zlm_b32_le4_s8000 / checkpoint-4500 /trainer_state.json
mikhail-panzo's picture
Training in progress, step 4500, checkpoint
f7261c5 verified
{
"best_metric": 0.3493475914001465,
"best_model_checkpoint": "mikhail-panzo/zlm_b32_le4_s8000/checkpoint-4500",
"epoch": 1.8848167539267016,
"eval_steps": 500,
"global_step": 4500,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.020942408376963352,
"grad_norm": 3.1775436401367188,
"learning_rate": 2.35e-06,
"loss": 1.1228,
"step": 50
},
{
"epoch": 0.041884816753926704,
"grad_norm": 4.18980073928833,
"learning_rate": 4.85e-06,
"loss": 0.8304,
"step": 100
},
{
"epoch": 0.06282722513089005,
"grad_norm": 4.5355024337768555,
"learning_rate": 7.35e-06,
"loss": 0.7701,
"step": 150
},
{
"epoch": 0.08376963350785341,
"grad_norm": 3.9590957164764404,
"learning_rate": 9.85e-06,
"loss": 0.7096,
"step": 200
},
{
"epoch": 0.10471204188481675,
"grad_norm": 5.419675350189209,
"learning_rate": 1.235e-05,
"loss": 0.6518,
"step": 250
},
{
"epoch": 0.1256544502617801,
"grad_norm": 3.1052777767181396,
"learning_rate": 1.485e-05,
"loss": 0.6326,
"step": 300
},
{
"epoch": 0.14659685863874344,
"grad_norm": 3.4649453163146973,
"learning_rate": 1.7349999999999998e-05,
"loss": 0.6241,
"step": 350
},
{
"epoch": 0.16753926701570682,
"grad_norm": 4.962624549865723,
"learning_rate": 1.985e-05,
"loss": 0.5723,
"step": 400
},
{
"epoch": 0.18848167539267016,
"grad_norm": 2.7586095333099365,
"learning_rate": 2.235e-05,
"loss": 0.5764,
"step": 450
},
{
"epoch": 0.2094240837696335,
"grad_norm": 3.5610299110412598,
"learning_rate": 2.485e-05,
"loss": 0.5552,
"step": 500
},
{
"epoch": 0.2094240837696335,
"eval_loss": 0.48825809359550476,
"eval_runtime": 277.5892,
"eval_samples_per_second": 30.581,
"eval_steps_per_second": 3.826,
"step": 500
},
{
"epoch": 0.23036649214659685,
"grad_norm": 3.3351809978485107,
"learning_rate": 2.7350000000000004e-05,
"loss": 0.5322,
"step": 550
},
{
"epoch": 0.2513089005235602,
"grad_norm": 3.7222373485565186,
"learning_rate": 2.985e-05,
"loss": 0.5186,
"step": 600
},
{
"epoch": 0.27225130890052357,
"grad_norm": 3.1811156272888184,
"learning_rate": 3.235e-05,
"loss": 0.5157,
"step": 650
},
{
"epoch": 0.2931937172774869,
"grad_norm": 2.2585642337799072,
"learning_rate": 3.485e-05,
"loss": 0.5089,
"step": 700
},
{
"epoch": 0.31413612565445026,
"grad_norm": 3.999460220336914,
"learning_rate": 3.735e-05,
"loss": 0.5117,
"step": 750
},
{
"epoch": 0.33507853403141363,
"grad_norm": 3.739990472793579,
"learning_rate": 3.9850000000000006e-05,
"loss": 0.5031,
"step": 800
},
{
"epoch": 0.35602094240837695,
"grad_norm": 4.251980781555176,
"learning_rate": 4.235e-05,
"loss": 0.5064,
"step": 850
},
{
"epoch": 0.3769633507853403,
"grad_norm": 2.770602226257324,
"learning_rate": 4.4850000000000006e-05,
"loss": 0.4921,
"step": 900
},
{
"epoch": 0.39790575916230364,
"grad_norm": 2.506974220275879,
"learning_rate": 4.735e-05,
"loss": 0.4839,
"step": 950
},
{
"epoch": 0.418848167539267,
"grad_norm": 2.2666189670562744,
"learning_rate": 4.9850000000000006e-05,
"loss": 0.4913,
"step": 1000
},
{
"epoch": 0.418848167539267,
"eval_loss": 0.4265913963317871,
"eval_runtime": 265.1599,
"eval_samples_per_second": 32.015,
"eval_steps_per_second": 4.005,
"step": 1000
},
{
"epoch": 0.4397905759162304,
"grad_norm": 3.1451058387756348,
"learning_rate": 5.235e-05,
"loss": 0.4722,
"step": 1050
},
{
"epoch": 0.4607329842931937,
"grad_norm": 3.197997570037842,
"learning_rate": 5.485e-05,
"loss": 0.4692,
"step": 1100
},
{
"epoch": 0.4816753926701571,
"grad_norm": 2.9112601280212402,
"learning_rate": 5.7350000000000005e-05,
"loss": 0.4738,
"step": 1150
},
{
"epoch": 0.5026178010471204,
"grad_norm": 3.036731243133545,
"learning_rate": 5.9850000000000005e-05,
"loss": 0.4521,
"step": 1200
},
{
"epoch": 0.5235602094240838,
"grad_norm": 5.021958351135254,
"learning_rate": 6.235000000000001e-05,
"loss": 0.4666,
"step": 1250
},
{
"epoch": 0.5445026178010471,
"grad_norm": 3.302204132080078,
"learning_rate": 6.485e-05,
"loss": 0.4625,
"step": 1300
},
{
"epoch": 0.5654450261780105,
"grad_norm": 3.2643635272979736,
"learning_rate": 6.735e-05,
"loss": 0.4683,
"step": 1350
},
{
"epoch": 0.5863874345549738,
"grad_norm": 1.7499467134475708,
"learning_rate": 6.985e-05,
"loss": 0.449,
"step": 1400
},
{
"epoch": 0.6073298429319371,
"grad_norm": 1.3616622686386108,
"learning_rate": 7.235000000000001e-05,
"loss": 0.4523,
"step": 1450
},
{
"epoch": 0.6282722513089005,
"grad_norm": 2.5826191902160645,
"learning_rate": 7.485e-05,
"loss": 0.446,
"step": 1500
},
{
"epoch": 0.6282722513089005,
"eval_loss": 0.3975289463996887,
"eval_runtime": 266.4222,
"eval_samples_per_second": 31.863,
"eval_steps_per_second": 3.986,
"step": 1500
},
{
"epoch": 0.6492146596858639,
"grad_norm": 3.6052303314208984,
"learning_rate": 7.735e-05,
"loss": 0.4449,
"step": 1550
},
{
"epoch": 0.6701570680628273,
"grad_norm": 3.4120566844940186,
"learning_rate": 7.985e-05,
"loss": 0.4477,
"step": 1600
},
{
"epoch": 0.6910994764397905,
"grad_norm": 2.187040090560913,
"learning_rate": 8.235000000000001e-05,
"loss": 0.4522,
"step": 1650
},
{
"epoch": 0.7120418848167539,
"grad_norm": 1.718518853187561,
"learning_rate": 8.485e-05,
"loss": 0.4431,
"step": 1700
},
{
"epoch": 0.7329842931937173,
"grad_norm": 1.8248894214630127,
"learning_rate": 8.735000000000001e-05,
"loss": 0.4496,
"step": 1750
},
{
"epoch": 0.7539267015706806,
"grad_norm": 2.733355760574341,
"learning_rate": 8.985e-05,
"loss": 0.4297,
"step": 1800
},
{
"epoch": 0.774869109947644,
"grad_norm": 2.5899884700775146,
"learning_rate": 9.235000000000001e-05,
"loss": 0.4588,
"step": 1850
},
{
"epoch": 0.7958115183246073,
"grad_norm": 1.5704914331436157,
"learning_rate": 9.485e-05,
"loss": 0.4345,
"step": 1900
},
{
"epoch": 0.8167539267015707,
"grad_norm": 5.18487548828125,
"learning_rate": 9.735000000000001e-05,
"loss": 0.4325,
"step": 1950
},
{
"epoch": 0.837696335078534,
"grad_norm": 2.4392499923706055,
"learning_rate": 9.985000000000001e-05,
"loss": 0.4222,
"step": 2000
},
{
"epoch": 0.837696335078534,
"eval_loss": 0.39489272236824036,
"eval_runtime": 267.0393,
"eval_samples_per_second": 31.789,
"eval_steps_per_second": 3.977,
"step": 2000
},
{
"epoch": 0.8586387434554974,
"grad_norm": 2.1138336658477783,
"learning_rate": 9.921666666666668e-05,
"loss": 0.4438,
"step": 2050
},
{
"epoch": 0.8795811518324608,
"grad_norm": 2.3158669471740723,
"learning_rate": 9.838333333333334e-05,
"loss": 0.421,
"step": 2100
},
{
"epoch": 0.900523560209424,
"grad_norm": 2.1935179233551025,
"learning_rate": 9.755000000000001e-05,
"loss": 0.4247,
"step": 2150
},
{
"epoch": 0.9214659685863874,
"grad_norm": 1.7845231294631958,
"learning_rate": 9.671666666666667e-05,
"loss": 0.4203,
"step": 2200
},
{
"epoch": 0.9424083769633508,
"grad_norm": 1.7287964820861816,
"learning_rate": 9.588333333333334e-05,
"loss": 0.4306,
"step": 2250
},
{
"epoch": 0.9633507853403142,
"grad_norm": 2.2276289463043213,
"learning_rate": 9.505e-05,
"loss": 0.4392,
"step": 2300
},
{
"epoch": 0.9842931937172775,
"grad_norm": 2.248389720916748,
"learning_rate": 9.421666666666668e-05,
"loss": 0.4304,
"step": 2350
},
{
"epoch": 1.0052356020942408,
"grad_norm": 1.5984878540039062,
"learning_rate": 9.338333333333333e-05,
"loss": 0.4125,
"step": 2400
},
{
"epoch": 1.0261780104712042,
"grad_norm": 2.6061534881591797,
"learning_rate": 9.255e-05,
"loss": 0.4229,
"step": 2450
},
{
"epoch": 1.0471204188481675,
"grad_norm": 2.127110481262207,
"learning_rate": 9.171666666666667e-05,
"loss": 0.4273,
"step": 2500
},
{
"epoch": 1.0471204188481675,
"eval_loss": 0.38292551040649414,
"eval_runtime": 263.8923,
"eval_samples_per_second": 32.168,
"eval_steps_per_second": 4.024,
"step": 2500
},
{
"epoch": 1.068062827225131,
"grad_norm": 1.9867545366287231,
"learning_rate": 9.088333333333334e-05,
"loss": 0.4164,
"step": 2550
},
{
"epoch": 1.0890052356020943,
"grad_norm": 2.1518051624298096,
"learning_rate": 9.005000000000001e-05,
"loss": 0.4038,
"step": 2600
},
{
"epoch": 1.1099476439790577,
"grad_norm": 2.1266932487487793,
"learning_rate": 8.921666666666668e-05,
"loss": 0.418,
"step": 2650
},
{
"epoch": 1.130890052356021,
"grad_norm": 1.9351108074188232,
"learning_rate": 8.838333333333334e-05,
"loss": 0.4213,
"step": 2700
},
{
"epoch": 1.1518324607329844,
"grad_norm": 1.4189810752868652,
"learning_rate": 8.755e-05,
"loss": 0.4201,
"step": 2750
},
{
"epoch": 1.1727748691099475,
"grad_norm": 3.590930223464966,
"learning_rate": 8.671666666666667e-05,
"loss": 0.4043,
"step": 2800
},
{
"epoch": 1.193717277486911,
"grad_norm": 3.049971103668213,
"learning_rate": 8.588333333333334e-05,
"loss": 0.4036,
"step": 2850
},
{
"epoch": 1.2146596858638743,
"grad_norm": 2.1481425762176514,
"learning_rate": 8.505000000000001e-05,
"loss": 0.4067,
"step": 2900
},
{
"epoch": 1.2356020942408377,
"grad_norm": 1.774927020072937,
"learning_rate": 8.421666666666666e-05,
"loss": 0.4126,
"step": 2950
},
{
"epoch": 1.256544502617801,
"grad_norm": 1.624089241027832,
"learning_rate": 8.338333333333333e-05,
"loss": 0.4028,
"step": 3000
},
{
"epoch": 1.256544502617801,
"eval_loss": 0.3674115538597107,
"eval_runtime": 263.8647,
"eval_samples_per_second": 32.172,
"eval_steps_per_second": 4.025,
"step": 3000
},
{
"epoch": 1.2774869109947644,
"grad_norm": 1.8440332412719727,
"learning_rate": 8.255e-05,
"loss": 0.4026,
"step": 3050
},
{
"epoch": 1.2984293193717278,
"grad_norm": 2.144713878631592,
"learning_rate": 8.171666666666667e-05,
"loss": 0.4007,
"step": 3100
},
{
"epoch": 1.3193717277486912,
"grad_norm": 1.9491198062896729,
"learning_rate": 8.088333333333334e-05,
"loss": 0.4087,
"step": 3150
},
{
"epoch": 1.3403141361256545,
"grad_norm": 2.0903196334838867,
"learning_rate": 8.005000000000001e-05,
"loss": 0.4034,
"step": 3200
},
{
"epoch": 1.3612565445026177,
"grad_norm": 2.226724863052368,
"learning_rate": 7.921666666666668e-05,
"loss": 0.4106,
"step": 3250
},
{
"epoch": 1.3821989528795813,
"grad_norm": 1.8056219816207886,
"learning_rate": 7.838333333333335e-05,
"loss": 0.3984,
"step": 3300
},
{
"epoch": 1.4031413612565444,
"grad_norm": 1.8196921348571777,
"learning_rate": 7.755e-05,
"loss": 0.4018,
"step": 3350
},
{
"epoch": 1.4240837696335078,
"grad_norm": 1.7930132150650024,
"learning_rate": 7.671666666666667e-05,
"loss": 0.395,
"step": 3400
},
{
"epoch": 1.4450261780104712,
"grad_norm": 2.525432586669922,
"learning_rate": 7.588333333333334e-05,
"loss": 0.399,
"step": 3450
},
{
"epoch": 1.4659685863874345,
"grad_norm": 1.2433278560638428,
"learning_rate": 7.505e-05,
"loss": 0.3941,
"step": 3500
},
{
"epoch": 1.4659685863874345,
"eval_loss": 0.3616171181201935,
"eval_runtime": 265.4304,
"eval_samples_per_second": 31.982,
"eval_steps_per_second": 4.001,
"step": 3500
},
{
"epoch": 1.486910994764398,
"grad_norm": 1.859680414199829,
"learning_rate": 7.421666666666666e-05,
"loss": 0.4023,
"step": 3550
},
{
"epoch": 1.5078534031413613,
"grad_norm": 1.9057739973068237,
"learning_rate": 7.338333333333333e-05,
"loss": 0.3938,
"step": 3600
},
{
"epoch": 1.5287958115183247,
"grad_norm": 1.6138437986373901,
"learning_rate": 7.255e-05,
"loss": 0.3995,
"step": 3650
},
{
"epoch": 1.5497382198952878,
"grad_norm": 1.9373960494995117,
"learning_rate": 7.171666666666667e-05,
"loss": 0.3934,
"step": 3700
},
{
"epoch": 1.5706806282722514,
"grad_norm": 1.7537634372711182,
"learning_rate": 7.088333333333334e-05,
"loss": 0.3969,
"step": 3750
},
{
"epoch": 1.5916230366492146,
"grad_norm": 2.310279607772827,
"learning_rate": 7.005000000000001e-05,
"loss": 0.3967,
"step": 3800
},
{
"epoch": 1.6125654450261782,
"grad_norm": 3.323341131210327,
"learning_rate": 6.921666666666668e-05,
"loss": 0.4058,
"step": 3850
},
{
"epoch": 1.6335078534031413,
"grad_norm": 1.784133791923523,
"learning_rate": 6.838333333333333e-05,
"loss": 0.3898,
"step": 3900
},
{
"epoch": 1.6544502617801047,
"grad_norm": 1.8042694330215454,
"learning_rate": 6.755e-05,
"loss": 0.3902,
"step": 3950
},
{
"epoch": 1.675392670157068,
"grad_norm": 1.8511914014816284,
"learning_rate": 6.671666666666667e-05,
"loss": 0.3871,
"step": 4000
},
{
"epoch": 1.675392670157068,
"eval_loss": 0.3518759310245514,
"eval_runtime": 268.3666,
"eval_samples_per_second": 31.632,
"eval_steps_per_second": 3.957,
"step": 4000
},
{
"epoch": 1.6963350785340314,
"grad_norm": 1.1159111261367798,
"learning_rate": 6.588333333333334e-05,
"loss": 0.3842,
"step": 4050
},
{
"epoch": 1.7172774869109948,
"grad_norm": 1.8107820749282837,
"learning_rate": 6.505e-05,
"loss": 0.3909,
"step": 4100
},
{
"epoch": 1.738219895287958,
"grad_norm": 1.5878946781158447,
"learning_rate": 6.421666666666666e-05,
"loss": 0.3974,
"step": 4150
},
{
"epoch": 1.7591623036649215,
"grad_norm": 3.3478894233703613,
"learning_rate": 6.338333333333333e-05,
"loss": 0.3874,
"step": 4200
},
{
"epoch": 1.7801047120418847,
"grad_norm": 2.9195618629455566,
"learning_rate": 6.255e-05,
"loss": 0.3893,
"step": 4250
},
{
"epoch": 1.8010471204188483,
"grad_norm": 1.6017109155654907,
"learning_rate": 6.171666666666667e-05,
"loss": 0.3827,
"step": 4300
},
{
"epoch": 1.8219895287958114,
"grad_norm": 2.468721628189087,
"learning_rate": 6.0883333333333334e-05,
"loss": 0.3835,
"step": 4350
},
{
"epoch": 1.8429319371727748,
"grad_norm": 1.5800402164459229,
"learning_rate": 6.005000000000001e-05,
"loss": 0.3767,
"step": 4400
},
{
"epoch": 1.8638743455497382,
"grad_norm": 2.512275218963623,
"learning_rate": 5.9216666666666665e-05,
"loss": 0.3835,
"step": 4450
},
{
"epoch": 1.8848167539267016,
"grad_norm": 1.4855787754058838,
"learning_rate": 5.8383333333333334e-05,
"loss": 0.3828,
"step": 4500
},
{
"epoch": 1.8848167539267016,
"eval_loss": 0.3493475914001465,
"eval_runtime": 265.6954,
"eval_samples_per_second": 31.95,
"eval_steps_per_second": 3.997,
"step": 4500
}
],
"logging_steps": 50,
"max_steps": 8000,
"num_input_tokens_seen": 0,
"num_train_epochs": 4,
"save_steps": 500,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": false
},
"attributes": {}
}
},
"total_flos": 2.014738293479155e+16,
"train_batch_size": 16,
"trial_name": null,
"trial_params": null
}