{ "config": { "vocab_size": 32009, "d_model": 1280, "n_layers": 23, "n_heads": 20, "d_ff": 3456, "max_seq_len": 768, "rope_theta": 10000.0, "rope_pct": 0.25, "dropout": 0.0, "tie_embeddings": true }, "train_args": { "steps": 500, "batch_size": 4, "grad_accum": 8, "seq_len": 768, "lr": 5e-05, "warmup": 20, "save_every": 100, "log_every": 10, "use_8bit_adam": true }, "total_params": 496998400, "steps": [ { "step": 0, "loss": 6.937812924385071, "ema_loss": 0.3468906462192536, "lr": 2.5e-06, "elapsed_s": 4.15720272064209, "steps_per_sec": 0.2405463642738951, "vram_gb": 9.164018176 }, { "step": 10, "loss": 2.739780366420746, "ema_loss": 1.9638136853068686, "lr": 2.75e-05, "elapsed_s": 40.6308798789978, "steps_per_sec": 0.2707300465251781, "vram_gb": 10.17390848 }, { "step": 20, "loss": 1.2084012106060982, "ema_loss": 1.882433597453757, "lr": 5e-05, "elapsed_s": 77.03947591781616, "steps_per_sec": 0.27258752412078047, "vram_gb": 10.17390848 }, { "step": 30, "loss": 1.0680485740303993, "ema_loss": 1.5722938780650435, "lr": 4.995182577286858e-05, "elapsed_s": 113.55863189697266, "steps_per_sec": 0.27298673365601217, "vram_gb": 10.17390848 }, { "step": 40, "loss": 0.7858129814267159, "ema_loss": 1.2813095330935025, "lr": 4.980750938091074e-05, "elapsed_s": 149.9709255695343, "steps_per_sec": 0.27338632367772026, "vram_gb": 10.17390848 }, { "step": 50, "loss": 0.6811780035495758, "ema_loss": 1.0847747264114023, "lr": 4.956766880907269e-05, "elapsed_s": 186.4907727241516, "steps_per_sec": 0.2734719753423769, "vram_gb": 10.17390848 }, { "step": 60, "loss": 0.6801792122423649, "ema_loss": 0.9220971225814172, "lr": 4.9233331091504034e-05, "elapsed_s": 222.90015697479248, "steps_per_sec": 0.27366512804608933, "vram_gb": 10.17390848 }, { "step": 70, "loss": 0.4824926480650902, "ema_loss": 0.8074420473552227, "lr": 4.8805927913639885e-05, "elapsed_s": 259.41771626472473, "steps_per_sec": 0.2736898659902916, "vram_gb": 10.17390848 }, { "step": 80, "loss": 0.4920233264565468, "ema_loss": 0.7012675245586312, "lr": 4.8287289481503954e-05, "elapsed_s": 295.8301751613617, "steps_per_sec": 0.2738057399175667, "vram_gb": 10.17390848 }, { "step": 90, "loss": 0.6347984820604324, "ema_loss": 0.6271864858820199, "lr": 4.767963668448549e-05, "elapsed_s": 332.34874868392944, "steps_per_sec": 0.2738087637168837, "vram_gb": 10.17390848 }, { "step": 100, "loss": 0.6809600777924061, "ema_loss": 0.5743832670414681, "lr": 4.6985571585149876e-05, "elapsed_s": 414.7752454280853, "steps_per_sec": 0.24350537095278896, "vram_gb": 10.17390848 }, { "step": 110, "loss": 0.4763348139822483, "ema_loss": 0.521244228570747, "lr": 4.620806627680727e-05, "elapsed_s": 451.26919865608215, "steps_per_sec": 0.24597291446118502, "vram_gb": 10.17390848 }, { "step": 120, "loss": 0.5240486040711403, "ema_loss": 0.494888873638692, "lr": 4.535045015655279e-05, "elapsed_s": 487.7379126548767, "steps_per_sec": 0.24808405674549147, "vram_gb": 10.17390848 }, { "step": 130, "loss": 0.286451730877161, "ema_loss": 0.46101876181189894, "lr": 4.4416395668276995e-05, "elapsed_s": 524.1727938652039, "steps_per_sec": 0.24991758735514977, "vram_gb": 10.17390848 }, { "step": 140, "loss": 0.35978504456579685, "ema_loss": 0.4399799286456892, "lr": 4.340990257669732e-05, "elapsed_s": 560.6376507282257, "steps_per_sec": 0.25149934153878484, "vram_gb": 10.17390848 }, { "step": 150, "loss": 0.3526405543088913, "ema_loss": 0.40866163158804786, "lr": 4.2335280839751554e-05, "elapsed_s": 597.072151184082, "steps_per_sec": 0.25290075864457046, "vram_gb": 10.17390848 }, { "step": 160, "loss": 0.3449416421353817, "ema_loss": 0.36948892724765275, "lr": 4.1197132152696215e-05, "elapsed_s": 633.5287528038025, "steps_per_sec": 0.2541321120587878, "vram_gb": 10.17390848 }, { "step": 170, "loss": 0.32148140482604504, "ema_loss": 0.3509791943321394, "lr": 4.000033024294105e-05, "elapsed_s": 669.9639327526093, "steps_per_sec": 0.2552376204752255, "vram_gb": 10.17390848 }, { "step": 180, "loss": 0.40944845974445343, "ema_loss": 0.3403033324976888, "lr": 3.875e-05, "elapsed_s": 706.4267771244049, "steps_per_sec": 0.2562190532142372, "vram_gb": 10.17390848 }, { "step": 190, "loss": 0.2508099125698209, "ema_loss": 0.31061072625747166, "lr": 3.7451495529927534e-05, "elapsed_s": 742.8610553741455, "steps_per_sec": 0.25711403043440195, "vram_gb": 10.17390848 }, { "step": 200, "loss": 0.2846630644053221, "ema_loss": 0.29097493865487023, "lr": 3.6110377228214525e-05, "elapsed_s": 823.1199431419373, "steps_per_sec": 0.2441928441592138, "vram_gb": 10.17390848 }, { "step": 210, "loss": 0.1874328227713704, "ema_loss": 0.2691413417018003, "lr": 3.473238796932114e-05, "elapsed_s": 859.5507295131683, "steps_per_sec": 0.24547707628554513, "vram_gb": 10.17390848 }, { "step": 220, "loss": 0.18770729657262564, "ema_loss": 0.2469535500748924, "lr": 3.332342851480672e-05, "elapsed_s": 896.0201919078827, "steps_per_sec": 0.24664622739073316, "vram_gb": 10.17390848 }, { "step": 230, "loss": 0.20278130657970905, "ema_loss": 0.2411670786782403, "lr": 3.1889532245362895e-05, "elapsed_s": 932.4535627365112, "steps_per_sec": 0.2477335164252839, "vram_gb": 10.17390848 }, { "step": 240, "loss": 0.24111063033342361, "ema_loss": 0.2326045524210834, "lr": 3.0436839324951166e-05, "elapsed_s": 968.919579744339, "steps_per_sec": 0.24873065323294502, "vram_gb": 10.17390848 }, { "step": 250, "loss": 0.2577468641102314, "ema_loss": 0.22216086817509004, "lr": 2.897157040767822e-05, "elapsed_s": 1005.3552136421204, "steps_per_sec": 0.24966300128956143, "vram_gb": 10.17390848 }, { "step": 260, "loss": 0.21209231484681368, "ema_loss": 0.2089893973874345, "lr": 2.75e-05, "elapsed_s": 1041.8542084693909, "steps_per_sec": 0.25051489726517534, "vram_gb": 10.17390848 }, { "step": 270, "loss": 0.16555726807564497, "ema_loss": 0.20300413108044887, "lr": 2.602842959232178e-05, "elapsed_s": 1078.29061460495, "steps_per_sec": 0.25132371211381216, "vram_gb": 10.17390848 }, { "step": 280, "loss": 0.20196319418027997, "ema_loss": 0.1959523809991305, "lr": 2.4563160675048846e-05, "elapsed_s": 1114.7595565319061, "steps_per_sec": 0.25207229519001423, "vram_gb": 10.17390848 }, { "step": 290, "loss": 0.1890458231791854, "ema_loss": 0.19173463181336906, "lr": 2.3110467754637118e-05, "elapsed_s": 1151.1741054058075, "steps_per_sec": 0.2527853941758165, "vram_gb": 10.17390848 }, { "step": 300, "loss": 0.14224259229376912, "ema_loss": 0.170971237515532, "lr": 2.1676571485193282e-05, "elapsed_s": 1231.5103888511658, "steps_per_sec": 0.2444153153111381, "vram_gb": 10.17390848 }, { "step": 310, "loss": 0.1485865106806159, "ema_loss": 0.17176215697934846, "lr": 2.026761203067887e-05, "elapsed_s": 1267.9178903102875, "steps_per_sec": 0.24528402223577067, "vram_gb": 10.17390848 }, { "step": 320, "loss": 0.17956801364198327, "ema_loss": 0.1700756775799304, "lr": 1.888962277178548e-05, "elapsed_s": 1304.4403862953186, "steps_per_sec": 0.2460825372876237, "vram_gb": 10.17390848 }, { "step": 330, "loss": 0.12682513892650604, "ema_loss": 0.15713850400483031, "lr": 1.7548504470072478e-05, "elapsed_s": 1340.8541731834412, "steps_per_sec": 0.24685756782495105, "vram_gb": 10.17390848 }, { "step": 340, "loss": 0.12957867234945297, "ema_loss": 0.1510191265222257, "lr": 1.6250000000000005e-05, "elapsed_s": 1377.3766162395477, "steps_per_sec": 0.24757208448258908, "vram_gb": 10.17390848 }, { "step": 350, "loss": 0.12622385285794735, "ema_loss": 0.14607439549355963, "lr": 1.4999669757058956e-05, "elapsed_s": 1413.787192106247, "steps_per_sec": 0.24826933074495, "vram_gb": 10.17390848 }, { "step": 360, "loss": 0.1389128677546978, "ema_loss": 0.1373236412001944, "lr": 1.3802867847303787e-05, "elapsed_s": 1450.3096079826355, "steps_per_sec": 0.24891236878872158, "vram_gb": 10.17390848 }, { "step": 370, "loss": 0.11274528503417969, "ema_loss": 0.1358663910950191, "lr": 1.2664719160248452e-05, "elapsed_s": 1486.721604347229, "steps_per_sec": 0.24954234801941552, "vram_gb": 10.17390848 }, { "step": 380, "loss": 0.12164316419512033, "ema_loss": 0.12515423124366815, "lr": 1.1590097423302684e-05, "elapsed_s": 1523.1735746860504, "steps_per_sec": 0.2501356420121259, "vram_gb": 10.17390848 }, { "step": 390, "loss": 0.10157549497671425, "ema_loss": 0.1216169996836354, "lr": 1.0583604331723012e-05, "elapsed_s": 1559.6568088531494, "steps_per_sec": 0.25069617737732386, "vram_gb": 10.17390848 }, { "step": 400, "loss": 0.10004387237131596, "ema_loss": 0.12196290672243944, "lr": 9.649549843447213e-06, "elapsed_s": 1641.5576264858246, "steps_per_sec": 0.24428018458203227, "vram_gb": 10.17390848 }, { "step": 410, "loss": 0.15670533804222941, "ema_loss": 0.11768572540664866, "lr": 8.79193372319273e-06, "elapsed_s": 1678.030652999878, "steps_per_sec": 0.24492997149083062, "vram_gb": 10.17390848 }, { "step": 420, "loss": 0.0967125273309648, "ema_loss": 0.12051557885116983, "lr": 8.01442841485013e-06, "elapsed_s": 1714.4823727607727, "steps_per_sec": 0.24555516387262588, "vram_gb": 10.17390848 }, { "step": 430, "loss": 0.08109346427954733, "ema_loss": 0.11255813170642848, "lr": 7.320363315514515e-06, "elapsed_s": 1750.9651441574097, "steps_per_sec": 0.24614995988820987, "vram_gb": 10.17390848 }, { "step": 440, "loss": 0.06976099917665124, "ema_loss": 0.10574265881842762, "lr": 6.712710518496049e-06, "elapsed_s": 1787.420247077942, "steps_per_sec": 0.24672429481591848, "vram_gb": 10.17390848 }, { "step": 450, "loss": 0.08222721493802965, "ema_loss": 0.09805770252970188, "lr": 6.194072086360125e-06, "elapsed_s": 1823.9047615528107, "steps_per_sec": 0.24727168298855356, "vram_gb": 10.17390848 }, { "step": 460, "loss": 0.09003248857334256, "ema_loss": 0.09465487723876762, "lr": 5.766668908495966e-06, "elapsed_s": 1860.354943037033, "steps_per_sec": 0.24780217437830257, "vram_gb": 10.17390848 }, { "step": 470, "loss": 0.050968108931556344, "ema_loss": 0.0877422503218975, "lr": 5.432331190927316e-06, "elapsed_s": 1896.8366348743439, "steps_per_sec": 0.24830815228914083, "vram_gb": 10.17390848 }, { "step": 480, "loss": 0.09658047696575522, "ema_loss": 0.08559591419795948, "lr": 5.192490619089267e-06, "elapsed_s": 1933.280499458313, "steps_per_sec": 0.24879990261877244, "vram_gb": 10.17390848 }, { "step": 490, "loss": 0.057105707470327616, "ema_loss": 0.08398783932396543, "lr": 5.048174227131422e-06, "elapsed_s": 1969.7602517604828, "steps_per_sec": 0.24926891461088543, "vram_gb": 10.17390848 }, { "step": 499, "loss": 0.0780253498815, "ema_loss": 0.08407452203461949, "lr": 5.000481912557093e-06, "elapsed_s": 2002.5768718719482, "steps_per_sec": 0.24967830549876227, "vram_gb": 10.17390848 } ] }