Microscopic-Mistral-6k-steps / trainer_state.json
DrNicefellow's picture
Upload 12 files
569953c verified
raw
history blame contribute delete
No virus
2.05 kB
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.04325558862204997,
"eval_steps": 100000000,
"global_step": 6000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"learning_rate": 1.0000000000000001e-07,
"loss": 10.7586,
"step": 1
},
{
"epoch": 0.0,
"learning_rate": 5e-05,
"loss": 6.5292,
"step": 500
},
{
"epoch": 0.01,
"learning_rate": 4.99998214600225e-05,
"loss": 4.7262,
"step": 1000
},
{
"epoch": 0.01,
"learning_rate": 4.9999285842640105e-05,
"loss": 4.1997,
"step": 1500
},
{
"epoch": 0.01,
"learning_rate": 4.999839315550315e-05,
"loss": 3.8892,
"step": 2000
},
{
"epoch": 0.02,
"learning_rate": 4.9997143411362066e-05,
"loss": 3.6618,
"step": 2500
},
{
"epoch": 0.02,
"learning_rate": 4.9995536628067196e-05,
"loss": 3.5132,
"step": 3000
},
{
"epoch": 0.03,
"learning_rate": 4.9993572828568536e-05,
"loss": 3.4092,
"step": 3500
},
{
"epoch": 0.03,
"learning_rate": 4.999125204091544e-05,
"loss": 3.3295,
"step": 4000
},
{
"epoch": 0.03,
"learning_rate": 4.998857429825615e-05,
"loss": 3.2596,
"step": 4500
},
{
"epoch": 0.04,
"learning_rate": 4.9985539638837424e-05,
"loss": 3.2068,
"step": 5000
},
{
"epoch": 0.04,
"learning_rate": 4.998214810600389e-05,
"loss": 3.163,
"step": 5500
},
{
"epoch": 0.04,
"learning_rate": 4.997839974819749e-05,
"loss": 3.1256,
"step": 6000
}
],
"logging_steps": 500,
"max_steps": 416130,
"num_input_tokens_seen": 0,
"num_train_epochs": 3,
"save_steps": 3000,
"total_flos": 2.197429420032e+18,
"train_batch_size": 16,
"trial_name": null,
"trial_params": null
}