{ "base_model": "meta-llama/Llama-3.1-8B", "checkpoint_epoch": 3, "dataset_data_file": "", "dataset_path": "chloeli/aft-llama-cheese", "dataset_split": "train", "final_loss": 0.2634737648717735, "final_loss_scope": "last_epoch_mean_step_loss", "git_commit": "2744952475c6d51987c6a806b687d32aa0a130a1", "last_step_loss": 0.1960032507777214, "lora": { "lora_alpha": 128, "lora_dropout": 0.0, "r": 64, "target_modules": [ "q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj" ] }, "loss": "causal_lm_cross_entropy_over_assistant_tokens_plus_eos", "loss_history": [ 1.674497276544571, 1.424118548631668, 1.2172126770019531, 0.9895333647727966, 0.9804797470569611, 0.8449891358613968, 0.8705039471387863, 0.8131226599216461, 0.7086513191461563, 0.8576440811157227, 0.8534276485443115, 0.715492844581604, 0.707272469997406, 0.6665292084217072, 0.7853038907051086, 0.7821626365184784, 0.7917397916316986, 0.8028056025505066, 0.7555173486471176, 0.7929783761501312, 0.7170834392309189, 0.6174379587173462, 0.6948337703943253, 0.6701837778091431, 0.7386516332626343, 0.73464235663414, 0.7026794850826263, 0.6100178137421608, 0.7647138237953186, 0.6368986219167709, 0.6608629748225212, 0.7625716030597687, 0.6585606038570404, 0.6864584535360336, 0.7305825799703598, 0.6575927436351776, 0.7425266802310944, 0.698588065803051, 0.7375385761260986, 0.6832857429981232, 0.5760672986507416, 0.7314791977405548, 0.5662346929311752, 0.683237761259079, 0.5718749314546585, 0.6281892955303192, 0.6270048916339874, 0.6319160535931587, 0.6945476531982422, 0.4989171773195267, 0.6054027453064919, 0.5484571978449821, 0.5098851323127747, 0.578282967209816, 0.5611839219927788, 0.6704423874616623, 0.6755621284246445, 0.6446011587977409, 0.6212967187166214, 0.6700150668621063, 0.605804368853569, 0.6849987730383873, 0.6438986361026764, 0.6158590316772461, 0.5576384887099266, 0.6274434924125671, 0.6196258440613747, 0.6213279366493225, 0.5476436838507652, 0.478846900165081, 0.5157266780734062, 0.6165014877915382, 0.5283321887254715, 0.5662178695201874, 0.524317778646946, 0.6116686463356018, 0.5614674985408783, 0.6579497307538986, 0.6378611698746681, 0.6100719124078751, 0.4799347370862961, 0.502042144536972, 0.47452619671821594, 0.5412955656647682, 0.6054418683052063, 0.6341072171926498, 0.6075669154524803, 0.6292348653078079, 0.5417201370000839, 0.5772567763924599, 0.6679643392562866, 0.6135343015193939, 0.5048784911632538, 0.5654781982302666, 0.6500979959964752, 0.548542208969593, 0.5479495972394943, 0.5252698585391045, 0.5274239182472229, 0.5279307588934898, 0.7067538946866989, 0.5134467482566833, 0.6220681518316269, 0.5735679417848587, 0.5978227704763412, 0.5640295669436455, 0.5638041347265244, 0.6589910835027695, 0.5192286223173141, 0.60391965508461, 0.5989179238677025, 0.6700718849897385, 0.5089113637804985, 0.5740328580141068, 0.5955520421266556, 0.5563981607556343, 0.45138001441955566, 0.6625692248344421, 0.6839468330144882, 0.5831266343593597, 0.6351960301399231, 0.5370437428355217, 0.5384622514247894, 0.5624337494373322, 0.5478598847985268, 0.5368805229663849, 0.6374900490045547, 0.5013832449913025, 0.5206728801131248, 0.5569872111082077, 0.5328838378190994, 0.5350202023983002, 0.5462082028388977, 0.6727063208818436, 0.5922725051641464, 0.6684836521744728, 0.502689890563488, 0.6078523173928261, 0.47100620716810226, 0.4847679063677788, 0.5584533661603928, 0.46530982851982117, 0.4517822116613388, 0.4732321724295616, 0.5928031280636787, 0.5177983343601227, 0.5342362597584724, 0.5351215079426765, 0.46405134350061417, 0.5742849260568619, 0.6058965623378754, 0.6007532849907875, 0.5192140638828278, 0.5684481263160706, 0.5518762916326523, 0.5977075770497322, 0.5833464711904526, 0.5938353538513184, 0.5923080667853355, 0.5855063423514366, 0.291910745203495, 0.31036364659667015, 0.3445335477590561, 0.4653793200850487, 0.42676743119955063, 0.3851235508918762, 0.4401862695813179, 0.4103095829486847, 0.3837387412786484, 0.3885618560016155, 0.3506322652101517, 0.41191332042217255, 0.32415204122662544, 0.33650926500558853, 0.35253313183784485, 0.44284647703170776, 0.4358452633023262, 0.43365541845560074, 0.38328836113214493, 0.4184325262904167, 0.3411296382546425, 0.38362909853458405, 0.3427336998283863, 0.3859572187066078, 0.34096013382077217, 0.5104679465293884, 0.3832346946001053, 0.41267383843660355, 0.31820689886808395, 0.38504786789417267, 0.314915306866169, 0.4017075449228287, 0.3661908954381943, 0.3611315190792084, 0.4527396187186241, 0.3885391727089882, 0.45560476183891296, 0.38864072412252426, 0.3669252246618271, 0.3932229354977608, 0.44473619759082794, 0.4246080815792084, 0.4368325024843216, 0.348334401845932, 0.40966174006462097, 0.4676578938961029, 0.39527183398604393, 0.3567265719175339, 0.33480511605739594, 0.3775472640991211, 0.4434337243437767, 0.38914725184440613, 0.41581474244594574, 0.33446600288152695, 0.4454450011253357, 0.36806925386190414, 0.3652227073907852, 0.39849936217069626, 0.44316042214632034, 0.4367980509996414, 0.3795015215873718, 0.3464418351650238, 0.4595238193869591, 0.3673698529601097, 0.4201071560382843, 0.4121697023510933, 0.4474613070487976, 0.3706766963005066, 0.4609332010149956, 0.3296925947070122, 0.3520989343523979, 0.34723304212093353, 0.45352353155612946, 0.3589302971959114, 0.4880386218428612, 0.4212202653288841, 0.42267151921987534, 0.40065161138772964, 0.4022143706679344, 0.37918292731046677, 0.4270031973719597, 0.34106292575597763, 0.39766620844602585, 0.3791928067803383, 0.35999079793691635, 0.4194788187742233, 0.377376027405262, 0.4246901422739029, 0.37102797627449036, 0.4210122898221016, 0.38083013892173767, 0.41633402556180954, 0.38353198021650314, 0.384234718978405, 0.40298980474472046, 0.3743172809481621, 0.4727516174316406, 0.4348696619272232, 0.44459711760282516, 0.4390462189912796, 0.3813927546143532, 0.4319426044821739, 0.33676718920469284, 0.4389949142932892, 0.4061638042330742, 0.4019470289349556, 0.4043594002723694, 0.3853309750556946, 0.3509000092744827, 0.3965689539909363, 0.33437757939100266, 0.2950911670923233, 0.33276114612817764, 0.4975763261318207, 0.44756609946489334, 0.34766971319913864, 0.42578204721212387, 0.32802601903676987, 0.4208790361881256, 0.4675574079155922, 0.450109027326107, 0.41079457849264145, 0.39477166533470154, 0.3906251937150955, 0.3529075011610985, 0.462636262178421, 0.3638685718178749, 0.43238307535648346, 0.41842425614595413, 0.42664729058742523, 0.3149915151298046, 0.4105845168232918, 0.38002239912748337, 0.4063094034790993, 0.3416373208165169, 0.4118576794862747, 0.49570276588201523, 0.2980887144804001, 0.43318992108106613, 0.45600610226392746, 0.4644649997353554, 0.38178467005491257, 0.43654313683509827, 0.44511647522449493, 0.3954680263996124, 0.4763437658548355, 0.42597514390945435, 0.4114840626716614, 0.39281921833753586, 0.39411572366952896, 0.4174913242459297, 0.3687373250722885, 0.450418658554554, 0.42525165528059006, 0.4495394751429558, 0.4290570616722107, 0.40731893479824066, 0.3982403725385666, 0.3695288933813572, 0.36221451312303543, 0.467012494802475, 0.5631162375211716, 0.275169912725687, 0.2769295312464237, 0.32038627192378044, 0.30277110636234283, 0.29848554730415344, 0.26729074865579605, 0.27991851046681404, 0.27027805149555206, 0.2398896925151348, 0.33135219663381577, 0.24011575058102608, 0.2620994560420513, 0.2357589676976204, 0.24159282818436623, 0.2784684933722019, 0.29401184618473053, 0.2600315771996975, 0.2757680304348469, 0.2954583987593651, 0.2190210409462452, 0.2407151721417904, 0.2440522238612175, 0.26953547820448875, 0.2277752161026001, 0.25821660831570625, 0.2169933281838894, 0.2616825960576534, 0.28293079882860184, 0.27060895785689354, 0.25377144291996956, 0.28574131429195404, 0.2421056590974331, 0.2531380355358124, 0.28092368319630623, 0.26702262088656425, 0.26329611614346504, 0.25108131021261215, 0.2451440878212452, 0.23318153619766235, 0.25508109852671623, 0.21028175950050354, 0.230103962123394, 0.29703231155872345, 0.31109801679849625, 0.3023436516523361, 0.21004461497068405, 0.23428110033273697, 0.24544857442378998, 0.25051964819431305, 0.27357492595911026, 0.27619923651218414, 0.28402502089738846, 0.224389236420393, 0.2580961212515831, 0.24802954867482185, 0.27879153937101364, 0.2875358685851097, 0.24999094381928444, 0.2273840270936489, 0.28602898865938187, 0.25376249477267265, 0.26688481494784355, 0.26999612152576447, 0.24909012392163277, 0.2765335962176323, 0.28363293781876564, 0.26537972688674927, 0.3266337960958481, 0.26171572133898735, 0.22582080960273743, 0.2566048949956894, 0.2597469910979271, 0.2545197866857052, 0.2278466857969761, 0.24357771128416061, 0.2466999776661396, 0.24694927409291267, 0.26240623742341995, 0.22974956035614014, 0.2696947753429413, 0.26243771240115166, 0.25917453691363335, 0.2958737276494503, 0.2829410582780838, 0.2374037653207779, 0.24049992486834526, 0.3016117662191391, 0.28854797407984734, 0.22929572314023972, 0.2542939856648445, 0.2615140788257122, 0.26986291259527206, 0.3086613491177559, 0.27161943912506104, 0.2593257650732994, 0.25777220726013184, 0.23684323579072952, 0.2502761147916317, 0.2774394042789936, 0.21241196244955063, 0.2672227807343006, 0.29368212074041367, 0.2314101718366146, 0.28621547296643257, 0.26108359545469284, 0.23656752333045006, 0.2402387037873268, 0.22915735840797424, 0.25281574577093124, 0.29678967222571373, 0.26573309302330017, 0.2802852839231491, 0.2689703069627285, 0.2451498582959175, 0.31286827474832535, 0.2536899000406265, 0.27442121878266335, 0.29743896424770355, 0.291730172932148, 0.33966264128685, 0.2926706150174141, 0.23463528603315353, 0.23385144770145416, 0.2607155703008175, 0.2499878890812397, 0.29127686098217964, 0.2218613550066948, 0.2550174184143543, 0.2221960499882698, 0.2820197455585003, 0.2809398025274277, 0.2843964621424675, 0.2505200430750847, 0.2493257112801075, 0.26728355884552, 0.2867716923356056, 0.23276104405522346, 0.27038122713565826, 0.25791164115071297, 0.24725092574954033, 0.30656392872333527, 0.2576770856976509, 0.28356923907995224, 0.21640633419156075, 0.25694428756833076, 0.24520491808652878, 0.30577392131090164, 0.29629358649253845, 0.2772160619497299, 0.2907728850841522, 0.2949519604444504, 0.2936568930745125, 0.23746906220912933, 0.2819291204214096, 0.2810583859682083, 0.2893615886569023, 0.2543134391307831, 0.2474202886223793, 0.2661695294082165, 0.25556614995002747, 0.1960032507777214 ], "mean_loss": 0.43162577979528877, "source": "stacked_lora_aft", "source_adapter": "chloeli/llama-3.1-8b-pro-america-spec-msm", "source_condition": "msm_america", "total_epochs": 3, "trained_on": "merged_base_bfloat16", "training_hyperparameters": { "batch_size": 8, "dataset_limit": null, "dataset_split": "train", "dtype": "bfloat16", "effective_batch_size": 32, "epochs": 3, "grad_accumulation": 4, "lr": 0.0001, "max_length": 256, "n_examples": 5129, "n_microbatches": 1926, "n_steps": 483, "optimizer": "AdamW", "seed": 0, "weight_decay": 0.0 } }