trainer_state.json

{
 "best_metric": 0.00019271760538686067,
 "best_model_checkpoint": "data/further_pretraining/xlnet-large-cased/checkpoint-4706",
 "epoch": 1.0,
 "eval_steps": 500,
 "global_step": 4706,
 "is_hyper_param_search": false,
 "is_local_process_zero": true,
 "is_world_process_zero": true,
 "log_history": [
 {
 "epoch": 0.11,
 "learning_rate": 2.6812579685507864e-05,
 "loss": 0.0002,
 "step": 500
 },
 {
 "epoch": 0.21,
 "learning_rate": 2.3625159371015725e-05,
 "loss": 0.0001,
 "step": 1000
 },
 {
 "epoch": 0.32,
 "learning_rate": 2.043773905652359e-05,
 "loss": 0.0001,
 "step": 1500
 },
 {
 "epoch": 0.42,
 "learning_rate": 1.7250318742031452e-05,
 "loss": 0.0077,
 "step": 2000
 },
 {
 "epoch": 0.53,
 "learning_rate": 1.4062898427539311e-05,
 "loss": 0.0006,
 "step": 2500
 },
 {
 "epoch": 0.64,
 "learning_rate": 1.0875478113047175e-05,
 "loss": 0.0006,
 "step": 3000
 },
 {
 "epoch": 0.74,
 "learning_rate": 7.688057798555037e-06,
 "loss": 0.0,
 "step": 3500
 },
 {
 "epoch": 0.85,
 "learning_rate": 4.500637484062899e-06,
 "loss": 0.0003,
 "step": 4000
 },
 {
 "epoch": 0.96,
 "learning_rate": 1.3132171695707608e-06,
 "loss": 0.0,
 "step": 4500
 },
 {
 "epoch": 1.0,
 "eval_loss": 0.00019271760538686067,
 "eval_runtime": 829.7336,
 "eval_samples_per_second": 90.73,
 "eval_steps_per_second": 5.672,
 "step": 4706
 }
 ],
 "logging_steps": 500,
 "max_steps": 4706,
 "num_input_tokens_seen": 0,
 "num_train_epochs": 1,
 "save_steps": 500,
 "total_flos": 1.89368250236928e+16,
 "train_batch_size": 16,
 "trial_name": null,
 "trial_params": null
}