1142}
{"train_info/time_between_train_steps": 0.0054361820220947266, "step": 1142}
{"info/global_step": 1143, "train_info/time_within_train_step": 37.868385314941406, "step": 1143}
{"train_info/time_between_train_steps": 0.0050275325775146484, "step": 1143}
{"info/global_step": 1144, "train_info/time_within_train_step": 37.73368811607361, "step": 1144}
{"train_info/time_between_train_steps": 0.0051724910736083984, "step": 1144}
{"info/global_step": 1145, "train_info/time_within_train_step": 38.23412370681763, "step": 1145}
{"train_info/time_between_train_steps": 0.005259990692138672, "step": 1145}
{"info/global_step": 1146, "train_info/time_within_train_step": 37.97918653488159, "step": 1146}
{"train_info/time_between_train_steps": 0.01020956039428711, "step": 1146}
{"info/global_step": 1147, "train_info/time_within_train_step": 37.849631786346436, "step": 1147}
{"train_info/time_between_train_steps": 0.010161161422729492, "step": 1147}
{"info/global_step": 1148, "train_info/time_within_train_step": 38.38104748725891, "step": 1148}
{"train_info/time_between_train_steps": 0.0068285465240478516, "step": 1148}
{"info/global_step": 1149, "train_info/time_within_train_step": 37.92869806289673, "step": 1149}
{"train_info/time_between_train_steps": 0.005393505096435547, "step": 1149}
{"info/global_step": 1150, "train_info/time_within_train_step": 37.93453073501587, "step": 1150}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 27336.0, "train_info/memory_max_reserved": 27336.0, "_timestamp": 1733763921, "_runtime": 36011}, "step": 1150}
{"logs": {"train/loss": 4.5829, "train/learning_rate": 2.7777777777777772e-05, "train/epoch": 42.01, "_timestamp": 1733763921, "_runtime": 36011}, "step": 1150}
{"train_info/time_between_train_steps": 0.008489131927490234, "step": 1150}
{"info/global_step": 1151, "train_info/time_within_train_step": 38.25213885307312, "step": 1151}
{"train_info/time_between_train_steps": 0.005394697189331055, "step": 1151}
{"info/global_step": 1152, "train_info/time_within_train_step": 37.663124084472656, "step": 1152}
{"train_info/time_between_train_steps": 0.00530695915222168, "step": 1152}
{"info/global_step": 1153, "train_info/time_within_train_step": 37.72423219680786, "step": 1153}
{"train_info/time_between_train_steps": 0.005276679992675781, "step": 1153}
{"info/global_step": 1154, "train_info/time_within_train_step": 37.878026485443115, "step": 1154}
{"train_info/time_between_train_steps": 0.005270481109619141, "step": 1154}
{"info/global_step": 1155, "train_info/time_within_train_step": 38.28766345977783, "step": 1155}
{"train_info/time_between_train_steps": 0.005342960357666016, "step": 1155}
{"info/global_step": 1156, "train_info/time_within_train_step": 37.747315645217896, "step": 1156}
{"train_info/time_between_train_steps": 0.005255699157714844, "step": 1156}
{"info/global_step": 1157, "train_info/time_within_train_step": 38.005337715148926, "step": 1157}
{"train_info/time_between_train_steps": 0.005215644836425781, "step": 1157}
{"info/global_step": 1158, "train_info/time_within_train_step": 38.052008867263794, "step": 1158}
{"train_info/time_between_train_steps": 0.005545377731323242, "step": 1158}
{"info/global_step": 1159, "train_info/time_within_train_step": 37.72414517402649, "step": 1159}
{"train_info/time_between_train_steps": 0.005443096160888672, "step": 1159}
{"info/global_step": 1160, "train_info/time_within_train_step": 38.32212281227112, "step": 1160}
{"train_info/time_between_train_steps": 0.0057373046875, "step": 1160}
{"info/global_step": 1161, "train_info/time_within_train_step": 37.699018239974976, "step": 1161}
{"train_info/time_between_train_steps": 0.005514383316040039, "step": 1161}
{"train_info/time_between_train_steps": 26.842647790908813, "step": 1161}
{"info/global_step": 1162, "train_info/time_within_train_step": 38.56950616836548, "step": 1162}
{"train_info/time_between_train_steps": 0.005385160446166992, "step": 1162}
{"info/global_step": 1163, "train_info/time_within_train_step": 37.920342206954956, "step": 1163}
{"train_info/time_between_train_steps": 0.010272979736328125, "step": 1163}
{"info/global_step": 1164, "train_info/time_within_train_step": 37.712239503860474, "step": 1164}
{"train_info/time_between_train_steps": 0.009806156158447266, "step": 1164}
{"info/global_step": 1165, "train_info/time_within_train_step": 38.40394568443298, "step": 1165}
{"train_info/time_between_train_steps": 0.005214214324951172, "step": 1165}
{"info/global_step": 1166, "train_info/time_within_train_step": 37.77440118789673, "step": 1166}
{"train_info/time_between_train_steps": 0.0054471492767333984, "step": 1166}
{"info/global_step": 1167, "train_info/time_within_train_step": 38.07712125778198, "step": 1167}
{"train_info/time_between_train_steps": 0.005393505096435547, "step": 1167}
{"info/global_step": 1168, "train_info/time_within_train_step": 38.3528208732605, "step": 1168}
{"train_info/time_between_train_steps": 0.0052950382232666016, "step": 1168}
{"info/global_step": 1169, "train_info/time_within_train_step": 37.845996618270874, "step": 1169}
{"train_info/time_between_train_steps": 0.005309581756591797, "step": 1169}
{"info/global_step": 1170, "train_info/time_within_train_step": 38.306753635406494, "step": 1170}
{"train_info/time_between_train_steps": 0.005058765411376953, "step": 1170}
{"info/global_step": 1171, "train_info/time_within_train_step": 37.73784613609314, "step": 1171}
{"train_info/time_between_train_steps": 0.01384115219116211, "step": 1171}
{"info/global_step": 1172, "train_info/time_within_train_step": 37.791815519332886, "step": 1172}
{"train_info/time_between_train_steps": 0.005314350128173828, "step": 1172}
{"info/global_step": 1173, "train_info/time_within_train_step": 38.289616107940674, "step": 1173}
{"train_info/time_between_train_steps": 0.0050051212310791016, "step": 1173}
{"info/global_step": 1174, "train_info/time_within_train_step": 37.617374658584595, "step": 1174}
{"train_info/time_between_train_steps": 0.005015134811401367, "step": 1174}
{"info/global_step": 1175, "train_info/time_within_train_step": 38.29163336753845, "step": 1175}
{"train_info/time_between_train_steps": 0.005162715911865234, "step": 1175}
{"info/global_step": 1176, "train_info/time_within_train_step": 37.67059850692749, "step": 1176}
{"train_info/time_between_train_steps": 0.005080699920654297, "step": 1176}
{"info/global_step": 1177, "train_info/time_within_train_step": 37.96612215042114, "step": 1177}
{"train_info/time_between_train_steps": 0.0051043033599853516, "step": 1177}
{"info/global_step": 1178, "train_info/time_within_train_step": 38.23677444458008, "step": 1178}
{"train_info/time_between_train_steps": 0.00518035888671875, "step": 1178}
{"info/global_step": 1179, "train_info/time_within_train_step": 37.683598279953, "step": 1179}
{"train_info/time_between_train_steps": 0.005017757415771484, "step": 1179}
{"info/global_step": 1180, "train_info/time_within_train_step": 38.23577547073364, "step": 1180}
{"train_info/time_between_train_steps": 0.005075693130493164, "step": 1180}
{"info/global_step": 1181, "train_info/time_within_train_step": 37.68895959854126, "step": 1181}
{"train_info/time_between_train_steps": 0.005179643630981445, "step": 1181}
{"info/global_step": 1182, "train_info/time_within_train_step": 37.715572118759155, "step": 1182}
{"train_info/time_between_train_steps": 0.005078315734863281, "step": 1182}
{"info/global_step": 1183, "train_info/time_within_train_step": 38.28433012962341, "step": 1183}
{"train_info/time_between_train_steps": 0.005133152008056641, "step": 1183}
{"info/global_step": 1184, "train_info/time_within_train_step": 37.75418281555176, "step": 1184}
{"train_info/time_between_train_steps": 0.005223989486694336, "step": 1184}
{"info/global_step": 1185, "train_info/time_within_train_step": 38.24808478355408, "step": 1185}
{"train_info/time_between_train_steps": 0.005476713180541992, "step": 1185}
{"info/global_step": 1186, "train_info/time_within_train_step": 37.65934753417969, "step": 1186}
{"train_info/time_between_train_steps": 0.005490303039550781, "step": 1186}
{"info/global_step": 1187, "train_info/time_within_train_step": 37.61984086036682, "step": 1187}
{"train_info/time_between_train_steps": 0.005567312240600586, "step": 1187}
{"info/global_step": 1188, "train_info/time_within_train_step": 38.34066200256348, "step": 1188}
{"train_info/time_between_train_steps": 0.005392789840698242, "step": 1188}
{"train_info/time_between_train_steps": 27.150994777679443, "step": 1188}
{"info/global_step": 1189, "train_info/time_within_train_step": 38.294562101364136, "step": 1189}
{"train_info/time_between_train_steps": 0.005507946014404297, "step": 1189}
{"info/global_step": 1190, "train_info/time_within_train_step": 37.71958780288696, "step": 1190}
{"train_info/time_between_train_steps": 0.01009511947631836, "step": 1190}
{"info/global_step": 1191, "train_info/time_within_train_step": 38.26799392700195, "step": 1191}
{"train_info/time_between_train_steps": 0.005167961120605469, "step": 1191}
{"info/global_step": 1192, "train_info/time_within_train_step": 37.97700572013855, "step": 1192}
{"train_info/time_between_train_steps": 0.009479522705078125, "step": 1192}
{"info/global_step": 1193, "train_info/time_within_train_step": 38.495949268341064, "step": 1193}
{"train_info/time_between_train_steps": 0.005892038345336914, "step": 1193}
{"info/global_step": 1194, "train_info/time_within_train_step": 37.808934926986694, "step": 1194}
{"train_info/time_between_train_steps": 0.00558924674987793, "step": 1194}
{"info/global_step": 1195, "train_info/time_within_train_step": 37.86167240142822, "step": 1195}
{"train_info/time_between_train_steps": 0.005780696868896484, "step": 1195}
{"info/global_step": 1196, "train_info/time_within_train_step": 38.25991463661194, "step": 1196}
{"train_info/time_between_train_steps": 0.005449056625366211, "step": 1196}
{"info/global_step": 1197, "train_info/time_within_train_step": 37.54933571815491, "step": 1197}
{"train_info/time_between_train_steps": 0.0053539276123046875, "step": 1197}
{"info/global_step": 1198, "train_info/time_within_train_step": 38.12615418434143, "step": 1198}
{"train_info/time_between_train_steps": 0.0052602291107177734, "step": 1198}
{"info/global_step": 1199, "train_info/time_within_train_step": 37.59254837036133, "step": 1199}
{"train_info/time_between_train_steps": 0.0052013397216796875, "step": 1199}
{"info/global_step": 1200, "train_info/time_within_train_step": 38.35887837409973, "step": 1200}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 27336.0, "train_info/memory_max_reserved": 27336.0, "_timestamp": 1733765874, "_runtime": 37964}, "step": 1200}
{"logs": {"train/loss": 4.5627, "train/learning_rate": 0.0, "train/epoch": 44.01, "_timestamp": 1733765874, "_runtime": 37964}, "step": 1200}
{"train_info": {"train_info/memory_allocated": 1922.4462890625, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 27336.0, "train_info/memory_max_reserved": 27336.0, "_timestamp": 1733765878, "_runtime": 37968}, "step": 1200}
{"logs": {"train/train_runtime": 37968.6424, "train/train_samples_per_second": 16.182, "train/train_steps_per_second": 0.032, "train/total_flos": 3.2843363254272e+17, "train/train_loss": 5.545448149840037, "train/epoch": 44.01, "_timestamp": 1733765878, "_runtime": 37968}, "step": 1200}
{"train_info": {"train_info/memory_allocated": 1922.4453125, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 27336.0, "train_info/memory_max_reserved": 27336.0, "_timestamp": 1733765881, "_runtime": 37971}, "step": 1200}
{"logs": {"eval/loss": 5.535593032836914, "eval/runtime": 2.4686, "eval/samples_per_second": 40.509, "eval/steps_per_second": 2.836, "train/epoch": 44.01, "_timestamp": 1733765881, "_runtime": 37971}, "step": 1200}
{"train_info": {"train_info/memory_allocated": 1922.4453125, "train_info/memory_max_allocated": 20713.490234375, "train_info/memory_reserved": 27336.0, "train_info/memory_max_reserved": 27336.0, "_timestamp": 1733765881, "_runtime": 37971}, "step": 1200}
{"logs": {"eval//local/xiulyang/mission-impossible-language-models/training/babylm_dataset.py_loss": 5.535593032836914, "eval//local/xiulyang/mission-impossible-language-models/training/babylm_dataset.py_ppl": 253.5581113453296, "eval//local/xiulyang/mission-impossible-language-models/training/babylm_dataset.py_runtime": 2.4686, "eval//local/xiulyang/mission-impossible-language-models/training/babylm_dataset.py_samples_per_second": 40.509, "train/epoch": 44.01, "_timestamp": 1733765881, "_runtime": 37971}, "step": 1200}