GaetanMichelet's picture
Model save
1d368a6 verified
{
"best_metric": 0.6932449340820312,
"best_model_checkpoint": "data/Gemma-2-2B_task-2_60-samples_config-2/checkpoint-34",
"epoch": 18.782608695652176,
"eval_steps": 500,
"global_step": 54,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.34782608695652173,
"grad_norm": 0.848487138748169,
"learning_rate": 1e-05,
"loss": 1.056,
"step": 1
},
{
"epoch": 0.6956521739130435,
"grad_norm": 0.7835727334022522,
"learning_rate": 2e-05,
"loss": 1.0008,
"step": 2
},
{
"epoch": 0.6956521739130435,
"eval_loss": 1.0610309839248657,
"eval_runtime": 5.2827,
"eval_samples_per_second": 2.272,
"eval_steps_per_second": 2.272,
"step": 2
},
{
"epoch": 1.391304347826087,
"grad_norm": 0.7976774573326111,
"learning_rate": 4e-05,
"loss": 0.9848,
"step": 4
},
{
"epoch": 1.7391304347826086,
"eval_loss": 1.0089634656906128,
"eval_runtime": 5.0231,
"eval_samples_per_second": 2.389,
"eval_steps_per_second": 2.389,
"step": 5
},
{
"epoch": 2.0869565217391304,
"grad_norm": 0.7716853022575378,
"learning_rate": 6e-05,
"loss": 0.9855,
"step": 6
},
{
"epoch": 2.782608695652174,
"grad_norm": 0.6007464528083801,
"learning_rate": 8e-05,
"loss": 0.8909,
"step": 8
},
{
"epoch": 2.782608695652174,
"eval_loss": 0.8771232962608337,
"eval_runtime": 5.0184,
"eval_samples_per_second": 2.391,
"eval_steps_per_second": 2.391,
"step": 8
},
{
"epoch": 3.4782608695652173,
"grad_norm": 0.5097690224647522,
"learning_rate": 0.0001,
"loss": 0.7601,
"step": 10
},
{
"epoch": 3.8260869565217392,
"eval_loss": 0.8055453300476074,
"eval_runtime": 5.0211,
"eval_samples_per_second": 2.39,
"eval_steps_per_second": 2.39,
"step": 11
},
{
"epoch": 4.173913043478261,
"grad_norm": 0.43442362546920776,
"learning_rate": 9.987820251299122e-05,
"loss": 0.7588,
"step": 12
},
{
"epoch": 4.869565217391305,
"grad_norm": 0.5163419246673584,
"learning_rate": 9.951340343707852e-05,
"loss": 0.7251,
"step": 14
},
{
"epoch": 4.869565217391305,
"eval_loss": 0.7676209807395935,
"eval_runtime": 5.0169,
"eval_samples_per_second": 2.392,
"eval_steps_per_second": 2.392,
"step": 14
},
{
"epoch": 5.565217391304348,
"grad_norm": 0.3666139245033264,
"learning_rate": 9.890738003669029e-05,
"loss": 0.6518,
"step": 16
},
{
"epoch": 5.913043478260869,
"eval_loss": 0.735206663608551,
"eval_runtime": 5.0198,
"eval_samples_per_second": 2.391,
"eval_steps_per_second": 2.391,
"step": 17
},
{
"epoch": 6.260869565217392,
"grad_norm": 0.2873038947582245,
"learning_rate": 9.806308479691595e-05,
"loss": 0.6247,
"step": 18
},
{
"epoch": 6.956521739130435,
"grad_norm": 0.2997960150241852,
"learning_rate": 9.698463103929542e-05,
"loss": 0.6256,
"step": 20
},
{
"epoch": 6.956521739130435,
"eval_loss": 0.7190656661987305,
"eval_runtime": 5.019,
"eval_samples_per_second": 2.391,
"eval_steps_per_second": 2.391,
"step": 20
},
{
"epoch": 7.6521739130434785,
"grad_norm": 0.3153882920742035,
"learning_rate": 9.567727288213005e-05,
"loss": 0.5985,
"step": 22
},
{
"epoch": 8.0,
"eval_loss": 0.7080118060112,
"eval_runtime": 5.0161,
"eval_samples_per_second": 2.392,
"eval_steps_per_second": 2.392,
"step": 23
},
{
"epoch": 8.347826086956522,
"grad_norm": 0.24715402722358704,
"learning_rate": 9.414737964294636e-05,
"loss": 0.5809,
"step": 24
},
{
"epoch": 8.695652173913043,
"eval_loss": 0.703488290309906,
"eval_runtime": 5.0196,
"eval_samples_per_second": 2.391,
"eval_steps_per_second": 2.391,
"step": 25
},
{
"epoch": 9.043478260869565,
"grad_norm": 0.27659326791763306,
"learning_rate": 9.24024048078213e-05,
"loss": 0.5496,
"step": 26
},
{
"epoch": 9.73913043478261,
"grad_norm": 0.42193853855133057,
"learning_rate": 9.045084971874738e-05,
"loss": 0.5221,
"step": 28
},
{
"epoch": 9.73913043478261,
"eval_loss": 0.6957711577415466,
"eval_runtime": 5.0162,
"eval_samples_per_second": 2.392,
"eval_steps_per_second": 2.392,
"step": 28
},
{
"epoch": 10.434782608695652,
"grad_norm": 0.23101910948753357,
"learning_rate": 8.83022221559489e-05,
"loss": 0.5052,
"step": 30
},
{
"epoch": 10.782608695652174,
"eval_loss": 0.6938557028770447,
"eval_runtime": 5.0227,
"eval_samples_per_second": 2.389,
"eval_steps_per_second": 2.389,
"step": 31
},
{
"epoch": 11.130434782608695,
"grad_norm": 0.2817871570587158,
"learning_rate": 8.596699001693255e-05,
"loss": 0.4982,
"step": 32
},
{
"epoch": 11.826086956521738,
"grad_norm": 0.30969250202178955,
"learning_rate": 8.345653031794292e-05,
"loss": 0.4727,
"step": 34
},
{
"epoch": 11.826086956521738,
"eval_loss": 0.6932449340820312,
"eval_runtime": 5.0202,
"eval_samples_per_second": 2.39,
"eval_steps_per_second": 2.39,
"step": 34
},
{
"epoch": 12.521739130434783,
"grad_norm": 0.27874162793159485,
"learning_rate": 8.07830737662829e-05,
"loss": 0.4384,
"step": 36
},
{
"epoch": 12.869565217391305,
"eval_loss": 0.6967836022377014,
"eval_runtime": 5.0207,
"eval_samples_per_second": 2.39,
"eval_steps_per_second": 2.39,
"step": 37
},
{
"epoch": 13.217391304347826,
"grad_norm": 0.3269486725330353,
"learning_rate": 7.795964517353735e-05,
"loss": 0.4157,
"step": 38
},
{
"epoch": 13.91304347826087,
"grad_norm": 0.29554519057273865,
"learning_rate": 7.500000000000001e-05,
"loss": 0.4074,
"step": 40
},
{
"epoch": 13.91304347826087,
"eval_loss": 0.7064465880393982,
"eval_runtime": 5.021,
"eval_samples_per_second": 2.39,
"eval_steps_per_second": 2.39,
"step": 40
},
{
"epoch": 14.608695652173914,
"grad_norm": 0.30623385310173035,
"learning_rate": 7.191855733945387e-05,
"loss": 0.3396,
"step": 42
},
{
"epoch": 14.956521739130435,
"eval_loss": 0.7209694385528564,
"eval_runtime": 5.0221,
"eval_samples_per_second": 2.389,
"eval_steps_per_second": 2.389,
"step": 43
},
{
"epoch": 15.304347826086957,
"grad_norm": 0.3516511619091034,
"learning_rate": 6.873032967079561e-05,
"loss": 0.3681,
"step": 44
},
{
"epoch": 16.0,
"grad_norm": 0.3707938492298126,
"learning_rate": 6.545084971874738e-05,
"loss": 0.3296,
"step": 46
},
{
"epoch": 16.0,
"eval_loss": 0.7495496273040771,
"eval_runtime": 5.0169,
"eval_samples_per_second": 2.392,
"eval_steps_per_second": 2.392,
"step": 46
},
{
"epoch": 16.695652173913043,
"grad_norm": 0.38637402653694153,
"learning_rate": 6.209609477998338e-05,
"loss": 0.2865,
"step": 48
},
{
"epoch": 16.695652173913043,
"eval_loss": 0.7751684784889221,
"eval_runtime": 5.0224,
"eval_samples_per_second": 2.389,
"eval_steps_per_second": 2.389,
"step": 48
},
{
"epoch": 17.391304347826086,
"grad_norm": 0.37249138951301575,
"learning_rate": 5.868240888334653e-05,
"loss": 0.2732,
"step": 50
},
{
"epoch": 17.73913043478261,
"eval_loss": 0.8187534213066101,
"eval_runtime": 5.0223,
"eval_samples_per_second": 2.389,
"eval_steps_per_second": 2.389,
"step": 51
},
{
"epoch": 18.08695652173913,
"grad_norm": 0.4434156119823456,
"learning_rate": 5.522642316338268e-05,
"loss": 0.256,
"step": 52
},
{
"epoch": 18.782608695652176,
"grad_norm": 0.4307098686695099,
"learning_rate": 5.174497483512506e-05,
"loss": 0.2207,
"step": 54
},
{
"epoch": 18.782608695652176,
"eval_loss": 0.880825936794281,
"eval_runtime": 5.0261,
"eval_samples_per_second": 2.388,
"eval_steps_per_second": 2.388,
"step": 54
},
{
"epoch": 18.782608695652176,
"step": 54,
"total_flos": 2.895737692094464e+16,
"train_loss": 0.5615027408908915,
"train_runtime": 1252.9206,
"train_samples_per_second": 1.836,
"train_steps_per_second": 0.08
}
],
"logging_steps": 2,
"max_steps": 100,
"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": 2.895737692094464e+16,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}