GaetanMichelet's picture
Model save
7257172 verified
{
"best_metric": 0.9190607070922852,
"best_model_checkpoint": "data/Llama-31-8B_task-1_60-samples_config-1_auto/checkpoint-34",
"epoch": 12.869565217391305,
"eval_steps": 500,
"global_step": 74,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.17391304347826086,
"grad_norm": 0.8832762241363525,
"learning_rate": 4.000000000000001e-06,
"loss": 1.0082,
"step": 1
},
{
"epoch": 0.34782608695652173,
"grad_norm": 0.9423871636390686,
"learning_rate": 8.000000000000001e-06,
"loss": 1.3579,
"step": 2
},
{
"epoch": 0.6956521739130435,
"grad_norm": 0.7188506126403809,
"learning_rate": 1.6000000000000003e-05,
"loss": 1.1017,
"step": 4
},
{
"epoch": 0.8695652173913043,
"eval_loss": 1.1986019611358643,
"eval_runtime": 5.0553,
"eval_samples_per_second": 2.374,
"eval_steps_per_second": 2.374,
"step": 5
},
{
"epoch": 1.0434782608695652,
"grad_norm": 0.8392509818077087,
"learning_rate": 2.4e-05,
"loss": 1.129,
"step": 6
},
{
"epoch": 1.391304347826087,
"grad_norm": 0.8663368225097656,
"learning_rate": 3.2000000000000005e-05,
"loss": 1.0328,
"step": 8
},
{
"epoch": 1.7391304347826086,
"grad_norm": 0.6638839244842529,
"learning_rate": 4e-05,
"loss": 1.0551,
"step": 10
},
{
"epoch": 1.9130434782608696,
"eval_loss": 1.0320765972137451,
"eval_runtime": 5.0292,
"eval_samples_per_second": 2.386,
"eval_steps_per_second": 2.386,
"step": 11
},
{
"epoch": 2.0869565217391304,
"grad_norm": 0.5207677483558655,
"learning_rate": 4.8e-05,
"loss": 1.0609,
"step": 12
},
{
"epoch": 2.4347826086956523,
"grad_norm": 0.5633108615875244,
"learning_rate": 5.6000000000000006e-05,
"loss": 0.9272,
"step": 14
},
{
"epoch": 2.782608695652174,
"grad_norm": 0.6757349967956543,
"learning_rate": 6.400000000000001e-05,
"loss": 0.9039,
"step": 16
},
{
"epoch": 2.9565217391304346,
"eval_loss": 0.9664216041564941,
"eval_runtime": 5.047,
"eval_samples_per_second": 2.378,
"eval_steps_per_second": 2.378,
"step": 17
},
{
"epoch": 3.130434782608696,
"grad_norm": 0.621843159198761,
"learning_rate": 7.2e-05,
"loss": 0.9263,
"step": 18
},
{
"epoch": 3.4782608695652173,
"grad_norm": 0.5294705629348755,
"learning_rate": 8e-05,
"loss": 0.8304,
"step": 20
},
{
"epoch": 3.8260869565217392,
"grad_norm": 0.5593620538711548,
"learning_rate": 8.800000000000001e-05,
"loss": 0.8058,
"step": 22
},
{
"epoch": 4.0,
"eval_loss": 0.9390919208526611,
"eval_runtime": 5.0492,
"eval_samples_per_second": 2.377,
"eval_steps_per_second": 2.377,
"step": 23
},
{
"epoch": 4.173913043478261,
"grad_norm": 0.5889991521835327,
"learning_rate": 9.6e-05,
"loss": 0.8395,
"step": 24
},
{
"epoch": 4.521739130434782,
"grad_norm": 0.6124499440193176,
"learning_rate": 9.999512620046522e-05,
"loss": 0.6935,
"step": 26
},
{
"epoch": 4.869565217391305,
"grad_norm": 0.6663640141487122,
"learning_rate": 9.995614150494293e-05,
"loss": 0.6678,
"step": 28
},
{
"epoch": 4.869565217391305,
"eval_loss": 0.9208094477653503,
"eval_runtime": 5.0497,
"eval_samples_per_second": 2.376,
"eval_steps_per_second": 2.376,
"step": 28
},
{
"epoch": 5.217391304347826,
"grad_norm": 0.6573657989501953,
"learning_rate": 9.987820251299122e-05,
"loss": 0.6198,
"step": 30
},
{
"epoch": 5.565217391304348,
"grad_norm": 0.8195709586143494,
"learning_rate": 9.976136999909156e-05,
"loss": 0.6435,
"step": 32
},
{
"epoch": 5.913043478260869,
"grad_norm": 0.8117626905441284,
"learning_rate": 9.96057350657239e-05,
"loss": 0.5587,
"step": 34
},
{
"epoch": 5.913043478260869,
"eval_loss": 0.9190607070922852,
"eval_runtime": 5.0444,
"eval_samples_per_second": 2.379,
"eval_steps_per_second": 2.379,
"step": 34
},
{
"epoch": 6.260869565217392,
"grad_norm": 0.6630845069885254,
"learning_rate": 9.941141907232765e-05,
"loss": 0.4912,
"step": 36
},
{
"epoch": 6.608695652173913,
"grad_norm": 0.8549047112464905,
"learning_rate": 9.917857354066931e-05,
"loss": 0.4736,
"step": 38
},
{
"epoch": 6.956521739130435,
"grad_norm": 0.8069887757301331,
"learning_rate": 9.890738003669029e-05,
"loss": 0.3836,
"step": 40
},
{
"epoch": 6.956521739130435,
"eval_loss": 1.0232114791870117,
"eval_runtime": 5.0523,
"eval_samples_per_second": 2.375,
"eval_steps_per_second": 2.375,
"step": 40
},
{
"epoch": 7.304347826086957,
"grad_norm": 0.7592400312423706,
"learning_rate": 9.859805002892732e-05,
"loss": 0.3201,
"step": 42
},
{
"epoch": 7.6521739130434785,
"grad_norm": 0.960902750492096,
"learning_rate": 9.825082472361557e-05,
"loss": 0.2681,
"step": 44
},
{
"epoch": 8.0,
"grad_norm": 0.9407598972320557,
"learning_rate": 9.786597487660337e-05,
"loss": 0.2717,
"step": 46
},
{
"epoch": 8.0,
"eval_loss": 1.2434306144714355,
"eval_runtime": 5.0508,
"eval_samples_per_second": 2.376,
"eval_steps_per_second": 2.376,
"step": 46
},
{
"epoch": 8.347826086956522,
"grad_norm": 1.0960992574691772,
"learning_rate": 9.744380058222483e-05,
"loss": 0.1522,
"step": 48
},
{
"epoch": 8.695652173913043,
"grad_norm": 0.9441701173782349,
"learning_rate": 9.698463103929542e-05,
"loss": 0.1582,
"step": 50
},
{
"epoch": 8.869565217391305,
"eval_loss": 1.3662136793136597,
"eval_runtime": 5.0455,
"eval_samples_per_second": 2.378,
"eval_steps_per_second": 2.378,
"step": 51
},
{
"epoch": 9.043478260869565,
"grad_norm": 1.1380761861801147,
"learning_rate": 9.648882429441257e-05,
"loss": 0.1482,
"step": 52
},
{
"epoch": 9.391304347826088,
"grad_norm": 0.7313019037246704,
"learning_rate": 9.595676696276172e-05,
"loss": 0.077,
"step": 54
},
{
"epoch": 9.73913043478261,
"grad_norm": 1.3297815322875977,
"learning_rate": 9.538887392664544e-05,
"loss": 0.0812,
"step": 56
},
{
"epoch": 9.91304347826087,
"eval_loss": 1.572574257850647,
"eval_runtime": 5.0512,
"eval_samples_per_second": 2.376,
"eval_steps_per_second": 2.376,
"step": 57
},
{
"epoch": 10.08695652173913,
"grad_norm": 0.8475630879402161,
"learning_rate": 9.478558801197065e-05,
"loss": 0.0834,
"step": 58
},
{
"epoch": 10.434782608695652,
"grad_norm": 0.6849445700645447,
"learning_rate": 9.414737964294636e-05,
"loss": 0.0373,
"step": 60
},
{
"epoch": 10.782608695652174,
"grad_norm": 1.016614317893982,
"learning_rate": 9.347474647526095e-05,
"loss": 0.0565,
"step": 62
},
{
"epoch": 10.956521739130435,
"eval_loss": 1.753727912902832,
"eval_runtime": 5.0516,
"eval_samples_per_second": 2.376,
"eval_steps_per_second": 2.376,
"step": 63
},
{
"epoch": 11.130434782608695,
"grad_norm": 1.0797334909439087,
"learning_rate": 9.276821300802534e-05,
"loss": 0.0427,
"step": 64
},
{
"epoch": 11.478260869565217,
"grad_norm": 0.9874071478843689,
"learning_rate": 9.202833017478422e-05,
"loss": 0.036,
"step": 66
},
{
"epoch": 11.826086956521738,
"grad_norm": 1.0409115552902222,
"learning_rate": 9.125567491391476e-05,
"loss": 0.0417,
"step": 68
},
{
"epoch": 12.0,
"eval_loss": 1.6627446413040161,
"eval_runtime": 5.043,
"eval_samples_per_second": 2.38,
"eval_steps_per_second": 2.38,
"step": 69
},
{
"epoch": 12.173913043478262,
"grad_norm": 0.40683284401893616,
"learning_rate": 9.045084971874738e-05,
"loss": 0.0247,
"step": 70
},
{
"epoch": 12.521739130434783,
"grad_norm": 0.5845514535903931,
"learning_rate": 8.961448216775954e-05,
"loss": 0.0311,
"step": 72
},
{
"epoch": 12.869565217391305,
"grad_norm": 0.8584285378456116,
"learning_rate": 8.874722443520899e-05,
"loss": 0.0509,
"step": 74
},
{
"epoch": 12.869565217391305,
"eval_loss": 1.666393756866455,
"eval_runtime": 5.0525,
"eval_samples_per_second": 2.375,
"eval_steps_per_second": 2.375,
"step": 74
},
{
"epoch": 12.869565217391305,
"step": 74,
"total_flos": 9819365392252928.0,
"train_loss": 0.4924258470434595,
"train_runtime": 725.5112,
"train_samples_per_second": 3.17,
"train_steps_per_second": 0.345
}
],
"logging_steps": 2,
"max_steps": 250,
"num_input_tokens_seen": 0,
"num_train_epochs": 50,
"save_steps": 25,
"stateful_callbacks": {
"EarlyStoppingCallback": {
"args": {
"early_stopping_patience": 7,
"early_stopping_threshold": 0.0
},
"attributes": {
"early_stopping_patience_counter": 0
}
},
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 9819365392252928.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}